Consumer Spending and Property Taxes

Consumer Spending and Property Taxes Abstract A sudden and temporary change to the Italian property tax system in 2011 generated significant variation in the amount of taxes paid across home-owners. Using new questions appositely added to the Survey on Household Income and Wealth, we exploit this cross-sectional variation to provide an unprecedented analysis of the consumption effects of a tax on housing wealth. A tax hike on the main dwelling leads to large expenditure cuts among mortgagors, who hold low liquid wealth despite owning sizable illiquid assets. In contrast, higher tax rates on other residential properties affect affluent households, thereby having a modest impact on their consumer spending. Our results provide novel and direct evidence in favor of recent theories that highlight the role of household debt in the transmission of economic policies. “Household debtors are frequently young families acquiring homes and furnishing before they earn incomes to pay for them outright; given the difficulty of borrowing against future wages, they are liquidity-constrained and have a high marginal propensity to consume.” (James Tobin, 1980, “Asset accumulation and economic activity”, p. 10) 1. Introduction What are the effects of housing taxes on consumer spending? And what groups of society bear most of its costs? Although a large body of research has made notable progress to quantify the effects of tax rebates, little is known about the impact of tax hikes on consumer spending and about whether property taxes are likely to distort household behavior less than income taxes. This seems particularly surprising in the light of a growing macroeconomic literature—surveyed by Mian and Sufi (2016, Chap. 5) and Piazzesi and Schneider (2016)—advocating a key role for housing over the business cycle. The present analysis fills this important gap in the literature exploiting the 2011 changes in the Italian property tax “Imposta Municipale Unica” (“IMU”). A newly appointed central government swiftly legislated and implemented a sizable fiscal consolidation plan whose main intervention temporarily redesigned the municipal system on housing taxes. The changes raised around €4.0 billion from taxes on the main dwelling and an additional €10.1 billion from taxes on other residential properties, for a total revenue increase of 0.90% of GDP. The IMU affected 25.8 million tax payers (or around 70% of households) with an average contribution per tax-paying household around €357 on the main dwelling and about €905 on other residential properties. Using a new set of questions (on the amount of IMU paid) appositely added to the Survey of Household Income and Wealth (SHIW) conducted by the Bank of Italy, we identify the effects of property taxes on consumer spending. We do so by comparing the difference in expenditure for IMU payers before and after the tax change to the difference in expenditure for non-IMU payers over the same period, employing a range of specifications that control for demographics, changes in house value, property characteristics, expectations on future household income, and expectations on future local house prices as well as regional fixed effects. In the most restrictive specifications, we look at home-owners only and therefore we focus exclusively on variation in the amount of property taxes paid. Our identification strategy builds upon four features of the 2011 changes in the municipal system of housing taxation in Italy. First, the central government introduced a new tax on the main dwelling and increased by an exogenous factor the (by then obsolete) land registry estimates of the rental values used to calculate the tax base for the main dwelling and other residential properties. Second, the timing and depth of the legislated changes were largely unanticipated. Third, municipal authorities were allowed to unilaterally modify the rates proposed by the central government and, as shown in Section 2, the geographical variation in property tax rates appears driven by political motives that were unrelated to past local economic conditions or other local economic policies. Fourth, the IMU tax changes were announced by the government as an experiment (whose possible extension would have been subject to government revision) and most SHIW respondents did not expect the changes to persist longer than five years. Indeed, the housing tax on the main dwelling was subsequently abolished in 2015. A household-level approach appears to offer two main advantages relative to a more macro strategy that relates changes in central government tax revenues to changes in aggregate consumption. First, macroeconomic interventions—like a change in residential property taxes—are often the endogenous policy response to conditions in the aggregate economy, thereby posing a reverse causality problem when using data from national statistics. In contrast, the cross-household and cross-municipality variation in property tax rates that we implicitly exploit for identification on micro data seems unlikely to be the policy response to specific circumstances at the individual household level, especially after controlling for demographics and property characteristics as we do here. On the other hand, aggregate circumstances or the effects of other economic policies may confound the evaluation of the impact of the 2011 property tax changes on household expenditure behavior. However, an extensive analysis in Section 3 reveals that the amount of IMU taxes paid is not systematically related to the household variables that were directly affected by other policy changes over the same period. A second main advantage of using survey data is that they allow us to explore potentially interesting dimensions of heterogeneity across liquid holdings and household debt positions, so as to shed light on the specific channel(s) of policy transmission. The empirical analysis isolates five major empirical regularities. First, the marginal propensity to consume (MPC) nondurable goods and services out of the IMU tax is around 0.05 whereas the MPC on durable goods is about 0.43. Second, these average effects mask pervasive heterogeneity across residential properties, with the taxes paid on the main dwelling associated with a large and significant MPC on durable goods and the taxes on other residential properties associated with a small and insignificant MPC on durable goods. In contrast, the MPCs on nondurable goods and services are statistically indistinguishable from zero, both across residential properties and across household groups. Third, the significant response to the main dwelling IMU tax is far more pronounced among home-owners with mortgage debt, who are shown to hold low liquid wealth relative to income despite owning sizable illiquid assets and thus appear to fit well the notion of “wealthy” hand-to-mouth consumers. Fourth, debtors concentrated their cuts on vehicles expenditure. Fifth, the direct negative consequences of the changes in the IMU residential property taxes are estimated to be around 0.11% of GDP in 2012 vis-à-vis an increase in tax revenues of 0.90% of GDP (or 1.80% of government revenues). On the other hand, the direct impact of the property taxes on the car industry was large, making a negative contribution around 11% (or about half of the overall decline) relative to the market size in 2011. Finally, the evidence in this paper compares favorably with a long standing tradition in economics that has advocated the use of age and income as proxies for the presence of liquidity constraints. We show that holding a mortgage is, in fact, a far stronger predictor for the liquidity shortage behind the observed sensitivity of consumption to temporary income changes, therefore highlighting the role of household debt as a novel source of violation of the permanent income hypothesis as well as a powerful amplification mechanism for the transmission of macroeconomic shocks. Contribution and Related Literature. Our analysis seeks to contribute to three main strands of the literature. First, a burgeoning line of theoretical research has emphasized the role that illiquid wealth (and especially housing) could play in the transmission of macroeconomic policies. Selected examples include Eggertsson and Krugman (2012), Kaplan and Violante (2014), Ragot (2014), Mitman (2012), and Andrés, Bosca, and Ferri (2011). Our analysis provides direct evidence in support of these theories by offering an unprecedented evaluation of the household expenditure effects of a housing wealth tax. Second, a theoretical literature pioneered by Browning and Crossley (2000) and extended by Aaronson, Agarwal, and French (2012) generate the testable predictions that not only durable expenditure should react more than nondurable consumption following a temporary change in household resources but also that, among the durable spending categories, goods requiring lower down payments for their purchase on credit—such as vehicles—should exhibit far larger marginal propensities to consume that other durable goods. Our findings speak in favor of this mechanism. Third, an important set of studies pioneered by Johnson, Parker, and Souleles (2006) and investigated further by Parker et al. (2013), Agarwal and Qian (2014), and Jappelli and Pistaferri (2014) look at household expenditure in response to a transitory increase in disposable income. In contrast, the presents analysis focuses on a decrease in disposable income by offering some of the earliest evidence on the consumer spending response to a fiscal austerity measure in Europe. Structure of the Paper. Section 2 describes the institutional design and the cross-sectional variation that we exploit for identification. Section 3 presents the data and the empirical specifications before assessing the role of other confounding factors. The main results on the IMU tax paid on the main dwelling and on other residential properties as well as the heterogeneous responses across household balance sheet positions are presented in Section 4, together with evidence that most mortgagors hold very low liquidity relative to income. We conclude this section by sketching a theoretical argument that generates the prediction of a larger response on durable goods expenditure (and vehicle purchases in particular). Estimates for different spending categories and for the role of credit are the focus of Section 5. Further results on age and income splits, the role of uncertainty, house prices and demographics as well as the response of income and hours worked are presented in Section 6. We conclude with some back of the envelope calculations that quantify the direct impact of the IMU tax changes on the Italian economy in 2012. 2. Institutional Design and Geographical Variation In this section, we first outline a brief history of housing taxation in Italy. We then describe the specific context in which the property tax changes were introduced in December 2011 and finally we describe the variation in the IMU rates that we exploit for identification in the econometric analysis. 2.1. A Brief History of Municipal Property Taxes in Italy The “Municipal Tax on Properties” (“Imposta Comunale sugli Immobili”, aka “ICI”) was introduced in the Italian legislation by the law by decree number 333 on July 11, 1992 and subsequently transformed into law on December 30, 1992.1 The ICI tax base included three main categories: buildings, building plots, and farmlands.2 Our analysis on household expenditure will focus on the “buildings” category. Under the ICI system, the tax base for “buildings” was the land registry value defined as an estimate of what the rental value of the property would have been in 1988–1989, which was used as a base biennium. This (rough) estimate, which was self-reported to the municipal registry by the buyer at the time of purchase, was based on the location and building type but did not account for other important dimensions such as the type of construction, the age of the building and more generally for the conditions of the property. Not surprisingly, the system became obsolete soon after its introduction but was left essentially unchanged in the following two decades against the backdrop of steadily growing house prices. In Figure A.1 of Online Appendix A, we show that the ratio of the estimated land registry values to the actual market values at the end of the ICI system averaged around 3.6 (see Bocci, Iommi, and Marinari 2012; IMF 2012 for similar evidence). The property tax rates were set independently by the municipal governments within the range of 0.4%–0.7%, according to local preferences. The ICI remained substantially unchanged until the end of 2007, when the government led by Prime Minister Romano Prodi approved an increase of the basic deduction of 0.133%.3 The policy change applied only to taxes on the main dwelling with a cap of €200. Finally, on March 27, 2008, the subsequent government led by Prime Minister Silvio Berlusconi abolished the ICI tax on the main dwellings (excluding three building categories corresponding to “luxury houses” (category “A1”), “villas” (category “A8”), and “castles” (category “A9”)) with the law by decree number 93/2008 whereas the ICI tax on other properties remained unchanged. 2.2. The “IMU” Tax On December 4, 2011, a newly appointed Italian government led by Prime Minister Mario Monti announced a fiscal consolidation plan that was meant to “ensure fiscal stability, growth and equity”. The plan was passed into law with immediate effect on December 22, 2011.4 Among the most sizable interventions, the government reformed the property tax system, abolished ICI, introduced a single municipal property tax under the heading of “Imposta Municipale Unica” (“IMU”), and presented the policy change to the public as an “experiment”. According to the official technical notes accompanying the law, the introduction of the IMU (which was levied only on property owners) accounted for three quarters of the increase in taxation associated with the 2011 consolidation plan. The swift implementation of Monti’s government IMU reform (in less than two months since the resignation of former Prime Minister Silvio Berlusconi), together with the frequency of the SHIW (conducted in 2010 and 2012), makes these property tax changes most likely unanticipated by households (especially back in 2010). Finally, in line with the government announcement back in 2011, the IMU tax on the main dwellings (subsequently extended to housing services under the new heading of “TASI”) was abolished in July 2015 by the government led by Prime Minister Matteo Renzi. The introduction of the IMU tax significantly reformed the property tax regime along three dimensions. First, it included the land registry value of the main dwelling in the tax base, previously excluded. Second, the land registry values (for both main dwellings and other properties) were scaled up by an exogenous factor (homogeneous across all municipalities and equal to 1.6 for residential dwellings), so as to increase the tax base by an average of 49% (see IMF 2012). Finally, the IMU system set the basic tax rate on primary (other) residences at 0.4% (0.76%) of the registry value but allowed municipalities to modify this rate within a 0.2% (0.3%) band. Furthermore, the government set the basic deduction at €200 plus an additional €50 deduction per children less than 26 years old (up to a maximum of an additional €400): whereas municipalities were allowed to modify this, around 98% of local governments chose the basic deduction of €200.5 Overall, the IMU system determined a sharp increase in residential property taxation: the revenues on the main properties increased from nothing in 2011 to €4.0 billion in 2012 while those on other properties increased from 7.8 billion in 2011 to 17.9 billion in 2012. Between 2011 and 2012, total tax revenues on residential properties increased by €14.1 billion corresponding to around 0.90% of Gross Domestic Product (GDP) in 2012.6 Our analysis exploits the fact that in the 2012 Italian Survey on Household Income and Wealth (SHIW), respondents were appositely asked for the first time to report the amount of recurrent housing taxes paid on both the main dwelling and other residential properties. In Figure 1, we plot the distribution of self-reported IMU payments per household from the SHIW, distinguishing between the amount of housing taxes paid on the main dwelling (in the first row) and the amount of taxes paid on other residential properties (in the second row). The first column displays the distribution of the absolute amount of euros paid whereas the second column reports this as a share of the household monthly income. Because of the deductions, 21.6% of home-owners did not pay the IMU tax on the main dwelling and 13.2% of home-owners with more than one property did not pay the IMU tax on other residential properties. The IMU affected 25.8 millions of tax payers (or around 70% of households). The average payment on the main dwelling was about €357 (or 14% of a household monthly income) whereas the average payment on all residential properties was €905 (or 36% of a household monthly income). It is worth noting that, as shown by Norregaard (2013), it is very hard to evade property taxes in a high-income country like Italy. Finally, about 30% of SHIW respondents reported a zero probability that the IMU tax would have been eliminated within five years and not replaced by another similar tax.7 Figure 1. View largeDownload slide IMU tax burden per household. The figures refer to owners, IMU tax payers only. The red line plots the Epanechnikov kernel density. Panel a1 (a2) refers to the amount paid on main dwellings in Euro per household (as a share of households’ monthly income), excluding 14 observations higher than €3,000. Panel b1 (b2) refers to the amount of IMU tax paid (as a share of monthly income) on other properties, excluding 129 observations higher than €3,000. Source: authors’ calculations on SHIW survey data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 1. View largeDownload slide IMU tax burden per household. The figures refer to owners, IMU tax payers only. The red line plots the Epanechnikov kernel density. Panel a1 (a2) refers to the amount paid on main dwellings in Euro per household (as a share of households’ monthly income), excluding 14 observations higher than €3,000. Panel b1 (b2) refers to the amount of IMU tax paid (as a share of monthly income) on other properties, excluding 129 observations higher than €3,000. Source: authors’ calculations on SHIW survey data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). 2.3. IMU Rate Variation and Local Business Cycle The variation in the amount of IMU tax paid across households stems from three main features of the law: demographics (and in particular the number of children eligible for deduction), property characteristics (including surface and building type, which determine the land registry rental value) and local tax rates (given that municipalities were allowed to vary the rates set by the government). In the SHIW, we observe demographics and property characteristics but—to preserve anonymity—we are only provided with the region (rather than the municipality) where a household lives in. This implies that controlling for demographics and property characteristics in a projection of household expenditure change on the income change stemming from the IMU taxes disbursement is likely to isolate variation in the amount of property taxes paid due either to geographical variation in the municipal tax rates or to unobserved characteristics that are not absorbed by our rich set of covariates. In 2012, 35.2% (57.3%) of municipalities chose to modify the tax rate on the main dwelling (other residential properties) set by the national government, with the vast majority opting for higher rates. In Figures B.1 and B.2 of Online Appendix B, we construct heat maps that illustrates the municipal variation in property tax rates on the main dwelling and other residential properties across the national territory. To interpret the coefficient on IMU paid as the causal effect of the tax change on private expenditure, we need to verify that the geographical variation in the tax rates was not the municipal government response to past local economic conditions. The concern is that property tax rates may have been consistently higher in municipalities with a higher concentration of households with certain (financial and economic) characteristics. To assess this hypothesis, Table 1 reports the correlation between the municipal IMU tax rates of 2012 and a number of indicator of local economic performance available at municipal level in 2010 and 2011, ranging from personal and business income to night light density.8 The main take away from this table is that there is little evidence of a systematic relation between the IMU tax rates and local economic conditions in the preceding years. Table 1. Correlation between IMU rates and local economic activity. Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Notes: Night lights density correlations exclude small municipalities (<5,000 inhabitants) and big cities (>300,000 inhabitants). Other measures of economic activity such as employment or unemployment data are not available at municipal level. a. IMU rates on both main dwelling and other residential properties refer to 2012. View Large Table 1. Correlation between IMU rates and local economic activity. Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Notes: Night lights density correlations exclude small municipalities (<5,000 inhabitants) and big cities (>300,000 inhabitants). Other measures of economic activity such as employment or unemployment data are not available at municipal level. a. IMU rates on both main dwelling and other residential properties refer to 2012. View Large This latter result is echoed by Figure 2, which records across municipalities the correlations between the share of votes to the left-wing coalition in the local elections immediately before the property tax reform and (i) the IMU tax rates on the main dwelling (top left panel), (ii) the IMU tax rates on other properties (top right panel), (iii) night light density (bottom left panel), and (iv) business income growth (bottom right panel). Although, on the one hand, the top row reveals a significant relation between the tax rates and political orientation, the findings in the bottom panels show that, on the other hand, political orientation is not systematically linked to local economic performance. Figure 2. View largeDownload slide Correlations political orientation—IMU rates—local business cycle. Each dot on the charts represents the average of the respective bin. “Votes to left-wing coalition” refers to the share of votes to the left-wing coalition in regional elections. For most municipalities the latest regional election before the IMU change was in 2010 (March 28th). Source: Authors’ calculations on IFEL data (available at: http://www.webifel.it/ICI/AliquoteIMU.cfm) and Ministry of Interior data. Figure 2. View largeDownload slide Correlations political orientation—IMU rates—local business cycle. Each dot on the charts represents the average of the respective bin. “Votes to left-wing coalition” refers to the share of votes to the left-wing coalition in regional elections. For most municipalities the latest regional election before the IMU change was in 2010 (March 28th). Source: Authors’ calculations on IFEL data (available at: http://www.webifel.it/ICI/AliquoteIMU.cfm) and Ministry of Interior data. In summary, this section contains two messages. First, municipal variation in property tax rates is not related to past local economic conditions. Second, political orientation at the municipal level is an important driver of variation in property tax rates. As there is no systematic difference in the economic performance of center-left and center-right governments across Italian municipalities, we regard this politically-driven variation in property tax rates as exogenous from the standpoint of household expenditure decisions. 2.4. Other Local Economic Policies To isolate the effects of property taxes (as opposed to the effects of other local economic policies), it seems important to establish how municipal governments have employed the extra resources made available by the IMU tax. We do so in this section by focusing on changes (between 2011 and 2012) in the local authorities balance sheets, whose descriptive statistics are reported in Table C.1 of Online Appendix C. More specifically, we project the change in (i) municipal government expenditure, (ii) other municipal tax revenues (net of IMU revenues), (iii) municipal property tax revenues, and (iv) local fiscal deficit on the municipal IMU tax rates using official data from the Ministry of the Interior. To control for the size of each municipality, the dependent variables are standardized by the number of inhabitants. We run these specifications either over the full sample of Italian municipalities or only for those with more than 2,000 inhabitants to ameliorate possible concerns on the quality of the balance sheets in smaller cities. The evidence from Table 2 suggests that there is no significant association between IMU rates and changes in either municipal public spending (in column (1)) or other municipal tax revenues (in column (2)). Also, there is no statistical association (not shown in Table 2) between IMU rates and other local government revenues (grants and dismissals). On the other hand, the estimates in column (3) suggest that increasing the IMU rate on the main dwelling by 0.1 percentage points brings IMU tax revenues up by €8.7 per capita (€9.3 per capita when restricting the sample to municipalities above 2,000 inhabitants). At the same time, an increase of 0.1 percentage points in the IMU rate on the other dwellings generates an average rise in IMU tax revenues of €22.4 per capita, consistent with the fact that IMU revenues on other residential properties are about three times larger than IMU revenues on the main dwelling at the aggregate level. Finally, these extra property tax revenues seem to translate fully into a reduction of the “municipal fiscal deficit” (in column (4)), which is the fiscal target as defined by the central government in the “Internal Stability Pact”.9 It is worth mentioning that the law requires to calculate the municipal fiscal deficit on a “mixed accrual-cash basis”, with current expenditure and revenues evaluated on an accrual basis and capital expenditure and revenues evaluated on a cash basis. For this reason (and because total revenues also include grants and dismissals), the first three columns of Table 2 do not sum up to the last column. Table 2. Other municipal government policies. Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Notes: Each column reports results of a regression where the left hand side variable is the per-capita change of the municipal government instrument in the title between 2012 and 2011 projected on the IMU rates on the main dwelling and other residential properties set by that very municipality in 2012. Municipal public expenditure refers to the variable “Total expenditure” (“Totale generale delle spese”), other municipal tax revenues refers to the sum of all municipal tax revenues net of IMU payments (specifically we consider the following taxes: “Scopo”, “Soggiorno”, “Pubblicita”, “Occupazione degli spazi pubbliche”, “Raccolta e smaltimento dei rifiuti”, “Tassa affissioni”, “Anagrafe”, “Uffici giudiziari”, “Polizia municipale”, “Istruzione elementare”, “Istruzione media”, “Assistenza scolastica”, “Biblioteche”, “Teatri, attività culturali”, “Piscine comunali”, “Stadio comunale, palazzo dello sport”, “Manifestazioni diverse”, “Servizi turistici”, “Viabilità”, “Trasporti pubblici locali”, “Urbanistica”, “Edilizia residenziale pubblica locale”, “Servizio idrico”, “tariffa igiene ambientale”, “Asili nido”, “Proventi servizi di prevenzione e riabilitazione”, “Ricovero per anziani”, “Assistenza, beneficenza pubblica”, “Servizio necroscopico e cimiteriale”, “Entrate da sanzioni amministrative”, “Mezzi pubblicitari”, “Proventi di bene”, “Scuola materna”, “Addizionale IRPEF”, “Segreteria”, “Ufficio tecnico”, “Servizi turistici”, “C.O.S.A.P.”, “Concessioni cimiteriali”). Fiscal deficit calculated as difference between municipal expenditures (current + capital) and municipal revenues (tax revenues + current grants + dismissals). Current revenues and expenditures are on accrual basis, capital revenues and expenditures are on a cash basis. For this reason, the sum of the coefficients in the first three columns cannot equal the coefficient in the last column. Robust standard errors clustered by provinces in brackets. ***Significant at 1%. Source: Ministry of Interior data (publicly available at: http://finanzalocale.interno.it/apps/floc.php/in/cod/4). View Large Table 2. Other municipal government policies. Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Notes: Each column reports results of a regression where the left hand side variable is the per-capita change of the municipal government instrument in the title between 2012 and 2011 projected on the IMU rates on the main dwelling and other residential properties set by that very municipality in 2012. Municipal public expenditure refers to the variable “Total expenditure” (“Totale generale delle spese”), other municipal tax revenues refers to the sum of all municipal tax revenues net of IMU payments (specifically we consider the following taxes: “Scopo”, “Soggiorno”, “Pubblicita”, “Occupazione degli spazi pubbliche”, “Raccolta e smaltimento dei rifiuti”, “Tassa affissioni”, “Anagrafe”, “Uffici giudiziari”, “Polizia municipale”, “Istruzione elementare”, “Istruzione media”, “Assistenza scolastica”, “Biblioteche”, “Teatri, attività culturali”, “Piscine comunali”, “Stadio comunale, palazzo dello sport”, “Manifestazioni diverse”, “Servizi turistici”, “Viabilità”, “Trasporti pubblici locali”, “Urbanistica”, “Edilizia residenziale pubblica locale”, “Servizio idrico”, “tariffa igiene ambientale”, “Asili nido”, “Proventi servizi di prevenzione e riabilitazione”, “Ricovero per anziani”, “Assistenza, beneficenza pubblica”, “Servizio necroscopico e cimiteriale”, “Entrate da sanzioni amministrative”, “Mezzi pubblicitari”, “Proventi di bene”, “Scuola materna”, “Addizionale IRPEF”, “Segreteria”, “Ufficio tecnico”, “Servizi turistici”, “C.O.S.A.P.”, “Concessioni cimiteriali”). Fiscal deficit calculated as difference between municipal expenditures (current + capital) and municipal revenues (tax revenues + current grants + dismissals). Current revenues and expenditures are on accrual basis, capital revenues and expenditures are on a cash basis. For this reason, the sum of the coefficients in the first three columns cannot equal the coefficient in the last column. Robust standard errors clustered by provinces in brackets. ***Significant at 1%. Source: Ministry of Interior data (publicly available at: http://finanzalocale.interno.it/apps/floc.php/in/cod/4). View Large Our findings are consistent with the evidence reported by Grembi, Nannicini, and Troiano (2012), who show that following a less stringent constraint on municipal fiscal deficits in Italy, local governments responded mainly by cutting real estate taxes and marginal income tax rates. Independent evidence on the lack of correlation between property taxes and other local taxes is provided in Figure C.1 of Online Appendix C, which scatter plots the IMU rates against the rates on the municipal component of income taxes (“IRPEF”). In summary, the geographical variation in IMU rates across municipalities does not seem to be associated with the cross-sectional variation in other municipal government economic policies. 3. Data and Empirical Framework In this section, we present the household survey data and outline the empirical specification that we use to link the income change induced by the IMU taxes paid to the expenditure change. As discussed in the previous section, we use a rich set of demographics and property-specific covariates to isolate exogenous variation across households at a similar stage of their life-cycle, owing properties with similar value and characteristics but living in (unobserved) municipalities with different tax rates. Finally, we discuss and evaluate the role of possible confounding factors, including other macroeconomic interventions, as well as run a placebo test over two waves of the SHIW that have witnessed no changes in municipal property taxes 3.1. The Household Survey Data Our dataset is based on the “Survey on Households Income and Wealth” (SHIW) conducted by the Bank of Italy. The survey is run every two years and covers around 8,000 households distributed over about 3,000 Italian municipalities. The data are available in anonymous form. Each survey is conducted at the end of the respective year during the last few weeks of December. On average, about half of the households that appear in one survey overlap in the following wave. Given that sampling design involves unequal stratum sampling fractions, the use of household sampling weights is necessary to obtain unbiased estimates of the corresponding aggregates. In our econometric analysis, we rely on two consecutive surveys (2010 and 2012), although in some of the analyses we consider also 2008. The 2010 survey covers 19,836 individuals grouped in 7,951 households whereas the 2012 survey covers 20,022 individuals grouped in 8,151 households. We use household level data and keep households who were surveyed both in 2010 and 2012 (about 56% of the 2012 survey). Then, we drop observations with missing values in some relevant variables (typically the market value or the surface of the main dwelling). Finally, to reduce the impact of compiling errors and outliers, we drop observations in the 0.5% tails of the distribution of total expenditure changes, leaving us with a sample of 4,002 observations. We report the descriptive statistics of our working dataset in Table 3, highlighting mean, median, 25th and 75th percentiles of the distribution of the variables of interest for the full regression sample (first three columns), home-owners only (middle panel) and mortgagors (last three columns). To correct for the under-reporting of financial assets (see D’Aurizio et al. 2006), we rescale this variable by the ratio between the value of financial assets for the whole economy calculated by the Bank of Italy on data from the national statistical agency (ISTAT) and its SHIW counterpart obtained by summing up the value of financial assets for all households in the survey using sampling weights. Home-owners are around 71.5% (27.9% by inheritance) of our regression sample whereas the share of mortgagors is 13.7% of the home-owners.10 As shown in Figure D.1 of Online Appendix D, these shares display a remarkable stability over time and whereas the fraction of households owning only one residential property has somewhat decreased between 2010 and 2012, we note that the start of this decline dates back to 2006. Table 3. Summary statistics, regression sample. Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Notes: “Age” and “Studio” refer to the age and the education level (1 = elementary or lower, 6 = postgraduate degree) of the head of the household. “$${\triangle }$$Y” refers to the change of household disposable income. “$${\triangle }$$C” refers to the change of household consumption. “$${\triangle }$$CD” refers to the change of household consumption on durables. “$${\triangle }$$CV” refers to the change of household consumption on vehicles. The entries for vehicles purchases are consistent with the data from ACI (http://www.aci.it), which show that in 2012, about 2.5 millions of vehicles were exchanged across all Italian households, which are around 24.5 millions. The number of mortgagors that bought a new care as a share of total number of mortgagors in 2012 was around 5% whereas in 2010 it was around 9.5%. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). “Real estate” refers to the variable “ar1” (“Real assets (housing, land, and other buildings)”) in database “ricf2012.dta”. “Mortgage debt” refers to variable “deb12a” in dataset “fami2012.dta”. View Large Table 3. Summary statistics, regression sample. Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Notes: “Age” and “Studio” refer to the age and the education level (1 = elementary or lower, 6 = postgraduate degree) of the head of the household. “$${\triangle }$$Y” refers to the change of household disposable income. “$${\triangle }$$C” refers to the change of household consumption. “$${\triangle }$$CD” refers to the change of household consumption on durables. “$${\triangle }$$CV” refers to the change of household consumption on vehicles. The entries for vehicles purchases are consistent with the data from ACI (http://www.aci.it), which show that in 2012, about 2.5 millions of vehicles were exchanged across all Italian households, which are around 24.5 millions. The number of mortgagors that bought a new care as a share of total number of mortgagors in 2012 was around 5% whereas in 2010 it was around 9.5%. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). “Real estate” refers to the variable “ar1” (“Real assets (housing, land, and other buildings)”) in database “ricf2012.dta”. “Mortgage debt” refers to variable “deb12a” in dataset “fami2012.dta”. View Large The net wealth of Italian households is among the highest in the world but it has a defining peculiarity: around 65% is represented by real assets. The median net wealth is around €270,000 (€348,000 among all home-owners and €289,000 among mortgagors) corresponding to a lower debt-to-income ratio than in other advanced economies.11 Relative to the full regression sample, which also include renters, home-owners tend to have a higher level of both net liquid and illiquid wealth. Relative to all home-owners, mortgagors tend to have a younger head, higher income, more volatile expenditure, lower net liquid wealth, and real estates with a higher value. 3.2. Empirical Specifications The goal of our analysis is to relate variation in disposable income stemming from cross-household variation in the IMU taxes to variation in household expenditure. As there was no tax on the main residential property in 2010 (and only a small tax amount was typically paid on other residential properties because of the obsolete land registry rental value then), we begin by looking at the effect of the tax paid on the main dwelling in 2012 on the household expenditure change between 2010 and 2012. Then, we turn our attention to the richer specification that also includes as a regressor the IMU paid on other residential properties in 2012.12 To ensure our empirical strategy isolates variation in the amount of taxes paid that is unrelated to household and property characteristics, a rich set of controls is featured in the following specification: \begin{equation} \triangle C_{i}=\alpha +{{\gamma }}\cdot \textit{IMUmain}{}_{i}+{\delta }\cdot \triangle HP_{i}+\boldsymbol {\theta }{\boldsymbol X}_{i}+\varepsilon _{i}, \end{equation} (1) where $${\triangle }$$Ci indicates the change in expenditure (on either nondurable goods and services or durable goods) of household i between 2010 and 2012 ($${\triangle }$$Ci = Ci, 2012 − Ci, 2010), $$\textit{IMUmain}_{i}$$ is the amount of IMU tax paid on the main dwelling in 2012, $${\triangle }$$HPi is the self-reported change in house price ($${\triangle }$$HPi = HPi, 2012 − HPi, 2010), $${\boldsymbol X}_{i}$$ contains a set of controls, and ϵi is an idiosyncratic shock.13 As covariates in matrix $${\boldsymbol X}_{i}$$, we add four sets of variables: (i) households demographics, including age and educational attainment of the household head, family size, number of children and their square values, two dummies that takes value of one for home-owners and mortgagors respectively, (ii) regional dummies, (iii) property characteristics including type of building, surface, number of owned properties and dummies for the type of neighborhood (city center, suburbs, etc.) and (iv) a set of dummy variables capturing expectations about future income and about future local house prices (see Online Appendix E for a detailed description). As we control for both demographics and property characteristics, which influence either directly (through the deductions) or indirectly (through the land registry rental value) the household-specific amount of municipal property tax paid, the coefficient γ on IMU is likely to capture the variation in household consumption due to the municipal variation in the IMU tax rates. As the latter appears unrelated both to other local economic policies and to past local economic conditions (as discussed in the previous section), equation (1) can be estimated using OLS and the coefficient γ can be interpreted as the causal effect of the IMU property tax on consumer spending. The coefficient δ captures the household-level association between changes in expenditure and changes in the subjective house value. Finally, our empirical strategy relies on the absence of dissimilar pretreatment trends in expenditure (between IMU payers and non-IMU payers) that may account for the post-treatment differences across the two groups. In Section 5.1, we present evidence consistent with this hypothesis. In the richer specification, we also consider the $$\textit{IMUother}_{i}$$ paid on other residential properties: \begin{equation} \triangle C_{i}=\alpha +{{\gamma _{1}}}\cdot \textit{IMUmain}{}_{i}+{{\gamma _{2}}}\cdot \textit{IMUother}{}_{i}+{\delta }\cdot \triangle HP_{i}+\boldsymbol {\theta } {\boldsymbol X}_{i}+\varepsilon _{i}, \end{equation} (2) where the coefficients of interest are now γ1 and γ2, representing the impact of the IMU tax on the main dwelling and the IMU tax on other residential properties. In our baseline specifications, equations (1) and (2) are estimated either over the full sample or for home-owners only, exploiting in the latter case exclusively variation in the amount of property taxes paid. To shed lights on the characteristics driving heterogeneity in the spending response, (1) and (2) will then be run splitting the sample into households without and with (mortgage) debt, showing that the latter display significantly higher MPCs. The expenditure changes from the housing tenure grouping are then compared in Section 6 to the findings from more traditional groupings based on age and income as well as to subsamples of households reporting different levels of uncertainty about their future income. In Section 5, we focus on the different categories of durable expenditure and find that most of the changes in debtors’ spending are concentrated in vehicle purchases. 3.3. Other Confounding Factors As shown in Section 2, the IMU tax hikes does not appear related neither to other local economic policies nor to past local economic conditions. Still, the availability of households survey data only in 2010 and 2012 poses the challenge that other macroeconomic developments may confound the inference one can draw about the effect of the property taxes on consumer spending. In this section, we take this challenge at face value and ask whether the IMU tax changes were correlated with any other significant macroeconomic change that may have occurred over these two years. Accordingly, we use specifications that are all like (1) and (2) but in which the dependent variable becomes: the amount of taxes paid on other non-IMU austerity macro interventions, changes in households transfers from the central government (including pensions), changes in house value, changes in the taxes paid on “super-cars” and changes in the expenditure for those nondurable goods and services whose VAT increased between 2010 and 2012. The dependent variable in the first column of Table 4 is calculated as the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the regional marginal tax rate on personal income. This is meant to capture the host of other austerity interventions that were passed together with the IMU tax changes. The columns on transfers and house value assess whether the change in municipal property taxation was associated, amplified or perhaps offset by other changes in the government budget, the household balance sheet or the tax base. This seems particularly important in the light of the Fornero reform of the Italian pension system, which was also part of the fiscal consolidation plan passed into law by Prime Minister Mario Monti’s government in December 2011. Given the very significant fall in vehicles expenditure associated with the property tax changes (reported in Section 5), in the fourth column we evaluate the relation between the amount of property taxes paid by each households and the taxes paid on supercars (defined as cars above 185 kW), whose tax rate was also changed in 2011.14 An additional confounding factor occurred in September 2011 when the government led by Prime Minister Silvio Berlusconi passed an increase in the Value Added Tax (VAT) rate from 20% to 21%. Accordingly, the last column of Table 4 reports the consumption response of those nondurable goods and services that were subject to the VAT rate change. Table 4. Confounding factors. Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Notes: Robust standard errors clustered by regions in brackets. IMU “main” and “other” refer to the IMU tax paid for the main dwelling and other properties, respectively. Because the variables have different magnitudes and variances, all left-hand side variables, IMU main and IMU other have been standardized. When running the same regressions on nonstandardized variables we obtain very similar results. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Austerity non-IMU” refers to the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the local (regional) marginal tax rate on personal income. “Transfers” refers to total transfers to households, including pensions. “Supercar” is a variable calculated as the product between the value of the car if above €40,000 and the average yearly tax rate of 1.26% on supercar (estimated using Automobile Club of Italy data). Finally, “VAT” refers to the consumption change on nondurable goods and services whose VAT rate changed in September 2011. View Large Table 4. Confounding factors. Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Notes: Robust standard errors clustered by regions in brackets. IMU “main” and “other” refer to the IMU tax paid for the main dwelling and other properties, respectively. Because the variables have different magnitudes and variances, all left-hand side variables, IMU main and IMU other have been standardized. When running the same regressions on nonstandardized variables we obtain very similar results. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Austerity non-IMU” refers to the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the local (regional) marginal tax rate on personal income. “Transfers” refers to total transfers to households, including pensions. “Supercar” is a variable calculated as the product between the value of the car if above €40,000 and the average yearly tax rate of 1.26% on supercar (estimated using Automobile Club of Italy data). Finally, “VAT” refers to the consumption change on nondurable goods and services whose VAT rate changed in September 2011. View Large Reassuringly, in each of the two panels and samples, there is little evidence that the amount of IMU taxes paid by each household was systematically related to any of the macro policy and economic changes summarized in Table 4. Furthermore, the Fornero reform on pensions affected younger generations evenly across housing tenure groups and we show in Section 6 that young households who paid the IMU taxes contracted their durable expenditure by a significantly larger amount than young households who did not pay the IMU taxes. Finally, the VAT rate changed both on vehicles and on all other (nonvehicle) durable goods: but, as we will show in Section 5.1, only the expenditure on vehicles witnessed a significant contraction, suggesting that the 2011 VAT rate change seems unlikely to have contributed to our findings. In Figure 3, we explore further the impact of the VAT change by reporting the evolution of three price indexes from national accounts: (i) items that experienced an increase in the VAT rate (dashed-dotted black line), (ii) items that did not experience an increase in the VAT rate (light gray solid line) and (iii) cars (red solid line), which were also subject to the VAT rate change. The vertical lines correspond to the dates of the introduction of the VAT rate increase and of the IMU reform respectively. Two main developments are apparent from Figure 3. First, following the VAT rate change of September 2011, both the increase in the price index on all items subject to the VAT rate hike and the increase in the price index on cars are far sharper than the mild increase in the price index on flat-VAT rate items. Second, the behavior of the price index on increased-VAT rate items decouples from the behavior of the car price index around December 2011 when the IMU tax changes were passed into law by Mr Monti’s government. Given we will show that vehicles purchases was the single most responsive and most declining spending category to the IMU taxes, we interpret the flat profile of the car price index after the introduction of the IMU (relative to the steadily rising profile of the price index on all increased-VAT rate items) as most likely stemming from the effect of the property taxes on consumer spending. Figure 3. View largeDownload slide Evolution of prices. The figure shows the evolution of prices for cars, items subject to the 2011 VAT increase, and items exempted from VAT increase (these items are subject to a 0% VAT, 4% VAT, or 10% VAT according to the category; these VAT rates were unchanged in the considered period). The aggregate indexes (for “Flat-VAT items” and “Increasing-VAT items”) are weighted averages of the respective subindexes. The relative weights are provided by ISTAT. Inflation for “cars” refer to the “motor cars” category (ISTAT code 711). Items excluded from VAT changes include: “education”, “food” (excluding “ready-made meals”), “restaurants and hotels”, “miscellaneous goods and services” (excluding “mineral or spring waters”), “actual rentals for housing”, “water supply and miscellaneous services relating to the dwelling”, “electricity, gas and other fuels”, “medical products, appliances and equipment”, “out-patient service”, “hospital services”, “transport services”, “postal services”, “recreational and cultural services”, “newspapers, books and stationery”. The share of items (including cars) subject to the VAT increase in 2011 was 40.6%. Source: Authors’ calculations on ISTAT data (available at: http://dati.istat.it/?lang=en). Figure 3. View largeDownload slide Evolution of prices. The figure shows the evolution of prices for cars, items subject to the 2011 VAT increase, and items exempted from VAT increase (these items are subject to a 0% VAT, 4% VAT, or 10% VAT according to the category; these VAT rates were unchanged in the considered period). The aggregate indexes (for “Flat-VAT items” and “Increasing-VAT items”) are weighted averages of the respective subindexes. The relative weights are provided by ISTAT. Inflation for “cars” refer to the “motor cars” category (ISTAT code 711). Items excluded from VAT changes include: “education”, “food” (excluding “ready-made meals”), “restaurants and hotels”, “miscellaneous goods and services” (excluding “mineral or spring waters”), “actual rentals for housing”, “water supply and miscellaneous services relating to the dwelling”, “electricity, gas and other fuels”, “medical products, appliances and equipment”, “out-patient service”, “hospital services”, “transport services”, “postal services”, “recreational and cultural services”, “newspapers, books and stationery”. The share of items (including cars) subject to the VAT increase in 2011 was 40.6%. Source: Authors’ calculations on ISTAT data (available at: http://dati.istat.it/?lang=en). Finally, the inference on the effects of an increase in property taxes may be distorted by a decline in central government expenditure, which as illustrated in Table G.1 of Online Appendix G mainly came in the form of a fall in government consumption or wages for public employees (see Born, Müller, and Pfeifer 2014). With respect to this hypothesis, we focus on two subgroups of households: those headed by a public employee and those not. We find no statistical differences in the coefficients on $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ across the two groups, with estimated responses being slightly stronger for nonpublic employees. In summary, the results in this section suggest that the effects of the IMU tax paid on household expenditure seem unlikely to be confounded by other nation-wide policies or macroeconomic factors that changed over the same period. 3.4. Placebo Test As a further empirical validation of the extent to which our framework is well-suited to capture the effect of the IMU taxes on consumer spending, we run placebo regressions that correlate the change in expenditure of each household between 2008 and 2010 with the IMU tax paid by that very household in 2012. If the IMU fiscal shock of December 2011 was unanticipated and was indeed the trigger of the significant expenditure decline in 2012 (which we document in the next section), then we would expect it to have no significant effect on expenditure before 2012, given that no actual changes in property taxes occurred between the end of 2008 and the end of 2010 when these two other waves of the SHIW were conducted. In this section (and in this section only), all other right hand side variables (including $${\triangle }$$HPi) refer to the period 2008–2010. In contrast, $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ refer to the amount of taxes paid by household i in 2012. The left hand side variable is the expenditure change of that same household i between 2008 and 2010. For the placebo analysis, we only rely on households who appear in all three waves. Accordingly, the regression sample is reduced from 4,002 to 2,480 observations. The results of the placebo test are shown in Table 5. Both $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ never affect significantly either nondurable consumption or durable expenditure and the estimated coefficients have often the wrong sign.15 On the other hand, the effect of house prices is highly significant for nondurable consumption (but not for durable expenditure), with magnitudes that are not statistically different from the point estimates we will present in the next section for 2010–2012. Table 5. Placebo test. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Notes: Robust standard errors clustered by regions in brackets. “Non durables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. ***Significant at 1%. View Large Table 5. Placebo test. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Notes: Robust standard errors clustered by regions in brackets. “Non durables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. ***Significant at 1%. View Large 4. The Response of Household Expenditure In this section, we present the main results of our analysis. We start with the baseline estimates in Table 6, which associate the IMU taxes paid on the main dwelling and other residential properties with nondurable and durable expenditure. Then, we explore the heterogeneity of these responses and show they vary significantly across household balance sheet positions as exemplified by the presence of mortgage debt. Furthermore, we show that the majority of households with mortgage debt hold very low liquid wealth relative to income and therefore are likely to face liquidity shortages in the wake of changes in their resources. Finally, we review the testable predictions of the theoretical literature on the spending categories that are most likely to respond to a temporary change in household resources. Table 6. Baseline results. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large Table 6. Baseline results. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large 4.1. Baseline Results The estimates of equation (1) and equation (2) are presented in Panel A and Panel B of Table 6, respectively.16 The two columns on the left refer to the full sample whereas those on the right focus on home-owners only. The odd columns display the relevant IMU and house price coefficients for a specification using nondurable consumption on the left hand side whereas in the even columns the dependent variable is the expenditure on durable goods. Four main empirical regularities emerge from these baseline estimates. First, the MPC associated with the IMU tax paid on the main dwelling in columns (1) and (3) is always very close to and never statistically different from zero. Second, the MPC on durable goods from $$\textit{IMUmain}_{i}$$ is always very significant and large, with point estimates around 0.43 in columns (2) and (4). Interestingly, Parker et al. (2013) report a marginal propensity to spend around 0.5 in response to the 2008 income tax rebate in the United States whereas Jappelli and Pistaferri (2014) document that for every hypothetical euro of transitory income, the average SHIW respondent would increase expenditure by 48 cents. Third, the results on $$\textit{IMUmain}_{i}$$ are robust to using a richer specification that also includes $$\textit{IMUother}_{i}$$ among the regressors. Fourth, and in sharp contrast to the main dwelling, the IMU tax paid on other residential properties in Panel B triggers neither a significant contraction in nondurable consumption nor a significant contraction in durable expenditure, with point estimates always in the neighborhood of zero.17 Of independent interest, both panels record also the estimates of the house price effect. In particular, the coefficient on $${\triangle }$$HPi is small and statistically indistinguishable from zero in columns (2) and (4) for durable expenditure. But the marginal propensity to consume out of housing wealth in columns (1) and (3) is always very significant and precisely estimated at around 1% (i.e., a €100 appreciation in house prices tends to be associated with a 97 cents increase in nondurable consumption). Although these estimates are in line with the effects reported by Guiso, Paiella, and Visco (2005) and Paiella and Pistaferri (2017) on earlier SHIW samples, they are sizably smaller than the 5%–7% reported by Mian, Rao, and Sufi (2013) for the United States or the 7% to 9% reported by Campbell and Cocco (2007) for the United Kingdom. It should be noted, however, that the scarcity of refinancing opportunities—and in particular of mortgage equity withdrawal—makes housing wealth in Italy significantly more illiquid (see IMF 2008; Grant and Peltonen 2008; Calza, Monacelli, and Stracca 2013). Accordingly, the statistical association between house prices and consumption in Italy seems more likely to reflect a direct wealth effect or a common factor driving both variables rather than a collateral channel. We will come back to the sensitivity of nondurable consumption to house prices in the extended analysis of Section 6.4 where we will, among other things, (i) add household income as a further control in an augmented version of specification (2) and (ii) exclude housing wealth and the number of children from the set of covariates. 4.2. Grouping Households by Mortgage Debt Position A growing strand of empirical studies, including Dynan (2012), Kaplan, Violante, and Weidner (2014), Cloyne and Surico (2017), and Acconcia, Corsetti, and Simonelli (2015) advocate a role for household balance sheet positions, and mortgage debt in particular, in the transmission of structural and policy shocks to consumption. The variation of IMU tax rates across households allows us an unprecedented evaluation of this hypothesis in the context of a tax hike on housing wealth. To this end, in Table 7 we group households according to whether they have debt (first two columns) or not (last two columns). In an effort to maximize the number of observations, in Panel A, we include secured and unsecured debt. In Panel B, we focus on mortgage debt only.18 Table 7. Debtors (mortgagors) versus nondebtors (nonmortgagors). Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with (secured or unsecured) debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large Table 7. Debtors (mortgagors) versus nondebtors (nonmortgagors). Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with (secured or unsecured) debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large The main take away from Table 7 is that the significant average effects on durable goods recorded in the previous tables are entirely driven by home-owners with debt, whose marginal propensities to spend (out of the taxes paid on the main dwelling) in column (4) tend to be larger and more significant than in Table 6, despite the far fewer number of observations. The results from columns (1) and (2) of Panel B reveal further that removing as few as some 400 mortgagors from the full sample yields very small and largely insignificant responses to both property taxes. In the next section, we will show that net vehicle purchases account for the lion share of the behavior of durable expenditure and that the magnitude of the MPC in Table 7 for households with debt is consistent with a down payment rate for buying a car around 10%. Interestingly, the house price effect in column (3) tends to be smaller for debtors in Panel A and not statistically different from zero (though imprecisely estimated) for mortgagors in Panel B, consistent with a shortage of refinancing opportunities in the Italian credit market. Finally, we note that some durable goods (such as cars, motorbikes, furniture, or electrical appliance) tend to be purchased using consumer credit. Accordingly, a worsened access to financial markets during 2011 and 2012 could—at least in principle—be partially responsible for the large adjustment on durable goods recorded between the two waves of the SHIW (at the end of 2010 and at the end of 2012) in Tables 6 and 7. As shown in Figure H.1 of Online Appendix H, however, there seems to be little evidence that over this biennium households were charged systematically higher interest rates for consumer credit (on purchases either below or above €5,000) relative to 2010, when these series begin. 4.3. Why do Mortgagors Have a Higher MPC? A large theoretical literature exemplified by Deaton (1992) and Zeldes (1992) has convincingly made the case that liquidity constrains could be a primary source of violation of the permanent income hypothesis. Although earlier empirical contributions have typically associated the presence of liquidity constraints with lower income, lower educational attainment and younger household head (see for instance Johnson et al. 2006), the theoretical mechanism in Kaplan and Violante (2014) suggests that also wealthy households may face liquidity shortages whenever a large durable purchase such as housing makes illiquid most of their wealth. In this section, we therefore evaluate the hypothesis that mortgagors could be “wealthy” hand-to-mouth by looking at their household balance sheet. In Figure 4, we compare, by number of dwellings, the distribution of the net saving rate—that is, disposable income minus total consumption as a share of disposable income—with the distribution of the debt service ratio,—that is, mortgage repayments as a share of disposable income. After expenditure and debt repayments, mortgagors with only one property are left with little disposable income as measured by the small distance between the median values of the net saving rate (17%) and the debt service ratio (16%). In contrast, mortgagors with more than one property appear less constrained, as they enjoy significantly larger saving rates relative to the debt service ratio distribution. Figure 4. View largeDownload slide Distribution of net saving rates and debt service ratio per number of property. The figure shows the distribution of net saving rates (using after tax income) and debt service ratio to net disposable income by number of properties. The bars span in between the 25th and 75th percentile of the distribution whereas the horizontal lines in each bar indicate the median of the distribution. Monthly saving rates defined as the ratio between variable “s2” and variable “y2” in database “cons12.dta”. Mortgage service payments is based on variables “tdebita11” plus “tdebita12” plus “tdebita13”, database “alld2_res.dta” in the 2012 survey. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 4. View largeDownload slide Distribution of net saving rates and debt service ratio per number of property. The figure shows the distribution of net saving rates (using after tax income) and debt service ratio to net disposable income by number of properties. The bars span in between the 25th and 75th percentile of the distribution whereas the horizontal lines in each bar indicate the median of the distribution. Monthly saving rates defined as the ratio between variable “s2” and variable “y2” in database “cons12.dta”. Mortgage service payments is based on variables “tdebita11” plus “tdebita12” plus “tdebita13”, database “alld2_res.dta” in the 2012 survey. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Complementary evidence on the liquidity shortage faced by mortgagors is recorded in Figure 5, which reports for each housing tenure group the share of households holding an amount of net liquid wealth below half of their monthly income.19 Two results emerge from this evidence: (i) about 65% of mortgagors appear hand-to-mouth and (ii) the shares of the other two groups are only a fraction of the mortgagor share, with the proportion of liquidity constrained households being around 25% among both outright home-owners and renters. Bearing in mind that renters represent about 30% of the Italian population and that renters tend to hold very little (if any) housing or financial wealth, our evidence reveals than about 8% of Italian households are “wealth-poor” hand-to-mouth. On the other hand, mortgagors and outright home-owners account for around 13% and 57% of the population, thereby making the share of “wealth-rich” hand-to-mouth households close to 22%. Figure 5. View largeDownload slide Share of hand-to-mouth households by housing tenure group. The figure plots the share of hand-to-mouth across housing tenure groups in 2012. Outright homeowners include households owning one property only. “Hand-to-mouth” refers to households with a net liquid wealth to income ratio lower than half month of income. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). Disposable income refers to variable “y2” in dataset “consXX.dta” (where the suffix XX indicates the year of the survey). Figure 5. View largeDownload slide Share of hand-to-mouth households by housing tenure group. The figure plots the share of hand-to-mouth across housing tenure groups in 2012. Outright homeowners include households owning one property only. “Hand-to-mouth” refers to households with a net liquid wealth to income ratio lower than half month of income. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). Disposable income refers to variable “y2” in dataset “consXX.dta” (where the suffix XX indicates the year of the survey). The findings of Figure 5 also offers a rationale for why grouping households only by their level of liquid wealth would produce less sharp and less significant evidence of heterogeneity (than looking at mortgage debt position) in the expenditure response to the temporary income changes triggered by the IMU property taxes. This can be seen both in Table I.2 and Figure I.1 of Online Appendix I, which report the MPCs for lower and higher liquid wealth households when alternative multiples of their monthly income are used as threshold for the group categorization. The reason is that although also a significant portion of renters as well as some outright home-owners tend to hold low liquid wealth (as shown in Figure 5), renters do not pay the property tax because they do not own a house whereas outright home-owners do no longer have a significant fraction of their expenditure precommitted into repaying the mortgage (which they have already repaid in full) and thus their MPC is not statistically different from zero (as shown in Table 7). Accordingly, in both Figure I.1 and Table I.2 of Online Appendix I, the standard errors are larger and the differences in point estimates across groups smaller than in Table 7, as grouping households into a single low liquidity group pools together consumers with very different MPCs.20 In summary, grouping households by their debt position, and in particular whether home-owners with only one property have a mortgage or not, seems to provide sharp(er) evidence of significant heterogeneity in the expenditure responses (than, for instance, simply looking at liquid wealth to income ratios).21 Inspection of the balance sheets of the different groups reveals further that, in each month, owner–occupier mortgagors can only spare little liquidity after meeting expenditure bills and mortgage repayments, consistent with Tobin’ conjecture (cited in the introduction) that debt makes households liquidity constrained and therefore leads to a high marginal propensity to consume. This is consistent with the notion that a significant portion of households with debt are hand-to-mouth despite owing sizable illiquid wealth, thereby providing a novel interpretation for the role of (il)liquid wealth in the transmission of macroeconomic shocks. 4.4. Why is the Response of Durable Expenditure Stronger? In Section 4.1, we have shown that the response of durable expenditure is stronger than the response of nondurable goods and services spending. In Section 4.2, we have reported that households with debt exhibit a larger marginal propensity to spend on durable goods whereas in Section 4.3 we have looked at the net liquid wealth to income ratio to document that most debtors are (wealthy) hand-to-mouth. In this section, we review a theoretical mechanism that can offer a rationale for why the changes in durable expenditure are larger than the changes in nondurable consumption. In an earlier contribution, Browning and Crossley (2000) prove that luxury goods have high elasticities of intertemporal substitution and therefore are easier to postpone in the face of temporary falls in disposable income. The intuition for this result comes from noting that, if goods are additively separable, the Frisch own price elasticities are proportional to the Marshallian income elasticities, with the latter implying that whenever total spending is cut then luxury goods expenditure will be cut by proportionally more. To the extent that most categories of durables are likely to involve more discretionary purchases and less necessity goods than many categories of nondurables, a corollary of their result is that the expenditure on durable goods should be easier to postpone than the consumption on nondurable goods and services. This is consistent with the estimates reported in Table 6. In a more recent work, Aaronson et al. (2012) study the problem of a household that maximizes the utility flows from the consumption of some nondurable goods, Ct, and from the services generated by a stock of durables, Dt, which are bundled together as \begin{equation*} E_{0}\sum _{t=0}^{T}\beta ^{t}\left(C_{t}^{1-\theta }D_{t}^{\theta }\right)^{1-\gamma }/(1-\gamma ). \end{equation*} The stock of durable goods, which depreciated at rate δ, can expand through further investment It, according to the following law of motion: \begin{equation} D_{t+1}=(1-\delta )D_{t}+I_{t} \end{equation} (3) whereas financial asset At accumulates in the form \begin{equation} A_{t+1}=(1+r)A_{t}+Y_{t}-C_{t}-I_{t} \end{equation} (4) with At+1 ≥ 0 and the interest rate denoted by r. The disposable income, represented by Yt, is made up of a deterministic life-cycle profile and a stochastic AR(1) process. A main feature of their analysis is to allow households to borrow against durable goods according to the constraint \begin{equation} -A_{t}\le (1-\pi )D_{t}, \end{equation} (5) where π is the down payment rate or the fraction of newly purchased durable goods that does not serve as collateral, in the spirit of Kiyotaki and Moore (1997). Aaronson et al. (2012) formally show that the marginal propensity to spend out of a temporary income change is far higher for durables than for nondurables. Furthermore, they show that, for goods purchased with higher down payments, consumer spending is less sensitive to income changes because higher down payments imply that fewer durable goods can be purchased with a given level of income. To develop an intuition for their results, assume that the borrowing constraint (5) always binds. Combining this binding constraint with the accumulation formula for financial assets (4) and the law of motion for durables (3) yields the following expression: \begin{equation} \pi I_{t}+C_{t}+(1-\pi )(r+\delta )D_{t}=Y_{t}. \end{equation} (6) Equation (6) makes it clear that although one dollar worth of nondurables requires one dollar of disposable income to be purchased, one dollar worth of durables only require a fraction π of a dollar. This finding not only offers a rationale for why the response to transitory income shocks may be concentrated on durable expenditure (consistent with the estimates reported in this section) but also generates two further testable predictions. First, durable goods requiring lower down payments for their purchase on credit—such as vehicles—should display a stronger response than the rest of durables, which most likely either require higher down payments—such as furniture—or for which collateralized financing may not be readily available—such as small appliances.22 Second, the MPC on durable goods with low down payment rates may exceed one (and indeed even two), as demonstrated by the quantitative analysis in Aaronson et al. (2012) (Table 8) using standard calibrations of the model described previously. In Section 5, we will provide evidence in support of these two further predictions by showing that the durable response is concentrated on vehicles, that the associated MPC is indeed well above one and that the credit taken for vehicles purchases declined significantly in response to the IMU tax changes. Table 8. Vehicles versus nonvehicles durable expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $$\triangle $$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large Table 8. Vehicles versus nonvehicles durable expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $$\triangle $$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large 5. Spending Categories and the Role of Credit In this section, we explore the extent of heterogeneity in the household responses to the IMU property taxes across spending categories. In particular, we show that net vehicle purchases (or lack thereof) are a main driver of the durable expenditure results in the previous section and that the magnitude of the coefficients on $$\textit{IMUmain}_{i}$$ reported in Table 7 for debtors is consistent with a typical down payment for buying a car. Furthermore, we show that IMU payers are significantly less likely to have taken out a loan for vehicles purchase after the tax policy change than non-IMU payers whereas no discernable patter across households is evident for credit to purchase other durable goods. Finally, we compare our estimates based on negative shocks on disposable income (triggered by the hike in property taxes) with the evidence based on positive income shocks in earlier studies. 5.1. Vehicles Versus Nonvehicles Expenditure To shed lights on the findings in the previous section, we rerun specification (2) over several categories of nondurable and durable expenditure. Based on question E02 of the 2012 (and 2010) SHIW survey, we consider as “durable good” precious objects, cars, other means of transport, furniture, furnishings, appliances, and “various equipment”.23 Nondurable expenditure is calculated as the difference between total expenditure net of rents or mortgage payments and expenditure on nondurable goods and services. Given the estimates in the previous tables, it should not come as a surprise that we find little evidence of heterogeneity among nondurable consumption categories. As for durable goods, we find that net vehicle expenditure (defined as the difference between vehicle purchases and vehicle sales) is the only component that displays large and significant responses to the IMU taxes. This is recorded in Table 8, which splits durable expenditure into vehicles (which account for about 70% of durable goods value) and every other durable goods. In the top panel, we report findings over the full-sample and for home-owners only whereas in the bottom panel we display results for debtors and mortgagors. The coefficients on net vehicles purchases in columns (2) and (4) of Panel A are similar (and statistically indistinguishable) from the coefficients on durable expenditure in Table 6. When net vehicles purchases are excluded from durable expenditure in columns (1) and (3), however, both IMU tax coefficients become insignificant, revealing that this durable category drives the total expenditure response. In Panel B of Table 8, we restrict our attention to indebted households, who display the strongest durable expenditure response, and show that their behavior is indeed driven by net vehicle purchases. The coefficients on $$\textit{IMUmain}_{i}$$ in columns (2) and (4) of Panel B appears in line with their durable expenditure counterparts in Table 7 whereas the responses of nonvehicle durables in columns (1) and (3) tend to be small and statistically indistinguishable from zero. Two points are worth emphasizing about the magnitude of the vehicle expenditure response in Table 8. First, Italian households paid a significant amount of property taxes in 2012, with an average around €357 and a significant portion of payers above €1000. This suggests that some households may have chosen to defer or even eliminate a large durable purchase, whose saving could offset the significant increase in property taxes over a multi-year horizon. Second, given the average per-year disbursement for the IMU tax on the main dwelling, a marginal propensity to spend around two for households with debt—while statistically close to one—is consistent with a down payment rate around 10% on a vehicle purchase. Interestingly, Misra and Surico (2014) show that, also in the context of the U.S. income tax rebates of 2001 and 2008, the aggregate consumption response was driven by a handful of vehicle purchases made by the mortgagors group, who displayed a marginal propensity to spend on this category around two. The finding that the response of total expenditure is driven by a few very large marginal propensities to spend on vehicles is also consistent with the evidence in Aaronson et al. (2012), who report estimated MPCs on vehicles around two for a small number of working households facing a minimum wage hike. In Figure 6, we provide a graphical counterpart of the results in Table 8. In particular, we show not only that (i) the average reduction in net vehicle purchases by IMU payers (dashed-dotted dark gray line) was larger than the average reduction by those home-owners who—because of the deductions—did not pay IMU taxes (dashed gray line), but also that (ii) the average reduction for mortgagor IMU payers (red line) was more pronounced that the average reduction across all IMU payers. Figure 6 provides graphical evidence that the pre-treatment trend was identical between mortgagors IMU-payers and home-owners non-IMU payers as well as that the decline in vehicles expenditure for all IMU payers between 2008 and 2010 (i.e., before the treatment) was, if anything, smaller than the decline for home-owners non-IMU payers over the same period. Finally, it is possible to show that the parallel trends extend back in time, although at the expense of reducing the sample size given the sample rotation in the SHIW survey. Figure 6. View largeDownload slide Expenditure on vehicles. The chart shows the expenditure on vehicles over time among home-owners in deviation from their respective means. The pretreatment trend is identical across groups and the same evidence applies to all variable of interest. If the time period is extended back in time (at the expenses of the sample size) the evidence of parallel trends remains identical. The chart is based on those households entering in the three waves of the SHIW survey reported. The expenditure on vehicles refers to variable “cd1” in database “consXX.dta” (where the suffix “XX” refers to the year of the survey). The households with a debt are identifies using the variable “deb12a” (“Amount of debts owed at the end of the year to banks or financial companies for the purchase or restructuring of buildings”) in database “famiXX.dta”. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 6. View largeDownload slide Expenditure on vehicles. The chart shows the expenditure on vehicles over time among home-owners in deviation from their respective means. The pretreatment trend is identical across groups and the same evidence applies to all variable of interest. If the time period is extended back in time (at the expenses of the sample size) the evidence of parallel trends remains identical. The chart is based on those households entering in the three waves of the SHIW survey reported. The expenditure on vehicles refers to variable “cd1” in database “consXX.dta” (where the suffix “XX” refers to the year of the survey). The households with a debt are identifies using the variable “deb12a” (“Amount of debts owed at the end of the year to banks or financial companies for the purchase or restructuring of buildings”) in database “famiXX.dta”. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). A complementary way to look at the impact of the IMU property taxes is to ask whether a larger tax disbursement is associated with a lower probability of making a durable purchase. To this end, we construct two binary variables that take the value of one if vehicle expenditure or other durable expenditure respectively are positive and zero otherwise. These become the dependent variables in two separate probit regressions that use otherwise the same regressors as in the specifications in the rest of the paper. Two main advantages of this approach is that a binary variable is less prone to measurement errors (relative to the exact euro amount of any durable purchase) and the probit specification is suited to handle nonlinearity in the data. The results in Table 9 reveals—consistently with the estimates in Table 8—that only for mortgagors the payment of the IMU tax on the main dwelling does significantly reduce the probability of buying a vehicle. Furthermore, Figure J.1 of Online Appendix J shows that the marginal effect of $$\textit{IMUmain}_{i}$$ (which is the probability of purchasing a vehicle following the payment of the IMU tax on the main dwelling) is monotonically decreasing in the amount of taxes paid, conditional on covariates. Table 9. Probit regressions on expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.11 −0.18 0.16 −0.16 [0.12] [0.11] [0.12] [0.11] IMU other −0.06 −0.03 −0.07* −0.03 [0.04] [0.05] [0.04] [0.05] $${\triangle }$$HP (€ ’00) 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01] Observations 4,002 4,002 3,122 3,122 Area under ROC 0.63 0.77 0.62 0.76 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.11 −0.18 0.16 −0.16 [0.12] [0.11] [0.12] [0.11] IMU other −0.06 −0.03 −0.07* −0.03 [0.04] [0.05] [0.04] [0.05] $${\triangle }$$HP (€ ’00) 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01] Observations 4,002 4,002 3,122 3,122 Area under ROC 0.63 0.77 0.62 0.76 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.15 −0.93*** 0.01 −0.95** [0.24] [0.27] [0.31] [0.47] IMU other −0.22** 0.02 −0.10 0.17 [0.09] [0.10] [0.18] [0.19] $${\triangle }$$HP (€ ’00) −0.01 −0.01 −0.01 −0.01*** [0.01] [0.01] [0.01] [0.01] Observations 877 877 414 414 Area under ROC 0.66 0.82 0.71 0.80 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.15 −0.93*** 0.01 −0.95** [0.24] [0.27] [0.31] [0.47] IMU other −0.22** 0.02 −0.10 0.17 [0.09] [0.10] [0.18] [0.19] $${\triangle }$$HP (€ ’00) −0.01 −0.01 −0.01 −0.01*** [0.01] [0.01] [0.01] [0.01] Observations 877 877 414 414 Area under ROC 0.66 0.82 0.71 0.80 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods, excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey).“Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Probit estimated via maximum likelihood. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large Table 9. Probit regressions on expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.11 −0.18 0.16 −0.16 [0.12] [0.11] [0.12] [0.11] IMU other −0.06 −0.03 −0.07* −0.03 [0.04] [0.05] [0.04] [0.05] $${\triangle }$$HP (€ ’00) 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01] Observations 4,002 4,002 3,122 3,122 Area under ROC 0.63 0.77 0.62 0.76 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.11 −0.18 0.16 −0.16 [0.12] [0.11] [0.12] [0.11] IMU other −0.06 −0.03 −0.07* −0.03 [0.04] [0.05] [0.04] [0.05] $${\triangle }$$HP (€ ’00) 0.01 0.01 0.01 0.01 [0.01] [0.01] [0.01] [0.01] Observations 4,002 4,002 3,122 3,122 Area under ROC 0.63 0.77 0.62 0.76 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.15 −0.93*** 0.01 −0.95** [0.24] [0.27] [0.31] [0.47] IMU other −0.22** 0.02 −0.10 0.17 [0.09] [0.10] [0.18] [0.19] $${\triangle }$$HP (€ ’00) −0.01 −0.01 −0.01 −0.01*** [0.01] [0.01] [0.01] [0.01] Observations 877 877 414 414 Area under ROC 0.66 0.82 0.71 0.80 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.15 −0.93*** 0.01 −0.95** [0.24] [0.27] [0.31] [0.47] IMU other −0.22** 0.02 −0.10 0.17 [0.09] [0.10] [0.18] [0.19] $${\triangle }$$HP (€ ’00) −0.01 −0.01 −0.01 −0.01*** [0.01] [0.01] [0.01] [0.01] Observations 877 877 414 414 Area under ROC 0.66 0.82 0.71 0.80 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods, excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey).“Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Probit estimated via maximum likelihood. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large The results in Tables 8 and 9 are further corroborated by Figure 7, which displays the volume of monthly transactions of new and used cars as published by the Italian automobile association (“Automobil Club d’Italia”). The vertical line denotes the launch of the IMU reform in December 2011 and this is also the month when the break in the mean of the time series is apparent.24 In Table J.1 of Online Appendix J, we use registration data from the Ministry of Infrastructure and Transport to show that this drop is more pronounced among new cars but evenly spread across national and international makers. Figure 7. View largeDownload slide Monthly sales of (new and used) cars. The figure shows the evolution of monthly sales of cars (new and used vehicles). The series refers to the seasonally adjusted sales as a share of the 2011 average level. The seasonal adjustment has been performed using an unobserved component model casted in the state-space (the Kalman filter has been initiated using a diffuse prior). The vertical red line indicates the month when the IMU tax was announced (December 2011). Source: Authors’ own calculations on ACI (“Automobile Club d’Italia”) data (available at: http://www.aci.it/laci/studi-e-ricerche/dati-e-statistiche/auto-trend.html). Figure 7. View largeDownload slide Monthly sales of (new and used) cars. The figure shows the evolution of monthly sales of cars (new and used vehicles). The series refers to the seasonally adjusted sales as a share of the 2011 average level. The seasonal adjustment has been performed using an unobserved component model casted in the state-space (the Kalman filter has been initiated using a diffuse prior). The vertical red line indicates the month when the IMU tax was announced (December 2011). Source: Authors’ own calculations on ACI (“Automobile Club d’Italia”) data (available at: http://www.aci.it/laci/studi-e-ricerche/dati-e-statistiche/auto-trend.html). To quantify the aggregate effects implied by the estimated marginal propensity to spend in Table 8, we note that the number of car sales recorded by ACI (see Figure 7) reveal an average fall of about 185,000 units per year between 2008 and 2011. But during 2012, car sales plummeted by around 346,000 units (or about 20% of the market size in 2011). Assuming an average cost for new vehicles of around €13,400, the total drop from 2011 to 2012 totaled to about €4.6 billion.25 Because the estimated average MPC on vehicle is 0.61 and the total IMU revenues on the main dwelling were €4 billion, the estimated impact on the car industry implied by our regressions is around €2.4 billion. Therefore, our analysis attributes around half of the 2012 fall, which is about −10.7% of the market size in 2011, to the introduction of the IMU tax. 5.2. The Role of Credit The theoretical set up of Section 4.4 highlights the role of credit in amplifying the (negative) consequences of the IMU taxes (hike) on durable goods purchases. In this section, we provide some direct evidence on this mechanism. More specifically, we exploit the section of the SHIW questionnaire that asks households to report whether, between 2010 and 2012, they have taken out a loan for vehicles purchase (question D40.c) or a loan for the purchase of other durable goods (question D40.d). We use these binary indicators as the dependent variables of two separate probit regressions that feature all the covariates used in the previous specifications as well as two dummies that take the value of one if a household paid the IMU tax on the main dwelling (on other residential properties) and zero otherwise. The results are reported in Table 10. The estimates in the first column reveals that the probability of having taken a new loan for vehicles purchase is about 17% lower for IMU tax payers on the main dwelling whereas paying the tax on other residential properties is associated with little difference across households. Furthermore, this pattern appears unique to vehicles as in the second column there is only a modest and insignificant gap in the probabilities of having taken a new loan for other durables purchase between payers and nonpayers of the property tax. Given that buying a vehicle typically requires a far smaller downpayment than buying other durable goods, the evidence in Table 10 appears supportive of the mechanism described in Section 4.4. Table 10. Probit regressions on the role of credit. Vehicles debt Other durables debt IMU main dummy −0.17** 0.03 [0.10] [0.22] IMU other dummy 0.04 0.37 [0.10] [0.23] Observations 4,002 4,002 R2 0.08 0.10 Vehicles debt Other durables debt IMU main dummy −0.17** 0.03 [0.10] [0.22] IMU other dummy 0.04 0.37 [0.10] [0.23] Observations 4,002 4,002 R2 0.08 0.10 Notes: Robust standard errors clustered by regions in brackets. The dependent variable is a dummy that takes value of one for households taking a loan for vehicle purchase (column (1)) or a loan for other durable purchase (column (2)) and zero otherwise. IMU “main” dummy and “other” dummy are dummy variables that take the value of one if a household paid the tax on the main dwelling and other properties, respectively. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies (omitted in this regression for collinearity issues—including the dummies does not alter the results but lowers the total number of observations), (iii) dummies of main dwelling commercial area (omitted in this regression for collinearity issues—including the dummies does not alter the results but lowers the total numer of observations), and (iv) expectations about household income and local house prices. **Significant at 7%. View Large Table 10. Probit regressions on the role of credit. Vehicles debt Other durables debt IMU main dummy −0.17** 0.03 [0.10] [0.22] IMU other dummy 0.04 0.37 [0.10] [0.23] Observations 4,002 4,002 R2 0.08 0.10 Vehicles debt Other durables debt IMU main dummy −0.17** 0.03 [0.10] [0.22] IMU other dummy 0.04 0.37 [0.10] [0.23] Observations 4,002 4,002 R2 0.08 0.10 Notes: Robust standard errors clustered by regions in brackets. The dependent variable is a dummy that takes value of one for households taking a loan for vehicle purchase (column (1)) or a loan for other durable purchase (column (2)) and zero otherwise. IMU “main” dummy and “other” dummy are dummy variables that take the value of one if a household paid the tax on the main dwelling and other properties, respectively. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies (omitted in this regression for collinearity issues—including the dummies does not alter the results but lowers the total number of observations), (iii) dummies of main dwelling commercial area (omitted in this regression for collinearity issues—including the dummies does not alter the results but lowers the total numer of observations), and (iv) expectations about household income and local house prices. **Significant at 7%. View Large 5.3. Asymmetric Expenditure Effects In a recent strand of research, Bunn et al. (2017) and Christelis et al. (2017) show that models of intertemporal choice with occasionally binding constraints or with income uncertainty, precautionary saving and credit limits generate the prediction that the consumption changes associated with transitory negative income shocks should be larger than the consumption changes associated with positive shocks. In the former class of models, negative shocks make households more likely to hit the borrowing constraint; in the latter framework, the liquidity increase associated with positive shocks triggers a smaller response for less-wealthy households because the prudence motive makes the MPC decline with liquidity. Unfortunately, the IMU property tax reform only provides us with negative shocks to disposable income. This implies that—like existing studies on positive shocks only—also our analysis falls short of the ideal empirical setting in which one can observe the spending responses to both positive and negative income changes for the very same household. Yet, it is interesting to note that the MPCs estimated in this paper using negative shocks tend to be some 30% larger than the MPCs out of positive income shocks recorded in earlier work. For instance, Parker et al. (2013) (Table 7, Panel E) report an average MPC of 0.48 for vehicles in the context of the 2008 U.S. income tax rebates whereas, in the top panels of Table 8, we report an average MPC on vehicles of 0.61. Similarly, Aaronson et al. (2012) (Table 4, last row) find that the American households who benefitted most by minimum wage hikes over the period 1992–2008 display a MPC on vehicles of about 1.8 (≈431/237) whereas, in the bottom panels of Table 8, we find that households with debt display a MPC on vehicles of around 2.4. In summary, when compared to the estimates in earlier work (based on positive shocks), our evidence (based on negative shocks) points to the presence of a possible asymmetry in the spending effects of temporary income shocks of different signs, consistent with models of credit constraints and precautionary saving 6. On Demographics, Savings, and other Covariates In this section, we present further results exploring whether other household characteristics may be driving the heterogeneity documented so far. The bottom line is that household head age and income grouping deliver a far milder heterogeneity (if any) than when grouping by mortgage debt position. Furthermore, we show that our results are robust to excluding housing wealth and number of children as covariates as well as to controlling for household income. Finally, our findings are not overturned when we account for heterogeneity in uncertainty and house price expectations. The results based on a sample split into high and low liquid wealth households are reported in I and have been discussed in Section 4.3. 6.1. Grouping by Age and Income The results on the age split are reported in Table K.2 of Online Appendix K where we categorize households into younger and older groups according to whether their head belongs or not to the youngest quartile of the household head age distribution.26 The top panel refers to the full sample whereas the bottom panel focuses on home-owners only. The results reveal that (i) the younger group tends to have a larger marginal propensity to spend than the older group, (ii) this is driven by vehicle purchases, whose estimates appear sharper in the home-owners sample, and (iii) the evidence of heterogeneous responses by age on durable goods is far weaker than when households are grouped according to their debt positions, in a combination of smaller point estimates (in absolute value) and larger standard errors than in Table 7. Although the results in Table K.2 of Online Appendix K are consistent with the evidence in Table 3 that debtors tend to be younger than nondebtors, it also suggests that age appears less likely than debt to be a primitive determinant of the heterogeneity we have documented so far. Furthermore, as noted in Section 3.3, the finding in the left panel of Table K.2 of Online Appendix K that young households paying (more of) housing taxes cut their durable expenditure by more than young households not paying at all (or paying less of) housing taxes suggests that our estimates are unlikely to reflect any possible effect of the Fornero pension reform, which was launched over the same period and should have affected evenly all young households independently of their housing tenure status or the payment of the IMU taxes. The analysis in Table K.3 of Online Appendix K focuses instead on income, using the 75th percentile of the household income distribution to categorize observations in to lower- (bottom 25 percentile) and higher- (top 75 percentile) groups. The MPCs of poorer households tend to be imprecisely estimated and their distribution largely overlap with the distributions of the MPC for the higher income group. The point estimates of the latter are marginally (more strongly) significant for durable (vehicle) expenditure, thought the estimates for either group or for either spending category tend to be far smaller than the coefficients in Table 7. On the other hand, the house price effect appears larger among households with a relatively lower income. Although inspection of Table 3 reveals that debtors tend to have higher income than the rest of the sample, also in Table K.3 of Online Appendix K the inference on the heterogeneous responses to the property taxes is far weaker than the inference one can draw from Section 4. This suggests that—unlike a household’s debt position—age and income appear only weakly correlated with the unobserved characteristics that drive the excess sensitivity of durable expenditure to the income change associated with the IMU property taxes. 6.2. Excluding Housing Wealth and Number of Children A main reason for why a property tax may have a different effect on consumption than an income tax is that the former may also affect house prices, which through a wealth effect may then have a further, indirect effect on spending. Indeed, Oliviero and Scognamiglio (2016) report that the 2011 IMU tax had a significant impact on house prices. This is, of course, not an issue for our baseline specification that controls for changes in housing wealth and focuses on the effects of the changes in household resources generated by the property taxes. In this section, we seek to evaluate whether the IMU tax had also a second round effect on household expenditure through a change in house prices, over and above the effect that we have already documented working through changes in disposable income. In Panel A of Table 11, we present results from a specification that is all alike our baseline model (2) but excludes changes in housing wealth, $${\triangle }$$HP, from the covariates. A comparison of the estimated MPC of −0.44 in Panel A and of −0.43 in Table 6 reveals little evidence of a possible wealth effect of the property taxes on consumer spending. Table 11. Removing HP and number of children. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A: Excluding $${\triangle }$$HP IMU main 0.12 −0.44** 0.13 −0.42* [0.66] [0.18] [0.61] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.02 Panel B: Excluding #children IMU main 0.11 −0.48** 0.09 −0.44** [0.67] [0.19] [0.62] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.23 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A: Excluding $${\triangle }$$HP IMU main 0.12 −0.44** 0.13 −0.42* [0.66] [0.18] [0.61] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.02 Panel B: Excluding #children IMU main 0.11 −0.48** 0.09 −0.44** [0.67] [0.19] [0.62] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.23 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey) “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. In Panel A the number of children is included as control whereas it is removed in Panel B. *Significant at 10%; **significant at 5%. View Large Table 11. Removing HP and number of children. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A: Excluding $${\triangle }$$HP IMU main 0.12 −0.44** 0.13 −0.42* [0.66] [0.18] [0.61] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.02 Panel B: Excluding #children IMU main 0.11 −0.48** 0.09 −0.44** [0.67] [0.19] [0.62] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.23 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A: Excluding $${\triangle }$$HP IMU main 0.12 −0.44** 0.13 −0.42* [0.66] [0.18] [0.61] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.02 Panel B: Excluding #children IMU main 0.11 −0.48** 0.09 −0.44** [0.67] [0.19] [0.62] [0.20] IMU other 0.05 0.07 0.05 0.10 [0.43] [0.09] [0.43] [0.09] Observations 4,002 4,002 3,122 3,122 R2 0.11 0.02 0.12 0.23 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey) “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. In Panel A the number of children is included as control whereas it is removed in Panel B. *Significant at 10%; **significant at 5%. View Large In Panel B of Table 11, we remove also the number of children from the list of controls in the specification behind the estimates in Panel A. The reason for this choice is that households received a €50 deduction per child of age below 26 years (up to an amount of €400) from their property tax bill. As the decision of having a child is predetermined relative to the reintroduction of the IMU tax, differences in the number of children across households represent a further independent source of variation to assess the effect of property taxes on consumer spending through this demographic channel. A comparison of the results from Panel B with the results in Panel A and in Table 11 suggests that the demographic channel makes a small contribution (namely the difference between −0.48 in Panel B and −0.44 in Panel A) to the overall effect of the property tax, although this appears to be statistically insignificant. 6.3. The Response of Savings In this section, we shed lights on the resources that nondebtors had to rely upon to pay their IMU taxes. In Table K.5 of Online Appendix K, we use the same regressors of our baseline specification but now have as dependent variable a measure of savings, defined as the fitted values of a projection of the difference between after-tax income and total expenditure on the answers to the question “Were you able to save this year? If so, how much did you save?”.27 Three main results emerge from this exercise. First, the reduction in savings for nondebtors in columns (1) and (3) of Table K.5 of Online Appendix K are large, significant and not statistically different from minus one. Second, the IMU tax paid on the main dwelling had a positive but insignificant impact on debtors’ savings, consistent with the finding of a larger-than-one MPC in Table 7. Third, the IMU taxes on other residential properties paid by the small number of debtors with more than one property were associated with a marginally significant reduction in savings, which was however not statistically different from (and less precisely estimated than) the corresponding responses for debtors. The evidence in Table K.5 of Online Appendix K reveals that the IMU taxes had little impact on the savings of mortgagors with only one property but made the rest of Italians drive significantly down their savings to resist any expenditure cut. As the tax on other residential properties was associated with saving reductions also among mortgagors, we conclude that the lower incidence of debt repayments on disposable income (shown in Figure 4) has likely made mortgagors with more than one property better placed (than mortgagors with only one property) to cope with an unanticipated negative income shock. On the other hand, the behavior of home-owners without debt appears consistent with Ricardian equivalence and the absence of liquidity constraints. 6.4. The Response of Income and Hours Worked In this section, we explore whether the IMU property taxes generated any general equilibrium effect in the local economy or stimulated a labor supply response. To this end, we report in Table K.4 of Online Appendix K the results from two regressions that are all alike our baseline specification (2) but the left hand side variable, which in turn becomes changes in income (left panel) and changes in hours worked. On the one hand, the estimates in Table K.4 of Online Appendix K reveal that neither household income nor hours worked respond significantly to the property taxes on either the main or other residential properties. On the other hand, changes in house prices display a strongly significant association with changes in income. The latter evidence is consistent with the notion that an external factor may be driving simultaneously both variables, with possible implications for the interpretation of the coefficient on house prices in the nondurable goods and service consumption regressions of Section 4. This hypothesis is explored in what follows. 6.5. Controlling for Household Income In Table K.6 of Online Appendix K, we run the specification (2) behind Table 6 but adding income change as a further control. This exercise makes clear that the estimates of the MPCs out of either property taxes are not sensitive to this addition. It is worth noting, however, that the coefficient on house prices in the nondurables regressions of columns (1) and (3) of Table K.6 in Online Appendix K gets halved relative to the Table 6 counterparts, based on a specification without income changes. We interpret the magnitude of the coefficient on $${\triangle }$$HP in Table K.6 of Online Appendix K as reflecting a genuine wealth effect and conjecture that the difference between the estimated house price coefficients in Table K.6 of Online Appendix K and Table 6 reflect a common factor driving both income and house price changes 6.6. The Role of Uncertainty In Table K.7 of Online Appendix K, we explore whether a different degree of uncertainty about the future resources available at the household level implies heterogeneous expenditure responses. In keeping with the dummies on income and house price expectations used in the baseline regressions, the “lower uncertainty” group comprises households who assigned a probability above 70% to any of the possible outcomes indicated in each question for at least one of the questions about (i) future household income, (ii) future local house prices, and (iii) future levels of the stock market index. Conversely, the “higher uncertainty” group is made of households who assigned probabilities equal or below 70% to all possible outcomes indicated in each of these three questions.28 Three main results emerge from this exercise. First, the coefficients on IMU main for nondurable, durable, and vehicles expenditure in the “lower uncertainty” group are never statistically different from their “higher uncertainty” group counterparts and, if any, they imply MPC point estimates that are slightly larger than those for the latter group. Second, the effects of the main dwelling tax on durable and vehicles expenditure in Table K.7 of Online Appendix K appears far smaller and less significant than the effects estimated in Table 7 across household debt positions, consistent with the fact that mortgagors appear evenly spread across the two uncertainty groups. Third, the IMU tax on other residential properties is never statistically different from zero whereas the house price effect working through changes in housing wealth is only slightly larger for households reporting a higher degree of uncertainty 7. Conclusions This paper offers an unprecedented evaluation of the heterogeneous effects of property taxes on consumer spending using a large and unanticipated tax hike on housing wealth, which took place in Italy at the end of 2011. Our analysis reveals that the taxes paid on the main dwelling triggered a large and very significant decline in household expenditure whereas the taxes paid on other residential properties caused a small and statistically insignificant change. The adjustment was mostly borne by home-owners with mortgage debt, who hold low liquid wealth relative to income and whose expenditure is therefore very sensitive to changes in household resources. Of independent interest, we show that our empirical results offer support for the predictions of newer theories of hand-to-mouth behavior that emphasize the role of household debt in the transmission of macroeconomic policies and shocks. Although the property tax change may have also generated nonnegligible general equilibrium effects, we can use our estimates together with data from national statistics reported in Figure 8 to provide some back of the envelope calculations for the direct effect of the legislated changes on the aggregate economy in 2012 along the lines of Johnson et al. (2006). The tax revenues on the main dwelling (other residential properties) for 2012 totaled €4.0 billion (10.7) or 0.3 (0.6)% of GDP. Bearing in mind an average marginal propensity to spend of 0.43 for the main dwelling tax and a coefficient statistically indistinguishable from zero for the other properties tax (see column (2) in Table 6), the direct recessionary effect of the IMU taxes on the Italian economy in 2012 was about 0.11% of GDP (or 0.21% of Personal Consumption Expenditure, PCE) vis-à-vis a tax revenue expansion around 0.90% of GDP (or 1.71% of PCE). Figure 8. View largeDownload slide Evolution over time of key variables. The series “GDP” refers to real Gross Domestic Product estimated by ISTAT (Italian National Institute of STATistics). The series “Durable” and “Nondurable” refer to real households consumption of durable and nondurable goods estimated by ISTAT. Finally, the series “Vehicles” refers to the total number of cars (new and used) and motorbikes (new and used) sold. Source: Authors’ own calculations on ISTAT data (available at http://www.istat.it), and ACI (“Automobile Club d’Italia”) data (available at: http://www.aci.it/laci/studi-e-ricerche/dati-e-statistiche/auto-trend.html). Figure 8. View largeDownload slide Evolution over time of key variables. The series “GDP” refers to real Gross Domestic Product estimated by ISTAT (Italian National Institute of STATistics). The series “Durable” and “Nondurable” refer to real households consumption of durable and nondurable goods estimated by ISTAT. Finally, the series “Vehicles” refers to the total number of cars (new and used) and motorbikes (new and used) sold. Source: Authors’ own calculations on ISTAT data (available at http://www.istat.it), and ACI (“Automobile Club d’Italia”) data (available at: http://www.aci.it/laci/studi-e-ricerche/dati-e-statistiche/auto-trend.html). As for the specific categories driving the aggregate result, our evidence points to a large drop in car expenditure: the introduction of the IMU tax led to a fall of about 11% in vehicle purchases over 2012. We conclude that although the short-run direct cost (in the form of foregone consumer spending) of the property taxes for the Italian economy was small relative to the amount of extra taxes raised, the negative consequences for the car industry were significant. This is consistent with the pattern in Figure 8: the decline in vehicle expenditure (red broken line) during 2012 (shaded area) appears abrupt and more pronounced than the steady decline visible in any other year since the Great Recession of 2007–2008. As for policy implications, the present analysis contributes to two important debates on the design of fiscal policy and housing taxes in particular. More specifically, our evidence suggests that setting a (carefully implemented) multi-year plan of higher property tax rates for nonowner-occupied dwellings as well as providing owner-occupier mortgagors with property tax deductions related to their level of outstanding debt (as currently done for instance in The Netherlands, Spain and Switzerland among other countries) can generate sizable government revenues over a relatively short period of time while minimizing the contractionary effects that levying a property tax may otherwise induce. Furthermore, our analysis provides both an instance in which a policy measure is highly recessionary (when borne by households with debt) but also another instance in which the same type of intervention is not recessionary at all (when borne by households without debt). This suggests that the decisions of what specific group(s) of society to target could (and perhaps should) become another relevant dimensions along which to evaluate the effectiveness and desirability of policy measures that are likely (if not meant) to influence consumers’ behavior. Footnotes 1 Respectively: “decreto legislativo 11 luglio 1992, n. 333” and “decreto legislativo 30 dicembre 1992, n. 504”. 2 The ICI (then IMU) was a yearly tax on housing wealth as measured by the land registry rental value of the property (based on its main characteristics of location and size). As such, it is conceptually (and administratively) very different either from a transaction tax on the sale price such as the British stamp duty or from a yearly tax on housing services (based on property characteristics) such as the British council tax. More specifically, and despite similarities in the way the tax base is imputed, a main difference between the Italian IMU and the British council tax is that the former is a progressive tax charged to the home-owner whereas the latter is a regressive tax charged to the occupier (independently of its housing tenure status). On the other hand, the Italian IMU appears conceptually closer to the American (local) property tax that is paid yearly by the owner on the basis of a periodically reassessed property value and the tax rates set by jurisdictions below the state level. 3 The law was officially passed on December 24th (“Legge 24 Dicembre 2007, n. 244”) and published on the “Gazzetta Ufficiale” on December 28th (“Gazzetta Ufficiale 28 Dicembre 2007”). 4 Law December 22, 2011, n. 24 (published on the “Gazzetta Ufficiale” on December 27, 2011, n. 300). 5 Source: “IFEL” (“Institute for Local Economics and Finance”—“Istituto per la Finanza e l’Economia Locale”) database (accessible at: http://www.webifel.it/ICI/AliquoteIMU.cfm). 6 The direct benefits for the fiscal position of the central government (in the form of either higher direct revenues or lower transfers to the municipal governments) totaled to about two thirds of the overall increase in the property taxes raised. To the extent that most municipal governments used the changes in IMU revenues to reduce their deficits (as shown in what follows), the consolidated balance sheet of the central government—which includes the net fiscal positions of all levels of governments—improved in 2012 by an amount close to the overall IMU revenue increase of around 0.90% of GDP. To give a sense for the magnitude of this intervention, we have calculated that a 1% increase in VAT could possibly generate a maximum increase in tax revenues of about 0.25% of GDP, under the assumption of no change in aggregate demand. 7 Because of an ambiguity in the wording of the SHIW question D37, however, this share is likely to overestimate significantly the proportion of households who regarded the tax change as permanent. The reason is that even respondents who either expected a significant reduction in at least one of the two property rates (as several political parties indicated during 2012) or expected a longer time horizon for the elimination of the IMU could have possibly responded “zero” to the specific question “In your opinion, which is the probability that the Municipal Property Tax (IMU) will be abolished within the next 5 years and not replaced by another similar tax?” (question D37 of the 2012 SHIW survey). 8 Data are collected by the US Air Force Weather Agency and distributed by the National Geophysical Data Center (accessible at: http://ngdc.noaa.gov/eog/dmsp/downloadV4composites.html). Technically, the nighttime lights is derived from the average visible band digital number (DN) of cloud-free light detections multiplied by the percent frequency of light detection. The inclusion of the percent frequency of detection term normalizes the resulting digital values for variations in the persistence of lighting. The Arc-GIS software used to elaborate the raster automatically calculates the average density of all pixels within a municipal territory on a continuous scale between 0 (“low density”—dark) to 62 (“high density”). Note that labor market data such as employment or unemployment are not available at the municipal level 9 By the budget law n. 488 in 1998, amended by the law n. 200 of December 13, 2010, the “Internal Stability Pact” assigns Italian municipalities with fiscal targets that ensure the country as a whole meet the fiscal targets imposed by the European Union. For 2011 (2012), the fiscal balance was imposed to be at least equal to the fixed amount of 14% (16%) of the “historical” expenditure, which was the average of 2006–2008 current expenditure. 10 The share of inherited dwelling is 29.5% for the main dwellings and 53.7% for other residential properties (as estimated using the variable “poss3” in the database “immp2012.dta”). All these figures (including the 13.7% share of mortgagors) refer to the unweighted averages of the regression sample. 11 In 2010 the median net wealth of Italian households was well above the euro area average and almost the double then the median in Germany—see IMF (2013). Also, the proportion of households with debt in Italy was less than half with respect to Spain, Germany, and France (see IMF 2013, p. 5). The full International Monetary Fund (IMF) report is available at: http://www.imf.org/external/pubs/ft/scr/2013/cr13348.pdf. 12 Unfortunately, the question on the amount of taxes paid on other residential properties was not asked in the 2010 SHIW. 13 Because the IMU tax paid in 2010 was zero, equation (1) is specified in differences rather than log-differences. 14 In Online Appendix F, we provide details on the construction of these variables. 15 Similar results are obtained using the change in vehicle expenditure between 2008 and 2010 as dependent variable. 16 For completeness, we also estimate a set of specifications including as only tax regressor $$\textit{IMUtotal}_{i}=\textit{IMUmain}_{i}+ \textit{IMUother}{}_{i}$$. The estimated coefficients are typically between −0.01 and −0.06 (reflecting the larger variation and larger incidence of $$\textit{IMUother}_{i}$$) whereas the standard errors are very high in all columns, consistent with the pervasive heterogeneity that we will show in what follows across the two taxes ($$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$). 17 To assess the influence of any possible under-reporting, we have verified that our findings are not sensitive to adjusting household expenditure on either nondurable goods and services or durable goods by the ratio between the corresponding aggregate variable from national statistics and its SHIW counterpart, which was aggregated using household weights. 18 Mortgage debt represents on average around 70% of total household debt in Italy. The majority of this is secured against the main dwelling with a typical loan-to-value around 50%. About half of all mortgages are on fixed rates but we have verified that the results in this section are robust to using mortgagors with either variable- or fixed-rate products only, though the standard errors increase substantially due to the very few number of observations in each subgroup. 19 Net liquid wealth is defined as the difference between total financial assets (variable “af” in dataset “ricfXX.dta”) minus total financial liabilities (variable “pf” in dataset “ricfXX.dta”). Disposable income refers to the variable “y2” in dataset “consXX.dta”. The suffix “XX” indicates the year of the survey. 20 Similar results are obtained replacing liquid wealth to income with liquid wealth to spending or with liquid wealth to illiquid (real estate) wealth. These results are reported in Tables I.1 and I.2 in Online Appendix I, respectively. 21 Table K.1 of Online Appendix K, and in particular the p-values for the null hypothesis of the interaction terms being equal to zero, provides formal statistical evidence in favor of the grouping strategy based on household debt. 22 According to a survey commissioned in 2015 by UnipolSai, the largest insurance group in Italy, around 70% of Italian households buy a car on credit whereas less than 30% reported to use credit to purchase other durable goods or services. 23 The complete list of “durable goods” included in question E02 of the survey is as follows: jewelry, coins, or gold, artwork, antiques including antique furniture, cars, motorcycles, caravans, boats, bicycles, furniture, furnishings carpets, lamps, small appliances, washing machines, dishwashers, vacuum cleaners, floor polishers, TVs, PCs, refrigerators, cookers, stoves, air conditioners, radio, video recorder, CD players, stereos, phones, fax machines, and cameras. 24 The first semester of 2016 has been the first period since the 2011 IMU reform in which most Italians have not paid the property tax on the main dwelling, which the government led by Mr. Renzi abolished during 2015. Although a systematic analysis of the effect of this latter property tax change must necessarily wait for the availability of new expenditure data (especially from the 2016 wave of the SHIW), we note that—consistent with our findings—during the 2016H1, the car industry in Italy has witnessed a strong performance with the highest growth rate in ten years. 25 The average cost of new vehicles is estimated as mean of variable “cd1” in the SHIW survey, considering only the observations above €5,000 in order to depurate this figure from motorbikes expenditure, which are not recorded in the ACI data. 26 Neither for age nor for income, results are significantly different using any other percentile between 60 and 90 as cutoff. 27 In the 2012 survey, we rely on the questions C42, C43, and C44. Question C42 asks “Please consider all of the sources of income for your household that you have told me about during this interview (employment income, rent, income from capital, etc.). Could you tell me if in 2012 your household (i) spent its entire yearly income and did not manage to save anything, (ii) spent less than its entire yearly income and succeeded in saving, (iii) spent more than its entire yearly income, drawing on savings or borrowing”. Question C43 asks “About how much did you save in 2012?”. Finally, question C44 asks “About how much more than your income did you spend in 2012?”. In the 2010 survey the questions were respectively the C43, C44, and C45. To the extent that more affluent households tend to under-report their annual income (but do not necessarily under-report their annual savings), this projection would isolate the common variation across the two series (their correlation is 0.37). To control for the pre-existing liquidity position, we also add household income in 2010 as additional regressor. 28 Similar results are obtained using a 60% or a 80% threshold. Acknowledgements We are grateful to Stephanie R. Aaronson, Benjamin Born, Nicola Borri, James Cloyne, Luigi Falcioni, Nicola Gennaioli, Luigi Guiso, Joanne Hsu, Dirk Krueger, Claudio Michelacci, Kurt Mitman, Elias Papaioannou, Richard Portes, Helene Rey, Andrea Stella, Amir Sufi, Hans-Joachim Voth, participants to the seminars at the Board of Governors of the Federal Reserve System, LBS, LSE, PSE, Brunel University, European Summer Symposius In Macroeconomics, the Society for Economic Dynamics (SED) meeting, and the Macroeconomic Dynamics Workshop at Bocconi University and six anonymous referees for very useful comments. We thank Francesca Proia and the IFEL institute for providing the data on the property tax rates and Felix Galbis-Reig for an outstanding support to collect municipal budget data. Surico gratefully acknowledges financial support from the European Research Council (Starting Grant 263429 and Consolidator Grant 647049). All errors and omissions remain ours. Disclaimer: the views expressed in this paper are those of the authors and do not necessarily reflect those of the Board of Governors or the Federal Reserve System. Surico is a Research Fellow at CEPR. Notes The editor in charge of this paper was Dirk Krueger. References Aaronson Daniel , Agarwal Sumit , French Eric ( 2012 ). “The Spending and Debt Response to Minimum Wage Hikes.” American Economic Review , 7 ( 102 ), 3111 – 3139 . Google Scholar CrossRef Search ADS Acconcia Antonio , Corsetti Giancarlo , Simonelli Saverio ( 2015 ). “The Consumption Response to Liquidity-Enhancing Transfers: Evidence from Italian Earthquakes.” CSEF Working Paper No. 396, Centre for Studies in Economics and Finance, University of Naples , Italy . Agarwal Sumit , Qian Wenlan ( 2014 ). “Consumption and Debt Response to Unanticipated Income Shocks: Evidence from a Natural Experiment in Singapore.” American Economic Review , 104 ( 12 ), 4205 – 4230 . Andrés J. , Bosca J. E. , Ferri J. ( 2011 ). “Household Leverage and Fiscal Multipliers.” Working Paper No. 1103, International Economics Institute, University of Valencia . Bocci Chiara , Iommi Sabrina , Marinari Donatella ( 2012 ). “Imu più equa con i valori di mercato.” Tech. rep., IRPET . Born Benjamin , Müller Gernot , Pfeifer Johannes ( 2014 ). “Does Austerity Pay Off?” SAFE Working Paper Series 77, Research Center SAFE -- Sustainable Architecture for Finance in Europe, Goethe University Frankfurt. Browning Martin , Crossley Thomas ( 2000 ). “Luxuries Are Easier to Postpone: A Proof.” Journal of Political Economy , 5 , 1022 – 1026 . Google Scholar CrossRef Search ADS Bunn Philip , Roux Jeanne Le , Reinold Kate , Surico Paolo ( 2017 ). “The Consumption Response to Positive and Negative Income Changes.” Working Paper No. 645, Bank of England . Google Scholar CrossRef Search ADS Calza Alessandro , Monacelli Tommaso , Stracca Livio ( 2013 ). “Housing Finance and Monetary Policy.” Journal of the European Economic Association , 11 , 101 – 122 . Google Scholar CrossRef Search ADS Campbell John Y. , Cocco Joao F. ( 2007 ). “How do House Prices Affect Consumption? Evidence from Micro Data.” Journal of Monetary Economics , 54 , 591 – 621 . Google Scholar CrossRef Search ADS Christelis Dimitris , Georgarakos Dimitris , Jappelli Tullio , Pistaferri Luigi , van Rooij Maarten ( 2017 ). “Asymmetric Consumption Effects of Transitory Income Shocks.” CSEF Working Paper No. 467, Centre for Studies in Economics and Finance, University of Naples , Italy . Cloyne James , Surico Paolo ( 2017 ). “Household Debt and the Dynamic Effects of Income Tax Changes.” Review of Economic Studies , 84 ( 1 ), 45 – 81 . Google Scholar CrossRef Search ADS D’Aurizio Leandro , Faiella Ivan , Iezzi Stefano , Neri Andrea ( 2006 ). “The Under-Reporting of Financial Wealth in the Survey on Household Income and Wealth.” Temi di discussione (Economic Working Papers) 610, Economic Research and International Relations Area, Bank of Italy . Google Scholar CrossRef Search ADS Deaton Angus ( 1992 ). “Understanding Consumption.” In OUP Catalogue . Oxford University Press . Dynan Karen ( 2012 ). “Is a Household Debt Overhang Holding Back Consumption.” Brookings Papers on Economic Activity , 44 , 299 – 362 . Google Scholar CrossRef Search ADS Eggertsson Gauti B. , Krugman Paul ( 2012 ). “Debt, Deleveraging, and the Liquidity Trap: A Fisher-Minsky-Koo Approach.” Quarterly Journal of Economics , 127 , 1469 – 1513 . Google Scholar CrossRef Search ADS Grant Charles , Peltonen Tuomas A. ( 2008 ). “Housing and Equity Wealth Effects of Italian Households.” Working Paper No. 0857, European Central Bank . Grembi Veronica , Nannicini Tommaso , Troiano Ugo ( 2012 ). “Policy Responses to Fiscal Restraints: A Difference-in-Discontinuities Design.” IZA Discussion Paper No. 6952, Institute for the Study of Labor (IZA) . Guiso Luigi , Paiella Monica , Visco Ignazio ( 2005 ). “Do Capital Gains Affect Consumption? Estimates of Wealth Effects from Italian Households' Behavior.” Temi di discussione (Economic Working Papers) 555, Economic Research and International Relations Area, Bank of Italy . Google Scholar CrossRef Search ADS IMF ( 2008 ). “The Changing Housing Cycle and the Implications for Monetary Policy.” Tech. rep., International Monetary Fund . IMF ( 2012 ). “Italy: Technical Assistance Report—The Delega Fiscale and the Strategic Orientation of Tax Reform.” Tech. rep., International Monetary Fund . IMF ( 2013 ). “Technical Note on the Financial Situation of Italian Households and Nonfinancial Corporations and Risks to the Banking System.” Tech. rep., International Monetary Fund . Jappelli Tullio , Pistaferri Luigi ( 2014 ). “Fiscal Policy and MPC Heterogeneity.” American Economic Journal: Macroeconomics , 6 , 107 – 136 . Google Scholar CrossRef Search ADS Johnson David S. , Parker Jonathan A. , Souleles Nicholas S. ( 2006 ). “Household Expenditure and the Income Tax Rebates of 2001.” American Economic Review , 96 ( 5 ), 1589 – 1610 . Google Scholar CrossRef Search ADS Kaplan Greg , Violante Gianluca , Weidner Justin ( 2014 ). “The Wealthy Hand-to-Mouth.” 2014 Meeting Paper No. 192, Society for Economic Dynamics. https://ideas.repec.org/p/red/sed014/192.html . Kaplan Greg , Violante Giovanni L. ( 2014 ). “A Model of the Consumption Response to Fiscal Stimulus Payments.” Econometrica , 82 , 1199 – 1239 . Google Scholar CrossRef Search ADS Kiyotaki Nobuhiro , Moore John ( 1997 ). “Credit Cycles.” Journal of Political Economy , 105 , 211 – 248 . Google Scholar CrossRef Search ADS Mian Atif , Rao Kamalesh , Sufi Amir ( 2013 ). “Household Balance Sheets, Consumption, and the Economic Slump.” Quarterly Journal of Economics , 128 , 1687 – 1726 . Google Scholar CrossRef Search ADS Mian Atif , Sufi Amir ( 2016 ). “Who Bears the Cost of Recessions? The Role of House Prices and Household Debt.” NBER Working Paper No. 22256, National Bureau of Economic Research, Cambridge, MA. Misra Kanishka , Surico Paolo ( 2014 ). “Consumption, Income Changes, and Heterogeneity: Evidence from Two Fiscal Stimulus Programs.” American Economic Journal: Macroeconomics , 6 , 84 – 106 . Google Scholar CrossRef Search ADS Mitman Kurt ( 2012 ). “Macroeconomic Effects of Bankruptcy and Foreclosure Policies.” 2012 Meeting Paper No. 563, Society for Economic Dynamics, Minneapolis . Norregaard John ( 2013 ). “Taxing Immovable Property Revenue Potential and Implementation Challenges.” IMF Working Paper No. 13/129, International Monetary Fund, Washington, DC . Google Scholar CrossRef Search ADS Oliviero Tommaso , Scognamiglio Annalisa ( 2016 ). “Property Tax and Property Values: Evidence from the 2012 Italian Tax Reform.” CSEF Working Paper No. 439, Centre for Studies in Economics and Finance (CSEF), University of Naples , Italy , https://ideas.repec.org/p/sef/csefwp/439.html . Google Scholar CrossRef Search ADS Paiella Monica , Pistaferri Luigi ( 2017 ). “Decomposing the Wealth Effect on Consumption.” The Review of Economics and Statistics , 99 , 710 – 721 . Google Scholar CrossRef Search ADS Parker Jonathan A. , Souleles Nicholas S. , Johnson David S. , McClelland Robert ( 2013 ). “Consumer Spending and the Economic Stimulus Payments of 2008.” American Economic Review , 103 ( 6 ), 2530 – 2553 . Google Scholar CrossRef Search ADS Piazzesi Monika , Schneider Martin ( 2016 ). “Housing and Macroeconomics.” NBER Working Paper No. 22354, National Bureau of Economic Research, Cambridge, MA. Ragot Xavier ( 2014 ). “The Case for a Financial Approach to Money Demand.” Journal of Monetary Economics , 62 , 94 – 107 . Google Scholar CrossRef Search ADS Zeldes Stephen P. ( 1992 ). “‘Housing Markets, Consumption and Financial Liberalisation in the Major Economies’: by D. Miles.” European Economic Review , 36 , 1131 – 1135 . Google Scholar CrossRef Search ADS Supplementary Data Supplementary data are available at JEEA online. Published by Oxford University Press on behalf of European Economic Association 2018. This work is written by (a) US Government employee(s) and is in the public domain in the US. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Journal of the European Economic Association Oxford University Press

Consumer Spending and Property Taxes

Loading next page...
 
/lp/ou_press/consumer-spending-and-property-taxes-ryeiGnFP6n
Publisher
Oxford University Press
Copyright
Published by Oxford University Press on behalf of European Economic Association 2018.
ISSN
1542-4766
eISSN
1542-4774
D.O.I.
10.1093/jeea/jvy008
Publisher site
See Article on Publisher Site

Abstract

Abstract A sudden and temporary change to the Italian property tax system in 2011 generated significant variation in the amount of taxes paid across home-owners. Using new questions appositely added to the Survey on Household Income and Wealth, we exploit this cross-sectional variation to provide an unprecedented analysis of the consumption effects of a tax on housing wealth. A tax hike on the main dwelling leads to large expenditure cuts among mortgagors, who hold low liquid wealth despite owning sizable illiquid assets. In contrast, higher tax rates on other residential properties affect affluent households, thereby having a modest impact on their consumer spending. Our results provide novel and direct evidence in favor of recent theories that highlight the role of household debt in the transmission of economic policies. “Household debtors are frequently young families acquiring homes and furnishing before they earn incomes to pay for them outright; given the difficulty of borrowing against future wages, they are liquidity-constrained and have a high marginal propensity to consume.” (James Tobin, 1980, “Asset accumulation and economic activity”, p. 10) 1. Introduction What are the effects of housing taxes on consumer spending? And what groups of society bear most of its costs? Although a large body of research has made notable progress to quantify the effects of tax rebates, little is known about the impact of tax hikes on consumer spending and about whether property taxes are likely to distort household behavior less than income taxes. This seems particularly surprising in the light of a growing macroeconomic literature—surveyed by Mian and Sufi (2016, Chap. 5) and Piazzesi and Schneider (2016)—advocating a key role for housing over the business cycle. The present analysis fills this important gap in the literature exploiting the 2011 changes in the Italian property tax “Imposta Municipale Unica” (“IMU”). A newly appointed central government swiftly legislated and implemented a sizable fiscal consolidation plan whose main intervention temporarily redesigned the municipal system on housing taxes. The changes raised around €4.0 billion from taxes on the main dwelling and an additional €10.1 billion from taxes on other residential properties, for a total revenue increase of 0.90% of GDP. The IMU affected 25.8 million tax payers (or around 70% of households) with an average contribution per tax-paying household around €357 on the main dwelling and about €905 on other residential properties. Using a new set of questions (on the amount of IMU paid) appositely added to the Survey of Household Income and Wealth (SHIW) conducted by the Bank of Italy, we identify the effects of property taxes on consumer spending. We do so by comparing the difference in expenditure for IMU payers before and after the tax change to the difference in expenditure for non-IMU payers over the same period, employing a range of specifications that control for demographics, changes in house value, property characteristics, expectations on future household income, and expectations on future local house prices as well as regional fixed effects. In the most restrictive specifications, we look at home-owners only and therefore we focus exclusively on variation in the amount of property taxes paid. Our identification strategy builds upon four features of the 2011 changes in the municipal system of housing taxation in Italy. First, the central government introduced a new tax on the main dwelling and increased by an exogenous factor the (by then obsolete) land registry estimates of the rental values used to calculate the tax base for the main dwelling and other residential properties. Second, the timing and depth of the legislated changes were largely unanticipated. Third, municipal authorities were allowed to unilaterally modify the rates proposed by the central government and, as shown in Section 2, the geographical variation in property tax rates appears driven by political motives that were unrelated to past local economic conditions or other local economic policies. Fourth, the IMU tax changes were announced by the government as an experiment (whose possible extension would have been subject to government revision) and most SHIW respondents did not expect the changes to persist longer than five years. Indeed, the housing tax on the main dwelling was subsequently abolished in 2015. A household-level approach appears to offer two main advantages relative to a more macro strategy that relates changes in central government tax revenues to changes in aggregate consumption. First, macroeconomic interventions—like a change in residential property taxes—are often the endogenous policy response to conditions in the aggregate economy, thereby posing a reverse causality problem when using data from national statistics. In contrast, the cross-household and cross-municipality variation in property tax rates that we implicitly exploit for identification on micro data seems unlikely to be the policy response to specific circumstances at the individual household level, especially after controlling for demographics and property characteristics as we do here. On the other hand, aggregate circumstances or the effects of other economic policies may confound the evaluation of the impact of the 2011 property tax changes on household expenditure behavior. However, an extensive analysis in Section 3 reveals that the amount of IMU taxes paid is not systematically related to the household variables that were directly affected by other policy changes over the same period. A second main advantage of using survey data is that they allow us to explore potentially interesting dimensions of heterogeneity across liquid holdings and household debt positions, so as to shed light on the specific channel(s) of policy transmission. The empirical analysis isolates five major empirical regularities. First, the marginal propensity to consume (MPC) nondurable goods and services out of the IMU tax is around 0.05 whereas the MPC on durable goods is about 0.43. Second, these average effects mask pervasive heterogeneity across residential properties, with the taxes paid on the main dwelling associated with a large and significant MPC on durable goods and the taxes on other residential properties associated with a small and insignificant MPC on durable goods. In contrast, the MPCs on nondurable goods and services are statistically indistinguishable from zero, both across residential properties and across household groups. Third, the significant response to the main dwelling IMU tax is far more pronounced among home-owners with mortgage debt, who are shown to hold low liquid wealth relative to income despite owning sizable illiquid assets and thus appear to fit well the notion of “wealthy” hand-to-mouth consumers. Fourth, debtors concentrated their cuts on vehicles expenditure. Fifth, the direct negative consequences of the changes in the IMU residential property taxes are estimated to be around 0.11% of GDP in 2012 vis-à-vis an increase in tax revenues of 0.90% of GDP (or 1.80% of government revenues). On the other hand, the direct impact of the property taxes on the car industry was large, making a negative contribution around 11% (or about half of the overall decline) relative to the market size in 2011. Finally, the evidence in this paper compares favorably with a long standing tradition in economics that has advocated the use of age and income as proxies for the presence of liquidity constraints. We show that holding a mortgage is, in fact, a far stronger predictor for the liquidity shortage behind the observed sensitivity of consumption to temporary income changes, therefore highlighting the role of household debt as a novel source of violation of the permanent income hypothesis as well as a powerful amplification mechanism for the transmission of macroeconomic shocks. Contribution and Related Literature. Our analysis seeks to contribute to three main strands of the literature. First, a burgeoning line of theoretical research has emphasized the role that illiquid wealth (and especially housing) could play in the transmission of macroeconomic policies. Selected examples include Eggertsson and Krugman (2012), Kaplan and Violante (2014), Ragot (2014), Mitman (2012), and Andrés, Bosca, and Ferri (2011). Our analysis provides direct evidence in support of these theories by offering an unprecedented evaluation of the household expenditure effects of a housing wealth tax. Second, a theoretical literature pioneered by Browning and Crossley (2000) and extended by Aaronson, Agarwal, and French (2012) generate the testable predictions that not only durable expenditure should react more than nondurable consumption following a temporary change in household resources but also that, among the durable spending categories, goods requiring lower down payments for their purchase on credit—such as vehicles—should exhibit far larger marginal propensities to consume that other durable goods. Our findings speak in favor of this mechanism. Third, an important set of studies pioneered by Johnson, Parker, and Souleles (2006) and investigated further by Parker et al. (2013), Agarwal and Qian (2014), and Jappelli and Pistaferri (2014) look at household expenditure in response to a transitory increase in disposable income. In contrast, the presents analysis focuses on a decrease in disposable income by offering some of the earliest evidence on the consumer spending response to a fiscal austerity measure in Europe. Structure of the Paper. Section 2 describes the institutional design and the cross-sectional variation that we exploit for identification. Section 3 presents the data and the empirical specifications before assessing the role of other confounding factors. The main results on the IMU tax paid on the main dwelling and on other residential properties as well as the heterogeneous responses across household balance sheet positions are presented in Section 4, together with evidence that most mortgagors hold very low liquidity relative to income. We conclude this section by sketching a theoretical argument that generates the prediction of a larger response on durable goods expenditure (and vehicle purchases in particular). Estimates for different spending categories and for the role of credit are the focus of Section 5. Further results on age and income splits, the role of uncertainty, house prices and demographics as well as the response of income and hours worked are presented in Section 6. We conclude with some back of the envelope calculations that quantify the direct impact of the IMU tax changes on the Italian economy in 2012. 2. Institutional Design and Geographical Variation In this section, we first outline a brief history of housing taxation in Italy. We then describe the specific context in which the property tax changes were introduced in December 2011 and finally we describe the variation in the IMU rates that we exploit for identification in the econometric analysis. 2.1. A Brief History of Municipal Property Taxes in Italy The “Municipal Tax on Properties” (“Imposta Comunale sugli Immobili”, aka “ICI”) was introduced in the Italian legislation by the law by decree number 333 on July 11, 1992 and subsequently transformed into law on December 30, 1992.1 The ICI tax base included three main categories: buildings, building plots, and farmlands.2 Our analysis on household expenditure will focus on the “buildings” category. Under the ICI system, the tax base for “buildings” was the land registry value defined as an estimate of what the rental value of the property would have been in 1988–1989, which was used as a base biennium. This (rough) estimate, which was self-reported to the municipal registry by the buyer at the time of purchase, was based on the location and building type but did not account for other important dimensions such as the type of construction, the age of the building and more generally for the conditions of the property. Not surprisingly, the system became obsolete soon after its introduction but was left essentially unchanged in the following two decades against the backdrop of steadily growing house prices. In Figure A.1 of Online Appendix A, we show that the ratio of the estimated land registry values to the actual market values at the end of the ICI system averaged around 3.6 (see Bocci, Iommi, and Marinari 2012; IMF 2012 for similar evidence). The property tax rates were set independently by the municipal governments within the range of 0.4%–0.7%, according to local preferences. The ICI remained substantially unchanged until the end of 2007, when the government led by Prime Minister Romano Prodi approved an increase of the basic deduction of 0.133%.3 The policy change applied only to taxes on the main dwelling with a cap of €200. Finally, on March 27, 2008, the subsequent government led by Prime Minister Silvio Berlusconi abolished the ICI tax on the main dwellings (excluding three building categories corresponding to “luxury houses” (category “A1”), “villas” (category “A8”), and “castles” (category “A9”)) with the law by decree number 93/2008 whereas the ICI tax on other properties remained unchanged. 2.2. The “IMU” Tax On December 4, 2011, a newly appointed Italian government led by Prime Minister Mario Monti announced a fiscal consolidation plan that was meant to “ensure fiscal stability, growth and equity”. The plan was passed into law with immediate effect on December 22, 2011.4 Among the most sizable interventions, the government reformed the property tax system, abolished ICI, introduced a single municipal property tax under the heading of “Imposta Municipale Unica” (“IMU”), and presented the policy change to the public as an “experiment”. According to the official technical notes accompanying the law, the introduction of the IMU (which was levied only on property owners) accounted for three quarters of the increase in taxation associated with the 2011 consolidation plan. The swift implementation of Monti’s government IMU reform (in less than two months since the resignation of former Prime Minister Silvio Berlusconi), together with the frequency of the SHIW (conducted in 2010 and 2012), makes these property tax changes most likely unanticipated by households (especially back in 2010). Finally, in line with the government announcement back in 2011, the IMU tax on the main dwellings (subsequently extended to housing services under the new heading of “TASI”) was abolished in July 2015 by the government led by Prime Minister Matteo Renzi. The introduction of the IMU tax significantly reformed the property tax regime along three dimensions. First, it included the land registry value of the main dwelling in the tax base, previously excluded. Second, the land registry values (for both main dwellings and other properties) were scaled up by an exogenous factor (homogeneous across all municipalities and equal to 1.6 for residential dwellings), so as to increase the tax base by an average of 49% (see IMF 2012). Finally, the IMU system set the basic tax rate on primary (other) residences at 0.4% (0.76%) of the registry value but allowed municipalities to modify this rate within a 0.2% (0.3%) band. Furthermore, the government set the basic deduction at €200 plus an additional €50 deduction per children less than 26 years old (up to a maximum of an additional €400): whereas municipalities were allowed to modify this, around 98% of local governments chose the basic deduction of €200.5 Overall, the IMU system determined a sharp increase in residential property taxation: the revenues on the main properties increased from nothing in 2011 to €4.0 billion in 2012 while those on other properties increased from 7.8 billion in 2011 to 17.9 billion in 2012. Between 2011 and 2012, total tax revenues on residential properties increased by €14.1 billion corresponding to around 0.90% of Gross Domestic Product (GDP) in 2012.6 Our analysis exploits the fact that in the 2012 Italian Survey on Household Income and Wealth (SHIW), respondents were appositely asked for the first time to report the amount of recurrent housing taxes paid on both the main dwelling and other residential properties. In Figure 1, we plot the distribution of self-reported IMU payments per household from the SHIW, distinguishing between the amount of housing taxes paid on the main dwelling (in the first row) and the amount of taxes paid on other residential properties (in the second row). The first column displays the distribution of the absolute amount of euros paid whereas the second column reports this as a share of the household monthly income. Because of the deductions, 21.6% of home-owners did not pay the IMU tax on the main dwelling and 13.2% of home-owners with more than one property did not pay the IMU tax on other residential properties. The IMU affected 25.8 millions of tax payers (or around 70% of households). The average payment on the main dwelling was about €357 (or 14% of a household monthly income) whereas the average payment on all residential properties was €905 (or 36% of a household monthly income). It is worth noting that, as shown by Norregaard (2013), it is very hard to evade property taxes in a high-income country like Italy. Finally, about 30% of SHIW respondents reported a zero probability that the IMU tax would have been eliminated within five years and not replaced by another similar tax.7 Figure 1. View largeDownload slide IMU tax burden per household. The figures refer to owners, IMU tax payers only. The red line plots the Epanechnikov kernel density. Panel a1 (a2) refers to the amount paid on main dwellings in Euro per household (as a share of households’ monthly income), excluding 14 observations higher than €3,000. Panel b1 (b2) refers to the amount of IMU tax paid (as a share of monthly income) on other properties, excluding 129 observations higher than €3,000. Source: authors’ calculations on SHIW survey data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 1. View largeDownload slide IMU tax burden per household. The figures refer to owners, IMU tax payers only. The red line plots the Epanechnikov kernel density. Panel a1 (a2) refers to the amount paid on main dwellings in Euro per household (as a share of households’ monthly income), excluding 14 observations higher than €3,000. Panel b1 (b2) refers to the amount of IMU tax paid (as a share of monthly income) on other properties, excluding 129 observations higher than €3,000. Source: authors’ calculations on SHIW survey data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). 2.3. IMU Rate Variation and Local Business Cycle The variation in the amount of IMU tax paid across households stems from three main features of the law: demographics (and in particular the number of children eligible for deduction), property characteristics (including surface and building type, which determine the land registry rental value) and local tax rates (given that municipalities were allowed to vary the rates set by the government). In the SHIW, we observe demographics and property characteristics but—to preserve anonymity—we are only provided with the region (rather than the municipality) where a household lives in. This implies that controlling for demographics and property characteristics in a projection of household expenditure change on the income change stemming from the IMU taxes disbursement is likely to isolate variation in the amount of property taxes paid due either to geographical variation in the municipal tax rates or to unobserved characteristics that are not absorbed by our rich set of covariates. In 2012, 35.2% (57.3%) of municipalities chose to modify the tax rate on the main dwelling (other residential properties) set by the national government, with the vast majority opting for higher rates. In Figures B.1 and B.2 of Online Appendix B, we construct heat maps that illustrates the municipal variation in property tax rates on the main dwelling and other residential properties across the national territory. To interpret the coefficient on IMU paid as the causal effect of the tax change on private expenditure, we need to verify that the geographical variation in the tax rates was not the municipal government response to past local economic conditions. The concern is that property tax rates may have been consistently higher in municipalities with a higher concentration of households with certain (financial and economic) characteristics. To assess this hypothesis, Table 1 reports the correlation between the municipal IMU tax rates of 2012 and a number of indicator of local economic performance available at municipal level in 2010 and 2011, ranging from personal and business income to night light density.8 The main take away from this table is that there is little evidence of a systematic relation between the IMU tax rates and local economic conditions in the preceding years. Table 1. Correlation between IMU rates and local economic activity. Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Notes: Night lights density correlations exclude small municipalities (<5,000 inhabitants) and big cities (>300,000 inhabitants). Other measures of economic activity such as employment or unemployment data are not available at municipal level. a. IMU rates on both main dwelling and other residential properties refer to 2012. View Large Table 1. Correlation between IMU rates and local economic activity. Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Variable Transform Main dwelling ratea Other properties ratea Personal income 2011 Level 0.154 0.121 Personal income 2010 Level 0.145 0.119 Personal income 2010–2011 % Change 0.056 0.016 Business income 2011 Level 0.051 −0.062 Business income 2010 Level 0.056 −0.031 Business income 2010–2011 % Change 0.022 0.004 Night lights density 2011 Level 0.115 0.205 Night lights density 2010 Level 0.125 0.198 Night lights density 2010–2011 % Change 0.021 0.107 IMU rate on other properties 0.323 1 Notes: Night lights density correlations exclude small municipalities (<5,000 inhabitants) and big cities (>300,000 inhabitants). Other measures of economic activity such as employment or unemployment data are not available at municipal level. a. IMU rates on both main dwelling and other residential properties refer to 2012. View Large This latter result is echoed by Figure 2, which records across municipalities the correlations between the share of votes to the left-wing coalition in the local elections immediately before the property tax reform and (i) the IMU tax rates on the main dwelling (top left panel), (ii) the IMU tax rates on other properties (top right panel), (iii) night light density (bottom left panel), and (iv) business income growth (bottom right panel). Although, on the one hand, the top row reveals a significant relation between the tax rates and political orientation, the findings in the bottom panels show that, on the other hand, political orientation is not systematically linked to local economic performance. Figure 2. View largeDownload slide Correlations political orientation—IMU rates—local business cycle. Each dot on the charts represents the average of the respective bin. “Votes to left-wing coalition” refers to the share of votes to the left-wing coalition in regional elections. For most municipalities the latest regional election before the IMU change was in 2010 (March 28th). Source: Authors’ calculations on IFEL data (available at: http://www.webifel.it/ICI/AliquoteIMU.cfm) and Ministry of Interior data. Figure 2. View largeDownload slide Correlations political orientation—IMU rates—local business cycle. Each dot on the charts represents the average of the respective bin. “Votes to left-wing coalition” refers to the share of votes to the left-wing coalition in regional elections. For most municipalities the latest regional election before the IMU change was in 2010 (March 28th). Source: Authors’ calculations on IFEL data (available at: http://www.webifel.it/ICI/AliquoteIMU.cfm) and Ministry of Interior data. In summary, this section contains two messages. First, municipal variation in property tax rates is not related to past local economic conditions. Second, political orientation at the municipal level is an important driver of variation in property tax rates. As there is no systematic difference in the economic performance of center-left and center-right governments across Italian municipalities, we regard this politically-driven variation in property tax rates as exogenous from the standpoint of household expenditure decisions. 2.4. Other Local Economic Policies To isolate the effects of property taxes (as opposed to the effects of other local economic policies), it seems important to establish how municipal governments have employed the extra resources made available by the IMU tax. We do so in this section by focusing on changes (between 2011 and 2012) in the local authorities balance sheets, whose descriptive statistics are reported in Table C.1 of Online Appendix C. More specifically, we project the change in (i) municipal government expenditure, (ii) other municipal tax revenues (net of IMU revenues), (iii) municipal property tax revenues, and (iv) local fiscal deficit on the municipal IMU tax rates using official data from the Ministry of the Interior. To control for the size of each municipality, the dependent variables are standardized by the number of inhabitants. We run these specifications either over the full sample of Italian municipalities or only for those with more than 2,000 inhabitants to ameliorate possible concerns on the quality of the balance sheets in smaller cities. The evidence from Table 2 suggests that there is no significant association between IMU rates and changes in either municipal public spending (in column (1)) or other municipal tax revenues (in column (2)). Also, there is no statistical association (not shown in Table 2) between IMU rates and other local government revenues (grants and dismissals). On the other hand, the estimates in column (3) suggest that increasing the IMU rate on the main dwelling by 0.1 percentage points brings IMU tax revenues up by €8.7 per capita (€9.3 per capita when restricting the sample to municipalities above 2,000 inhabitants). At the same time, an increase of 0.1 percentage points in the IMU rate on the other dwellings generates an average rise in IMU tax revenues of €22.4 per capita, consistent with the fact that IMU revenues on other residential properties are about three times larger than IMU revenues on the main dwelling at the aggregate level. Finally, these extra property tax revenues seem to translate fully into a reduction of the “municipal fiscal deficit” (in column (4)), which is the fiscal target as defined by the central government in the “Internal Stability Pact”.9 It is worth mentioning that the law requires to calculate the municipal fiscal deficit on a “mixed accrual-cash basis”, with current expenditure and revenues evaluated on an accrual basis and capital expenditure and revenues evaluated on a cash basis. For this reason (and because total revenues also include grants and dismissals), the first three columns of Table 2 do not sum up to the last column. Table 2. Other municipal government policies. Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Notes: Each column reports results of a regression where the left hand side variable is the per-capita change of the municipal government instrument in the title between 2012 and 2011 projected on the IMU rates on the main dwelling and other residential properties set by that very municipality in 2012. Municipal public expenditure refers to the variable “Total expenditure” (“Totale generale delle spese”), other municipal tax revenues refers to the sum of all municipal tax revenues net of IMU payments (specifically we consider the following taxes: “Scopo”, “Soggiorno”, “Pubblicita”, “Occupazione degli spazi pubbliche”, “Raccolta e smaltimento dei rifiuti”, “Tassa affissioni”, “Anagrafe”, “Uffici giudiziari”, “Polizia municipale”, “Istruzione elementare”, “Istruzione media”, “Assistenza scolastica”, “Biblioteche”, “Teatri, attività culturali”, “Piscine comunali”, “Stadio comunale, palazzo dello sport”, “Manifestazioni diverse”, “Servizi turistici”, “Viabilità”, “Trasporti pubblici locali”, “Urbanistica”, “Edilizia residenziale pubblica locale”, “Servizio idrico”, “tariffa igiene ambientale”, “Asili nido”, “Proventi servizi di prevenzione e riabilitazione”, “Ricovero per anziani”, “Assistenza, beneficenza pubblica”, “Servizio necroscopico e cimiteriale”, “Entrate da sanzioni amministrative”, “Mezzi pubblicitari”, “Proventi di bene”, “Scuola materna”, “Addizionale IRPEF”, “Segreteria”, “Ufficio tecnico”, “Servizi turistici”, “C.O.S.A.P.”, “Concessioni cimiteriali”). Fiscal deficit calculated as difference between municipal expenditures (current + capital) and municipal revenues (tax revenues + current grants + dismissals). Current revenues and expenditures are on accrual basis, capital revenues and expenditures are on a cash basis. For this reason, the sum of the coefficients in the first three columns cannot equal the coefficient in the last column. Robust standard errors clustered by provinces in brackets. ***Significant at 1%. Source: Ministry of Interior data (publicly available at: http://finanzalocale.interno.it/apps/floc.php/in/cod/4). View Large Table 2. Other municipal government policies. Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Municipal public expenditure Other municipal tax revenues IMU revenues Municipal fiscal deficit Panel A: All municipalities IMU main rate 8.18 −1.85 8.71*** −7.56*** [10.00] [1.84] [1.75] [2.23] IMU other rate 1.85 −2.12 22.40*** −20.81*** [5.85] [1.35] [1.51] [2.21] Observations 7,355 7,355 7,355 7,355 R2 0.02 0.04 0.17 0.06 Panel B: Municipalities above 2,000 inhabitants IMU main rate 6.25 −3.38 9.32*** −9.72*** [7.59] [2.44] [2.02] [2.35] IMU other rate 0.37 −2.06 21.47*** −19.81*** [5.04] [1.43] [1.53] [1.93] Observations 4,258 4,258 4,258 4,258 R2 0.04 0.06 0.22 0.12 Notes: Each column reports results of a regression where the left hand side variable is the per-capita change of the municipal government instrument in the title between 2012 and 2011 projected on the IMU rates on the main dwelling and other residential properties set by that very municipality in 2012. Municipal public expenditure refers to the variable “Total expenditure” (“Totale generale delle spese”), other municipal tax revenues refers to the sum of all municipal tax revenues net of IMU payments (specifically we consider the following taxes: “Scopo”, “Soggiorno”, “Pubblicita”, “Occupazione degli spazi pubbliche”, “Raccolta e smaltimento dei rifiuti”, “Tassa affissioni”, “Anagrafe”, “Uffici giudiziari”, “Polizia municipale”, “Istruzione elementare”, “Istruzione media”, “Assistenza scolastica”, “Biblioteche”, “Teatri, attività culturali”, “Piscine comunali”, “Stadio comunale, palazzo dello sport”, “Manifestazioni diverse”, “Servizi turistici”, “Viabilità”, “Trasporti pubblici locali”, “Urbanistica”, “Edilizia residenziale pubblica locale”, “Servizio idrico”, “tariffa igiene ambientale”, “Asili nido”, “Proventi servizi di prevenzione e riabilitazione”, “Ricovero per anziani”, “Assistenza, beneficenza pubblica”, “Servizio necroscopico e cimiteriale”, “Entrate da sanzioni amministrative”, “Mezzi pubblicitari”, “Proventi di bene”, “Scuola materna”, “Addizionale IRPEF”, “Segreteria”, “Ufficio tecnico”, “Servizi turistici”, “C.O.S.A.P.”, “Concessioni cimiteriali”). Fiscal deficit calculated as difference between municipal expenditures (current + capital) and municipal revenues (tax revenues + current grants + dismissals). Current revenues and expenditures are on accrual basis, capital revenues and expenditures are on a cash basis. For this reason, the sum of the coefficients in the first three columns cannot equal the coefficient in the last column. Robust standard errors clustered by provinces in brackets. ***Significant at 1%. Source: Ministry of Interior data (publicly available at: http://finanzalocale.interno.it/apps/floc.php/in/cod/4). View Large Our findings are consistent with the evidence reported by Grembi, Nannicini, and Troiano (2012), who show that following a less stringent constraint on municipal fiscal deficits in Italy, local governments responded mainly by cutting real estate taxes and marginal income tax rates. Independent evidence on the lack of correlation between property taxes and other local taxes is provided in Figure C.1 of Online Appendix C, which scatter plots the IMU rates against the rates on the municipal component of income taxes (“IRPEF”). In summary, the geographical variation in IMU rates across municipalities does not seem to be associated with the cross-sectional variation in other municipal government economic policies. 3. Data and Empirical Framework In this section, we present the household survey data and outline the empirical specification that we use to link the income change induced by the IMU taxes paid to the expenditure change. As discussed in the previous section, we use a rich set of demographics and property-specific covariates to isolate exogenous variation across households at a similar stage of their life-cycle, owing properties with similar value and characteristics but living in (unobserved) municipalities with different tax rates. Finally, we discuss and evaluate the role of possible confounding factors, including other macroeconomic interventions, as well as run a placebo test over two waves of the SHIW that have witnessed no changes in municipal property taxes 3.1. The Household Survey Data Our dataset is based on the “Survey on Households Income and Wealth” (SHIW) conducted by the Bank of Italy. The survey is run every two years and covers around 8,000 households distributed over about 3,000 Italian municipalities. The data are available in anonymous form. Each survey is conducted at the end of the respective year during the last few weeks of December. On average, about half of the households that appear in one survey overlap in the following wave. Given that sampling design involves unequal stratum sampling fractions, the use of household sampling weights is necessary to obtain unbiased estimates of the corresponding aggregates. In our econometric analysis, we rely on two consecutive surveys (2010 and 2012), although in some of the analyses we consider also 2008. The 2010 survey covers 19,836 individuals grouped in 7,951 households whereas the 2012 survey covers 20,022 individuals grouped in 8,151 households. We use household level data and keep households who were surveyed both in 2010 and 2012 (about 56% of the 2012 survey). Then, we drop observations with missing values in some relevant variables (typically the market value or the surface of the main dwelling). Finally, to reduce the impact of compiling errors and outliers, we drop observations in the 0.5% tails of the distribution of total expenditure changes, leaving us with a sample of 4,002 observations. We report the descriptive statistics of our working dataset in Table 3, highlighting mean, median, 25th and 75th percentiles of the distribution of the variables of interest for the full regression sample (first three columns), home-owners only (middle panel) and mortgagors (last three columns). To correct for the under-reporting of financial assets (see D’Aurizio et al. 2006), we rescale this variable by the ratio between the value of financial assets for the whole economy calculated by the Bank of Italy on data from the national statistical agency (ISTAT) and its SHIW counterpart obtained by summing up the value of financial assets for all households in the survey using sampling weights. Home-owners are around 71.5% (27.9% by inheritance) of our regression sample whereas the share of mortgagors is 13.7% of the home-owners.10 As shown in Figure D.1 of Online Appendix D, these shares display a remarkable stability over time and whereas the fraction of households owning only one residential property has somewhat decreased between 2010 and 2012, we note that the start of this decline dates back to 2006. Table 3. Summary statistics, regression sample. Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Notes: “Age” and “Studio” refer to the age and the education level (1 = elementary or lower, 6 = postgraduate degree) of the head of the household. “$${\triangle }$$Y” refers to the change of household disposable income. “$${\triangle }$$C” refers to the change of household consumption. “$${\triangle }$$CD” refers to the change of household consumption on durables. “$${\triangle }$$CV” refers to the change of household consumption on vehicles. The entries for vehicles purchases are consistent with the data from ACI (http://www.aci.it), which show that in 2012, about 2.5 millions of vehicles were exchanged across all Italian households, which are around 24.5 millions. The number of mortgagors that bought a new care as a share of total number of mortgagors in 2012 was around 5% whereas in 2010 it was around 9.5%. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). “Real estate” refers to the variable “ar1” (“Real assets (housing, land, and other buildings)”) in database “ricf2012.dta”. “Mortgage debt” refers to variable “deb12a” in dataset “fami2012.dta”. View Large Table 3. Summary statistics, regression sample. Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Full sample Home-owners Mortgagors Variable Unit Mean Median 25% 75% Mean Median 25% 75% Mean Median 25% 75% Education Index 3.1 3.0 2.0 4.0 3.2 3.0 2.0 4.0 3.6 4.0 3.0 4.0 # components Units 2.4 2.0 1.0 3.0 2.4 2.0 2.0 3.0 3.1 3.0 2.0 4.0 Age Years 60.9 62.0 50.0 73.0 62.1 62.0 51.0 73.0 50.6 49.0 43.0 57.0 Children Units 0.5 0.0 0.0 1.0 0.5 0.0 0.0 1.0 1.1 1.0 0.0 2.0 Income Euro (’000) 36.4 30.9 20.2 47.1 40.8 35.7 24.3 51.9 46.5 42.2 29.2 57.2 $${\triangle }$$Y Euro (’000) 1.6 1.7 −2.9 6.5 2.0 2.1 −2.9 7.4 2.5 2.6 −2.9 8.8 $${\triangle }$$C Euro (’000) 0.6 0.6 −3.6 5.1 0.8 0.9 −3.6 5.9 1.7 1.5 −4.7 7.6 $${\triangle }$$CD Euro (’000) −0.4 0.0 −0.4 0.0 −0.3 0.0 −0.5 0.0 −0.2 0.0 −0.8 0.3 $${\triangle }$$CV Euro (’000) −0.2 0.0 0.0 0.0 −0.2 0.0 0.0 0.0 0.0 0.0 0.0 0.0 $${\triangle }$$CN Euro (’000) 1.0 0.9 −3.0 4.9 1.2 1.2 −3.0 5.4 1.9 1.6 −3.2 6.9 Net liquid wealth Euro (’000) 15.9 5.0 0.4 16.0 19.1 7.0 1.3 20.0 1.6 −0.4 −4.9 8.5 Real estate Euro (’000) 219.4 172.4 60.9 304.4 281.2 213.1 142.0 355.1 284.6 233.3 152.1 355.0 Mortgage debt Euro (’000) 6.2 0.0 0.0 0.0 7.9 0.0 0.0 0.0 59.1 50.0 18.0 90.0 IMU main Euro 208.9 120.0 0.0 300.0 267.7 200.0 60.0 390.0 249.2 200.0 60.0 350.0 IMU other Euro 150.2 0.0 0.0 0.0 192.5 0.0 0.0 44.0 185.8 0.0 0.0 100.0 $${\triangle }$$House Price Euro (’000) −3.4 0.0 −35.0 20.0 −4.0 0.0 −50.0 50.0 −0.1 0.0 −50.0 40.0 # properties Units 1.2 1.0 1.0 2.0 1.6 1.0 1.0 2.0 1.6 1.0 1.0 2.0 Number of observations 4,002 3,122 420 Notes: “Age” and “Studio” refer to the age and the education level (1 = elementary or lower, 6 = postgraduate degree) of the head of the household. “$${\triangle }$$Y” refers to the change of household disposable income. “$${\triangle }$$C” refers to the change of household consumption. “$${\triangle }$$CD” refers to the change of household consumption on durables. “$${\triangle }$$CV” refers to the change of household consumption on vehicles. The entries for vehicles purchases are consistent with the data from ACI (http://www.aci.it), which show that in 2012, about 2.5 millions of vehicles were exchanged across all Italian households, which are around 24.5 millions. The number of mortgagors that bought a new care as a share of total number of mortgagors in 2012 was around 5% whereas in 2010 it was around 9.5%. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). “Real estate” refers to the variable “ar1” (“Real assets (housing, land, and other buildings)”) in database “ricf2012.dta”. “Mortgage debt” refers to variable “deb12a” in dataset “fami2012.dta”. View Large The net wealth of Italian households is among the highest in the world but it has a defining peculiarity: around 65% is represented by real assets. The median net wealth is around €270,000 (€348,000 among all home-owners and €289,000 among mortgagors) corresponding to a lower debt-to-income ratio than in other advanced economies.11 Relative to the full regression sample, which also include renters, home-owners tend to have a higher level of both net liquid and illiquid wealth. Relative to all home-owners, mortgagors tend to have a younger head, higher income, more volatile expenditure, lower net liquid wealth, and real estates with a higher value. 3.2. Empirical Specifications The goal of our analysis is to relate variation in disposable income stemming from cross-household variation in the IMU taxes to variation in household expenditure. As there was no tax on the main residential property in 2010 (and only a small tax amount was typically paid on other residential properties because of the obsolete land registry rental value then), we begin by looking at the effect of the tax paid on the main dwelling in 2012 on the household expenditure change between 2010 and 2012. Then, we turn our attention to the richer specification that also includes as a regressor the IMU paid on other residential properties in 2012.12 To ensure our empirical strategy isolates variation in the amount of taxes paid that is unrelated to household and property characteristics, a rich set of controls is featured in the following specification: \begin{equation} \triangle C_{i}=\alpha +{{\gamma }}\cdot \textit{IMUmain}{}_{i}+{\delta }\cdot \triangle HP_{i}+\boldsymbol {\theta }{\boldsymbol X}_{i}+\varepsilon _{i}, \end{equation} (1) where $${\triangle }$$Ci indicates the change in expenditure (on either nondurable goods and services or durable goods) of household i between 2010 and 2012 ($${\triangle }$$Ci = Ci, 2012 − Ci, 2010), $$\textit{IMUmain}_{i}$$ is the amount of IMU tax paid on the main dwelling in 2012, $${\triangle }$$HPi is the self-reported change in house price ($${\triangle }$$HPi = HPi, 2012 − HPi, 2010), $${\boldsymbol X}_{i}$$ contains a set of controls, and ϵi is an idiosyncratic shock.13 As covariates in matrix $${\boldsymbol X}_{i}$$, we add four sets of variables: (i) households demographics, including age and educational attainment of the household head, family size, number of children and their square values, two dummies that takes value of one for home-owners and mortgagors respectively, (ii) regional dummies, (iii) property characteristics including type of building, surface, number of owned properties and dummies for the type of neighborhood (city center, suburbs, etc.) and (iv) a set of dummy variables capturing expectations about future income and about future local house prices (see Online Appendix E for a detailed description). As we control for both demographics and property characteristics, which influence either directly (through the deductions) or indirectly (through the land registry rental value) the household-specific amount of municipal property tax paid, the coefficient γ on IMU is likely to capture the variation in household consumption due to the municipal variation in the IMU tax rates. As the latter appears unrelated both to other local economic policies and to past local economic conditions (as discussed in the previous section), equation (1) can be estimated using OLS and the coefficient γ can be interpreted as the causal effect of the IMU property tax on consumer spending. The coefficient δ captures the household-level association between changes in expenditure and changes in the subjective house value. Finally, our empirical strategy relies on the absence of dissimilar pretreatment trends in expenditure (between IMU payers and non-IMU payers) that may account for the post-treatment differences across the two groups. In Section 5.1, we present evidence consistent with this hypothesis. In the richer specification, we also consider the $$\textit{IMUother}_{i}$$ paid on other residential properties: \begin{equation} \triangle C_{i}=\alpha +{{\gamma _{1}}}\cdot \textit{IMUmain}{}_{i}+{{\gamma _{2}}}\cdot \textit{IMUother}{}_{i}+{\delta }\cdot \triangle HP_{i}+\boldsymbol {\theta } {\boldsymbol X}_{i}+\varepsilon _{i}, \end{equation} (2) where the coefficients of interest are now γ1 and γ2, representing the impact of the IMU tax on the main dwelling and the IMU tax on other residential properties. In our baseline specifications, equations (1) and (2) are estimated either over the full sample or for home-owners only, exploiting in the latter case exclusively variation in the amount of property taxes paid. To shed lights on the characteristics driving heterogeneity in the spending response, (1) and (2) will then be run splitting the sample into households without and with (mortgage) debt, showing that the latter display significantly higher MPCs. The expenditure changes from the housing tenure grouping are then compared in Section 6 to the findings from more traditional groupings based on age and income as well as to subsamples of households reporting different levels of uncertainty about their future income. In Section 5, we focus on the different categories of durable expenditure and find that most of the changes in debtors’ spending are concentrated in vehicle purchases. 3.3. Other Confounding Factors As shown in Section 2, the IMU tax hikes does not appear related neither to other local economic policies nor to past local economic conditions. Still, the availability of households survey data only in 2010 and 2012 poses the challenge that other macroeconomic developments may confound the inference one can draw about the effect of the property taxes on consumer spending. In this section, we take this challenge at face value and ask whether the IMU tax changes were correlated with any other significant macroeconomic change that may have occurred over these two years. Accordingly, we use specifications that are all like (1) and (2) but in which the dependent variable becomes: the amount of taxes paid on other non-IMU austerity macro interventions, changes in households transfers from the central government (including pensions), changes in house value, changes in the taxes paid on “super-cars” and changes in the expenditure for those nondurable goods and services whose VAT increased between 2010 and 2012. The dependent variable in the first column of Table 4 is calculated as the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the regional marginal tax rate on personal income. This is meant to capture the host of other austerity interventions that were passed together with the IMU tax changes. The columns on transfers and house value assess whether the change in municipal property taxation was associated, amplified or perhaps offset by other changes in the government budget, the household balance sheet or the tax base. This seems particularly important in the light of the Fornero reform of the Italian pension system, which was also part of the fiscal consolidation plan passed into law by Prime Minister Mario Monti’s government in December 2011. Given the very significant fall in vehicles expenditure associated with the property tax changes (reported in Section 5), in the fourth column we evaluate the relation between the amount of property taxes paid by each households and the taxes paid on supercars (defined as cars above 185 kW), whose tax rate was also changed in 2011.14 An additional confounding factor occurred in September 2011 when the government led by Prime Minister Silvio Berlusconi passed an increase in the Value Added Tax (VAT) rate from 20% to 21%. Accordingly, the last column of Table 4 reports the consumption response of those nondurable goods and services that were subject to the VAT rate change. Table 4. Confounding factors. Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Notes: Robust standard errors clustered by regions in brackets. IMU “main” and “other” refer to the IMU tax paid for the main dwelling and other properties, respectively. Because the variables have different magnitudes and variances, all left-hand side variables, IMU main and IMU other have been standardized. When running the same regressions on nonstandardized variables we obtain very similar results. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Austerity non-IMU” refers to the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the local (regional) marginal tax rate on personal income. “Transfers” refers to total transfers to households, including pensions. “Supercar” is a variable calculated as the product between the value of the car if above €40,000 and the average yearly tax rate of 1.26% on supercar (estimated using Automobile Club of Italy data). Finally, “VAT” refers to the consumption change on nondurable goods and services whose VAT rate changed in September 2011. View Large Table 4. Confounding factors. Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Austerity non-IMU Transfers $${\triangle }$$HP Supercar VAT Panel A IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Panel B IMU main 0.02 0.01 0.03 0.01 −0.01 [0.01] [0.02] [0.03] [0.02] [0.02] IMU other 0.01 0.01 0.04 −0.03 −0.00 [0.04] [0.01] [0.03] [0.03] [0.03] Observations 4,002 4,002 4,002 4,002 4,002 R2 0.14 0.03 0.22 0.03 0.09 Notes: Robust standard errors clustered by regions in brackets. IMU “main” and “other” refer to the IMU tax paid for the main dwelling and other properties, respectively. Because the variables have different magnitudes and variances, all left-hand side variables, IMU main and IMU other have been standardized. When running the same regressions on nonstandardized variables we obtain very similar results. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Austerity non-IMU” refers to the sum of the increase in taxation on electricity bills, the increase in taxation on cooking gas, the increase in taxation on motor fuel, and the increase of the local (regional) marginal tax rate on personal income. “Transfers” refers to total transfers to households, including pensions. “Supercar” is a variable calculated as the product between the value of the car if above €40,000 and the average yearly tax rate of 1.26% on supercar (estimated using Automobile Club of Italy data). Finally, “VAT” refers to the consumption change on nondurable goods and services whose VAT rate changed in September 2011. View Large Reassuringly, in each of the two panels and samples, there is little evidence that the amount of IMU taxes paid by each household was systematically related to any of the macro policy and economic changes summarized in Table 4. Furthermore, the Fornero reform on pensions affected younger generations evenly across housing tenure groups and we show in Section 6 that young households who paid the IMU taxes contracted their durable expenditure by a significantly larger amount than young households who did not pay the IMU taxes. Finally, the VAT rate changed both on vehicles and on all other (nonvehicle) durable goods: but, as we will show in Section 5.1, only the expenditure on vehicles witnessed a significant contraction, suggesting that the 2011 VAT rate change seems unlikely to have contributed to our findings. In Figure 3, we explore further the impact of the VAT change by reporting the evolution of three price indexes from national accounts: (i) items that experienced an increase in the VAT rate (dashed-dotted black line), (ii) items that did not experience an increase in the VAT rate (light gray solid line) and (iii) cars (red solid line), which were also subject to the VAT rate change. The vertical lines correspond to the dates of the introduction of the VAT rate increase and of the IMU reform respectively. Two main developments are apparent from Figure 3. First, following the VAT rate change of September 2011, both the increase in the price index on all items subject to the VAT rate hike and the increase in the price index on cars are far sharper than the mild increase in the price index on flat-VAT rate items. Second, the behavior of the price index on increased-VAT rate items decouples from the behavior of the car price index around December 2011 when the IMU tax changes were passed into law by Mr Monti’s government. Given we will show that vehicles purchases was the single most responsive and most declining spending category to the IMU taxes, we interpret the flat profile of the car price index after the introduction of the IMU (relative to the steadily rising profile of the price index on all increased-VAT rate items) as most likely stemming from the effect of the property taxes on consumer spending. Figure 3. View largeDownload slide Evolution of prices. The figure shows the evolution of prices for cars, items subject to the 2011 VAT increase, and items exempted from VAT increase (these items are subject to a 0% VAT, 4% VAT, or 10% VAT according to the category; these VAT rates were unchanged in the considered period). The aggregate indexes (for “Flat-VAT items” and “Increasing-VAT items”) are weighted averages of the respective subindexes. The relative weights are provided by ISTAT. Inflation for “cars” refer to the “motor cars” category (ISTAT code 711). Items excluded from VAT changes include: “education”, “food” (excluding “ready-made meals”), “restaurants and hotels”, “miscellaneous goods and services” (excluding “mineral or spring waters”), “actual rentals for housing”, “water supply and miscellaneous services relating to the dwelling”, “electricity, gas and other fuels”, “medical products, appliances and equipment”, “out-patient service”, “hospital services”, “transport services”, “postal services”, “recreational and cultural services”, “newspapers, books and stationery”. The share of items (including cars) subject to the VAT increase in 2011 was 40.6%. Source: Authors’ calculations on ISTAT data (available at: http://dati.istat.it/?lang=en). Figure 3. View largeDownload slide Evolution of prices. The figure shows the evolution of prices for cars, items subject to the 2011 VAT increase, and items exempted from VAT increase (these items are subject to a 0% VAT, 4% VAT, or 10% VAT according to the category; these VAT rates were unchanged in the considered period). The aggregate indexes (for “Flat-VAT items” and “Increasing-VAT items”) are weighted averages of the respective subindexes. The relative weights are provided by ISTAT. Inflation for “cars” refer to the “motor cars” category (ISTAT code 711). Items excluded from VAT changes include: “education”, “food” (excluding “ready-made meals”), “restaurants and hotels”, “miscellaneous goods and services” (excluding “mineral or spring waters”), “actual rentals for housing”, “water supply and miscellaneous services relating to the dwelling”, “electricity, gas and other fuels”, “medical products, appliances and equipment”, “out-patient service”, “hospital services”, “transport services”, “postal services”, “recreational and cultural services”, “newspapers, books and stationery”. The share of items (including cars) subject to the VAT increase in 2011 was 40.6%. Source: Authors’ calculations on ISTAT data (available at: http://dati.istat.it/?lang=en). Finally, the inference on the effects of an increase in property taxes may be distorted by a decline in central government expenditure, which as illustrated in Table G.1 of Online Appendix G mainly came in the form of a fall in government consumption or wages for public employees (see Born, Müller, and Pfeifer 2014). With respect to this hypothesis, we focus on two subgroups of households: those headed by a public employee and those not. We find no statistical differences in the coefficients on $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ across the two groups, with estimated responses being slightly stronger for nonpublic employees. In summary, the results in this section suggest that the effects of the IMU tax paid on household expenditure seem unlikely to be confounded by other nation-wide policies or macroeconomic factors that changed over the same period. 3.4. Placebo Test As a further empirical validation of the extent to which our framework is well-suited to capture the effect of the IMU taxes on consumer spending, we run placebo regressions that correlate the change in expenditure of each household between 2008 and 2010 with the IMU tax paid by that very household in 2012. If the IMU fiscal shock of December 2011 was unanticipated and was indeed the trigger of the significant expenditure decline in 2012 (which we document in the next section), then we would expect it to have no significant effect on expenditure before 2012, given that no actual changes in property taxes occurred between the end of 2008 and the end of 2010 when these two other waves of the SHIW were conducted. In this section (and in this section only), all other right hand side variables (including $${\triangle }$$HPi) refer to the period 2008–2010. In contrast, $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ refer to the amount of taxes paid by household i in 2012. The left hand side variable is the expenditure change of that same household i between 2008 and 2010. For the placebo analysis, we only rely on households who appear in all three waves. Accordingly, the regression sample is reduced from 4,002 to 2,480 observations. The results of the placebo test are shown in Table 5. Both $$\textit{IMUmain}_{i}$$ and $$\textit{IMUother}_{i}$$ never affect significantly either nondurable consumption or durable expenditure and the estimated coefficients have often the wrong sign.15 On the other hand, the effect of house prices is highly significant for nondurable consumption (but not for durable expenditure), with magnitudes that are not statistically different from the point estimates we will present in the next section for 2010–2012. Table 5. Placebo test. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Notes: Robust standard errors clustered by regions in brackets. “Non durables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. ***Significant at 1%. View Large Table 5. Placebo test. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main 0.13 0.21 0.19 0.15 [0.78] [0.54] [0.83] [0.54] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Panel B IMU main 0.12 0.23 0.18 0.17 [0.79] [0.54] [0.84] [0.55] IMU other 0.08 −0.13 0.06 −0.13 [0.19] [0.09] [0.21] [0.10] $${\triangle }$$HP (€ ’00) 0.90*** 0.06 0.89*** 0.07 [0.18] [0.07] [0.19] [0.07] Observations 2,480 2,480 2,419 2,419 R2 0.15 0.02 0.15 0.02 Notes: Robust standard errors clustered by regions in brackets. “Non durables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta” where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. ***Significant at 1%. View Large 4. The Response of Household Expenditure In this section, we present the main results of our analysis. We start with the baseline estimates in Table 6, which associate the IMU taxes paid on the main dwelling and other residential properties with nondurable and durable expenditure. Then, we explore the heterogeneity of these responses and show they vary significantly across household balance sheet positions as exemplified by the presence of mortgage debt. Furthermore, we show that the majority of households with mortgage debt hold very low liquid wealth relative to income and therefore are likely to face liquidity shortages in the wake of changes in their resources. Finally, we review the testable predictions of the theoretical literature on the spending categories that are most likely to respond to a temporary change in household resources. Table 6. Baseline results. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large Table 6. Baseline results. Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Full sample Home-owners Nondurables Durables Nondurables Durables Panel A IMU main −0.05 −0.43** −0.09 −0.42** [0.57] [0.18] [0.53] [0.20] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Panel B IMU main −0.05 −0.44** −0.08 −0.43** [0.57] [0.18] [0.53] [0.20] IMU other −0.05 0.06 −0.06 0.09 [0.44] [0.09] [0.44] [0.09] $${\triangle }$$HP (€ ’00) 0.97*** 0.03 0.97*** 0.04 [0.10] [0.05] [0.10] [0.05] Observations 4,002 4,002 3,122 3,122 R2 0.15 0.02 0.16 0.02 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large 4.1. Baseline Results The estimates of equation (1) and equation (2) are presented in Panel A and Panel B of Table 6, respectively.16 The two columns on the left refer to the full sample whereas those on the right focus on home-owners only. The odd columns display the relevant IMU and house price coefficients for a specification using nondurable consumption on the left hand side whereas in the even columns the dependent variable is the expenditure on durable goods. Four main empirical regularities emerge from these baseline estimates. First, the MPC associated with the IMU tax paid on the main dwelling in columns (1) and (3) is always very close to and never statistically different from zero. Second, the MPC on durable goods from $$\textit{IMUmain}_{i}$$ is always very significant and large, with point estimates around 0.43 in columns (2) and (4). Interestingly, Parker et al. (2013) report a marginal propensity to spend around 0.5 in response to the 2008 income tax rebate in the United States whereas Jappelli and Pistaferri (2014) document that for every hypothetical euro of transitory income, the average SHIW respondent would increase expenditure by 48 cents. Third, the results on $$\textit{IMUmain}_{i}$$ are robust to using a richer specification that also includes $$\textit{IMUother}_{i}$$ among the regressors. Fourth, and in sharp contrast to the main dwelling, the IMU tax paid on other residential properties in Panel B triggers neither a significant contraction in nondurable consumption nor a significant contraction in durable expenditure, with point estimates always in the neighborhood of zero.17 Of independent interest, both panels record also the estimates of the house price effect. In particular, the coefficient on $${\triangle }$$HPi is small and statistically indistinguishable from zero in columns (2) and (4) for durable expenditure. But the marginal propensity to consume out of housing wealth in columns (1) and (3) is always very significant and precisely estimated at around 1% (i.e., a €100 appreciation in house prices tends to be associated with a 97 cents increase in nondurable consumption). Although these estimates are in line with the effects reported by Guiso, Paiella, and Visco (2005) and Paiella and Pistaferri (2017) on earlier SHIW samples, they are sizably smaller than the 5%–7% reported by Mian, Rao, and Sufi (2013) for the United States or the 7% to 9% reported by Campbell and Cocco (2007) for the United Kingdom. It should be noted, however, that the scarcity of refinancing opportunities—and in particular of mortgage equity withdrawal—makes housing wealth in Italy significantly more illiquid (see IMF 2008; Grant and Peltonen 2008; Calza, Monacelli, and Stracca 2013). Accordingly, the statistical association between house prices and consumption in Italy seems more likely to reflect a direct wealth effect or a common factor driving both variables rather than a collateral channel. We will come back to the sensitivity of nondurable consumption to house prices in the extended analysis of Section 6.4 where we will, among other things, (i) add household income as a further control in an augmented version of specification (2) and (ii) exclude housing wealth and the number of children from the set of covariates. 4.2. Grouping Households by Mortgage Debt Position A growing strand of empirical studies, including Dynan (2012), Kaplan, Violante, and Weidner (2014), Cloyne and Surico (2017), and Acconcia, Corsetti, and Simonelli (2015) advocate a role for household balance sheet positions, and mortgage debt in particular, in the transmission of structural and policy shocks to consumption. The variation of IMU tax rates across households allows us an unprecedented evaluation of this hypothesis in the context of a tax hike on housing wealth. To this end, in Table 7 we group households according to whether they have debt (first two columns) or not (last two columns). In an effort to maximize the number of observations, in Panel A, we include secured and unsecured debt. In Panel B, we focus on mortgage debt only.18 Table 7. Debtors (mortgagors) versus nondebtors (nonmortgagors). Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with (secured or unsecured) debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large Table 7. Debtors (mortgagors) versus nondebtors (nonmortgagors). Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel A: Total debt Nondebtors Debtors Nondurables Durables Nondurables Durables IMU main −0.09 0.13 0.16 −2.71*** [0.56] [0.22] [1.42] [0.56] IMU other −0.12 0.08 0.15 0.13 [0.54] [0.12] [0.90] [0.45] $${\triangle }$$HP (€ ’00) 1.02*** 0.01 0.72*** 0.03 [0.13] [0.04] [0.23] [0.11] Observations 3,121 3,121 881 881 R2 0.15 0.03 0.22 0.07 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Panel B: Mortgage debt Nonmortgagors Mortgagors Nondurables Durables Nondurables Durables IMU main −0.01 −0.17 −0.34 −2.60** [0.66] [0.21] [2.55] [1.02] IMU other −0.02 0.07 0.24 −0.09 [0.47] [0.10] [1.47] [0.75] $${\triangle }$$HP (€ ’00) 0.98*** 0.02 0.67 0.06 [0.12] [0.05] [0.42] [0.19] Observations 3,582 3,582 420 420 R2 0.14 0.02 0.27 0.11 Notes: Robust standard errors clustered by regions in brackets. “Nondurables” refers to the change in household expenditure on nondurable goods (variable “cn” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Durables” refers to the change in household expenditure on durable goods (variable “cd” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $${\triangle }$$HP refers to the change of (self-reported) market value of all properties owned. “Debtors” refer to households with (secured or unsecured) debt at the end of 2012 (meaning with positive entry of the variable “pf” in database “ricf12.dta”). “Mortgagors” refer to households with mortgage debt at the end of 2012 (meaning with positive entry of the variable “deb12a” in database “fami12.dta”). Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. **Significant at 5%; ***significant at 1%. View Large The main take away from Table 7 is that the significant average effects on durable goods recorded in the previous tables are entirely driven by home-owners with debt, whose marginal propensities to spend (out of the taxes paid on the main dwelling) in column (4) tend to be larger and more significant than in Table 6, despite the far fewer number of observations. The results from columns (1) and (2) of Panel B reveal further that removing as few as some 400 mortgagors from the full sample yields very small and largely insignificant responses to both property taxes. In the next section, we will show that net vehicle purchases account for the lion share of the behavior of durable expenditure and that the magnitude of the MPC in Table 7 for households with debt is consistent with a down payment rate for buying a car around 10%. Interestingly, the house price effect in column (3) tends to be smaller for debtors in Panel A and not statistically different from zero (though imprecisely estimated) for mortgagors in Panel B, consistent with a shortage of refinancing opportunities in the Italian credit market. Finally, we note that some durable goods (such as cars, motorbikes, furniture, or electrical appliance) tend to be purchased using consumer credit. Accordingly, a worsened access to financial markets during 2011 and 2012 could—at least in principle—be partially responsible for the large adjustment on durable goods recorded between the two waves of the SHIW (at the end of 2010 and at the end of 2012) in Tables 6 and 7. As shown in Figure H.1 of Online Appendix H, however, there seems to be little evidence that over this biennium households were charged systematically higher interest rates for consumer credit (on purchases either below or above €5,000) relative to 2010, when these series begin. 4.3. Why do Mortgagors Have a Higher MPC? A large theoretical literature exemplified by Deaton (1992) and Zeldes (1992) has convincingly made the case that liquidity constrains could be a primary source of violation of the permanent income hypothesis. Although earlier empirical contributions have typically associated the presence of liquidity constraints with lower income, lower educational attainment and younger household head (see for instance Johnson et al. 2006), the theoretical mechanism in Kaplan and Violante (2014) suggests that also wealthy households may face liquidity shortages whenever a large durable purchase such as housing makes illiquid most of their wealth. In this section, we therefore evaluate the hypothesis that mortgagors could be “wealthy” hand-to-mouth by looking at their household balance sheet. In Figure 4, we compare, by number of dwellings, the distribution of the net saving rate—that is, disposable income minus total consumption as a share of disposable income—with the distribution of the debt service ratio,—that is, mortgage repayments as a share of disposable income. After expenditure and debt repayments, mortgagors with only one property are left with little disposable income as measured by the small distance between the median values of the net saving rate (17%) and the debt service ratio (16%). In contrast, mortgagors with more than one property appear less constrained, as they enjoy significantly larger saving rates relative to the debt service ratio distribution. Figure 4. View largeDownload slide Distribution of net saving rates and debt service ratio per number of property. The figure shows the distribution of net saving rates (using after tax income) and debt service ratio to net disposable income by number of properties. The bars span in between the 25th and 75th percentile of the distribution whereas the horizontal lines in each bar indicate the median of the distribution. Monthly saving rates defined as the ratio between variable “s2” and variable “y2” in database “cons12.dta”. Mortgage service payments is based on variables “tdebita11” plus “tdebita12” plus “tdebita13”, database “alld2_res.dta” in the 2012 survey. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 4. View largeDownload slide Distribution of net saving rates and debt service ratio per number of property. The figure shows the distribution of net saving rates (using after tax income) and debt service ratio to net disposable income by number of properties. The bars span in between the 25th and 75th percentile of the distribution whereas the horizontal lines in each bar indicate the median of the distribution. Monthly saving rates defined as the ratio between variable “s2” and variable “y2” in database “cons12.dta”. Mortgage service payments is based on variables “tdebita11” plus “tdebita12” plus “tdebita13”, database “alld2_res.dta” in the 2012 survey. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Complementary evidence on the liquidity shortage faced by mortgagors is recorded in Figure 5, which reports for each housing tenure group the share of households holding an amount of net liquid wealth below half of their monthly income.19 Two results emerge from this evidence: (i) about 65% of mortgagors appear hand-to-mouth and (ii) the shares of the other two groups are only a fraction of the mortgagor share, with the proportion of liquidity constrained households being around 25% among both outright home-owners and renters. Bearing in mind that renters represent about 30% of the Italian population and that renters tend to hold very little (if any) housing or financial wealth, our evidence reveals than about 8% of Italian households are “wealth-poor” hand-to-mouth. On the other hand, mortgagors and outright home-owners account for around 13% and 57% of the population, thereby making the share of “wealth-rich” hand-to-mouth households close to 22%. Figure 5. View largeDownload slide Share of hand-to-mouth households by housing tenure group. The figure plots the share of hand-to-mouth across housing tenure groups in 2012. Outright homeowners include households owning one property only. “Hand-to-mouth” refers to households with a net liquid wealth to income ratio lower than half month of income. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). Disposable income refers to variable “y2” in dataset “consXX.dta” (where the suffix XX indicates the year of the survey). Figure 5. View largeDownload slide Share of hand-to-mouth households by housing tenure group. The figure plots the share of hand-to-mouth across housing tenure groups in 2012. Outright homeowners include households owning one property only. “Hand-to-mouth” refers to households with a net liquid wealth to income ratio lower than half month of income. “Net liquid wealth” calculated as the difference between liquid financial assets and unsecured financial liabilities plus mortgage service (variables “af1”, “af2”, “pf2”, “pf3”, and “tmutuoab” in SHIW survey). Disposable income refers to variable “y2” in dataset “consXX.dta” (where the suffix XX indicates the year of the survey). The findings of Figure 5 also offers a rationale for why grouping households only by their level of liquid wealth would produce less sharp and less significant evidence of heterogeneity (than looking at mortgage debt position) in the expenditure response to the temporary income changes triggered by the IMU property taxes. This can be seen both in Table I.2 and Figure I.1 of Online Appendix I, which report the MPCs for lower and higher liquid wealth households when alternative multiples of their monthly income are used as threshold for the group categorization. The reason is that although also a significant portion of renters as well as some outright home-owners tend to hold low liquid wealth (as shown in Figure 5), renters do not pay the property tax because they do not own a house whereas outright home-owners do no longer have a significant fraction of their expenditure precommitted into repaying the mortgage (which they have already repaid in full) and thus their MPC is not statistically different from zero (as shown in Table 7). Accordingly, in both Figure I.1 and Table I.2 of Online Appendix I, the standard errors are larger and the differences in point estimates across groups smaller than in Table 7, as grouping households into a single low liquidity group pools together consumers with very different MPCs.20 In summary, grouping households by their debt position, and in particular whether home-owners with only one property have a mortgage or not, seems to provide sharp(er) evidence of significant heterogeneity in the expenditure responses (than, for instance, simply looking at liquid wealth to income ratios).21 Inspection of the balance sheets of the different groups reveals further that, in each month, owner–occupier mortgagors can only spare little liquidity after meeting expenditure bills and mortgage repayments, consistent with Tobin’ conjecture (cited in the introduction) that debt makes households liquidity constrained and therefore leads to a high marginal propensity to consume. This is consistent with the notion that a significant portion of households with debt are hand-to-mouth despite owing sizable illiquid wealth, thereby providing a novel interpretation for the role of (il)liquid wealth in the transmission of macroeconomic shocks. 4.4. Why is the Response of Durable Expenditure Stronger? In Section 4.1, we have shown that the response of durable expenditure is stronger than the response of nondurable goods and services spending. In Section 4.2, we have reported that households with debt exhibit a larger marginal propensity to spend on durable goods whereas in Section 4.3 we have looked at the net liquid wealth to income ratio to document that most debtors are (wealthy) hand-to-mouth. In this section, we review a theoretical mechanism that can offer a rationale for why the changes in durable expenditure are larger than the changes in nondurable consumption. In an earlier contribution, Browning and Crossley (2000) prove that luxury goods have high elasticities of intertemporal substitution and therefore are easier to postpone in the face of temporary falls in disposable income. The intuition for this result comes from noting that, if goods are additively separable, the Frisch own price elasticities are proportional to the Marshallian income elasticities, with the latter implying that whenever total spending is cut then luxury goods expenditure will be cut by proportionally more. To the extent that most categories of durables are likely to involve more discretionary purchases and less necessity goods than many categories of nondurables, a corollary of their result is that the expenditure on durable goods should be easier to postpone than the consumption on nondurable goods and services. This is consistent with the estimates reported in Table 6. In a more recent work, Aaronson et al. (2012) study the problem of a household that maximizes the utility flows from the consumption of some nondurable goods, Ct, and from the services generated by a stock of durables, Dt, which are bundled together as \begin{equation*} E_{0}\sum _{t=0}^{T}\beta ^{t}\left(C_{t}^{1-\theta }D_{t}^{\theta }\right)^{1-\gamma }/(1-\gamma ). \end{equation*} The stock of durable goods, which depreciated at rate δ, can expand through further investment It, according to the following law of motion: \begin{equation} D_{t+1}=(1-\delta )D_{t}+I_{t} \end{equation} (3) whereas financial asset At accumulates in the form \begin{equation} A_{t+1}=(1+r)A_{t}+Y_{t}-C_{t}-I_{t} \end{equation} (4) with At+1 ≥ 0 and the interest rate denoted by r. The disposable income, represented by Yt, is made up of a deterministic life-cycle profile and a stochastic AR(1) process. A main feature of their analysis is to allow households to borrow against durable goods according to the constraint \begin{equation} -A_{t}\le (1-\pi )D_{t}, \end{equation} (5) where π is the down payment rate or the fraction of newly purchased durable goods that does not serve as collateral, in the spirit of Kiyotaki and Moore (1997). Aaronson et al. (2012) formally show that the marginal propensity to spend out of a temporary income change is far higher for durables than for nondurables. Furthermore, they show that, for goods purchased with higher down payments, consumer spending is less sensitive to income changes because higher down payments imply that fewer durable goods can be purchased with a given level of income. To develop an intuition for their results, assume that the borrowing constraint (5) always binds. Combining this binding constraint with the accumulation formula for financial assets (4) and the law of motion for durables (3) yields the following expression: \begin{equation} \pi I_{t}+C_{t}+(1-\pi )(r+\delta )D_{t}=Y_{t}. \end{equation} (6) Equation (6) makes it clear that although one dollar worth of nondurables requires one dollar of disposable income to be purchased, one dollar worth of durables only require a fraction π of a dollar. This finding not only offers a rationale for why the response to transitory income shocks may be concentrated on durable expenditure (consistent with the estimates reported in this section) but also generates two further testable predictions. First, durable goods requiring lower down payments for their purchase on credit—such as vehicles—should display a stronger response than the rest of durables, which most likely either require higher down payments—such as furniture—or for which collateralized financing may not be readily available—such as small appliances.22 Second, the MPC on durable goods with low down payment rates may exceed one (and indeed even two), as demonstrated by the quantitative analysis in Aaronson et al. (2012) (Table 8) using standard calibrations of the model described previously. In Section 5, we will provide evidence in support of these two further predictions by showing that the durable response is concentrated on vehicles, that the associated MPC is indeed well above one and that the credit taken for vehicles purchases declined significantly in response to the IMU tax changes. Table 8. Vehicles versus nonvehicles durable expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $$\triangle $$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large Table 8. Vehicles versus nonvehicles durable expenditure. Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel A Full sample Home-owners Nonvehicles Vehicles Nonvehicles Vehicles IMU main 0.16 −0.61*** 0.18 −0.61*** [0.12] [0.17] [0.12] [0.19] IMU other −0.01 0.08 −0.01 0.09 [0.07] [0.09] [0.08] [0.08] $${\triangle }$$HP (€ ’00) 0.05* −0.02 0.05* −0.01 [0.03] [0.03] [0.03] [0.03] Observations 4,002 4,002 3,122 3,122 R2 0.02 0.02 0.03 0.02 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Panel B Debtors Mortgagors Nonvehicles Vehicles Nonvehicles Vehicles IMU main −0.33 −2.38*** −0.06 −2.54** [0.36] [0.55] [0.55] [1.02] IMU other −0.09 0.22 −0.47 0.38 [0.15] [0.36] [0.46] [0.65] $${\triangle }$$HP (€ ’00) 0.11 −0.07 0.23 −0.17 [0.07] [0.10] [0.14] [0.13] Observations 881 881 420 420 R2 0.06 0.07 0.10 0.10 Notes: Robust standard errors clustered by regions in brackets. “Nonvehicles” refers to the change in household expenditure on durable goods excluding vehicles (variable “cd2” in dataset “consXX.dta”, where the suffix “XX” indicates the year of the survey). “Vehicles” refers to the change in household expenditure on vehicles (variable “cd1” in dataset “consXX.dta”). IMU “main” and “other” refer to the tax on the main dwelling and other properties, respectively. $$\triangle $$HP refers to the change of (self-reported) market value of all properties owned. Control variables (omitted for brevity) include: (i) households demographics, (ii) geographical dummies, (iii) dummies of main dwelling commercial area, and (iv) expectations about household income and local house prices. *Significant at 10%; **significant at 5%; ***significant at 1%. View Large 5. Spending Categories and the Role of Credit In this section, we explore the extent of heterogeneity in the household responses to the IMU property taxes across spending categories. In particular, we show that net vehicle purchases (or lack thereof) are a main driver of the durable expenditure results in the previous section and that the magnitude of the coefficients on $$\textit{IMUmain}_{i}$$ reported in Table 7 for debtors is consistent with a typical down payment for buying a car. Furthermore, we show that IMU payers are significantly less likely to have taken out a loan for vehicles purchase after the tax policy change than non-IMU payers whereas no discernable patter across households is evident for credit to purchase other durable goods. Finally, we compare our estimates based on negative shocks on disposable income (triggered by the hike in property taxes) with the evidence based on positive income shocks in earlier studies. 5.1. Vehicles Versus Nonvehicles Expenditure To shed lights on the findings in the previous section, we rerun specification (2) over several categories of nondurable and durable expenditure. Based on question E02 of the 2012 (and 2010) SHIW survey, we consider as “durable good” precious objects, cars, other means of transport, furniture, furnishings, appliances, and “various equipment”.23 Nondurable expenditure is calculated as the difference between total expenditure net of rents or mortgage payments and expenditure on nondurable goods and services. Given the estimates in the previous tables, it should not come as a surprise that we find little evidence of heterogeneity among nondurable consumption categories. As for durable goods, we find that net vehicle expenditure (defined as the difference between vehicle purchases and vehicle sales) is the only component that displays large and significant responses to the IMU taxes. This is recorded in Table 8, which splits durable expenditure into vehicles (which account for about 70% of durable goods value) and every other durable goods. In the top panel, we report findings over the full-sample and for home-owners only whereas in the bottom panel we display results for debtors and mortgagors. The coefficients on net vehicles purchases in columns (2) and (4) of Panel A are similar (and statistically indistinguishable) from the coefficients on durable expenditure in Table 6. When net vehicles purchases are excluded from durable expenditure in columns (1) and (3), however, both IMU tax coefficients become insignificant, revealing that this durable category drives the total expenditure response. In Panel B of Table 8, we restrict our attention to indebted households, who display the strongest durable expenditure response, and show that their behavior is indeed driven by net vehicle purchases. The coefficients on $$\textit{IMUmain}_{i}$$ in columns (2) and (4) of Panel B appears in line with their durable expenditure counterparts in Table 7 whereas the responses of nonvehicle durables in columns (1) and (3) tend to be small and statistically indistinguishable from zero. Two points are worth emphasizing about the magnitude of the vehicle expenditure response in Table 8. First, Italian households paid a significant amount of property taxes in 2012, with an average around €357 and a significant portion of payers above €1000. This suggests that some households may have chosen to defer or even eliminate a large durable purchase, whose saving could offset the significant increase in property taxes over a multi-year horizon. Second, given the average per-year disbursement for the IMU tax on the main dwelling, a marginal propensity to spend around two for households with debt—while statistically close to one—is consistent with a down payment rate around 10% on a vehicle purchase. Interestingly, Misra and Surico (2014) show that, also in the context of the U.S. income tax rebates of 2001 and 2008, the aggregate consumption response was driven by a handful of vehicle purchases made by the mortgagors group, who displayed a marginal propensity to spend on this category around two. The finding that the response of total expenditure is driven by a few very large marginal propensities to spend on vehicles is also consistent with the evidence in Aaronson et al. (2012), who report estimated MPCs on vehicles around two for a small number of working households facing a minimum wage hike. In Figure 6, we provide a graphical counterpart of the results in Table 8. In particular, we show not only that (i) the average reduction in net vehicle purchases by IMU payers (dashed-dotted dark gray line) was larger than the average reduction by those home-owners who—because of the deductions—did not pay IMU taxes (dashed gray line), but also that (ii) the average reduction for mortgagor IMU payers (red line) was more pronounced that the average reduction across all IMU payers. Figure 6 provides graphical evidence that the pre-treatment trend was identical between mortgagors IMU-payers and home-owners non-IMU payers as well as that the decline in vehicles expenditure for all IMU payers between 2008 and 2010 (i.e., before the treatment) was, if anything, smaller than the decline for home-owners non-IMU payers over the same period. Finally, it is possible to show that the parallel trends extend back in time, although at the expense of reducing the sample size given the sample rotation in the SHIW survey. Figure 6. View largeDownload slide Expenditure on vehicles. The chart shows the expenditure on vehicles over time among home-owners in deviation from their respective means. The pretreatment trend is identical across groups and the same evidence applies to all variable of interest. If the time period is extended back in time (at the expenses of the sample size) the evidence of parallel trends remains identical. The chart is based on those households entering in the three waves of the SHIW survey reported. The expenditure on vehicles refers to variable “cd1” in database “consXX.dta” (where the suffix “XX” refers to the year of the survey). The households with a debt are identifies using the variable “deb12a” (“Amount of debts owed at the end of the year to banks or financial companies for the purchase or restructuring of buildings”) in database “famiXX.dta”. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). Figure 6. View largeDownload slide Expenditure on vehicles. The chart shows the expenditure on vehicles over time among home-owners in deviation from their respective means. The pretreatment trend is identical across groups and the same evidence applies to all variable of interest. If the time period is extended back in time (at the expenses of the sample size) the evidence of parallel trends remains identical. The chart is based on those households entering in the three waves of the SHIW survey reported. The expenditure on vehicles refers to variable “cd1” in database “consXX.dta” (where the suffix “XX” refers to the year of the survey). The households with a debt are identifies using the variable “deb12a” (“Amount of debts owed at the end of the year to banks or financial companies for the purchase or restructuring of buildings”) in database “famiXX.dta”. Source: Authors’ calculations on Bank of Italy SHIW surveys data (available at: https://www.bancaditalia.it/statistiche/tematiche/indagini-famiglie-imprese/bilanci-famiglie/). A complementary way to look at the impact of the IMU property taxes is to ask whether a larger tax disbursement is associated with a lower probability of making a durable purchase. To this end, we construct two binary variables that take the value of one if vehicle expenditure or other durable expenditure respectively are positive and zero otherwise. These become the dependent variables in two separate probit regressions that use otherwise the same regressors as in the specifications in the rest of the paper. Two main advantages of this approach is that a binary variable is less prone to measurement errors (relative to the exact euro amount of any durable purchase) and the probit specification is suited to handle nonlinearity in the data. The results in Table 9 reveals—consistently with the estimates in Table 8—that only for mortgagors the payment of the IMU tax on the main dwelling does significantly reduce the probability of buying a vehicle. Furthermore, Figure J.1 of Online Appendix J sho