Fiscal consolidation after the Great Recession: the role of composition

Fiscal consolidation after the Great Recession: the role of composition Abstract We have examined the fiscal consolidation episodes in a group of OECD countries from 2009 to 2014. The range of the estimated short-term fiscal multiplier runs from 1.2 to 2.0, larger than those obtained in more ‘normal times’, implying that the contractionary effect has been larger in depressed environments. Nevertheless, we also found that revenue measures have a higher and more persistent real impact than expenditure measures, which is more consistent with the influence of current consolidations on the expectations about the future path of fiscal policies (the expectations channel). This result suggests that expenditure cuts are less harmful for the economy than tax hikes. 1. Introduction In the wake of the global financial crisis, the fall in revenues, the fiscal stimulus, and the realization of contingent liabilities, mostly related to the support provided to the financial system, triggered a considerable increase in the public deficit in advanced economies, reaching 9% of GDP and a debt ratio over 90% in 2009 (IMF, 2014). The long-term costs of high public debt are well known from the literature, explaining why so many countries subsequently committed to fiscal adjustment programs to ensure the sustainability of their public finances. By 2015 the deficits have been reduced significantly, but public debt to GDP ratios in the advanced economies have still not stabilized and stand at historical levels (IMF, 2014). Nevertheless, the fiscal policy debate has focused on the short-term effects of fiscal consolidation and, consequently, on the appropriate speed of debt reduction and the composition between revenue and expenditure. This paper analyses the effects of fiscal consolidation for a group of OECD countries in the period 2009–2014 and compares it with previous consolidation periods. There are several possible reasons why the current episode of large fiscal retrenchment is so very different from previous ones. Six years after the onset of the crisis, output gaps and cyclical unemployment still loom large in many advanced economies. Monetary policy is very expansionary, but given the zero lower bound of interest rates, financial conditions remain tight. Moreover, a synchronized fiscal adjustment across several major economies may adversely impact the recovery. In fact, Blanchard and Leigh (2013) have already shown that a negative relationship between fiscal consolidation forecasts and subsequent growth forecast errors has been responsible for a slower than expected recovery in a number of advanced economies during the period 2010–2011. Empirically, the problem is to correctly identify the effects of fiscal policy on output. When analysing cross-country evidence, a standard method has been to use the cyclically adjusted primary balance (CAPB) to approximate discretionary changes in fiscal policy. From another standpoint, traditional VAR methods assume that fiscal changes are uncorrelated with other determinants of output. However, as pointed out by Romer and Romer (2010), these fiscal variables may include non-policy changes correlated with output. That is particularly relevant when using annual data, since agents may be responding to the fiscal change within the year and that correlation would bias the analysis against the contractionary effects of fiscal consolidation. To address these identification problems, the literature has used historical records—the narrative approach—to look for fiscal policy actions that are aimed at reducing the budget deficit rather than in response to current economic conditions. To that end we have constructed a dataset of government spending cuts and revenue increases from a broad set of policy documents for a group of advanced economies. Alesina et al. (2015) analysed historical records up to 2007 to simulate multi-year fiscal plans in 2009–2013. In our case we have looked to the cost of fiscal consolidation for a wider set of advanced economies using a new dataset from the period 2009–2014. The collected measures are ex-post outcomes and we do not try to separate the expected from the unexpected component of the fiscal change. Against this background, we are able to compare our data with the fiscal consolidation records in more ‘normal times’ (1979–2009) previously studied in the literature (Guajardo et al., 2014). Moreover, following a line of the literature, we study how the current fiscal policies affect the expected future path of fiscal policies (the expectation channel) in the most recent period when considering revenue and expenditure consolidations in a separated way. The next section of the paper first presents a brief discussion of the theoretical predictions of fiscal consolidations, and second describes some characteristics of the fiscal consolidation taking place after 2009, showing that the efforts have been higher than in previous episodes. We also present a test of exogeneity of fiscal actions compared with standard CAPB measures. Section 3 reports the main econometric results. We estimate the dynamic fiscal multiplier on output in a single-equation specification and compare it with an alternative VAR specification. We also report the differences between revenue and expenditure, considering only large consolidations. The evidence points to higher fiscal multiplier effects after the financial crisis and to higher and more persistent impact of revenue-based rather than expenditure-based actions. The latter evidence goes in the same direction as the one reported by Alesina et al. (2015). Moreover, we found that revenue multipliers are larger while expenditure multipliers become non-significant when only large consolidations are considered. Section 4 presents different robustness exercises. First, we add to the VAR analysis some financial variables that the literature has found relevant in explaining the idiosyncratic country characteristics of the fiscal multipliers. The results from the previous section then become clearer. Second, given the interdependence of monetary and fiscal policy, we analyse whether it has any significant influence on the estimated fiscal multipliers. Changes in policy interest rates are shaping the magnitude of the fiscal actions, especially when the most recent period, affected by the zero lower bound, is excluded. Third, we study if there are any specific fiscal effects for the euro area countries, since they have been subject to a common monetary policy and fiscal framework. The confidence channel, in the presence of a financial stress environment, may explain the lower real effects of fiscal actions in the euro area. To conclude, Section 5 presents a summary of the main results together with some open issues for future research. 2. Fiscal consolidations: theory and the most recent data 2.1. Theoretical predictions There is a general view that, in normal times, fiscal consolidations and the resulting government debt reduction contributes to long-term growth. However, there is no consensus regarding the short-term effects of fiscal austerity. The prediction of traditional Keynesian models is that cutting government spending or raising taxes reduces the economic activity in the short term. In the IS-LM model, consumers behave in a non-Ricardian fashion and their consumption is a fraction of their current disposable income. The size of the fiscal multiplier depends mainly on the marginal propensity to consume. But other factors may also be relevant, like how accommodative is the monetary policy, the investment response to the fiscal shock, or the fiscal sustainability of the country. The contractionary effect of fiscal adjustments would be consistent with the empirical evidence analysing aggregate time-series models of a short-term spending multiplier that lies between 0.6 and 1.8 (Ramey, 2011). In general, New Keynesian models with sticky-prices and neoclassical foundations that feature infinitely lived Ricardian households tend to reduce the size of the multiplier (Smets and Wouters, 2007), whereas the introduction of rule-of-thumb consumers increases the multiplier (Galí et al., 2007). The efficacy of fiscal policy may also be affected by the special circumstances we have seen in many advanced economies after the global financial crisis. This is the case of the zero lower bound of short-term nominal interest rates. Christiano et al. (2011) show that in a deflationary setup with monetary authorities committed to keep interest rates constant for a considerable period of time, an expansionary fiscal policy leads inflation expectations to rise and the multiplier is likely to be larger1. Another line of the literature has highlighted the importance of composition between revenue and expenditure when studying consolidation efforts and the existence of non-Keynesian effects. Giavazzi and Pagano (1990) were the first to show that large expenditure-based fiscal adjustments could be expansionary. More recently, Alesina and Ardagna (2010) found that spending cuts were much more effective than tax increases on large fiscal consolidations, and that they were associated with economic expansion across a panel of OECD countries. Significant fiscal consolidation today (particularly if implemented with large expenditure cuts) may reduce the need for future fiscal action, raising the current confidence of households and firms. For example, Bertola and Drazen (1993) outlined a model where the expectations about future fiscal policies affected the effect of current fiscal policies. Thus, when this expectations channel dominates the pure multiplier effect, consolidation will have a positive effect on output. Given the scale of the fiscal efforts across countries in the wake of the global crisis and the sizeable revenue increases, it is an open issue if there is any evidence of ‘expansionary fiscal contractions’ when considering the most recent time period. 2.2. Fiscal consolidations after the financial crisis: 2009–2014 After the global financial crisis many economies have been involved in a process of fiscal consolidation.2 The main objective was to reduce the fiscal deficit that increased significantly as a consequence of the 2008 financial crisis. For example, the average discretionary fiscal stimulus in the G-20 economies was around 2% of GDP in 2009 and 2010 and their average fiscal deficit reached 6% in 2010 (see IMF, 2010). In consequence, at the Toronto Summit in 2009 they committed to fiscal plans that would halve deficits by 2013. We follow the historical approach proposed by Romer and Romer (2010) to identify fiscal consolidations after the financial crisis.3 The historical episode narrative aims to separate policy changes from those arising from non-policy developments. The set of documents reviewed for this purpose relies on Stability and Convergence Programmes submitted annually to the European Commission, national budgets, OECD Economic Surveys, and IMF Staff Reports, as well as national sources such as the Congressional Budget Office (CBO) and several Memorandums of Understanding (MoUs). Where possible we have reviewed the measures announced, using the most up-to-date document available, and have checked those documents using the retrospective analysis included in the IMF’s Article IV reviews and in the OECD’s ‘Restoring Public Finances’ documents. All the budgetary measures are fully credible and implemented in the year assigned by the official documents. The sample includes the revenue-based and expenditure-based fiscal actions taken by 27 economies in the period 2009–2014 (see Appendix 1, Table A1). These are 25 OECD countries plus Latvia and Lithuania, which are now euro area countries. Thus, the majority are advanced countries, but there are also two emerging economies (Mexico and Turkey). In total we have identified 101 cases of action where these countries took some budgetary measures. During the period 2010–2014, more than 70% of the economies included in the sample were immersed in fiscal consolidation. The median of the adjustments is 1% of GDP and the average is 1.76% of GDP, with a standard deviation of 0.18 pp. The range runs from –3.0% to 9.75% of GDP (see Fig. 1). The negative figures correspond to the expiration of temporary fiscal measures, for example Estonia in 2010 and 2011, and the figures over 9% of GDP correspond to Estonia and Latvia in 2009. Fig. 1. View largeDownload slide Narrative episodes of fiscal consolidation 2009–2014: size distribution Fig. 1. View largeDownload slide Narrative episodes of fiscal consolidation 2009–2014: size distribution The current fiscal adjustment episodes are very different from the previous ones studied by Guajardo et al. (2014). During the period 1978–2009, the fiscal adjustments of 15 advanced economies averaged 0.99% with a standard deviation of 0.94. Thus, although the number of observations and of years in our sample is smaller, it contains more countries, the average size of fiscal consolidation is larger, and it has lower variance. If we define large consolidations as consolidation efforts amounting to more than 1.5% of GDP, as in Alesina and Ardagna (2010), we find a total of 67 large fiscal consolidations. Most observations in the sample are concentrated between 2011 and 2013 (see Fig. 2). The 2009 data are driven by the Baltic economies that suffered a sudden stop and a credit crunch in 2008. In 2009 they started to consolidate very significantly (an average of 8%), in a year dominated by spending cuts (see Fig. 3). In 2010 there was an increase in the number of countries tightening their budgets, including the euro area economies subject to market pressure. Finally, in 2014 the pace of fiscal consolidation abated, with fewer countries making consolidation efforts and expenditure cuts exceeding tax increases. Fig. 2. View largeDownload slide Narrative episodes of fiscal consolidation, 1978–2014 Fig. 2. View largeDownload slide Narrative episodes of fiscal consolidation, 1978–2014 Fig. 3. View largeDownload slide Average size of narrative fiscal consolidations Fig. 3. View largeDownload slide Average size of narrative fiscal consolidations The composition of fiscal adjustments is critical for their effects on the economy. Figs 2 and 3 show that the adjustments made in the 2009–2014 period were fairly equally split between tax-based and expenditure-based actions. However, our sample contains several examples of important consolidations affecting both expenditure and revenue, so it will be crucial to estimate the impact of both types of consolidations jointly. This observation contrasts with the fiscal adjustment episodes examined by Alesina et al. (2015). They find that most consolidation efforts made in the 2009–2013 period have been based on expenditure cuts. We have found that the calculated total consolidation is fairly similar for six out of the eleven countries their dataset contains. The starkest difference occurs in Spain, where we have detected a total consolidation of 10.5% of GDP, fairly balanced between expenditure cuts and tax hikes, while Alesina and co-authors computed a total consolidation of more than 15% of GDP, almost entirely based on adjustments in expenditure. Our balanced consolidation is consistent with some country-specific studies about fiscal consolidation in Spain, with some authors putting more weight on the revenue side (Hernández de Cos and López Rodríguez, 2014). Moreover, with respect to 2012–2013, several official documents identify a balanced path of consolidation, with total consolidation amounting to around 6.5%–7.5% of GDP (see e.g. IMF, 2013), in contrast with more than 9% of anticipated and unanticipated measures in Alesina et al. (2015). Finally, we have also observed significant differences from one region to another. Overall, advanced non-stressed euro area and emerging economies are the regions that have consolidated less, whereas Baltic and stressed euro area countries are the ones that have made most efforts. In relative terms, taxes have been raised more in the emerging countries and expenditure has been cut more in the Baltic and the advanced non-European countries. 2.3. Comparison of fiscal policy measures We have said that the conventional approach to measure policy-driven changes in fiscal policy is through the cyclically adjusted primary balance (CAPB) that excludes changes in fiscal variables induced by business cycle fluctuations. To have a uniform indicator, we selected the variable constructed by the IMF’s Fiscal Monitor database. In total there are 101 cases where these 27 countries took budgetary actions as measured by CAPB. The median of the adjustments is 0.77% of GDP and the average is 1.2%, with a standard deviation of 0.12 pp. Thus, in general, the CAPB adjustments are smaller than those identified in the narrative approach. Before analysing the real effects of the fiscal narrative measures, we compare them with the more standard CAPB measures. Figure 4 plots the observations of the two fiscal consolidation indicators in the period 2009–2014. There are no major discrepancies observed between the two, and in principle it would be impossible to say which measure provides a more reliable identification. However, we have found a total of eight episodes with large differences, defined as more than 3 pp of difference between the two approaches. And we have identified five large one-off accounting measures that could explain the inaccuracies in the CAPB (Ireland 2009, Lithuania 2010, Hungary 2011, Spain 2012, and Slovakia 2013). Fig. 4. View largeDownload slide Narrative episodes vs. CAPB: 2009–2014 Fig. 4. View largeDownload slide Narrative episodes vs. CAPB: 2009–2014 The econometric findings for the period before 2009 are that the narrative episodes are more exogenous to output than the CAPB.4 But if the financial crisis period is included, the CAPB measure may possibly be more accurate. In order to check for contemporaneous orthogonality with output, we perform a test of weak exogeneity of both the CAPB and the narrative measures. Following Guajardo et al. (2014), we construct a measure of economic surprises, based on the IMF’s GDP forecast revisions. We define the economic news index ( Newsit) as the log-difference between GDP at time t in the October World Economic Outlook (WEO) of year t and GDP at time t of the WEO at time t-1 for each country. We perform the following regression:   ΔFit= μi+λt+β Newsit+uit (1) where ΔFit is the fiscal consolidation measure (CAPB or narrative-based), μi is an unobservable country-fixed effect, and λt is a common year effect for all economies. In order to avoid potential small-sample bias, we merge our sample with the 1978–2009 period, giving 373 observations for 15 economies that had consolidation episodes in both periods.5,6 This exercise does not test for orthogonality with past (and future) output developments. In fact, the current fiscal actions are likely to be correlated with past large fiscal deficits and will be used as instruments to predict output movements. Table 1 presents the results. The β coefficient of the news index on the narrative measure equation is negative and only weakly significant, while the coefficient on the CAPB equation is positive and strongly significant. This result is consistent with that obtained for the pre-crisis period. Moreover, the explanatory power of the narrative-based measures (0.26) is lower than that of the CAPB measures (0.48), supporting the theory of the greater exogeneity of the narrative-based approach to contemporaneous changes to output movements. In addition, when we perform the same test for narrative-based revenue and expenditure, we find that the coefficient related to the tax changes is near zero, it is not significant, and it is less correlated with output than with government spending changes. The greater exogeneity of tax changes signals that spending actions may respond to tax actions or other variables correlated with output, and therefore that revenue actions will be more help in the identification strategy. Another avenue, as in Ramey and Shapiro (1998), would be to rely on specific government spending items that do not respond to economic events. Table 1. An exogeneity test of fiscal policy changes (1978–2014) Equation: ΔFit=μi+λt+βNewsit+uit Dependent variable: ΔFit  Narrative fiscal consolidations  CAPB  Narrative revenue consolidations  Narrative expenditure consolidations  β coefficient  –0.120  0.254  –0.037  –0.082    [0.067]*  [0.081]***  [0.032]  [0.039]*  Obervations  373  371  373  373  No. of countries  5  15  15  15  R2  0.256  0.482  0.205  0.228  Dependent variable: ΔFit  Narrative fiscal consolidations  CAPB  Narrative revenue consolidations  Narrative expenditure consolidations  β coefficient  –0.120  0.254  –0.037  –0.082    [0.067]*  [0.081]***  [0.032]  [0.039]*  Obervations  373  371  373  373  No. of countries  5  15  15  15  R2  0.256  0.482  0.205  0.228  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. In order to test for the exogeneity of our more recent sample, we exploit the fact that narrative fiscal measures after the Great Recession were in general of a bigger magnitude. Therefore, we estimated a quantile regression of eq. (1). The results are shown in Table 2. Innovations to GDP growth are not significant at different points of the distribution. Moreover, they are not significant at the higher percentiles, which are mostly stacked in the more recent period. Table 2. Exogeneity test by quantile Dependent variable: ΔFit  Narrative fiscal consolidations  β coefficient    20% percentile  0.018  40% percentile  –0.144  60% percentile  –0.095  80% percentile  –0.346  90% percentile  –0.264  Observations  373  No. of countries  15  Dependent variable: ΔFit  Narrative fiscal consolidations  β coefficient    20% percentile  0.018  40% percentile  –0.144  60% percentile  –0.095  80% percentile  –0.346  90% percentile  –0.264  Observations  373  No. of countries  15  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. *, **, *** denote significance at the 10%, 5%, and 1% level, respectively. The coefficients are estimated following Powell (2015). 3. The estimated effects of fiscal consolidation on economic activity The estimation strategy first examines the output effect on a single-equation specification and later on a vector autoregressive (VAR) model. In order to make the fiscal multipliers comparable across specifications, we normalize the fiscal change so that the CAPB to GDP ratio rises by 1% of GDP on impact. In the VAR model, we will report the short-term fiscal multiplier after one year and the medium-term fiscal multiplier after four years, dividing the cumulated output response by the CAPB response. The first section presents the results for the aggregated fiscal measures and analyses the real effects before and after the financial crisis. The second section presents the composition results, splitting the fiscal actions between revenues and expenditures, and considers also large fiscal consolidations as defended by the non-Keynesian view. 3.1. Fiscal multipliers Following the empirical literature in this area, we estimate the effect of fiscal consolidation on a single-equation specification that regresses real output growth ( ΔYit) on lagged output growth and the contemporaneous values and lags of the fiscal changes ( ΔFit−s). It takes the following form:   ΔYit= μi+λt+∑s=12γsΔYit−s+∑s=02δsΔFit−s+ uit (2) Including past output allows us to control for the state of the economy in each country and thus helps estimate the discretionary part of fiscal action. As is standard in the panel data approach, the estimation includes country-fixed effects and year-fixed effects, and based on the information criteria we choose two lags in the dynamic specification.7 We performed panel unit root tests for GDP, CAPB, and the narrative fiscal measures, showing that the variables are all (weakly) stationary. Denoting ΔFit the CAPB-based fiscal measures, the estimation by ordinary least squares (OLS) of the regression coefficients would be biased by several factors. First, the presence of an unobservable country-fixed effect would be correlated with the error term. In order to deal with country heterogeneity, we will estimate our set of parameters using a fixed-effect estimator, although this estimator will be biased under the presence of the lagged output on the right-hand side for a small T specification. Second, if there is endogeneity in the CAPB measures, it will be correlated with the error term, and therefore an instrumental variables approach should be used. In particular, we use a two-stage least squares (TSLS) estimator with our exogenous narrative measures as an instrument for the CAPB. Third, when the number of observations per country is small (a problem present mainly in the 2009–2014 sample), a Nickell bias related to the correlation between the lagged dependent variable and the error term is more likely to be found. Against this background, this bias will overestimate the negative sign of the fiscal multiplier. To overcome this problem, we use a Generalized Method of Moments (GMM) estimator. However, the small number of countries in our sample precludes us from using the full amount of moment restrictions available, i.e. all the lags of the dependent variable (and the CAPB), as instruments, and in consequence we use the Anderson-Hsiao estimator. Our fourth specification is a natural variation of the first one. We run a three-vector auto-regression model (VAR) with the change in GDP, the change in the CAPB ratio, and the narrative fiscal measures.8 Thus, lagged output and past cyclically adjusted primary balance also affect the current cyclically adjusted primary balance. This specification also includes two lags and a full set of country and time effects. Consistent with the tested exogeneity of the historical episodes of the fiscal consolidations and the TSLS estimation in the single-equation approach, the narrative measures are ordered first in the VAR and second in the CAPB, allowing it to have a contemporaneous effect on output (the last variable in the system equation) when considering a Cholesky orthogonalization of the residuals. The estimation results under the three single-equation specifications are shown in Table 3. The first column presents the fiscal multipliers in the OLS case with fixed effects, the second column the results with the TSLS estimator, and the third column the results with the GMM estimator. We show the effect of a 1% change of fiscal consolidation in GDP, at the time of the consolidation, and calculate the dynamic response function one year later. Table 3. Estimated dynamic output effect of a 1% CAPB change: single-equation specification Equation: ΔYit=μi+λt+∑s=12γsΔYit−s+∑s=02δsΔFit−s+ uit     OLS  TSLS  GMM  Observations  2009–2014 (25 countries)  (1)  –0.691  –1.951  –0.961  150  [0.221]***  [0.521]***  [0.23]***  1978–2009 (15 countries)  (2)  0.246  –0.871  –  445  [0.121]**  [0.335]***  1978–2014 (15 countries)  (3)  0.138  –1.180  –  520  [0.115]  [0.33]***  (4)  –  –1.560  –  520  [0.467]***      OLS  TSLS  GMM  Observations  2009–2014 (25 countries)  (1)  –0.691  –1.951  –0.961  150  [0.221]***  [0.521]***  [0.23]***  1978–2009 (15 countries)  (2)  0.246  –0.871  –  445  [0.121]**  [0.335]***  1978–2014 (15 countries)  (3)  0.138  –1.180  –  520  [0.115]  [0.33]***  (4)  –  –1.560  –  520  [0.467]***  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. Accumulated output effect after one year. The instruments in the two-stage least squares (TSLS) estimation are the narrative fiscal measures. In panel (4) the instruments are the narrative fiscal revenue measures. In the GMM specification, narrative measures are included as external instruments and the second lags of GDP and CAPB are used as GMM instruments (18 instruments in total). Panel 1 shows the results for the period 2009–2014, after the Great Recession, with the full set of 25 countries. The OLS estimator shows an effect of –0.69 pp one year after the consolidation. The effect is statistically significant, and after that period the responses on output stabilize. In column 2, CAPB is instrumented with the narrative measures.9 The effect after one year is a highly significant 1.95 pp, almost three times more than with the OLS. As expected, the GMM estimator lowers the calculated fiscal multiplier to 0.96.10 In Panel 2 we report the results with a smaller sample of countries (15 instead of 25) for the period 1978–2009, given our interest in comparing the fiscal consolidation effects after the financial crisis with the international episodes before 2009. The estimates reproduce the stark differences between OLS and TSLS found by Guajardo et al. (2014):11 the positive OLS estimate (0.24) with the CAPB measure is consistent with the ‘expansionary austerity’ found previously in the literature; the instrumental estimate with the narrative shocks is negative (–0.87) and significant. Note that this discrepancy between estimation methods was not present when studying only the crisis period in Panel 1. In Panel 3 the analysis is extended to include the crisis period. The qualitative results are maintained, but the short-term effects on real activity are larger (–1.18 with TSLS). This is primary evidence that the contractionary effect of consolidations was larger in depressed environments and is consistent with recent findings that have stressed the relevance of the state of the economy when measuring the size of fiscal multipliers (e.g. Auerbach and Gorodnichenko, 2012). Figure 5 presents the same TSLS estimation in Panel 3, but in a rolling window compressing 14 years. The fiscal multiplier shows a remarkable stability until the last three years are included in the estimation, when the multiplier drops by almost 0.2. This is suggestive evidence of a different effect of consolidations in the most recent sample period.12 Fig. 5. View largeDownload slide Responses of output to fiscal shocks, rolling window Notes: All specifications contain a full set of country- and time-fixed effects. TSLS estimation of the effect of CAPB on GDP, using the narrative fiscal measures as instruments. Dashed lines represent two-standard-deviation confidence intervals. The figure shows the accumulated output effect after one year. Fig. 5. View largeDownload slide Responses of output to fiscal shocks, rolling window Notes: All specifications contain a full set of country- and time-fixed effects. TSLS estimation of the effect of CAPB on GDP, using the narrative fiscal measures as instruments. Dashed lines represent two-standard-deviation confidence intervals. The figure shows the accumulated output effect after one year. Finally, in Panel 4, we use the CAPB only with the narrative measures from the revenue side. This is based on the idea that the main motivation for tax changes, usually implemented through new legislation, may be to reduce the inherited fiscal deficit, whereas expenditure changes may respond more to other developments affecting output. The estimated fiscal multiplier in the TSLS estimation (–1.56) supports the possibility of a downward bias towards zero in the above multipliers because the expenditure measures are less exogenous than the revenue ones, which is consistent with the exogeneity test shown in Section 2. Figure 6 depicts the impulse response function from the benchmark VAR specification together with one-standard-deviation confidence bands.13 The size of the shock on the narrative fiscal measures is normalized to generate a response on impact (t = 0) of 1% of GDP on the CAPB. That allows the VAR response to be comparable with the single-equation results. We can see first that fiscal consolidation has a significant negative effect on output. Second, the chart shows that the response is more contractionary when the international sample includes the most recent period after the financial crisis. And finally, it is clear that the real response to the fiscal shock is more persistent with the full sample, since there is a stronger tendency for output to return to normal when the sample ends in 2008. Fig. 6. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Fig. 6. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Table 4 shows the value of the output response after one year and the corresponding standard error in the VAR. Together with the accumulated response of output to a narrative shock, we report the fiscal multipliers both in the short and the long run, calculated as the accumulated response of output divided by the accumulated response of CAPB. The results shown for the two sample periods (columns 1 and 2 in Table 3) coincide with the instrumental variable estimation in the single-equation specification: there is a significant contractionary effect of fiscal consolidation, and that effect is larger when considering the most recent period.14 Table 4. Response of output to a narrative fiscal shock: VAR specification   Benchmark VAR  Benchmark VAR    1978–2009  1978–2014  Response of output to a shock of 1% of CAPB  –1.715  –2.076  [0.420]***  [0.343]***  Fiscal multiplier (after one year)  –1.004  –1.143  Fiscal multiplier (after four years)  –0.731  –0.989  Observations  445  520    Benchmark VAR  Benchmark VAR    1978–2009  1978–2014  Response of output to a shock of 1% of CAPB  –1.715  –2.076  [0.420]***  [0.343]***  Fiscal multiplier (after one year)  –1.004  –1.143  Fiscal multiplier (after four years)  –0.731  –0.989  Observations  445  520  Notes: Robust standard errors are obtained using Monte Carlo simulation. All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation. The accumulated output response is after 1 year. The fiscal multiplier is the ratio between the cumulated effect on output and the cumulated effect on CAPB. In order to test for parameter stability in the more recent period, we perform a test of whether overall parameter values are unchanged after and before 2009, when we think a break date may exist. Against this background, we perform a sample-split test by introducing in the VAR model an interacted dummy variable in the right-hand side of all the regressions, and we compute our structural break statistic as in Sims (1980):   (T−k)log⁡|Σre|−log⁡|Σun| (3) where T is the number of observations, k is the number of regressors, and Σre and Σun are the residual covariance matrices for the restricted and the unrestricted model. Under the null hypothesis of parameter stability, the test statistic is asymptotically chi-squared with the degrees of freedom equal to the total number of constraints. The value of the statistic χ2(18)=42.88 corresponds to a significance level below 0.005, rejecting the null. However, this result could be driven by the important economic changes (and not just fiscal) as a consequence of the crisis. In order to check if the fiscal variables are drivers of the structural change, we have corroborated that the coefficient of the dummy variable interacted with the narrative consolidations is negative and significant in the regression with GDP on the left-hand side. 3.2. The composition effect The standard literature that supports ‘expansionary fiscal contractions’ has emphasized the role of composition (e.g. Alesina and Ardagna, 2010). Their evidence shows that fiscal adjustments based on spending cuts are more effective than tax increases in stabilising the debt ratio and avoiding economic contraction. The approach has been to identify historical cases of fiscal retrenchment, looking at the cyclically adjusted changes in fiscal variables. Here we investigate the relevance of composition by including the narrative fiscal expenditure and revenue measures in the previous VAR framework.15 We consider both sets of measures at the same time given the importance of consolidations involving both expenditure and revenue measures, in order to account for the combined effect. In the 4-variable VAR, revenue and expenditure are first and second in the order of the system, which is consistent with our exogeneity evidence in Section 2. Figure 7 depicts the impulse response functions of revenue and expenditure shocks to output for the 4-variable VAR. We find a significant negative contemporaneous impact of the expenditure shock, returning to the baseline after two years. By contrast the revenue shock has a stronger impact contemporaneously on GDP and a much more persistent effect. Fig. 7. View largeDownload slide The response of output to narrative fiscal shocks: the composition effect Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Fig. 7. View largeDownload slide The response of output to narrative fiscal shocks: the composition effect Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Table 5 confirms the previous results when taking into account the CAPB response and calculating the fiscal multiplier dividing the accumulated output response by the accumulated CAPB ratio response. After one year the expenditure multiplier (–0.67) is significantly smaller than the revenue multiplier (–1.54). Similarly, four years after the shock, the expenditure multiplier (–0.43) is much smaller than the revenue multiplier (–1.50). Table 5. Response of output to a narrative fiscal revenue/expenditure shock (1978–2014): VAR specification   4-variable VAR  4-variable VAR (large consolidations)  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.561  [0.529]***  [0.704]***   Fiscal multiplier (after one year)  –1.545  –2.117   Fiscal multiplier (after four years)  –1.498  –2.315  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –0.839  [0.476]**  [0.51]*   Fiscal multiplier (after one year)  –0.673  –0.478   Fiscal multiplier (after four years)  –0.434  –0.260   Observations  520  520    4-variable VAR  4-variable VAR (large consolidations)  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.561  [0.529]***  [0.704]***   Fiscal multiplier (after one year)  –1.545  –2.117   Fiscal multiplier (after four years)  –1.498  –2.315  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –0.839  [0.476]**  [0.51]*   Fiscal multiplier (after one year)  –0.673  –0.478   Fiscal multiplier (after four years)  –0.434  –0.260   Observations  520  520  Note: See Table 3. As in Guajardo et al. (2014), fiscal consolidations are contractionary even when they are based on spending cuts. But here we also find that for certain specifications expenditure adjustments are significantly different from revenue increases. Thus, these results are also consistent with Alesina et al. (2015), suggesting that expenditure cuts are less harmful for the economy than tax hikes. Similarly, Beetsma et al. (2015) report that consolidation affects consumer confidence negatively but more significantly through the revenue component than the spending component. The non-Keynesian view is that large fiscal adjustments, especially expenditure-based ones, are more effective in avoiding economic downturns, based on the argument that cutting sensitive items such as transfer programmes or government consumption may signal a credible commitment to long-term debt reduction. To test that hypothesis, we re-estimated the 4-variable VAR with the narrative revenue measures greater than 1% of GDP, and 0 otherwise, and similarly for the narrative expenditure measures. We find nine cases where both variables detect a large consolidation; we also find 35 more cases of large expenditure consolidations and 18 cases of only large revenue consolidations. Of the total 71 cases of large consolidations, 23 were detected in the 2009–2014 period. However, given the limited number of observations, the following results should be treated with caution. Column 2 in Table 5 summarizes the results. We find a significant negative response of large revenue-related fiscal consolidations, amounting to an output multiplier that stands at –2.1. The interesting result is in the effect of large expenditure-related consolidations. The effect after one year is negative but not different from zero. Specifically, we find that large expenditure consolidations, after a negative contemporaneous real effect, have a non-significant effect after one year and onwards. Although the data limitation problem becomes more of a concern if we investigate this effect in the 1978–2009 period, it seems that both large expenditure-related and revenue-related consolidations are more contractionary when the recent crisis period is considered. Thus, consistent with the ‘expansionary fiscal contraction’ literature, we find evidence that the composition of fiscal consolidation matters. The large expenditure-based adjustments performed when considering the Great Recession have fiscal multipliers that are not significantly different from zero, whereas large revenue-related consolidations are highly contractionary and very persistent. 4. Robustness exercises 4.1. Controlling for financial factors We first test the robustness of our findings by increasing the benchmark VAR with potentially relevant idiosyncratic country characteristics like the financial position. The fiscal consolidation periods are in many cases related to situations of large public debt and/or financial stress combined with other macroeconomic imbalances that may be perceived as affecting sovereign risk. In order to control for such factors, we have considered two additional variables, namely government debt to GDP in the previous period and an index of the sovereign ratings16 that are included at the end in the ordering of the VAR. We aim to identify fiscal sustainability problems reflected in financial variables, rather than indicators that may be a sign of short-run concerns. In any case, the sovereign rating indicator attempts to control for fundamental problems (low growth prospects, high external debt, weak banking sector, etc.) that could have fiscal sustainability implications. Figure 8 presents the impulse response function from the 5-variable VAR specification. The greater real effect when considering the most recent sample period is more evident when controlling for financial variables. Fig. 8. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. The 5-variable VAR includes lagged debt and sovereign rating index. Dashed lines represent one-standard-deviation confidence intervals. Fig. 8. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. The 5-variable VAR includes lagged debt and sovereign rating index. Dashed lines represent one-standard-deviation confidence intervals. Table 6 quantifies the fiscal multipliers after controlling for financial factors (column 2). The results are robust to the inclusion of these variables since the short-run fiscal multiplier stays around 1.2, a value similar to that obtained in the instrumental variable estimation and the benchmark VAR (column 1). We see a slightly larger multiplier in the short run when controlling for public debt, consistent with the results in Burriel et al. (2009). Table 6. Response of output to a narrative fiscal revenue/expenditure shock (1978–2014): VAR specification with financial variables as controls   Benchmark VAR  Additional variables: debt, sovereign rating index  Narrative consolidations       Response of output to a shock of 1% of CAPB  –2.076  –2.159  [0.343]***  [0.564]***   Fiscal multiplier (after one year)  –1.143  –1.220   Fiscal multiplier (after four years)  –0.989  –0.900  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.060  [0.529]***  [0.730]***   Fiscal multiplier (after one year)  –1.545  –1.590   Fiscal multiplier (after four years)  –1.498  –1.354  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –1.275  [0.476]**  [0.878]   Fiscal multiplier (after one year)  –0.673  –0.753   Fiscal multiplier (after four years)  –0.434  –0.060   Observations  520  391    Benchmark VAR  Additional variables: debt, sovereign rating index  Narrative consolidations       Response of output to a shock of 1% of CAPB  –2.076  –2.159  [0.343]***  [0.564]***   Fiscal multiplier (after one year)  –1.143  –1.220   Fiscal multiplier (after four years)  –0.989  –0.900  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.060  [0.529]***  [0.730]***   Fiscal multiplier (after one year)  –1.545  –1.590   Fiscal multiplier (after four years)  –1.498  –1.354  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –1.275  [0.476]**  [0.878]   Fiscal multiplier (after one year)  –0.673  –0.753   Fiscal multiplier (after four years)  –0.434  –0.060   Observations  520  391  Interestingly, the differences are more marked when we focus on the composition effect. Figure 9 presents the output response to the expenditure and the revenue shock whereas the lower panels in Table 6 compare the estimated fiscal multipliers in the 6-VAR specification (column 2) with the ones in the previous 4-VAR specification (column 1). Now, the differences between the two shocks increase since the responses to the expenditure shock are not significant after one year. Fig. 9. View largeDownload slide Responses of output to a narrative revenue/expenditure shock, with financial variables as controls Fig. 9. View largeDownload slide Responses of output to a narrative revenue/expenditure shock, with financial variables as controls 4.2. Monetary policy influence The interaction between fiscal and monetary policy can greatly affect the size of fiscal multipliers. For example, in a Keynesian framework, monetary policy may react to fiscal consolidation episodes by reducing interest rates because inflationary pressures diminish. Moreover, under a Taylor-rule based monetary reaction, fiscal consolidation could produce a negative output gap, therefore leading to a drop in interest rates. Similarly, the probable response of the exchange rate could help cushion the impact of fiscal retrenchment on domestic demand. However, as Christiano et al. (2011) have shown, the counteracting effect of monetary policy could be less noticeable in a context where the economy hits the zero lower bound (ZLB) of interest rates since the space for more accommodative policies is exhausted. Our sample includes an important period where the ZLB is present for most economies, and therefore we expect a weaker response of monetary policy to fiscal developments. Consequently, we anticipate a lower fiscal multiplier if we control for interest rates in the period previous to 2009. To check that out, a policy interest rate17 is included in a 4-variable VAR of fiscal consolidations, with the results presented in Table 7. The estimated multiplier, –1.0, is not very much affected by the inclusion of monetary policy rates, which is consistent with a less responsive monetary policy when the crisis period is considered. That result is even stronger when the debt ratio and the sovereign rating index are considered in the VAR. In that case, the short-run fiscal multiplier over –1.0 remains robust, while in the 1978–2009 period the multiplier shrinks from –1.0 to less than –0.5. Table 7. Fiscal multipliers: the influence of monetary policy. VAR specification   1978–2009   1978–2014     4-VAR  6-VAR  4-VAR  6-VAR  Response of output to a shock of 1% of CAPB  –1.363  –0.691  –1.811  –1.953    [0.392]***  [0.488]  [0.344]***  [0.515]***  Fiscal multiplier (after one year)  –0.815  –0.486  –1.022  –1.132  Fiscal multiplier (after four years)  –0.544  –0.164  –0.844  –0.864  Observations  420  298  495  391    1978–2009   1978–2014     4-VAR  6-VAR  4-VAR  6-VAR  Response of output to a shock of 1% of CAPB  –1.363  –0.691  –1.811  –1.953    [0.392]***  [0.488]  [0.344]***  [0.515]***  Fiscal multiplier (after one year)  –0.815  –0.486  –1.022  –1.132  Fiscal multiplier (after four years)  –0.544  –0.164  –0.844  –0.864  Observations  420  298  495  391  Notes: See Table 3. With respect to the benchmark, 4-VAR includes an intervention interest rate. The 6-VAR incorporates the lagged debt-to-GDP ratio and the sovereign rating index. Thus, the ZLB on interest rates may have precluded the authorities from adopting a more accommodative monetary policy during the crisis that would have reduced the magnitude of the fiscal multipliers. Nevertheless, this analysis is largely limited by the inclusion of other variables reflecting the effects of non-conventional monetary policy actions after 2009 in many of the countries of the sample. Another line of investigation was the role of monetary policy in relation to the composition of the fiscal adjustments. We have obtained (not shown) that the short-run multiplier of revenue consolidation is not affected if we include monetary policy in the 1978–2014 period, while the expenditure multiplier is slightly reduced. This effect is stronger in the period 1978–2009, suggesting a more accommodative monetary policy stance for expenditure consolidations in the pre-crisis period, as suggested by the IMF (2010). 4.3. Euro area countries Under financial stress, the confidence channel may be more present in fiscal consolidation, for several reasons. First, consolidation today could avoid more extensive and more harmful consolidation in the future, as in the model presented in Bertola and Drazen (1993); second, risk premia could be reduced by the consolidation, reflecting a lower financial risk of sovereign debt. We could test these hypotheses by restricting our sample to the stressed euro area countries that received external financial support after the crisis. However, given the low number of countries in that group, we prefer to focus on all the euro area economies.18 Although these economies had very different fiscal positions, the consolidations in the euro area took place in a more financially restricted environment, with higher debt-to-GDP ratios. Against this background, the probability of an unstable sovereign risk scenario was greater. Additionally, the euro area economies are also highly interconnected, so spillovers from a large number of countries pursuing a consolidation of public finances at the same time could impact significantly on the size of the fiscal multiplier for the whole group. The results are summarized in Table 8. The one-year multiplier is close to unity in the benchmark 3-variable VAR (column 1). This –0.98 multiplier is lower than the estimated –1.14 for the whole sample (in Table 3). However, if we include lagged debt and the sovereign rating index, the multiplier is greatly affected, becoming non-significant. This difference of estimates between specifications in the euro area contrasts with the more stable multiplier found for the whole sample. Table 8. Fiscal multipliers: the euro area effect, VAR specification   Benchmark VAR  Additional variables: debt, sovereign rating index  Response of output to a shock of 1% of CAPB  –1.656  –0.771  [0.442]***  [0.53]  Fiscal multiplier (after one year)  –0.984  –0.457  Fiscal multiplier (after four years)  –0.733  –0.201  Observations  345  252    Benchmark VAR  Additional variables: debt, sovereign rating index  Response of output to a shock of 1% of CAPB  –1.656  –0.771  [0.442]***  [0.53]  Fiscal multiplier (after one year)  –0.984  –0.457  Fiscal multiplier (after four years)  –0.733  –0.201  Observations  345  252  Notes: See Table 3. The euro area economies included are Austria, Belgium, Finland, France, Germany, Italy, Netherlands, Portugal, and Spain. The estimation results are consistent with those of Guajardo et al. (2014), reported for a sample not including the recent crisis: fiscal consolidations preceded by high perceived sovereign default risk are less contractionary. Nevertheless, the estimation carried out in Table 8, only for the euro area countries, presents some instability depending on the chosen specification, and that may be due to the inclusion of the most recent years and the loss of observations after restricting the sample. 5. Conclusions We have examined the fiscal consolidation episodes that have taken place in a group of OECD countries after the global financial crisis (2009–2014). For that purpose we have constructed a dataset of policy actions—a narrative approach—from a broad set of official documents. Compared with previous periods of fiscal consolidation, during this episode the average size of the adjustment was larger, with more countries consolidating at the same time and with a strong focus on tax measures. Using dynamic panel data estimation, we are interested in the short-term effects of fiscal consolidation on economic activity. The different specifications—from single-equation to VAR systems—take into account the possible endogeneity of the regressors. Across all estimation methods the fiscal multiplier is negative and significant, in contrast to the results found previously with standard cyclically adjusted fiscal balance measures. Moreover, the average multiplier after one year is between –1.2 and –2.0, a higher multiplier than that found with historical episodes before 2009. We also obtain a significant real effect of revenue measures of around –1.6, while expenditure consolidations have an effect close to –0.7. These differences are even higher when looking at large consolidation episodes, with expenditure cuts having a non-significant effect after one year under certain specifications. This evidence showing the importance of the composition is closer to the expansionary fiscal contractions hypothesis, since it supports the view that spending cuts are more effective in stabilizing debt and avoiding economic downturns. In the last section we also present some evidence in favour of the need to consider non-conventional monetary policies to obtain a more accurate fiscal multiplier after the financial crisis and of the existence of a confidence channel for specific countries under financial stress that reduces the cost of fiscal consolidation. Finally, we believe there are two other natural extensions of this paper that need to be pursued. First, the current fiscal consolidation episodes are still ongoing in many economies and it is not yet possible to determine whether they have been successful in stabilizing and reducing high public debt ratios. Thus, more time observations will be needed to obtain a better assessment of this ongoing fiscal adjustment process. Second, our investigation has only disaggregated between revenues and expenditures. Efficiency arguments would also demand an analysis of current expenditure versus public investment and of direct versus indirect taxes. Supplementary material Supplementary material—the Appendix and the Data files—are available online at the OUP website. Footnotes 1 DeLong and Summers (2012) present the case of fiscal consolidations that reduce output in the medium term (self-defeated consolidation), because of the permanent effects of the recession. 2 According to the IMF (2016), mostly advanced economies have been the ones trying to stabilize their debt levels. By contrast, emerging economies, on average, had rising deficits and debt ratios after 2011. 3 We identify fiscal policy changes using historical documents as in Devries et al. (2011). 4 See, for example, Hernández de Cos and Moral-Benito (2013) and Jordá and Taylor (2013). 5 The simple contemporaneous correlation between the surprises on output and CAPB change for the 2009–2014 period is 0.59. The correlation with the narrative measure is –0.11. 6 The 15 countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, Spain, the UK, and the USA. 7 GDP data taken from the OECD Economic Outlook (November 2014) and Eurostat. 8 In order to check whether the multicollinearity between the CAPB and the narrative fiscal measure could inflate the variance of the results, we estimated a Variance Inflation Factor (VIF) for eq. (2). All VIF were lower than 1.6, suggesting that multicollinearity does not appear to affect our estimations. 9 The F-test of the first stage has a p-value of less than 0.05 in all the equations, reinforcing the explanatory power of our narrative measures on CAPB. 10 The Hansen test (p-value = 0.331) provides further evidence of the exogeneity of our instruments. 11 The sample comprises the same 15 countries as in Section 2. 12 This differential effect during the recent period may be caused by the cyclical environment (proxied by the rolling window estimation), the composition of the adjustment (see Section 3.2), or the size of the consolidation. Preliminary evidence based on quantile regressions points to a lesser role of the latter. 13 The Akaike information criterion pointed to a lag structure with two lags. 14 We computed this estimation accounting for the build-up of fiscal consolidation episodes prior to the adoption of the euro in 1999. The results were unchanged. 15 CAPB data for revenue and expenditure to compare the VAR with the instrumental variable estimation is not available for all countries. 16 The variable is taken from Broto and Molina (2014). 17 Policy interest rates for countries are taken from Datastream. 18 The euro area countries in this sample are Austria, Belgium, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, and Spain. We consider the whole 1978–2014 period, although in the first part of that period each economy had its own independent monetary policy. Acknowledgements We would like to thank participants at the Banco de España and the ESM seminars for their helpful comments. The views expressed in this paper are those of the authors alone and do not necessarily reflect the views of the Banco de España. References Alesina A., Ardagna S. ( 2010) Large changes in fiscal policy: taxes versus spending, in Brown J.R. (ed.) Tax Policy and the Economy , vol. 24, National Bureau of Economic Research, Cambridge, MA, 35– 68. Alesina A., Barbiero O., Favero C., Giavazzi F., Paradisi M. ( 2015) Austerity in 2009–2013, Working Paper No. 20827, National Bureau of Economic Research, Cambridge, MA. Auerbach A., Gorodnichenko Y. ( 2012) Fiscal multipliers in recession and expansion, in Alesina A., Giavazzi F. (eds) Fiscal Policy After the Financial Crisis , University of Chicago Press, Chicago, IL, 98– 102. Beetsma R., Cimadomo J., Furtuna O., Giuliodori M. ( 2015) The confidence effects of fiscal consolidations, Working Paper Series No. 1770, European Central Bank, Frankfurt-am-Main. Google Scholar CrossRef Search ADS   Bertola G., Drazen A. ( 1993) Trigger points and budget cuts: explaining the effects of fiscal austerity, American Economic Review , 83, 11– 26. Blanchard O., Leigh D. ( 2013) Growth forecast errors and fiscal multipliers, IMF Working Paper, No. 13/1, Washington, DC. Broto C., Molina L. ( 2014) Sovereign ratings and their asymmetric response to fundamentals, Documentos de Trabajo Banco de España, No. 1428, Madrid. Burriel P., de Castro F., Garrote D., Gordo E., Paredes J., Pérez J. ( 2009) Fiscal policy shocks in the euro area and the US: an empirical assessment, Working Papers Series No. 1133, European Central Bank, Frankfurt-am-Main. Christiano L., Eichenbaum M., Rebelo S. ( 2011) When is the government spending multiplier large?, Journal of Political Economy , 119, 78– 121. Google Scholar CrossRef Search ADS   DeLong J.B., Summers L.H. ( 2012) Fiscal policy in a depressed economy, Brookings Papers on Economic Activity, Spring 2012, 233–297, Washington, DC. Devries P., Guajardo J., Leigh D., Pescatori A. ( 2011) An action-based analysis of fiscal consolidation in OECD countries, IMF Working Papers No. 11/128, Washington, DC. Galí J., López-Salido J.D., Vallés J. ( 2007) Understanding the effects of government spending on consumption, Journal of the European Economic Association , 5, 227– 70. Google Scholar CrossRef Search ADS   Giavazzi F., Pagano M. ( 1990) Can severe fiscal consolidations be expansionary? Tales of two small European countries, in Blanchard O.J., Fischer S. (eds) NBER Macroeconomics Annual , vol. 5, National Bureau of Economic Research, Cambridge, MA, 75– 122. Guajardo J., Leigh D., Pescatori A. ( 2014) Expansionary austerity? International evidence, Journal of European Economic Association , 12, 949– 68. Google Scholar CrossRef Search ADS   Hernández de Cos P., López Rodríguez D. ( 2014) Estructura impositiva y capacidad recaudatoria en España: un análisis comparado con la UE, Documentos Ocasionales No. 1406, Banco de España, Madrid. Hernández de Cos P., Moral-Benito E. ( 2013) Fiscal consolidations and economic growth, Fiscal Studies , 34, 491– 515. Google Scholar CrossRef Search ADS   IMF ( 2010) Fiscal Monitor , May, Washington, DC. IMF ( 2013) Staff report for the 2013 article IV consultation for Spain, IMF Country Report No. 13/244, Washington, DC. IMF ( 2014) Fiscal Monitor , April, Washington, DC. Jordá O., Taylor A. ( 2013) The time of austerity: estimating the average treatment effect of fiscal policy, Working Paper No. 2013‐25, Federal Reserve Bank of San Francisco, San Francisco, CA. Powell D. ( 2015) Quantile regression with nonadditive fixed effects, RAND Working Paper, Santa Monica, CA. Ramey V.A. ( 2011) Can government purchases stimulate the economy?, Journal of Economic Literature , 49, 673– 85. Google Scholar CrossRef Search ADS   Ramey V.A., Shapiro M.D. ( 1998) Costly capital reallocation and the effects of government spending, Carnegie-Rochester Conference Series on Public Policy , 48, 145– 94. Google Scholar CrossRef Search ADS   Romer C., Romer D. ( 2010) The macroeconomic effects of tax changes: estimates based on a new measure of fiscal shocks, American Economic Review , 100, 763– 801. Google Scholar CrossRef Search ADS   Sims C. ( 1980) Macroeconomics and reality, Econometrica , 48, 1– 48. Google Scholar CrossRef Search ADS   Smets F., Wouters R. ( 2007) Shocks and frictions in US business cycles: a Bayesian DSGE approach, American Economic Review , 97, 586– 606. Google Scholar CrossRef Search ADS   Appendix 1 Table A1. Narrative consolidations Country  Year  Consolidation  Revenue  Expenditure  Country  Year  Consolidation  Revenue  Expenditure  Australia  2012  0.3  0.2  0.1  Latvia  2009  9.75  3.8  5.95  Australia  2013  0.4  0.2  0.2  Latvia  2010  4.7  2.2  2.5  Austria  2011  0.7  0.4  0.3  Latvia  2011  2.3  1.6  0.7  Austria  2012  0.5  0.3  0.2  Latvia  2012  0.7  0.3  0.4  Austria  2013  0.7  0.2  0.5  Lithuania  2009  7.4  1.6  5.8  Austria  2014  0.4  –0.1  0.5  Lithuania  2010  4.5  0.5  4  Belgium  2010  0.4  0.1  0.3  Lithuania  2011  1.9  0.1  1.8  Belgium  2011  0.5  0.2  0.3  Lithuania  2012  1.3  0  1.3  Belgium  2012  1.9  1.1  0.8  Mexico  2010  0.7  0.7  0  Belgium  2013  0.9  0.5  0.4  Mexico  2011  0.8  0.8  0  Canada  2011  0.1  0.05  0.05  Mexico  2012  0.2  0.2  0  Canada  2012  0.1  0  0.1  Netherlands  2011  1.8  0.2  1.6  Canada  2013  0.3  0.05  0.25  Netherlands  2012  0.4  0.2  0.2  Canada  2014  0.5  0.1  0.4  Netherlands  2013  2.1  1.1  1  Czech Republic  2010  2.6  1.7  0.9  Netherlands  2014  1  0.5  0.5  Czech Republic  2011  1.6  0.8  0.8  New Zealand  2011  0.4  0  0.4  Czech Republic  2012  1.4  0.8  0.6  New Zealand  2012  0.9  0  0.9  Czech Republic  2013  1.1  0.8  0.3  New Zealand  2013  0.9  0  0.9  Denmark  2011  1.3  0.4  0.9  New Zealand  2014  0.9  0  0.9  Denmark  2012  0.5  0.3  0.2  Poland  2010  0.6  0  0.6  Denmark  2013  1.1  0.4  0.7  Poland  2011  2.4  1.3  1.1  Estonia  2009  9.2  3  6.2  Poland  2012  0.5  0.3  0.2  Finland  2010  0.2  0.1  0.1  Poland  2013  0.2  0.1  0.1  Finland  2011  0.6  0.7  –0.1  Poland  2014  0.1  –0.3  0.4  Finland  2012  0.3  0.3  0  Portugal  2010  2.2  1.7  0.5  Finland  2013  1.3  0.7  0.6  Portugal  2011  3.4  1.6  1.8  France  2011  0.9  0.4  0.5  Portugal  2012  6  2.2  3.8  France  2012  1.4  0.8  0.6  Portugal  2013  3.5  2.8  0.7  France  2013  2  1.4  0.6  Portugal  2014  1.9  0.5  1.4  France  2014  0.7  0.3  0.4  Slovakia  2011  1.9  1.1  0.8  Germany  2011  0.6  0.1  0.5  Slovakia  2012  1  0.3  0.7  Germany  2012  0.6  0.2  0.4  Slovakia  2013  3.9  2.6  1.3  Germany  2013  0.4  0  0.4  Slovenia  2010  2.6  0  2.6  Greece  2010  7.8  4.1  3.7  Slovenia  2011  0.7  0.1  0.6  Greece  2011  2.6  1  1.6  Slovenia  2012  2.9  0.5  2.4  Greece  2012  3.5  2  1.5  Slovenia  2013  2  1  1  Greece  2013  1.6  0.7  0.9  Spain  2010  0.9  0.7  0.2  Hungary  2010  4.1  0.6  3.5  Spain  2011  2.1  0.5  1.6  Hungary  2011  0.8  0  0.8  Spain  2012  4  1.6  2.4  Hungary  2012  3.3  2.1  1.2  Spain  2013  3.5  2  1.5  Hungary  2013  1  0.3  0.7  Spain  2014  1.2  0.7  0.5  Ireland  2009  5.8  3.6  2.2  Turkey  2010  1  0.8  0.2  Ireland  2010  1  0.2  0.8  Turkey  2011  1  0.8  0.2  Ireland  2011  3.26  0.86  2.4  Turkey  2012  1  0.8  0.2  Ireland  2012  2  0.8  1.2  United Kingdom  2010  0.3  0.3  0  Ireland  2013  2  0.8  1.2  United Kingdom  2011  1.7  1.4  0.3  Ireland  2014  1.3  0.6  0.7  United Kingdom  2012  1.1  0.8  0.3  Italy  2011  1  0.4  0.6  United Kingdom  2013  0.6  0.2  0.4  Italy  2012  2.8  2.3  0.5  United States  2012  0.2  0.1  0.1  Italy  2013  0.8  0.2  0.6  United States  2013  0.5  0.4  0.1  Italy  2014  1  0  1            Country  Year  Consolidation  Revenue  Expenditure  Country  Year  Consolidation  Revenue  Expenditure  Australia  2012  0.3  0.2  0.1  Latvia  2009  9.75  3.8  5.95  Australia  2013  0.4  0.2  0.2  Latvia  2010  4.7  2.2  2.5  Austria  2011  0.7  0.4  0.3  Latvia  2011  2.3  1.6  0.7  Austria  2012  0.5  0.3  0.2  Latvia  2012  0.7  0.3  0.4  Austria  2013  0.7  0.2  0.5  Lithuania  2009  7.4  1.6  5.8  Austria  2014  0.4  –0.1  0.5  Lithuania  2010  4.5  0.5  4  Belgium  2010  0.4  0.1  0.3  Lithuania  2011  1.9  0.1  1.8  Belgium  2011  0.5  0.2  0.3  Lithuania  2012  1.3  0  1.3  Belgium  2012  1.9  1.1  0.8  Mexico  2010  0.7  0.7  0  Belgium  2013  0.9  0.5  0.4  Mexico  2011  0.8  0.8  0  Canada  2011  0.1  0.05  0.05  Mexico  2012  0.2  0.2  0  Canada  2012  0.1  0  0.1  Netherlands  2011  1.8  0.2  1.6  Canada  2013  0.3  0.05  0.25  Netherlands  2012  0.4  0.2  0.2  Canada  2014  0.5  0.1  0.4  Netherlands  2013  2.1  1.1  1  Czech Republic  2010  2.6  1.7  0.9  Netherlands  2014  1  0.5  0.5  Czech Republic  2011  1.6  0.8  0.8  New Zealand  2011  0.4  0  0.4  Czech Republic  2012  1.4  0.8  0.6  New Zealand  2012  0.9  0  0.9  Czech Republic  2013  1.1  0.8  0.3  New Zealand  2013  0.9  0  0.9  Denmark  2011  1.3  0.4  0.9  New Zealand  2014  0.9  0  0.9  Denmark  2012  0.5  0.3  0.2  Poland  2010  0.6  0  0.6  Denmark  2013  1.1  0.4  0.7  Poland  2011  2.4  1.3  1.1  Estonia  2009  9.2  3  6.2  Poland  2012  0.5  0.3  0.2  Finland  2010  0.2  0.1  0.1  Poland  2013  0.2  0.1  0.1  Finland  2011  0.6  0.7  –0.1  Poland  2014  0.1  –0.3  0.4  Finland  2012  0.3  0.3  0  Portugal  2010  2.2  1.7  0.5  Finland  2013  1.3  0.7  0.6  Portugal  2011  3.4  1.6  1.8  France  2011  0.9  0.4  0.5  Portugal  2012  6  2.2  3.8  France  2012  1.4  0.8  0.6  Portugal  2013  3.5  2.8  0.7  France  2013  2  1.4  0.6  Portugal  2014  1.9  0.5  1.4  France  2014  0.7  0.3  0.4  Slovakia  2011  1.9  1.1  0.8  Germany  2011  0.6  0.1  0.5  Slovakia  2012  1  0.3  0.7  Germany  2012  0.6  0.2  0.4  Slovakia  2013  3.9  2.6  1.3  Germany  2013  0.4  0  0.4  Slovenia  2010  2.6  0  2.6  Greece  2010  7.8  4.1  3.7  Slovenia  2011  0.7  0.1  0.6  Greece  2011  2.6  1  1.6  Slovenia  2012  2.9  0.5  2.4  Greece  2012  3.5  2  1.5  Slovenia  2013  2  1  1  Greece  2013  1.6  0.7  0.9  Spain  2010  0.9  0.7  0.2  Hungary  2010  4.1  0.6  3.5  Spain  2011  2.1  0.5  1.6  Hungary  2011  0.8  0  0.8  Spain  2012  4  1.6  2.4  Hungary  2012  3.3  2.1  1.2  Spain  2013  3.5  2  1.5  Hungary  2013  1  0.3  0.7  Spain  2014  1.2  0.7  0.5  Ireland  2009  5.8  3.6  2.2  Turkey  2010  1  0.8  0.2  Ireland  2010  1  0.2  0.8  Turkey  2011  1  0.8  0.2  Ireland  2011  3.26  0.86  2.4  Turkey  2012  1  0.8  0.2  Ireland  2012  2  0.8  1.2  United Kingdom  2010  0.3  0.3  0  Ireland  2013  2  0.8  1.2  United Kingdom  2011  1.7  1.4  0.3  Ireland  2014  1.3  0.6  0.7  United Kingdom  2012  1.1  0.8  0.3  Italy  2011  1  0.4  0.6  United Kingdom  2013  0.6  0.2  0.4  Italy  2012  2.8  2.3  0.5  United States  2012  0.2  0.1  0.1  Italy  2013  0.8  0.2  0.6  United States  2013  0.5  0.4  0.1  Italy  2014  1  0  1            © Oxford University Press 2017 All rights reserved This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Oxford Economic Papers Oxford University Press

Fiscal consolidation after the Great Recession: the role of composition

Loading next page...
 
/lp/ou_press/fiscal-consolidation-after-the-great-recession-the-role-of-composition-bonYWq9IoG
Publisher
Oxford University Press
Copyright
© Oxford University Press 2017 All rights reserved
ISSN
0030-7653
eISSN
1464-3812
D.O.I.
10.1093/oep/gpx032
Publisher site
See Article on Publisher Site

Abstract

Abstract We have examined the fiscal consolidation episodes in a group of OECD countries from 2009 to 2014. The range of the estimated short-term fiscal multiplier runs from 1.2 to 2.0, larger than those obtained in more ‘normal times’, implying that the contractionary effect has been larger in depressed environments. Nevertheless, we also found that revenue measures have a higher and more persistent real impact than expenditure measures, which is more consistent with the influence of current consolidations on the expectations about the future path of fiscal policies (the expectations channel). This result suggests that expenditure cuts are less harmful for the economy than tax hikes. 1. Introduction In the wake of the global financial crisis, the fall in revenues, the fiscal stimulus, and the realization of contingent liabilities, mostly related to the support provided to the financial system, triggered a considerable increase in the public deficit in advanced economies, reaching 9% of GDP and a debt ratio over 90% in 2009 (IMF, 2014). The long-term costs of high public debt are well known from the literature, explaining why so many countries subsequently committed to fiscal adjustment programs to ensure the sustainability of their public finances. By 2015 the deficits have been reduced significantly, but public debt to GDP ratios in the advanced economies have still not stabilized and stand at historical levels (IMF, 2014). Nevertheless, the fiscal policy debate has focused on the short-term effects of fiscal consolidation and, consequently, on the appropriate speed of debt reduction and the composition between revenue and expenditure. This paper analyses the effects of fiscal consolidation for a group of OECD countries in the period 2009–2014 and compares it with previous consolidation periods. There are several possible reasons why the current episode of large fiscal retrenchment is so very different from previous ones. Six years after the onset of the crisis, output gaps and cyclical unemployment still loom large in many advanced economies. Monetary policy is very expansionary, but given the zero lower bound of interest rates, financial conditions remain tight. Moreover, a synchronized fiscal adjustment across several major economies may adversely impact the recovery. In fact, Blanchard and Leigh (2013) have already shown that a negative relationship between fiscal consolidation forecasts and subsequent growth forecast errors has been responsible for a slower than expected recovery in a number of advanced economies during the period 2010–2011. Empirically, the problem is to correctly identify the effects of fiscal policy on output. When analysing cross-country evidence, a standard method has been to use the cyclically adjusted primary balance (CAPB) to approximate discretionary changes in fiscal policy. From another standpoint, traditional VAR methods assume that fiscal changes are uncorrelated with other determinants of output. However, as pointed out by Romer and Romer (2010), these fiscal variables may include non-policy changes correlated with output. That is particularly relevant when using annual data, since agents may be responding to the fiscal change within the year and that correlation would bias the analysis against the contractionary effects of fiscal consolidation. To address these identification problems, the literature has used historical records—the narrative approach—to look for fiscal policy actions that are aimed at reducing the budget deficit rather than in response to current economic conditions. To that end we have constructed a dataset of government spending cuts and revenue increases from a broad set of policy documents for a group of advanced economies. Alesina et al. (2015) analysed historical records up to 2007 to simulate multi-year fiscal plans in 2009–2013. In our case we have looked to the cost of fiscal consolidation for a wider set of advanced economies using a new dataset from the period 2009–2014. The collected measures are ex-post outcomes and we do not try to separate the expected from the unexpected component of the fiscal change. Against this background, we are able to compare our data with the fiscal consolidation records in more ‘normal times’ (1979–2009) previously studied in the literature (Guajardo et al., 2014). Moreover, following a line of the literature, we study how the current fiscal policies affect the expected future path of fiscal policies (the expectation channel) in the most recent period when considering revenue and expenditure consolidations in a separated way. The next section of the paper first presents a brief discussion of the theoretical predictions of fiscal consolidations, and second describes some characteristics of the fiscal consolidation taking place after 2009, showing that the efforts have been higher than in previous episodes. We also present a test of exogeneity of fiscal actions compared with standard CAPB measures. Section 3 reports the main econometric results. We estimate the dynamic fiscal multiplier on output in a single-equation specification and compare it with an alternative VAR specification. We also report the differences between revenue and expenditure, considering only large consolidations. The evidence points to higher fiscal multiplier effects after the financial crisis and to higher and more persistent impact of revenue-based rather than expenditure-based actions. The latter evidence goes in the same direction as the one reported by Alesina et al. (2015). Moreover, we found that revenue multipliers are larger while expenditure multipliers become non-significant when only large consolidations are considered. Section 4 presents different robustness exercises. First, we add to the VAR analysis some financial variables that the literature has found relevant in explaining the idiosyncratic country characteristics of the fiscal multipliers. The results from the previous section then become clearer. Second, given the interdependence of monetary and fiscal policy, we analyse whether it has any significant influence on the estimated fiscal multipliers. Changes in policy interest rates are shaping the magnitude of the fiscal actions, especially when the most recent period, affected by the zero lower bound, is excluded. Third, we study if there are any specific fiscal effects for the euro area countries, since they have been subject to a common monetary policy and fiscal framework. The confidence channel, in the presence of a financial stress environment, may explain the lower real effects of fiscal actions in the euro area. To conclude, Section 5 presents a summary of the main results together with some open issues for future research. 2. Fiscal consolidations: theory and the most recent data 2.1. Theoretical predictions There is a general view that, in normal times, fiscal consolidations and the resulting government debt reduction contributes to long-term growth. However, there is no consensus regarding the short-term effects of fiscal austerity. The prediction of traditional Keynesian models is that cutting government spending or raising taxes reduces the economic activity in the short term. In the IS-LM model, consumers behave in a non-Ricardian fashion and their consumption is a fraction of their current disposable income. The size of the fiscal multiplier depends mainly on the marginal propensity to consume. But other factors may also be relevant, like how accommodative is the monetary policy, the investment response to the fiscal shock, or the fiscal sustainability of the country. The contractionary effect of fiscal adjustments would be consistent with the empirical evidence analysing aggregate time-series models of a short-term spending multiplier that lies between 0.6 and 1.8 (Ramey, 2011). In general, New Keynesian models with sticky-prices and neoclassical foundations that feature infinitely lived Ricardian households tend to reduce the size of the multiplier (Smets and Wouters, 2007), whereas the introduction of rule-of-thumb consumers increases the multiplier (Galí et al., 2007). The efficacy of fiscal policy may also be affected by the special circumstances we have seen in many advanced economies after the global financial crisis. This is the case of the zero lower bound of short-term nominal interest rates. Christiano et al. (2011) show that in a deflationary setup with monetary authorities committed to keep interest rates constant for a considerable period of time, an expansionary fiscal policy leads inflation expectations to rise and the multiplier is likely to be larger1. Another line of the literature has highlighted the importance of composition between revenue and expenditure when studying consolidation efforts and the existence of non-Keynesian effects. Giavazzi and Pagano (1990) were the first to show that large expenditure-based fiscal adjustments could be expansionary. More recently, Alesina and Ardagna (2010) found that spending cuts were much more effective than tax increases on large fiscal consolidations, and that they were associated with economic expansion across a panel of OECD countries. Significant fiscal consolidation today (particularly if implemented with large expenditure cuts) may reduce the need for future fiscal action, raising the current confidence of households and firms. For example, Bertola and Drazen (1993) outlined a model where the expectations about future fiscal policies affected the effect of current fiscal policies. Thus, when this expectations channel dominates the pure multiplier effect, consolidation will have a positive effect on output. Given the scale of the fiscal efforts across countries in the wake of the global crisis and the sizeable revenue increases, it is an open issue if there is any evidence of ‘expansionary fiscal contractions’ when considering the most recent time period. 2.2. Fiscal consolidations after the financial crisis: 2009–2014 After the global financial crisis many economies have been involved in a process of fiscal consolidation.2 The main objective was to reduce the fiscal deficit that increased significantly as a consequence of the 2008 financial crisis. For example, the average discretionary fiscal stimulus in the G-20 economies was around 2% of GDP in 2009 and 2010 and their average fiscal deficit reached 6% in 2010 (see IMF, 2010). In consequence, at the Toronto Summit in 2009 they committed to fiscal plans that would halve deficits by 2013. We follow the historical approach proposed by Romer and Romer (2010) to identify fiscal consolidations after the financial crisis.3 The historical episode narrative aims to separate policy changes from those arising from non-policy developments. The set of documents reviewed for this purpose relies on Stability and Convergence Programmes submitted annually to the European Commission, national budgets, OECD Economic Surveys, and IMF Staff Reports, as well as national sources such as the Congressional Budget Office (CBO) and several Memorandums of Understanding (MoUs). Where possible we have reviewed the measures announced, using the most up-to-date document available, and have checked those documents using the retrospective analysis included in the IMF’s Article IV reviews and in the OECD’s ‘Restoring Public Finances’ documents. All the budgetary measures are fully credible and implemented in the year assigned by the official documents. The sample includes the revenue-based and expenditure-based fiscal actions taken by 27 economies in the period 2009–2014 (see Appendix 1, Table A1). These are 25 OECD countries plus Latvia and Lithuania, which are now euro area countries. Thus, the majority are advanced countries, but there are also two emerging economies (Mexico and Turkey). In total we have identified 101 cases of action where these countries took some budgetary measures. During the period 2010–2014, more than 70% of the economies included in the sample were immersed in fiscal consolidation. The median of the adjustments is 1% of GDP and the average is 1.76% of GDP, with a standard deviation of 0.18 pp. The range runs from –3.0% to 9.75% of GDP (see Fig. 1). The negative figures correspond to the expiration of temporary fiscal measures, for example Estonia in 2010 and 2011, and the figures over 9% of GDP correspond to Estonia and Latvia in 2009. Fig. 1. View largeDownload slide Narrative episodes of fiscal consolidation 2009–2014: size distribution Fig. 1. View largeDownload slide Narrative episodes of fiscal consolidation 2009–2014: size distribution The current fiscal adjustment episodes are very different from the previous ones studied by Guajardo et al. (2014). During the period 1978–2009, the fiscal adjustments of 15 advanced economies averaged 0.99% with a standard deviation of 0.94. Thus, although the number of observations and of years in our sample is smaller, it contains more countries, the average size of fiscal consolidation is larger, and it has lower variance. If we define large consolidations as consolidation efforts amounting to more than 1.5% of GDP, as in Alesina and Ardagna (2010), we find a total of 67 large fiscal consolidations. Most observations in the sample are concentrated between 2011 and 2013 (see Fig. 2). The 2009 data are driven by the Baltic economies that suffered a sudden stop and a credit crunch in 2008. In 2009 they started to consolidate very significantly (an average of 8%), in a year dominated by spending cuts (see Fig. 3). In 2010 there was an increase in the number of countries tightening their budgets, including the euro area economies subject to market pressure. Finally, in 2014 the pace of fiscal consolidation abated, with fewer countries making consolidation efforts and expenditure cuts exceeding tax increases. Fig. 2. View largeDownload slide Narrative episodes of fiscal consolidation, 1978–2014 Fig. 2. View largeDownload slide Narrative episodes of fiscal consolidation, 1978–2014 Fig. 3. View largeDownload slide Average size of narrative fiscal consolidations Fig. 3. View largeDownload slide Average size of narrative fiscal consolidations The composition of fiscal adjustments is critical for their effects on the economy. Figs 2 and 3 show that the adjustments made in the 2009–2014 period were fairly equally split between tax-based and expenditure-based actions. However, our sample contains several examples of important consolidations affecting both expenditure and revenue, so it will be crucial to estimate the impact of both types of consolidations jointly. This observation contrasts with the fiscal adjustment episodes examined by Alesina et al. (2015). They find that most consolidation efforts made in the 2009–2013 period have been based on expenditure cuts. We have found that the calculated total consolidation is fairly similar for six out of the eleven countries their dataset contains. The starkest difference occurs in Spain, where we have detected a total consolidation of 10.5% of GDP, fairly balanced between expenditure cuts and tax hikes, while Alesina and co-authors computed a total consolidation of more than 15% of GDP, almost entirely based on adjustments in expenditure. Our balanced consolidation is consistent with some country-specific studies about fiscal consolidation in Spain, with some authors putting more weight on the revenue side (Hernández de Cos and López Rodríguez, 2014). Moreover, with respect to 2012–2013, several official documents identify a balanced path of consolidation, with total consolidation amounting to around 6.5%–7.5% of GDP (see e.g. IMF, 2013), in contrast with more than 9% of anticipated and unanticipated measures in Alesina et al. (2015). Finally, we have also observed significant differences from one region to another. Overall, advanced non-stressed euro area and emerging economies are the regions that have consolidated less, whereas Baltic and stressed euro area countries are the ones that have made most efforts. In relative terms, taxes have been raised more in the emerging countries and expenditure has been cut more in the Baltic and the advanced non-European countries. 2.3. Comparison of fiscal policy measures We have said that the conventional approach to measure policy-driven changes in fiscal policy is through the cyclically adjusted primary balance (CAPB) that excludes changes in fiscal variables induced by business cycle fluctuations. To have a uniform indicator, we selected the variable constructed by the IMF’s Fiscal Monitor database. In total there are 101 cases where these 27 countries took budgetary actions as measured by CAPB. The median of the adjustments is 0.77% of GDP and the average is 1.2%, with a standard deviation of 0.12 pp. Thus, in general, the CAPB adjustments are smaller than those identified in the narrative approach. Before analysing the real effects of the fiscal narrative measures, we compare them with the more standard CAPB measures. Figure 4 plots the observations of the two fiscal consolidation indicators in the period 2009–2014. There are no major discrepancies observed between the two, and in principle it would be impossible to say which measure provides a more reliable identification. However, we have found a total of eight episodes with large differences, defined as more than 3 pp of difference between the two approaches. And we have identified five large one-off accounting measures that could explain the inaccuracies in the CAPB (Ireland 2009, Lithuania 2010, Hungary 2011, Spain 2012, and Slovakia 2013). Fig. 4. View largeDownload slide Narrative episodes vs. CAPB: 2009–2014 Fig. 4. View largeDownload slide Narrative episodes vs. CAPB: 2009–2014 The econometric findings for the period before 2009 are that the narrative episodes are more exogenous to output than the CAPB.4 But if the financial crisis period is included, the CAPB measure may possibly be more accurate. In order to check for contemporaneous orthogonality with output, we perform a test of weak exogeneity of both the CAPB and the narrative measures. Following Guajardo et al. (2014), we construct a measure of economic surprises, based on the IMF’s GDP forecast revisions. We define the economic news index ( Newsit) as the log-difference between GDP at time t in the October World Economic Outlook (WEO) of year t and GDP at time t of the WEO at time t-1 for each country. We perform the following regression:   ΔFit= μi+λt+β Newsit+uit (1) where ΔFit is the fiscal consolidation measure (CAPB or narrative-based), μi is an unobservable country-fixed effect, and λt is a common year effect for all economies. In order to avoid potential small-sample bias, we merge our sample with the 1978–2009 period, giving 373 observations for 15 economies that had consolidation episodes in both periods.5,6 This exercise does not test for orthogonality with past (and future) output developments. In fact, the current fiscal actions are likely to be correlated with past large fiscal deficits and will be used as instruments to predict output movements. Table 1 presents the results. The β coefficient of the news index on the narrative measure equation is negative and only weakly significant, while the coefficient on the CAPB equation is positive and strongly significant. This result is consistent with that obtained for the pre-crisis period. Moreover, the explanatory power of the narrative-based measures (0.26) is lower than that of the CAPB measures (0.48), supporting the theory of the greater exogeneity of the narrative-based approach to contemporaneous changes to output movements. In addition, when we perform the same test for narrative-based revenue and expenditure, we find that the coefficient related to the tax changes is near zero, it is not significant, and it is less correlated with output than with government spending changes. The greater exogeneity of tax changes signals that spending actions may respond to tax actions or other variables correlated with output, and therefore that revenue actions will be more help in the identification strategy. Another avenue, as in Ramey and Shapiro (1998), would be to rely on specific government spending items that do not respond to economic events. Table 1. An exogeneity test of fiscal policy changes (1978–2014) Equation: ΔFit=μi+λt+βNewsit+uit Dependent variable: ΔFit  Narrative fiscal consolidations  CAPB  Narrative revenue consolidations  Narrative expenditure consolidations  β coefficient  –0.120  0.254  –0.037  –0.082    [0.067]*  [0.081]***  [0.032]  [0.039]*  Obervations  373  371  373  373  No. of countries  5  15  15  15  R2  0.256  0.482  0.205  0.228  Dependent variable: ΔFit  Narrative fiscal consolidations  CAPB  Narrative revenue consolidations  Narrative expenditure consolidations  β coefficient  –0.120  0.254  –0.037  –0.082    [0.067]*  [0.081]***  [0.032]  [0.039]*  Obervations  373  371  373  373  No. of countries  5  15  15  15  R2  0.256  0.482  0.205  0.228  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. In order to test for the exogeneity of our more recent sample, we exploit the fact that narrative fiscal measures after the Great Recession were in general of a bigger magnitude. Therefore, we estimated a quantile regression of eq. (1). The results are shown in Table 2. Innovations to GDP growth are not significant at different points of the distribution. Moreover, they are not significant at the higher percentiles, which are mostly stacked in the more recent period. Table 2. Exogeneity test by quantile Dependent variable: ΔFit  Narrative fiscal consolidations  β coefficient    20% percentile  0.018  40% percentile  –0.144  60% percentile  –0.095  80% percentile  –0.346  90% percentile  –0.264  Observations  373  No. of countries  15  Dependent variable: ΔFit  Narrative fiscal consolidations  β coefficient    20% percentile  0.018  40% percentile  –0.144  60% percentile  –0.095  80% percentile  –0.346  90% percentile  –0.264  Observations  373  No. of countries  15  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. *, **, *** denote significance at the 10%, 5%, and 1% level, respectively. The coefficients are estimated following Powell (2015). 3. The estimated effects of fiscal consolidation on economic activity The estimation strategy first examines the output effect on a single-equation specification and later on a vector autoregressive (VAR) model. In order to make the fiscal multipliers comparable across specifications, we normalize the fiscal change so that the CAPB to GDP ratio rises by 1% of GDP on impact. In the VAR model, we will report the short-term fiscal multiplier after one year and the medium-term fiscal multiplier after four years, dividing the cumulated output response by the CAPB response. The first section presents the results for the aggregated fiscal measures and analyses the real effects before and after the financial crisis. The second section presents the composition results, splitting the fiscal actions between revenues and expenditures, and considers also large fiscal consolidations as defended by the non-Keynesian view. 3.1. Fiscal multipliers Following the empirical literature in this area, we estimate the effect of fiscal consolidation on a single-equation specification that regresses real output growth ( ΔYit) on lagged output growth and the contemporaneous values and lags of the fiscal changes ( ΔFit−s). It takes the following form:   ΔYit= μi+λt+∑s=12γsΔYit−s+∑s=02δsΔFit−s+ uit (2) Including past output allows us to control for the state of the economy in each country and thus helps estimate the discretionary part of fiscal action. As is standard in the panel data approach, the estimation includes country-fixed effects and year-fixed effects, and based on the information criteria we choose two lags in the dynamic specification.7 We performed panel unit root tests for GDP, CAPB, and the narrative fiscal measures, showing that the variables are all (weakly) stationary. Denoting ΔFit the CAPB-based fiscal measures, the estimation by ordinary least squares (OLS) of the regression coefficients would be biased by several factors. First, the presence of an unobservable country-fixed effect would be correlated with the error term. In order to deal with country heterogeneity, we will estimate our set of parameters using a fixed-effect estimator, although this estimator will be biased under the presence of the lagged output on the right-hand side for a small T specification. Second, if there is endogeneity in the CAPB measures, it will be correlated with the error term, and therefore an instrumental variables approach should be used. In particular, we use a two-stage least squares (TSLS) estimator with our exogenous narrative measures as an instrument for the CAPB. Third, when the number of observations per country is small (a problem present mainly in the 2009–2014 sample), a Nickell bias related to the correlation between the lagged dependent variable and the error term is more likely to be found. Against this background, this bias will overestimate the negative sign of the fiscal multiplier. To overcome this problem, we use a Generalized Method of Moments (GMM) estimator. However, the small number of countries in our sample precludes us from using the full amount of moment restrictions available, i.e. all the lags of the dependent variable (and the CAPB), as instruments, and in consequence we use the Anderson-Hsiao estimator. Our fourth specification is a natural variation of the first one. We run a three-vector auto-regression model (VAR) with the change in GDP, the change in the CAPB ratio, and the narrative fiscal measures.8 Thus, lagged output and past cyclically adjusted primary balance also affect the current cyclically adjusted primary balance. This specification also includes two lags and a full set of country and time effects. Consistent with the tested exogeneity of the historical episodes of the fiscal consolidations and the TSLS estimation in the single-equation approach, the narrative measures are ordered first in the VAR and second in the CAPB, allowing it to have a contemporaneous effect on output (the last variable in the system equation) when considering a Cholesky orthogonalization of the residuals. The estimation results under the three single-equation specifications are shown in Table 3. The first column presents the fiscal multipliers in the OLS case with fixed effects, the second column the results with the TSLS estimator, and the third column the results with the GMM estimator. We show the effect of a 1% change of fiscal consolidation in GDP, at the time of the consolidation, and calculate the dynamic response function one year later. Table 3. Estimated dynamic output effect of a 1% CAPB change: single-equation specification Equation: ΔYit=μi+λt+∑s=12γsΔYit−s+∑s=02δsΔFit−s+ uit     OLS  TSLS  GMM  Observations  2009–2014 (25 countries)  (1)  –0.691  –1.951  –0.961  150  [0.221]***  [0.521]***  [0.23]***  1978–2009 (15 countries)  (2)  0.246  –0.871  –  445  [0.121]**  [0.335]***  1978–2014 (15 countries)  (3)  0.138  –1.180  –  520  [0.115]  [0.33]***  (4)  –  –1.560  –  520  [0.467]***      OLS  TSLS  GMM  Observations  2009–2014 (25 countries)  (1)  –0.691  –1.951  –0.961  150  [0.221]***  [0.521]***  [0.23]***  1978–2009 (15 countries)  (2)  0.246  –0.871  –  445  [0.121]**  [0.335]***  1978–2014 (15 countries)  (3)  0.138  –1.180  –  520  [0.115]  [0.33]***  (4)  –  –1.560  –  520  [0.467]***  Notes: Country- and time-fixed effects included. Robust standard errors in brackets. Accumulated output effect after one year. The instruments in the two-stage least squares (TSLS) estimation are the narrative fiscal measures. In panel (4) the instruments are the narrative fiscal revenue measures. In the GMM specification, narrative measures are included as external instruments and the second lags of GDP and CAPB are used as GMM instruments (18 instruments in total). Panel 1 shows the results for the period 2009–2014, after the Great Recession, with the full set of 25 countries. The OLS estimator shows an effect of –0.69 pp one year after the consolidation. The effect is statistically significant, and after that period the responses on output stabilize. In column 2, CAPB is instrumented with the narrative measures.9 The effect after one year is a highly significant 1.95 pp, almost three times more than with the OLS. As expected, the GMM estimator lowers the calculated fiscal multiplier to 0.96.10 In Panel 2 we report the results with a smaller sample of countries (15 instead of 25) for the period 1978–2009, given our interest in comparing the fiscal consolidation effects after the financial crisis with the international episodes before 2009. The estimates reproduce the stark differences between OLS and TSLS found by Guajardo et al. (2014):11 the positive OLS estimate (0.24) with the CAPB measure is consistent with the ‘expansionary austerity’ found previously in the literature; the instrumental estimate with the narrative shocks is negative (–0.87) and significant. Note that this discrepancy between estimation methods was not present when studying only the crisis period in Panel 1. In Panel 3 the analysis is extended to include the crisis period. The qualitative results are maintained, but the short-term effects on real activity are larger (–1.18 with TSLS). This is primary evidence that the contractionary effect of consolidations was larger in depressed environments and is consistent with recent findings that have stressed the relevance of the state of the economy when measuring the size of fiscal multipliers (e.g. Auerbach and Gorodnichenko, 2012). Figure 5 presents the same TSLS estimation in Panel 3, but in a rolling window compressing 14 years. The fiscal multiplier shows a remarkable stability until the last three years are included in the estimation, when the multiplier drops by almost 0.2. This is suggestive evidence of a different effect of consolidations in the most recent sample period.12 Fig. 5. View largeDownload slide Responses of output to fiscal shocks, rolling window Notes: All specifications contain a full set of country- and time-fixed effects. TSLS estimation of the effect of CAPB on GDP, using the narrative fiscal measures as instruments. Dashed lines represent two-standard-deviation confidence intervals. The figure shows the accumulated output effect after one year. Fig. 5. View largeDownload slide Responses of output to fiscal shocks, rolling window Notes: All specifications contain a full set of country- and time-fixed effects. TSLS estimation of the effect of CAPB on GDP, using the narrative fiscal measures as instruments. Dashed lines represent two-standard-deviation confidence intervals. The figure shows the accumulated output effect after one year. Finally, in Panel 4, we use the CAPB only with the narrative measures from the revenue side. This is based on the idea that the main motivation for tax changes, usually implemented through new legislation, may be to reduce the inherited fiscal deficit, whereas expenditure changes may respond more to other developments affecting output. The estimated fiscal multiplier in the TSLS estimation (–1.56) supports the possibility of a downward bias towards zero in the above multipliers because the expenditure measures are less exogenous than the revenue ones, which is consistent with the exogeneity test shown in Section 2. Figure 6 depicts the impulse response function from the benchmark VAR specification together with one-standard-deviation confidence bands.13 The size of the shock on the narrative fiscal measures is normalized to generate a response on impact (t = 0) of 1% of GDP on the CAPB. That allows the VAR response to be comparable with the single-equation results. We can see first that fiscal consolidation has a significant negative effect on output. Second, the chart shows that the response is more contractionary when the international sample includes the most recent period after the financial crisis. And finally, it is clear that the real response to the fiscal shock is more persistent with the full sample, since there is a stronger tendency for output to return to normal when the sample ends in 2008. Fig. 6. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Fig. 6. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Table 4 shows the value of the output response after one year and the corresponding standard error in the VAR. Together with the accumulated response of output to a narrative shock, we report the fiscal multipliers both in the short and the long run, calculated as the accumulated response of output divided by the accumulated response of CAPB. The results shown for the two sample periods (columns 1 and 2 in Table 3) coincide with the instrumental variable estimation in the single-equation specification: there is a significant contractionary effect of fiscal consolidation, and that effect is larger when considering the most recent period.14 Table 4. Response of output to a narrative fiscal shock: VAR specification   Benchmark VAR  Benchmark VAR    1978–2009  1978–2014  Response of output to a shock of 1% of CAPB  –1.715  –2.076  [0.420]***  [0.343]***  Fiscal multiplier (after one year)  –1.004  –1.143  Fiscal multiplier (after four years)  –0.731  –0.989  Observations  445  520    Benchmark VAR  Benchmark VAR    1978–2009  1978–2014  Response of output to a shock of 1% of CAPB  –1.715  –2.076  [0.420]***  [0.343]***  Fiscal multiplier (after one year)  –1.004  –1.143  Fiscal multiplier (after four years)  –0.731  –0.989  Observations  445  520  Notes: Robust standard errors are obtained using Monte Carlo simulation. All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation. The accumulated output response is after 1 year. The fiscal multiplier is the ratio between the cumulated effect on output and the cumulated effect on CAPB. In order to test for parameter stability in the more recent period, we perform a test of whether overall parameter values are unchanged after and before 2009, when we think a break date may exist. Against this background, we perform a sample-split test by introducing in the VAR model an interacted dummy variable in the right-hand side of all the regressions, and we compute our structural break statistic as in Sims (1980):   (T−k)log⁡|Σre|−log⁡|Σun| (3) where T is the number of observations, k is the number of regressors, and Σre and Σun are the residual covariance matrices for the restricted and the unrestricted model. Under the null hypothesis of parameter stability, the test statistic is asymptotically chi-squared with the degrees of freedom equal to the total number of constraints. The value of the statistic χ2(18)=42.88 corresponds to a significance level below 0.005, rejecting the null. However, this result could be driven by the important economic changes (and not just fiscal) as a consequence of the crisis. In order to check if the fiscal variables are drivers of the structural change, we have corroborated that the coefficient of the dummy variable interacted with the narrative consolidations is negative and significant in the regression with GDP on the left-hand side. 3.2. The composition effect The standard literature that supports ‘expansionary fiscal contractions’ has emphasized the role of composition (e.g. Alesina and Ardagna, 2010). Their evidence shows that fiscal adjustments based on spending cuts are more effective than tax increases in stabilising the debt ratio and avoiding economic contraction. The approach has been to identify historical cases of fiscal retrenchment, looking at the cyclically adjusted changes in fiscal variables. Here we investigate the relevance of composition by including the narrative fiscal expenditure and revenue measures in the previous VAR framework.15 We consider both sets of measures at the same time given the importance of consolidations involving both expenditure and revenue measures, in order to account for the combined effect. In the 4-variable VAR, revenue and expenditure are first and second in the order of the system, which is consistent with our exogeneity evidence in Section 2. Figure 7 depicts the impulse response functions of revenue and expenditure shocks to output for the 4-variable VAR. We find a significant negative contemporaneous impact of the expenditure shock, returning to the baseline after two years. By contrast the revenue shock has a stronger impact contemporaneously on GDP and a much more persistent effect. Fig. 7. View largeDownload slide The response of output to narrative fiscal shocks: the composition effect Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Fig. 7. View largeDownload slide The response of output to narrative fiscal shocks: the composition effect Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. Dashed lines represent one-standard-deviation confidence intervals. Table 5 confirms the previous results when taking into account the CAPB response and calculating the fiscal multiplier dividing the accumulated output response by the accumulated CAPB ratio response. After one year the expenditure multiplier (–0.67) is significantly smaller than the revenue multiplier (–1.54). Similarly, four years after the shock, the expenditure multiplier (–0.43) is much smaller than the revenue multiplier (–1.50). Table 5. Response of output to a narrative fiscal revenue/expenditure shock (1978–2014): VAR specification   4-variable VAR  4-variable VAR (large consolidations)  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.561  [0.529]***  [0.704]***   Fiscal multiplier (after one year)  –1.545  –2.117   Fiscal multiplier (after four years)  –1.498  –2.315  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –0.839  [0.476]**  [0.51]*   Fiscal multiplier (after one year)  –0.673  –0.478   Fiscal multiplier (after four years)  –0.434  –0.260   Observations  520  520    4-variable VAR  4-variable VAR (large consolidations)  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.561  [0.529]***  [0.704]***   Fiscal multiplier (after one year)  –1.545  –2.117   Fiscal multiplier (after four years)  –1.498  –2.315  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –0.839  [0.476]**  [0.51]*   Fiscal multiplier (after one year)  –0.673  –0.478   Fiscal multiplier (after four years)  –0.434  –0.260   Observations  520  520  Note: See Table 3. As in Guajardo et al. (2014), fiscal consolidations are contractionary even when they are based on spending cuts. But here we also find that for certain specifications expenditure adjustments are significantly different from revenue increases. Thus, these results are also consistent with Alesina et al. (2015), suggesting that expenditure cuts are less harmful for the economy than tax hikes. Similarly, Beetsma et al. (2015) report that consolidation affects consumer confidence negatively but more significantly through the revenue component than the spending component. The non-Keynesian view is that large fiscal adjustments, especially expenditure-based ones, are more effective in avoiding economic downturns, based on the argument that cutting sensitive items such as transfer programmes or government consumption may signal a credible commitment to long-term debt reduction. To test that hypothesis, we re-estimated the 4-variable VAR with the narrative revenue measures greater than 1% of GDP, and 0 otherwise, and similarly for the narrative expenditure measures. We find nine cases where both variables detect a large consolidation; we also find 35 more cases of large expenditure consolidations and 18 cases of only large revenue consolidations. Of the total 71 cases of large consolidations, 23 were detected in the 2009–2014 period. However, given the limited number of observations, the following results should be treated with caution. Column 2 in Table 5 summarizes the results. We find a significant negative response of large revenue-related fiscal consolidations, amounting to an output multiplier that stands at –2.1. The interesting result is in the effect of large expenditure-related consolidations. The effect after one year is negative but not different from zero. Specifically, we find that large expenditure consolidations, after a negative contemporaneous real effect, have a non-significant effect after one year and onwards. Although the data limitation problem becomes more of a concern if we investigate this effect in the 1978–2009 period, it seems that both large expenditure-related and revenue-related consolidations are more contractionary when the recent crisis period is considered. Thus, consistent with the ‘expansionary fiscal contraction’ literature, we find evidence that the composition of fiscal consolidation matters. The large expenditure-based adjustments performed when considering the Great Recession have fiscal multipliers that are not significantly different from zero, whereas large revenue-related consolidations are highly contractionary and very persistent. 4. Robustness exercises 4.1. Controlling for financial factors We first test the robustness of our findings by increasing the benchmark VAR with potentially relevant idiosyncratic country characteristics like the financial position. The fiscal consolidation periods are in many cases related to situations of large public debt and/or financial stress combined with other macroeconomic imbalances that may be perceived as affecting sovereign risk. In order to control for such factors, we have considered two additional variables, namely government debt to GDP in the previous period and an index of the sovereign ratings16 that are included at the end in the ordering of the VAR. We aim to identify fiscal sustainability problems reflected in financial variables, rather than indicators that may be a sign of short-run concerns. In any case, the sovereign rating indicator attempts to control for fundamental problems (low growth prospects, high external debt, weak banking sector, etc.) that could have fiscal sustainability implications. Figure 8 presents the impulse response function from the 5-variable VAR specification. The greater real effect when considering the most recent sample period is more evident when controlling for financial variables. Fig. 8. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. The 5-variable VAR includes lagged debt and sovereign rating index. Dashed lines represent one-standard-deviation confidence intervals. Fig. 8. View largeDownload slide Responses of output to a narrative fiscal shock: 1978–2009 vs. 1978–2014 Notes: All specifications contain a full set of country- and time-fixed effects. The shock on output is an orthogonalized narrative fiscal innovation, normalized to 1% of CAPB. The 5-variable VAR includes lagged debt and sovereign rating index. Dashed lines represent one-standard-deviation confidence intervals. Table 6 quantifies the fiscal multipliers after controlling for financial factors (column 2). The results are robust to the inclusion of these variables since the short-run fiscal multiplier stays around 1.2, a value similar to that obtained in the instrumental variable estimation and the benchmark VAR (column 1). We see a slightly larger multiplier in the short run when controlling for public debt, consistent with the results in Burriel et al. (2009). Table 6. Response of output to a narrative fiscal revenue/expenditure shock (1978–2014): VAR specification with financial variables as controls   Benchmark VAR  Additional variables: debt, sovereign rating index  Narrative consolidations       Response of output to a shock of 1% of CAPB  –2.076  –2.159  [0.343]***  [0.564]***   Fiscal multiplier (after one year)  –1.143  –1.220   Fiscal multiplier (after four years)  –0.989  –0.900  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.060  [0.529]***  [0.730]***   Fiscal multiplier (after one year)  –1.545  –1.590   Fiscal multiplier (after four years)  –1.498  –1.354  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –1.275  [0.476]**  [0.878]   Fiscal multiplier (after one year)  –0.673  –0.753   Fiscal multiplier (after four years)  –0.434  –0.060   Observations  520  391    Benchmark VAR  Additional variables: debt, sovereign rating index  Narrative consolidations       Response of output to a shock of 1% of CAPB  –2.076  –2.159  [0.343]***  [0.564]***   Fiscal multiplier (after one year)  –1.143  –1.220   Fiscal multiplier (after four years)  –0.989  –0.900  Revenue       Response of output to a shock of 1% of CAPB  –2.990  –3.060  [0.529]***  [0.730]***   Fiscal multiplier (after one year)  –1.545  –1.590   Fiscal multiplier (after four years)  –1.498  –1.354  Expenditure       Response of output to a shock of 1% of CAPB  –1.111  –1.275  [0.476]**  [0.878]   Fiscal multiplier (after one year)  –0.673  –0.753   Fiscal multiplier (after four years)  –0.434  –0.060   Observations  520  391  Interestingly, the differences are more marked when we focus on the composition effect. Figure 9 presents the output response to the expenditure and the revenue shock whereas the lower panels in Table 6 compare the estimated fiscal multipliers in the 6-VAR specification (column 2) with the ones in the previous 4-VAR specification (column 1). Now, the differences between the two shocks increase since the responses to the expenditure shock are not significant after one year. Fig. 9. View largeDownload slide Responses of output to a narrative revenue/expenditure shock, with financial variables as controls Fig. 9. View largeDownload slide Responses of output to a narrative revenue/expenditure shock, with financial variables as controls 4.2. Monetary policy influence The interaction between fiscal and monetary policy can greatly affect the size of fiscal multipliers. For example, in a Keynesian framework, monetary policy may react to fiscal consolidation episodes by reducing interest rates because inflationary pressures diminish. Moreover, under a Taylor-rule based monetary reaction, fiscal consolidation could produce a negative output gap, therefore leading to a drop in interest rates. Similarly, the probable response of the exchange rate could help cushion the impact of fiscal retrenchment on domestic demand. However, as Christiano et al. (2011) have shown, the counteracting effect of monetary policy could be less noticeable in a context where the economy hits the zero lower bound (ZLB) of interest rates since the space for more accommodative policies is exhausted. Our sample includes an important period where the ZLB is present for most economies, and therefore we expect a weaker response of monetary policy to fiscal developments. Consequently, we anticipate a lower fiscal multiplier if we control for interest rates in the period previous to 2009. To check that out, a policy interest rate17 is included in a 4-variable VAR of fiscal consolidations, with the results presented in Table 7. The estimated multiplier, –1.0, is not very much affected by the inclusion of monetary policy rates, which is consistent with a less responsive monetary policy when the crisis period is considered. That result is even stronger when the debt ratio and the sovereign rating index are considered in the VAR. In that case, the short-run fiscal multiplier over –1.0 remains robust, while in the 1978–2009 period the multiplier shrinks from –1.0 to less than –0.5. Table 7. Fiscal multipliers: the influence of monetary policy. VAR specification   1978–2009   1978–2014     4-VAR  6-VAR  4-VAR  6-VAR  Response of output to a shock of 1% of CAPB  –1.363  –0.691  –1.811  –1.953    [0.392]***  [0.488]  [0.344]***  [0.515]***  Fiscal multiplier (after one year)  –0.815  –0.486  –1.022  –1.132  Fiscal multiplier (after four years)  –0.544  –0.164  –0.844  –0.864  Observations  420  298  495  391    1978–2009   1978–2014     4-VAR  6-VAR  4-VAR  6-VAR  Response of output to a shock of 1% of CAPB  –1.363  –0.691  –1.811  –1.953    [0.392]***  [0.488]  [0.344]***  [0.515]***  Fiscal multiplier (after one year)  –0.815  –0.486  –1.022  –1.132  Fiscal multiplier (after four years)  –0.544  –0.164  –0.844  –0.864  Observations  420  298  495  391  Notes: See Table 3. With respect to the benchmark, 4-VAR includes an intervention interest rate. The 6-VAR incorporates the lagged debt-to-GDP ratio and the sovereign rating index. Thus, the ZLB on interest rates may have precluded the authorities from adopting a more accommodative monetary policy during the crisis that would have reduced the magnitude of the fiscal multipliers. Nevertheless, this analysis is largely limited by the inclusion of other variables reflecting the effects of non-conventional monetary policy actions after 2009 in many of the countries of the sample. Another line of investigation was the role of monetary policy in relation to the composition of the fiscal adjustments. We have obtained (not shown) that the short-run multiplier of revenue consolidation is not affected if we include monetary policy in the 1978–2014 period, while the expenditure multiplier is slightly reduced. This effect is stronger in the period 1978–2009, suggesting a more accommodative monetary policy stance for expenditure consolidations in the pre-crisis period, as suggested by the IMF (2010). 4.3. Euro area countries Under financial stress, the confidence channel may be more present in fiscal consolidation, for several reasons. First, consolidation today could avoid more extensive and more harmful consolidation in the future, as in the model presented in Bertola and Drazen (1993); second, risk premia could be reduced by the consolidation, reflecting a lower financial risk of sovereign debt. We could test these hypotheses by restricting our sample to the stressed euro area countries that received external financial support after the crisis. However, given the low number of countries in that group, we prefer to focus on all the euro area economies.18 Although these economies had very different fiscal positions, the consolidations in the euro area took place in a more financially restricted environment, with higher debt-to-GDP ratios. Against this background, the probability of an unstable sovereign risk scenario was greater. Additionally, the euro area economies are also highly interconnected, so spillovers from a large number of countries pursuing a consolidation of public finances at the same time could impact significantly on the size of the fiscal multiplier for the whole group. The results are summarized in Table 8. The one-year multiplier is close to unity in the benchmark 3-variable VAR (column 1). This –0.98 multiplier is lower than the estimated –1.14 for the whole sample (in Table 3). However, if we include lagged debt and the sovereign rating index, the multiplier is greatly affected, becoming non-significant. This difference of estimates between specifications in the euro area contrasts with the more stable multiplier found for the whole sample. Table 8. Fiscal multipliers: the euro area effect, VAR specification   Benchmark VAR  Additional variables: debt, sovereign rating index  Response of output to a shock of 1% of CAPB  –1.656  –0.771  [0.442]***  [0.53]  Fiscal multiplier (after one year)  –0.984  –0.457  Fiscal multiplier (after four years)  –0.733  –0.201  Observations  345  252    Benchmark VAR  Additional variables: debt, sovereign rating index  Response of output to a shock of 1% of CAPB  –1.656  –0.771  [0.442]***  [0.53]  Fiscal multiplier (after one year)  –0.984  –0.457  Fiscal multiplier (after four years)  –0.733  –0.201  Observations  345  252  Notes: See Table 3. The euro area economies included are Austria, Belgium, Finland, France, Germany, Italy, Netherlands, Portugal, and Spain. The estimation results are consistent with those of Guajardo et al. (2014), reported for a sample not including the recent crisis: fiscal consolidations preceded by high perceived sovereign default risk are less contractionary. Nevertheless, the estimation carried out in Table 8, only for the euro area countries, presents some instability depending on the chosen specification, and that may be due to the inclusion of the most recent years and the loss of observations after restricting the sample. 5. Conclusions We have examined the fiscal consolidation episodes that have taken place in a group of OECD countries after the global financial crisis (2009–2014). For that purpose we have constructed a dataset of policy actions—a narrative approach—from a broad set of official documents. Compared with previous periods of fiscal consolidation, during this episode the average size of the adjustment was larger, with more countries consolidating at the same time and with a strong focus on tax measures. Using dynamic panel data estimation, we are interested in the short-term effects of fiscal consolidation on economic activity. The different specifications—from single-equation to VAR systems—take into account the possible endogeneity of the regressors. Across all estimation methods the fiscal multiplier is negative and significant, in contrast to the results found previously with standard cyclically adjusted fiscal balance measures. Moreover, the average multiplier after one year is between –1.2 and –2.0, a higher multiplier than that found with historical episodes before 2009. We also obtain a significant real effect of revenue measures of around –1.6, while expenditure consolidations have an effect close to –0.7. These differences are even higher when looking at large consolidation episodes, with expenditure cuts having a non-significant effect after one year under certain specifications. This evidence showing the importance of the composition is closer to the expansionary fiscal contractions hypothesis, since it supports the view that spending cuts are more effective in stabilizing debt and avoiding economic downturns. In the last section we also present some evidence in favour of the need to consider non-conventional monetary policies to obtain a more accurate fiscal multiplier after the financial crisis and of the existence of a confidence channel for specific countries under financial stress that reduces the cost of fiscal consolidation. Finally, we believe there are two other natural extensions of this paper that need to be pursued. First, the current fiscal consolidation episodes are still ongoing in many economies and it is not yet possible to determine whether they have been successful in stabilizing and reducing high public debt ratios. Thus, more time observations will be needed to obtain a better assessment of this ongoing fiscal adjustment process. Second, our investigation has only disaggregated between revenues and expenditures. Efficiency arguments would also demand an analysis of current expenditure versus public investment and of direct versus indirect taxes. Supplementary material Supplementary material—the Appendix and the Data files—are available online at the OUP website. Footnotes 1 DeLong and Summers (2012) present the case of fiscal consolidations that reduce output in the medium term (self-defeated consolidation), because of the permanent effects of the recession. 2 According to the IMF (2016), mostly advanced economies have been the ones trying to stabilize their debt levels. By contrast, emerging economies, on average, had rising deficits and debt ratios after 2011. 3 We identify fiscal policy changes using historical documents as in Devries et al. (2011). 4 See, for example, Hernández de Cos and Moral-Benito (2013) and Jordá and Taylor (2013). 5 The simple contemporaneous correlation between the surprises on output and CAPB change for the 2009–2014 period is 0.59. The correlation with the narrative measure is –0.11. 6 The 15 countries are Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, Spain, the UK, and the USA. 7 GDP data taken from the OECD Economic Outlook (November 2014) and Eurostat. 8 In order to check whether the multicollinearity between the CAPB and the narrative fiscal measure could inflate the variance of the results, we estimated a Variance Inflation Factor (VIF) for eq. (2). All VIF were lower than 1.6, suggesting that multicollinearity does not appear to affect our estimations. 9 The F-test of the first stage has a p-value of less than 0.05 in all the equations, reinforcing the explanatory power of our narrative measures on CAPB. 10 The Hansen test (p-value = 0.331) provides further evidence of the exogeneity of our instruments. 11 The sample comprises the same 15 countries as in Section 2. 12 This differential effect during the recent period may be caused by the cyclical environment (proxied by the rolling window estimation), the composition of the adjustment (see Section 3.2), or the size of the consolidation. Preliminary evidence based on quantile regressions points to a lesser role of the latter. 13 The Akaike information criterion pointed to a lag structure with two lags. 14 We computed this estimation accounting for the build-up of fiscal consolidation episodes prior to the adoption of the euro in 1999. The results were unchanged. 15 CAPB data for revenue and expenditure to compare the VAR with the instrumental variable estimation is not available for all countries. 16 The variable is taken from Broto and Molina (2014). 17 Policy interest rates for countries are taken from Datastream. 18 The euro area countries in this sample are Austria, Belgium, Finland, France, Germany, Ireland, Italy, Netherlands, Portugal, and Spain. We consider the whole 1978–2014 period, although in the first part of that period each economy had its own independent monetary policy. Acknowledgements We would like to thank participants at the Banco de España and the ESM seminars for their helpful comments. The views expressed in this paper are those of the authors alone and do not necessarily reflect the views of the Banco de España. References Alesina A., Ardagna S. ( 2010) Large changes in fiscal policy: taxes versus spending, in Brown J.R. (ed.) Tax Policy and the Economy , vol. 24, National Bureau of Economic Research, Cambridge, MA, 35– 68. Alesina A., Barbiero O., Favero C., Giavazzi F., Paradisi M. ( 2015) Austerity in 2009–2013, Working Paper No. 20827, National Bureau of Economic Research, Cambridge, MA. Auerbach A., Gorodnichenko Y. ( 2012) Fiscal multipliers in recession and expansion, in Alesina A., Giavazzi F. (eds) Fiscal Policy After the Financial Crisis , University of Chicago Press, Chicago, IL, 98– 102. Beetsma R., Cimadomo J., Furtuna O., Giuliodori M. ( 2015) The confidence effects of fiscal consolidations, Working Paper Series No. 1770, European Central Bank, Frankfurt-am-Main. Google Scholar CrossRef Search ADS   Bertola G., Drazen A. ( 1993) Trigger points and budget cuts: explaining the effects of fiscal austerity, American Economic Review , 83, 11– 26. Blanchard O., Leigh D. ( 2013) Growth forecast errors and fiscal multipliers, IMF Working Paper, No. 13/1, Washington, DC. Broto C., Molina L. ( 2014) Sovereign ratings and their asymmetric response to fundamentals, Documentos de Trabajo Banco de España, No. 1428, Madrid. Burriel P., de Castro F., Garrote D., Gordo E., Paredes J., Pérez J. ( 2009) Fiscal policy shocks in the euro area and the US: an empirical assessment, Working Papers Series No. 1133, European Central Bank, Frankfurt-am-Main. Christiano L., Eichenbaum M., Rebelo S. ( 2011) When is the government spending multiplier large?, Journal of Political Economy , 119, 78– 121. Google Scholar CrossRef Search ADS   DeLong J.B., Summers L.H. ( 2012) Fiscal policy in a depressed economy, Brookings Papers on Economic Activity, Spring 2012, 233–297, Washington, DC. Devries P., Guajardo J., Leigh D., Pescatori A. ( 2011) An action-based analysis of fiscal consolidation in OECD countries, IMF Working Papers No. 11/128, Washington, DC. Galí J., López-Salido J.D., Vallés J. ( 2007) Understanding the effects of government spending on consumption, Journal of the European Economic Association , 5, 227– 70. Google Scholar CrossRef Search ADS   Giavazzi F., Pagano M. ( 1990) Can severe fiscal consolidations be expansionary? Tales of two small European countries, in Blanchard O.J., Fischer S. (eds) NBER Macroeconomics Annual , vol. 5, National Bureau of Economic Research, Cambridge, MA, 75– 122. Guajardo J., Leigh D., Pescatori A. ( 2014) Expansionary austerity? International evidence, Journal of European Economic Association , 12, 949– 68. Google Scholar CrossRef Search ADS   Hernández de Cos P., López Rodríguez D. ( 2014) Estructura impositiva y capacidad recaudatoria en España: un análisis comparado con la UE, Documentos Ocasionales No. 1406, Banco de España, Madrid. Hernández de Cos P., Moral-Benito E. ( 2013) Fiscal consolidations and economic growth, Fiscal Studies , 34, 491– 515. Google Scholar CrossRef Search ADS   IMF ( 2010) Fiscal Monitor , May, Washington, DC. IMF ( 2013) Staff report for the 2013 article IV consultation for Spain, IMF Country Report No. 13/244, Washington, DC. IMF ( 2014) Fiscal Monitor , April, Washington, DC. Jordá O., Taylor A. ( 2013) The time of austerity: estimating the average treatment effect of fiscal policy, Working Paper No. 2013‐25, Federal Reserve Bank of San Francisco, San Francisco, CA. Powell D. ( 2015) Quantile regression with nonadditive fixed effects, RAND Working Paper, Santa Monica, CA. Ramey V.A. ( 2011) Can government purchases stimulate the economy?, Journal of Economic Literature , 49, 673– 85. Google Scholar CrossRef Search ADS   Ramey V.A., Shapiro M.D. ( 1998) Costly capital reallocation and the effects of government spending, Carnegie-Rochester Conference Series on Public Policy , 48, 145– 94. Google Scholar CrossRef Search ADS   Romer C., Romer D. ( 2010) The macroeconomic effects of tax changes: estimates based on a new measure of fiscal shocks, American Economic Review , 100, 763– 801. Google Scholar CrossRef Search ADS   Sims C. ( 1980) Macroeconomics and reality, Econometrica , 48, 1– 48. Google Scholar CrossRef Search ADS   Smets F., Wouters R. ( 2007) Shocks and frictions in US business cycles: a Bayesian DSGE approach, American Economic Review , 97, 586– 606. Google Scholar CrossRef Search ADS   Appendix 1 Table A1. Narrative consolidations Country  Year  Consolidation  Revenue  Expenditure  Country  Year  Consolidation  Revenue  Expenditure  Australia  2012  0.3  0.2  0.1  Latvia  2009  9.75  3.8  5.95  Australia  2013  0.4  0.2  0.2  Latvia  2010  4.7  2.2  2.5  Austria  2011  0.7  0.4  0.3  Latvia  2011  2.3  1.6  0.7  Austria  2012  0.5  0.3  0.2  Latvia  2012  0.7  0.3  0.4  Austria  2013  0.7  0.2  0.5  Lithuania  2009  7.4  1.6  5.8  Austria  2014  0.4  –0.1  0.5  Lithuania  2010  4.5  0.5  4  Belgium  2010  0.4  0.1  0.3  Lithuania  2011  1.9  0.1  1.8  Belgium  2011  0.5  0.2  0.3  Lithuania  2012  1.3  0  1.3  Belgium  2012  1.9  1.1  0.8  Mexico  2010  0.7  0.7  0  Belgium  2013  0.9  0.5  0.4  Mexico  2011  0.8  0.8  0  Canada  2011  0.1  0.05  0.05  Mexico  2012  0.2  0.2  0  Canada  2012  0.1  0  0.1  Netherlands  2011  1.8  0.2  1.6  Canada  2013  0.3  0.05  0.25  Netherlands  2012  0.4  0.2  0.2  Canada  2014  0.5  0.1  0.4  Netherlands  2013  2.1  1.1  1  Czech Republic  2010  2.6  1.7  0.9  Netherlands  2014  1  0.5  0.5  Czech Republic  2011  1.6  0.8  0.8  New Zealand  2011  0.4  0  0.4  Czech Republic  2012  1.4  0.8  0.6  New Zealand  2012  0.9  0  0.9  Czech Republic  2013  1.1  0.8  0.3  New Zealand  2013  0.9  0  0.9  Denmark  2011  1.3  0.4  0.9  New Zealand  2014  0.9  0  0.9  Denmark  2012  0.5  0.3  0.2  Poland  2010  0.6  0  0.6  Denmark  2013  1.1  0.4  0.7  Poland  2011  2.4  1.3  1.1  Estonia  2009  9.2  3  6.2  Poland  2012  0.5  0.3  0.2  Finland  2010  0.2  0.1  0.1  Poland  2013  0.2  0.1  0.1  Finland  2011  0.6  0.7  –0.1  Poland  2014  0.1  –0.3  0.4  Finland  2012  0.3  0.3  0  Portugal  2010  2.2  1.7  0.5  Finland  2013  1.3  0.7  0.6  Portugal  2011  3.4  1.6  1.8  France  2011  0.9  0.4  0.5  Portugal  2012  6  2.2  3.8  France  2012  1.4  0.8  0.6  Portugal  2013  3.5  2.8  0.7  France  2013  2  1.4  0.6  Portugal  2014  1.9  0.5  1.4  France  2014  0.7  0.3  0.4  Slovakia  2011  1.9  1.1  0.8  Germany  2011  0.6  0.1  0.5  Slovakia  2012  1  0.3  0.7  Germany  2012  0.6  0.2  0.4  Slovakia  2013  3.9  2.6  1.3  Germany  2013  0.4  0  0.4  Slovenia  2010  2.6  0  2.6  Greece  2010  7.8  4.1  3.7  Slovenia  2011  0.7  0.1  0.6  Greece  2011  2.6  1  1.6  Slovenia  2012  2.9  0.5  2.4  Greece  2012  3.5  2  1.5  Slovenia  2013  2  1  1  Greece  2013  1.6  0.7  0.9  Spain  2010  0.9  0.7  0.2  Hungary  2010  4.1  0.6  3.5  Spain  2011  2.1  0.5  1.6  Hungary  2011  0.8  0  0.8  Spain  2012  4  1.6  2.4  Hungary  2012  3.3  2.1  1.2  Spain  2013  3.5  2  1.5  Hungary  2013  1  0.3  0.7  Spain  2014  1.2  0.7  0.5  Ireland  2009  5.8  3.6  2.2  Turkey  2010  1  0.8  0.2  Ireland  2010  1  0.2  0.8  Turkey  2011  1  0.8  0.2  Ireland  2011  3.26  0.86  2.4  Turkey  2012  1  0.8  0.2  Ireland  2012  2  0.8  1.2  United Kingdom  2010  0.3  0.3  0  Ireland  2013  2  0.8  1.2  United Kingdom  2011  1.7  1.4  0.3  Ireland  2014  1.3  0.6  0.7  United Kingdom  2012  1.1  0.8  0.3  Italy  2011  1  0.4  0.6  United Kingdom  2013  0.6  0.2  0.4  Italy  2012  2.8  2.3  0.5  United States  2012  0.2  0.1  0.1  Italy  2013  0.8  0.2  0.6  United States  2013  0.5  0.4  0.1  Italy  2014  1  0  1            Country  Year  Consolidation  Revenue  Expenditure  Country  Year  Consolidation  Revenue  Expenditure  Australia  2012  0.3  0.2  0.1  Latvia  2009  9.75  3.8  5.95  Australia  2013  0.4  0.2  0.2  Latvia  2010  4.7  2.2  2.5  Austria  2011  0.7  0.4  0.3  Latvia  2011  2.3  1.6  0.7  Austria  2012  0.5  0.3  0.2  Latvia  2012  0.7  0.3  0.4  Austria  2013  0.7  0.2  0.5  Lithuania  2009  7.4  1.6  5.8  Austria  2014  0.4  –0.1  0.5  Lithuania  2010  4.5  0.5  4  Belgium  2010  0.4  0.1  0.3  Lithuania  2011  1.9  0.1  1.8  Belgium  2011  0.5  0.2  0.3  Lithuania  2012  1.3  0  1.3  Belgium  2012  1.9  1.1  0.8  Mexico  2010  0.7  0.7  0  Belgium  2013  0.9  0.5  0.4  Mexico  2011  0.8  0.8  0  Canada  2011  0.1  0.05  0.05  Mexico  2012  0.2  0.2  0  Canada  2012  0.1  0  0.1  Netherlands  2011  1.8  0.2  1.6  Canada  2013  0.3  0.05  0.25  Netherlands  2012  0.4  0.2  0.2  Canada  2014  0.5  0.1  0.4  Netherlands  2013  2.1  1.1  1  Czech Republic  2010  2.6  1.7  0.9  Netherlands  2014  1  0.5  0.5  Czech Republic  2011  1.6  0.8  0.8  New Zealand  2011  0.4  0  0.4  Czech Republic  2012  1.4  0.8  0.6  New Zealand  2012  0.9  0  0.9  Czech Republic  2013  1.1  0.8  0.3  New Zealand  2013  0.9  0  0.9  Denmark  2011  1.3  0.4  0.9  New Zealand  2014  0.9  0  0.9  Denmark  2012  0.5  0.3  0.2  Poland  2010  0.6  0  0.6  Denmark  2013  1.1  0.4  0.7  Poland  2011  2.4  1.3  1.1  Estonia  2009  9.2  3  6.2  Poland  2012  0.5  0.3  0.2  Finland  2010  0.2  0.1  0.1  Poland  2013  0.2  0.1  0.1  Finland  2011  0.6  0.7  –0.1  Poland  2014  0.1  –0.3  0.4  Finland  2012  0.3  0.3  0  Portugal  2010  2.2  1.7  0.5  Finland  2013  1.3  0.7  0.6  Portugal  2011  3.4  1.6  1.8  France  2011  0.9  0.4  0.5  Portugal  2012  6  2.2  3.8  France  2012  1.4  0.8  0.6  Portugal  2013  3.5  2.8  0.7  France  2013  2  1.4  0.6  Portugal  2014  1.9  0.5  1.4  France  2014  0.7  0.3  0.4  Slovakia  2011  1.9  1.1  0.8  Germany  2011  0.6  0.1  0.5  Slovakia  2012  1  0.3  0.7  Germany  2012  0.6  0.2  0.4  Slovakia  2013  3.9  2.6  1.3  Germany  2013  0.4  0  0.4  Slovenia  2010  2.6  0  2.6  Greece  2010  7.8  4.1  3.7  Slovenia  2011  0.7  0.1  0.6  Greece  2011  2.6  1  1.6  Slovenia  2012  2.9  0.5  2.4  Greece  2012  3.5  2  1.5  Slovenia  2013  2  1  1  Greece  2013  1.6  0.7  0.9  Spain  2010  0.9  0.7  0.2  Hungary  2010  4.1  0.6  3.5  Spain  2011  2.1  0.5  1.6  Hungary  2011  0.8  0  0.8  Spain  2012  4  1.6  2.4  Hungary  2012  3.3  2.1  1.2  Spain  2013  3.5  2  1.5  Hungary  2013  1  0.3  0.7  Spain  2014  1.2  0.7  0.5  Ireland  2009  5.8  3.6  2.2  Turkey  2010  1  0.8  0.2  Ireland  2010  1  0.2  0.8  Turkey  2011  1  0.8  0.2  Ireland  2011  3.26  0.86  2.4  Turkey  2012  1  0.8  0.2  Ireland  2012  2  0.8  1.2  United Kingdom  2010  0.3  0.3  0  Ireland  2013  2  0.8  1.2  United Kingdom  2011  1.7  1.4  0.3  Ireland  2014  1.3  0.6  0.7  United Kingdom  2012  1.1  0.8  0.3  Italy  2011  1  0.4  0.6  United Kingdom  2013  0.6  0.2  0.4  Italy  2012  2.8  2.3  0.5  United States  2012  0.2  0.1  0.1  Italy  2013  0.8  0.2  0.6  United States  2013  0.5  0.4  0.1  Italy  2014  1  0  1            © Oxford University Press 2017 All rights reserved This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Oxford Economic PapersOxford University Press

Published: Apr 1, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off