The Liquid Hand-to-Mouth: Evidence from Personal Finance Management Software

The Liquid Hand-to-Mouth: Evidence from Personal Finance Management Software Abstract We use a very accurate panel of all individual spending, income, balances, and credit limits from a personal finance software to document spending responses to the arrival of both regular and irregular income. These payday responses are robust and homogeneous for all income and spending categories throughout the income distribution. Moreover, we find that few people hold little or no liquidity. We then analyze whether people hold liquidity cushions to cope with future liquidity constraints. However, we find that peoples’ responses are consistent with standard models without illiquid savings, in which neither present nor future liquidity constraints are frequently binding. Received May 31, 2016; editorial decision September 30, 2017 by Editor Itay Goldstein. Standard economic theory states that consumption should not respond to the timing of predictable changes in disposable income.1 However, a number of empirical studies report that consumption responds to disposable income or that it is “excessively sensitive” to income.2 This excess sensitivity and the mechanisms behind it are important for understanding the effectiveness of short-term stimulus payments among other policy prescriptions. Recent advances in the literature explain excess sensitivity with households’ financial structures. In the presence of illiquid savings, many households consume hand-to-mouth because they hold little or no liquid wealth Kaplan et al. 2014, Kaplan and Violante 2014b, Laibson et al. 2015. Using very accurate data on spending, income, balances, and credit limits, this paper shows that (1) spending is significantly excessively sensitive to income payments for almost all people; (2) less than 3% of people have less than 1 day of their average spending left in liquidity before receiving their paychecks; and (3) individual liquidity “cushions” are at least three times greater than predicted by the model of Kaplan and Violante (2014b). However, it is difficult to empirically determine whether individual liquidity cushions are great enough, because they depend on individual economic circumstances. This makes it difficult to infer whether people are liquidity constrained. To overcome this measurement problem, we analyze whether cash-holding responses indicate the presence of insufficient liquidity cushions and future liquidity constraints inspired by the cash-flow sensitivity of cash work in the corporate finance literature Almeida et al. 2004. However, we find that peoples’ cash-holding responses do not indicate the presence of insufficient liquidity cushions or future liquidity constraints. In line with previous studies, we start by documenting significant spending responses on paydays of both regular and irregular income. These payday responses are clean, robust, and homogeneous for all income and spending categories throughout the income distribution. When we split the sample into ten income deciles, we observe a monotonic decrease in the initial spending response from 70% to 40% above the average daily spending. We then undertake multiple steps to ensure that we do not pick up naturally coincident or coordinated timing of consumption and income such as restricting ourselves to subgroups of people (for instance, people whose income schedules do not coincide with typical patterns), subgroups of spending categories, or exogenous income arrivals. Moreover, we can directly measure liquidity constraints via balances and credit limits or the presence or absence of household spending on (discretionary) goods and services immediately before their payday. Sorting people into liquidity groups, cross-sectionally or relative to their own history of liquidity holdings, again results in payday responses that are decreasing but large even for the most liquid people. Moreover, almost all people hold a substantial amount of liquidity the morning of their paychecks. We thus conclude that the fraction of constrained households is too small to quantitatively generate the degree of excess sensitivity documented. Payday responses could be explained by our measure of liquidity constraints not capturing whether households actually “feel” liquidity constrained. More specifically, the measurement of liquidity constraints via balances and credit limits is not applicable if households hold cash or liquidity cushions either to cope with unforeseen expenses or to save for foreseen expenses. If households have insufficient liquidity cushions, they feel liquidity constrained because they worry about binding future liquidity constraints. Such insufficient liquidity cushions or potentially binding future liquidity constraints may explain payday responses even when present liquidity constraints are not binding because people want to wait for extra liquidity before they spend. To address this conjecture, we examine cash-holding responses to income payments. The basic idea is that people should have a high propensity to hold on to cash upon receiving income payments, if they are worried about binding liquidity constraints in the future. More formally, we consider three models: (1) a standard model in which people hold their lifetime savings in cash, such that the marginal propensity to hold on to cash is simply the reverse of the marginal propensity to consume; (2) a model with liquid and illiquid savings in which people optimally hold little or no cash; and (3) a model with liquid and illiquid savings in which future liquidity constraints bind frequently. The third model captures insufficient liquidity cushions and predicts a decreasing relationship between cash-holding responses and individual liquidity. This is a new testable prediction of future liquidity constraints or insufficient liquidity cushions. When we examine the empirical patterns of individual cash-holding responses to income payments, we find that cash-holding responses correspond to the first standard model, which assumes neither present nor future liquidity constraints and thus cannot explain the baseline high marginal propensities to consume out of income payments. Thus, we conclude that neither current nor future liquidity constraints (or “hard” and “soft” liquidity constraints) seem to explain payday responses.3 A natural question arises regarding the economic importance of understanding these payday responses. After all, the payday effects are small with around $${\$}$$30 of additional spending on paydays. The calculations of Browning and Crossley (2001) show that the utility loss from setting consumption equal to income (instead of smoothing it perfectly) is second order in a plausibly parameterized life-cycle buffer stock model.4 Thus, our documented payday effects only cause a negligible welfare loss in terms of imperfect consumption smoothing under standard assumptions. However, the standard model is clearly inconsistent with the payday responses we observe; therefore, its welfare predictions may not be the right ones to apply. But, even under other assumptions, welfare effects are probably small.5 Nevertheless, we still think that payday effects are important for two main reasons. First, fiscal policy makers want to understand not only the rate at which fiscal stimulus payments are consumed by households but also the mechanisms behind the effectiveness of tax rebates as short-term stimuli for consumption. Despite very comprehensive data, we cannot confirm the presence of hard or soft liquidity constraints as an explanation for payday responses.6 Moreover, we point to an important difficulty in measuring which people are liquidity constrained. To figure out whether people are liquidity constrained, as opposed to just having low liquid wealth, is of great importance: after all, liquidity constraints call for policy measures that expand credit versus low liquid wealth may be caused by overconsumption problems in which case credit should be restricted. In this context, we also document the interesting finding that less liquid people tend to decrease their overdraft limits around paydays, whereas more liquid people do not. Thus, instead of being liquidity constrained, people may prefer to restrict their access to credit because they have an overconsumption problem. Second, we document payday responses that are so clean and homogeneous throughout a population holding substantial liquidity that they teach us something about how people think about spending and income and appear to point toward a shortcoming in the way that we currently model economic behavior in a life-cycle consumption context. A powerful psychological reason to spend cash inflows appears to exist. Thus, people do not intertemporally optimize, but, instead, they use heuristics to decide how much to consume and save. In this paper, we remain agnostic about which environmental or preference-related theories drive hand-to-mouth behavior, and we assume that this behavior may be caused by any preference or cognitive, computational, and time limits of the household. However, we believe that our results call attention to an important issue: the lack of rigorous, portable, and generally-applicable models of such behavior. An early example of such a theory is Campbell and Mankiw (1989), who simply assume that a fraction of income goes to hand-to-mouth consumers who consume part of their disposable income each period. Beyond this approach, the only existing theory that rationalizes our findings is modeled in Addoum et al. (2015), who assume that peoples’ marginal utilities of consumption increase upon the arrival of income because they feel they have a license to spend. We follow Gelman et al. (2014), Baker (2013), Kuchler (2015), and Kueng (2015) and use data from a financial aggregation and service app; doing so overcomes the accuracy, scope, and frequency limitations of the existing data sources of consumption and income as it is derived from actual transactions and account balances. We follow Gelman et al. (2014) in documenting payday responses, but observe even cleaner and more homogeneous responses for all income levels and every income and spending category. This is because our data are from Iceland for the years 2011 to 2015 and exceptionally thorough with respect to capturing all income and spending. More specifically, (1) the income and spending data are precategorized (and the categorization is very thorough and accurate), (2) the app is marketed through banks and supplied for their customers (thus covering a fairly representative sample of the population), and (3) the data are basically free of one important shortcoming of all transaction-level data—the absence of cash transactions (in Iceland, consumers almost exclusively use electronic means of payment). Previous work on payday effects has restricted its attention to subpopulations. These papers document that expenditures and the caloric intake of poor households increase on payday Sims e.g., 2003, Huffman and Barenstein e.g., 2005, Shapiro e.g., 2005. More specifically, Sims (2003) and Mastrobuoni and Weinberg (2009) find that both consumption expenditures and consumption are higher in the week after Social Security checks are distributed than in the week before. Shapiro (2005) also rejects the exponential discounting model by showing that food stamp recipients consume 10% to 15% fewer calories the week before food stamps are disbursed. With respect to behavior and cognitive function around paydays, Carvalho et al. (2016) fail to find before-after payday differences in risk-taking, the quality of decision-making, the performance in cognitive function tasks, or in heuristic judgments. Other empirical papers that examine transitory payments to test the permanent income hypothesis include Shapiro and Slemrod (1995), Shapiro and Slemrod (2003a), Parker (1999), among many others. However, these studies on the share of hand-to-mouth consumers are based on surveys that make “following the money” of consumers difficult because respondents may have little incentive to answer the questions accurately, may not understand the wording of the questions, or may behave differently in practice and forget their reported behavior. Moreover, such measurement error or noise in the data generated by surveys can increase with the length of the recall period de Nicola and Giné 2014. Additionally, surveys can produce biased (rather than merely noisy) data if respondents have justification bias, concerns about surveyors sharing the information, or stigma about their consumption habits Karlan and Zinman 2008. Overall, the conclusions of this literature regarding liquidity constraints are very mixed. Shapiro and Slemrod (2009) document that poor households, which are arguably more likely to be liquidity constrained, did not spend most of the 2008 tax rebate as the fiscal stimulus package intended. Shapiro and Slemrod (1995) conclude that liquidity constraints do not motivate the spending behavior of the 43% of households that report that the timing of tax payments will affect their consumption. Souleles (1999) examines the responses in nondurable and durable consumption. The author finds that constrained households are more likely to spend their tax refunds on nondurable consumption; the picture is reversed for durable consumption. Thus, liquidity-unconstrained households are not overwithholding to force themselves to save up enough for durable consumption goods because they could easily undo any forced saving by drawing down their liquid assets. As noted by Kaplan and Violante (2014b), wealthy people may engage in hand-to-mouth behavior due to illiquid wealth. Recent theoretical examples of models with liquid and illiquid assets are Angeletos et al. (2001), Laibson et al. (2003), Flavin and Nakagawa (2008), Chetty and Szeidl (2007), Alvarez et al. (2012), Huntley and Michelangeli (2014), and Kaplan and Violante (2014a). Angeletos et al. (2001) and Laibson et al. (2003) show that households with hyperbolic-discounting preferences optimally decide to lock their wealth in the illiquid asset in order to cope with self-control problems. Kaplan and Violante (2014a) do not need to assume that households have hyperbolic-discounting preferences and still generate a high marginal propensity to consume out of transitory shocks. In a one-asset environment, Koszegi and Rabin (2009) show that, in an environment with little to no uncertainty, agents with reference-dependent preferences may consume entire windfall gains, and Pagel (2013) shows that the preferences also rationalize excess smoothness in consumption. Moreover, Reis (2006) assumes that agents face costs when processing information and thus optimally decide to update their consumption plans sporadically, resulting in excessively smooth consumption that is shown to matter in the aggregate by Gabaix and Laibson (2002). Additionally, Tutoni (2010) assumes that consumers are rationally inattentive, and Attanasio and Pavoni (2011) show that excessively smooth consumption results from incomplete consumption insurance due to a moral hazard problem. 1. Data and Summary Statistics 1.1 Data This paper exploits new data from Iceland generated by Meniga, a financial aggregation software provider to European banks and financial institutions. Meniga has become Europe’s leading financial management (PFM) provider. Meniga’s account aggregation platform allows bank customers to manage all their bank accounts and credit cards across multiple banks in one place by aggregating data from various sources (internal and external). Meniga’s financial feed documents consumers’ budgets in a social media style. Figure 1 displays screenshots of the app’s user interface. The first screenshot shows the background characteristics that the user provides; the second one shows transactions; the third one shows bank account information; and the fourth one shows a sample of accounts that can be added. Figure 1 View largeDownload slide Screenshots of the financial aggregation app Figure 1 View largeDownload slide Screenshots of the financial aggregation app In October 2014, the Icelandic population was 331,310, and 20% of Icelandic households were using the Meniga app. Because the app is marketed through banks and automatically supplied to customers using online banking, the sample of Icelandic users is fairly representative. Each day, the application automatically records all bank and credit card transactions (including descriptions as well as balances), overdraft limits, and credit limits. We use the entire de-identified population of active users in Iceland and the data derived from their records from 2011 to 2015. We perform the analysis on normalized and aggregated user-level data for different income and spending categories. Additionally, the app collects demographic information, such as age, gender, marital status, and postal code. Moreover, we can infer employment status, real estate ownership, and the presence of young children in the household from the data. We have the following regular income categories: child support, benefits, child benefits, interest income, invalidity benefits, parental leave, pension income, housing benefits, rental benefits, rental income, salaries, student loans, and unemployment benefits. In addition, we have the following irregular income categories: damages, grants, other income, insurance claims, investment transactions, reimbursements, tax rebates, and travel allowances. The spending categories are groceries, fuel, alcohol, ready-made food, home improvements, transportation, clothing and accessories, sports and activities, and pharmacies. We can observe expenditures on alcohol that is not purchased in bars or restaurants because a state-owned company, the State Alcohol and Tobacco Company, has a monopoly on the sale of alcohol in Iceland. We exclude all recurring spending such as rent and bill payments. 1.2 Summary statistics Table 1 displays summary statistics of the Icelandic users, including not only income and spending in U.S. dollars but also some demographic statistics. We can see that the average user is 40 years old; 15% of the users are pensioners; 50% are female; 20% have children; and 8% are unemployed. For comparison, Statistics Iceland reports that the average age in Iceland is 37 years; 12% of Icelanders are pensioners; 48% are female; 33% have children; and 6% are unemployed. Thus, our demographic statistics are representative for those of the overall Icelandic population. Moreover, our sample characteristics are very similar to U.S. data. The average age in the U.S. population is 38; the percentage of women in the United States is 51%; and the mean income in the U.S. population in 2015 dollars per adult member is $${\$}$$3,266. In our sample, the individual monthly mean income is $${\$}$$3,256.7 This is not the case for other studies using app data that observe a user population more likely than the overall population to be young, financially secure, male, and tech savvy. Table 1 Summary statistics    Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12     Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12  All income, salary, and spending numbers are in U.S. dollars. Parentheses indicate that data categories do not match perfectly. Table 1 Summary statistics    Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12     Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12  All income, salary, and spending numbers are in U.S. dollars. Parentheses indicate that data categories do not match perfectly. The representative national household expenditure survey conducted by Statistics Iceland also reports income and spending statistics. In Table 1, parentheses indicate when spending categories do not match perfectly with the data. We can see that the income and spending figures are remarkably similar for the categories that match well. Figures 2 to 4 show the distribution of regular, salary, and irregular income payments over the course of a month. Approximately 85% of the people in the sample were paid on a monthly basis, whereas the remaining people are paid on a more frequent basis. This variation allows us to also consider people who are paid on unusual schedules. Additionally, the irregular payments are distributed rather evenly over the course of the month. Figure 2 View largeDownload slide The distribution of regular income arrivals over a month Figure 2 View largeDownload slide The distribution of regular income arrivals over a month Figure 3 View largeDownload slide The distribution of paycheck arrivals over a month Figure 3 View largeDownload slide The distribution of paycheck arrivals over a month Figure 4 View largeDownload slide The distribution of irregular income arrivals over a month Figure 4 View largeDownload slide The distribution of irregular income arrivals over a month 2. Analysis In this study, we estimate payday effects by running the following regression   \begin{equation} \label{one} x_{it} = \sum\limits_{k=-7}^{7}\beta_{k}I_i(Paid_{t+k}) + \delta_{dow} + \phi_{wom} + \psi_{my} + \eta_i + \epsilon_{it} \end{equation} (1) where $$x_{it}$$ is the ratio of spending by individual $$i$$ to his or her average daily spending on date $$t$$, $$\delta_{dow}$$ are day-of-week fixed effects, $$\phi_{wom}$$ are week-of-the-month fixed effects, $$\psi_{my}$$ are month-by-year fixed effects, $$\eta_{i}$$ are individual fixed effects, and $$I_i(Paid_{t+k})$$ is an indicator that is equal to $$1$$ if $$i$$ receives a payment at time $$t+k$$ and that is equal to $$0$$ otherwise. The $$\beta_{k}$$ coefficients thus measure the fraction by which individual spending deviates from the average daily spending in the days surrounding the receipt of a payment. We use indicator variables for income payments to alleviate potential endogeneity concerns at the income level. The individual fixed effects control for all observable or unobservable individual characteristics. The day-of-week dummies capture within-week patterns for both income and spending. Beyond the day-of-week fixed effects, the week-of-the-month fixed effects control for some mechanical effects due to fixed expense cycles at the beginning of each month. Finally, the month-by-year dummies capture all slow-moving trends. Standard errors are clustered at the individual level. We will initially differentiate between the arrival of regular and irregular income and separate households into ten income deciles. All the following figures and regression tables display the $$\beta_{k}$$ coefficients for different spending categories, income categories, and sample splits. 2.1 Regular income payments Figure 5 displays the spending responses to regular income payments of households in ten different income deciles, as measured by their regular salaries. Both poor and rich households clearly respond to the receipt of their income, with the poorest households spending 70% more than they would on an average day and the richest households spending 40% more. Even for the richest households, we thus observe a surprisingly high spending response. Table 2 presents all regression results for four income quartiles and four types of spending. While grocery and fuel spending can be regarded as necessary, ready-made food (such as restaurants) and alcohol spending can be regarded as discretionary. Figures A.1 and A.2 in the Appendix separately display the spending responses to income for all necessary categories and all discretionary categories. People are equally inclined to spend on necessary and discretionary goods and services upon receiving their income. In terms of magnitudes in dollars, we observe an additional spending of approximately $${\$}$$30 on paydays. Figure 5 View largeDownload slide The effects of regular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day dummy: the deviation of spending from average daily spending. Figure 5 View largeDownload slide The effects of regular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day dummy: the deviation of spending from average daily spending. Table 2 The impact of payments on household spending by income quartiles    (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)     (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01. Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. The salary arrival responses are estimated by salary quartiles, whereas the responses to any payments, regular payments, and irregular payments are estimated by total income quartiles. The outcome is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. Table 2 The impact of payments on household spending by income quartiles    (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)     (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01. Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. The salary arrival responses are estimated by salary quartiles, whereas the responses to any payments, regular payments, and irregular payments are estimated by total income quartiles. The outcome is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. There is no change in permanent income on paydays and there is no new information because paydays are perfectly predictable. While a buffer stock model can potentially explain sensitivity to surprising large payments or changes in permanent income, it cannot explain sensitivity to regular paydays. Thus, these payday responses are inconsistent with standard models of consumption and savings. Although we focus on nonrecurring spending (exclude all rent and bill payments) and control for day-of-week and week-of-month fixed effects, this spending response to regular income might stem from the coincident timing of regular income and irregular spending. Therefore, we will now examine irregular income. 2.2 Irregular income payments Figure 6 displays the spending responses to irregular income payments of households in ten different income deciles, as measured by their regular salaries. Again, we observe both poor and rich households responding to the receipt of their income, and poor households’ spending responses are somewhat more pronounced. Again, even for rich households, the spending response on payday is large and significant, at approximately 40%. Thus, we do not conclude that the bulk of the spending responses to income or the excess sensitivity of consumption is due to poor households or the coincident timing of regular income and spending, as proposed in Gelman et al. (2014). Figure 6 View largeDownload slide The effects of irregular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 6 View largeDownload slide The effects of irregular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.3 Heterogeneity We are interested in the question of whether the payday responses are prevalent for a large fraction of the population or are being driven by a small fraction of the population. To do so, we run a regression for each individual in four income quartiles and display their individual payday coefficients in Figure 7. Approximately 22% of people have a payday coefficient equal to zero. A greater number of positive coefficients outweigh some negative coefficients, leading to an average coefficient of approximately 0.6 for the lowest quartile and 0.4 for the highest quartile. Therefore, at least half of the population, rather than a small fraction of the population, displays significantly positive and large payday responses. Figure 7 View largeDownload slide The distribution of payday coefficients for people by income and salary quartiles The payday coefficient is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. Figure 7 View largeDownload slide The distribution of payday coefficients for people by income and salary quartiles The payday coefficient is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. 2.4 Intensive versus extensive spending We are interested in the question of whether payday responses are an intensive or extensive phenomenon in the sense of people spending more when they go shopping or making an additional shopping trip. In Table 3, we display the results of regressions that estimate how much more likely people are to buy in different categories, such as groceries, fuel, and restaurants, on their payday. For instance, people are 11% more likely to go on any shopping trip on paydays. In a second set of regressions, we then compare how much they spend if they shop on a payday relative to any other day. People spend $${\$}$$21 more on all shopping trips on their paydays. Because people spend, on average, $${\$}$$50 every day on nonrecurring consumption and approximately $${\$}$$80 on paydays, this $${\$}$$21 increase corresponds to approximately 80% of the increase in spending on paydays ($${\$}$$30). Thus, people are more likely to go shopping and, if they go shopping, they spend more than they would on a shopping day when they are paid. Table 3 Intensive and extensive spending reaction    (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)     (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01 Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. Panel A shows the effect on the probability of buying the goods under consideration on payday. Panel B compares the expenditure on shopping days when consumers are paid to those when they do not. Table 3 Intensive and extensive spending reaction    (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)     (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01 Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. Panel A shows the effect on the probability of buying the goods under consideration on payday. Panel B compares the expenditure on shopping days when consumers are paid to those when they do not. 2.5 Financial sophistication We observe a number of potential proxies for financial sophistication: age, pensions, employment, benefit payments, number of logins, voluntary reductions of overdraft limits, banking fees paid, payday loans, simultaneous savings and overdraft debt, large checking account balances that do not pay interest, and whether people link their spouse. We first examine simultaneous savings and overdraft debt, which can be considered a mistake because overdrafts cost more interest than savings yield. Figure 8 shows the spending responses of people sorted by how much interest is lost by holding overdrafts and savings simultaneously. People who lose less have less pronounced spending responses than those who lose more. A possible reason for this result is that wealthier people lose more by having savings and overdrafts simultaneously. The spending responses of people sorted by our other measures of financial sophistication show a similar picture. For instance, the same is true when we sort people by a summary measure of how much they lose in banking fees, interest, and payday loans in Figure A.3. We thus conclude that our measures of financial sophistication do not predict peoples’ propensities for payday responses well. Figure 8 View largeDownload slide The effects of regular income arrival on spending by amount lost due to holding overdrafts and savings simultaneously The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 8 View largeDownload slide The effects of regular income arrival on spending by amount lost due to holding overdrafts and savings simultaneously The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.6 Other explanations for payday effects Thus far we cannot pinpoint an explanation for payday responses as we fail to find any population group or income or spending category for which the payday responses are absent or less pronounced. Because irregular income responses may be unanticipated, payday responses are not necessarily inconsistent with the standard model. Nevertheless, confirming the existence of payday responses for irregular income rules out many alternative explanations for payday responses to regular income, such as naturally recurring spending and income or coordination stories that would not be picked up by our fixed effects. We will now review all potential other explanations. We perform a number of additional exercises to figure out potential mechanisms behind payday responses. First, we take a closer look at the characteristics of the people in the lowest income decile because their spending responses appear to be slightly different from the other income deciles. However, we do not observe unusual characteristics. For instance, the mean income of people in the lowest income decile is approximately $${\$}$$750 and their mean age is 34 years, while the second decile’s mean income and age are approximately $${\$}$$1,000 and 34 years, respectively. Second, we examine the responses for the ready-made-food category because such spending is instantly consumed. Third, we examine only people who are paid on unusual days. In doing so, we also ensure that we observe payday responses for all categories and not simply categories that are likely to be consumed alongside coworkers who are paid on the same unusual paydays (such as restaurants and alcohol). Fourth, we examine only tax rebates and exogenous wealth shocks (described below) to control for potential endogeneity of income. Additionally, we can restrict our sample to income payments that arrive at a certain day of the month and include day-of-month fixed effects, which obtains exogenous variation in the arrival of payments due to weekends and holidays. Fifth, we examine people who have linked their spouse to ensure that the responses are not driven by intra-household bargaining. Sixth, any price-discriminatory response of firms does not explain the magnitude of the observed effects Hastings and Washington 2010 and does not apply to people with unusual paydays or irregular income. Seventh, we sort people by how often they log into the app to ensure that app usage is unrelated to payday responses. Moreover, we can restrict the sample to pre-2014, the year when the smartphone app was introduced (before people could access only via a desktop or laptop computer), but we do not find any differences in payday responses. All of these exercises yield similar payday responses. Thus, we conclude that spending responses to income payments constitute a very robust phenomenon, which is cleanly estimated and prevalent throughout the population. Given the robustness of these payday responses, we think that attempting to better understand what is driving them is a valuable exercise. People plausibly incur adjustment costs when they spend. Nevertheless, we do not believe that adjustment costs can explain our payday effects for the following reasons: (1) Plausible magnitudes of adjustment cost depend on the category of consumption under consideration. If one thinks of housing or car expenses, high adjustment costs are plausible. However, plausible tangible or intangible costs to changing nondurable consumption are probably low. Moreover, these costs may vary with the consumption category, for instance, dependent on whether these are durables or nondurables. However, the payday effects are very similar across the different consumption categories. This suggests to us that adjustment costs are eaten up by the calendar fixed effects and the payday effects pick up a residual but strong psychology to spend cash inflows. (2) The payday effects do not depend in a systematic way on the size of payments. For instance, irregular payments are much smaller than regular payments but yield the same magnitude in payday responses. (3) We observe sizable and significant payday effects at both the intensive and extensive margin. The extensive margin is difficult to reconcile with fixed adjustment costs. Nevertheless, if people hold plenty of cash or liquidity (as we will show below), they could have done any shopping trip a day early rather than the day they are paid. All income payments are made via direct deposit in Iceland, and we do not observe any check transactions. Moreover, just one clearing house in the country exists, so all transactions are recorded without delay. (4) More generally, if people hold plenty of cash or liquidity, they would not react to a perfectly predictable regular income payment, an irregular income payment, or an exogenous income payment (such as a lottery payment). However, we find the same magnitudes in payday responses for all of these income categories. Using large exogenous wealth shocks, we can also estimate the marginal propensity to consume in response to fiscal stimulus payments of our sample population. The shocks that we use originate from a debt relief ruling that resulted in large repayments from banks to thousands of Icelandic households holding foreign-indexed debt. In this natural experiment, Icelandic lenders had to pay out as much as $${\$}$$4.3 billion, that is, one-third of the economy’s gross domestic product (GDP), after a court found that some foreign loans were illegal. These foreign loans were the largest single loan category of the banks, with a value of approximately $${\$}$$7.2 billion.After the financial crisis, the Icelandic Supreme Court ruled on June 16, 2010, that loans indexed to foreign currency rates were illegal in three cases involving private car loans and a corporate property loan. This decision meant that borrowers with such loans were only obliged to repay the principal in Icelandic krona, making the lenders liable for currency losses of approximately $${\$}$$28 billion in debt because the krona’s value against the Japanese yen and Swiss franc declined by one-third since September 2008.8 After the debt-relief ruling, banks had to repay their customers, which we consider to be exogenous wealth shocks. We obtain marginal propensities to consume that are perfectly in line with existing papers, such as Agarwal and Qian (2014), who analyze Singaporean consumers’ responses to a fiscal stimulus announcement and payout, and Kueng (2015), who uses payments originating from the Alaska Permanent Fund.9 We thus conclude that Icelanders do not exhibit larger spending responses to income windfalls than do people in other countries. 2.7 Comparison to Gelman et al. (2014) We now briefly discuss the differences between the payday effects of Gelman et al. (2014) and ours. Four main differences stand out: First, our payday response is concentrated on the payday. In Gelman et al. (2014), there appears a second larger spike a few days after the payday. The reason is probably the absence of check transactions in the Icelandic data. In Iceland, everybody is paid by direct deposit. In contrast, many people in the United States receive paper paychecks and that appears to cause a spending boost and then another one once the paycheck is deposited. Second, we observe payday responses for many income categories whereas Gelman et al. (2014) restrict themselves to regular paycheck and Social Security payments. Thus, we are able to only look at people who are paid on unusual days and document payday responses to irregular payments. Third, we observe payday responses for many spending categories, whereas Gelman et al. (2014) restrict themselves to recurring, nonrecurring, and coffee shop and fast food spending. The coffee shop and fast food measure is identified using keywords from the transaction descriptions. In contrast, when we receive the data, it is already categorized by a three-tiered approach: system rules and user and community rules. The system rules are applied in instances where codes from the transactions systems clearly indicate the type of transaction being categorized. For example, when transactions in the Icelandic banking system contain the value “04” in a field named “Text key” the payer has indicated payment of salary. User rules apply if no system rules are in place and when a user repeatedly categorizes transactions with certain text or code attributes to a specific category. In those instances the system will automatically create a rule which is applied to all further such transactions. If neither system rules nor user rules apply, the system can sometimes detect identical categorization rules from multiple users which allows for the generation of a community rule. While Gelman et al. (2014) do not observe payday responses to coffee shop and fast food spending, we observe responses of similar magnitude for restaurant spending, which isolates a discretionary, nondurable, and divisible form of spending. To further understand why our payday responses appear to be cleaner, we reran the regressions using only 300 consecutive days in 2012 and 2013, as Gelman et al. (2014) use, but we find very similar responses. We thus conclude that our categorization and measurement of spending and income make a difference. The Diary of Consumer Payment Choice (DCPC), conducted in October 2012 by the Boston, Richmond, and San Francisco Federal Reserve Banks, shows that cash makes up the single largest share of consumer transaction activity at 40%, followed by debit cards at 25%, and credit cards at 17%. Electronic methods (online banking bill pay and bank account number payments) account for 7%, while checks make up 7%. All other payments represent less than 5% of monthly transaction activity, with text and mobile payments barely registering at less than one-half of 1%. By value, cash accounts for 14% of total consumer transaction activity, while electronic methods make up 27%, and checks 19%. These findings suggest that cash is used quite often, but primarily for low-value transactions such as coffee shop and fast food spending. Overall, just from eyeballing the responses by Gelman et al. (2014), we feel that our responses are considerably more clean, homogeneous, and robust. 2.8 Examining liquidity constraints Thus far, our results suggest that hand-to-mouth behavior is prevalent across all income groups, which casts doubt on liquidity constraints as the only explanation for such behavior. To further establish that liquidity-constrained households are not alone in exhibiting spending responses, we now examine different measures of liquidity constraints. The first is simply cash holdings in checking and savings accounts. The second concerns liquidity or maximum borrowing capacity and equals cash holdings in checking and savings accounts minus credit card balances plus credit limits and overdraft limits. The third is credit utilization and the fourth is spending on (discretionary) goods and services immediately before income payments. The consumer credit landscape in Iceland is slightly different from the United States. Most importantly, rolling over credit card debt is possible, but people in Iceland seldom do so. Instead Icelanders typically pay off the balance in full at the end of the month using their current account. An overdraft occurs when withdrawals from a current account exceed the available balance. This means that the balance is negative and hence that the bank is providing credit to the account holder and interest is charged at the agreed rate. Virtually all current accounts in Iceland offer a pre-agreed overdraft facility, the size of which is based upon affordability and credit history. This overdraft facility can be used at any time without consulting the bank and can be maintained indefinitely (subject to ad hoc reviews). Although an overdraft facility is authorized, technically the money is repayable on demand by the bank. In reality, this is a rare occurrence as the overdrafts are profitable for the bank and expensive for the customer. Moreover, unlike those in the United States, current accounts do not feature minimum balance requirements. All liquidity measures are normalized by each individual’s average spending; that is, we measure individual liquidity in individual average consumption days rather than absolute liquidity. Figures 9, 10, and A.4 compare the spending responses to regular and irregular income for three terciles of our standard measures of liquidity: cash holdings (checking and savings account balances), liquidity (overdraft and credit limits plus checking and savings accounts balances minus credit card balances), and credit utilization. Additionally, instead of sorting cross-sectionally, Figure 11 sorts people into terciles of cash or liquidity relative to their own history of cash or liquidity holdings; that is, for each individual we generate personal liquidity terciles to compare them within their own histories. Figures A.5 and A.6 compare the spending responses to regular and irregular income for three terciles of our alternative measure of liquidity: whether people spend on (discretionary) goods and services prior to income payments. Overall, we see that households exhibit spending responses, even in the highest tercile of all liquidity measures. Figure 9 View largeDownload slide The effects of regular and irregular income on spending by cash (measured by the median number of consumption days held in cash (checking and savings account balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 9 View largeDownload slide The effects of regular and irregular income on spending by cash (measured by the median number of consumption days held in cash (checking and savings account balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 10 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by the median number of consumption days held in liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 10 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by the median number of consumption days held in liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 11 View largeDownload slide The effects of regular and irregular income on spending by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) relative to peoples’ own histories of liquidity holdings The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 11 View largeDownload slide The effects of regular and irregular income on spending by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) relative to peoples’ own histories of liquidity holdings The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Furthermore, we examine the distribution of cash holdings and liquidity before paydays in Figure 12. We see that cash holdings fall discontinuously at zero when overdrafts start to cost interest and that approximately 10% of people hold fewer than 10 days of cash in their checking and savings accounts. Moreover, approximately 10% of people hold fewer than 10 days of liquidity. In turn, Figure 13 provides a breakdown by 1 to 10 days of spending for cash and liquidity for people who hold less than 10 days of cash or liquidity, respectively. Here, we see that less than 3% of people hold less than 1 day of spending in cash and that less than 3% hold less than 1 day of spending in liquidity. Thus, according to our measures, the fraction of liquidity-constrained people is quantitatively too small to explain the observed spending responses to income. Thus, we conclude that liquidity constraints in the literal sense are unlikely to explain payday responses. Figure 12 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) the morning of paydays Figure 12 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) the morning of paydays Figure 13 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) before paydays of those people with less than 10 consumption days in cash or liquidity Figure 13 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) before paydays of those people with less than 10 consumption days in cash or liquidity Additionally, Table 4 displays summary statistics for the three terciles of liquidity in consumption days. We can see that even the least liquid households hold considerable liquidity of approximately 38 days of spending, while the most liquid tercile of people holds approximately 546 days of spending in liquidity. When we compare these numbers to the state-of-the-art model developed by Kaplan and Violante (2014b) to explain high marginal propensities to consume out of tax rebates, we see a discrepancy between the theoretical predictions and the empirical evidence on the amount of liquid assets that people hold as we explain next. Table 4 Summary statistics by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) in consumption days    (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546     (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546  All income, salary, spending, cash, and liquidity numbers are in U.S. dollars. Table 4 Summary statistics by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) in consumption days    (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546     (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546  All income, salary, spending, cash, and liquidity numbers are in U.S. dollars. Figure 15 shows the life-cycle profiles of liquidity normalized by quarterly consumption for five quintiles of the distribution of agents in the model of Kaplan and Violante (2014b). We see that the liquid asset holdings of the bottom three quintiles are basically zero for all over the simulated agents’ lives. The top two quintiles of agents hold, on average, approximately 4 quarters of consumption in liquidity. By contrast, empirically, the most liquid tercile of people holds, on average, 6.1 quarters of consumption in liquidity, while the middle and least liquid terciles hold 1.37 and 0.41 quarters of consumption in liquidity, respectively, all of which far exceeds the predictions of the model.10 Moreover, if the Kaplan and Violante (2014b) model is to generate the amount of liquidity that we observe in the data, the fixed costs of illiquid assets must be very low, which implies that people can easily adjust their illiquid asset holdings, which reduces their marginal propensity to consume out of fiscal stimulus payments.11 Figure 15 View largeDownload slide Life-cycle profiles of liquid asset in consumption (quarterly) as predicted by the model in Kaplan and Violante (2014b) Figure 15 View largeDownload slide Life-cycle profiles of liquid asset in consumption (quarterly) as predicted by the model in Kaplan and Violante (2014b) Clearly, the Kaplan and Violante (2014b) life-cycle liquidity profiles are a function of medicaid, social security, unemployment insurance, and any institution that partially insures income risk differentially over the life cycle. Iceland is one of the signature countries of the Nordic model. The Nordic model refers to the economic and social policies common to the Nordic countries (Denmark, Finland, Norway, Iceland, and Sweden) including a combination of free market capitalism with a comprehensive welfare state and collective bargaining at the national level. Nevertheless, if anything, the more comprehensive welfare system of Iceland would reduce the need for liquidity and cash cushions in Iceland relative to the United States. According to our measures, the fraction of liquidity-constrained people is quantitatively too small to explain the observed spending responses to income. Nevertheless, many people hold rollover debt; the lowest tercile holds an average of 38 days of their average spending in debt. Obviously, low resources do not necessarily imply liquidity constraints that are determined by future income or other assets that the agent wishes to but cannot borrow against or collateralize. These results suggest that liquidity constraints are not straightforward to document empirically. Almost none of our people are liquidity constrained in the literal sense, that is, they live from paycheck to paycheck and have no ability to consume before their paycheck. Nevertheless, many households may hold a cash or liquidity cushion for either unforeseen adverse expenditure shocks or foreseen expenses. However, they may still be liquidity constrained inasmuch as they would consume or invest more if they could borrow more because they expect higher income in the future or have other assets they cannot collateralize. We are thus left wondering: how can we define liquidity-constrained people and identify them empirically? The average liquidity holdings in our data are pretty similar to the average liquidity holdings in the U.S. data. Table 5 shows a direct comparison of monthly income, cash, and liquidity for three terciles of the liquidity distribution of our data with U.S. data from the Survey of Consumer Finances (SCF) Kaplan and Violante using the 2001 survey as 2014b.12 In fact, average liquidity holdings in the United States are slightly higher than in Iceland. Again, institutions and social norms are very different but the numbers of both countries raise the same question: when do we label an individual as liquidity constrained if he or she holds substantial debt but also liquidity? Table 5 Summary statistics by terciles of liquidity: Comparison to the United States (SCF 2001 data like in Kaplan and Violante (2014b))    (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0     (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0  All numbers are in U.S. dollars. For the SCF data, strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). Cash equals strict liquid wealth minus credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. Table 5 Summary statistics by terciles of liquidity: Comparison to the United States (SCF 2001 data like in Kaplan and Violante (2014b))    (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0     (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0  All numbers are in U.S. dollars. For the SCF data, strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). Cash equals strict liquid wealth minus credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. The theoretical literature has explicitly considered wealthy households to be liquidity constrained when they lock their wealth in illiquid assets Laibson et al. 2003, Kaplan and Violante 2014b. However, empirically, we find that almost all households hold large amounts of cash and only very few hit a liquidity constraint even right before their paychecks. Because Kaplan and Violante (2014b) use Survey of Consumer Finances data, the authors do not observe liquidity holdings before paychecks but only average liquidity holdings. They classify people as hand-to-mouth consumers when their average liquid wealth is less than half of their earnings, which they find to be the case for 30% of the U.S. population. For comparison, using their definition, we find that in our population 58% of households live hand-to-mouth. However, because people have sufficient liquidity at the end of their pay cycles, this finding cannot explain payday responses to income. People who choose to hold a significant amount of liquidity could “feel” liquidity constrained because they hold an insufficient cash or liquidity cushion. A potential approach to assess whether payday responses are driven by these people is the following: people who have just received a large exogenous wealth shock should not exhibit payday responses, as they are exogenously more liquid. In Figure 14, we thus show that people exhibit substantial payday responses even in the months after which they received a large exogenous wealth shock from a court ruling (explained in Subsection 2.6). Therefore, endogenous liquidity holdings due to insufficient liquidity cushions seemingly do not explain payday responses. Additionally, if we think that people hold liquidity cushions because of the possibility of expense shocks, we would observe a gradual increase in spending shortly before the arrival of income payments (after all, the probability of encountering a big expense shock declines over time). However, we observe a dip in spending if anything. Nevertheless, to examine the question further, we now look at cash-holding responses to income payments. Figure 14 View largeDownload slide The effects of regular income arrival on spending by people who did not or did receive a large exogenous wealth shock in that month The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 14 View largeDownload slide The effects of regular income arrival on spending by people who did not or did receive a large exogenous wealth shock in that month The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.9 Examining cash-holding responses to income payments Given the difficulties of measuring liquidity constraints in the presence of liquidity cushions, we are interested in a different method that considers the potential existence of liquidity cushions. To this end, inspired by the corporate finance literature, we think about a measure of liquidity constraints derived from peoples’ demands for liquidity. The methodology follows the one in the influential paper by Almeida et al. (2004). In this paper, the authors develop a new test measuring the effect of financial constraints on corporate policies. This effect of financial constraints is captured by a firm’s propensity to save cash out of cash inflows. The authors hypothesize that constrained firms should have a positive cash-flow sensitivity of cash but that unconstrained firms’ cash savings should not be systematically related to cash flows. In the household context, we empirically assess households’ propensities to increase liquidity cushions after cash inflows and the ways in which this propensity is related to liquidity. If a household feels liquidity constrained (even if its “hard” liquidity constraint is not literally binding), it will try to increase its cash or liquidity cushion after cash inflows. In corporate finance, analyses of the effects of financial constraints on firm behavior and the manner in which firms implement financial management have a long tradition. The paper by Almeida et al. (2004) states that firms want to have a liquid balance sheet to undertake valuable projects when they arise. However, if a firm has unrestricted access to external capital—that is, if a firm is financially unconstrained—there is no need to safeguard against future investment needs; thus, corporate liquidity becomes irrelevant. In contrast, when a firm faces financing frictions, liquidity management is a key issue for corporate policy. Thus, there exists a link between financial constraints and corporate liquidity demand, which has been ignored by the prior literature focusing on corporate investment demand Fazzari et al. 1988. Accordingly, the authors examine the influence of financing frictions on corporate investment by comparing the empirical sensitivity of investment to cash flow across groups of firms that are sorted by a proxy for financial constraints. Follow-up research, however, has identified several problems with that strategy regarding the theoretical and empirical robustness of the implications. In a household context, the study by Fazzari et al. (1988) may correspond with the analysis of household spending or investment in response to cash inflows. Households may spend or invest more in response to cash inflows because they are currently liquidity constrained. However, we find that people hold too much cash relative to the predictions of state-of-the-art economic models. In turn, we want to examine whether peoples’ payday responses stem from a concern about insufficient liquidity cushions or future liquidity constraints, which would be reflected in a high marginal propensity to hold on to cash or liquidity. To formalize these ideas, Figure 16 shows the marginal propensities to hold on to cash implied by three different simple models as well as additional model details. First, we consider a standard consumption-savings model without illiquid savings. In this model, the marginal propensity to hold on to cash (MPCash) equals one minus the marginal propensity to consume (MPCons), that is, MPCash = 1-MPCons, as the agent holds his entire lifetime wealth in cash. Because the MPCons in this model is always decreasing in income or liquidity, the MPCash will always be increasing. Furthermore, the MPCash is higher when the agent’s horizon increases, as he consumes only a small amount of his income and saves most of it. Figure 16 View largeDownload slide Marginal propensities to consume, save illiquidly, and save liquidly as implied by models with and without illiquid savings and future binding liquidity constraints We consider a three-period toy model with stochastic income, a liquid asset, and an illiquid asset. We follow Carroll (1997), who specify income $$Y_{t}$$ as log-normal and characterized by transitory shocks $$N_{t}^{T}$$. Doing so allows for a low probability $$p$$ of unemployment; that is, $$N_{t}^{T}=e^{s_{t}^{T}}$$ with $$s_{t}^{T}\sim N(\mu_{T},\sigma_{T}^{2})$$ or $$N_{t}^{T}=0$$ with probability $$p$$. We assume a probability of 8% for unemployment in line with our sample’s summary statistics. Additionally, we calibrate the model to a quarterly frequency and assume an annualized transitory income shock volatility of 0.2 following Carroll (1997) such that $$\sigma_{T}=0.2(\sqrt{0.25})$$. We also assume an annualized exponential discount factor of 0.98 such that $$\delta=0.98^{0.25}$$ and a simple power-utility specification $$u(C)=\frac{C^{1-\theta}}{1-\theta}$$ with $$\theta=2$$. The interest on illiquid savings is given by an annualized 2% such that $$r=0.02(0.25)$$ while liquid savings earn no interest. To simulate a long horizon in a three-period model, we assume that the last period is $$T=10$$ years ahead and no intermediate consumption takes place. Moreover, to illustrate the effects of liquidity constraints, we assume that the agent can collateralize only 30% of his illiquid asset holdings and can only access 30% of his transitory income shock in the second period right away. Thus, the agent expects to be liquidity constrained in the second period. In the last period, the agent consumes his liquid savings from the second period in addition to his illiquid savings. In turn, the model is solved by backward induction using standard numerical optimization techniques. The graphs depict the marginal propensities to consume, save illiquidly, and save liquidly in period 1. Figure 16 View largeDownload slide Marginal propensities to consume, save illiquidly, and save liquidly as implied by models with and without illiquid savings and future binding liquidity constraints We consider a three-period toy model with stochastic income, a liquid asset, and an illiquid asset. We follow Carroll (1997), who specify income $$Y_{t}$$ as log-normal and characterized by transitory shocks $$N_{t}^{T}$$. Doing so allows for a low probability $$p$$ of unemployment; that is, $$N_{t}^{T}=e^{s_{t}^{T}}$$ with $$s_{t}^{T}\sim N(\mu_{T},\sigma_{T}^{2})$$ or $$N_{t}^{T}=0$$ with probability $$p$$. We assume a probability of 8% for unemployment in line with our sample’s summary statistics. Additionally, we calibrate the model to a quarterly frequency and assume an annualized transitory income shock volatility of 0.2 following Carroll (1997) such that $$\sigma_{T}=0.2(\sqrt{0.25})$$. We also assume an annualized exponential discount factor of 0.98 such that $$\delta=0.98^{0.25}$$ and a simple power-utility specification $$u(C)=\frac{C^{1-\theta}}{1-\theta}$$ with $$\theta=2$$. The interest on illiquid savings is given by an annualized 2% such that $$r=0.02(0.25)$$ while liquid savings earn no interest. To simulate a long horizon in a three-period model, we assume that the last period is $$T=10$$ years ahead and no intermediate consumption takes place. Moreover, to illustrate the effects of liquidity constraints, we assume that the agent can collateralize only 30% of his illiquid asset holdings and can only access 30% of his transitory income shock in the second period right away. Thus, the agent expects to be liquidity constrained in the second period. In the last period, the agent consumes his liquid savings from the second period in addition to his illiquid savings. In turn, the model is solved by backward induction using standard numerical optimization techniques. The graphs depict the marginal propensities to consume, save illiquidly, and save liquidly in period 1. Second, we consider a consumption-savings model in which the agent can save in a liquid or an illiquid asset that pays higher interest. In such a model, the MPCash may be either increasing or decreasing in liquidity or income because the MPCash equals one minus the MPCons minus the marginal propensity to invest in the illiquid asset (MPIllInv); that is, MPCash = 1-MPCons-MPIllInv. While the MPCons is always decreasing in liquidity, the MPIllInv is increasing, which implies that the MPCash is either increasing or decreasing. However, the MPCash is always small like in the model of Kaplan and Violante (2014b), because agents have little reason to hold cash or liquid savings. If one would add transitory income shocks to the Kaplan et al. (2014) in the spirit of the buffer stock model of Carroll (2001) then agents would hold more liquidity but their marginal propensity to consume (MPC) out of fiscal stimulus payments would be low again. Therefore, we analyze a model with both transitory shocks and expected liquidity constraints to increase the MPC out of fiscal stimulus payments. One way to add future liquidity constraints is to assume that the agent receives news about income shocks in the future but that he or she will not be able to consume that income immediately. Frequently binding future liquidity constraints generate the equivalent prediction from the corporate finance literature: an MPCash that is decreasing in liquidity. This theoretical result is robust to different assumptions about the environment, such as social security payments. In fact, the prediction that the MPCash is decreasing in liquidity in anticipation of liquidity constraints appears to hold quite generally: suppose people have some type of threshold rule for how much rollover debt they allow themselves to hold. In that case, they would display a high marginal propensity to hold on to cash when they are close to that threshold. Because we include individual fixed effects, we can pick up heterogeneity along the lines of where people set their rules. Therefore, we only utilize variations from comparing people with low or high liquidity holdings to themselves. In the high-frequency consumption setting that we consider, illiquid savings could be interpreted as purchasing bulk consumption goods for instance. Figure 17 displays peoples’ cash-holding responses to regular and irregular income payments for three terciles of liquidity. We can see that less liquidity-constrained people have more pronounced cash-holding responses than more liquidity-constrained people. Moreover, cash responses are larger than spending responses. Both of these findings are predicted by a standard consumption-savings model. Thus, we conclude that cash responses do not seem to indicate the presence of illiquid savings, future liquidity constraints, or insufficient liquidity cushions. Instead of sorting people cross-sectionally, we could again analyze individual behavior for liquidity terciles relative to own histories of liquidity holdings. While the responses are somewhat flatter in this case, we do not find a decreasing pattern in the responses either. Figure 17 View largeDownload slide The effects of regular and irregular income arrival on liquidity by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction liquidity deviates from average liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances). Figure 17 View largeDownload slide The effects of regular and irregular income arrival on liquidity by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction liquidity deviates from average liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances). Even for deciles, all of the pictures show an increasing relationship between the propensity to hold on to cash and liquidity constraints as well as a very high propensity to hold on to cash, one that is much higher than the propensity to consume. We again use indicator variables for income payments to alleviate potential endogeneity concerns, but we can also estimate the MPCash directly and obtain the same relationship with liquidity. These findings are thus consistent with the standard consumption-savings model without illiquid savings. However, this model is not consistent with a high marginal propensity to consume out of transitory income shocks. We thus conclude that neither current nor future liquidity constraints can account for the observed payday responses to income payments. People can reduce overdraft limits relatively easily, while any credit limit increases have to be approved by the bank. Examining changes in overdraft limits around paydays yields very interesting results. Figure A.7 shows that people with less liquidity tend to reduce their overdraft limits around paydays, whereas people with high liquidity do not engage in such behavior. In itself, this is evidence against liquidity constraints being a problem in our sample and points toward the existence of overconsumption problems. After all, standard economic theory predicts that people should never reduce their limits, as borrowing opportunities are always weakly welfare increasing. However, we clearly see that less liquid people tend to reduce their limits after paydays. A potential explanation for this tendency is that people want to restrict their future selves from borrowing or that they want to reduce their mental borrowing accounts. To ensure that the documented increasing payday liquidity responses do not stem from low-liquidity peoples’ tendencies to reduce their limits after paydays, we also examine peoples’ balances—that is, their checking and savings account balance minus their credit balance—in Figure A.8. We again observe high and increasing responses that are consistent with a model without illiquid savings or future binding liquidity constraints. While the Icelandic financial crisis undoubtedly affected people, we believe that our qualitative results do not depend on the financial crisis or are otherwise country-specific. Even if the crisis had an effect on cash or liquidity holdings (instead of investments into durable goods, houses, cars or stocks and bonds) in Iceland today, we firmly believe we would be left with the same puzzling observation: significant payday responses in the presence of substantial liquidity. Iceland recovered very quickly and experienced high GDP growth and low unemployment during our sample period. Moreover, we include individual and month-by-year effects controlling for all individual (un)observables as well as any slow-moving trends. Additionally, we redid the entire analysis restricting ourselves to the year 2015 and do not find any differences. The OECD Economic Survey of Iceland from June 2011 states that the economic contraction and rise in unemployment appear to have been stopped by late 2010 with growth under way in mid-2011. The Icelandic government was successfully able to raise $${\$}$$1 billion with a bond issue in June 2011, indicating that international investors have given the government and the new banking system a clean bill of health. By mid-2012, Iceland was regarded as a recovery success story with 2 years of economic growth and unemployment down to 6.3%. Moreover, we find quantitatively similar payday responses (around 50% for the average household) to Gelman et al. (2014), who use U.S. data of the same kind. Of course, one has to keep in mind that there are large institutional differences between the United States and Iceland, as well as differences in average income, liquidity, and demographics. All of these differences make the magnitudes not perfectly comparable. 3. Conclusion We use data from a financial account aggregation provider in Iceland to evaluate whether spending or consumption results from an intertemporal optimization problem and is thus independent of income. The spending and income data are characterized by outstanding accuracy and comprehensiveness because of Icelanders’ nearly exclusive use of electronic payments. In line with previous studies, we find significant spending responses to the receipt of regular and irregular income on paydays. Moreover, we are in a unique position to simultaneously analyze credit and overdraft balances and limits at the same high frequency. This allows us to convincingly rule out liquidity constraints as an explanation. In contrast to previous studies, we thus argue that hand-to-mouth behavior is not limited to liquidity-constrained households, because we show that non-liquidity-constrained households exhibit hand-to-mouth behavior through various measures of liquidity constraints: balances and credit limits, spending on discretionary goods and services, and spending immediately before income payments. Overall, less than 3% of people have less than 1 day of average spending left in liquidity before their paydays. Because people may either hold liquidity cushions or save for foreseen expenses, we also examine cash-holding responses to income payments, inspired by the corporate finance literature. We notice that a model with liquid and illiquid savings and future liquidity constraints makes a joint prediction about the marginal propensity to consume and the marginal propensity to hold on to cash: both are decreasing in liquidity. We test this joint prediction in our data, however we do not find evidence for it. Because the cash-holding responses are most consistent with the standard consumption-savings problem without illiquid savings or future binding liquidity constraints, we argue that the evidence is not consistent with either present or future liquidity constraints. Moreover, our findings highlight a general difficulty to measure liquidity constraints. To determine whether people with liquidity cushions and rollover debt are liquidity constrained, however, is of great importance: after all, liquidity constraints are not equivalent to low financial resources. For policy purposes, liquidity constraints call for expanding credit, whereas low resources due to overconsumption problems call for restricting credit. The latter measure is also supported by our finding that low-liquidity households tend to voluntarily reduce their overdraft limits around paydays. While we have no way of telling whether the excess consumption on paydays is due to a time-inconsistent overconsumption problem, it seems to be generally believed nowadays that the extent of credit card borrowing in the United States (and Icelanders hold similar amounts of credit) is not consistent with standard models and must be explained by time-inconsistent overconsumption and the use of illiquid savings as a commitment device Laibson et al. 2003, Carroll 2001. We thank Itay Goldstein and two anonymous referees. We also thank Charlie Calomiris, Gianluca Violante, Steffen Andersen, Jonathan Parker, Dan Silverman, Jeremy Tobacman, David Laibson, Xavier Gabaix, Shachar Kariv, Botond Koszegi, Steve Zeldes, Michael Woodford, Ted O’Donoghue, Wei Jiang, Dimitri Vayanos, Emir Kamenica, Jialan Wang, Adam Szeidl, Matthew Rabin, Sumit Agarwal, Paul Tetlock, Tom Chang, and Justin Sydnor for their insightful comments. We also thank the seminar and the conference participants at the Behavioral Economics Annual Meeting (BEAM) 2016, the CFPB research conference, TAU, AFA, University of British Columbia, Boston Fed, Imperial College London, UC Berkeley, University of Zurich, Carnegie Mellon University, Columbia University, Gerzensee, Goethe University Frankfurt, University of Lugano, Stockholm School of Economics, University of St. Gallen, Huntsman, University of Colorado Boulder, and University of Utah for all their constructive questions and remarks. Nathen Huang and Guangyu Wang provided excellent research assistance. We are indebted to Ágúst Schweitz Eriksson and Meniga for providing and helping with the data. An earlier draft of this paper was circulating under the title “The Only Day Better than Friday is Payday!” Appendix Figure A.1 View largeDownload slide The effects of paycheck arrival on necessary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.1 View largeDownload slide The effects of paycheck arrival on necessary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.2 View largeDownload slide The effects of paycheck arrival on discretionary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.2 View largeDownload slide The effects of paycheck arrival on discretionary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.3 View largeDownload slide The effects of regular income arrival on spending by people costs in banking fees, interest, and payday loans The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.3 View largeDownload slide The effects of regular income arrival on spending by people costs in banking fees, interest, and payday loans The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.4 View largeDownload slide The effects of regular and irregular income on spending by terciles of credit utilization The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.4 View largeDownload slide The effects of regular and irregular income on spending by terciles of credit utilization The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.5 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.5 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.6 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend on discretionary goods and services as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.6 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend on discretionary goods and services as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.7 View largeDownload slide The effects of regular and irregular payments on overdraft limits by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: how overdraft limits change. Figure A.7 View largeDownload slide The effects of regular and irregular payments on overdraft limits by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: how overdraft limits change. Figure A.8 View largeDownload slide The effects of regular income and salary arrival on cash minus credit balances by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction the balances deviate from their average. Figure A.8 View largeDownload slide The effects of regular income and salary arrival on cash minus credit balances by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction the balances deviate from their average. Footnotes 1 This is true for both the standard consumption-savings model Friedman 1957, Hall 1978 and the more recent “buffer-stock” versions Deaton 1991, Carroll 1997. 2 Examples of micro evidence on excess sensitivity are Parker (1999), Souleles (1999), Shapiro and Slemrod (2003a), Shapiro and Slemrod (2003b), Shapiro and Slemrod (2009), Johnson et al. (2006), Parker et al. (2013), and Broda and Parker (2014), as surveyed in Jappelli and Pistaferri (2010) and Fuchs-Schundeln and Hassan (2015). Macro evidence is provided by Campbell and Mankiw (1989) and Campbell and Mankiw (1990) in response to the seminal paper by Flavin (1981). 3 As additional evidence, we present large spending responses by people who recently received a large exogenous wealth shock due to a court ruling. 4 Cochrane (1989) and Krusell and Smith (1996) perform similar calculations in a representative agent environment. 5 More substantial welfare effects of excess sensitivity to consumption are documented by Ganong and Noel (2016) and Baker and Yannelis (2017). 6 We also confirm the finding of Parker (2014) that liquidity appears to be a very persistent household trait rather than the product of swings due to transitory income shocks, as predicted in the Kaplan and Violante (2014b) model. 7 All numbers stem from the U.S. Census Bureau’s American Community Survey (ACS) in 2015. 8 Iceland’s 2008 financial crisis was exacerbated by banks that borrowed in Japanese yen or Swiss francs to take advantage of lower interest rates and then repackaged the loans in krona before passing them on to clients. This exchange-rate indexation of loans meant that the total amounts owed in Icelandic krona varied according to its exchange rate against the currencies in which the loans were issued. Such loans had been aggressively promoted by Icelandic banks in previous years and left many diligent car and homeowners with debts greater than the original amount despite paying their bills every month. 9 Other studies examining fiscal stimulus payments are Johnson et al. (2006), Parker et al. (2013), Parker (2014), and Jappelli and Pistaferri (2014), as surveyed by Jappelli and Pistaferri (2010). 10 As we will explain below, we believe that the observed discrepancy between the model-predicted and observed liquidity holdings are not due to the Icelandic financial crisis or otherwise country-specific. The economy has been booming in the sample period and many households have large amounts of rollover debt (inefficient financial markets should restrict borrowing). Moreover, Iceland is characterized by well-functioning health care, social security, and unemployment insurance systems. Additionally, as mentioned, there are no minimum balance restrictions associated with bank accounts that cause people to hold liquidity or cash in Iceland. 11 Liquidity holdings are increased in a model with transitory income shocks in addition to permanent shocks, which are a standard feature of life-cycle models and for which much evidence exists Carroll refer to 2001, Carroll et al. refer to 1992. However, in that case, the model is not able to generate a high MPC out of fiscal stimulus payments any more as people hold enough liquidity to deal with transitory income shocks. 12 Strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). We then compute cash by subtracting credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. References Addoum, J. M., Delikouras, S., and Korniotis, G. M. License to spend: Consumption-income sensitivity and portfolio choice. Available at SSRN 2409209 , 2015. Agarwal, S. and Qian, W. Consumption and Debt Response to Unanticipated Income Shocks: Evidence from a Natural Experiment in Singapore. American Economic Review , 104 12: 4205– 4230, 2014. Google Scholar CrossRef Search ADS   Almeida, H., Campello, M., and Weisbach, M. The Cash Flow Sensitivity of Cash. Journal of Finance , 59 4: 1777– 1804, 2004. Google Scholar CrossRef Search ADS   Alvarez, F., Guiso, L., and Lippi, F. Durable Consumption and Asset Management with Transaction and Observation Costs. American Economic Review , 5 102: 2272– 2300, 2012. Google Scholar CrossRef Search ADS   Angeletos, G. M., Laibson, D., Repetto, A., and Weinberg, S. The Hyperbolic Consumption Model: Calibration, Simulation, and Empirical Evaluation. Journal of Economic Perspectives , 15 3: 47, 2001. Google Scholar CrossRef Search ADS   Attanasio, . and Pavoni, N. Risk Sharing in Private Information Models with Asset Accumulation: Explaining the Excess Smoothness of Consumption. Econometrica , 79 4: 1027– 1068, 2011. Google Scholar CrossRef Search ADS   Baker, S. R. Debt and the Consumption Response to Household Income Shocks. Mimeo , 2013. Baker, S. R. and Yannelis, C. Income changes and consumption: Evidence from the 2013 federal government shutdown. Review of Economic Dynamics , 23: 99– 124, 2017. Google Scholar CrossRef Search ADS   Broda, C. and Parker, J. The Economic Stimulus Payments of 2008 and the Aggregate Demand for Consumption. Mimeo , 2014. Browning, M. and Crossley, T. F. The Life-Cycle Model of Consumption and Saving. Journal of Economic Perspectives , 15 3: 3– 22, 2001. Google Scholar CrossRef Search ADS   Campbell, J. Y. and Mankiw, N. G. Consumption, Income and Interest Rates: Reinterpreting the Time Series Evidence. NBER Macroeconomics Annual 1989, Volume 4, NBER Chapters National Bureau of Economic Research, Inc. , pages 185– 246, 1989. Campbell, J. Y. and Mankiw, N. G. Permanent Income, Current Income, and Consumption. Journal of Business and Economic Statistics , 8 3: 265– 279, 1990. Carroll, C. D. Buffer Stock Saving and the Life Cycle/Permanent Income Hypothesis. Quarterly Journal of Economics , CXII: 1– 56, 1997. Google Scholar CrossRef Search ADS   Carroll, C. D. A theory of the consumption function, with and without liquidity constraints. The Journal of Economic Perspectives , 15 3: 23– 45, 2001. Google Scholar CrossRef Search ADS   Carroll, C. D., Hall, R. E., and Zeldes, S. P. The buffer-stock theory of saving: Some macroeconomic evidence. Brookings papers on economic activity , 1992 2: 61– 156, 1992. Google Scholar CrossRef Search ADS   Carvalho, L. S., Meier, S., and Wang, S. W. Poverty and economic decision-making: Evidence from changes in financial resources at payday. American Economic Review , 106 2: 260– 84, 2016. Google Scholar CrossRef Search ADS PubMed  Chetty, R. and Szeidl, A. Consumption Commitments and Risk Preferences. Quarterly Journal of Economics , 122 2: 831– 877, 2007. Google Scholar CrossRef Search ADS   Cochrane, J. H. The Sensitivity of Tests of the Intertemporal Allocation of Consumption to Near-Rational Alternatives. American Economic Review , 79 3: 909– 936, 1989. de Nicola, F. and Giné, X. How accurate are recall data? evidence from coastal india. Journal of Development Economics , 106 0: 52– 65, 2014. ISSN 0304-3878. http://dx.doi.org/10.1016/j.jdeveco.2013.08.008. URL http://www.sciencedirect.com/science/article/pii/S0304387813001235. Google Scholar CrossRef Search ADS   Deaton, A. Saving and Liquidity Constraints. Econometrica , 59: 1221– 1248, 1991. Google Scholar CrossRef Search ADS   Fazzari, S., Hubbard, G., and Petersen, B. Financing Constraints and Corporate Investment. Brookings Papers on Economic Activity , 1: 141– 206, 1988. Google Scholar CrossRef Search ADS   Flavin, M. The Adjustment of Consumption to Changing Expectations about Future Income. Journal of Political Economy , 89: 974– 1009, 1981. Google Scholar CrossRef Search ADS   Flavin, M. and Nakagawa, S. A Model of Housing in the Presence of Adjustment Costs: A Structural Interpretation of Habit Persistence. American Economic Review , 98 1: 474– 495, 2008. Google Scholar CrossRef Search ADS   Friedman, M. A Theory of the Consumption Function . Number frie57-1 in NBER Books. National Bureau of Economic Research, Inc, 1957. URL http://ideas.repec.org/b/nbr/nberbk/frie57-1.html. Fuchs-Schundeln, N. and Hassan, T. Natural Experiments in Macroeconomics. Working Paper , 2015. Gabaix, X. and Laibson, D. The 6D Bias and the Equity-Premium Puzzle. NBER Macroeconomics Annual , 16: 257– 312, 2002. Google Scholar CrossRef Search ADS   Ganong, P. and Noel, P. How does unemployment affect consumer spending? Technical report, Working paper , 2016. Gelman, M., Kariv, S., Shapiro, M. D., Silverman, D., and Tadelis, S. Harnessing naturally occurring data to measure the response of spending to income. Science , 345 6193: 212– 215, 2014. Google Scholar CrossRef Search ADS PubMed  Hall, R. E. Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence. Journal of Political Economy , 86 6: 971– 987, 1978. Google Scholar CrossRef Search ADS   Hastings, J. and Washington, E. The first of the month effect: Consumer behavior and store responses. American Economic Journal: Economic Policy , 2 2:pp. 142– 162, 2010. ISSN 19457731. URL http://www.jstor.org/stable/25760066. Google Scholar CrossRef Search ADS   Huffman, D. and Barenstein, M. A Monthly Struggle for Self-Control? Hyperbolic Discounting, Mental Accounting, and the Fall in Consumption Between Paydays. Working Paper IZA Bonn , 2005. Huntley, J. and Michelangeli, V. Can Tax Rebates Stimulate Spending in a Life-Cycle Model? American Economic Journal: Macroeconomics , 1 6: 162– 189, 2014. Google Scholar CrossRef Search ADS   Jappelli, T. and Pistaferri, L. The Consumption Response to Income Changes. Annual Review of Economics , 2: 479– 506, 2010. Google Scholar CrossRef Search ADS   Jappelli, T. and Pistaferri, L. Fiscal policy and mpc heterogeneity. American Economic Journal: Macroeconomics , 6 4, 2014. URL http://ideas.repec.org/p/sef/csefwp/325.html. Johnson, D., Parker, J., and Souleles, N. Household Expenditure and the Income Tax Rebates of 2001. American Economic Review , 96 5: 1598– 1610, 2006. Google Scholar CrossRef Search ADS   Kaplan, G. and Violante, G. A Tale of Two Stimulus Payments: 2001 vs 2008. American Economic Review (Papers and Proceedings) , 5 104: 162– 189, 2014a. Kaplan, G. and Violante, G. A model of the consumption response to fiscal stimulus payments. Econometrica , 82: 1199– 1239, 2014b. Google Scholar CrossRef Search ADS   Kaplan, G., Violante, G., and Weidner, J. The wealthy hand-to-mouth. Brookings Papers on Economic Activity , 2014. Karlan, D. and Zinman, J. Lying about borrowing. Journal of the European Economic Association , 6( 2/3):pp. 510– 521, 2008. ISSN 15424766. URL http://www.jstor.org/stable/40282660. Google Scholar CrossRef Search ADS   Koszegi, B. and Rabin, M. Reference-Dependent Consumption Plans. American Economic Review , 99 3: 909– 936, 2009. Google Scholar CrossRef Search ADS   Krusell, P. and Smith, A. Rules of Thumb in Macroeconomic Equilibrium: A Quantitative Analysis. Journal of Economic Dynamics and Control , 20 4: 527– 558, 1996. Google Scholar CrossRef Search ADS   Kuchler, T. Sticking To Your Plan: Empirical Evidence on the Role of Present Bias for Credit Card Debt Paydown. Working Paper , 2015. Kueng, L. Explaining Consumption Excess Sensitivity with Near-Rationality: Evidence from Large Predetermined Payments. Working Paper , 2015. Laibson, D., Repetto, A., and Tobacman, J. A Debt Puzzle. in Aghion, Philippe Frydman, Roman Stiglitz, Joseph and Woodford, Michael eds., Knowledge, Information, and Expectations in Modern Economics: In Honor of Edmund S. Phelps , Princeton: Princeton University Press, 2003. Laibson, D., Repetto, A., and Tobacman, J. Estimating Discount Functions with Consumption Choices over the Lifecycle. Working Paper , 2015. Mastrobuoni, G. and Weinberg, M. Heterogeneity in intra-monthly consumption patterns, self-control, and savings at retirement. American Economic Journal.Economic Policy , 1 2: 163– 189, 08 2009. URL http://search.proquest.com/docview/872053305?accountid=39039. Google Scholar CrossRef Search ADS   Pagel, M. Expectations-based Reference-Dependent Life-Cycle Consumption. Mimeo , 2013. Parker, J. A. The Reaction of Household Consumption to Predictable Changes in Social Security Taxes. American Economic Review , 89 4: 959– 973, 1999. Google Scholar CrossRef Search ADS   Parker, J. A. Why Don’t Households Smooth Consumption? Evidence from a 25 million dollar experiment. Working Paper , 2014. Parker, J. A., Souleles, N., Johnson, D., and McClelland, R. Consumer Spending and the Economic Stimulus Payments of 2008. American Economic Review , 103 6: 2530– 53, 2013. Google Scholar CrossRef Search ADS   Reis, R. Inattentive Consumers. Journal of Monetary Economics , 53: 1761– 1800, 2006. Google Scholar CrossRef Search ADS   Shapiro, J. Is there a daily discount rate? Evidence from the food stamp nutrition cycle. Journal of Public Economics , 89: 303– 325, 2005. Shapiro, M. and Slemrod, J. Consumer Response to Tax Rebates. American Economic Review , 93 1: 381– 396, 2003a. Google Scholar CrossRef Search ADS   Shapiro, M. and Slemrod, J. Did the 2001 Tax Rebate Stimulate Spending? Evidence from Taxpayer Surveys. Tax Policy and the Economy , 17: 83– 109, 2003b. Google Scholar CrossRef Search ADS   Shapiro, M. and Slemrod, J. Did the 2008 Tax Rebates Stimulate Spending? American Economic Review , 99 2: 374– 379, 2009. Google Scholar CrossRef Search ADS PubMed  Shapiro, M. D. and Slemrod, J. Consumer response to the timing of income: Evidence from a change in tax withholding. American Economic Review , 85 1: 274– 83, March 1995. URL http://ideas.repec.org/a/aea/aecrev/v85y1995i1p274-83.html. Sims, C. Implications of Rational Inattention. Journal of Monetary Economics , 50 3: 665– 690, 2003. Google Scholar CrossRef Search ADS   Souleles, N. The Response of Household Consumption to Income Tax Refunds. American Economic Review , 89 4: 947– 958, 1999. Google Scholar CrossRef Search ADS   Tutoni, A. Rationally Inattentive Consumption Choices. Review of Economic Dynamics , 16: 421– 439, 2010. Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. 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 The Review of Financial Studies Oxford University Press

The Liquid Hand-to-Mouth: Evidence from Personal Finance Management Software

Loading next page...
 
/lp/ou_press/the-liquid-hand-to-mouth-evidence-from-personal-finance-management-MmbA5mX2Te
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com.
ISSN
0893-9454
eISSN
1465-7368
D.O.I.
10.1093/rfs/hhy055
Publisher site
See Article on Publisher Site

Abstract

Abstract We use a very accurate panel of all individual spending, income, balances, and credit limits from a personal finance software to document spending responses to the arrival of both regular and irregular income. These payday responses are robust and homogeneous for all income and spending categories throughout the income distribution. Moreover, we find that few people hold little or no liquidity. We then analyze whether people hold liquidity cushions to cope with future liquidity constraints. However, we find that peoples’ responses are consistent with standard models without illiquid savings, in which neither present nor future liquidity constraints are frequently binding. Received May 31, 2016; editorial decision September 30, 2017 by Editor Itay Goldstein. Standard economic theory states that consumption should not respond to the timing of predictable changes in disposable income.1 However, a number of empirical studies report that consumption responds to disposable income or that it is “excessively sensitive” to income.2 This excess sensitivity and the mechanisms behind it are important for understanding the effectiveness of short-term stimulus payments among other policy prescriptions. Recent advances in the literature explain excess sensitivity with households’ financial structures. In the presence of illiquid savings, many households consume hand-to-mouth because they hold little or no liquid wealth Kaplan et al. 2014, Kaplan and Violante 2014b, Laibson et al. 2015. Using very accurate data on spending, income, balances, and credit limits, this paper shows that (1) spending is significantly excessively sensitive to income payments for almost all people; (2) less than 3% of people have less than 1 day of their average spending left in liquidity before receiving their paychecks; and (3) individual liquidity “cushions” are at least three times greater than predicted by the model of Kaplan and Violante (2014b). However, it is difficult to empirically determine whether individual liquidity cushions are great enough, because they depend on individual economic circumstances. This makes it difficult to infer whether people are liquidity constrained. To overcome this measurement problem, we analyze whether cash-holding responses indicate the presence of insufficient liquidity cushions and future liquidity constraints inspired by the cash-flow sensitivity of cash work in the corporate finance literature Almeida et al. 2004. However, we find that peoples’ cash-holding responses do not indicate the presence of insufficient liquidity cushions or future liquidity constraints. In line with previous studies, we start by documenting significant spending responses on paydays of both regular and irregular income. These payday responses are clean, robust, and homogeneous for all income and spending categories throughout the income distribution. When we split the sample into ten income deciles, we observe a monotonic decrease in the initial spending response from 70% to 40% above the average daily spending. We then undertake multiple steps to ensure that we do not pick up naturally coincident or coordinated timing of consumption and income such as restricting ourselves to subgroups of people (for instance, people whose income schedules do not coincide with typical patterns), subgroups of spending categories, or exogenous income arrivals. Moreover, we can directly measure liquidity constraints via balances and credit limits or the presence or absence of household spending on (discretionary) goods and services immediately before their payday. Sorting people into liquidity groups, cross-sectionally or relative to their own history of liquidity holdings, again results in payday responses that are decreasing but large even for the most liquid people. Moreover, almost all people hold a substantial amount of liquidity the morning of their paychecks. We thus conclude that the fraction of constrained households is too small to quantitatively generate the degree of excess sensitivity documented. Payday responses could be explained by our measure of liquidity constraints not capturing whether households actually “feel” liquidity constrained. More specifically, the measurement of liquidity constraints via balances and credit limits is not applicable if households hold cash or liquidity cushions either to cope with unforeseen expenses or to save for foreseen expenses. If households have insufficient liquidity cushions, they feel liquidity constrained because they worry about binding future liquidity constraints. Such insufficient liquidity cushions or potentially binding future liquidity constraints may explain payday responses even when present liquidity constraints are not binding because people want to wait for extra liquidity before they spend. To address this conjecture, we examine cash-holding responses to income payments. The basic idea is that people should have a high propensity to hold on to cash upon receiving income payments, if they are worried about binding liquidity constraints in the future. More formally, we consider three models: (1) a standard model in which people hold their lifetime savings in cash, such that the marginal propensity to hold on to cash is simply the reverse of the marginal propensity to consume; (2) a model with liquid and illiquid savings in which people optimally hold little or no cash; and (3) a model with liquid and illiquid savings in which future liquidity constraints bind frequently. The third model captures insufficient liquidity cushions and predicts a decreasing relationship between cash-holding responses and individual liquidity. This is a new testable prediction of future liquidity constraints or insufficient liquidity cushions. When we examine the empirical patterns of individual cash-holding responses to income payments, we find that cash-holding responses correspond to the first standard model, which assumes neither present nor future liquidity constraints and thus cannot explain the baseline high marginal propensities to consume out of income payments. Thus, we conclude that neither current nor future liquidity constraints (or “hard” and “soft” liquidity constraints) seem to explain payday responses.3 A natural question arises regarding the economic importance of understanding these payday responses. After all, the payday effects are small with around $${\$}$$30 of additional spending on paydays. The calculations of Browning and Crossley (2001) show that the utility loss from setting consumption equal to income (instead of smoothing it perfectly) is second order in a plausibly parameterized life-cycle buffer stock model.4 Thus, our documented payday effects only cause a negligible welfare loss in terms of imperfect consumption smoothing under standard assumptions. However, the standard model is clearly inconsistent with the payday responses we observe; therefore, its welfare predictions may not be the right ones to apply. But, even under other assumptions, welfare effects are probably small.5 Nevertheless, we still think that payday effects are important for two main reasons. First, fiscal policy makers want to understand not only the rate at which fiscal stimulus payments are consumed by households but also the mechanisms behind the effectiveness of tax rebates as short-term stimuli for consumption. Despite very comprehensive data, we cannot confirm the presence of hard or soft liquidity constraints as an explanation for payday responses.6 Moreover, we point to an important difficulty in measuring which people are liquidity constrained. To figure out whether people are liquidity constrained, as opposed to just having low liquid wealth, is of great importance: after all, liquidity constraints call for policy measures that expand credit versus low liquid wealth may be caused by overconsumption problems in which case credit should be restricted. In this context, we also document the interesting finding that less liquid people tend to decrease their overdraft limits around paydays, whereas more liquid people do not. Thus, instead of being liquidity constrained, people may prefer to restrict their access to credit because they have an overconsumption problem. Second, we document payday responses that are so clean and homogeneous throughout a population holding substantial liquidity that they teach us something about how people think about spending and income and appear to point toward a shortcoming in the way that we currently model economic behavior in a life-cycle consumption context. A powerful psychological reason to spend cash inflows appears to exist. Thus, people do not intertemporally optimize, but, instead, they use heuristics to decide how much to consume and save. In this paper, we remain agnostic about which environmental or preference-related theories drive hand-to-mouth behavior, and we assume that this behavior may be caused by any preference or cognitive, computational, and time limits of the household. However, we believe that our results call attention to an important issue: the lack of rigorous, portable, and generally-applicable models of such behavior. An early example of such a theory is Campbell and Mankiw (1989), who simply assume that a fraction of income goes to hand-to-mouth consumers who consume part of their disposable income each period. Beyond this approach, the only existing theory that rationalizes our findings is modeled in Addoum et al. (2015), who assume that peoples’ marginal utilities of consumption increase upon the arrival of income because they feel they have a license to spend. We follow Gelman et al. (2014), Baker (2013), Kuchler (2015), and Kueng (2015) and use data from a financial aggregation and service app; doing so overcomes the accuracy, scope, and frequency limitations of the existing data sources of consumption and income as it is derived from actual transactions and account balances. We follow Gelman et al. (2014) in documenting payday responses, but observe even cleaner and more homogeneous responses for all income levels and every income and spending category. This is because our data are from Iceland for the years 2011 to 2015 and exceptionally thorough with respect to capturing all income and spending. More specifically, (1) the income and spending data are precategorized (and the categorization is very thorough and accurate), (2) the app is marketed through banks and supplied for their customers (thus covering a fairly representative sample of the population), and (3) the data are basically free of one important shortcoming of all transaction-level data—the absence of cash transactions (in Iceland, consumers almost exclusively use electronic means of payment). Previous work on payday effects has restricted its attention to subpopulations. These papers document that expenditures and the caloric intake of poor households increase on payday Sims e.g., 2003, Huffman and Barenstein e.g., 2005, Shapiro e.g., 2005. More specifically, Sims (2003) and Mastrobuoni and Weinberg (2009) find that both consumption expenditures and consumption are higher in the week after Social Security checks are distributed than in the week before. Shapiro (2005) also rejects the exponential discounting model by showing that food stamp recipients consume 10% to 15% fewer calories the week before food stamps are disbursed. With respect to behavior and cognitive function around paydays, Carvalho et al. (2016) fail to find before-after payday differences in risk-taking, the quality of decision-making, the performance in cognitive function tasks, or in heuristic judgments. Other empirical papers that examine transitory payments to test the permanent income hypothesis include Shapiro and Slemrod (1995), Shapiro and Slemrod (2003a), Parker (1999), among many others. However, these studies on the share of hand-to-mouth consumers are based on surveys that make “following the money” of consumers difficult because respondents may have little incentive to answer the questions accurately, may not understand the wording of the questions, or may behave differently in practice and forget their reported behavior. Moreover, such measurement error or noise in the data generated by surveys can increase with the length of the recall period de Nicola and Giné 2014. Additionally, surveys can produce biased (rather than merely noisy) data if respondents have justification bias, concerns about surveyors sharing the information, or stigma about their consumption habits Karlan and Zinman 2008. Overall, the conclusions of this literature regarding liquidity constraints are very mixed. Shapiro and Slemrod (2009) document that poor households, which are arguably more likely to be liquidity constrained, did not spend most of the 2008 tax rebate as the fiscal stimulus package intended. Shapiro and Slemrod (1995) conclude that liquidity constraints do not motivate the spending behavior of the 43% of households that report that the timing of tax payments will affect their consumption. Souleles (1999) examines the responses in nondurable and durable consumption. The author finds that constrained households are more likely to spend their tax refunds on nondurable consumption; the picture is reversed for durable consumption. Thus, liquidity-unconstrained households are not overwithholding to force themselves to save up enough for durable consumption goods because they could easily undo any forced saving by drawing down their liquid assets. As noted by Kaplan and Violante (2014b), wealthy people may engage in hand-to-mouth behavior due to illiquid wealth. Recent theoretical examples of models with liquid and illiquid assets are Angeletos et al. (2001), Laibson et al. (2003), Flavin and Nakagawa (2008), Chetty and Szeidl (2007), Alvarez et al. (2012), Huntley and Michelangeli (2014), and Kaplan and Violante (2014a). Angeletos et al. (2001) and Laibson et al. (2003) show that households with hyperbolic-discounting preferences optimally decide to lock their wealth in the illiquid asset in order to cope with self-control problems. Kaplan and Violante (2014a) do not need to assume that households have hyperbolic-discounting preferences and still generate a high marginal propensity to consume out of transitory shocks. In a one-asset environment, Koszegi and Rabin (2009) show that, in an environment with little to no uncertainty, agents with reference-dependent preferences may consume entire windfall gains, and Pagel (2013) shows that the preferences also rationalize excess smoothness in consumption. Moreover, Reis (2006) assumes that agents face costs when processing information and thus optimally decide to update their consumption plans sporadically, resulting in excessively smooth consumption that is shown to matter in the aggregate by Gabaix and Laibson (2002). Additionally, Tutoni (2010) assumes that consumers are rationally inattentive, and Attanasio and Pavoni (2011) show that excessively smooth consumption results from incomplete consumption insurance due to a moral hazard problem. 1. Data and Summary Statistics 1.1 Data This paper exploits new data from Iceland generated by Meniga, a financial aggregation software provider to European banks and financial institutions. Meniga has become Europe’s leading financial management (PFM) provider. Meniga’s account aggregation platform allows bank customers to manage all their bank accounts and credit cards across multiple banks in one place by aggregating data from various sources (internal and external). Meniga’s financial feed documents consumers’ budgets in a social media style. Figure 1 displays screenshots of the app’s user interface. The first screenshot shows the background characteristics that the user provides; the second one shows transactions; the third one shows bank account information; and the fourth one shows a sample of accounts that can be added. Figure 1 View largeDownload slide Screenshots of the financial aggregation app Figure 1 View largeDownload slide Screenshots of the financial aggregation app In October 2014, the Icelandic population was 331,310, and 20% of Icelandic households were using the Meniga app. Because the app is marketed through banks and automatically supplied to customers using online banking, the sample of Icelandic users is fairly representative. Each day, the application automatically records all bank and credit card transactions (including descriptions as well as balances), overdraft limits, and credit limits. We use the entire de-identified population of active users in Iceland and the data derived from their records from 2011 to 2015. We perform the analysis on normalized and aggregated user-level data for different income and spending categories. Additionally, the app collects demographic information, such as age, gender, marital status, and postal code. Moreover, we can infer employment status, real estate ownership, and the presence of young children in the household from the data. We have the following regular income categories: child support, benefits, child benefits, interest income, invalidity benefits, parental leave, pension income, housing benefits, rental benefits, rental income, salaries, student loans, and unemployment benefits. In addition, we have the following irregular income categories: damages, grants, other income, insurance claims, investment transactions, reimbursements, tax rebates, and travel allowances. The spending categories are groceries, fuel, alcohol, ready-made food, home improvements, transportation, clothing and accessories, sports and activities, and pharmacies. We can observe expenditures on alcohol that is not purchased in bars or restaurants because a state-owned company, the State Alcohol and Tobacco Company, has a monopoly on the sale of alcohol in Iceland. We exclude all recurring spending such as rent and bill payments. 1.2 Summary statistics Table 1 displays summary statistics of the Icelandic users, including not only income and spending in U.S. dollars but also some demographic statistics. We can see that the average user is 40 years old; 15% of the users are pensioners; 50% are female; 20% have children; and 8% are unemployed. For comparison, Statistics Iceland reports that the average age in Iceland is 37 years; 12% of Icelanders are pensioners; 48% are female; 33% have children; and 6% are unemployed. Thus, our demographic statistics are representative for those of the overall Icelandic population. Moreover, our sample characteristics are very similar to U.S. data. The average age in the U.S. population is 38; the percentage of women in the United States is 51%; and the mean income in the U.S. population in 2015 dollars per adult member is $${\$}$$3,266. In our sample, the individual monthly mean income is $${\$}$$3,256.7 This is not the case for other studies using app data that observe a user population more likely than the overall population to be young, financially secure, male, and tech savvy. Table 1 Summary statistics    Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12     Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12  All income, salary, and spending numbers are in U.S. dollars. Parentheses indicate that data categories do not match perfectly. Table 1 Summary statistics    Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12     Mean  Standard deviation  Statistics Iceland  Monthly total income  3,256  3,531  3,606  Monthly salary  2,701  2,993  2,570  Monthly spending:           $$\quad$$ Total  1,315.1  1,224.3     $$\quad$$ Groceries  468.29  389.29  490  $$\quad$$ Fuel  235.88  258.77  (359)  $$\quad$$ Alcohol  61.75  121.43  85  $$\quad$$ Ready-made food  170.19  172.64  (252)  $$\quad$$ Home improvement  150.16  464.94  (229)  $$\quad$$ Transportations  58.33  700.06  66  $$\quad$$ Clothing and accessories  86.62  181.27  96  $$\quad$$ Sports and activities  44.29  148.41  (36)  $$\quad$$ Pharmacies  39.62  62.08  42  Age  40.6  11.5  37.2  Female  0.45  –  0.48  Unemployed  0.08  –  0.06  Parent  0.23  –  0.33  Pensioner  0.15  –  0.12  All income, salary, and spending numbers are in U.S. dollars. Parentheses indicate that data categories do not match perfectly. The representative national household expenditure survey conducted by Statistics Iceland also reports income and spending statistics. In Table 1, parentheses indicate when spending categories do not match perfectly with the data. We can see that the income and spending figures are remarkably similar for the categories that match well. Figures 2 to 4 show the distribution of regular, salary, and irregular income payments over the course of a month. Approximately 85% of the people in the sample were paid on a monthly basis, whereas the remaining people are paid on a more frequent basis. This variation allows us to also consider people who are paid on unusual schedules. Additionally, the irregular payments are distributed rather evenly over the course of the month. Figure 2 View largeDownload slide The distribution of regular income arrivals over a month Figure 2 View largeDownload slide The distribution of regular income arrivals over a month Figure 3 View largeDownload slide The distribution of paycheck arrivals over a month Figure 3 View largeDownload slide The distribution of paycheck arrivals over a month Figure 4 View largeDownload slide The distribution of irregular income arrivals over a month Figure 4 View largeDownload slide The distribution of irregular income arrivals over a month 2. Analysis In this study, we estimate payday effects by running the following regression   \begin{equation} \label{one} x_{it} = \sum\limits_{k=-7}^{7}\beta_{k}I_i(Paid_{t+k}) + \delta_{dow} + \phi_{wom} + \psi_{my} + \eta_i + \epsilon_{it} \end{equation} (1) where $$x_{it}$$ is the ratio of spending by individual $$i$$ to his or her average daily spending on date $$t$$, $$\delta_{dow}$$ are day-of-week fixed effects, $$\phi_{wom}$$ are week-of-the-month fixed effects, $$\psi_{my}$$ are month-by-year fixed effects, $$\eta_{i}$$ are individual fixed effects, and $$I_i(Paid_{t+k})$$ is an indicator that is equal to $$1$$ if $$i$$ receives a payment at time $$t+k$$ and that is equal to $$0$$ otherwise. The $$\beta_{k}$$ coefficients thus measure the fraction by which individual spending deviates from the average daily spending in the days surrounding the receipt of a payment. We use indicator variables for income payments to alleviate potential endogeneity concerns at the income level. The individual fixed effects control for all observable or unobservable individual characteristics. The day-of-week dummies capture within-week patterns for both income and spending. Beyond the day-of-week fixed effects, the week-of-the-month fixed effects control for some mechanical effects due to fixed expense cycles at the beginning of each month. Finally, the month-by-year dummies capture all slow-moving trends. Standard errors are clustered at the individual level. We will initially differentiate between the arrival of regular and irregular income and separate households into ten income deciles. All the following figures and regression tables display the $$\beta_{k}$$ coefficients for different spending categories, income categories, and sample splits. 2.1 Regular income payments Figure 5 displays the spending responses to regular income payments of households in ten different income deciles, as measured by their regular salaries. Both poor and rich households clearly respond to the receipt of their income, with the poorest households spending 70% more than they would on an average day and the richest households spending 40% more. Even for the richest households, we thus observe a surprisingly high spending response. Table 2 presents all regression results for four income quartiles and four types of spending. While grocery and fuel spending can be regarded as necessary, ready-made food (such as restaurants) and alcohol spending can be regarded as discretionary. Figures A.1 and A.2 in the Appendix separately display the spending responses to income for all necessary categories and all discretionary categories. People are equally inclined to spend on necessary and discretionary goods and services upon receiving their income. In terms of magnitudes in dollars, we observe an additional spending of approximately $${\$}$$30 on paydays. Figure 5 View largeDownload slide The effects of regular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day dummy: the deviation of spending from average daily spending. Figure 5 View largeDownload slide The effects of regular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day dummy: the deviation of spending from average daily spending. Table 2 The impact of payments on household spending by income quartiles    (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)     (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01. Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. The salary arrival responses are estimated by salary quartiles, whereas the responses to any payments, regular payments, and irregular payments are estimated by total income quartiles. The outcome is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. Table 2 The impact of payments on household spending by income quartiles    (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)     (1)  (2)  (3)  (4)  (5)     Total spending  Groceries  Fuel  RMF  Alcohol  A. First salary quartile  $$I(Payment_{it} > 0)$$  0.800***  0.679***  0.756***  0.560***  0.875***  (0.0088)  (0.0085)  (0.0124)  (0.0084)  (0.0162)  $$I(Regular payment_{it} > 0)$$  0.880***  0.768***  0.880***  0.599***  0.992***  (0.0135)  (0.0126)  (0.0195)  (0.0120)  (0.0243)  $$I(Irregular payment_{it} > 0)$$  0.727***  0.595***  0.653***  0.507***  0.814***  (0.0099)  (0.0097)  (0.0132)  (0.0100)  (0.0201)  $$I(Salary_{it} > 0)$$  0.815***  0.722***  0.825***  0.548***  0.862***  (0.0151)  (0.0147)  (0.0231)  (0.0140)  (0.0307)  B. Second salary quartile  $$I(Payment_{it} > 0)$$  0.529***  0.434***  0.508***  0.318***  0.627***  (0.0099)  (0.0091)  (0.0168)  (0.0094)  (0.0205)  $$I(Regular payment_{it} > 0)$$  0.590***  0.516***  0.649***  0.332***  0.750***  (0.0142)  (0.0130)  (0.0269)  (0.0123)  (0.0282)  $$I(Irregular payment_{it} > 0)$$  0.464***  0.348***  0.377***  0.287***  0.533***  (0.0110)  (0.0099)  (0.0159)  (0.0117)  (0.0265)  $$I(Salary_{it} > 0)$$  0.560***  0.457***  0.654***  0.283***  0.678***  (0.0134)  (0.0121)  (0.0261)  (0.0128)  (0.0309)  C. Third salary quartile  $$I(Payment_{it} > 0)$$  0.430***  0.314***  0.429***  0.241***  0.522***  (0.0104)  (0.0087)  (0.0176)  (0.0090)  (0.0202)  $$I(Regular payment_{it} > 0)$$  0.436***  0.358***  0.544***  0.248***  0.572***  (0.0136)  (0.0119)  (0.0272)  (0.0120)  (0.0275)  $$I(Irregular payment_{it} > 0)$$  0.418***  0.260***  0.339***  0.225***  0.474***  (0.0130)  (0.0101)  (0.0182)  (0.0121)  (0.0270)  $$I(Salary_{it} > 0)$$  0.448***  0.364***  0.529***  0.210***  0.580***  (0.0116)  (0.0104)  (0.0234)  (0.0114)  (0.0287)  D. Fourth salary quartile  $$I(Payment_{it} > 0)$$  0.350***  0.230***  0.418***  0.155***  0.430***  (0.0111)  (0.0092)  (0.0229)  (0.0096)  (0.0219)  $$I(Regular payment_{it} > 0)$$  0.343***  0.245***  0.530***  0.139***  0.467***  (0.0148)  (0.0121)  (0.0356)  (0.0127)  (0.0294)  $$I(Irregular payment_{it} > 0)$$  0.348***  0.208***  0.294***  0.160***  0.372***  (0.0152)  (0.0112)  (0.0206)  (0.0130)  (0.0301)  $$I(Salary_{it} > 0)$$  0.405***  0.318***  0.513***  0.184***  0.690***  (0.0106)  (0.0097)  (0.0231)  (0.0105)  (0.0259)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01. Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. The salary arrival responses are estimated by salary quartiles, whereas the responses to any payments, regular payments, and irregular payments are estimated by total income quartiles. The outcome is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. There is no change in permanent income on paydays and there is no new information because paydays are perfectly predictable. While a buffer stock model can potentially explain sensitivity to surprising large payments or changes in permanent income, it cannot explain sensitivity to regular paydays. Thus, these payday responses are inconsistent with standard models of consumption and savings. Although we focus on nonrecurring spending (exclude all rent and bill payments) and control for day-of-week and week-of-month fixed effects, this spending response to regular income might stem from the coincident timing of regular income and irregular spending. Therefore, we will now examine irregular income. 2.2 Irregular income payments Figure 6 displays the spending responses to irregular income payments of households in ten different income deciles, as measured by their regular salaries. Again, we observe both poor and rich households responding to the receipt of their income, and poor households’ spending responses are somewhat more pronounced. Again, even for rich households, the spending response on payday is large and significant, at approximately 40%. Thus, we do not conclude that the bulk of the spending responses to income or the excess sensitivity of consumption is due to poor households or the coincident timing of regular income and spending, as proposed in Gelman et al. (2014). Figure 6 View largeDownload slide The effects of irregular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 6 View largeDownload slide The effects of irregular income on spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.3 Heterogeneity We are interested in the question of whether the payday responses are prevalent for a large fraction of the population or are being driven by a small fraction of the population. To do so, we run a regression for each individual in four income quartiles and display their individual payday coefficients in Figure 7. Approximately 22% of people have a payday coefficient equal to zero. A greater number of positive coefficients outweigh some negative coefficients, leading to an average coefficient of approximately 0.6 for the lowest quartile and 0.4 for the highest quartile. Therefore, at least half of the population, rather than a small fraction of the population, displays significantly positive and large payday responses. Figure 7 View largeDownload slide The distribution of payday coefficients for people by income and salary quartiles The payday coefficient is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. Figure 7 View largeDownload slide The distribution of payday coefficients for people by income and salary quartiles The payday coefficient is the fraction by which individual spending in each category deviates from average daily spending on the day of income arrival. 2.4 Intensive versus extensive spending We are interested in the question of whether payday responses are an intensive or extensive phenomenon in the sense of people spending more when they go shopping or making an additional shopping trip. In Table 3, we display the results of regressions that estimate how much more likely people are to buy in different categories, such as groceries, fuel, and restaurants, on their payday. For instance, people are 11% more likely to go on any shopping trip on paydays. In a second set of regressions, we then compare how much they spend if they shop on a payday relative to any other day. People spend $${\$}$$21 more on all shopping trips on their paydays. Because people spend, on average, $${\$}$$50 every day on nonrecurring consumption and approximately $${\$}$$80 on paydays, this $${\$}$$21 increase corresponds to approximately 80% of the increase in spending on paydays ($${\$}$$30). Thus, people are more likely to go shopping and, if they go shopping, they spend more than they would on a shopping day when they are paid. Table 3 Intensive and extensive spending reaction    (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)     (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01 Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. Panel A shows the effect on the probability of buying the goods under consideration on payday. Panel B compares the expenditure on shopping days when consumers are paid to those when they do not. Table 3 Intensive and extensive spending reaction    (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)     (1)  (2)  (3)  (4)  (5)  (6)  (7)  (8)  (9)  (10)  (11)  Category:  Any  Groceries  Fuel  Alcohol  Ready-made food  Home improvements  Home security  Vehicles  Clothing and accessories  Sports and activities  Pharmacies  A. Extensive  Payment  0.121***  0.082***  0.052***  0.021***  0.058***  0.020***  0.001***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0007)  (0.0006)  (0.0003)  (0.0006)  (0.0003)  (0.0001)  (0.0002)  (0.0002)  (0.0002)  (0.0003)  Regular  0.107***  0.075***  0.047***  0.023***  0.050***  0.017***  0.002***  0.015***  0.010***  0.009***  0.015***     (0.0009)  (0.0008)  (0.0007)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  Irregular  0.122***  0.081***  0.053***  0.018***  0.059***  0.021***  0.001***  0.014***  0.010***  0.008***  0.014***     (0.0012)  (0.0009)  (0.0008)  (0.0004)  (0.0009)  (0.0003)  (0.0001)  (0.0003)  (0.0002)  (0.0002)  (0.0003)  Salary  0.103***  0.069***  0.046***  0.024***  0.049***  0.015***  0.002***  0.015***  0.009***  0.008***  0.013***     (0.0010)  (0.0009)  (0.0008)  (0.0004)  (0.0007)  (0.0003)  (0.0002)  (0.0003)  (0.0002)  (0.0003)  (0.0003)  B. Intensive  Payment  20.2***  6.0***  9.0***  4.4***  1.6***  15.2***  2.1***  50.7***  4.5***  7.9***  1.6***     (0.4)  (0.1)  (0.3)  (0.2)  (0.1)  (0.9)  (1.9)  (4.8)  (0.5)  (0.7)  (0.1)  Regular  19.2***  7.6***  11.8***  3.8***  1.6***  6.7***  3.4***  20.3***  4.3***  3.5***  1.8***     (0.5)  (0.1)  (0.5)  (0.2)  (0.1)  (1.0)  (3.4)  (5.1)  (0.6)  (0.8)  (0.1)  Irregular  20.2 ***  4.3***  6.3***  4.8***  1.6***  19.7***  0.1***  70.8***  3.9 ***  10.8***  1.4***     (0.6)  (0.1)  (0.3)  (0.3)  (0.1)  (1.2)  (0.5)  (7.0)  (0.6)  (0.9)  (0.1)  Salary  18.0***  6.9***  12.2***  3.9***  1.5***  7.2***  4.1***  13.8***  4.2***  3.7***  1.4***     (0.5)  (0.2)  (0.5)  (0.2)  (0.1)  (1.1)  (3.9)  (5.0)  (0.6)  (0.9)  (0.1)  * p$$<$$0.1, ** p$$<$$0.05, and *** p$$<$$0.01 Standard errors are clustered at the individual level and are within parentheses. Each entry is a separate regression. Panel A shows the effect on the probability of buying the goods under consideration on payday. Panel B compares the expenditure on shopping days when consumers are paid to those when they do not. 2.5 Financial sophistication We observe a number of potential proxies for financial sophistication: age, pensions, employment, benefit payments, number of logins, voluntary reductions of overdraft limits, banking fees paid, payday loans, simultaneous savings and overdraft debt, large checking account balances that do not pay interest, and whether people link their spouse. We first examine simultaneous savings and overdraft debt, which can be considered a mistake because overdrafts cost more interest than savings yield. Figure 8 shows the spending responses of people sorted by how much interest is lost by holding overdrafts and savings simultaneously. People who lose less have less pronounced spending responses than those who lose more. A possible reason for this result is that wealthier people lose more by having savings and overdrafts simultaneously. The spending responses of people sorted by our other measures of financial sophistication show a similar picture. For instance, the same is true when we sort people by a summary measure of how much they lose in banking fees, interest, and payday loans in Figure A.3. We thus conclude that our measures of financial sophistication do not predict peoples’ propensities for payday responses well. Figure 8 View largeDownload slide The effects of regular income arrival on spending by amount lost due to holding overdrafts and savings simultaneously The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 8 View largeDownload slide The effects of regular income arrival on spending by amount lost due to holding overdrafts and savings simultaneously The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.6 Other explanations for payday effects Thus far we cannot pinpoint an explanation for payday responses as we fail to find any population group or income or spending category for which the payday responses are absent or less pronounced. Because irregular income responses may be unanticipated, payday responses are not necessarily inconsistent with the standard model. Nevertheless, confirming the existence of payday responses for irregular income rules out many alternative explanations for payday responses to regular income, such as naturally recurring spending and income or coordination stories that would not be picked up by our fixed effects. We will now review all potential other explanations. We perform a number of additional exercises to figure out potential mechanisms behind payday responses. First, we take a closer look at the characteristics of the people in the lowest income decile because their spending responses appear to be slightly different from the other income deciles. However, we do not observe unusual characteristics. For instance, the mean income of people in the lowest income decile is approximately $${\$}$$750 and their mean age is 34 years, while the second decile’s mean income and age are approximately $${\$}$$1,000 and 34 years, respectively. Second, we examine the responses for the ready-made-food category because such spending is instantly consumed. Third, we examine only people who are paid on unusual days. In doing so, we also ensure that we observe payday responses for all categories and not simply categories that are likely to be consumed alongside coworkers who are paid on the same unusual paydays (such as restaurants and alcohol). Fourth, we examine only tax rebates and exogenous wealth shocks (described below) to control for potential endogeneity of income. Additionally, we can restrict our sample to income payments that arrive at a certain day of the month and include day-of-month fixed effects, which obtains exogenous variation in the arrival of payments due to weekends and holidays. Fifth, we examine people who have linked their spouse to ensure that the responses are not driven by intra-household bargaining. Sixth, any price-discriminatory response of firms does not explain the magnitude of the observed effects Hastings and Washington 2010 and does not apply to people with unusual paydays or irregular income. Seventh, we sort people by how often they log into the app to ensure that app usage is unrelated to payday responses. Moreover, we can restrict the sample to pre-2014, the year when the smartphone app was introduced (before people could access only via a desktop or laptop computer), but we do not find any differences in payday responses. All of these exercises yield similar payday responses. Thus, we conclude that spending responses to income payments constitute a very robust phenomenon, which is cleanly estimated and prevalent throughout the population. Given the robustness of these payday responses, we think that attempting to better understand what is driving them is a valuable exercise. People plausibly incur adjustment costs when they spend. Nevertheless, we do not believe that adjustment costs can explain our payday effects for the following reasons: (1) Plausible magnitudes of adjustment cost depend on the category of consumption under consideration. If one thinks of housing or car expenses, high adjustment costs are plausible. However, plausible tangible or intangible costs to changing nondurable consumption are probably low. Moreover, these costs may vary with the consumption category, for instance, dependent on whether these are durables or nondurables. However, the payday effects are very similar across the different consumption categories. This suggests to us that adjustment costs are eaten up by the calendar fixed effects and the payday effects pick up a residual but strong psychology to spend cash inflows. (2) The payday effects do not depend in a systematic way on the size of payments. For instance, irregular payments are much smaller than regular payments but yield the same magnitude in payday responses. (3) We observe sizable and significant payday effects at both the intensive and extensive margin. The extensive margin is difficult to reconcile with fixed adjustment costs. Nevertheless, if people hold plenty of cash or liquidity (as we will show below), they could have done any shopping trip a day early rather than the day they are paid. All income payments are made via direct deposit in Iceland, and we do not observe any check transactions. Moreover, just one clearing house in the country exists, so all transactions are recorded without delay. (4) More generally, if people hold plenty of cash or liquidity, they would not react to a perfectly predictable regular income payment, an irregular income payment, or an exogenous income payment (such as a lottery payment). However, we find the same magnitudes in payday responses for all of these income categories. Using large exogenous wealth shocks, we can also estimate the marginal propensity to consume in response to fiscal stimulus payments of our sample population. The shocks that we use originate from a debt relief ruling that resulted in large repayments from banks to thousands of Icelandic households holding foreign-indexed debt. In this natural experiment, Icelandic lenders had to pay out as much as $${\$}$$4.3 billion, that is, one-third of the economy’s gross domestic product (GDP), after a court found that some foreign loans were illegal. These foreign loans were the largest single loan category of the banks, with a value of approximately $${\$}$$7.2 billion.After the financial crisis, the Icelandic Supreme Court ruled on June 16, 2010, that loans indexed to foreign currency rates were illegal in three cases involving private car loans and a corporate property loan. This decision meant that borrowers with such loans were only obliged to repay the principal in Icelandic krona, making the lenders liable for currency losses of approximately $${\$}$$28 billion in debt because the krona’s value against the Japanese yen and Swiss franc declined by one-third since September 2008.8 After the debt-relief ruling, banks had to repay their customers, which we consider to be exogenous wealth shocks. We obtain marginal propensities to consume that are perfectly in line with existing papers, such as Agarwal and Qian (2014), who analyze Singaporean consumers’ responses to a fiscal stimulus announcement and payout, and Kueng (2015), who uses payments originating from the Alaska Permanent Fund.9 We thus conclude that Icelanders do not exhibit larger spending responses to income windfalls than do people in other countries. 2.7 Comparison to Gelman et al. (2014) We now briefly discuss the differences between the payday effects of Gelman et al. (2014) and ours. Four main differences stand out: First, our payday response is concentrated on the payday. In Gelman et al. (2014), there appears a second larger spike a few days after the payday. The reason is probably the absence of check transactions in the Icelandic data. In Iceland, everybody is paid by direct deposit. In contrast, many people in the United States receive paper paychecks and that appears to cause a spending boost and then another one once the paycheck is deposited. Second, we observe payday responses for many income categories whereas Gelman et al. (2014) restrict themselves to regular paycheck and Social Security payments. Thus, we are able to only look at people who are paid on unusual days and document payday responses to irregular payments. Third, we observe payday responses for many spending categories, whereas Gelman et al. (2014) restrict themselves to recurring, nonrecurring, and coffee shop and fast food spending. The coffee shop and fast food measure is identified using keywords from the transaction descriptions. In contrast, when we receive the data, it is already categorized by a three-tiered approach: system rules and user and community rules. The system rules are applied in instances where codes from the transactions systems clearly indicate the type of transaction being categorized. For example, when transactions in the Icelandic banking system contain the value “04” in a field named “Text key” the payer has indicated payment of salary. User rules apply if no system rules are in place and when a user repeatedly categorizes transactions with certain text or code attributes to a specific category. In those instances the system will automatically create a rule which is applied to all further such transactions. If neither system rules nor user rules apply, the system can sometimes detect identical categorization rules from multiple users which allows for the generation of a community rule. While Gelman et al. (2014) do not observe payday responses to coffee shop and fast food spending, we observe responses of similar magnitude for restaurant spending, which isolates a discretionary, nondurable, and divisible form of spending. To further understand why our payday responses appear to be cleaner, we reran the regressions using only 300 consecutive days in 2012 and 2013, as Gelman et al. (2014) use, but we find very similar responses. We thus conclude that our categorization and measurement of spending and income make a difference. The Diary of Consumer Payment Choice (DCPC), conducted in October 2012 by the Boston, Richmond, and San Francisco Federal Reserve Banks, shows that cash makes up the single largest share of consumer transaction activity at 40%, followed by debit cards at 25%, and credit cards at 17%. Electronic methods (online banking bill pay and bank account number payments) account for 7%, while checks make up 7%. All other payments represent less than 5% of monthly transaction activity, with text and mobile payments barely registering at less than one-half of 1%. By value, cash accounts for 14% of total consumer transaction activity, while electronic methods make up 27%, and checks 19%. These findings suggest that cash is used quite often, but primarily for low-value transactions such as coffee shop and fast food spending. Overall, just from eyeballing the responses by Gelman et al. (2014), we feel that our responses are considerably more clean, homogeneous, and robust. 2.8 Examining liquidity constraints Thus far, our results suggest that hand-to-mouth behavior is prevalent across all income groups, which casts doubt on liquidity constraints as the only explanation for such behavior. To further establish that liquidity-constrained households are not alone in exhibiting spending responses, we now examine different measures of liquidity constraints. The first is simply cash holdings in checking and savings accounts. The second concerns liquidity or maximum borrowing capacity and equals cash holdings in checking and savings accounts minus credit card balances plus credit limits and overdraft limits. The third is credit utilization and the fourth is spending on (discretionary) goods and services immediately before income payments. The consumer credit landscape in Iceland is slightly different from the United States. Most importantly, rolling over credit card debt is possible, but people in Iceland seldom do so. Instead Icelanders typically pay off the balance in full at the end of the month using their current account. An overdraft occurs when withdrawals from a current account exceed the available balance. This means that the balance is negative and hence that the bank is providing credit to the account holder and interest is charged at the agreed rate. Virtually all current accounts in Iceland offer a pre-agreed overdraft facility, the size of which is based upon affordability and credit history. This overdraft facility can be used at any time without consulting the bank and can be maintained indefinitely (subject to ad hoc reviews). Although an overdraft facility is authorized, technically the money is repayable on demand by the bank. In reality, this is a rare occurrence as the overdrafts are profitable for the bank and expensive for the customer. Moreover, unlike those in the United States, current accounts do not feature minimum balance requirements. All liquidity measures are normalized by each individual’s average spending; that is, we measure individual liquidity in individual average consumption days rather than absolute liquidity. Figures 9, 10, and A.4 compare the spending responses to regular and irregular income for three terciles of our standard measures of liquidity: cash holdings (checking and savings account balances), liquidity (overdraft and credit limits plus checking and savings accounts balances minus credit card balances), and credit utilization. Additionally, instead of sorting cross-sectionally, Figure 11 sorts people into terciles of cash or liquidity relative to their own history of cash or liquidity holdings; that is, for each individual we generate personal liquidity terciles to compare them within their own histories. Figures A.5 and A.6 compare the spending responses to regular and irregular income for three terciles of our alternative measure of liquidity: whether people spend on (discretionary) goods and services prior to income payments. Overall, we see that households exhibit spending responses, even in the highest tercile of all liquidity measures. Figure 9 View largeDownload slide The effects of regular and irregular income on spending by cash (measured by the median number of consumption days held in cash (checking and savings account balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 9 View largeDownload slide The effects of regular and irregular income on spending by cash (measured by the median number of consumption days held in cash (checking and savings account balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 10 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by the median number of consumption days held in liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 10 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by the median number of consumption days held in liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances)) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 11 View largeDownload slide The effects of regular and irregular income on spending by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) relative to peoples’ own histories of liquidity holdings The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 11 View largeDownload slide The effects of regular and irregular income on spending by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) relative to peoples’ own histories of liquidity holdings The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Furthermore, we examine the distribution of cash holdings and liquidity before paydays in Figure 12. We see that cash holdings fall discontinuously at zero when overdrafts start to cost interest and that approximately 10% of people hold fewer than 10 days of cash in their checking and savings accounts. Moreover, approximately 10% of people hold fewer than 10 days of liquidity. In turn, Figure 13 provides a breakdown by 1 to 10 days of spending for cash and liquidity for people who hold less than 10 days of cash or liquidity, respectively. Here, we see that less than 3% of people hold less than 1 day of spending in cash and that less than 3% hold less than 1 day of spending in liquidity. Thus, according to our measures, the fraction of liquidity-constrained people is quantitatively too small to explain the observed spending responses to income. Thus, we conclude that liquidity constraints in the literal sense are unlikely to explain payday responses. Figure 12 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) the morning of paydays Figure 12 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) the morning of paydays Figure 13 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) before paydays of those people with less than 10 consumption days in cash or liquidity Figure 13 View largeDownload slide The distribution of cash (checking and savings account balances) and liquidity (cash plus overdraft and credit card limits minus credit card balances) before paydays of those people with less than 10 consumption days in cash or liquidity Additionally, Table 4 displays summary statistics for the three terciles of liquidity in consumption days. We can see that even the least liquid households hold considerable liquidity of approximately 38 days of spending, while the most liquid tercile of people holds approximately 546 days of spending in liquidity. When we compare these numbers to the state-of-the-art model developed by Kaplan and Violante (2014b) to explain high marginal propensities to consume out of tax rebates, we see a discrepancy between the theoretical predictions and the empirical evidence on the amount of liquid assets that people hold as we explain next. Table 4 Summary statistics by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) in consumption days    (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546     (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546  All income, salary, spending, cash, and liquidity numbers are in U.S. dollars. Table 4 Summary statistics by terciles of liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances) in consumption days    (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546     (1)  (2)  (3)  Monthly income  3,119.3  4,268.0  5,158.8  Age  36.0  41.0  45.0  Spouse  0.2  0.2  0.2  Savings account balance  176.0  665.8  9655.2  Checking account balance  –1,898.8  –1,288.3  2,850.1  Credit card balance  –1,137.9  –1,866.1  –1,911.7  Checking account limit  2,677.3  3,730.1  3,784.5  Credit card limit  2,073.1  5,386.0  8,833.0  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  2,3211.1  Credit utilization  0.5  0.4  0.3  Checking account utilization  0.4  0.3  0.1  Payday loan  41.0  4.0  0.0  Gender  0.5  0.5  0.4  Average daily spending  47.8  54.1  49.3  Number of days held in cash  –38  –14  214  Number of days held in liquidity  38  123  546  All income, salary, spending, cash, and liquidity numbers are in U.S. dollars. Figure 15 shows the life-cycle profiles of liquidity normalized by quarterly consumption for five quintiles of the distribution of agents in the model of Kaplan and Violante (2014b). We see that the liquid asset holdings of the bottom three quintiles are basically zero for all over the simulated agents’ lives. The top two quintiles of agents hold, on average, approximately 4 quarters of consumption in liquidity. By contrast, empirically, the most liquid tercile of people holds, on average, 6.1 quarters of consumption in liquidity, while the middle and least liquid terciles hold 1.37 and 0.41 quarters of consumption in liquidity, respectively, all of which far exceeds the predictions of the model.10 Moreover, if the Kaplan and Violante (2014b) model is to generate the amount of liquidity that we observe in the data, the fixed costs of illiquid assets must be very low, which implies that people can easily adjust their illiquid asset holdings, which reduces their marginal propensity to consume out of fiscal stimulus payments.11 Figure 15 View largeDownload slide Life-cycle profiles of liquid asset in consumption (quarterly) as predicted by the model in Kaplan and Violante (2014b) Figure 15 View largeDownload slide Life-cycle profiles of liquid asset in consumption (quarterly) as predicted by the model in Kaplan and Violante (2014b) Clearly, the Kaplan and Violante (2014b) life-cycle liquidity profiles are a function of medicaid, social security, unemployment insurance, and any institution that partially insures income risk differentially over the life cycle. Iceland is one of the signature countries of the Nordic model. The Nordic model refers to the economic and social policies common to the Nordic countries (Denmark, Finland, Norway, Iceland, and Sweden) including a combination of free market capitalism with a comprehensive welfare state and collective bargaining at the national level. Nevertheless, if anything, the more comprehensive welfare system of Iceland would reduce the need for liquidity and cash cushions in Iceland relative to the United States. According to our measures, the fraction of liquidity-constrained people is quantitatively too small to explain the observed spending responses to income. Nevertheless, many people hold rollover debt; the lowest tercile holds an average of 38 days of their average spending in debt. Obviously, low resources do not necessarily imply liquidity constraints that are determined by future income or other assets that the agent wishes to but cannot borrow against or collateralize. These results suggest that liquidity constraints are not straightforward to document empirically. Almost none of our people are liquidity constrained in the literal sense, that is, they live from paycheck to paycheck and have no ability to consume before their paycheck. Nevertheless, many households may hold a cash or liquidity cushion for either unforeseen adverse expenditure shocks or foreseen expenses. However, they may still be liquidity constrained inasmuch as they would consume or invest more if they could borrow more because they expect higher income in the future or have other assets they cannot collateralize. We are thus left wondering: how can we define liquidity-constrained people and identify them empirically? The average liquidity holdings in our data are pretty similar to the average liquidity holdings in the U.S. data. Table 5 shows a direct comparison of monthly income, cash, and liquidity for three terciles of the liquidity distribution of our data with U.S. data from the Survey of Consumer Finances (SCF) Kaplan and Violante using the 2001 survey as 2014b.12 In fact, average liquidity holdings in the United States are slightly higher than in Iceland. Again, institutions and social norms are very different but the numbers of both countries raise the same question: when do we label an individual as liquidity constrained if he or she holds substantial debt but also liquidity? Table 5 Summary statistics by terciles of liquidity: Comparison to the United States (SCF 2001 data like in Kaplan and Violante (2014b))    (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0     (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0  All numbers are in U.S. dollars. For the SCF data, strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). Cash equals strict liquid wealth minus credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. Table 5 Summary statistics by terciles of liquidity: Comparison to the United States (SCF 2001 data like in Kaplan and Violante (2014b))    (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0     (1)  (2)  (3)  Iceland           Monthly income  3,119.3  4,268.0  5,158.8  Cash  –1,722.8  –622.5  12,505.3  Liquidity  1,889.7  6,627.4  23,211.1  United States           Monthly income  2,655.2  3,741.9  6,112.9  Cash  –2,923.0  2,415.9  38,615.6  Liquidity  5,159.9  12,658.8  62,508.0  All numbers are in U.S. dollars. For the SCF data, strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). Cash equals strict liquid wealth minus credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. The theoretical literature has explicitly considered wealthy households to be liquidity constrained when they lock their wealth in illiquid assets Laibson et al. 2003, Kaplan and Violante 2014b. However, empirically, we find that almost all households hold large amounts of cash and only very few hit a liquidity constraint even right before their paychecks. Because Kaplan and Violante (2014b) use Survey of Consumer Finances data, the authors do not observe liquidity holdings before paychecks but only average liquidity holdings. They classify people as hand-to-mouth consumers when their average liquid wealth is less than half of their earnings, which they find to be the case for 30% of the U.S. population. For comparison, using their definition, we find that in our population 58% of households live hand-to-mouth. However, because people have sufficient liquidity at the end of their pay cycles, this finding cannot explain payday responses to income. People who choose to hold a significant amount of liquidity could “feel” liquidity constrained because they hold an insufficient cash or liquidity cushion. A potential approach to assess whether payday responses are driven by these people is the following: people who have just received a large exogenous wealth shock should not exhibit payday responses, as they are exogenously more liquid. In Figure 14, we thus show that people exhibit substantial payday responses even in the months after which they received a large exogenous wealth shock from a court ruling (explained in Subsection 2.6). Therefore, endogenous liquidity holdings due to insufficient liquidity cushions seemingly do not explain payday responses. Additionally, if we think that people hold liquidity cushions because of the possibility of expense shocks, we would observe a gradual increase in spending shortly before the arrival of income payments (after all, the probability of encountering a big expense shock declines over time). However, we observe a dip in spending if anything. Nevertheless, to examine the question further, we now look at cash-holding responses to income payments. Figure 14 View largeDownload slide The effects of regular income arrival on spending by people who did not or did receive a large exogenous wealth shock in that month The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure 14 View largeDownload slide The effects of regular income arrival on spending by people who did not or did receive a large exogenous wealth shock in that month The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. 2.9 Examining cash-holding responses to income payments Given the difficulties of measuring liquidity constraints in the presence of liquidity cushions, we are interested in a different method that considers the potential existence of liquidity cushions. To this end, inspired by the corporate finance literature, we think about a measure of liquidity constraints derived from peoples’ demands for liquidity. The methodology follows the one in the influential paper by Almeida et al. (2004). In this paper, the authors develop a new test measuring the effect of financial constraints on corporate policies. This effect of financial constraints is captured by a firm’s propensity to save cash out of cash inflows. The authors hypothesize that constrained firms should have a positive cash-flow sensitivity of cash but that unconstrained firms’ cash savings should not be systematically related to cash flows. In the household context, we empirically assess households’ propensities to increase liquidity cushions after cash inflows and the ways in which this propensity is related to liquidity. If a household feels liquidity constrained (even if its “hard” liquidity constraint is not literally binding), it will try to increase its cash or liquidity cushion after cash inflows. In corporate finance, analyses of the effects of financial constraints on firm behavior and the manner in which firms implement financial management have a long tradition. The paper by Almeida et al. (2004) states that firms want to have a liquid balance sheet to undertake valuable projects when they arise. However, if a firm has unrestricted access to external capital—that is, if a firm is financially unconstrained—there is no need to safeguard against future investment needs; thus, corporate liquidity becomes irrelevant. In contrast, when a firm faces financing frictions, liquidity management is a key issue for corporate policy. Thus, there exists a link between financial constraints and corporate liquidity demand, which has been ignored by the prior literature focusing on corporate investment demand Fazzari et al. 1988. Accordingly, the authors examine the influence of financing frictions on corporate investment by comparing the empirical sensitivity of investment to cash flow across groups of firms that are sorted by a proxy for financial constraints. Follow-up research, however, has identified several problems with that strategy regarding the theoretical and empirical robustness of the implications. In a household context, the study by Fazzari et al. (1988) may correspond with the analysis of household spending or investment in response to cash inflows. Households may spend or invest more in response to cash inflows because they are currently liquidity constrained. However, we find that people hold too much cash relative to the predictions of state-of-the-art economic models. In turn, we want to examine whether peoples’ payday responses stem from a concern about insufficient liquidity cushions or future liquidity constraints, which would be reflected in a high marginal propensity to hold on to cash or liquidity. To formalize these ideas, Figure 16 shows the marginal propensities to hold on to cash implied by three different simple models as well as additional model details. First, we consider a standard consumption-savings model without illiquid savings. In this model, the marginal propensity to hold on to cash (MPCash) equals one minus the marginal propensity to consume (MPCons), that is, MPCash = 1-MPCons, as the agent holds his entire lifetime wealth in cash. Because the MPCons in this model is always decreasing in income or liquidity, the MPCash will always be increasing. Furthermore, the MPCash is higher when the agent’s horizon increases, as he consumes only a small amount of his income and saves most of it. Figure 16 View largeDownload slide Marginal propensities to consume, save illiquidly, and save liquidly as implied by models with and without illiquid savings and future binding liquidity constraints We consider a three-period toy model with stochastic income, a liquid asset, and an illiquid asset. We follow Carroll (1997), who specify income $$Y_{t}$$ as log-normal and characterized by transitory shocks $$N_{t}^{T}$$. Doing so allows for a low probability $$p$$ of unemployment; that is, $$N_{t}^{T}=e^{s_{t}^{T}}$$ with $$s_{t}^{T}\sim N(\mu_{T},\sigma_{T}^{2})$$ or $$N_{t}^{T}=0$$ with probability $$p$$. We assume a probability of 8% for unemployment in line with our sample’s summary statistics. Additionally, we calibrate the model to a quarterly frequency and assume an annualized transitory income shock volatility of 0.2 following Carroll (1997) such that $$\sigma_{T}=0.2(\sqrt{0.25})$$. We also assume an annualized exponential discount factor of 0.98 such that $$\delta=0.98^{0.25}$$ and a simple power-utility specification $$u(C)=\frac{C^{1-\theta}}{1-\theta}$$ with $$\theta=2$$. The interest on illiquid savings is given by an annualized 2% such that $$r=0.02(0.25)$$ while liquid savings earn no interest. To simulate a long horizon in a three-period model, we assume that the last period is $$T=10$$ years ahead and no intermediate consumption takes place. Moreover, to illustrate the effects of liquidity constraints, we assume that the agent can collateralize only 30% of his illiquid asset holdings and can only access 30% of his transitory income shock in the second period right away. Thus, the agent expects to be liquidity constrained in the second period. In the last period, the agent consumes his liquid savings from the second period in addition to his illiquid savings. In turn, the model is solved by backward induction using standard numerical optimization techniques. The graphs depict the marginal propensities to consume, save illiquidly, and save liquidly in period 1. Figure 16 View largeDownload slide Marginal propensities to consume, save illiquidly, and save liquidly as implied by models with and without illiquid savings and future binding liquidity constraints We consider a three-period toy model with stochastic income, a liquid asset, and an illiquid asset. We follow Carroll (1997), who specify income $$Y_{t}$$ as log-normal and characterized by transitory shocks $$N_{t}^{T}$$. Doing so allows for a low probability $$p$$ of unemployment; that is, $$N_{t}^{T}=e^{s_{t}^{T}}$$ with $$s_{t}^{T}\sim N(\mu_{T},\sigma_{T}^{2})$$ or $$N_{t}^{T}=0$$ with probability $$p$$. We assume a probability of 8% for unemployment in line with our sample’s summary statistics. Additionally, we calibrate the model to a quarterly frequency and assume an annualized transitory income shock volatility of 0.2 following Carroll (1997) such that $$\sigma_{T}=0.2(\sqrt{0.25})$$. We also assume an annualized exponential discount factor of 0.98 such that $$\delta=0.98^{0.25}$$ and a simple power-utility specification $$u(C)=\frac{C^{1-\theta}}{1-\theta}$$ with $$\theta=2$$. The interest on illiquid savings is given by an annualized 2% such that $$r=0.02(0.25)$$ while liquid savings earn no interest. To simulate a long horizon in a three-period model, we assume that the last period is $$T=10$$ years ahead and no intermediate consumption takes place. Moreover, to illustrate the effects of liquidity constraints, we assume that the agent can collateralize only 30% of his illiquid asset holdings and can only access 30% of his transitory income shock in the second period right away. Thus, the agent expects to be liquidity constrained in the second period. In the last period, the agent consumes his liquid savings from the second period in addition to his illiquid savings. In turn, the model is solved by backward induction using standard numerical optimization techniques. The graphs depict the marginal propensities to consume, save illiquidly, and save liquidly in period 1. Second, we consider a consumption-savings model in which the agent can save in a liquid or an illiquid asset that pays higher interest. In such a model, the MPCash may be either increasing or decreasing in liquidity or income because the MPCash equals one minus the MPCons minus the marginal propensity to invest in the illiquid asset (MPIllInv); that is, MPCash = 1-MPCons-MPIllInv. While the MPCons is always decreasing in liquidity, the MPIllInv is increasing, which implies that the MPCash is either increasing or decreasing. However, the MPCash is always small like in the model of Kaplan and Violante (2014b), because agents have little reason to hold cash or liquid savings. If one would add transitory income shocks to the Kaplan et al. (2014) in the spirit of the buffer stock model of Carroll (2001) then agents would hold more liquidity but their marginal propensity to consume (MPC) out of fiscal stimulus payments would be low again. Therefore, we analyze a model with both transitory shocks and expected liquidity constraints to increase the MPC out of fiscal stimulus payments. One way to add future liquidity constraints is to assume that the agent receives news about income shocks in the future but that he or she will not be able to consume that income immediately. Frequently binding future liquidity constraints generate the equivalent prediction from the corporate finance literature: an MPCash that is decreasing in liquidity. This theoretical result is robust to different assumptions about the environment, such as social security payments. In fact, the prediction that the MPCash is decreasing in liquidity in anticipation of liquidity constraints appears to hold quite generally: suppose people have some type of threshold rule for how much rollover debt they allow themselves to hold. In that case, they would display a high marginal propensity to hold on to cash when they are close to that threshold. Because we include individual fixed effects, we can pick up heterogeneity along the lines of where people set their rules. Therefore, we only utilize variations from comparing people with low or high liquidity holdings to themselves. In the high-frequency consumption setting that we consider, illiquid savings could be interpreted as purchasing bulk consumption goods for instance. Figure 17 displays peoples’ cash-holding responses to regular and irregular income payments for three terciles of liquidity. We can see that less liquidity-constrained people have more pronounced cash-holding responses than more liquidity-constrained people. Moreover, cash responses are larger than spending responses. Both of these findings are predicted by a standard consumption-savings model. Thus, we conclude that cash responses do not seem to indicate the presence of illiquid savings, future liquidity constraints, or insufficient liquidity cushions. Instead of sorting people cross-sectionally, we could again analyze individual behavior for liquidity terciles relative to own histories of liquidity holdings. While the responses are somewhat flatter in this case, we do not find a decreasing pattern in the responses either. Figure 17 View largeDownload slide The effects of regular and irregular income arrival on liquidity by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction liquidity deviates from average liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances). Figure 17 View largeDownload slide The effects of regular and irregular income arrival on liquidity by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction liquidity deviates from average liquidity (checking and savings account balances plus overdraft and credit card limits minus credit card balances). Even for deciles, all of the pictures show an increasing relationship between the propensity to hold on to cash and liquidity constraints as well as a very high propensity to hold on to cash, one that is much higher than the propensity to consume. We again use indicator variables for income payments to alleviate potential endogeneity concerns, but we can also estimate the MPCash directly and obtain the same relationship with liquidity. These findings are thus consistent with the standard consumption-savings model without illiquid savings. However, this model is not consistent with a high marginal propensity to consume out of transitory income shocks. We thus conclude that neither current nor future liquidity constraints can account for the observed payday responses to income payments. People can reduce overdraft limits relatively easily, while any credit limit increases have to be approved by the bank. Examining changes in overdraft limits around paydays yields very interesting results. Figure A.7 shows that people with less liquidity tend to reduce their overdraft limits around paydays, whereas people with high liquidity do not engage in such behavior. In itself, this is evidence against liquidity constraints being a problem in our sample and points toward the existence of overconsumption problems. After all, standard economic theory predicts that people should never reduce their limits, as borrowing opportunities are always weakly welfare increasing. However, we clearly see that less liquid people tend to reduce their limits after paydays. A potential explanation for this tendency is that people want to restrict their future selves from borrowing or that they want to reduce their mental borrowing accounts. To ensure that the documented increasing payday liquidity responses do not stem from low-liquidity peoples’ tendencies to reduce their limits after paydays, we also examine peoples’ balances—that is, their checking and savings account balance minus their credit balance—in Figure A.8. We again observe high and increasing responses that are consistent with a model without illiquid savings or future binding liquidity constraints. While the Icelandic financial crisis undoubtedly affected people, we believe that our qualitative results do not depend on the financial crisis or are otherwise country-specific. Even if the crisis had an effect on cash or liquidity holdings (instead of investments into durable goods, houses, cars or stocks and bonds) in Iceland today, we firmly believe we would be left with the same puzzling observation: significant payday responses in the presence of substantial liquidity. Iceland recovered very quickly and experienced high GDP growth and low unemployment during our sample period. Moreover, we include individual and month-by-year effects controlling for all individual (un)observables as well as any slow-moving trends. Additionally, we redid the entire analysis restricting ourselves to the year 2015 and do not find any differences. The OECD Economic Survey of Iceland from June 2011 states that the economic contraction and rise in unemployment appear to have been stopped by late 2010 with growth under way in mid-2011. The Icelandic government was successfully able to raise $${\$}$$1 billion with a bond issue in June 2011, indicating that international investors have given the government and the new banking system a clean bill of health. By mid-2012, Iceland was regarded as a recovery success story with 2 years of economic growth and unemployment down to 6.3%. Moreover, we find quantitatively similar payday responses (around 50% for the average household) to Gelman et al. (2014), who use U.S. data of the same kind. Of course, one has to keep in mind that there are large institutional differences between the United States and Iceland, as well as differences in average income, liquidity, and demographics. All of these differences make the magnitudes not perfectly comparable. 3. Conclusion We use data from a financial account aggregation provider in Iceland to evaluate whether spending or consumption results from an intertemporal optimization problem and is thus independent of income. The spending and income data are characterized by outstanding accuracy and comprehensiveness because of Icelanders’ nearly exclusive use of electronic payments. In line with previous studies, we find significant spending responses to the receipt of regular and irregular income on paydays. Moreover, we are in a unique position to simultaneously analyze credit and overdraft balances and limits at the same high frequency. This allows us to convincingly rule out liquidity constraints as an explanation. In contrast to previous studies, we thus argue that hand-to-mouth behavior is not limited to liquidity-constrained households, because we show that non-liquidity-constrained households exhibit hand-to-mouth behavior through various measures of liquidity constraints: balances and credit limits, spending on discretionary goods and services, and spending immediately before income payments. Overall, less than 3% of people have less than 1 day of average spending left in liquidity before their paydays. Because people may either hold liquidity cushions or save for foreseen expenses, we also examine cash-holding responses to income payments, inspired by the corporate finance literature. We notice that a model with liquid and illiquid savings and future liquidity constraints makes a joint prediction about the marginal propensity to consume and the marginal propensity to hold on to cash: both are decreasing in liquidity. We test this joint prediction in our data, however we do not find evidence for it. Because the cash-holding responses are most consistent with the standard consumption-savings problem without illiquid savings or future binding liquidity constraints, we argue that the evidence is not consistent with either present or future liquidity constraints. Moreover, our findings highlight a general difficulty to measure liquidity constraints. To determine whether people with liquidity cushions and rollover debt are liquidity constrained, however, is of great importance: after all, liquidity constraints are not equivalent to low financial resources. For policy purposes, liquidity constraints call for expanding credit, whereas low resources due to overconsumption problems call for restricting credit. The latter measure is also supported by our finding that low-liquidity households tend to voluntarily reduce their overdraft limits around paydays. While we have no way of telling whether the excess consumption on paydays is due to a time-inconsistent overconsumption problem, it seems to be generally believed nowadays that the extent of credit card borrowing in the United States (and Icelanders hold similar amounts of credit) is not consistent with standard models and must be explained by time-inconsistent overconsumption and the use of illiquid savings as a commitment device Laibson et al. 2003, Carroll 2001. We thank Itay Goldstein and two anonymous referees. We also thank Charlie Calomiris, Gianluca Violante, Steffen Andersen, Jonathan Parker, Dan Silverman, Jeremy Tobacman, David Laibson, Xavier Gabaix, Shachar Kariv, Botond Koszegi, Steve Zeldes, Michael Woodford, Ted O’Donoghue, Wei Jiang, Dimitri Vayanos, Emir Kamenica, Jialan Wang, Adam Szeidl, Matthew Rabin, Sumit Agarwal, Paul Tetlock, Tom Chang, and Justin Sydnor for their insightful comments. We also thank the seminar and the conference participants at the Behavioral Economics Annual Meeting (BEAM) 2016, the CFPB research conference, TAU, AFA, University of British Columbia, Boston Fed, Imperial College London, UC Berkeley, University of Zurich, Carnegie Mellon University, Columbia University, Gerzensee, Goethe University Frankfurt, University of Lugano, Stockholm School of Economics, University of St. Gallen, Huntsman, University of Colorado Boulder, and University of Utah for all their constructive questions and remarks. Nathen Huang and Guangyu Wang provided excellent research assistance. We are indebted to Ágúst Schweitz Eriksson and Meniga for providing and helping with the data. An earlier draft of this paper was circulating under the title “The Only Day Better than Friday is Payday!” Appendix Figure A.1 View largeDownload slide The effects of paycheck arrival on necessary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.1 View largeDownload slide The effects of paycheck arrival on necessary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.2 View largeDownload slide The effects of paycheck arrival on discretionary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.2 View largeDownload slide The effects of paycheck arrival on discretionary spending The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.3 View largeDownload slide The effects of regular income arrival on spending by people costs in banking fees, interest, and payday loans The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.3 View largeDownload slide The effects of regular income arrival on spending by people costs in banking fees, interest, and payday loans The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.4 View largeDownload slide The effects of regular and irregular income on spending by terciles of credit utilization The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.4 View largeDownload slide The effects of regular and irregular income on spending by terciles of credit utilization The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.5 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.5 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.6 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend on discretionary goods and services as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.6 View largeDownload slide The effects of regular and irregular income on spending by liquidity (measured by how much people spend on discretionary goods and services as compared to their average in the last 4 days prior to income arrival) The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction spending deviates from average daily spending. Figure A.7 View largeDownload slide The effects of regular and irregular payments on overdraft limits by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: how overdraft limits change. Figure A.7 View largeDownload slide The effects of regular and irregular payments on overdraft limits by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: how overdraft limits change. Figure A.8 View largeDownload slide The effects of regular income and salary arrival on cash minus credit balances by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction the balances deviate from their average. Figure A.8 View largeDownload slide The effects of regular income and salary arrival on cash minus credit balances by terciles of consumption days from current liquidity The horizontal axis displays 2 weeks around paydays. The vertical axis shows the estimated coefficients for each day: the fraction the balances deviate from their average. Footnotes 1 This is true for both the standard consumption-savings model Friedman 1957, Hall 1978 and the more recent “buffer-stock” versions Deaton 1991, Carroll 1997. 2 Examples of micro evidence on excess sensitivity are Parker (1999), Souleles (1999), Shapiro and Slemrod (2003a), Shapiro and Slemrod (2003b), Shapiro and Slemrod (2009), Johnson et al. (2006), Parker et al. (2013), and Broda and Parker (2014), as surveyed in Jappelli and Pistaferri (2010) and Fuchs-Schundeln and Hassan (2015). Macro evidence is provided by Campbell and Mankiw (1989) and Campbell and Mankiw (1990) in response to the seminal paper by Flavin (1981). 3 As additional evidence, we present large spending responses by people who recently received a large exogenous wealth shock due to a court ruling. 4 Cochrane (1989) and Krusell and Smith (1996) perform similar calculations in a representative agent environment. 5 More substantial welfare effects of excess sensitivity to consumption are documented by Ganong and Noel (2016) and Baker and Yannelis (2017). 6 We also confirm the finding of Parker (2014) that liquidity appears to be a very persistent household trait rather than the product of swings due to transitory income shocks, as predicted in the Kaplan and Violante (2014b) model. 7 All numbers stem from the U.S. Census Bureau’s American Community Survey (ACS) in 2015. 8 Iceland’s 2008 financial crisis was exacerbated by banks that borrowed in Japanese yen or Swiss francs to take advantage of lower interest rates and then repackaged the loans in krona before passing them on to clients. This exchange-rate indexation of loans meant that the total amounts owed in Icelandic krona varied according to its exchange rate against the currencies in which the loans were issued. Such loans had been aggressively promoted by Icelandic banks in previous years and left many diligent car and homeowners with debts greater than the original amount despite paying their bills every month. 9 Other studies examining fiscal stimulus payments are Johnson et al. (2006), Parker et al. (2013), Parker (2014), and Jappelli and Pistaferri (2014), as surveyed by Jappelli and Pistaferri (2010). 10 As we will explain below, we believe that the observed discrepancy between the model-predicted and observed liquidity holdings are not due to the Icelandic financial crisis or otherwise country-specific. The economy has been booming in the sample period and many households have large amounts of rollover debt (inefficient financial markets should restrict borrowing). Moreover, Iceland is characterized by well-functioning health care, social security, and unemployment insurance systems. Additionally, as mentioned, there are no minimum balance restrictions associated with bank accounts that cause people to hold liquidity or cash in Iceland. 11 Liquidity holdings are increased in a model with transitory income shocks in addition to permanent shocks, which are a standard feature of life-cycle models and for which much evidence exists Carroll refer to 2001, Carroll et al. refer to 1992. However, in that case, the model is not able to generate a high MPC out of fiscal stimulus payments any more as people hold enough liquidity to deal with transitory income shocks. 12 Strict liquid wealth equals money market, checking, savings, and call accounts plus currency holdings (assumed to be $${\$}$$69 per individual by Kaplan and Violante (2014b)). We then compute cash by subtracting credit card debt. Liquidity equals individual’s maximum credit capacity plus cash. References Addoum, J. M., Delikouras, S., and Korniotis, G. M. License to spend: Consumption-income sensitivity and portfolio choice. Available at SSRN 2409209 , 2015. Agarwal, S. and Qian, W. Consumption and Debt Response to Unanticipated Income Shocks: Evidence from a Natural Experiment in Singapore. American Economic Review , 104 12: 4205– 4230, 2014. Google Scholar CrossRef Search ADS   Almeida, H., Campello, M., and Weisbach, M. The Cash Flow Sensitivity of Cash. Journal of Finance , 59 4: 1777– 1804, 2004. Google Scholar CrossRef Search ADS   Alvarez, F., Guiso, L., and Lippi, F. Durable Consumption and Asset Management with Transaction and Observation Costs. American Economic Review , 5 102: 2272– 2300, 2012. Google Scholar CrossRef Search ADS   Angeletos, G. M., Laibson, D., Repetto, A., and Weinberg, S. The Hyperbolic Consumption Model: Calibration, Simulation, and Empirical Evaluation. Journal of Economic Perspectives , 15 3: 47, 2001. Google Scholar CrossRef Search ADS   Attanasio, . and Pavoni, N. Risk Sharing in Private Information Models with Asset Accumulation: Explaining the Excess Smoothness of Consumption. Econometrica , 79 4: 1027– 1068, 2011. Google Scholar CrossRef Search ADS   Baker, S. R. Debt and the Consumption Response to Household Income Shocks. Mimeo , 2013. Baker, S. R. and Yannelis, C. Income changes and consumption: Evidence from the 2013 federal government shutdown. Review of Economic Dynamics , 23: 99– 124, 2017. Google Scholar CrossRef Search ADS   Broda, C. and Parker, J. The Economic Stimulus Payments of 2008 and the Aggregate Demand for Consumption. Mimeo , 2014. Browning, M. and Crossley, T. F. The Life-Cycle Model of Consumption and Saving. Journal of Economic Perspectives , 15 3: 3– 22, 2001. Google Scholar CrossRef Search ADS   Campbell, J. Y. and Mankiw, N. G. Consumption, Income and Interest Rates: Reinterpreting the Time Series Evidence. NBER Macroeconomics Annual 1989, Volume 4, NBER Chapters National Bureau of Economic Research, Inc. , pages 185– 246, 1989. Campbell, J. Y. and Mankiw, N. G. Permanent Income, Current Income, and Consumption. Journal of Business and Economic Statistics , 8 3: 265– 279, 1990. Carroll, C. D. Buffer Stock Saving and the Life Cycle/Permanent Income Hypothesis. Quarterly Journal of Economics , CXII: 1– 56, 1997. Google Scholar CrossRef Search ADS   Carroll, C. D. A theory of the consumption function, with and without liquidity constraints. The Journal of Economic Perspectives , 15 3: 23– 45, 2001. Google Scholar CrossRef Search ADS   Carroll, C. D., Hall, R. E., and Zeldes, S. P. The buffer-stock theory of saving: Some macroeconomic evidence. Brookings papers on economic activity , 1992 2: 61– 156, 1992. Google Scholar CrossRef Search ADS   Carvalho, L. S., Meier, S., and Wang, S. W. Poverty and economic decision-making: Evidence from changes in financial resources at payday. American Economic Review , 106 2: 260– 84, 2016. Google Scholar CrossRef Search ADS PubMed  Chetty, R. and Szeidl, A. Consumption Commitments and Risk Preferences. Quarterly Journal of Economics , 122 2: 831– 877, 2007. Google Scholar CrossRef Search ADS   Cochrane, J. H. The Sensitivity of Tests of the Intertemporal Allocation of Consumption to Near-Rational Alternatives. American Economic Review , 79 3: 909– 936, 1989. de Nicola, F. and Giné, X. How accurate are recall data? evidence from coastal india. Journal of Development Economics , 106 0: 52– 65, 2014. ISSN 0304-3878. http://dx.doi.org/10.1016/j.jdeveco.2013.08.008. URL http://www.sciencedirect.com/science/article/pii/S0304387813001235. Google Scholar CrossRef Search ADS   Deaton, A. Saving and Liquidity Constraints. Econometrica , 59: 1221– 1248, 1991. Google Scholar CrossRef Search ADS   Fazzari, S., Hubbard, G., and Petersen, B. Financing Constraints and Corporate Investment. Brookings Papers on Economic Activity , 1: 141– 206, 1988. Google Scholar CrossRef Search ADS   Flavin, M. The Adjustment of Consumption to Changing Expectations about Future Income. Journal of Political Economy , 89: 974– 1009, 1981. Google Scholar CrossRef Search ADS   Flavin, M. and Nakagawa, S. A Model of Housing in the Presence of Adjustment Costs: A Structural Interpretation of Habit Persistence. American Economic Review , 98 1: 474– 495, 2008. Google Scholar CrossRef Search ADS   Friedman, M. A Theory of the Consumption Function . Number frie57-1 in NBER Books. National Bureau of Economic Research, Inc, 1957. URL http://ideas.repec.org/b/nbr/nberbk/frie57-1.html. Fuchs-Schundeln, N. and Hassan, T. Natural Experiments in Macroeconomics. Working Paper , 2015. Gabaix, X. and Laibson, D. The 6D Bias and the Equity-Premium Puzzle. NBER Macroeconomics Annual , 16: 257– 312, 2002. Google Scholar CrossRef Search ADS   Ganong, P. and Noel, P. How does unemployment affect consumer spending? Technical report, Working paper , 2016. Gelman, M., Kariv, S., Shapiro, M. D., Silverman, D., and Tadelis, S. Harnessing naturally occurring data to measure the response of spending to income. Science , 345 6193: 212– 215, 2014. Google Scholar CrossRef Search ADS PubMed  Hall, R. E. Stochastic Implications of the Life Cycle-Permanent Income Hypothesis: Theory and Evidence. Journal of Political Economy , 86 6: 971– 987, 1978. Google Scholar CrossRef Search ADS   Hastings, J. and Washington, E. The first of the month effect: Consumer behavior and store responses. American Economic Journal: Economic Policy , 2 2:pp. 142– 162, 2010. ISSN 19457731. URL http://www.jstor.org/stable/25760066. Google Scholar CrossRef Search ADS   Huffman, D. and Barenstein, M. A Monthly Struggle for Self-Control? Hyperbolic Discounting, Mental Accounting, and the Fall in Consumption Between Paydays. Working Paper IZA Bonn , 2005. Huntley, J. and Michelangeli, V. Can Tax Rebates Stimulate Spending in a Life-Cycle Model? American Economic Journal: Macroeconomics , 1 6: 162– 189, 2014. Google Scholar CrossRef Search ADS   Jappelli, T. and Pistaferri, L. The Consumption Response to Income Changes. Annual Review of Economics , 2: 479– 506, 2010. Google Scholar CrossRef Search ADS   Jappelli, T. and Pistaferri, L. Fiscal policy and mpc heterogeneity. American Economic Journal: Macroeconomics , 6 4, 2014. URL http://ideas.repec.org/p/sef/csefwp/325.html. Johnson, D., Parker, J., and Souleles, N. Household Expenditure and the Income Tax Rebates of 2001. American Economic Review , 96 5: 1598– 1610, 2006. Google Scholar CrossRef Search ADS   Kaplan, G. and Violante, G. A Tale of Two Stimulus Payments: 2001 vs 2008. American Economic Review (Papers and Proceedings) , 5 104: 162– 189, 2014a. Kaplan, G. and Violante, G. A model of the consumption response to fiscal stimulus payments. Econometrica , 82: 1199– 1239, 2014b. Google Scholar CrossRef Search ADS   Kaplan, G., Violante, G., and Weidner, J. The wealthy hand-to-mouth. Brookings Papers on Economic Activity , 2014. Karlan, D. and Zinman, J. Lying about borrowing. Journal of the European Economic Association , 6( 2/3):pp. 510– 521, 2008. ISSN 15424766. URL http://www.jstor.org/stable/40282660. Google Scholar CrossRef Search ADS   Koszegi, B. and Rabin, M. Reference-Dependent Consumption Plans. American Economic Review , 99 3: 909– 936, 2009. Google Scholar CrossRef Search ADS   Krusell, P. and Smith, A. Rules of Thumb in Macroeconomic Equilibrium: A Quantitative Analysis. Journal of Economic Dynamics and Control , 20 4: 527– 558, 1996. Google Scholar CrossRef Search ADS   Kuchler, T. Sticking To Your Plan: Empirical Evidence on the Role of Present Bias for Credit Card Debt Paydown. Working Paper , 2015. Kueng, L. Explaining Consumption Excess Sensitivity with Near-Rationality: Evidence from Large Predetermined Payments. Working Paper , 2015. Laibson, D., Repetto, A., and Tobacman, J. A Debt Puzzle. in Aghion, Philippe Frydman, Roman Stiglitz, Joseph and Woodford, Michael eds., Knowledge, Information, and Expectations in Modern Economics: In Honor of Edmund S. Phelps , Princeton: Princeton University Press, 2003. Laibson, D., Repetto, A., and Tobacman, J. Estimating Discount Functions with Consumption Choices over the Lifecycle. Working Paper , 2015. Mastrobuoni, G. and Weinberg, M. Heterogeneity in intra-monthly consumption patterns, self-control, and savings at retirement. American Economic Journal.Economic Policy , 1 2: 163– 189, 08 2009. URL http://search.proquest.com/docview/872053305?accountid=39039. Google Scholar CrossRef Search ADS   Pagel, M. Expectations-based Reference-Dependent Life-Cycle Consumption. Mimeo , 2013. Parker, J. A. The Reaction of Household Consumption to Predictable Changes in Social Security Taxes. American Economic Review , 89 4: 959– 973, 1999. Google Scholar CrossRef Search ADS   Parker, J. A. Why Don’t Households Smooth Consumption? Evidence from a 25 million dollar experiment. Working Paper , 2014. Parker, J. A., Souleles, N., Johnson, D., and McClelland, R. Consumer Spending and the Economic Stimulus Payments of 2008. American Economic Review , 103 6: 2530– 53, 2013. Google Scholar CrossRef Search ADS   Reis, R. Inattentive Consumers. Journal of Monetary Economics , 53: 1761– 1800, 2006. Google Scholar CrossRef Search ADS   Shapiro, J. Is there a daily discount rate? Evidence from the food stamp nutrition cycle. Journal of Public Economics , 89: 303– 325, 2005. Shapiro, M. and Slemrod, J. Consumer Response to Tax Rebates. American Economic Review , 93 1: 381– 396, 2003a. Google Scholar CrossRef Search ADS   Shapiro, M. and Slemrod, J. Did the 2001 Tax Rebate Stimulate Spending? Evidence from Taxpayer Surveys. Tax Policy and the Economy , 17: 83– 109, 2003b. Google Scholar CrossRef Search ADS   Shapiro, M. and Slemrod, J. Did the 2008 Tax Rebates Stimulate Spending? American Economic Review , 99 2: 374– 379, 2009. Google Scholar CrossRef Search ADS PubMed  Shapiro, M. D. and Slemrod, J. Consumer response to the timing of income: Evidence from a change in tax withholding. American Economic Review , 85 1: 274– 83, March 1995. URL http://ideas.repec.org/a/aea/aecrev/v85y1995i1p274-83.html. Sims, C. Implications of Rational Inattention. Journal of Monetary Economics , 50 3: 665– 690, 2003. Google Scholar CrossRef Search ADS   Souleles, N. The Response of Household Consumption to Income Tax Refunds. American Economic Review , 89 4: 947– 958, 1999. Google Scholar CrossRef Search ADS   Tutoni, A. Rationally Inattentive Consumption Choices. Review of Economic Dynamics , 16: 421– 439, 2010. Google Scholar CrossRef Search ADS   © The Author(s) 2018. Published by Oxford University Press on behalf of The Society for Financial Studies. All rights reserved. For permissions, please e-mail: journals.permissions@oup.com. 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

The Review of Financial StudiesOxford University Press

Published: Apr 25, 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