Ebook Credit Card Use After The Final Mortgage Payment: Does The Magnitude Of Income Shocks Matter?
The Life Cycle/Permanent Income Hypothesis (PIH) predicts that individuals should smooth consumption over time if future income shocks are predictable. For example, if an individual knew with certainty that she would receive $1000 in 6 months time, the PIH predicts that she should borrow today and then pay off this debt when she receives the predictable income shock in the future. This is so she can smooth consumption both before as well as after the date she receives the income shock.
However, even though the PIH is central to much of modern consumption theory, and in spite of a very large number of empirical studies on consumption smoothing, no consensus has emerged on whether consumption smoothing does or does not hold empirically. It remains a major outstanding puzzle to explain why consumption smoothing is sometimes accepted and sometimes rejected by the data.
A variety of authors (e.g. Kreinin, 1961, Souleles, 1999, Browning and Collado, 2001, Hsieh, 2003, Coulibaly and Li, 2006, Stephens, 2008) have suggested that one possible solution to this puzzle involves the magnitude of the predictable income shock. This argument (which we term the “magnitude hypothesis”) states that consumption smoothing will hold if the size of the predictable income shock is large enough, but will not hold if these predictable income shocks are small.
One popular explanation for the magnitude hypothesis in is bounded rationality. Browning and Collado (2001) argue that individuals “do smooth (consumption) …if there are large and predictable income changes” (p. 682) but that they “will not bother to adjust optimally to small income changes since the utility cost of doing so is small” (p. 690). Similarly, Hsieh (2003) summarizes the bounded rationality argument by noting that there may be “costs associated with the mental processing of these forecastable income changes” (p. 404).
To extend our example above, if the amount of the certain future income shock was small (say $100), then the magnitude hypothesis suggests that the individual may “not bother” to arrange the credit needed to smooth consumption, or to engage in the “mental processing” needed to work out her optimal consumption patterns. On the other hand, if the magnitude of the future income shock was large (say $5000), then the magnitude hypothesis suggests that the individual is much more likely to smooth consumption by making use of credit and working out her optimal stream of consumption over time.
Table 1 provides a summary of some of the literature testing the PIH using identifiable income shocks. Panel A includes papers that discuss the magnitude hypothesis as a possible explanation for their results, while panel B includes papers that do not. An interesting observation from Table 1 is that many of the papers who have discussed the magnitude hypothesis as an explanation (panel A) find results that are consistent with the PIH, while a majority of papers who do not discuss the magnitude hypothesis (panel B) find results that are not consistent with the PIH.
This would seem to suggest the importance of the magnitude hypothesis as an explanation for the PIH. Some of the papers listed in panel A, such as Browning and Collado, (2001), Hsieh, (2003) and Coulibaly and Li, (2006) do not set out to formally test the magnitude hypothesis, but rather suggest that their results may be consistent with the magnitude hypothesis, because the PIH tends to hold following income shocks that the authors consider are “large”. Other papers such as Kreinin, (1961) and Souleles (1999) do formally test the magnitude hypothesis, but as we argue in Section 2 below, use data that is subject to important data concerns. The aim of our paper is to provide a new test of the magnitude hypothesis using a high quality new database.
Our data consists of a confidential individual level database provided by a Canadian bank. The data consists of monthly statement data for approximately 20 000 individuals for both their credit card as well as their mortgage accounts, over 19 months. We follow Coulibaly and Li (2006) and Stephens (2008) in arguing that the final mortgage payment of an individual can be analyzed as an expected disposable income shock.
Our aim is to examine how credit card usage is impacted by the expected disposable income shock of a final mortgage payment. We measure the expected disposable income shock using our mortgage data, and we measure the individual’s consumption and debt response using our credit card data. Our main test of the magnitude hypothesis examines if consumption and debt responses are different for individuals with high compared to low expected disposable income shocks (i.e. the cessation of high versus low monthly mortgage payments).
Our use of monthly credit card data to examine issues around consumption smoothing follows a variety of recent papers including Gross and Souleles (2002a) and Agarwal, Liu and Souleles (2007) etc. We believe that our data set is unique, however, because our monthly credit card data is matched to monthly mortgage balance data. This allows us to be the first to use monthly bank account data to specifically test the magnitude hypothesis. Because our monthly mortgage account data is matched with monthly credit card statement data, we are able to exploit the wide variance in the magnitude of final mortgage payments over individuals, in order to test how the magnitude of an expected disposable income shock impacts credit card consumption.
There are a number of important advantages in using this database and research design to test the magnitude hypothesis. First, because we have monthly data on each individual’s mortgage balance as it declines towards zero, we are able to isolate exactly which month a mortgage holder finally pays off their mortgage as well as the exact amount of the monthly payments. In other words we have a remarkably precise measure of both the timing and magnitude of each individual’s expected future income shock as measured by the final monthly mortgage payment.
This differs from those papers in the literature that have identified either the timing or magnitude of income shocks using survey based databases (such as the Consumer Expenditure Survey (CEX) or the Survey of Consumer Finance (SCF)), which are subject to various well known measurement issues inherent in the use of survey based data. Essentially, the monthly bank account data we observe concerning the timing and magnitude of the income shock is the same data observed by the individual in the study.
Second, we exploit the fact that the dates of final mortgage payments are randomly distributed across individuals over time. In this regard, our use of final mortgage payments as an expected income shock differs from examining government payments (e.g. tax rebate payments or fiscal stimulus payments) which have been extensively examined in the consumption smoothing literature (see Table 1). As highlighted by Agarwal, Liu and Souleles (2007), government payments tend to be clustered for all individuals in a few months of the year, thus it may be difficult to disentangle whether each individual’s consumption on that date was responding to that specific government payment, or to any other macroeconomic factor that occurred at the same time, e.g. stock exchange or interest rate developments.
In our research design, we are able to exploit the random distribution of the date of the final mortgage payment across individuals to identify exactly when specific individuals received this disposable income shock relative to all other individuals in our sample. Furthermore, we are able to use our data to only include instances where the date of an individual’s final mortgage payment is predetermined, an important element of identification in our tests.
Third, we are able to extend the existing monthly credit card statement based literature (e.g. Gross and Souleles, 2002a, Agarwal et al 2007 etc) by adding census based measures of variables such as income. Our monthly bank statement data includes the Canadian postal code of each individual in our sample, thus we are able to match this data with Canadian Census data which provides post code level data on a variety of demographic variables. In particular, we use the post code level income data to test a variation of the magnitude hypothesis that income shocks should be classified as large or small relative to the agent’s income.
Our research also has important policy implications. An important motivation of the large literature examining consumption and debt responses to income shocks, concerns measuring the impacts of fiscal policies such as tax rebates and cuts, as well as fiscal stimuli programs (such as the 2008 US Government stimulus package, where each individual was sent a check in an attempt to increase consumption). However, as described above, very little evidence exists on how the dollar magnitude of such expected income shocks impact individual consumption and debt response.
In other words, an important policy issues is whether a larger stimulus payment check (of say $1000) will have different impacts than a smaller stimulus payment check (of say $300). Stephens (2008), argues that “the smaller income changes, that are not smoothed, are typically the focus of stimulative fiscal measures (e.g. tax rebates and permanent tax cuts)” (italics added p. 241). Even though this paper uses data on mortgage payments, rather than fiscal stimuli, we argue that the policy relevance of the magnitude hypothesis in the context of fiscal stimuli payments is an important additional motivation for our research.
Posted in :
