'The Effect of Payday Lending Restrictions on Liquor Sales'

This is a summary of new research from two of our former graduate students here at the University of Oregon, Harold Cuffe and Chris Gibbs (link to full paper):

The effect of payday lending restrictions on liquor sales – Synopsis, by Harold Cuffe and Chris Gibbs: The practice of short-term consumer financing known as payday lending remains controversial because the theoretical gains in welfare from greater credit access stand in opposition to anecdotal evidence that many borrowers are made worse off. Advocates for the industry assert that the loans fill a gap in credit access for underserved individuals facing temporary financial hardship. Opponents, who include many state legislatures and the Obama administration, argue that lenders target financially vulnerable individuals with little ability to pay down their principal, who may end up paying many times the borrowed amount in interest and fees.

Regulations restricting both payday loan and liquor access seek to minimize the potential for overuse. To justify intervention in the two markets, policy makers note a host of negative externalities associated with each product, and cite behavioral motivations underlying individuals' consumption decisions. In particular, researchers have shown that the same models of impulsivity and dynamically inconsistent decision making - hyperbolic preferences and the cue theory of consumption - used to describe the demand for alcohol, also describe patterns of payday loan usage. In these models, individuals can objectively benefit from a restricted choice set that limits their access to loans and liquor. The overlap in behavioral characteristics of over-users of both products suggests that liquor sales is a reasonable and interesting place to test the effectiveness of payday lending regulations.

To identify the causal effect of lending restrictions on liquor sales, we exploit a change in payday lending laws in the State of Washington. Leveraging lender- and liquor store-level data, we estimate a difference-in-differences model comparing Washington to the neighboring State of Oregon, which did not experience a change in payday lending laws during this time. We find that the law change leads to a significant reduction in liquor sales, with the largest decreases occurring at liquor stores located very near to payday lenders at the time the law took effect. Our results provide compelling evidence on how credit constraints affect consumer spending, suggest a behavioral mechanism that may underlie some payday loan usage, and provide evidence that the Washington’s payday lending regulations reduced one form of loan misuse.

Background

Washington State enacted HB 1709 on January, 1st 2010, which introduced three new major restrictions to the payday loan industry. First the law limited the size of a payday loan to 30% of a person's monthly income or $700, whichever is less. Second the law created a state-wide database to track the issuance of payday loans in order to set a hard cap on the number of loans an individual could obtain in a twelve month period to eight, and eliminated multiple concurrent loans. This effectively prohibited the repayment of an existing loan with a new one. In the year prior to the law, the State of Washington estimated that roughly one third of all payday loan borrowers took out more than eight loans. Finally, the law mandated that borrowers were entitled to a 90 day instalment plan to pay back loans of $400 or less or 180 days for loans over $400.

The effect of the law on the industry was severe. There were 603 payday loan locations active in Washington in 2009 that were responsible for 3.24 million loans worth $1.366 billion according to Washington Division of Financial Institutions. In the year following the law change, the number of payday lenders dropped to 424, and loan volume fell to 1.09 million loans worth only $434 million. The following year the number of locations fell again to 256 with a loan volume of roughly 900,000 worth $330 million. Today there are fewer than 200 lenders in Washington and the total loan volume and value has stabilized close to the 2011 values.

A crucial feature of our estimation strategy involves accounting for potentially endogenous supply side factors that challenge efforts to separately identify changes in demand from the store response to the change. To do so, we focus on liquor control states, in which the state determines the number and location of liquor stores, the products offered, and harmonizes prices across stores to regulate and restrict liquor access. Oregon and Washington were both liquor control states until June of 2012 (Washington privatized liquor sales in June 2012).

Main Results

For this study, we use monthly store-level sales data provided by Oregon's and Washington's respective liquor control agencies from July 2008 through March 2012. Figure 4 plots estimated residuals from a regression of log liquor store sales on a set of store-by-month fixed effects, averaged over state and quarter. The graph possesses three notable features. First, prior to Washington's lending restrictions (indicated by the vertical dashed line), the states' log sales are trending in parallel, which confirming the plausibility of the ``common trends'' assumption of the DD model. Second, a persistent gap in the states' sales appears in the same quarter as the law change. This gap is the result of a relatively large downward movement in Washington's sales compared to Oregon's, consistent with a negative effect of the law on sales. Finally, the effect appears to be primarily a level shift as sales in both states maintain a common upward trend.

Our regression estimates indicate that the introduction of payday lending restrictions reduced liquor store sales by approximately 3.6% (statistically significant at the 1% level). As average Washington liquor sales were approximately $163,000 in the months prior to the law change, this represents a $5,900 decline per store each month. At the state level, the point estimate implies a $23.5 million dollar annual decrease in liquor sales. As Washington State reported that the law decreased payday loans by $932 million from 2009 to 2010, this decline represents approximately 2.5% of the change in total value of loans issued.

We see two primary explanations (not mutually exclusive) for the decline in Washington liquor sales in response to the law change. First, the effect may represent a wider permanent reduction in consumption as households lose their ability to cope with unforeseen negative income shocks. Alternatively, the drop in spending may indicate a more direct financing of liquor purchases by individuals with present-biased preferences. The first explanation implies that restrictions on payday lending negatively affect consumer welfare, while the second allows for a positive impact, since individuals with present-biased preferences may be made objectively better off with a restricted choice set.

Zinman (2013) highlights Laibson (2001) theory of Pavlovian cues as a particularly intriguing explanation for payday loan usage. In these models, consumer ``impulsivity'' makes instant gratification a special case during dynamic utility maximization, where exposure to a cue can explain dynamically inconsistent behavior. Indeed, Laibson uses liquor as a prime example of a consumption good thought to be influenced by cues, and subsequent experimental research on liquor uncovers evidence consistent with this hypothesis (MacKillop et al (2010)). In situations where payday lenders locate very near to liquor stores, individuals may be exposed to a cue for alcohol, and then see the lender as a means to satisfy the urge to make an immediate purchase. A lender and liquor store separated by even a brief walk may be far enough apart to allow an individual to resist the urge to obtain both the loan and liquor. Of course, cue-theory of consumption makes lender-liquor store distance relevant even in circumstances where individuals experience a cue only after borrowing. Lenders locating near liquor stores increase the likelihood that an individual exposed to a cue is financially liquid, and able to act on an impulse.

To investigate liquor store and lender proximity, we geocode the stores' and lenders' street addresses, and calculate walking distances for all liquor store-lender pairs within two kilometers of one another. We then repeatedly estimate our preferred specification with a full set of controls on an ever expanding window of liquor stores beginning with the stores that were located within a ten meter walking distance of a lender in the month prior to the law change, then within 100 meters, within 200 meters, etc., to two kilometres. These estimates are presented in Figure 5. The graph demonstrates a negative effect of 9.2% on those liquor stores that had a payday lender located within ten meters in the month before the law change (significant at the 1% levels), an effect almost three times as large as that overall.  The larger effect rapidly declines in distance suggesting that even a small degree of separation is significant. The degree of nonlinearity in the relationship between distance and liquor sales supports the behavioral explanation of demand.

Conclusion

Our analysis provides the first empirical evidence of the connection between payday lending and spending on liquor. We uncover a clear reduction in liquor sales resulting from payday lending restrictions. In addition, we find that those liquor stores located very near to lenders at the time of the law change experience declines in sales almost three times as large as the overall average.

This finding is significant because it highlights that a segment of borrowers may be willing to assume significant risk by borrowing in order to engage in alcohol consumption - an activity which carries significant personal risk of its own. The connection between payday lending restrictions and reduced liquor purchases, therefore, suggests that the benefits to payday lending restrictions extend beyond personal finance and may be large.

Effective payday loan regulation should recognize the potential for greater credit access to help or harm consumers. As Carrell and Zinman (2014) highlight, heterogeneity likely exists within the pool of payday loan users, and external factors will influence the ratio of ``productive and counter-productive borrowers.'' Lending restrictions can seek to reduce the proportion of counterproductive borrowers through the prohibition of practices known to harm consumers, including those that rely upon leveraging behavioral responses such as addiction and impulsivity. The behavioral overlap identified in the literature between counterproductive payday loan borrowers and heavy alcohol users suggests that there exists a link between the two markets. The decline in liquor sales documented here provides evidence that these regulations may be effective in promoting productive borrowing.

References

1. Carrell, Scott and Jonathan Zinman, “In harm's way? Payday loan access and military personnel performance," Review of Financial Studies, 2014, 27(9), 2805-2840.

2. Laibson, David, “A cue-theory of consumption," Quarterly Journal of Economics, 2001, pp. 81-119.

3. MacKillop, James, Sean O'Hagen, Stephen A Lisman, James G Murphy, Lara A Ray, Jennifer W Tidey, John E McGeary, and Peter M Monti, “Behavioral economic analysis of cue-elicited craving for alcohol," Addiction, 2010, 105 (9), 1599-1607.

4. Zinman, Jonathan, “Consumer Credit: Too Much or Too Little (or Just Right)?," Working Paper 19682, National Bureau of Economic Research November 2013.