Abstract
Subprime consumers often use small-dollar credit products, such as payday loans, to meet short-term financial needs over pay cycles. However, relatively little is known about the income sensitivity of demand for credit in this market. This paper provides a causal estimate of the effect of tax rebates on the demand for small-dollar credit, using a unique proprietary loan-level dataset. Identification relies on variation in state Earned Income Tax Credit (EITC) generosity for areas within the same commuting zones that span state borders. The results show that a $100 increase in EITC benefits leads to an 8.3% reduction in the number of loan applications and a 6.6% reduction in the number of borrowers. This could translate into sizable reductions in loan volume and savings in financial charges. More broadly, the results suggest that public programs with income benefits could help recipients with consumption smoothing in the presence of credit market frictions.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
Subprime borrowers typically have below-average credit scores and are charged with higher interest rates for loans or credit cards. Subprime borrowers are often identified by having a FICO credit score below 640, and this threshold varies slightly by lenders, across financial products, and over time.
This is a convenient way to compare the cost of short-term credit to other financial services, such as credit card and auto loans. However, using APR to measure the cost of short-term credit might not be entirely accurate. One might argue it is not reasonable to compound the charges of short-term credit over the period of a whole year.
Most subprime consumers have multiple jobs. They have incentives to report their income strategically to gain more access to credit and better loan terms. Common characteristics of subprime consumers include limited or inferior credit history, volatile income or employment history, and facing relatively high financial uncertainty. Consumer advocates often argue that small-dollar loans are too costly and should only be restricted to people who have the ability to pay. Accurate income measures and elasticity of demand for credit to income shocks would also be a key factor, when lenders build underwriting models and identify the population with sufficient ability to pay. For example, CFPB is considering imposing regulations on rollover or renewal of payday loans at national level, based on borrowers’ income or other characteristics (see proposal from CFPB at http://www.consumerfinance.gov/newsroom/cfpb-considers-proposal-to-end-payday-debt-traps/). Pew institute has proposed a 5% loan-to-income ratio as the benchmark for underwriting (see details from the Pew institute at http://www.pewtrusts.org/~/media/legacy/uploadedfiles/pcs_assets/2013/PewPaydayOverviewandRecommendationspdf).
These papers also highlight the importance of liquidity, especially for low-income families. Related studies can also be found in sociology. For example, Sykes et al. (2015) offers detailed accounts of the use of tax refunds through in-depth interviews.
Full report is available from the CFSA at http://cfsaa.com/Portals/0/cfsa2014_conference/Presentations/CFSA2014_THURSDAY_GeneralSession_JohnHecht_Stephens.
To get a loan, a borrower gives a payday lender a postdated check (e.g., dated on the borrower’s next payday) and receives cash right away. The borrower will pay the full loan amount and fees on the next payday. The lender then holds the check until the borrower’s next payday, which generally falls anywhere from less than a week to a month later.
Lanning et al. (2014) document that the payday loans are often rolled over or followed by another loan within 14 days.
See state-level regulations on payday loans from NCSL at http://www.ncsl.org/research/financial-services-and-commerce/payday-lending-state-statutes.aspx.
The annual percentage rate is the periodic interest rate applied to outstanding balances multiplied by the number of periods in a year. The finance charge is the total dollar amount of all interest payments. Other disclosures for payday loan transactions include the amount of the loan (amount financed), the total of payments (for payday loans, the check amount) and the schedule of payments.
Thirty-two states enacted safe harbor legislation for payday lenders and permit loans based on checks written from consumers’ bank accounts at triple digit interest rates, or with no rate cap at all. These states include: Alabama, Alaska, California, Delaware, Florida, Hawaii, Idaho, Illinois, Indiana, Iowa, Kansas, Kentucky, Louisiana, Michigan, Minnesota, Mississippi, Missouri, Nebraska, Nevada, New Mexico, North Dakota, Oklahoma, Rhode Island, South Carolina, South Dakota, Tennessee, Texas, Utah, Virginia, Washington, Wisconsin and Wyoming. See statutes from NCSL for details at http://www.ncsl.org/research/financial-services-and-commerce/payday-lending-state-statutes.aspx.
In the online small-dollar credit market, lenders can operate under two different legal models. They can choose to obtain state licenses and comply with state-level regulations, or they can choose to locate and register their company at tribal areas or offshore locations and bypass any regulations. State-licensed lenders make up the most part of the industry. In the baseline analysis, I focus on state-licensed lenders and provide results including offshore or tribal lenders as robustness checks.
Size limits frequently range between $300 and $500 per advance. Some states directly limit the size of the advance. Others limit the size of the check, which includes the amount of the advance plus the finance charge. Montana has a variation on size limits that restricts advances to the lesser of $300 or 25% of the customer’s net monthly income. Nevada limits the amount of the advance to one-third of the customer’s net monthly income. Many states also limit the aggregate amount of advances to a customer at a company, which is generally the same as the size of the maximum advance. The intent of these restrictions on nonprice terms is to force consumers to use payday loans for short-term needs and to keep the consumers from falling too far into debt.
For more details on construction of commuting zones, see documentations from USDA at https://catalog.data.gov/dataset/commuting-zones-and-labor-market-areas.
For most updated information on eligibility and generosity of EITC, see details from IRS at http://www.eitc.irs.gov/EITC-Central/abouteitc/ranges.
People who file both state and federal tax returns would get both federal and state EITC benefits, if eligible. In particular, most people use paid services (for example, H&R Block), when filing tax returns. The paid tax preparation services have strong incentives to encourage people to take up and file for both federal and state EITC benefits, if eligible.
I use CZs updated in 2000 and a crosswalk between ZIP code areas and CZs to identify borrowers’ location from loan-level data.
The Brookings Institution uses original aggregated-level data from the IRS. The difference is that the Brookings Institution re-assigned residence to more accurate and consistent ZIP code areas and counties. Data from the Brookings Institution also includes some additional information on EITC-eligible tax filers. See more details at http://www.brookings.edu/research/interactives/eitc.
Since the tax return data only include federal tax filings, I assume that people who file federal tax returns would also file state tax returns. This is a reasonable assumption for the following reasons. First, a substantial part of the population eligible for EITC use paid tax filing services, which would mostly likely assist taxpayers to file and claim any refunds both at the federal and state level. Second, people using electronic tax filing would file for both federal and state returns automatically. All states offer some form of e-filing options for taxpayers. Third, taxpayers would have strong incentives to file for both, if eligible for any public assistance.
In this paper, I restrict data to lenders who have been reporting to this credit bureau consistently throughout the period of study.
A comprehensive report of online small-dollar credit borrowers observed in data is available at https://www.nonprime101.com/wp-content/uploads/2015/02/Profiling-Internet-Small-Dollar-Lending-Final.
The payday loan application process does not involve a traditional credit check, and payday borrowing activity is not reported to the national credit bureaus, Equifax, Experian or TransUnion. This means that payday borrowing is not a factor that directly affects one’s traditional credit score. Although some alternative credit data vendors (for example, LexisNexis) cover this segment, it is not yet widely used for underwriting in traditional financial products. Thus, usage of small-dollar credit is unlikely to affect borrowers’ subsequent access to credit from other sources. If borrowers worry about the impact on credit scores when applying for payday loans, one might be concerned that demand for small-dollar credit could be correlated with the availability of other types of liquidity from more traditional sources or demand for credit in the future, as credit score is often used for underwriting for auto loans, mortgages, etc.
There are several advantages of focusing on loans from the online credit market. First, online credit is on average more expensive than the traditional storefront lenders. People who turn to online borrowing could be relatively more credit constrained. I will capture the population who have a high demand for credit. Second, another key advantage of focusing on the online credit market is to avoid confounding factors from the supply side. People in different locations basically have the same access to credit, regardless of their residential address and distance to locations of financial services. This allows me to ignore the traveling distance to the physical lending sites. This helps with the identification strategy, which relies on discontinuity in EITC generosity across state borders.
Storefront lenders are only allowed to operate in states that allow payday loans. However, for online lenders, there are two options. They can choose to be state-licensed and obey all state-level regulations. Or lenders could be incorporated or registered in tribal or offshore areas, which would allow them to operate even in states than ban payday loans. Due to the presence of tribal or offshore lenders, I observe loans originated for borrowers living in states that ban payday loans in the dataset. However, with the presence of these two types of lenders, we would expect the market conditions in states that ban payday loans to be very different from the other states.
According to the distribution of distance to the closest state borders for 5-digit ZIP code areas, 10 km and 20 km is about 25th percentile and median of the distribution, respectively.
R-squared is high (0.86) for the state-level analysis, and lower (about 0.4) for ZIP5 level. This makes sense since there would be more confounding factors at more detailed geographic levels (ZIP5 level), which leads to a lower value for the R-squared.
The measure of default here depends on lenders’ reporting. A caveat is that lenders could report “default” status for loans differently or selectively due to accounting purpose and lag in reporting time to credit bureau.
Data are available at http://www.federalreserve.gov/releases/housedebt/.
To measure state income tax, I use a national representative sample from the CPS, run the sample through TAXSIM and assign state of residence as Alabama, Alaska, and so on. Then, I calculate the average tax rate for each state.
Cutoffs for the deciles of average EITC benefits (in $) are 1906, 1990, 2057, 2132, 2164, 2194, 2231, 2306, and 2476. Cutoffs for the deciles of fraction of borrowers (in %) are .16, .29, .48, .66, .81, 1.06, 1.26, 1.80, and 2.56.
References
Aaronson D, Agarwal S, French E (2012) The spending and debt response to minimum wage hikes. Am Econ Rev 102(7):3111–3139
Ackerman RA, Fries G, Windle RA (2012) Changes in US family finances from 2007 to 2010: evidence from the survey of consumer finances. Fed Reserve Bull 100:1–80
Agarwal S, Chunlin L, Souleles NS (2007) The reaction of consumer spending and debt to tax rebates—evidence from consumer credit data. J Polit Econ 115(6):986–1019
Barrow L, McGranahan L (2000) The effects of the earned income credit on the seasonality of household expenditures. Natl Tax J 53(4):1211–1244
Bertrand M, Morse A (2009) What do high-interest borrowers do with their tax rebate? Am Econ Rev 99(2):418–423
Bhutta N, Skiba P, Tobacman J (2015) Payday loan choices and consequences. J Money Credit Bank 47(2–3):223–260
Carrell S, Zinman J (2008) In harms way?. Payday loan access and military personnel performance, Federal Reserve Bank of Philadelphia, Philadelphia
Caskey J (2010) Payday lending: new research and the big question. SSRN working paper series
Coulibaly B, Li G (2006) Do homeowners increase consumption after the last mortgage payment? An alternative test of the permanent income hypothesis. Rev Econ Stat 88(1):10–19
Currie J, Gruber J (1996) Health insurance eligibility, utilization of medical care, and child health. Q J Econ 111(2):431–466
Evans WN, Moore TJ (2012) Liquidity, economic activity, and mortality. Rev Econ Stat 94(2):400–418
Gross T, Tobacman J (2014) Dangerous liquidity and the demand for health care evidence from the 2008 stimulus payments. J Hum Resour 49(2):424–445
Gross T, Notowidigdo MJ, Wang J (2014) Liquidity constraints and consumer bankruptcy: evidence from tax rebates. Rev Econ Stat 96(3):431–443
Gruber J (1997) The consumption smoothing benefits of unemployment insurance. Am Econ Rev 87(1):192–205
Holt S et al (2009) Beyond lump sum: periodic payment of the earned income tax credit. The Brookings Institution, Washington
Karger H (2005) Shortchanged: life and debt in the fringe economy. Berrett-Koehler Publishers, Oakland
LaLumia S (2013) The EITC, tax refunds, and unemployment spells. Am Econ J Econ Policy 5(2):188–221
Lanning J, Leary J, Burke K, Wang J (2014) Payday lending, CFPB data point. Consumer Financial Protection Bureau, Washington
Manoli DS, Turner N (2014) Cash-on-hand & college enrollment: evidence from population tax data and policy nonlinearities. Working paper 19836, National Bureau of Economic Research, Jan 2014
Melzer BT (2011) The real costs of credit access: evidence from the payday lending market. Q J Econ 126(1):517–555
Michelmore K (2013) The effect of income on educational attainment: evidence from state earned income tax credit expansions. Available at SSRN 2356444
Morse A (2011) Payday lenders: heroes or villains? J Financ Econ 102(1):28–44
Parker JA, Souleles NS, Johnson DS, McClelland R (2013) Consumer spending and the economic stimulus payments of 2008. Am Econ Rev 103(6):2530–2553
Parsons CA, Van Wesep ED (2013) The timing of pay. J Financ Econ 109(2):373–397
Pew (2012) Payday lending in America: who borrows, where they borrow, and why. Pew Charitable Trusts Safe Small-Dollar Loans Research Project
Shapiro MD, Slemrod J (2003) Consumer response to tax rebates. Am Econ Rev 93(1):381–396
Skiba PM (2014) Tax rebates and the cycle of payday borrowing. Am Law Econ Rev 16(2):550–576
Stegman MA (2007) Payday lending. J Econ Perspect 21(1):169–190
Sykes J, Križ K, Edin K, Halpern-Meekin S (2015) Dignity and dreams what the earned income tax credit (EITC) means to low-income families. Am Sociol Rev 80(2):243–267
Acknowledgements
I am grateful to Marika Cabral and Day Manoli for advice and support, and to Clemens Sialm, Mike Geruso, Sandra Black, Rich Murphy, and seminar participants at the UT Austin for their comments. I would also like to thank Rick Hackett for helpful discussions and Clarity Services Inc. for data access for this project. All errors are my own.
Author information
Authors and Affiliations
Corresponding author
Appendix: Additional analysis
Appendix: Additional analysis
1.1 Reduced form analysis
I present the reduced form analysis as stated in Eq. 3 using the instrument variable directly in Table 15. I present the results on loan applications first. As EITC top-up rates increases by 10%, the number of borrowers decreased by 6.3%, and the number of loan applications decreased by 8.1%. For originated loans, number of borrowers with funded loans decreased by 7.1%, similar to the loan application results although the estimate is insignificant. This is likely due to the reduced sample size in the originated loan data set, as the majority of loan applications are denied or unfunded by lenders (Figs. 4, 5).
A specific example on the identification strategy Notes: ZIP5 areas on this map are located in two different states, New Mexico and Texas, but all belong to the same commuting zone. New Mexico offers 5% additional EITC benefits, while Texas does not offer any additional benefits. Assuming people in these areas are similar in demographic and socioeconomic characteristics, those in New Mexico would have additional income compared to the others. Using this variation in income due to generosity of state EITC benefits, I analyze whether people in New Mexico would have less demand for small-dollar credit
Distribution of t-statistics from a permutation test by randomly assigning generosity of state EITC benefits Notes: For all the states that allow payday loans, I assigned each state a random draw from the distribution of the simulated instrument and run the reduced-form specification using this randomly assigned instrument and I repeated this process 2000 times
a Distribution of borrowers for small-dollar loans across counties for year 2010 by decile, with darker color indicating more borrowers. b Distribution of average amount of EITC benefits received across counties for year 2010 by decile, with darker color indicating higher average EITC benefits. (Color figure online)
1.2 Cross-county distribution of EITC benefits and loan applications
Figure 6 shows the distribution of loan applications and the amount of EITC benefits received across counties. Figure (a) shows the fraction of borrowers I observed in loan-level data among the county-level population by decile, with the darker colors indicating higher fractions. Figure (b) shows the actual amount of EITC benefits received at county level by decile, with the darker colors indicating higher amounts. Comparing these two figures, we can see that higher EITC benefits and a higher density of borrowers are located in southern and western areas, which could be due to common local characteristics or economic shocks.Footnote 31
1.3 Additional checks on balance of characteristics across state borders
In Tables 12 and 13, I present additional checks on the balance of characteristics across state borders. Other than the demographic and socioeconomic characteristics, regulations in the small-dollar credit market could be different across states as well. I ran a similar specification using payday loan regulation as outcomes. The first panel uses an indicator for the presence of regulations on loan size, interest rate, etc. The second panel focuses on the specific regulations, for example, the maximum loan size or interest rate allowed. The results in Table 12 show that the regulations are uncorrelated with the instrument, the generosity of state EITC benefits. Additionally, one might also be concerned that people in different states might have different tax filing behaviors or use of financial services. I also show in Table 13 that the use of direct deposit and refund anticipation loans (RAL) or refund anticipation checks (RAC) is balanced across borders in the same commuting zone.
1.4 Additional placebo analysis
I implemented an additional placebo check to ensure that the estimates presented in previous sections are valid. First, I constructed placebo state boundaries to verify the estimates are robust to omitted characteristics of geographical areas. Specifically, I divide each CZ-state area in cross-border CZs into two pieces: the border area within 20 km of the state boundary and the remainder of the CZ-state area not directly bordering the neighboring state. Thus, the newly created “placebo boundary” is entirely inside the state. Using this “placebo boundary” as the state border, I assign the border area a counterfactual instrument (the simulated instrument for EITC benefits generosity) equal to that of neighboring state and keep the true value of the simulated instrument for the other area. Then, I run the same specifications using this newly created instrument and control for CZ-state fixed effects, instead of CZ fixed effects. The goal of this test is to show that the estimates are robust to omitted variables that could be trending geographically from one state to another. The results are reported in Table 14. As expected, estimates are insignificant, indicating that it is unlikely that other characterizes trending geographically confound the estimates (Tables 15, 16).
1.5 Additional analysis on default rate
I look at the default rate among all originated loans for the offshore/tribal lenders. Offshore/tribal lenders on average have a higher default rate than state-licensed lenders (32.7 vs. 26.3%). Table 17 shows that $100 additional EITC benefits reduce the default rate by 2.3 percentage points for offshore/tribal lenders. This is slightly higher than the effects on state-licensed lenders shown in the text.
1.6 Loan volume by the date of origination
I look at the pattern of loan volume by date of origination. Figure 7 shows the percentage of loans originated for each week. The first vertical line indicates beginning of February, which is the peak of EITC filing (see LaLumia 2013). The second vertical line is right after Apr 15, which is due date of tax filing, though extensions are allowed. There is clearly a reduction in loan volume during the tax refunds season, suggesting that people borrow less when they have additional liquidity from EITC tax refunds.
Pattern of loan volume by the date of originationNotes: The vertical axis is the percentage of loan originated in a certain calendar week. The horizontal axis shows the calendar week. The first vertical line indicates beginning of February, which is the peak of EITC filing (see LaLumia 2013). The second vertical line is right after April 15, which is due date of tax filing, though extensions are allowed. The gray horizontal line indicates the percentage of loans each week if origination was evenly spread out through out the whole year
Rights and permissions
About this article
Cite this article
Cui, C. Cash-on-hand and demand for credit. Empir Econ 52, 1007–1039 (2017). https://doi.org/10.1007/s00181-016-1213-2
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s00181-016-1213-2