Shale shocked: Cash windfalls and household debt repayment

https://doi.org/10.1016/j.jfineco.2022.09.008Get rights and content

Abstract

Using individual credit bureau data matched with cash windfalls from fracking, we estimate that windfall recipients reduce debt-to-income by 2.4 percentage points relative to no-windfall controls. Debt repayment effects are 3 times stronger for subprime individuals than for prime individuals. Based on the timing of upfront versus continuing cash payments, debt repayment coincides with the timing of payments but not with news about future payments. These findings present a challenge for purely forward-looking models of debt. Indeed, when we incorporate a windfall shock into a forward-looking model, the model predicts an increase in debt that runs counter to our evidence of debt repayment.

Introduction

Conventional economic models have difficulty rationalizing how much debt is held by households (Laibson, Repetto, Tobacman, 2003, Bertaut, Haliassos, Reiter, 2009). For instance, many households revolve high-cost unsecured debts (Gross and Souleles, 2002), the use of which has increased over time and across many countries (Müller, 2018). Moreover, the wealthiest individuals often hold the most of all types of debts (Mason, 2018), which suggests that debtholders are not often financially constrained. Motivated by households’ puzzling debt behaviors, we study how households repay or accumulate debt after they receive unexpected cash windfalls.

We identify the effects of unanticipated cash windfalls by studying debt choices for a novel sample of individuals who receive unexpected and continuing payments from the discovery and extraction of natural gas at the onset of the Fracking Revolution. Specifically, we study how $1.2 billion in royalty windfalls over 11 years affect the debt repayment decisions of 215,639 individuals who own subsurface mineral rights in the Barnett Shale in Texas. The cash windfalls in our sample vary widely, ranging from $78 in the first percentile to $46,245 in the 99th percentile. Further, our sample of windfall recipients includes individuals who live in every U.S. state. The nonlocal individuals in our sample receive payments because subsurface rights are separate from surface rights (e.g., land), and are inherited through generational family ties.

With this uniquely detailed information, we attribute variation in household debt repayment to the cash received by individuals, which in our setting, is distinct from changes to local conditions after the discovery of shale. Further, our sample contains people in every credit score category, including 43,829 subprime individuals. Our sample has significant overlap with demographics of the U.S. population. Crucially, these payments were unexpected at the beginning of our sample. And, the timing and amount of payments are primarily driven by factors external to the individuals we study because payments are determined by the price of natural gas, the number of wells drilled, and production stemming from those wells. In this way, our setting enables us to address the most salient endogeneity issues in estimating the impact of cash inflows on debt repayment, as well as understanding sources of heterogeneity in debt repayment decisions.

Our main difference-in-differences specification evaluates how unexpected cash windfalls affect recipients’ debt, measured as debt payments over income (DTI) or annualized debt payments.1 We estimate that receiving cash windfalls leads individuals to reduce DTI by 2.4 percentage points and reduce annual debt payments by $912 on average. This is substantial debt repayment relative to the average cash windfall of $5684 in our sample. We estimate dynamic treatment effects and find that treatment and control individuals exhibit parallel trends in debt outcomes prior to the first cash payments. After receiving the first windfall, treatment individuals gradually reduce debt relative to control individuals. We estimate that recipients of cash windfalls reduce DTI (annual debt payments) by 0.76 percentage points ($257) in the year after first payment, but by 3.1 percentage points ($1338) as of six years after payments begin. These differences level off beyond six years after the first cash payment.

There is substantial heterogeneity in debt repayment that depends on the initial creditworthiness of individuals receiving the unexpected windfall. We consider heterogeneity by whether an individual is initially subprime (credit score <620) versus initially prime (credit score >720). For subprime individuals, we estimate that cash windfalls reduce DTI by 3.47 percentage points, whereas for prime individuals, cash windfalls reduce DTI by only 1.22 percentage points. We observe similar heterogeneity by subprime versus prime individuals when we employ annual debt payments as the dependent variable, or when we employ a treatment intensity specification that also uses variation in the amount of cash windfalls. Thus, the discrepancy between subprime and prime debt repayment is unexplained by differences in incomes, nor is it explained by the two samples receiving different amounts of cash payments. Linking back to our motivation, our finding that windfall recipients repay rather than accumulate debt out of cash windfalls is consistent with households behaving as though they were liquidity constrained when they receive the cash payments. Our finding that debt repayment is stronger among subprime individuals, who are likely more constrained in accessing credit markets than prime individuals,2 further reinforces the view that debt repayment is driven by variation in financial constraints.3

Next, we evaluate whether liquidity or wealth channels can explain the debt repayment effects of cash windfalls. To do so, we exploit the fact that cash payments from shale arrive in two components – an upfront bonus payment received when mineral rights are leased, and continuing royalty payments that begin once the well begins to produce. There are at least 3 years between the upfront bonus payment and the beginning of production for 40% of cash windfall recipients. For this subsample, we can separately observe the effect around the bonus payment date from the first royalty date. Using an event window centered on the first royalty payment year, we observe the average effect of an upfront bonus payment is to decrease DTI by approximately 0.5 percentage points with little change in DTI until after royalty payments begin. Most debt repayment occurs gradually after royalty payments begin (up to a 2.4 percentage reduction by 6 years after the first royalty payment). This effect takes the form of a trend break that leads DTI to diverge over time as payments continue. These findings imply a tight link between debt repayment out of cash windfalls in the years when payments are received, but a weaker link between debt repayment and the arrival of news about future cash flows (bonus payment date and first royalty payment date). Overall, these findings point strongly to a liquidity mechanism rather than a wealth mechanism that drives debt repayment.

Another potential mechanism is that debt repayment could be driven by changes to the local economy caused by fracking activity rather than the cash windfalls directly. To confront this explanation, we exploit another novel feature of our data – that many windfall recipients are not local to the shale extraction area. When we repeat the analysis on only nonlocal individuals and their matched controls, we obtain similar estimates for the effect of cash windfalls on debt repayment, despite reducing the sample size by approximately 90%. Collectively, these findings point to an individual-level impact of cash windfalls on debt repayment. The effects are uniformly stronger for initially subprime individuals and are tightly linked to the time periods in which people receive cash payments.

Motivated by these findings in the annual panel, we quantify the amount of long-term debt repayment using a treatment intensity difference-in-differences specification. To estimate the long-run marginal propensity to repay debt out of cash windfalls, the treatment intensity in this specification is the size of the windfalls scaled by income (i.e., payments/income, annualized), which places the payments on the same scale as household leverage (debt payments/income). Overall, we estimate the marginal propensity to repay debt (MPR) out of the mineral rights windfalls to be 0.33. This implies that 33% of an additional dollar of mineral payments goes toward repaying existing debts for the average individual in our sample. In these specifications, we also control for the amount of mineral acres owned by the individual, as well as ZIP3-year fixed effects and individual fixed effects. Our estimated magnitudes are insensitive to the choice of control variables. We also repeat the analysis using only the nonlocal windfall recipients, obtaining a similar elasticity. Similar to the evidence at the annual frequency, we find that there is substantial heterogeneity in the MPR, with initially subprime individuals exhibiting a MPR that is roughly 5 times the MPR for the prime individuals. Specifically, we estimate that 77% of the wealth shock is used for debt repayment for subprime individuals, but only 14% is used for debt repayment for prime individuals.

Digging deeper into the difference between subprime and prime debt repayment, we examine how cash windfalls affect debt balances for different categories of debt. These specifications attribute much of the overall difference between subprime and prime debt repayment to different effects of cash flows on revolving debt balances (e.g., credit cards). For subprime individuals, a standard deviation increase in payments (1.42 pp) leads to roughly a 5.5% reduction in revolving debt ($462 on an initial base of $8345). By contrast, prime individuals increase revolving debt balances by approximately 7% (nearly $600) in response to a standard deviation increase in cash windfall. These differences are consistent with subprime individuals using the cash flow shock to repay revolving debts, whereas prime individuals increase their monthly credit card balances, presumably because they increase consumption.

Finally, we relate our findings to a dynamic model of consumption in which consumers can hold unsecured debt (Fulford, 2015). Departing from standard models of intertemporal consumption choice, the key feature of Fulford (2015) is that individuals hedge against the uncertainty of future borrowing limits by consuming out of debt and carrying high-cost debt balances across periods of time. Individuals borrow in good times because they are unable to do so during downturns when credit markets tighten.

In the model, exogenous variation in credit limits is the crucial factor that motivates how much individuals consume out of cash or out of debt. As such, to align our setting with the model, we need to estimate shocks to credit limits and use such estimates for the model calibration. We find that the average royalty recipient experiences a 30% increase in their unsecured credit limits. Moreover, we use the consumption response to cash windfalls to further validate the model. In our setting, we estimate individuals’ consumption response to cash windfalls in addition to our main analysis of debt repayment. Because the credit bureau data does not contain direct measures of consumption, we estimate the effects on automobile purchases inferred from individuals’ credit histories (Benmelech et al., 2017 and Dupor et al., 2019 validate this measure by showing that changes in automobile credit lines tracks actual automobile purchases). Our estimates confirm that subprime individuals exhibit a positive and significant sensitivity of automobile consumption to cash flow shocks, whereas prime individuals do not.

Within the Fulford (2015) model, we estimate consumption and debt responses to increased borrowing limits that result from cash windfalls. We find that the model generates a positive consumption response to cash windfalls, which is consistent with the reduced form estimates on automobile consumption. However, the model predicts that individuals should accumulate debt after the windfall, which contrasts sharply with our finding that windfall recipients repay debt. This finding stems from the forward-looking aspect of the model in which consumers borrow more when credit lines are offered to them as a precaution against reduced credit access in the future. As our tests around the timing of upfront bonus versus continuing royalty payments show, the pattern of debt repayment out of cash windfalls is generically inconsistent with a forward-looking mechanism. Together, this evidence suggests that the debt repayment effects we see are due to the liquidity benefits of cash. Given this, there is no minimal departure from the workhorse model that can replicate our findings on debt, despite matching consumption changes.

The primary contribution of our paper is to provide evidence on how household debt depends on unexpected cashflows. We contribute to the literature on determinants of household leverage (Bailey, Dávila, Kuchler, Stroebel, 2019, Barrot, Loualiche, Plosser, Sauvagnat, 2017, Brown, Cookson, Heimer, 2019, Kalda, 2020) and also to the literature that attempts to understand why households hold as much debt as they do (e.g., Gathergood, Weber, 2014, Gorbachev, Luengo-Prado, 2019, Olafsson, Gathergood, 2021). Our research most closely relates to work that connects household budget shocks to household leverage choices. For example, previous work has examined factors that affect household liquidity, such as resets to adjustable rate mortgages (Fuster, Willen, 2017, Di Maggio, Kermani, Keys, Piskorski, Ramcharan, Seru, Yao, 2017) and government policies targeted to mortgage owners (Scharlemann, Shore, 2016, Abel, Fuster, 2020), or changes to the value of household assets via house price appreciation (Cloyne et al., 2019). Our paper shows the debt repayment effects of persistent cash transfers.4 Interestingly, most of these papers find that households increase debt in response to positive shocks to their budget constraint – e.g., through a collateral channel mechanism (Cloyne et al., 2019). In contrast, we find that liquid cash windfalls lead to debt repayment, with greater debt repayment for subpopulations who are ex ante credit constrained. This finding suggests that subtle differences in the way the household budget constraint is shocked can have qualitatively different implications.

In addition, our findings emphasize the pervasive role of heterogeneity in creditworthiness. Our finding that subprime consumers repay more debt than prime consumers is consistent with the perspective that debt repayment is motivated, in part, by liquidity considerations. There are fundamental differences between our paper and prior research on debt repayment. Agarwal et al. (2007) and Agarwal and Qian (2014) study debt repayment out of one-time income shocks, which are qualitatively different than the unexpected and persistent stream of royalties from shale natural gas extraction. Dynamic theories of consumption choice predict different responses depending on whether the shocks are transitory or persistent. As such, our analysis of initially unexpected and continuing cash flows provides guidance for policies that involve recurring payments, such as extended unemployment insurance or a universal basic income, while prior work is best suited to study one-time transfers, such as tax rebates or stimulus checks. Similar to our work, some papers estimate the effects of sustained shocks to wealth on household leverage, using fiscal shocks (Demyanyk et al., 2019) and shocks to import competition (Barrot et al., 2017). However, these papers rely on regional aggregates, which cannot distinguish local economic effects (e.g., employment opportunities) from the effect of the shock. Our analysis of household debt repayment is distinct from this literature in at least three ways. First, our individual-level tests more cleanly distinguish debt repayment out of individual cash payments from local area confounding variation, such as changes to employment opportunities. Second, our individually matched data enable us to evaluate rich individual-level heterogeneity. Third, the uneven timing of bonus versus royalty payments allows us to speak to the importance of timing of payments separately from news about future payments.

Furthermore, our findings provide a fresh perspective on whether unexpected cash windfalls primarily affect households via their liquidity effects versus their wealth effects. Because of the intrinsic connection between household debt and consumption choices, our paper relates to literature that studies the consumption response to cashflow shocks. According to the canonical permanent income hypothesis (e.g., Jappelli and Pistaferri, 2010), cash windfalls influence household decisions through how their arrival affects expectations – i.e., the present value of wealth. Our evidence that the timing of cash inflows, not news about future cash inflows, matters for debt repayment provides novel evidence that household debt decisions are governed by liquidity motives as well as debt. In this respect, our research relates to the literature on how consumption behaviors deviate from the permanent income hypothesis. For example, papers have argued for the role of credit and liquidity constraints (e.g., Zeldes, 1989, Baker, 2018) and a litany of behavioral rules (e.g., Campbell, Mankiw, 1989, Kueng, 2018). When applied to policy, recent papers estimate the consumption response over several years to government spending (Dupor et al., 2019), tax rebates (Sahm, Shapiro, Slemrod, 2015, Baugh, Ben-David, Park, Parker, 2018), housing variables (e.g., Berger, Guerrieri, Lorenzoni, Vavra, 2017, Ganong, Noel, 2018), and employment (e.g., Ganong and Noel, 2019). Despite this growing body of research, popular perspectives on household leverage (e.g., Mason, 2018, Laibson, Repetto, Tobacman, 2003, Kaplan, Violante, 2014), and workhorse models that motivate consumer debt choices (e.g., Laibson, Repetto, Tobacman, 2003, Fulford, 2015) maintain that debt is primarily driven by forward-looking behavior. Like the related work in consumption behavior, our findings on debt repayment and its timing with respect to cash inflows present an instructive counterpoint to the forward-looking view of household balance sheets. Importantly, our evidence on debt behaviors contrasts with the forward-looking view, despite consumption behavior in our setting being rationalizable in a forward looking model. Our results suggest that there is value to taking into account both consumption and debt responses.

Finally, our paper relates to a growing literature on the economic effects of shale development. Existing literature documents that natural gas shale development in specific localities has led to job growth (Feyrer et al., 2017), lending (Gilje, Loutskina, Strahan, 2016, Gilje, 2019), and changes in house prices (Muehlenbachs et al., 2017). Our work advances this literature by using individual level data on cash payments from shale extraction to trace out the effects of shale development on household outcomes.5 More closely related to our study, Brown (2021) examines the local effects of the oil and gas boom in the U.S. on consumer debt accumulation, while some papers study the effects on financial distress (Heimer, Stehulak, Treanor, 2016, Haughwout, Lee, Scally, van der Klaauw, 2016, Cunningham, Gerardi, Shen, 2021). Some of these studies use the FRBNY - CCP/Equifax panel data set to measure credit outcomes. However, these studies rely on local aggregates of drilling intensity, rather than trace out the effects of individual cash royalty payments. As such, our paper is in a unique position to evaluate the role of ex-ante heterogeneity in households’ balance sheets and consider the effects on nonlocal beneficiaries of the Shale boom.

Section snippets

Data and institutional setting

The analysis uses several data sets that are novel to the literature. Below we outline the data sources underlying our empirical tests and their construction.

Main results

In this section, we provide evidence on how wealth shocks from mineral payments affect household leverage. In addition to estimating how cash windfalls affect the propensity to repay debt, the tests in this section also provide evidence on the heterogeneity in the repayment propensities with respect to initial creditworthiness.

Connection to theory

In this section, we attempt to reconcile the findings in the prior section with an intertemporal model of consumption and debt choice. There is a vast literature on consumption models that studies intertemporal choices of savings versus consumption (see e.g., Jappelli and Pistaferri, 2010). A longstanding puzzle in this literature is to explain the amount of high-cost unsecured debt held by households (Gross, Souleles, 2002, Laibson, Repetto, Tobacman, 2003). Since our focus is on how debt is

Conclusion

We study the propensity to repay debt out of cash windfalls using novel data on royalty payments to individuals from the extraction of shale natural gas. We find that cash windfalls lead to debt repayment. Windfall recipients repay more debt in the years when they receive the cash inflows, not necessarily upon receiving news about future cash windfalls, a finding that suggests liquidity constraints affect debt repayment. Subprime individuals repay more debt, consistent with these households

References (63)

  • J. Abel et al.

    How do mortgage refinances affect debt, default, and spending? Evidence from HARP

    Am. Econ. J.: Macroeconomics

    (2020)
  • S. Agarwal et al.

    The reaction of consumer spending and debt to tax rebates – evidence from consumer credit data

    J. Polit. Economy

    (2007)
  • S. Agarwal et al.

    Consumption and debt response to unanticipated income shocks: evidence from a natural experiment in Singapore

    Am. Econ. Rev.

    (2014)
  • B.S. Argyle et al.

    Monthly payment targeting and the demand for maturity

    Rev. Financ. Stud.

    (2020)
  • M. Bailey et al.

    House price beliefs and mortgage leverage choice

    Rev. Econ. Stud.

    (2019)
  • A.C. Baker et al.

    How much should we trust staggered difference-in-differences estimates?

    J. Financ. Econ.

    (2022)
  • S.R. Baker

    Debt and the response to household income shocks: validation and application of linked financial account data

    J. Polit. Economy

    (2018)
  • J.-N. Barrot et al.

    Import Competition and Household Debt

    FRB of New York Staff Reports 822

    (2017)
  • B. Baugh et al.

    Asymmetric Consumption Response of Households to Positive and Negative Anticipated Cash Flows

    Technical Report

    (2018)
  • A. Bellon et al.

    Personal wealth, self-employment, and business ownership

    Rev. Financ. Stud.

    (2021)
  • E. Benmelech et al.

    The real effects of liquidity during the financial crisis: evidence from automobiles

    Q. J. Econ.

    (2017)
  • D. Berger et al.

    House prices and consumer spending

    Rev. Econ. Stud.

    (2017)
  • Bermejo, V. J., Ferreira, M. A., Wolfenzon, D., Zambrana, R., 2020. Entrepreneurship and regional windfall gains:...
  • C.C. Bertaut et al.

    Credit card debt puzzles and debt revolvers for self control

    Rev. Financ.

    (2009)
  • J.S. Briggs et al.

    Windfall gains and stock market participation

    J. Financ. Econ.

    (2020)
  • J.P. Brown

    Response of consumer debt to income shocks: the case of energy booms and busts

    J. Money Credit Bank.

    (2021)
  • J.R. Brown et al.

    Growing up without finance

    J. Financ. Econ.

    (2019)
  • B. Callaway et al.

    Difference-in-differences with multiple time periods

    J. Econom.

    (2021)
  • J.Y. Campbell et al.

    Consumption, income, and interest rates: reinterpreting the time series evidence

    NBER Macroecon. Annu.

    (1989)
  • C.D. Carroll et al.

    The nature of precautionary wealth

    J. Monet. Econ.

    (1997)
  • C. Cunningham et al.

    The double trigger for mortgage default: evidence from the fracking boom

    Manage. Sci.

    (2021)
  • Cited by (3)

    • Coal busts and urban recovery: Evidence from China

      2023, Regional Science and Urban Economics
    • The Impact Mechanism of Household Financial Debt on Physical Health in China

      2023, International Journal of Environmental Research and Public Health

    Toni Whited was the editor for this article. We thank the following sources for providing funding support for this project: Wharton Dean’s Research Fund, the Wharton Alternative Investments Initiative, the Rodney L. White Center for Financial Research, the Jacobs Levy Equity Management Center for Quantitative Financial Research, the National Bureau of Economic Research Household Finance Working Group and the Sloan Foundation, and the Ewing Marion Kauffman Foundation. In addition, we are grateful to individual feedback from an anonymous referee, Asaf Bernstein, John Campbell, Sudheer Chava, Henrik Cronqvist, Julia Fonseca, Andreas Fuster, Julio Galvez, Ali Hortacsu, Sasha Indarte, Mark Jansen, Yaron Levi, Yueran Ma, Mike Palazzolo, Mikael Passo, Carlos Parra, Matthew Plosser, Toni Whited, and Vincent Yao, as well as the participants at seminars and workshops at the University of California-Berkeley, Boston College, CEPR Household Finance Virtual Seminar Series, FRB-Chicago, FRB-Kansas City, Notre Dame, Penn State University, University of Kansas, Depaul University, University of Colorado–Boulder, Emory University, PUC Chile, University of Delaware, and feedback from conference presentations at the 2018 University of Chicago Empirical Asset Pricing Conference, the 2018 London Business School Summer Symposium (Early Ideas), the 2018 FRB-Philadelphia Conference on Consumer Behavior in Credit and Payments Markets, the 2019 Boulder Summer Conference on Consumer Financial Decision Making, the 2019 Frontiers in Finance Conference, the 2019 Cornell IBHF Symposium, the 2019 University of Kentucky Finance Conference, the 2019 UNC-Duke Corporate Finance Conference, the 2019 FIRS Conference, the 2019 Helsinki Finance Summit, the 2019 European Finance Association Conference, the 2019 Denver University Summer Conference, the 2019 Northern Finance Association Conference, the 2019 Washington University Finance Conference, the 2019 Tel Aviv University Finance Conference, the 2020 CFP Board Academic Research Colloquium, the 2020 SFS Cavalcade Conference, and the 2021 Western Finance Association Conference. The paper was previously circulated as ”Shale Shocked: The Long Run Effect of Wealth on Household Debt.”

    View full text