VoxEU Column Global crisis

How did household balance sheets affect consumption during the Great Recession?

The dramatic fall in consumption during the Great Recession was accompanied by an equally dramatic increase in household debt in the years preceding it. This column examines the relationship between household debt and consumption behaviour, and the channels through which this link operates. The column concludes that the relationship is driven almost entirely by the presence of financial constraints, such as liquidity or borrowing limits.

The presence of substantial amounts of household debt in 2007 has prompted many economists and policymakers to link debt to the depth of the recession in the following years. The possibility that higher levels of household debt induce deeper or longer recessions has important implications for both financial regulation and the size of the social safety net. More broadly, a better understanding of the dynamic relationship between a household's spending decisions, income process, and balance sheet is imperative to accurately describe microeconomic drivers of business cycles.

In recent years, economists such as Mian and Sufi (2013) have noted that areas with high household debt saw relatively poorer economic performance during the 2007-09 recession. Others have suggested that households became more debt-averse, leading to consumption cutbacks among the most indebted households. In my Job Market Paper (Baker 2013), I seek to clarify this relationship using novel household financial data, and to more precisely isolate the channels that drive it.

Leveraging household-level financial data

Despite the importance of understanding the nature of the relationship between household balance sheets and consumption behaviour, clear analysis is often difficult to due to endogeneity concerns, as well as limited data covering the entirety of household finances. For this paper, I use data of an anonymous set of users of a large personal financial website that allows these users to link all of their financial accounts to a single location. Thus, I am able to observe income and spending as well as assets, debt, and credit across all of a household’s accounts in a high-frequency setting.

Care must be taken into eliminating or controlling for households that have not linked all of their financial accounts, and to reweigh sample households by demographic characteristics and location in order to match the distribution of households across the country. After making these adjustments, I find that observable household financial characteristics like retail spending and asset and debt holdings match the distributions observed across all American households. Overall, the data represent a huge new potential source of disaggregated financial information covering millions of households and billions of transactions, allowing for a more thorough investigation of a wide range of questions regarding household finance and consumer behaviour.

Using natural language processing techniques on the text descriptions accompanying direct deposit transactions, I am able to link households to large and publicly traded employers.

  • With these links, I can exploit shocks, such as large and surprising earnings reports, merger announcements, or layoff announcements that occur at a firm level as exogenous drivers of household income. I find that these shocks are unanticipated by households and cause persistent changes in household income.

Similarly, I employ variation in geographically determined housing supply elasticity to isolate exogenous changes in debt accumulated by households at the beginning of my sample period based on where the household was located. For instance, the housing supply in Miami, which is bounded on all sides by water or the Everglades, is highly inelastic, but the housing supply in Houston, which can expand in virtually any direction, is highly elastic.

  • In the years leading up to the Great Recession, housing prices rose much more in areas with inelastic housing supplies, leading households in those areas to take on higher levels of debt through mortgage or home equity loans.

Debt and the household consumption response to income shocks

With this framework, I am able to highlight a strong relationship between measures of household debt and the responsiveness of consumption to changes in income.

  • In particular, I find that a household with a one standard-deviation higher level of debt adjusts its consumption by 25% more than the median household in response to an unexpected change in its income.

This result is robust to controlling for a myriad of household-level demographic or financial controls as well as when instrumenting for both changes in income and for measures of debt.

While my findings suggest a strong link between debt and household consumption behaviour, I am able to attribute this link to two types of financial constraints: liquidity and credit constraints. The impact of financial constraints on household behaviour has been noted in recent theoretical frameworks posed by Kaplan and Violante (2013) and Huntley and Michelangeli (2014), as well as a range of empirical work such as that of Johnson, Parker, and Souleles (2006). In my paper, I confirm the importance of liquidity and credit constraints, demonstrating that they can account for the heterogeneity in consumption responses to income shocks across households in recent years.

Separately considering impacts of liquid and illiquid savings against debt, I find much larger effects of liquid savings on consumption behaviour. In other words, converting one dollar of housing wealth to one dollar in a savings account moves households farther from liquidity constraints and reduces the sensitivity of consumption to changes in income. In addition, I find that households that have greater access to credit – in the form of room to borrow on credit cards or possessing high credit scores – are also less impacted by higher levels of debt.

Implications during the Great Recession

These results have important implications for household financial behaviour during periods akin to the Great Recession. The presence of substantial declines in asset prices as well as a sudden tightening of the credit supply drove many households close to borrowing constraints. Consistent with recent theoretical work by Eggertsson and Krugman (2012), my findings suggest that these constraints greatly exacerbated the declines in consumption seen across the US. While the Federal Reserve’s potential responses were limited due to the zero lower bound during the recent recession, other temporary government interventions designed to boost household credit access or liquidity may be useful tools to limit declines in aggregate demand.

References

Baker, Scott R (2013), “Debt and the Consumption Response to Household Income Shocks” Stanford University Job Market Paper. 

Eggertsson, Gauti B. and Paul Krugman (2012), “Debt, Deleveraging, and the Liquidity Trap: A Fisher-Minsky-Koo approach”, Quarterly Journal of Economics.

Huntley , Jonathan and Valentina Michelangeli (2014), "Can Tax Rebates Stimulate Consumption Spending in a Life-Cycle Model?" American Economic Journal: Macroeconomics.

Johnson DS, Parker JA, Souleles NS (2006), “Household expenditure and the income tax rebates of 2001”, American Economic Review.

Kaplan, Greg and Giovanni Violante (2013), “A Model of the Consumption Response to Fiscal Stimulus Payments”, Econometrica (forthcoming).

Mian, Atif R., Kamelesh Rao and Amir Sufi (2013), “Household Balance Sheets, Consumption, and the Economic Slump”, Quarterly Journal of Economics.

1,995 Reads