Collateral Damage: The Impact of Mortgage Debt on U.S. Savings

Abstract

This article contributes to the literature on saving by empirically investigating the determinants of the saving rate in the United States, with a special focus on the role of mortgage debt. Using data from 1987 to 2013, we find that mortgage payments have a substantial negative impact on both personal and private saving rates in the United States. An increase of 10 percentage points in mortgage payments leads to a 9.1-percentage-point drop in the personal saving rate and a 12.4-percentage-point drop in the private saving rate. In addition, including mortgage debt as an explanatory variable leads to significant changes in the impact of other variables, which further reinforces our claim that mortgage debt is important for the analysis of the saving rate. Comparing mortgage payments with nonmortgage consumer debt payments, we find that mortgage payments have a larger impact on the private saving rate whereas nonmortgage consumer debt payments have a larger impact on the personal saving rate. We also find a partial but robust crowding-out effect of public saving rate on the two saving rates. Our results have implications for monetary policy and government policies that encourage mortgage borrowing.

Keywords:

• Saving
• household saving
• mortgage debt

Acknowledgments

We thank the seminar participants at the University of Southern California, the Central Bank of Republic of Turkey, 2014 AREUEA-ASSA Meeting, and 2014 Midwest Macro Conference for their helpful comments.

Notes

1. Dynan and Kohn (2007) claim that the rise in household debt in the United States, including mortgage debt, is due to the increase in house prices, financial innovations, and demographic changes. Barnes and Young (2003) list shocks to interest rate, income, and expectations as the additional explanations for the rise in household debt.

2. If the borrower takes out an adjustable mortgage loan, she will face uncertainty about the mortgage payments in future years. However, this uncertainty is still less significant than the price uncertainty she would face in the absence of a mortgage loan.

3. Although mortgage financing existed in the United States long before the beginning of our sample period, this argument applies to the impact of the magnitude of access to mortgage financing, as well as the impact of availability of mortgage financing, on saving behavior. An increase in the mortgage debt/GDP ratio represents easier access to mortgage financing, including home equity loans, and possibly lower down-payment requirements. When mortgage financing goes up, as a percentage of GDP, we should see, at least partially, a similar impact on savings as when mortgage financing is introduced the first time.

4. Aiyagari (1994) reports that for sufficiently high variability and persistency in earnings, the aggregate saving rate could be higher by as much as 7 to 14 percentage points.

5. It can be argued that mortgage borrowing increases construction activity and real estate investment. Furthermore, it is also possible that a dollar increase in construction investment and real estate investment adds more to the economy than the loss because of a dollar decrease in other types of investment activity. This important question is outside the scope of the current article.

6. Carroll, Overland, and Weil (2000) incorporate habit formation in consumption into a standard growth model to show that high growth leads to high saving, not the other way around.

7. Mortgage equity withdrawal is defined as the amount of equity that the borrower extracts from the underlying asset when the asset price appreciates.

8. There is also a growing literature on the cross-country determinants of personal and private saving rates. An important study by Loayza, Schmidt-Hebbel, and Servén (2000), using panel data of 69 countries from 1965 through 1994, shows that private saving rate exhibits inertia; public savings partially crowd out private savings; and income, inflation, dependency ratio, and financial deepness have a sizable impact on the private saving rate. Other examples include Ang (2011), Athukorala and Sen (2004), Choi, Lugauer, and Mark (2014), Edwards (1996), Harris, Loundes, and Webster (2002), Hondroyiannis (2006), Horioka and Wan (2007), Masson, Bayoumi, and Samiei (1998), Ozcan, Gunay, and Ertac (2010), and Paiva and Jahan (2003).

9. There are also studies that investigate saving behavior of individuals using micro level data. A recent summary of this literature can be found in Cronqvist and Siegel (2015).

10. This precautionary motive for saving has attracted great deal of attention in the literature. Examples include Skinner (1988) and Zeldes (1989). In more recent studies, Alan, Crossley, and Low (2012), Carroll, Slacalek, and Sommer (2012), Fogli and Perri (2015), Gourio (2012), and Nakamura, Steinson, Barro, and Ursua (2013) report empirical evidence that saving rates increase with more macroeconomic volatility and risk. One exception is Aizenman, Cavallo, and Noy (2015) who find a negative correlation between risk and the private saving rate in cross-country comparisons, particularly in developing countries. They attribute their result to weaker institutions and informal labor markets in those countries, and argue that household saving decisions in such economies are intertwined with firms’ investment decisions.

11. The mortgage interest rate is the National Average Contract Mortgage Rate provided by the Federal Housing Finance Agency (FHFA).

12. http://www.nber.org/taxsim

13. We perform a Wald test to study whether there is a true structural break in the series of the saving rate in 2008Q1. The test results suggest rejection of the null hypothesis of no structural break. It is also worth mentioning that the results are very similar if we do not split the sample but instead interact the mortgage payment variable with a dummy variable that takes the value of 1 for the 2008Q1–2013Q3 period, and 0 otherwise.

14. One possible explanation is that high-income households were more likely to recover from the recession and qualify to obtain a mortgage to purchase a home, and their saving behavior is less sensitive to mortgage borrowing than that of low-income households is.

15. We also tested for Granger causality between mortgage payments and the two saving rates. We find that both before- and after-tax mortgage payments Granger-cause each of the two saving rates, whereas the reverse is generally not true.