Comment on "The Impact of Housing Markets on Consumer Debt"

Transcription

1 Federal Reserve Board From the SelectedWorks of Karen M. Pence March, 2015 Comment on "The Impact of Housing Markets on Consumer Debt" Karen M. Pence Available at:

2 KAREN PENCE Discussion of Brown, Stein, and Zafar HOUSEHOLD WEALTH AND THE composition of household balance sheets changed dramatically during the 2000s. House prices, as measured by the CoreLogic index, rose 113% from 1999 to 2006 before plunging 31% from 2006 to Home equity lines of credit (HELOCs) held on the balance sheets of banks quintupled between 1999 and the beginning of 2009, but have been steadily contracting at an annual rate of nearly 5% since then and now stand 26% below their peak. 1 Credit card and auto debt rose more modestly than HELOCs about 45% each between 1999 and the mid-2000s. 2 Both types of debt contracted thereafter; auto loans have rebounded, but credit card debt remains 12% below its peak. Student loans outstanding, meanwhile, have surged in recent years and registered a 150% increase between 2006 and the second quarter of The article by Brown, Stein, and Zafar (2015) explores the relationship between the changes in house prices and consumer debt, defined here to include HELOCs, credit cards, auto loans, and student loans, but not other loans collateralized by real estate. The authors draw upon the New York Federal Reserve Consumer Credit Panel, sourced from Equifax, which includes detailed information on the debt portfolios of around 5% of Americans with credit records. The authors link a house price index, measured at the level of the individual s ZIP code, to each credit record. These rich data allow the authors to unpack some of the individual-level dynamics underlying the aggregate changes in house prices and household debt. The authors consider three different time periods: the prehousing boom ( ), the boom ( ), and the bust ( ). They identify homeowners as those individuals with home-secured debt on their credit files at the beginning of I am grateful to Brett McCully for excellent research assistance. KAREN PENCE is at the Federal Reserve Board ( Karen.pence@frb.gov). Received October 27, 2014; and accepted in revised form October 27, Assets and Liabilities of Commercial Banks in the United States. H.8 Statistical Release, October Consumer Credit. G.19 Statistical Release, October 7. g19/current/default.htm. Journal of Money, Credit and Banking, Supplement to Vol. 47, No. 1 (March April 2015) Published This article is a U.S. Government work and is in the public domain in the USA

3 216 : MONEY, CREDIT AND BANKING each period. They relate the changes in several components of debt over each of these periods to the ZIP-code-level changes in house prices, the county-level average income, the county-level change in employment, and the change in the borrower s credit score. The authors are concerned that an unobserved factor such as an improvement in employment prospects might underlie both a change in house prices and a change in consumer debt. To control for this possibility, they instrument for house prices using a supply-elasticity measure developed by Albert Saiz and a house-price volatility measure developed by Christopher Palmer. These instruments are intended to capture the part of house price movements that are exogenous to local economic conditions and other factors. To demonstrate the robustness of their results, the authors also develop a second approach: a difference-in-difference specification that compares owners and renters in high-house-price-appreciation ZIP codes to owners and renters in low-appreciation ZIP codes. This specification assumes that the double-differencing controls for any such unobserved factors. The results demonstrate that the positive relationship suggested in the aggregate data between home prices and HELOC balances is a widespread phenomenon. In the authors data, homeowners living in ZIP codes where house prices rose substantially also increased their HELOC balances significantly. Likewise, homeowners HELOC balances contracted in ZIP codes where house prices decreased. This relationship is fairly constant across the three time periods that the authors examine, and is also apparent across different age and credit score buckets. The relationship between house prices and other forms of consumer credit, however, varied across time, subgroups, and type of credit. In high-house-priceappreciation areas in the period, credit card debt decreased for all age and credit score buckets, and auto debt decreased for homeowners with prime credit scores. In the period, credit card balances did not vary much with house prices, whereas auto loan balances increased for all homeowners, particularly those with near-prime and subprime scores. In the period, in areas where house prices fell, auto loan and credit card balances contracted for near-prime and subprime borrowers, whereas prime borrowers appear to have increased their credit card balances a bit. The authors are not able to find any robust patterns between house price changes and student loan balances, despite exploring a variety of creative empirical approaches. Although a relationship between house prices and HELOC balances is intuitive, it is less clear why consumer debt should shift with house prices, and why that relationship might change over time. To fix ideas, consider three pathways by which a house price increase could affect consumer debt. First, an increase affects household wealth and thereby household spending. Households, as always, will finance some of that spending with debt. Second, house price changes affect a borrower s credit worthiness and the willingness of lenders to extend credit. As a result, households may have a greater ability to borrow. Third, house price changes affect the availability and desirability of financing consumption with an HELOC. Households may substitute HELOCs for other forms of debt.

4 KAREN PENCE : 217 The period provides support for the substitution story: HELOC balances increased, but auto loan balances (for prime borrowers) and credit card balances decreased. The and periods provide support for the wealth effects and credit constraints stories: HELOC, credit card (for nonprime and subprime borrowers), and auto loan balances increased, and then decreased, with the boom and bust in house prices. The fact that auto loan and credit card balances expanded more for subprime and nonprime than prime borrowers during the boom is further evidence for the credit constraints story. These results spark further questions, however: what is the mechanism by which the substitution occurred between HELOCs and consumer debt, and by which borrowing constraints eased? For example, auto loans might decrease if borrowers drew upon their HELOCs to purchase cars, if borrowers paid down outstanding auto debt with HELOC proceeds, or if borrowers simply had more cash on hand, and less need for an auto loan, because other demands on their budgets were financed with HE- LOCs. Likewise, auto loans might increase because auto lenders loosened their terms and standards in areas where house prices rose, or because an easing of mortgage credit constraints enabled households to spend more on automobiles. The question is particularly interesting in the case of autos because of recent work by Mian, Rao, and Sufi (2013) and Mian and Sufi (2014) that shows a strong relationship between auto sales and house price changes. To shed some light on these issues, I show tabulations from special questions that the Federal Reserve fields on the Reuters/Michigan Survey of Consumers (Table 1). Since 2003, the Federal Reserve has asked these survey respondents whether they have purchased an automobile in the previous 6 months, and if so, how they paid for the purchases. The Federal Reserve also asked respondents whether they have done a cash-out mortgage refinancing in the previous 6 months, and if so, what they did with the proceeds. I assume here that borrowers use the proceeds of a cash-out refinancing and an HELOC in roughly similar ways, although one salient distinction is that a borrower can withdraw home equity multiple times with an HELOC, but only once with any given cash-out refinancing. The results indicate that about 13% of households purchased a new or used automobile in the 6 months preceding the survey. About half of these respondents financed their purchase with an auto loan, about a third used cash from savings or investments, and about 10% used cash from another source. During the housing boom, about 3% used an HELOC or the proceeds from a cash-out refinancing to purchase a car; that number dropped to 1% during the bust. A much smaller share of households about 2% refinanced their homes and took out cash in the 6 months before the survey. Not surprisingly, this share is higher during the housing boom than the bust. Around half used the proceeds to pay down debt, around a third used the proceeds for home improvements, 10% bought a car or another big expense, and a small number paid for educational or medical expenses. Canner, Dynan, and Passmore (2002) document similar patterns in these same data for the period.

5 218 : MONEY, CREDIT AND BANKING TABLE 1 AUTO PURCHASES FINANCED WITH HOME EQUITY Years All Share with an auto purchase 14% 13% 13% Share who funded the purchase with: Auto loan 54% 49% 51% Cash from savings or investment 30% 36% 34% Cash from HELOC 2% 1% 2% Cash from mortgage refinancing 1% 0% 0% Cash from other 13% 13% 13% N 683 1,370 2,053 Share with a cash-out mortgage refinancing 3% 1% 2% Share who used the proceeds to: Pay down debt 54% 47% 51% Home improvements 37% 32% 35% Car or other big expense 12% 7% 10% Educational or medical expense 4% 2% 3% Business expense 1% 4% 3% Other 14% 14% 14% N SOURCE: Author s tabulations from the Reuters/Michigan Surveys of Consumers. Taken together, these tabulations suggest that borrowers typically fund vehicles and education at least initially with the loan products that are designed for these purchases. A small number of borrowers appear to use home equity extraction products for these purchases, and this share was larger during the housing boom than the bust. The more typical means of substitution between home equity and consumer credit appears to be paying off existing debt. Bhutta and Keys (2014), however, suggest that homeowners are much more likely to pay off credit card debt than auto debt with the proceeds from home equity extraction. These tabulations do not address the question of whether auto lenders eased or tightened terms in response to changes in house prices. In other work with the Reuters/Michigan data, Johnson, Pence, and Vine (2014) show that households in higher-house-price appreciation areas are more likely to report that it is a good time to buy a car because interest rates are low. Very little is known on this topic otherwise, and more research would be helpful. The authors make two data decisions that may affect their results. The first is that they include HELOCs in their empirical work, but not closed-end home equity loans or cash-out refinancings. The authors chose this definition because they believe that it is too difficult, for these other types of loans, to isolate the part of the loan used for consumption instead of housing services. Bhutta and Keys (2014), however, show that only about 35% of home equity is extracted through HELOCs, 45% comes from cash-out refinancings, 10% from second liens, and 10% from other products. In addition, HELOCs are primarily originated to high-credit-score borrowers (Lee, Mayer, and Tracey 2012). (See figure 3 in the paper.) By choosing this more restrictive

6 KAREN PENCE : 219 TABLE 2 HOMEOWNERS WITH HOME-SECURED DEBT Share of homeowners with home-secured debt Share of renters who own homes All families 65% 66% 71% 41% 42% 39% Age of head (years) Less than 35 85% 89% 92% 9% 7% 5% % 88% 90% 21% 20% 16% % 78% 85% 38% 41% 34% % 59% 68% 61% 67% 58% % 39% 50% 75% 74% 75% 75 or more 15% 12% 18% 74% 74% 73% SOURCE: Survey of Consumer Finances. Estimates derived from nominal.xls. definition, the authors may be overstating the role of substitution in the debt dynamics of the 2000s, especially among subprime borrowers. The second decision is that the authors define anyone with home-secured debt on the credit records, including first and second liens and HELOCs, as a homeowner. Many homeowners, however, do not have debt secured by their homes. As shown by the first set of columns in Table 2, 65 70% of homeowners in the 1998, 2002, and 2007 Surveys of Consumer Finances corresponding roughly to the beginning of the time periods examined by the authors held mortgages. The authors definition of renter is an individual without such debt on the credit record, which combines true renters with homeowners with no mortgage debt. As indicated by the second set of columns in Table 2, around 40% of renters, under that definition, are actually homeowners. This misclassification may affect the authors results, as homeowners without mortgages will likely respond differently to house price changes than homeowners with mortgages. For example, homeowners without mortgages may be less likely to be credit constrained, as well as less likely to want to rebalance their debt portfolios. However, such homeowners may still want to change their consumption because of the wealth effects. Although the authors cannot identify these homeowners without debt in their data, one possible solution is to limit the sample to younger individuals, perhaps around 50 years or younger. The results in Table 2 suggest that the share of households with mortgages drops dramatically after this age, and the renter sample becomes majority homeowner. Overall, though, this paper makes an important contribution toward our understanding of the household debt dynamics of the 2000s. As the authors demonstrate with their rich and comprehensive data, there was considerable heterogeneity in how homeowners adjusted their debt portfolios in response to house price changes. Prime borrowers, for example, were more likely to rebalance their debt portfolios in response to house price increases, whereas subprime and nonprime borrowers were more likely to expand their overall borrowing. Understanding the role that household-level

7 220 : MONEY, CREDIT AND BANKING dynamics play in macroeconomic aggregates will continue to be a fruitful avenue for research. LITERATURE CITED Bhutta, Neil, and Benjamin Keys. (2014) Interest Rates and Equity Extraction during the Housing Boom. The University of Chicago Law School Kreisman Working Papers on Housing Law and Policy. Canner, Glenn, Karen Dynan, and Wayne Passmore. (2002) Mortgage Refinancing in 2001 and Early Federal Reserve Bulletin, 88, Johnson, Kathleen, Karen Pence, and Daniel Vine. (2014) Auto Sales and Credit Supply. Finance and Economics Discussion Series Paper No Lee, Donghoon, Christopher Mayer, and Joseph Tracey. (2012) A New Look at Second Liens. Federal Reserve Bank of New York Staff Reports, 569, August. (accessed October 24, 2014). Mian, Atif, Kamalesh Rao, and Amir Sufi. (2013) Household Balance Sheets, Consumption, and the Economic Slump. Quarterly Journal of Economics, 128, Mian, Atif, and Amir Sufi. (2014) House Price Gains and U.S. Household Spending from 2002 to Fama-Miller Working Paper, University of Chicago.