AdobeStock_143772480.jpeg
VoxEU Column Global crisis Microeconomic regulation

The secular decline of co-borrowers in the mortgage market is not harmless

Mortgage markets are dynamic in nature, which sometimes comes at a cost. This column shows that over the last few decades, the US mortgage market experienced a secular decline in co-borrowers. Having a co-borrower minimises the exposure and effects of adverse income shocks and thus should enhance mortgage performance. The authors show that this yet unexplored decline in co-borrowers therefore has non-trivial implications for the financial stability of the mortgage market and regional economic outcomes. 

Since the 1990s, developments in technology and regulation have transformed the US mortgage market. Lenders have become more sophisticated both in terms of underwriting credit and financing it (Mian et al. 2010, Fostel and Geanakoplos 2012, Favara and Imbs 2015, Foote et al. 2018). These changes on the supply side contributed to changes in borrower characteristics and affected the stability of the mortgage market. For instance, before 2007 a significant portion of credit was issued to risky borrowers, which played an important role in causing the financial crisis (Mian and Sufi 2009).

A vast literature has documented changes in the mortgage market over the last few decades and the implications of those changes, but it has largely ignored one ubiquitous aspect, namely the presence of a co-borrower, i.e. an additional borrower responsible for the repayment of the mortgage. Since the early 1990s, the number of mortgages with a co-borrower has declined sharply. In recent research (Jakucionyte and Singh 2020), we compute the co-borrower share as a ratio of total loan amount with a co-borrower to the total loan amount over the period 1990-2016. Figure 1 shows that the co-borrower share has declined from about 75% in 1990 to 40% in 2006. 

In this column we discuss the evidence that a low share of co-borrowers in local mortgage markets is associated with lower financial stability, worse credit market outcomes, and lower housing prices. 

Figure 1 Share of mortgage applications with a co-borrower

Source: HMDA dataset 1990-2016
Note: The sample is restricted to conventional mortgage applications made with the intent to purchase an owner occupied house. The figure shows the trend since 1990 for all applications (solid blue line) and only accepted applications (dashed red line). The co-borrower share is defined as the ratio of loan amount associated with mortgage applications with a co-borrower and total loan amount in a particular year. All nominal values are deflated by the Consumer Price Index (CPI), base year 2000.

Implications for mortgage default

Co-borrowers are likely to have close personal ties or even belong to the same household. So, having a co-borrower should allow for better smoothing of adverse shocks, unless the income of the borrower and her co-borrower are perfectly correlated. Bhutta et al. (2017), Schelkle (2018) and others provide evidence that borrowers are considerably more likely to default on underwater mortgages if negative home equity coincides with adverse shocks such as unemployment or sickness. This is also known as the double-trigger hypothesis. Having a co-borrower minimises the exposure and the effects of adverse income shocks and thus should enhance mortgage performance as predicted by the double-trigger hypothesis. 

We confirm the performance differential for mortgages with and without co-borrowers by estimating a mortgage default model. Figure 2a plots the estimated co-borrower effect on the probability of mortgage default. Regardless of the origination year and thus potentially time-varying lending standards, the presence of co-borrower affects the probability of default negatively. For instance, mortgages with a co-borrower originated in 2004 were 2% less likely to default than mortgages without a co-borrower, which constitutes 63% of average default probability in this cohort. 

Figure 2a Co-borrower coefficient 

Figure 2b Co-borrower coefficient, prime and subprime samples

Source: FMFM dataset 2000 -- 2015 
Note: Panel (a) plots the co-borrower estimate obtained from year-by-year regressions. Panel (b) plots the co-borrower estimate obtained from year-by-year regressions for two samples: subprime loans and prime loans. Subprime (prime) loans are loans with a FICO credit score of less (greater) than 660. The blue shaded area shows 95% CI.

Two other observations follow. First, the co-borrower effect is time-varying and is largest during the Global Crisis. Compared to mortgages originated in 2004, the presence of a co-borrower reduces the default probability for mortgages originated in 2007 twice as much – by 4%. Since in economic downturns the occurrence of adverse shocks is the largest, this result upholds the role of the double-trigger hypothesis. Second, the co-borrower effect is not restricted to subprime mortgages. Figure 2b shows that the co-borrower effect estimated in the sample of prime mortgages and the sample of subprime mortgages separately is statistically significant and negative in both samples.

Regional implications for housing prices and credit market outcomes

The decreasing share of mortgages with a co-borrower raises the question of financial stability in general, but it turns out that the co-borrower share is unevenly distributed across the regions in the US. Since 1990 the rate of decline in the co-borrower share varied across different regions and thus the co-borrower share diverged regionally. For instance, Figure 3 shows that the measures of dispersion of the co-borrower share at the county level has gradually increased over time and reached its peak in 2006. Could the diverging presence of co-borrowers have contributed to regionally diverse economic outcomes? We answer positively.

Figure 3 Measures of dispersion

Source: HMDA dataset 1990-2016

Extrapolating the relationship between the presence of a co-borrower and the default probability suggests that regions with a lower co-borrower share should have higher default rates. Moreover, the regionally diverse effects on other economic outcomes should follow. We focus on year 2006 as the point of highest dispersion in the co-borrower share and the year that also precedes the Global Crisis. The economic downturn that followed allows us to capture the effects on financial stability and other outcomes more clearly. Precisely, we show that regions with a lower co-borrower share in 2006 had lower credit growth, lower refinancing growth, and house price growth over 2007-10.

To show this, we estimate a county-level regression model that relates regional outcomes to the co-borrower share prior to the crisis. The co-borrower share is measured as an indicator with value one if the co-borrower share in the county in 2006 was below the median. Such counties are labelled here as counties with a lower co-borrower share. We interact the co-borrower share indicator with year effects to follow the dynamics of the effect. All estimations obtain significant estimates for the co-borrower effect relating a lower co-borrower share at the county level to lower house price growth, lower mortgage credit growth, and lower refinancing growth.

Figure 4 plots the obtained estimates for house price growth, mortgage credit growth and refinancing growth in the county-level regression. Since we interact the co-borrower share indicator with time effects, the plot shows the estimates for these interactions for different origination years. In addition, the estimates are normalised to year 2006. The figure shows that before 2006 credit market outcomes and house price growth closely track each other in both samples of counties, but diverge after 2006. House price growth, mortgage credit growth and refinancing growth are significantly lower in counties with a lower co-borrower share and this difference in all presented outcomes persist at least for three years. The difference is economically significant as it is close to one third of a standard deviation of the respective outcomes. For instance, house price growth in counties with a lower co-borrower share in 2007 is lower by approximately 3%, which constitutes 34% of the standard deviation of house price growth. 

Figure 4 County-level outcomes before and after the financial crisis

a) House price growth

b) Purchase mortgage growth

c) Refinance mortgage growth

Source: HMDA dataset 2001--2010, Federal Housing Finance Agency's House Price Index Dataset 2001--2010 
Note: The figure reports the economic outcomes of low co-borrower share counties relative to the high co-borrower share counties before and after the financial crisis with 2006 as the base year. Time varying county controls are included. The solid line reports the actual estimate and the shaded region refers to 95% confidence interval. 

Conclusions

The presented stylised and empirical facts show that the presence of co-borrowers in the mortgage market has important implications. It can explain a substantial share of mortgage defaults, especially in economic downturns. Further, the uneven presence of co-borrowers across different regions before 2007 is associated with differences in house price growth, mortgage credit growth, and refinancing growth during the crisis. Thus, the regionally diverse presence of co-borrowers seems to be yet another factor explaining regional differences. So far, we have known about the effects arising from the geographic distribution of home equity only (Mian et al. 2013, Mian and Sufi 2014, Beraja et al. 2019).

However, it is still an open question why the co-borrower share declined so dramatically in the last three decades. As shown in Figure 1, within a span of one and a half decades, the co-borrower share declined by 35%. The magnitude of the decline rules out demographic factors as the only drivers of the change. An array of factors including changes in securitisation is most likely contributing to this decline.

References

Beraja, M, A Fuster, E Hurst, E, and J Vavra (2019), “Regional heterogeneity and the refinancing channel of monetary policy”, The Quarterly Journal of Economics 134(1): 109-183.

Bhutta, N, J Dokko, and H Shan (2017), “Consumer ruthlessness and mortgage default during the 2007 to 2009 housing bust”, The Journal of Finance 72(6): 2433-2466.

Favara, G, and J Imbs (2015), “Credit supply and the price of housing”, American Economic Review 105(3): 958-92.

Foote, C L, L Loewenstein, and P Willen (2018), “Technological innovation in mortgage underwriting and the growth in credit: 1985–2015”, FRB of Cleveland Working Paper No. 18-16.

Fostel, A, and J Geanakoplos (2012), “Tranching, CDS, and asset prices: How financial innovation can cause bubbles and crashes”, American Economic Journal: Macroeconomics 4(1): 190-225. 

Jakucionyte, E, and S Singh (2020), “Bowling Alone, Buying Alone: The Decline of Co-borrowers in the US Mortgage Market”, Bank of Lithuania Working Paper No. 78.

Mian, A, and A Sufi (2009), “The consequences of mortgage credit expansion: Evidence from the US mortgage default crisis”, The Quarterly Journal of Economics 124(4): 1449-1496 (see on VoxEU).

Mian, A, and A Sufi (2014), “What explains the 2007–2009 drop in employment?”, Econometrica 82(6): 2197-2223.

Mian, A, K Rao, and A Sufi (2013), “Household balance sheets, consumption, and the economic slump”, The Quarterly Journal of Economics 128(4): 1687-1726.

Mian, A, A Sufi, and F Trebbi (2010), “The political economy of the US mortgage default crisis”, American Economic Review 100(5): 1967-1998 (see on VoxEU).

Schelkle, T (2018), “Mortgage default during the US mortgage crisis”, Journal of Money, Credit and Banking 50(6): 1101-1137.

210 Reads