Risks and Incentives in Underserved Mortgage Markets

Brent W. Ambrose

University of Pennsylvania Wharton School of Business, University of Wisconsin, Milwaukee

and

William N. Goetzmann(1)

Yale School of Management

November 1, 1996

Abstract

Subsidized loans may help increase home ownership in low income neighborhoods with positive social benefits, however there are risks and costs to the homeowners themselves. Home ownership increases incentives to maintain property and neighborhood, as well as decreasing the outflow of rents from low-income zones. These benefits, however are not costless to participants. With a mortgage comes the possibility of a default, the financial demands of maintenance, the reduction in alternate investment opportunities, an increased exposure to fluctuations in local economic conditions, and a drastic reduction in the liquidity of personal wealth. In this paper we examine the role of the owner-occupied house in the asset allocation decision of a family living in an area characterized as a low income neighborhood. We find that the current subsidies are likely to be too low relative to the costs. In particular, the tax law makes home ownership relatively less attractive to low-income families. This may explain a lack of home-ownership and thus, mortgage lending in low-income neighborhoods.

Introduction

In response to the Federal Housing Enterprise Safety and Soundness Act of 1992 (the 1992 GSE legislation) requiring Fannie Mae and Freddie Mac to meet performance goals with respect to affordable housing, the U.S. Department of Housing and Urban Development has targeted certain low-income and minority neighborhoods as deserving additional mortgage funding from the GSEs. These areas are defined as "Underserved" mortgage markets. Recent research by Carr, Megbolugbe, Cho and Buist (1994) provide empirical support for this designation. They find a strong correlation between a common measure of underservice, the ratio of mortgage originations to housing units, to income and racial characteristics of census tracts. The implication, consistent with efforts to identify evidence of mortgage discrimination, is that lenders are denying mortgage credit or 'redlining' areas with high concentrations of low-income or minority populations. However, strong correlation is not sufficient to establish causation. Thus the issue of whether individuals living in those areas are denied access to the mortgage market or fail to apply due to a lack of demand remains unclear. Rather than engage in the debate surrounding mortgage discrimination, we focus on the implications and likely outcomes of using these targeted lending programs.

Low mortgage originations in low-income neighborhoods, whether due to discriminatory lending practices or lower demand for home loans, presumably reflect a low rate of owner-occupied housing. The rational behind the 1992 GSE legislation which requires Fannie Mae and Freddie Mac to meet goals with respect to underserved markets is that increasing the proportion of owner-occupied housing is widely regarded as socially beneficial, particularly in low-income neighborhoods. First, it aligns the interest of the home-owner with the interest of society in maintaining and improving individual properties and neighborhoods. Home-owners have a stake in the value of their home, and will take actions to maintain and increase that value. Improving home values are in turn a reflection of improving conditions in low-income neighborhoods in general. Second, homeownership represents an opportunity for low-income families to build wealth through investment in an asset.

A home typically represents a large proportion of a family investment portfolio. For example, Kennickell and Shack-Marquez (1992) report that in 1989 the median net worth of US homeowners was $97,000, of which $70,000 represented their principal residence. For low-income families, it might represent its entirety -- at least before non-traded "assets" like human capital are considered.(2)

Given the high transaction costs associated with housing, it is not unreasonable to expect that many low-income families may be unable to purchase a home.

Like all assets, home-ownership has risk and return characteristics that influence the degree to which it is an attractive investment. While it might be attractive to some, it is probably not attractive for all. Home-owner's equity in a mortgaged property fluctuates dramatically as local housing prices change. Home-owners have only a limited ability to improve property values in the face of declining city-wide or regional economic conditions. While homes may be good investments (c.f. Goetzmann, 1993), they have certain drawbacks as well. Thus, homes are typically illiquid assets with high transactions costs. Perhaps most importantly, they are not portable. Home-owners are captive to the risk and cost of neighborhood conditions. Crime, education, city services and retail all vary within metropolitan areas. Home-ownership is a pre-commitment to a geographical exposure to these variables.

In this paper, we examine the drawbacks and benefits of home ownership to potential "Underserved" home-owners in low-income neighborhoods. We find that the drawbacks in many cases may outweigh the benefits. The implication of our analysis is that regulatory agencies should not be surprised to find lower demand for and hence lower rates of mortgage originations in low-income neighborhoods. As a result, more aggressive lending, without attempts to insure against risks associated with home ownership are likely to be ineffectual. In fact, it may lead to "crowding-out" of low-income families by higher income home-owners for whom these risks are more acceptable. This paper does not directly address the question of whether low mortgage originations are due to discriminatory lending practices or due to lower demand, rather we demonstrate that lower demand for mortgage originations is rational for persons living in areas defined as "underserved".

Agency Issues in Owner-Occupied Housing

There is a broad consensus that owner-occupied housing is desirable. There is theoretical and empirical support for the proposition that people who own their houses take care of them. In addition, homeowners make good citizens, since improved public services increase housing prices. The theoretical justification for this idea is that when an agent holds an equity share in an asset, he is motivated through self-interest to take actions to improve the value of that asset. This is the justification given for executive stock option programs, for instance.(3)

Boards of directors compensate CEOs with a package of salary and stock options not merely to make them rich, but to align their incentives more closely to the incentives of shareholders. The more their livelihood depends upon the share price of the company, the more they will work to keep that share price high.

By the same token, encouraging citizens in low-income neighborhoods to own an equity stake in their city will encourage actions that will increase property values city-wide.(4)

Presumably, anything that makes owner-occupied housing cheaper or easier to purchase will have positive social benefits, not only for the home-owner, but for city occupants as a whole. Indeed, this agency argument provides the theoretical justification for targeted lending in undesirable neighborhoods. It may make sense from a social policy perspective to provide below-market rates to home-owners because the benefits of this action could far outweigh the cost. What kind of target lending would make sense from the potential homeowner's perspective? Is "fairness in lending" sufficient to induce people to buy their own homes in low-income neighborhoods? To examine these questions, we consider the house purchase decision from the perspective of the potential home-owner.

Nest Eggs and Down Payments

Housing is both a consumption and an investment decision. Within the context of housing as an investment, the household home purchase choice is an asset allocation decision. Every household, large or small faces financial trade-offs. For instance, should savings be invested in housing, education or financial assets? If the household has no savings but can borrow, how should this credit capacity be used? The optimal allocation of financial resources depends upon the current and future needs of the family, as well as upon characteristics of the assets themselves. Goetzmann (1993) examines housing as an investment and finds that in four major metropolitan markets, Atlanta, Chicago, Dallas and San Francisco, housing represents a significant proportion of the minimum variance investment portfolio for families considering the optimal mixtures of stocks, bonds and the home. While investment characteristics of housing differ across and within cities and through time, a home owned free and clear generally has low risk ( 9% to 13% annual standard deviation) and an expected return between that of stocks and bonds (7% to 12% annual mean return over the period 1971 to 1985). This would appear to make the home an attractive investment -- even for risk-averse investors. The attractiveness of the home investment is not only due to the risk - return ratio. It is also due to the fact that housing has low correlation to traditional investment assets such as stocks and bonds. The home is a good hedge against fluctuations in the financial markets, and vice-versa.

The relative attractiveness of housing as an investment changes when taxes, maintenance, imputed rent and mortgage are considered. Using data inputs from Goetzmann (1993) we can construct an efficient frontier of investment portfolios formed of stocks, bonds and home equity under the assumption that complete credit availability in low-income areas would result in these areas matching the overall MSA appreciation rate.(5)

For instance, under the assumption of 80% financing and maximum use of the tax shield due to the deductibility of the home mortgage, the expected return and the risk of the house investment both increase. Figure 1 shows the range of efficient portfolios. The minimum variance portfolio is the leftmost point in the efficient frontier. It is composed of 34% stocks, 41% bonds and 25% home. In other words a diversified portfolio of financial and real estate assets provides the investor with minimal standard deviation of investment returns. This may be interpreted as an investment policy mix. Suppose a family had $10,000 saved and they wished to choose a safe portfolio. This analysis would recommend $3,400 in stocks, $4,100 in bonds and $2,500 in the home.

Low-income families are likely to be low-wealth families and thus, our analysis should also reflect their very real investment constraints. According to the Census Department's 1993 Survey of Asset Ownership, 25.3% of persons in the lowest income quintile had a negative net worth while 24.6% had a net worth between $1 and $4,999. Thus approximately 50% of all low-income individuals in 1993 a net worth below $5,000. In contrast, only 5.9% of all individuals in the highest income quintile had a net worth below $5,000. Individuals with net wealth at these levels may have trouble finding affordable housing to purchase. For example, with 80% financing, the $10,000 wealth constraint and the optimal policy mix above would imply a capped housing investment of $12,500.(6)

There may be no homes for sale for that price! Reducing the required down payment may allow the family to buy a more expensive home, but it also changes the efficient frontier. With additional leverage, expected returns and risk for the housing asset both increase, with the result being a reduction in the proportion of the portfolio represented by the home. The increased leverage can be thought of as the bank lending families more money to speculate in the local housing market. The ability to lever up is only attractive if the family is not risk averse.

Suppose a family with a modest savings of $10,000 placed all of their money in a home. Note that requiring 100% of the family wealth be invested in the home increases the volatility of their investment portfolio from 7.5% to over 19% per year! This is a huge difference in risk, despite the fact that the sums of money are relatively small. From the perspective of a low-income family seeking to preserve and grow their modest nest egg, this risk differential should not be taken lightly. Saving just enough money to buy a home might not be an optimal investment strategy, unless one is prepared to expose the family savings to significant risk, or alternatively, to buy a home outright with no mortgage.(7)

Are lower-income families more or less risk-averse than average? This is difficult to answer without empirical evidence, however unless they have consistently lower levels of risk aversion we can presume that they will not consistently choose the rightmost points along the efficient frontier. Economic theory tells us that investors will demand compensation for taking on greater levels of risk. The greater the risk-aversion, the greater is the demanded compensation.

Will Targeted Lending Work?

In the above analysis, we calculated the minimum variance portfolio and noted the historically attractive returns available from housing using returns from a MSA-wide housing index. However, recent research has indicated that house price appreciation rates can vary dramatically across neighborhoods within an MSA. For example, Smith and Tesarek (1991) found divergent price appreciation rates across 21 Houston submarkets based on initial home values. Similarly, Knight, Sirmans and Turnbull (1996) also find significantly different price appreciation rates across neighborhood locations for Baton Rouge, Louisiana. In a formal analysis of price appreciation rates correlated with race using neighborhood price indexes created from the 1994 City of Milwaukee Master Property (MPROP) file, Kim (1995) found significant variation across neighborhoods (defined at the census tract level). For example, Kim reports that the average annual nominal Milwaukee neighborhood home price increase between 1971 and 1993 was 4.72% but varied across neighborhoods from 8.75% to -0.82%. Not surprisingly, the highest rate of price appreciation was concentrated in a small high-income, predominately white area while the lowest price appreciation rates (actually declines) were located in predominately inner-city minority areas. Based on research on property values at the neighborhood level, it appears that housing returns in low-income and minority neighborhoods may be significantly lower than the average overall rate for the MSA. This appears to suggest that housing in these areas may not be a great investment.

Unfortunately, we cannot separate out the causal effect between low mortgage origination activity and neighborhood price appreciation. For example, it is possible that lender reluctance to lend in minority areas resulted in fewer potential buyers and thus depressed house prices. This is the implicit assumption behind defining certain areas as "underserved" and having the GSEs target lending to these areas. Thus in order to estimate the impact of this policy it is rational to use the overall mean MSA return in calculating efficient portfolios under the assumption that with additional targeted lending, returns in these areas will more closely match the overall MSA returns in the future. However, if the historically low property returns are the result of other factors (such as crime, age of housing stock, etc.) rather than lower availability of credit, then housing returns may not improve. The implication is that our analysis may underestimate the need for direct subsidies to increase the mortgage demand in these areas.

Compensation for Increased Risk

Given that owner-occupied housing is one goal of our housing policy, how much would we have to pay a wealth-constrained potential homeowner to more than double their home equity stake? One approach to this problem, in the framework of the Markowitz model, is to characterize risk aversion in terms of a threshold return. Portfolios along the efficient frontier can be ranked according to the probability they provide for returns dropping below a chosen threshold. For simplicity, let us take as our objective the minimization of the chance that our portfolio will lose money. The portfolio that minimizes the chance of loss is the tangency portfolio identified by the ray extending from the origin to the efficient frontier.(8)

The tangency portfolio represented in Figure 2 is composed of approximately equal weights in stocks, bonds and the home. What would it take to induce an investor to change from a portfolio with 34% home equity to a portfolio with 63% home equity? A portfolio with 63% in housing can be found in about the middle of the efficient frontier. Notice from Figure 2 that the volatility of the investment portfolio increases from about 8% to about 13% per year. The tangency line identifies all portfolios, feasible or infeasible, with equal probability of loss. The point on this line with 13% volatility is considerably above the existing efficient frontier. In order to maintain the same probability of loss as the tangency portfolio, an additional compensation of 6% per year is necessary. As the proportion of investment in housing increases to 100%, the compensation demanded for maintaining constant probability of loss increases at an increasing rate.

Requiring a resident of a low-income neighborhood to place 100% of his or her wealth into housing in that neighborhood may be desirable from the perspective of a society which seeks to align incentives for home and neighborhood maintenance, however mean-variance analysis suggests that this may be difficult to do without substantial subsidies. It is unclear whether subsidies targeted at increasing mortgage availability in targeted areas will be sufficient to increase demand in these areas. Direct subsidies may be necessary unless residents of low-income neighborhoods are risk-neutral. Insurance against loss of principal is another possible mechanism to induce investors to take on the larger risk of higher leverage and lower diversification. Given the poor historic performance of housing in predominately minority and low-income areas, it would appear that subsidies, such as loss insurance, in addition to increased targeted lending would be required to increase demand in these areas.

Finally, it is also possible that targeted lending in underserved markets may actually result in the displacement of low-income residents since our tax system does not provide incentives for low-income families to purchase homes. Interest payments on home mortgages are deductible and are a possible medium of governmental subsidization of home purchases in low-income neighborhoods. Our graduated tax structure means, however, that a high-income family receives a larger subsidy for home purchase than a low-income family. Consider who will benefit from targeted mortgage lending in a low-income neighborhood. All other things equal, families that stand to receive greater tax shields will be more likely to take out a mortgage than families whose incomes are too low to benefit from the subsidy. Indeed they can actually pay more for the home because their net costs of home ownership are lower, bidding up the prices of homes in the targeted areas. Thus, while aggressive targeting of underserved markets may attract borrowers to that neighborhood and increase property values, it may be at the expense of lower-income families. The implication is that lower-income families may be crowded out by higher-income families because homes are relatively more expensive for them to purchase!

Conclusion

Targeting underserved mortgage markets has the laudable goal of increasing owner-occupied housing in low-income neighborhoods. Insufficient consideration is given, however, to the household investment decision processes which may lead to under-utilization of mortgage opportunities. Lower-income families may be wealth-constrained, and thus, if they buy a home, the home equity is likely to represent a substantial portion of their asset portfolio. A large allocation to home equity may not be optimal from the investor's perspective, and reduced loan-to-value requirements will not make the home purchase more attractive because with increased leverage comes increased risk.

Using the Markowitz asset allocation framework, and statistical inputs from a major metropolitan area, Atlanta, we show that investors with a reasonable desire to minimize their risk of loss will demand a substantial subsidy to increase the proportion of their housing investment. While the tax code would seem to represent one potential mechanism for this subsidy, it is relatively unfavorable to low-income households. Because of this, we find that the targeting of underserved mortgage markets may accomplish the goal of raising property values but at the same time crowd out lower income residents and lead to gentrification.

References



Carr, J.H., I.F. Megbolugbe, M. Cho and H Buist, 1994, Underserved Residential Mortgage Markets, Working Paper.

Goetzmann, William, 1993, The Single Family Home in the Investment Portfolio, Journal of Real Estate Finance and Economics, 6, 201-222.

Jensen, Michael C. and Kevin J. Murphy, "Performance Pay and Top-Management Incentives," Journal of Political Economy 98 (April 1990), pp. 225-264.

Kennickell, Arthur and Janice Shack-Marquez, 1992, Changes in Family Finances from 1983 to 1989: Evidence from the Survey of Consumer Finances, Federal Reserve Bulletin 78(1), 1-18.

Kim, Sunwoong, 1995, Race and Home Price Appreciation in Urban Neighborhoods, University of Wisconsin-Milwaukee, Working paper.

Knight, John R., C.F. Sirmans, and Geoffrey K. Turnbull, 1996, List Price Information in Residential Appraisal and Underwriting, Journal of Real Estate Research, (forthcoming).

Miller, Merton H., and Myron S. Scholes, "Executive Compensation, Taxes, and Incentives," in William F. Sharpe and Cathryn M. Cootner, eds., Financial Economics: Essays in Honor of Paul Cootner (Prentice-Hall, Englewood Cliffs, NJ, 1982), pp. 139-157.

Murphy, Kevin J., "Corporate Performance and Managerial Remuneration: An Empirical Analysis," Journal of Accounting and Economics 7 (April 1985), pp. 11-42.

Smith, Barton A. and William P. Tesarek, 1991, House Prices and Regional Real Estate Cycles: Market Adjustments in Houston, AREUEA Journal 19(3), 396-416.


Figure 1: Efficient Frontier for Stocks, Bonds and Home Equity

After-tax portfolios of positive mixtures of stocks, bonds and home equity for an Atlanta property are represented on the efficient frontier. The home investment returns are adjusted by accounting for an 80% mortgage, imputed rental income and maintenance fees for the median home buyer. Taxes on stock and bond portfolios and capital gains taxes on the home are assumed to be deferred. Frontier created with the Ibbotson Associates EnCorr Optimizer.


Figure 2: Optimal Portfolio Choice: Minimizing the Probability of Loss

The tangency line represents the feasible and infeasible portfolios that minimize the probability of loss. The tangency portfolio is composed of 33% stocks, 32% bonds and 34% home. For an investor to double their exposure in home equity to 67% and maintain the same risk of loss would require an additional compensation of 6% expected return per year.

1. Please direct correspondence to William N. Goetzmann, Yale School of Management, Box 208200, New Haven, CT 06520-8200, (203) 432-5950, william.goetzmann@yale.edu, http://viking.som.yale.edu.

2.

2 According to the Census Department 1993 Survey of Households, the median net worth for persons in the lowest income quintile was $4,249 while Kennickell and Shack-Marquez (1992) report that the 1989 median net worth for renters was $2,000. Equity requirements for home ownership could easily exceed these wealth levels.

3.

3 See Miller and Scholes (1982), Murphy (1985), and Jensen and Murphy (1990) among others for the theoretical development of firm compensation policies.

4.

4 However, unlike corporate officers, homeowners are somewhat restricted in their ability to increase the value of their home given that exogenous factors have a significant impact on price.

5.

5 See Goetzmann (1993) for details of the calculation for means, standard deviations and correlations of asset classes, including the home. Statistics for the housing investment are calculated from adjustments to a repeat-sales return index calculated from Atlanta repeat-sales data provided by Robert Shiller. Means for stocks, bonds and house are 9.72, 7.95 and 17.67, respectively. Standard deviations for stocks, bonds and house are 14.03, 12.65 and 19.71 respectively. The correlation between stocks and bonds is .11, and the correlation of the home to stocks and bonds is -.18 and -.15 respectively. Adjustments for taxes, maintenance and imputed rent are described in Goetzmann (1993).

6.

6 Approximately, 43.6% of individuals in the lowest-income quintile have a net worth greater than $10,000.

7.

7 The risks become even greater if we use the historic neighborhood appreciation rates rather than the overall MSA rate.

8.

8 This is a straightforward implication of the assumption that returns are normally distributed. Minimizing the probability of loss is analogous to maximizing the probability that the chosen portfolio return is greater than zero. Thus, portfolios p = 1...N can be ranked by their t-statistics: Rp/sp. The feasible portfolio with the greatest t-statistic is the tangency portfolio. All portfolios, along the ray from the origin -- feasible or not -- share the same probability of loss.