Hostname: page-component-cd9895bd7-jn8rn Total loading time: 0 Render date: 2024-12-25T07:33:18.788Z Has data issue: false hasContentIssue false
  • English
  • Français

Interstate Differences in Mortgage Lending Risks: An Analysis of the Causes

Published online by Cambridge University Press:  19 October 2009

George M. von Furstenberg
Get access Rights & Permissions [Opens in a new window]

Extract

Researchers and political analysts concerned with the inter-regional flow of mortgage funds have often pointed to the existence of yield differentials as prima facie evidence of misallocation of capital and national resources. Limited information and myopic lending horizons, with market imperfections reinforced by state laws and institutional segmentation, have been postulated. They are regarded as responsible for costly “frictions” in the export of capitalto the fast-growing, generally low-income, states, particularly those of the South. Both federal and state legislative action, intensified private arbitrage, and better secondary market facilities and instruments are then urged to improve inter-regional financial mediation to reduce or eliminate the yield differentials.

Type
Research Article
Information
Journal of Financial and Quantitative Analysis , Volume 5 , Issue 2 , June 1970 , pp. 229 - 242
Copyright
Copyright © School of Business Administration, University of Washington 1970

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 The historical record is described in Grebler, Leo, Blank, David M., and Winnick, Louis, Capital Formation in Residential Real Estate, National Bureau of Economic Research (Princeton: Princeton University Press, 1956)Google Scholar, ch. 15. Small but, compared to intraregional variations, statistically significant inter-regional yield spreads of only 12 basis points on FHA-insured mortgages have been found at a high level of aggregation and for a period of comparatively easy money (1959) by Halbert C. Smith. See his articles, “A Note on Regional Mortgage Yield Differentials,” Journal of Risk and Insurance, Vol. 32 (June 1965), pp. 283–285CrossRefGoogle Scholar, and “Institutional Aspects of Inter-regional Mortgage Investment,” Journal of Finance, Vol. 23 (May 1968), p. 353. By contrast, the maximum spread estimated in early 1969 by FHA and FHLBB, assuming prepayment of 25-year conventional mortgages on new houses after 10 or 12 years, was 65 basis points between the West (7.95 percent) and North Central regions (7.30 percent). The range between metropolitan areas was even wider with a yield of around 8 percent reported for New Orleans, Houston, and Seattle compared to just below 7 percent for Chicago, Detroit, and Philadelphia.

2 Regional differences in the structure of the mortgage lending industry and the size of and competition within local lending markets are used to explain yield differentials by Meyer, Paul A., “Price Discrimination, Regional Loan Rates, and the Structure of the Banking Industry,” Journal of Finance, Vol. 22 (March 1967), pp. 37–48Google Scholar; and Conard, Joseph W., The Behavior of Interest Rates, National Bureau of Economic Research (New York: Columbia University Press, 1966), pp. 21–22Google Scholar. Further causes of market imperfections are analyzed in Klaman, Saul B., The Postwar Residential Mortgage Market, National Bureau of Economic Research (Princeton: Princeton University Press, 1961), pp. 95–97Google Scholar; and McKie, James W., “Credit Gaps and Federal Credit Programs,” in Federal Credit Programs, Commission on Money and Credit (Englewood Cliffs, N.J.: Prentice-Hall, 1963), pp. 331–332Google Scholar. See also Jones, Oliver and Grebler, Leo, The Secondary Mortgage Market (Los Angeles: The Regents of the University of California, 1961), pp. 44–51.Google Scholar

3 It has been found that “regions with net import surpluses (capital deficits) tend to grow faster in terms of both total and per capita income than regions with net export surpluses.” Romans, J. Thomas, Capital Exports and Growth Among U.S. Regions (Middletown, Connecticut: Wesleyan University Press, 1965), p. 118Google Scholar. In the mortgage market, capital imports take the form of purchases of FHAinsured or VA-guaranteed instruments by out-of-state lenders, primarily eastern insurance companies and mutual savings banks. For a “scalp” of one-half percent of the annual mortgage balance, servicing may remain with the savings and loan association, mortgage company, or commercial bank which originated the loan for sale to permanent investors.

4 Effects of FNMA secondary market transactions have been evaluated by Guttentag, Jack M., “The Federal National Mortgage Association,” in Federal Credit Agencies, Commission on Money and Credit (Englewood Cliffs, N.J.: Prentice—Hall, 1963), pp. 67–158Google Scholar; and Break, George F., The Economic Impact of Federal Loan Insurance (Washington: National Planning Association, 1961)Google Scholar, ch. 5. Klaman, The Postwar Residential Mortgage Market, p. 226, shows that up to 1955, FNMA purchased mortgages in amounts over 3 times that of sales in California, Texas, and Florida, while purchases exceeded sales by only 66 percent for the United States as a whole. As of June 30, 1968, mortgages secured by homes in these three states accounted for over one-third of the total secondary market portfolio held by FNMA.

Recently, calls for mortgage-backed securities have been issued by Jones, Oliver, “Private Secondary Market Facilities,” Journal of Finance, Vol. 23 (May 1968), pp. 359–366CrossRefGoogle Scholar; and Report of the President's Committee on Urban Housing, A Decent Home (U.S. Government Printing Office, 1969), pp. 22–24Google Scholar, 131–133. See also Brimmer, Andrew F., “Central Banking and Residential Mortgage Credit,” Savings and Loan Annals 1968 (Chicago: U.S. Savings and Loan League, 1969), pp. 32–39.Google Scholar

5 Pratt, Richard T., “Discussion,” Journal of Finance, Vol. 23 (May 1968), p. 379Google Scholar.

6 See Morton, Joseph E., Urban Mortgage Lending: Comparative Markets and Experience, National Bureau of Economic Research (Princeton: Princeton University Press, 1956), pp. 95–97Google Scholar; Saulnier, Raymond, Halcrow, Harold G., and Jacoby, Neil H., Federal Lending and Loan Insurance, National Bureau of Economic Research (Princeton: Princeton University Press, 1958), p. 329Google Scholar; and Maisel, Sherman J., Financing Real Estate (New York: McGraw Hill, 1965), pp. 194–195Google Scholar.

7 See Page, Alfred N., “The Variation in Mortgage Interest Rates,” Journal of Business, Vol. 37 (July 1964), pp. 280–294CrossRefGoogle Scholar. See also von Furstenberg, George M., “Default Risk on FHA-insured Home Mortgages as a Function of the Terms of Financing: A Quantitative Analysis,” Journal of Finance, Vol. 24 (June 1969), pp. 459–477.CrossRefGoogle Scholar

8 This has been emphasized by McFarland, M. Carter, “Major Developments in the Financing of Residential Construction since World War II,” Journal of Finance, Vol. 21 (March 1966), pp. 382–394CrossRefGoogle Scholar. The same inter-relation is assumed in some time series studies where changing terms on FHA mortgages are used as proxies for ” the whole mortgage market. See, for instance, Huang, David S., “The Short-Run Flows of Nonfarm Residential Mortgage Credit,” Econometrica, Vol. 34 (April 1966), p. 347.CrossRefGoogle Scholar

9 As defined by FHA, net effective income is generally the equivalent of takehome pay.

10 See U.S. Department of Housing and Urban Development, Federal Housing Administration, FHA Homes in 1966: Data for States and Selected Areas, Table 27S. The standard error of the percentages available for 48 states is 3.6.

11 U.S. Bureau of the Census, U.S. Census of Housing: 1960, States and Small Areas, United States Summary, Vol. I, Final Report HC(1)–1 (1963), pp. 22–27Google Scholar. The percentages calculated refer to homeowners who moved into their (then) present units between January 1957 and March 1960.

12 For statistical evidence, see Projector, Dorothy S. and Weiss, Gertrude S., Survey of Financial Characteristics of Consumers (Washington: The Board of Governors of the Federal Reserve System, 1966), pp. 96–97.Google Scholar

The inter-relations between the age of family head and asset holding, mortgage indebtedness, and mobility are traced in Riley, Matilda W. and Foner, Anne, Aging and Society (New York: Russell Sage Foundation, 1968), pp. 69–95, 131–132, 143–147.Google Scholar

13 See Winger, Alan R., “Interarea Variations in Vacancy Rates,” Land Economics, Vol. 43 (February 1967), pp. 84–90CrossRefGoogle Scholar. Winger even sees an accelerator mechanism at work in that disequilibria caused by an unexpected lag in growth rates which tends to be self-aggravating by lowering the vacancy requirements of the community. For further references and discussion, see Grebler, Leo and Maisel, Sherman J., “The Determinants of Residential Construction: A Review of Present Knowledge,” in Impacts of Monetary Policy, studies prepared for the Commission on Money and Credit (Englewood Cliffs, N.J.: Prentice-Hall, 1963), pp. 551–575.Google Scholar

14 Cf. Kendall, Leon T., “The Quality of Credit,” in Conference on Savings and Residential Finance, Proceedings 1964 (Chicago: U.S. Savings and Loan League, 1964), pp. 58–70.Google Scholar

15 Mortgage Discounts: A Report by the Department of Housing and Urban Development, Senate Committee on Banking and Currency, 90 Cong. 1st Sess. (1967), pp. 35–37. There is normally a lag of several months between the initial delinquency, the exhaustion of forebearance, and the time when title to the property is eventually conveyed to FHA and default is registered. For this reason, measured defaults in 1966 can hardly have been affected by the “crunch” starting in August of that year. We therefore retained the 1966 observations.

16 In FHA terminology, both previously occupied homes and those hitherto unoccupied homes which were not approved as proposed construction and subsequently inspected are classified as “existing.” The percentage of “new” home mortgages in the FHA sample fell somewhat irregularly from 36.5 percent in 1959 to 25.9 percent in 1963. PN is particularly low in the District of Columbia and the New England states. Insurance in force cannot be broken down into its originally “new” and “existing” components by states. We therefore used the percentage of “new” homes in the annual endorsements lagged 3 years from the date of default as an appropriate proxy. PN was also used to derive a weighted average of the L/V, M, and median income characteristics reported for “new” and “existing” homes separately by states in the source cited in footnote 10. The resulting state averages could then be used to explain default rates on all insured mortgages in force taken from U.S. Department of Housing and Urban Development, Statistical Yearbook 1966 (1968), pp. 109–110, and earlier Annual Reports of HUD or HHFA, or from computer run FHA S–6.

17 The simple correlation between ln(l−L/V) and M is -0.398; between M and PN, 0.476; but between ln(l−L/V) and PN, it is barely above zero. Average terms on “new” (“existing”) home mortgages in 1958, 1962, and 1966 were 27.4 (24.2), 30.4 (27.5), and 30.2 (28.2) years, respectively. At the same dates, the average L/V grew from 89.1 (88.5) to 92.9 (92.6) and 92.7 (93.2) according to the 30–50 percent samples of all purchase transactions described in the introduction to the annual FHA publication, Data for States and Selected Areas.

18 Muth is agnostic on the effect of lengthening maturities although, ceteris paribus, the greater the term, the slower the equity buildup and hence the longer the period during which default may be cheaper than prepayment. However, he unequivocally expects used home financing to be more risky than new home mortgages. This could not be confirmed in our estimates since the coefficient of PN is subsequently found to be completely insignificant. Muth, Richard F., “Interest Rates, Contract Terms, and the Allocation of Mortgage Funds,” Journal of Finance, Vol. 17 (March 1962), pp. 63–80 (esp. pp. 67, 78).Google Scholar

19 This is, of course, equal to the cumulative insurance volume written to December 31, 1961 divided by the number of mortgages endorsed up to December 31, 1957. The average value of the indicated ratio (G) is equal to 1.47. Over the complete residential construction cycle between the two dates starting and ending about 1.5 years before a respective peak, the number of mortgages ever insured thus increased by almost 50 percent.

20 The distribution of default rates over the age of mortgages is strongly skewed to the right so that the average age of insured mortgages in force lies well to the right of the peak in annual default rates. Reducing this age, therefore, moves us up the tail toward the peak.

21 Official vacancy data by states are available only at the decennial Census dates. The source is U.S. Bureau of the Census, U.S. Census of Housing: 1960, Vol. 1, pp. 4–9.

22 The correlation (r) between V and ln(l-L/V) is -0.209. When overbuilding would otherwise threaten, reducing down payments will move houses, while at the same time lowering the mortgagor's incentive not to lose them. Also, FHA found that “the highest proportion of acquisitions (foreclosures) to total cases insured was for properties approved prior to the start of construction. Since seven-eighths of new construction commitments are issued to builders, this experience is specifically applicable to mortgage risks associated with speculative construction.” See FHA Mortgage Foreclosures, Hearings Before a Subcommittee of the Senate Committee on Banking and Currency, 88 Cong. 2 Sess. (1964), p. 254.

23 For the nation as a whole, the median annual effective income for “new” (“existing”) home occupant transactions in the sample was $6,933 ($6,587) and $7,565 ($7,293) in 1959 and 1963, respectively. The percentage increase is twice the negligible rise in the consumer price index (5.4 percent) over this period. Total effective income excludes those components that are not expected to last over the early period of mortgage risk so that Y, on the average, falls 11 percent below the total current income of mortgagors.

24 The insignificance of the first variable may be explained by insufficient variations in M. Its coefficient of variation, defined as the ratio of standard error to mean, is only 7 percent compared to-13 percent for G, 28 percent for V, and 122 percent for ln[10(l−L/V)]. The finding that PN is also insignificant is compatible with results derived in our earlier paper in which default rates could be estimated on a national basis for separate maturity groups. “Existing” home 25- and 30-year mortgages were found to be about as risky at given loan-value ratios as 30-year “new” home mortgages. Ninety percent of all FHA mortgages insured in recent years fall into these three maturity groups, and redistribution within this single risk class should not substantially affect default rates.

25 The weights (wi) were constructed by dividing the number of endorsements in force in any state and year (Ei) by the total cumulative number of endorsements in force reported at the end of years 1961 through 1965 to which annual defaults were related from 1962 through 1966. In short, wi = Ei/Sum(Ei).

26 The relative importance of each variable can be determined by multiplying regression coefficients by the standard deviation of the variable to which they refer. For ln[10(l−L/V)], we then obtain -0.5363 compared to -0.4751 for lnY. From the absolute value of these standardized coefficients, it appears that the relative importance of the income variable has now become almost as great as that of the equity-value ratio.

27 If Y is deflated by a budget cost index for a moderate standard of living derived for states by Lawrence R. Tharp of Columbia University, the income variable is no longer significant in the unweighted run, as shown below..

Even if the coefficient of Y is to be believed, the absolute value of the income elasticity of default rates is now well below unity.

28 Though higher-income, mortgagors will frequently make larger percentage down payments, the correlation between lnY and ln(l-L/V) is only 0.153 in the unweighted regressions, but higher in the weighted regression (0.246).

29 To see whether locational characteristics by themselves are significant for the explanation of default rates, states were grouped as far as possible into contiguous blocks. We tentatively created three risk classes to see whether the resulting grouping would suggest any obvious geographic or economic explanation for interstate variations in default rates. For states in the worst group, default rates were generally between 4 and 1 percent; states in the next highest class showed rates between 1 and 1/2 percent; and the best states have less than 1/2 percent of their section 203(b) mortgage insurance in force terminated through default in all or most of the years 1962–1966. On this basis, all states bordering the Gulf of Mexico and a few adjoining states, such as South Carolina, Georgia, Oklahoma, and Kansas, were classified as high risk, with dummy HR, and states in the vicinity of Washington, D.C. as well as the New England, Northern Mountain, and Pacific states were classified as low risk. The middle belt of states, ranging from New Jersey and North Carolina to Nevada, was medium risk with dummy MR.

Although the high-risk states are generally southern, the geographic, demographic, and economic characteristics of the states in the other groups are most diverse. For instance, while the lowest risk group contains most of the states with the highest per capita personal income, such as California and New York, it also includes states, such as North Dakota and Maine, which have per capita income levels comparable to those in the South. [See Survey of Current Business, Vol. A3 (August 1963), p. 9.] We therefore had little hope that regional dummies or general economic characteristics could contribute significantly to the explanation of default rates. Regressing the logarithm of annual default rates on the dummies HR and MR captured only 36 percent of the total variation, and no theory could be devised which would give these dummies explanatory rather than merely descriptive meaning.

30 The importance of L/V and the insignificance of location in either the microeconomic or regional sense have been stressed in FHA Mortgage Foreclosures, pp. 242–246, 275, and Kendall, Leon T., Anatomy of the Residential Mortgage, Occasional Paper No. 2 (Chicago: U.S. Savings and Loan League, 1964), pp. 17–20Google Scholar. The results of both studies are derived from simple cross-tabulations, however, with little or no attempt being made to hold “other things constant.”

31 For instance, raising G by one standard deviation from its mean of 1.47 to 1.67, meaning that relative to the insurance volume ever written through 1957 the number of mortgages insured in the next 4 years is to rise by 20 percent more than average in a particular state, causes expected default rates to increase by 17 percent.

32 Nondefault disinsurance rates would decrease by over 20 percent in the same situation according to (4) below. In view of FHA's invariant premium of 1/2 percent per year on the (average 12-monthly) mortgage balance outstanding, one would expect all factors raising risk and default rates (D/E) to reduce nondefault termination rates (ND/E) during the early period of mortgage risk. Consequently, corroborating evidence that both mortgagors and mortgagees recognize the effect of V, G, and (1−L/V) on the need for and desirability of remaining insured is provided by all regression coefficients in (4) being significant but with the opposite sign of those in (1).

33 Standardizing the regression coefficients in (1) as before (see note 26), we obtain 0.6196 for ln[10(l−L/V)] and 0.1867 for V. The relative importance of the equity-value ratio in explaining variations in D/E is therefore over three times as great as that of the vacancy rate. Raising state vacancy rates with mean 1.52 and standard deviation of 0.42 by one percentage point, or from 1 to 2, causes default rates to increase by 56 percent. The down payment percentage, with geometric mean of around 8 percent, would have to be reduced by only about 1.4 percentage points, or from 0.80 to 0.66 in our specification to achieve the same effect. This reduction of 17.5 percent is equal to only three-quarters of the standard deviation of the equity-value ratios.

34 Conard, The Behavior of Interest Rates, p. 21.

35 While frequently recommended, FNMA has not hitherto made use of its authority to operate in conventional mortgages under certain conditions. Nevertheless, heavy concentration of net purchases of government-insured or guaranteed mortgages in the capital deficit states may create temporary imbalances in the supply also of conventional mortgage credit. The unassisted private market may not be instantly self-adjusting, although it may be efficient in providing for transfers of capital over longer periods.

36 Conard, op. cit.