United States
Selected Issues

This Selected Issues paper on the United States examines the effect of the structure of the mortgage market on real housing activity and housing prices. The market-based financial structure has reduced the volatility of mortgage lending. Changes in the structure of the mortgage market have coincided with lower volatility of real housing activity. Regional income growth and unemployment rates have statistically significant and correct signed effects on housing prices. Tests of the relative importance of mortgage market structure and macroeconomic variables suggest an important effect from the financial structure.

Abstract

This Selected Issues paper on the United States examines the effect of the structure of the mortgage market on real housing activity and housing prices. The market-based financial structure has reduced the volatility of mortgage lending. Changes in the structure of the mortgage market have coincided with lower volatility of real housing activity. Regional income growth and unemployment rates have statistically significant and correct signed effects on housing prices. Tests of the relative importance of mortgage market structure and macroeconomic variables suggest an important effect from the financial structure.

PART I: REAL DEVELOPMENTS

I. Boom-Bust Cycles in Housing: The Changing Role of Financial Structure1

1. This paper examines the effect of the structure of the U.S. mortgage market on real housing activity and housing prices. Following a brief review of structural changes in mortgage finance, the paper analyzes their impact on real housing activity and housing prices, and carries out an econometric investigation of house price dynamics. In addition, it explores whether the current system of housing finance may have reduced risks of a speculative boom-bust housing cycle.

A. Structural Changes in Mortgage Finance

2. Through the mid-1980s, mortgage lending was mainly a local business, prone to local and regional shocks. Residential mortgages were originated locally, and most mortgages were kept on the balance sheets of local lenders—such as savings and loan institutions (S&Ls)—for the lifetime of the loan. The availability of mortgage credit in any region depended largely on local financial conditions, including the quality of bank and thrift loan portfolios and levels of capital, as restrictions on interstate banking inhibited the flow of lending between regions.

3. Monetary policy and bank regulations contributed to the volatility of housing finance. Monetary policy had a first-order impact on housing activity (the mortgage lending channel) through the prominent role of depositories’ balance sheets in mortgage flows. The supply of credit fluctuated more severely—including through credit rationing under tight monetary conditions—than would have occurred through changes in interest rates alone. Bank regulations—including Regulation Q (Reg Q), which limited the interest rate to be paid on bank deposits—also restricted mortgage finance.

4. Under such a market structure, local markets were subject to a boom-bust financing cycle. Real mortgage growth slowed sharply during several episodes in the 1960s when Reg Q ceilings on deposit rates became binding and banks and S&Ls were unable to retain deposits (Figure 1). Lending shrank in response to the declining funding base, and would subsequently rebound as monetary policy eased and deposit growth resumed.

Figure 1:
Figure 1:

Real Mortgage Growth and Real M1 Deposit Growth

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Board of Governors, Federal Reserve.

5. The S&L crisis precipitated a shift toward nationwide funding of housing activities based on mortgage securitization. S&Ls had funded long-term fixed-rate mortgage loans with short-term deposits, incurring large interest rate exposures. They suffered major capital losses when interest rates rose in the early 1980s, which contributed to the failure of many institutions. In their place, a system evolved under which loans originated by mortgage banks were subsequently pooled and securitized by Government Sponsored Enterprises (GSEs). Subsequently, home mortgages held by depositories dropped from above 70 percent of the market in the 1970s to below 40 percent in recent years (Figure 2).

Figure 2.
Figure 2.

Mortgage Market Structure

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Board of Governors, Federal Reserve.

6. This market-based financial structure has reduced the volatility of mortgage lending. The availability of funds is no longer limited by conditions at local depository institutions or the strength of regional economies, and the supply of mortgages is less impacted by policy-induced fluctuations in high-powered money expanding or contracting bank balance sheets. Tighter or looser monetary conditions still affect the market interest rate and the general willingness to lend, of course, but this has a smaller direct impact on the mortgage lending channel.2

7. Changes in the structure of the mortgage market have coincided with lower volatility of real housing activity. In the past, housing construction and transactions tended to fluctuate strongly with the cycle, as evidenced by the boom-bust phases between the late 1960s and 1980s (Figure 3). Residential investment spending also exhibited pronounced cycles, with growth rates of 40 percent or more not uncommon during booms, and declines of nearly the same magnitude during busts. This cyclical volatility has diminished markedly, however, and housing starts have grown at a relatively stable pace since 1990. Although starts have risen somewhat faster in the past few years, their growth still remains well below rates reached in previous booms.

Figure 3:
Figure 3:

Housing Starts and Residential Investment

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: BEA and Census Bureau.

8. Housing activity has also become more synchronized across regions. Cycles of housing starts in past decades often moved independently in different parts of the country, especially during the early 1960s and the 1980s (Figure 4). Beginning in 1990, however, growth trends in housing starts have been similar across all major regions.

Figure 4.
Figure 4.

Regional Housing Starts

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Census Bureau.

9. Housing prices have also converged toward more steady growth since the early 1990s. Amid high volatility, housing prices increased strongly in the 1970s, both in nominal terms and relative to the overall consumer price index (Figure 5). Nominal gains slowed sharply during the recessions in the early 1980s and prices declined in real terms in most regions. Since the late 1980s, price swings have had lower amplitude, and price trends have converged across regions.

Figure 5:
Figure 5:

Regional House Prices

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Office of Federal Housing Enterprise Oversight.

B. House Prices and Fundamentals

10. The smoothing of the boom-bust cycle in mortgage lending, real activity, and housing prices raises two questions. Have structural changes in mortgage finance improved the process of price formation? And what do these developments tell us about the risk of a sharp collapse in prices following the ongoing housing boom? To address the first question, an econometric approach is used to analyze the impact of financial market changes on price formation in the housing market. The results are subsequently used to address the second question.

11. Economic fundamentals for housing prices include regional variables for income, unemployment rates, and the labor force for the nine regions in Office of Federal Housing Enterprise Oversight’s (OFHEO) housing price index (HPI). A number of national variables are included to control for business cycle dynamics, namely interest rates, the CPI, and GDP growth. The model was estimated allowing coefficients to vary by region in a set of stand-alone regressions, as well as in a fixed-effects panel in which only constants vary by location.3

12. The results suggest that fundamental factors generally have the expected impact on housing prices (Table 1).4 Regional income growth and unemployment rates have statistically significant and correctly signed effects on housing prices. The panel estimates imply that a 10 percent rise in regional incomes is associated with a 2.5 percent increase in housing prices. Similarly, a 1 percentage point rise in the unemployment rate depresses housing prices by about 1 percent. National business cycle indicators are not estimated to have a significant impact, suggesting that their effect is captured by the regional measures. The coefficient on CPI inflation is small and insignificant as price effects may be captured through nominal incomes. Somewhat surprisingly, interest rates are not estimated to have a significant effect, which could reflect endogeneity problems. Results of the region-by-region regressions are similar in character, although the standard errors are larger.

Table 1.

House Price Regressions1

article image
Sources: Bureau of Economic Analysis; Bureau of Labor Statistics; Office of Federal Housing Enterprise Oversight; and Fund staff calculations.

Absolute value of t-statistics in parenthesis; * significant at 5 percent;** significant at 1 percent.

13. The model’s fit improves considerably when estimated separately over the periods corresponding to different mortgage market structures. For example, the model explains more than twice as much of the overall variation in housing prices over 1990–2004 compared to estimates over the entire time horizon (Columns 3 and 5). Moreover, the interest rate is now found to have the expected negative effect (also significant at the 1 percent level) in both the earlier and later time periods (Columns 4 and 5), although the different coefficients suggest that housing prices have become more sensitive to long-term interest rates as the importance of quantity restrictions on mortgage lending has waned.

14. Assuming that estimation residuals represent pricing errors, the results also indicate significant improvements in the pricing process. Pricing errors in the earlier years often were quite large but decreased through the mid-1980s and 1990s (Figure 6, Table 2). Price movements also appear to have converged across regions as the financial structure changed from a local and regional to a national basis (Figure 7).

Figure 6:
Figure 6:

Pricing Errors from Panel Regression

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: Office of Federal Housing Enterprise Oversight; BEA; BLS; and IMF staff estimates.
Table 2.

Pricing Errors

(Panel regressions in percent)

article image
Sources: Office of Federal Housing Enterprise Oversight; and Fund staff calculations.
Figure 7:
Figure 7:

Pricing Errors from Panel Regression

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: Office of Federal Housing Enterprise Oversight; BEA; BLS; and IMF staff estimates.

15. Tests of the relative importance of mortgage market structure and macroeconomic variables suggest an important effect from financial structure. Average absolute pricing errors from the housing price model were regressed on the securitized share of the mortgage market (a proxy for structural change in housing finance) and the 4-quarter rate of CPI inflation (Table 3). The results indicate that changes in mortgage market structure are associated with improvements in the house price formation process: higher securitization tends to reduce pricing errors. The estimates suggest that the results are economically significant, with a 10 percentage point rise in securitized share, all else equal, lowering the average absolute pricing error by 0.8 percentage point. The rise in securitization from 10 percent of the total mortgage market to 60 percent or more between the mid-1970s and 2004 would thus be associated with a 4 percentage point decline in absolute pricing errors—essentially all of the improvement that is estimated to have occurred. Inflation, in contrast, was not found to have a significant effect on the price formation process.

Table 3.

Pricing Error Regressions1

article image
Sources: Bureau of Economic Analysis; Bureau of Labor Statistics; Office of Federal Housing Enterprise Oversight; and Fund staff calculations.

Absolute value of t -statistics in parenthesis; * significant at 5 percent;** significant at 1 percent.

16. Pricing errors may have resulted from rational decisions by home buyers potentially facing credit rationing at some later date. When mortgage lending was subject to quantity restrictions, potential home buyers who otherwise would qualify for a loan may at times have been unable to obtain financing. Knowing this, a potential homebuyer may have rationally paid a premium over a house’s fundamental value during periods when financing was readily available. Conversely, homeowners forced to sell during periods when potential buyers had difficulty obtaining financing may have reduced sales prices.

17. An alternative test using earlier historical data was performed to confirm the importance of the impact of mortgage market structure for the pricing process. While reliable data on house prices are available only for the most recent three decades, data on real and nominal residential investment spending are available since 1948. As inflation and macroeconomic volatility remained low until the late 1960s, spending data from 1948–1968 provide a control period for testing whether the elevated volatility of residential investment and price changes from the late 1960s through the early 1980s was more related to macroeconomic factors or to the structure of the mortgage market.

18. Before the 1990s, boom-bust cycles were similar during the low and high inflation decades, suggesting that the structure of the mortgage market was mainly responsible for generating large upswings and downswings in activity. Residential investment was highly procyclical from 1948 through the late 1960s (Figure 8, previous page). There were four episodes in which residential investment grew at more than a 25 percent rate over 4 or more quarters, with a surge of nearly 50 percent in the early 1950s. During bust periods, spending often declined at a 20 percent rate or faster. These swings in real activity are comparable to (and at times greater than) those that occurred during the higher-inflation years of the 1970s and 1980s.

Figure 8:
Figure 8:

Real Residential Investment and Mortgage Growth

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: BEA; BLS; and Board of Governors, Federal Reserve.

C. Will the Current Boom Be Followed by a Bust?

19. There are widespread concerns about the recent run-up in housing prices. While financing in a nationwide securitized market may have reduced the stop-go cycle of mortgage flows, it could be argued that the freer flow of financing may aid in the development of a nationwide bubble. Indeed, nationwide housing prices have risen 50 percent over the past five years, with some metropolitan and regional markets soaring much higher. While some of these gains may reflect a catch-up following little or no appreciation in previous years, recent increases have been particularly rapid, and may be ahead of fundamentals.

20. Several indicators point to speculative pressures on prices. Purchases of second homes, often for investment purposes, have risen, as has the use of interest-only mortgages that allow a more expensive purchase for a given monthly payment. Surveys suggest home buyers have extrapolated past gains into their expectations for future appreciation. The price-to-rent ratio has increased, suggesting that in some markets the valuations can only be justified by expectations of rapid appreciations. These warning signs could foretell a drop in prices, or an adjustment through a long period of slow nominal gains until real valuations came back in line with fundamentals.5

21. Other factors, however, mute some of these risks. Estimated pricing errors from the model presented earlier are not particularly large, suggesting that much of the recent gains can be explained by rising incomes, rising employment, and low interest rates. Moreover, the positive surprises in housing prices over the past five years follow a decade of negative surprises, especially on the East and West Coasts (Figure 9). These patterns suggest that much of the recent gains may be a catch-up after a prolonged period of prices lagging fundamentals. In addition, there have been other changes in homebuilders’ behavior since the 1980s—including in speculative housing starts and in the accumulation of inventories of new homes—that may moderate these risks.

Figure 9.
Figure 9.

Regional Pricing Errors, 1978–2004

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: Office of Federal Housing Enterprise Oversight; BEA; BLS; and IMF staff estimates.

22. Speculative homebuilding used to dominate housing activity. Using data on the completion status of a new home sale, one can identify speculative building (sales of houses already completed or under construction) versus nonspeculative building (construction not yet started at the time of sale). From the 1960s through the 1980s, builders often engaged in speculative starts, largely due to the necessity to build an inventory of homes for sale in advance of a (relatively short) “hot” market (Figure 10).

Figure 10.
Figure 10.

Construction status at time of sale

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Census Bureau.

23. Speculative construction has declined since the 1980s. With builders’ demand no longer subject to stop-go cycles, the incentives to build an inventory of new homes in anticipation of a surge in demand are much diminished, limiting the overhang of new homes available for sale. Indeed, inventories of new homes have only recently returned to the levels of the early 1970s, despite a doubling of sales since then (Figure 11).

Figure 11:
Figure 11:

New Home Sales and Supply

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Source: Census Bureau.

24. During previous boom-bust cycles, a buildup of inventories preceded price collapses when demand waned. The amount of new home inventories divided by the monthly sales pace gives an indication of the vulnerability of housing markets to a drop-off in demand, and how long it may take to work off any excess inventories. During a typical cycle, an early increase in housing prices would have prompted more rapid building and a rise in months’ supply. As demand would wane later in the cycle—most often due to a constraints from the supply of mortgage finance—prices would decline to absorb the excess supply. For example, new home inventories reached nearly 10 months’ supply on the west coast during the early 1980s, and a high of 15 months’ supply during the boom in the northeast in the late1980s (Figures 12 and 13). Prices subsequently fell as much as 10 percent in order to work off excess stocks of new homes.

Figure 12:
Figure 12:

Housing Supply and Prices in the Pacific/West Region

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Figure 13:
Figure 13:

Housing Supply and Prices in the North East/New England Region

Citation: IMF Staff Country Reports 2005, 258; 10.5089/9781451839647.002.A001

Sources: Census Bureau, Office of Federal Housing Oversight, and IMF staff estimates.

25. During the current boom, however, months’ supply has remained near historic lows even in the regions where housing markets are particularly strong. Current inventories are at less than 4 months’ supply in the northeast and less than 3 months’ on the west coast. These levels are below those at which housing prices stabilized during previous bust cycles, and suggest that housing supply has not gotten far ahead of demand.

D. Conclusion

26. The change in mortgage market structure from a system based on balance sheet lending by depositories to a market-based system of securitized mortgage finance has damped the volatility of financing flows and real activity. With funding conditions now determined in a national market, trends in real activity and prices have become less cyclical and converged across all regions of the United States. As a result, a model of housing prices based on economic fundamentals finds that pricing errors—the deviations of actual prices from those estimated in the model—have fallen by half. Moreover, a change in homebuilders’ behavior—in particular, a move away from speculative starts and a reduction of levels of inventories of new homes—has reduced the risk of a sharp decline in housing prices, although some indicators continue to suggest speculative pressures in a number of metropolitan areas.

References

  • Angell, C., and N. Williams, 2005, “U.S. Home Prices: Does Bust Always Follow Boom?FYI: An Update on Emerging Issues in Banking (Washington: Federal Deposit Insurance Corporation). Available on the Internet at http://www.fdic.gov/bank/analytical/fyi/.

    • Search Google Scholar
    • Export Citation
  • Case, K.E., and R.J. Shiller, 2003, “Is There a Bubble in the Housing Market?Brookings Papers on Economic Activity, Vol. 2, pp. 299342.

    • Search Google Scholar
    • Export Citation
  • Davis, M.A., and J. Heathcote, 2004, “The Price and Quantity of Residential Land in the United States” (unpublished; Washington: Federal Reserve Board).

    • Search Google Scholar
    • Export Citation
  • Gallin, J., 2003, “The Long-Run Relationship between House Prices and Income: Evidence from Local Housing Markets” (unpublished; Washington: Federal Reserve Board).

    • Search Google Scholar
    • Export Citation
  • Genesove, D., and C. Mayer, 2001, “Loss Aversion and Seller Behavior: Evidence from the Housing Market,” Quarterly Journal of Economics, Vol. 116, pp. 123360.

    • Search Google Scholar
    • Export Citation
  • Glaeser, E.L., and J. Gyourko, 2003, “The Impact of Building Restrictions on Housing Affordability,” FRBNY Economic Policy Review, Vol. 9, June (New York: Federal Reserve Board).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2003a, “Are U.S. House Prices Overvalued?United States: Selected Issues, IMF Country Report No. 03/245.

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2003b, “Real and Financial Effects of Bursting Asset Price Bubbles,” World Economic Outlook, April (Washington).

    • Search Google Scholar
    • Export Citation
  • International Monetary Fund, 2004, “The Global House Price Boom,” World Economic Outlook, September (Washington).

  • Macroeconomic Advisers, 2004, “A Bubble in House Prices?Economic Outlook, June 18 (St. Louis).

  • McCarthy, J., and R.W. Peach, 2004, “Are Home Prices the Next ‘Bubble’?FRBNY Economic Policy Review, Vol. 10, December (New York: Federal Reserve Board).

    • Search Google Scholar
    • Export Citation
  • Peek, J., and J.A. Wilcox, 2005, “Secondary Mortgage Markets, GSEs and the Changing Cyclicality of Mortgage Flows” (unpublished; University of California at Berkeley).

    • Search Google Scholar
    • Export Citation
1

Prepared by Calvin Schnure.

2

See Peek and Wilcox (2005) for further discussion of how the GSEs and secondary mortgage markets have moderated the cyclicality of mortgage flows.

3

In the short run the housing stock is fixed, and most home sales involve existing dwellings rather than new construction. The availability of open land, as well as the presence of zoning restrictions, has been found to have an important effect on housing prices, explaining a portion of the more rapid appreciation in recent years in the mature cities of the East Coast and California (see McCarthy and Peach, 2004, and Glaeser and Gyourko, 2003).

4

The model is estimated in log differences, as is common in the literature on housing prices. Tests of regional housing prices and income do not find evidence of a cointegrating relationship, suggesting that a cointegrating equation may not be appropriate. See also Gallin (2003), who finds no cointegration between housing prices and income in metropolitan areas.

5

Many observers have noted that the rapid rise in housing prices may more likely be followed by slow appreciation than a price collapse. See, for example, Case and Shiller (2003), Genesove and Mayer (2001), IMF (2003), IMF (2004), Macroeconomic Advisers (2004), McCarthy and Peach (2004). Angell and Williams (2005) find that booms may be followed by busts, but that “this pattern may be more the exception than the rule” (FDIC 2005).

United States: Selected Issues
Author: International Monetary Fund