## _wp05200

## Source details

**Canonical URL:** [_wp05200](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05200.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05200.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05200.pdf.json)

---

### I. Introduction
- Research question: Why are housing markets prone to boom-bust cycles?
- Focus: Role of mortgage-market structure in boom-bust cycles in real housing activity and housing prices.
- Key mechanisms highlighted:
  - Pre-mid-1980s U.S. financial structure produced a stop-and-go credit cycle in response to changes in monetary policy and local banking conditions.
  - During easy-money periods: rapid expansions in real housing activity.
  - During tightening: depositories bore the brunt of monetary policy, contracting mortgage finance (cost of housing finance rising sharply or lending spigot shut off), causing real housing activity to contract and pent-up demand to build.
  - Price dynamics: tight-money episodes could push prices below long-term fundamental values; easing could bid prices above fundamentals, producing excess price volatility.
- Natural experiment: Post–Savings and Loan (S&L) crisis structural shift from local/regional balance-sheet lending to a national, market-based system of securitized mortgage finance.

### II. Structural changes in mortgage finance
- Pre-mid-1980s structure:
  - Mortgages mainly originated and held on balance sheets of local lenders (S&Ls/thrifts).
  - Availability of mortgage credit depended on local financial conditions and was constrained by restrictions on interstate banking.
  - Monetary policy and bank regulations (e.g., Regulation Q) amplified volatility in housing finance by affecting depositories’ funding and deposit rates.
  - Historical episodes: Real mortgage growth slowed sharply during several episodes in the 1960s when Reg Q ceilings became binding and banks and S&Ls were unable to retain deposits.
  - Under that structure, lending contracted with declining funding and rebounded as monetary policy eased and deposit growth resumed.
- S&L crisis and transition:
  - S&Ls had funded long-term fixed-rate mortgages with short-term deposits, creating unhedged interest rate exposures; many institutions failed when interest rates rose in the early 1980s.
  - Transition to nationwide funding via mortgage securitization: loans originated by mortgage banks were pooled and securitized by Government Sponsored Enterprises (GSEs).
  - Market-share shift: home mortgages held by depositories dropped from above 70 percent of the market in the 1970s to about 30 percent in recent years.
- Consequences of market-based structure:
  - Reduced volatility of mortgage lending: availability of funds no longer limited by local depository conditions or regional economic strength.
  - Monetary policy no longer has the same direct effect on the supply of mortgages through expansion of bank balance sheets.
  - Reported change in correlation: 0.72 between 1960 and 1984, to 0.06 between 1985 and 2004.

### III. Empirical approach and scope
- Empirical strategy:
  - Describe structural changes in mortgage finance and changing patterns of real housing activity and housing prices.
  - Present an econometric model of housing price dynamics based on economic fundamentals to analyze changes in price formation over time.
  - Explore whether the current securitized system reduces risks of speculative boom-bust cycles, including effects on homebuilders’ behavior and incentives to build inventories of new houses in advance of anticipated demand increases.
  - Analyze the role of residential investment in the decline of the volatility of real GDP.
- Claimed outcomes of the structural shift:
  - Smoother boom-bust cycles in lending flows, real housing activity, and housing prices.
  - Reduction in pricing errors relative to fundamental valuations.
  - Lower volatility of residential investment appears to contribute to decreased macroeconomic volatility.

### IV. House price model and regression findings
- Model specification (log differences):
  - hpi_{i,t} = β0 + β1 y_{i,t} + β2 U_{i,t} + β3 lf_{i,t} + β4 i_{t} + β5 cpi_{t} + β6 gdp_{t} + ε_{i,t}
  - hpi: OFHEO repeat sales housing price index for nine major regions.
  - y, U, lf: regional income, unemployment rates, labor force.
  - i, cpi, gdp: national mortgage interest rate, CPI inflation, GDP growth.
- Estimation approaches:
  - Stand-alone region regressions (coefficients vary by region).
  - Fixed-effects panel with location-specific constants.
- Key regression findings:
  - A 10 percent rise in regional incomes associated with a 2.5 percent increase in housing prices (columns 1–3).
  - A 1 percentage point rise in the unemployment rate depresses housing prices by about 1 percent.
  - GDP growth and CPI inflation generally not significant in panel estimates (CPI small and insignificant).
  - Interest rates not significant over full sample in some specifications; when estimated separately over time periods, interest rates show the expected negative effect and are significant at the 1 percent level in both earlier and later periods.
  - Model explains more than twice as much of overall variation in housing prices over 1990–2004 compared to 1978–2004:
    - Adjusted R-squared (panel regressions) reported as 0.11, 0.11, 0.15 (columns 1–3), 0.20 (1978–89), 0.39 (1990–2004).

### V. Pricing errors, securitization, and regional convergence
- Interpretation: estimation residuals treated as pricing errors; declines indicate improved price formation.
- Pricing errors by period (panel regressions, percent):
  - 1978–82: Inflation = High; Market Structure = Depository; Average Absolute = 6.2; Median Absolute = 4.8; Standard Deviation = 5.6.
  - 1983–90: Inflation = Low; Market Structure = Mixed; Average Absolute = 4.9; Median Absolute = 3.7; Standard Deviation = 4.2.
  - 1991–2004: Inflation = Low; Market Structure = Market; Average Absolute = 2.7; Median Absolute = 2.0; Standard Deviation = 2.5.
- Regional convergence:
  - Standard deviation of pricing errors across regions fell from 5.6 percent in the first part of the sample to 2.5 percent in the 1990s.
- Pricing error regressions (MBS share coefficient estimates, OLS):
  - Column (1): -0.064 (t-stat 4.59)**; Adjusted R-squared 0.31.
  - Column (2): -0.079 (t-stat 6.01)**; Adjusted R-squared 0.42.
  - Column (3): -0.083 (t-stat 5.88)**; Adjusted R-squared 0.40.
  - Column (4): -0.106 (t-stat 8.48)**; Adjusted R-squared 0.57.
- Economic magnitude:
  - A 10 percentage point rise in securitized share, all else equal, lowers the average absolute pricing error by 0.8 percentage points.
  - Rise in securitization from 10 percent to 60 percent or more between mid-1970s and 2004 associated with a 4 percentage point decline in absolute pricing errors — essentially accounting for the estimated improvement.
- CPI coefficients not statistically different from zero in all specifications.

### VI. Residential investment history and volatility
- Data: Real and nominal residential investment available since 1948.
- Historical volatility patterns:
  - High procyclicality from 1948 through late 1960s with episodes of more than 25 percent growth over 4 or more quarters; early 1950s surge nearly 50 percent.
  - Busts featured declines often at a 20 percent rate or faster.
- Recent patterns:
  - Residential investment and housing starts became markedly less cyclical since 1990.
  - Housing activity became more synchronized across regions beginning in 1990.
  - Since 1990, starts and investment spending have grown at a relatively stable pace, with recent starts growth still well below previous boom rates.
- Role of secondary markets:
  - GSEs and secondary mortgage markets have moderated the cyclicality of mortgage flows.

### VII. Will the current boom be followed by a bust?
- Recent price run-up:
  - Nationwide housing prices have risen 50 percent over the past five years.
  - Some metropolitan and regional markets experienced higher gains.
- Indicators suggesting speculative pressures:
  - Rise in purchases of second homes (investment purposes).
  - Increased use of interest-only mortgages.
  - Surveys and anecdotes of buyers extrapolating past gains into expectations.
  - Increased price-to-rent ratio in some markets.
- Mitigating factors and inventory dynamics:
  - Estimated pricing errors from the model are not particularly large; much recent gain explained by rising incomes, rising employment, and low interest rates.
  - Positive surprises over the past five years follow a decade of negative surprises in many regions (e.g., East and West Coasts), suggesting a catch-up.
  - Homebuilders’ behavior changed since 1980s:
    - Decline in speculative construction.
    - Inventories of new homes have only recently returned to early 1970s levels despite a doubling of sales.
  - Months’ supply dynamics:
    - During prior busts, months’ supply reached nearly 10 months on the west coast (early 1980s) and 15 months in the northeast (late 1980s); prices then fell as much as 10 percent to clear inventories.
    - Current inventories: less than 4 months’ supply in the northeast and less than 3 months on the west coast — below levels at which prices stabilized in earlier busts, suggesting supply has not greatly outpaced demand.

### VIII. Mortgage market structure and macroeconomic volatility
- Decline in GDP volatility:
  - Standard deviation of 4-quarter GDP growth decreased from 2.8 percent between 1960 and 1983 to 1.6 percent between 1984 and 2004.
- Contribution of residential investment:
  - Among GDP components, residential investment exhibited more than twice the volatility of any other item.
  - Annual standard deviation of residential investment growth fell from 17 percent to 7 percent.
- Mechanism and implication:
  - Shift from depository balance-sheet lending to securitization and national funding eliminated stop-go credit cycles in mortgage finance.
  - Smoother mortgage finance reduced volatility in real housing investment and turnover, and likely dampened volatility in consumption of durables linked to housing activity.
  - Structural change in mortgage markets provides a plausible and potentially enduring explanation for part of the decline in U.S. output volatility since the mid-1980s.

### IX. Conclusions — key takeaways
- Change from depository-based balance-sheet mortgage lending to market-based securitized finance:
  - Dampened volatility of mortgage financing flows and real housing activity.
  - Produced convergence of regional trends in real activity and prices.
  - Reduced pricing errors in a fundamentals-based housing price model by about half (median absolute errors fell from nearly 5 percent pre-1982 to 2 percent in 1991–2004).
- Securitization’s measured effect:
  - A 10 percentage point increase in securitized share reduces average absolute pricing error by 0.8 percentage points; the rise from 10 percent to 60 percent+ explains an estimated 4 percentage point decline in errors.
- Risk assessment:
  - Reduced speculative construction and lower inventories of new homes mitigate risk of a sharp nationwide price collapse, although speculative indicators are present in some metropolitan areas.
- Macroeconomic implication:
  - Structural changes in mortgage markets likely contributed materially to the reduction in GDP growth volatility and may make that stabilization more enduring.

*Source: _wp05200 (excerpt from IMF staff estimates and analysis).*

### References..............................................................................................................

### _wp05200 - References

### I. INTRODUCTION
- Research question: Why are housing markets prone to boom-bust cycles?
- Focus: Role of mortgage-market structure in boom-bust cycles in real housing activity and housing prices.
- Key mechanisms highlighted:
  - Pre-mid-1980s U.S. financial structure produced a stop-and-go credit cycle in response to changes in monetary policy and local banking conditions.
  - During easy-money periods: rapid expansions in real housing activity.
  - During tightening: depositories bore the brunt of monetary policy, contracting mortgage finance (cost of housing finance rising sharply or lending spigot shut off), causing real housing activity to contract and pent-up demand to build.
  - Price dynamics: tight-money episodes could push prices below long-term fundamental values; easing could bid prices above fundamentals, producing excess price volatility.
- Natural experiment: Post–Savings and Loan (S&L) crisis structural shift from local/regional balance-sheet lending to a national, market-based system of securitized mortgage finance.

### II. STRUCTURAL CHANGES IN MORTGAGE FINANCE
- Pre-mid-1980s structure:
  - Mortgages were mainly a local business; originated and held on balance sheets of local lenders (S&Ls/thrifts).
  - Availability of mortgage credit depended on local financial conditions and was constrained by restrictions on interstate banking.
  - Monetary policy and bank regulations (e.g., Regulation Q) amplified volatility in housing finance by affecting depositories’ funding and deposit rates.
  - Historical episodes: Real mortgage growth slowed sharply during several episodes in the 1960s when Reg Q ceilings became binding and banks and S&Ls were unable to retain deposits.
  - Under that structure, lending contracted with declining funding and rebounded as monetary policy eased and deposit growth resumed.
- S&L crisis and transition:
  - S&Ls had funded long-term fixed-rate mortgages with short-term deposits, creating unhedged interest rate exposures; many institutions failed when interest rates rose in the early 1980s.
  - Transition to nationwide funding via mortgage securitization: loans originated by mortgage banks were pooled and securitized by Government Sponsored Enterprises (GSEs).
  - Market-share shift: home mortgages held by depositories dropped from above 70 percent of the market in the 1970s to about 30 percent in recent years (Figure 2).
- Consequences of market-based structure:
  - Reduced volatility of mortgage lending: availability of funds no longer limited by local depository conditions or regional economic strength.
  - Monetary policy no longer has the same direct effect on the supply of mortgages through expansion of bank balance sheets.
  - Result noted: the correlation between growth of real deposits and real mortgage flows dropped (text ends at this point).

### III. Analytical Approach and Scope (as described)
- Empirical strategy:
  - Description of structural changes in mortgage finance and changing patterns of real housing activity and housing prices.
  - Presentation of an econometric model of housing price dynamics based on economic fundamentals to analyze changes in price formation over time.
  - Exploration of whether the current securitized system reduces risks of speculative boom-bust cycles, including effects on homebuilders’ behavior and incentives to build inventories of new houses in advance of anticipated demand increases.
  - Final analysis of the role of residential investment in the decline of the volatility of real GDP.

- Claimed outcomes of the structural shift:
  - Smoother boom-bust cycles in lending flows, real housing activity, and housing prices.
  - Reduction in pricing errors relative to fundamental valuations.
  - Lower volatility of residential investment appears to contribute to decreased macroeconomic volatility.

*Source: _wp05200 - References (excerpt).*

### 0.72 between 1960 and 1984, to 0.06 between 1985 and 2004. Tighter or looser monetary

### _wp05200 - 0.72 between 1960 and 1984, to 0.06 between 1985 and 2004. Tighter or looser monetary

### Changes in mortgage market structure and housing volatility
- Mortgage market funding shifted from local on-balance-sheet depository lending toward a national securitized market (including GSEs and securitization).
- Coincident changes:
  - Residential investment and housing starts became markedly less cyclical since 1990.
  - Housing activity became more synchronized across regions beginning in 1990.
  - Housing prices converged toward steadier growth since the early 1990s; regional price swings declined in amplitude compared with the 1970s and 1980s.
- Historical patterns:
  - Boom-bust phases between the late 1960s and 1980s: residential investment growth rates of 40 percent or more during booms and declines of nearly the same magnitude during busts.
  - Since 1990, starts and investment spending have grown at a relatively stable pace, with recent starts growth still well below previous boom rates.
- Role of secondary markets:
  - GSEs and secondary mortgage markets have moderated the cyclicality of mortgage flows (Peek and Wilcox (2005) referenced).

### House prices and fundamentals — empirical approach
- Model estimated in log differences; no evidence of cointegration between regional housing prices and income (reference to Gallin (2003)).
- Baseline regression specification:
  - hpi_{i,t} = β0 + β1 y_{i,t} + β2 U_{i,t} + β3 lf_{i,t} + β4 i_{t} + β5 cpi_{t} + β6 gdp_{t} + ε_{i,t}
  - hpi: OFHEO repeat sales housing price index for nine major regions.
  - y, U, lf: regional income, unemployment rates, labor force.
  - i, cpi, gdp: national mortgage interest rate, CPI inflation, GDP growth.
- Estimation approaches:
  - Stand-alone region regressions (coefficients vary by region).
  - Fixed-effects panel with location-specific constants.

### Regression findings and fit across periods (Table 1 highlights)
- Fundamentals effects (panel estimates):
  - A 10 percent rise in regional incomes associated with a 2.5 percent increase in housing prices (columns 1–3).
  - A 1 percentage point rise in the unemployment rate depresses housing prices by about 1 percent.
- National business cycle indicators:
  - GDP growth and CPI inflation generally not significant in panel estimates (CPI small and insignificant).
- Interest rates:
  - Interest rates not significant over full sample in some specifications, possibly reflecting endogeneity.
  - When estimated separately over time periods, interest rates show the expected negative effect and are significant at the 1 percent level in both earlier and later periods.
- Model fit improvements:
  - Model explains more than twice as much of overall variation in housing prices over 1990–2004 compared to estimates over the entire 1978–2004 horizon.
  - Adjusted R-squared for panel regressions: 0.11, 0.11, 0.15 (columns 1–3), 0.20 (1978–89), 0.39 (1990–2004).

### Pricing errors, securitization, and regional convergence (Tables 2–3; Figures 6–7)
- Interpretation: estimation residuals treated as pricing errors; declines indicate improved price formation.
- Pricing errors by period (Table 2, Panel regressions, percent):
  - 1978–82: Inflation = High; Market Structure = Depository; Average Absolute = 6.2; Median Absolute = 4.8; Standard Deviation = 5.6.
  - 1983–90: Inflation = Low; Market Structure = Mixed; Average Absolute = 4.9; Median Absolute = 3.7; Standard Deviation = 4.2.
  - 1991–2004: Inflation = Low; Market Structure = Market; Average Absolute = 2.7; Median Absolute = 2.0; Standard Deviation = 2.5.
- Regional convergence:
  - Standard deviation of pricing errors across regions fell from 5.6 percent in the first part of the sample to 2.5 percent in the 1990s.
- Pricing error regressions (Table 3):
  - Dependent variables: Average Absolute Pricing Error by regions; Variance of Pricing Error by regions; Average Absolute Pricing Error from panel; Variance from panel.
  - MBS (securitized) share coefficient estimates (OLS):
    - Column (1): -0.064 (t-stat 4.59)**; Adjusted R-squared 0.31.
    - Column (2): -0.079 (t-stat 6.01)**; Adjusted R-squared 0.42.
    - Column (3): -0.083 (t-stat 5.88)**; Adjusted R-squared 0.40.
    - Column (4): -0.106 (t-stat 8.48)**; Adjusted R-squared 0.57.
  - CPI coefficients not statistically different from zero in all specifications.
- Economic magnitude:
  - A 10 percentage point rise in securitized share, all else equal, lowers the average absolute pricing error by 0.8 percentage points.
  - Rise in securitization from 10 percent to 60 percent or more between mid-1970s and 2004 associated with a 4 percentage point decline in absolute pricing errors — essentially accounting for the estimated improvement.

### Historical test using residential investment (Figure 8)
- Real and nominal residential investment data available since 1948.
- Residential investment volatility:
  - High procyclicality from 1948 through late 1960s with episodes of more than 25 percent growth over 4 or more quarters; early 1950s surge nearly 50 percent.
  - Busts featured declines often at a 20 percent rate or faster.
- Implication: elevated volatility in prior decades likely linked to mortgage market structure rather than solely to inflation or macro shocks.

### Will the current boom be followed by a bust?
- Recent price run-up:
  - Nationwide housing prices have risen 50 percent over the past five years (text statement).
  - Some metropolitan and regional markets experienced higher gains.
- Indicators suggesting speculative pressures:
  - Rise in purchases of second homes (investment purposes).
  - Increased use of interest-only mortgages.
  - Surveys and anecdotes of buyers extrapolating past gains into expectations.
  - Increased price-to-rent ratio in some markets.
- Mitigating factors:
  - Estimated pricing errors from the model are not particularly large; much recent gain explained by rising incomes, rising employment, and low interest rates.
  - Positive surprises over the past five years follow a decade of negative surprises in many regions (e.g., East and West Coasts), suggesting a catch-up.
  - Homebuilders’ behavior changed since 1980s:
    - Decline in speculative construction (Figure 10).
    - Inventories of new homes have only recently returned to early 1970s levels despite a doubling of sales (Figure 11).
  - Months’ supply dynamics:
    - During prior busts, months’ supply reached nearly 10 months on the west coast (early 1980s) and 15 months in the northeast (late 1980s); prices then fell as much as 10 percent to clear inventories.
    - Current inventories: less than 4 months’ supply in the northeast and less than 3 months on the west coast — below levels at which prices stabilized in earlier busts, suggesting supply has not greatly outpaced demand.

### Mortgage market structure and macroeconomic volatility
- Decline in GDP volatility:
  - Standard deviation of 4-quarter GDP growth decreased from 2.8 percent between 1960 and 1983 to 1.6 percent between 1984 and 2004.
- Contribution of residential investment:
  - Among GDP components, residential investment exhibited more than twice the volatility of any other item (Figure 14).
  - Annual standard deviation of residential investment growth fell from 17 percent to 7 percent (Figure 15).
- Mechanism:
  - Shift from depository balance-sheet lending to securitization and national funding eliminated stop-go credit cycles in mortgage finance.
  - Smoother mortgage finance reduced volatility in real housing investment and turnover, and likely dampened volatility in consumption of durables linked to housing activity.
- Implication:
  - Structural change in mortgage markets provides a plausible and potentially enduring explanation for part of the decline in U.S. output volatility since the mid-1980s.

### Conclusions (key takeaways)
- Change from depository-based balance-sheet mortgage lending to market-based securitized finance:
  - Dampened volatility of mortgage financing flows and real housing activity.
  - Produced convergence of regional trends in real activity and prices.
  - Reduced pricing errors in a fundamentals-based housing price model by about half (median absolute errors fell from nearly 5 percent pre-1982 to 2 percent in 1991–2004).
- Securitization’s measured effect:
  - A 10 percentage point increase in securitized share reduces average absolute pricing error by 0.8 percentage points; the rise from 10 percent to 60 percent+ explains an estimated 4 percentage point decline in errors.
- Reduced speculative construction and lower inventories of new homes mitigate risk of a sharp nationwide price collapse, although speculative indicators are present in some metropolitan areas.
- Structural changes in mortgage markets likely contributed materially to the reduction in GDP growth volatility and may make that stabilization more enduring.

*Source: IMF staff estimates and analysis (PDF content: _wp05200).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2005/_wp05200.pdf_
