## _wp09252

## Source details

**Canonical URL:** [_wp09252](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09252.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09252.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09252.pdf.json)

---

### I. Introduction — purpose and high-level findings
- Purpose: Examine historical evidence for consistent macroeconomic patterns that could serve as reliable leading indicators of asset price busts.
- High-level findings:
  - Inflation and output do not typically display unusual behavior ahead of asset price busts.
  - Credit, the share of investment in GDP, current account deficits, and asset prices typically rise before busts and provide useful leading indicators.
  - Since 1985, the stance of monetary policy—as measured by real policy rates or deviations from a Taylor rule—has generally not been a good leading indicator of future house price busts.
  - Some evidence exists of loose monetary policy in the years leading up to the current crisis in some countries, but it is not the main, systematic cause of the boom and consequent bust.
  - Monetary policymakers are more plausibly culpable for acting too narrowly and not reacting strongly enough to growing financial vulnerability.

### II. Asset price busts — definitions, sample, frequency, severity
- Definition of a bust:
  - A bust occurs when the four-quarter trailing moving average of the annual growth rate of the asset price, in real terms, falls below a specified threshold.
  - Thresholds used:
    - House prices: –5 percent
    - Stock prices: –20 percent
  - Operational rule: If the condition x < g_{t-3}+g_{t-2}+g_{t-1}+g_t (with the appropriate threshold x) holds, periods t-3 through t are labeled as a bust.
- Sample and data:
  - Quarterly observations for 21 advanced economies from 1970 to 2008.
  - Countries include: Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and United States.
- Frequency and severity (1970–2008):
  - Identified busts:
    - House price busts: 47
    - Stock price busts: 98
  - Duration:
    - Average house price bust lasts for two and a half years.
    - Average stock price bust lasts for about one and a half years.
  - Output loss:
    - Cumulative decline in output below trend is roughly 4¼ percent for the first year after the onset of a house price bust.
    - Cumulative decline is 1¼ percent after stock price busts.
  - Distribution:
    - Asset price busts are relatively evenly distributed before and after 1985.
    - Clusters in 1974-75, 1983, 1992, and 2008; the 2008 cluster is the most widespread.

### III. Detrending approach and methodology
- Primary detrending method:
  - Trailing eight-year moving average used as filter to isolate abnormal movements.
  - Rationale: easily reproducible; trends are fairly slow moving; one-sided filter does not include future information.
- Caveats of one-sided filter:
  - For temporary but persistent increases in growth, deviation from trailing moving average eventually shrinks as trend catches up, potentially masking ongoing high growth.
  - Eight-year window mitigates this by lengthening period required for trend catch-up.
- Robustness:
  - Rolling HP filter with smoothness parameter set to 400,000 following Borio and Lowe (2004) used as a check; qualitatively similar patterns obtained.

### IV. Macroeconomic patterns around busts
- General common run-up patterns (two to three years before a bust):
  - Significant expansion in domestic credit.
  - Asset price booms.
  - Rising investment shares (including residential investment).
  - Current account deficits (notably for house price busts).
- House price busts — post-1985 run-ups:
  - Higher-than-normal growth rates of credit relative to GDP.
  - Large deteriorations in current account balances.
  - Higher-than-normal ratios of investment to GDP (including residential investment).
  - Asset prices grow faster than the eight-year trend, though not significantly different from zero within two years before busts.
  - Output growth does not display significant deviation from trend.
  - Inflation is below its eight-year moving average.
- House price busts — pre-1985 run-ups:
  - No pattern of rapid increases in credit relative to GDP or deteriorating current account balances.
  - Large deviations in inflation coinciding with the two oil crises.
- Stock price busts:
  - Post-1985: large increases in credit growth and in the ratio of investment to GDP during the run-up to stock price busts.
  - Key differences vs house price busts:
    - Output growth tends to be significantly higher than trend during the run-up to stock price busts.
    - No deterioration in current account balances comparable to that seen for house price busts; median current account balance deteriorates in the year leading up to a stock price bust but level not significantly different from zero.

### V. Alarm-based predictive methodology (KLR approach) and predictive performance
- Alarm methodology:
  - Large movements defined as deviations from an eight-year moving average; alarms raised when deviation exceeds a percentile-based threshold.
  - Percentiles considered: 70th, 75th, 80th and 90th; thresholds chosen to minimize noise-to-signal ratio.
  - Real-time percentile computation uses observations over the previous 15 years; statistics calculated for post-1985 period.
  - Time horizon for a legitimate alarm: bust occurs within three years with a minimum lead time of one year.
  - Classification matrix cells: A (alarm and bust within horizon), B (alarm but no bust), C (no alarm but bust), D (no alarm, no bust).
  - Summary statistics:
    - Conditional probability of a bust given an alarm: A/(A+B).
    - Proportion of missed busts: C/(A+C).
- Unconditional probabilities (1–3 years ahead, 1985–2008):
  - House price bust: 14 percent.
  - Stock price bust: 29 percent.
- Indicators with strong predictive power for house price busts (post-1985):
  - Large deviations in credit-to-GDP, current-account-to-GDP, residential-investment-to-GDP, and house prices.
  - Example: large deviations in credit-to-GDP are associated with a 28 percent probability of a house price bust one to three years in the future (twice the unconditional 14 percent).
- Combined signals:
  - When credit, current account, and residential investment all raise alarms simultaneously, 56 percent of these occasions were associated with a house price bust one to three years in the future.
- Limitations:
  - Even the best single indicator (credit) raised alarms only in one-half of all cases; others raised alarms only one-quarter to one-half of the time.
  - Large deviations in these macro variables still fail to predict roughly one-half of busts since 1985.

### VI. Probit analysis — marginal effects and statistical significance
- Model setup:
  - Dependent variable = 1 if a bust occurs between one and three years ahead, 0 otherwise.
  - Variables measured as deviations from an eight-year moving average.
  - Marginal probabilities computed at mean values.
- Key marginal effects (post-1985 sample, evaluated at means):
  - House price busts:
    - A 10 percentage point increase in the credit-to-GDP ratio relative to an 8-year moving average increases the probability of a house price bust by 4.4 percent (about one third higher than the unconditional probability of about 15 percent).
    - A 1½ percentage point deterioration of the current account relative to its 8-year moving average implies a one third increase in the probability of a house price bust over the unconditional probability.
    - Current account balances, residential investment, and house price growth are significant predictors.
    - Output growth and inflation are not significantly associated with the likelihood of a house price bust.
  - Stock price busts:
    - A 10 percentage point increase in the credit-to-GDP ratio is associated with a 6.4 percent increase in the probability of a stock price bust (roughly 20 percent higher than the unconditional probability of a stock price bust).
    - The coefficient on current account balances for stock price busts appears to have the wrong sign for the latter half of the sample.
- Selected exact probit marginal probabilities and significance (excerpt from Table 4; marginal probabilities at mean values; z-statistics in parentheses; ***, **, * at 1, 5, 10 percent):
  - Full Sample (N = 1699; Pseudo R2 = 0.14)
    - Credit/GDP (house price bust): 0.241** (2.180)
    - Credit/GDP (stock price bust): 0.546*** (4.070)
    - Current account balance (house price bust): –3.910*** (–7.560)
    - Residential investment/GDP (stock price bust): 6.392*** (5.280)
    - House price growth (house price bust): 0.798*** (5.240)
    - Output growth (stock price bust): 1.686** (2.540)
    - Inflation (house price bust): 2.511*** (7.180)
  - 1985–2008 (N = 1264; Pseudo R2 = 0.15 for house price bust; 0.10 for stock price bust)
    - Credit/GDP (house price bust): 0.443*** (4.280)
    - Credit/GDP (stock price bust): 0.638*** (4.210)
    - Current account balance (house price bust): –3.191*** (–5.440)
    - Residential investment/GDP (house price bust): 2.456* (1.930)
    - Residential investment/GDP (stock price bust): 7.327*** (5.290)
    - House price growth (house price bust): 0.455*** (2.910)
    - Stock price growth (stock price bust): 0.349*** (4.660)
    - Inflation (stock price bust): 3.732*** (4.130)
  - Note (exact text): "Dependent variable takes a value of 1 if there is a bust between 12 and 4 quarters ahead and zero otherwise. Estimation is carried out using robust standard errors. Z-statistics are reported in parentheses. ***, **, and * refer to significance at the 1, 5, and 10 percent level, respectively. Marginal probabilities computed at the mean values of other variables are reported. Variables are measured as deviations from an eight-year moving average."

### VII. Macroeconomic patterns ahead of the recent crisis (2001:Q4–2008:Q4)
- House prices and cross-country variation:
  - Real house prices rose strongly on average up until 2007; subsequent severe falls in most economies.
  - 2001:Q4–2006:Q3 real house price rises were strong in Ireland, New Zealand, Spain; fell in Austria, Germany, Japan.
  - Correlation between house price rises and subsequent falls (excluding Germany, Japan): 0.79.
- Indicators in run-up to recent house price busts:
  - Signs of residential investment booms present in at least half the economies one to three years before bust onset.
  - Credit growth unusually high in roughly half the economies over almost the entire three-year period.
  - Current account alarms more muted until about one year ahead, when raised for nearly half the countries.
- Cross-country associations:
  - Economies with largest house price appreciations (2001:Q4–2006:Q3) had large increases in residential investment share of GDP, large current account deficits as a share of GDP, and large expansions of credit relative to output growth.
  - Stronger credit growth typically matched by more severe deteriorations in household balance sheets (liabilities to liquid assets quick ratio highly associated with house price growth).

### VIII. Monetary policy and the recent boom-bust
- Two criticisms of monetary policy:
  1. Policy too loose from 2002 to 2006—policy rates below levels implied by simple Taylor-type rules.
  2. Monetary policy focused too narrowly on CPI inflation and output gap and should have leaned against asset price booms/financial vulnerability.
- Evidence reported:
  - Real interest rates and Taylor residuals in the United States, euro area, and other advanced economies (except Japan) were low and even negative leading up to the crisis (Figure 11 referenced).
  - Core inflation stayed within 1–3 percent during the period of credit expansion and asset price booms.
  - Cross-country relationships between monetary policy stance and house price gains are weak:
    - Within the euro area there is at best a weak association.
    - In full sample associations are weak or virtually nonexistent (examples: Ireland and Spain low real short-term rates & large house price rises; Australia, New Zealand, UK relatively high real rates & large house price rises).
    - Association between monetary policy stance and real stock price growth is extremely weak.
    - Association between monetary policy conditions and credit expansion rates is very weak.
- Interpretation:
  - Monetary policy was not the main or systematic source of excesses leading to recent price busts.
  - Evidence consistent with criticism that policymakers paid too little attention to emerging macrofinancial risks (acted too narrowly).

### IX. Policy implications and trade-offs
- Indicators to monitor beyond inflation and output:
  - Excessive credit expansion (credit-to-GDP deviations).
  - Overinvestment, particularly residential-investment-to-GDP share.
  - Deteriorating external balances (current-account-to-GDP).
- Limitations of policy rates as a tool:
  - Not immediately clear that policy rates are the appropriate instrument to address these vulnerabilities.
  - A macroprudential tool that works directly on lending margins can be useful to tackle problems in financial markets and limit the need for aggressive monetary policy reactions (companion paper explored).
- Institutional implications:
  - Coordination of monetary and macroprudential policy is important.
  - May require expansion of central bank mandates to include concern for financial vulnerabilities.
- Trade-off and realism:
  - Even the best leading indicators are imperfect; in attempting to reduce bust probability, central banks may raise costly false alarms.
  - Policy design must weigh the trade-off between reducing the probability of a dangerous bust and the risk of raising costly false alarms.

### X. Key tables and figures (selected exact statistics)
- Table 1 highlights (mean values, 1970–2008):
  - Total number of busts: House prices 47; Stock prices 98.
  - Before 1985: House prices 22; Stock prices 41.
  - 1985–2008: House prices 25; Stock prices 57.
  - Cumulative decline in prices (percent): House prices –17.71; Stock prices –37.38 (Full sample).
  - Duration (quarters): House prices 10.02; Stock prices 6.98 (Full sample).
  - Cumulative decline in output (percent relative to trend): House prices –4.27; Stock prices –1.31 (Full sample).
- Table 3 (percentiles used as thresholds where noise-to-signal ratio minimized):
  - Credit/GDP: House Price Bust 90; Stock Price Bust 90.
  - Current account/GDP: House Price Bust 90; Stock Price Bust 90.
  - Residential investment/GDP: House Price Bust 90; Stock Price Bust 90.
  - House price growth: House Price Bust 90; Stock Price Bust 70.
  - Stock price growth: House Price Bust 70; Stock Price Bust 70.
  - Growth: House Price Bust 75; Stock Price Bust 80.
  - Inflation: House Price Bust 90; Stock Price Bust 90.
- Figures report median percent deviations from trailing eight-year moving averages and other correlations (selected reported R-squared values):
  - R  = 0.54 2 for Real House Prices and Residential Investment.
  - R  = 0.24 2 for Real House Prices and Current Account.
  - R  = 0.37 2 for Real House Prices and Growth Rate of Nominal Credit relative to GDP.
  - R  = 0.60 2 for Growth Rate of Nominal Credit relative to GDP and Household Quick Ratio.
  - Scatterplot correlations between house prices/stock prices and monetary conditions reported as low (examples: R  = 0.17 2, R  = 0.03 2, R  = 0.05 2, R  = 0.19 2, R = 0.00 2, R = 0.007 2).
  - Growth Rate of Nominal Credit Relative to GDP and Real Policy Rates reported R = 0.02 2.

### XI. Summary conclusions
- Asset price busts have typically been preceded by rising investment, expanding credit, and deteriorating current account balances.
- Large deviations in these variables from local trends have some value as indicators of future asset price busts, often increasing the probability of a bust by more than twice the unconditional probability when alarms are raised.
- Nonetheless, asset price busts remain difficult to predict: even the best indicator failed to raise an alarm one to three years ahead of roughly one-half of all busts since 1985.
- Monetary policy stance (real rates, Taylor residuals), and conventional measures of inflation and output, are poor leading indicators of asset price busts.
- If avoiding asset price busts is a policy concern, a narrow focus on inflation and output is insufficient; monitoring and tools addressing credit expansion, overinvestment, and external imbalances are warranted, and macroprudential instruments plus coordination with monetary policy deserve greater emphasis.

*Source: _wp09252 - References .................................................................................................17*

### References .................................................................................................17

### _wp09252 - References .................................................................................................17

### I. INTRODUCTION
- Purpose: Examine historical evidence for consistent macroeconomic patterns that could serve as reliable leading indicators of asset price busts.
- Key high-level findings:
  - Inflation and output do not typically display unusual behavior ahead of asset price busts.
  - Credit, the share of investment in GDP, current account deficits, and asset prices typically rise before busts and provide useful leading indicators.
  - Since 1985, the stance of monetary policy—as measured by real policy rates or deviations from a Taylor rule—has generally not been a good leading indicator of future house price busts.
  - Some evidence exists of loose monetary policy in the years leading up to the current crisis in some countries, but it is not the main, systematic cause of the boom and consequent bust. Monetary policymakers are more plausibly culpable for acting too narrowly and not reacting strongly enough to growing financial vulnerability.
- Contribution and relation to literature:
  - Focuses separately on house prices and stock prices, identifying recurring patterns such as deteriorating current account balances in the run-up to house price busts and differences in patterns before and after 1985.
  - Links to literature on asset price booms and early warning systems (Kaminsky, Lizondo and Reinhart (1998); Kaminsky and Reinhart (1999); Borio and Lowe (2002); Gerdesmeier, Reimers and Roffia (2009); Adalid and Detken (2007); Detken and Smets (2004); Mendoza and Terrones (2008); Berg et. al., 2000).

### II. ASSET PRICE BUSTS IN THE MODERN ERA — DEFINITIONS AND SAMPLE
- Definition of an asset price bust:
  - A bust occurs when the four-quarter trailing moving average of the annual growth rate of the asset price, in real terms, falls below a specified threshold.
  - Thresholds used:
    - House prices: –5 percent
    - Stock prices: –20 percent
  - Operational rule: If the condition x < g_{t-3}+g_{t-2}+g_{t-1}+g_t (with the appropriate threshold x) holds, periods t-3 through t are labeled as a bust.
  - Rationale: higher absolute threshold for stock prices because stock prices are typically more volatile.
- Sample and data:
  - Quarterly observations on asset prices and macroeconomic variables for 21 advanced economies from 1970 to 2008.
  - Countries included (subject to data limitations): Australia, Austria, Belgium, Canada, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Japan, Netherlands, New Zealand, Norway, Portugal, Spain, Sweden, Switzerland, United Kingdom, and United States.
- Frequency and severity of busts (1970–2008):
  - Identified busts:
    - House price busts: 47
    - Stock price busts: 98
  - Duration:
    - Average house price bust lasts for two and a half years.
    - Average stock price bust lasts for about one and a half years.
  - Output loss:
    - Cumulative decline in output below trend is roughly 4¼ percent for the first year after the onset of a house price bust.
    - Cumulative decline is 1¼ percent after stock price busts.
  - Distribution:
    - Asset price busts are relatively evenly distributed before and after 1985.
    - Clusters of episodes occurred in 1974-75, 1983, 1992, and 2008; the 2008 cluster is the most widespread for both house and stock price busts.

### III. DETRENDING APPROACH AND METHODOLOGY
- Need to correct for slow-moving trends (e.g., lower inflation in 1990s vs. 1970s; diminishing output growth).
- Primary detrending method:
  - Trailing eight-year moving average used as filter to isolate abnormal movements.
  - Reasons for choice:
    - Easily reproducible.
    - Trends are fairly slow moving.
    - One-sided filter does not include future information.
- Caveats of one-sided filter:
  - For variables with temporary but persistent increases in growth, deviation from trailing moving average eventually shrinks as trend catches up, potentially masking ongoing high growth.
  - Eight-year window mitigates this by lengthening period required for trend catch-up.
- Robustness check:
  - Analysis also carried out using a rolling HP filter with a slow-moving trend.
  - Smoothness parameter set to 400,000 following Borio and Lowe (2004).
  - Qualitatively similar patterns obtained (Figures 16 and 17).

### IV. PATTERNS IN MACROECONOMIC VARIABLES AROUND BUSTS
- General observation:
  - Theory suggests timing of large asset price drops is hard to predict; nevertheless, systematic patterns in macro variables may indicate increased likelihood of a bust.
- Findings for house price busts (split pre-1985 and 1985 onward):
  - Post-1985 run-ups to house price busts feature:
    - Higher-than-normal growth rates of credit relative to GDP.
    - Large deteriorations in current account balances.
    - Higher-than-normal ratios of investment to GDP.
    - Asset prices grow faster than the eight-year trend, though not significantly different from zero within two years before busts.
    - Output growth does not display significant deviation from trend.
    - Inflation is below its eight-year moving average.
  - Pre-1985 run-ups:
    - No pattern of rapid increases in credit relative to GDP or deteriorating current account balances in the run-up to busts.
    - Large deviations in inflation coinciding with the two oil crises.
- Findings for stock price busts:
  - Post-1985 period shows similar patterns as house price busts: large increases in credit growth and in the ratio of investment to GDP during the run-up to stock price busts.

### V. MONETARY POLICY AND THE CURRENT CRISIS
- Questions posed to policymakers:
  - Did policymakers create the current crisis by reacting insufficiently to growing inflationary pressure?
  - Did policymakers increase the likelihood of an asset price bust by placing insufficient weight on credit and asset prices when setting interest rates?
- Evidence and interpretation:
  - Some evidence of loose monetary policy in years leading up to the current crisis in some countries.
  - Loose monetary policy is not identified as the main, systematic cause of the boom and consequent bust.
  - If monetary policymakers bear blame, it is primarily for acting too narrowly—failing to react strongly enough to signs of growing financial vulnerability.

### VI. STRUCTURE OF THE PAPER (AS PRESENTED)
- Section 2: Establishes stylized facts about asset price booms and busts and looks for potential leading indicators of asset price busts.
- Section 3: Assesses predictive power of candidate indicators.
- Section 4: Examines macroeconomic patterns leading up to the current crisis, with emphasis on monetary policy’s role.
- Section 5: Conclusions.
- Appendix: Data sources and details.

*Source: _wp09252 - References .................................................................................................17*

### 3. There are, however, two notable differences between the behavior of macroeconomic

### 3. There are, however, two notable differences between the behavior of macroeconomic

### Key differences: stock price busts vs house price busts
- Output growth tends to be significantly higher than trend during the run-up to stock price busts.
- There is no deterioration in current account balances for stock price busts comparable to that seen for house price busts; the median current account balance deteriorates in the year leading up to a stock price bust but the level is not significantly different from zero.

### Macroeconomic patterns preceding asset price busts (general)
- Common run-up patterns (two to three years before a bust):
  - Significant expansion in domestic credit.
  - Asset price booms.
  - Rising investment shares (including residential investment).
  - Current account deficits (notably for house price busts).
- For particularly costly house price busts (1985–2008), the following are higher in costly busts than in less costly ones:
  - Growth rate of credit relative to output.
  - Share of residential investment in GDP.
  - Rate of increase of house prices.
- No significant difference in inflation and output growth in the run-up to high-cost vs other house price busts.

### Monetary policy indicators as leading signals
- Since 1985, monetary policy conditions are generally not good leading indicators of house price or stock price busts.
- Two measures tracked: real policy rates and deviation from a standard Taylor rule (weights: 1.5 on inflation deviation; 0.5 on output gap).
- Evidence:
  - Pre-1985: some loose monetary policy in the run-up to busts.
  - Post-1985: real policy rates were typically above trend in the run-up to house price busts and high relative to Taylor-rule-implied rates; real rates often increased leading up to a bust.
- Conclusion: monetary policy stance (real rates, Taylor residuals) has limited predictive value for busts over the post-1985 period.

### Alarm-based predictive methodology (Kaminsky–Lizondo–Reinhart approach)
- Large movements defined as deviations from an eight-year moving average; alarms raised when deviation exceeds a percentile-based threshold.
- Percentiles considered in grid search: 70th, 75th, 80th and 90th.
- Thresholds chosen to minimize the noise-to-signal ratio (ratio of false alarms to legitimate alarms).
- Real-time percentile computation uses observations over the previous 15 years; statistics calculated only for the post-1985 period.
- Time horizon for a legitimate alarm: bust occurs within three years with a minimum lead time of one year.
- Classification matrix: A (alarm and bust within horizon), B (alarm but no bust), C (no alarm but bust), D (no alarm, no bust).
- Two summary statistics:
  - Conditional probability of a bust given an alarm: A/(A+B).
  - Proportion of missed busts: C/(A+C).

### Predictive performance (post-1985 results)
- Unconditional probabilities (one to three years ahead):
  - House price bust: 14 percent.
  - Stock price bust: 29 percent.
- Indicators with strong predictive power for house price busts:
  - Large deviations in credit-to-GDP, current-account-to-GDP, residential-investment-to-GDP, and house prices.
  - Example: large deviations in credit-to-GDP are associated with a 28 percent probability of a house price bust one to three years in the future (twice the unconditional 14 percent).
- For stock price busts:
  - Output and inflation perform slightly better than for house price busts, but credit, current account balance, and residential investment remain important predictors.
- Combined signals:
  - When credit, current account, and residential investment all raise alarms simultaneously, 56 percent of these occasions were associated with a house price bust one to three years in the future.
- Limitations:
  - Even the best single indicator (credit) raised alarms only in one-half of all cases; others raised alarms only one-quarter to one-half of the time.
  - Thus, large deviations in these macro variables substantially increase bust probability but still fail to predict roughly one-half of busts since 1985.

### Probit analysis (addressing correlation and statistical significance)
- Probit model explains a binary outcome: 1 if a bust occurs between one and three years ahead, 0 otherwise.
- Variables measured as deviations relative to the 8-quarter trailing moving average.
- Key marginal effects (post-1985 sample, evaluated at means):
  - For house price busts:
    - A 10 percentage point increase in the credit-to-GDP ratio relative to an 8-year moving average increases the probability of a house price bust by 4.4 percent (about one third higher than the unconditional probability of about 15 percent).
    - A 1½ percentage point deterioration of the current account relative to its 8-year moving average implies a one third increase in the probability of a house price bust over the unconditional probability.
    - Current account balances, residential investment, and house price growth are significant predictors.
    - Output growth and inflation are not significantly associated with the likelihood of a house price bust.
  - For stock price busts:
    - A 10 percentage point increase in the credit-to-GDP ratio is associated with a 6.4 percent increase in the probability of a stock price bust (roughly 20 percent higher than the unconditional probability of a stock price bust).
    - The coefficient on current account balances for stock price busts appears to have the wrong sign for the latter half of the sample.

### Macroeconomic patterns ahead of the current crisis (2001:Q4–2008:Q4 period)
- House prices:
  - Real house prices rose strongly on average up until 2007; subsequent severe falls in most economies.
  - Cross-country variation: 2001:Q4–2006:Q3 real house price rises were strong in Ireland, New Zealand, Spain; fell in Austria, Germany, Japan.
  - Correlation between house price rises and subsequent falls (excluding Germany, Japan): 0.79.
- Indicators in the run-up to recent house price busts:
  - Signs of residential investment booms (sometimes funded by current account declines) present in at least half the economies one to three years before bust onset.
  - Credit growth unusually high in roughly half the economies over almost the entire three-year period.
  - Current account alarms more muted until about one year ahead, when raised for nearly half the countries.
- Cross-country associations:
  - Economies with largest house price appreciations (2001:Q4–2006:Q3) had large increases in residential investment share of GDP, large current account deficits as a share of GDP, and large expansions of credit relative to output growth.
  - Stronger credit growth typically matched by more severe deteriorations in household balance sheets (liabilities to liquid assets quick ratio highly associated with house price growth).

### The role of monetary policy in the recent boom-bust
- Two criticisms of monetary policy:
  1. Policy was too loose from 2002 to 2006—policy rates below levels implied by simple rules reacting to output gap and inflation (Taylor rule argument).
  2. Monetary policy focused too narrowly on CPI inflation and output gap and should have leaned against asset price booms/financial vulnerability at the cost of more inflation/output variability.
- Evidence:
  - Figure indications: real interest rates and Taylor residuals in the United States, euro area, and other advanced economies (except Japan) were low and even negative leading up to the crisis.
  - Core inflation stayed within 1–3 percent during the period of credit expansion and asset price booms.
  - Cross-country relationships between monetary policy stance and house price gains are weak:
    - Within the euro area there is at best a weak association.
    - In the full sample, associations are weak or virtually nonexistent (examples: Ireland and Spain low real short-term rates & large house price rises; Australia, New Zealand, UK relatively high real rates & large house price rises).
    - Association between monetary policy stance and real stock price growth is extremely weak.
    - Association between monetary policy conditions and credit expansion rates is very weak.
- Interpretation:
  - Monetary policy was not the main or systematic source of excesses leading to recent price busts.
  - Evidence is more consistent with the criticism that monetary policymakers paid too little attention to emerging macrofinancial risks (i.e., acted too narrowly).

### Policy implications and recommendations
- Indicators to monitor beyond inflation and output:
  - Excessive credit expansion (credit-to-GDP deviations).
  - Overinvestment, particularly residential-investment-to-GDP share.
  - Deteriorating external balances (current-account-to-GDP).
- Limitations of using policy rates as the primary tool:
  - It is not immediately clear that policy rates are the appropriate instrument to address these vulnerabilities.
  - A macroprudential tool that works directly on lending margins can be useful to tackle problems in financial markets and limit the need for aggressive monetary policy reactions (as explored in a companion paper).
- Institutional implications:
  - Coordination of monetary and macroprudential policy is important.
  - May require an expansion of central bank mandates to include concern for financial vulnerabilities.

### Summary conclusions
- Asset price busts have typically been preceded by rising investment, expanding credit, and deteriorating current account balances.
- Large deviations in these variables from local trends have some value as indicators of future asset price busts, often increasing the probability of a bust by more than twice the unconditional probability when alarms are raised.
- Nonetheless, asset price busts remain difficult to predict: even the best indicator failed to raise an alarm one to three years ahead of roughly one-half of all busts since 1985.
- Monetary policy stance (real rates, Taylor residuals), and conventional measures of inflation and output, are poor leading indicators of asset price busts.
- If avoiding asset price busts is a policy concern, a narrow focus on inflation and output is insufficient; monitoring and tools addressing credit expansion, overinvestment, and external imbalances are warranted, and macroprudential instruments plus coordination with monetary policy deserve greater emphasis.

*Source: content unit _wp09252 - 3. There are, however, two notable differences between the behavior of macroeconomic*

### chapter implies it is important to be realistic about what could be achieved with a broader

### _wp09252 - chapter implies it is important to be realistic about what could be achieved with a broader

### Main implication
- "chapter implies it is important to be realistic about what could be achieved with a broader approach to monetary policy:"
- "even the best leading indicators of asset price busts are imperfect, such that, in the process of trying to reduce the probability of a dangerous bust, central banks may raise costly false alarms."

### Policy considerations
- Recognize limits of leading indicators of asset price busts.
- Weigh the trade-off between reducing the probability of a dangerous bust and the risk of raising costly false alarms.

### Key item
- Page: 17

*Source: _wp09252 - chapter implies it is important to be realistic about what could be achieved with a broader*

### REFERENCES

### _wp09252 - REFERENCES

### References (selected)
- Adalid, Ramón, and Carsten Detken, 2007, “Liquidity Shocks and Asset Price Boom/Bust Cycles,” ECB Working Paper No. 732.
- Ahrend, Rudiger, Boris Cournède, and Robert Price, 2008, “Monetary Policy, Market Excesses and Financial Turmoil,” OECD Economics Department Working Paper No. 597.
- Alessi, Lucia, and Carsten Detken, 2009, “`Real Time’ Early Warning Indicators for Costly Asset Price Boom/Bust Cycles: A Role for Global Liquidity,” ECB Working Paper No. 1039.
- Berg, Andrew, Eduardo Borensztein, Gian Maria Milesi-Ferretti, and Catherine Pattillo, 2000, “Anticipating Balance of Payments Crises: The Role of Early Warning Systems,” IMF Occasional Paper No. 186.
- Borio, Claudio, and Philip Lowe, 2002, “Assessing the Risk of Banking Crises,” BIS Quarterly Review (December), pp. 43–54.
- Borio, Claudio, and Ilhyock Shim, 2007, “What Can (Macro-)Prudential Policy Do to Support Monetary Policy?” BIS Working Paper No. 242.
- Cecchetti, Stephen G., Hans Genberg, John Lipsky and Sushil Wadhwani, 2000, Asset Prices and Central Bank Policy, Geneva Reports on the World Economy 2.
- Claessens, Stijn, Ayhan Kose, and Marco Terrones, 2008, “What Happens During Recessions, Crunches, and Busts?” IMF Working Paper 08/274.
- Detken, Carsten, and Frank Smets, 2004, “Asset Price Booms and Monetary Policy,” ECB Working Paper No. 364.
- Frankel, Jeffrey A., and Andrew K. Rose, 1996, “Currency Crashes in Emerging Markets: An Empirical Treatment,” Journal of International Economics, Vol. 41 (November), pp. 351–66.
- Galí, Jordi, and Luca Gambetti, 2009, “On the Sources of the Great Moderation,” American Economic Journal: Macroeconomics, Vol. 1, No. 1, pp. 26–57.
- Kaminsky, Graciela, Saul Lizondo, and Carmen Reinhart, 1998, “Leading Indicators of Currency Crisis,” IMF Staff Papers, Vol. 45 (March), pp. 1–48.
- Kannan, Prakash, Pau Rabanal and Alasdair Scott, 2009, “Monetary and Macroprudential Rules in a Model with House Price Booms,” forthcoming IMF Working Paper.
- Mendoza, Enrique G., and Marco E. Terrones, 2008, “An Anatomy of Credit Booms: Evidence from Macro Aggregates and Micro Data,” IMF Working Paper 08/226.
- Taylor, John B., 1993, “Discretion versus Policy Rules in Practice,” Carnegie-Rochester Conference Series on Public Policy, Vol. 39, No. 1, pp. 195–214.
- White, William R., 2006, “Procyclicality in the Financial System: Do We Need a New Macrofinancial Stabilisation Framework?” BIS Working Paper No. 193.

### Data appendix — Data sources and variable mapping
- Nominal house prices: Bank of International Settlements (BIS), Haver Analytics, OECD.
- Real house prices: Organization for Economic Cooperation and Development.
- Real stock prices: Bloomberg Financial Markets, International Financial Statistics (IFS) database.
- Real credit: International Financial Statistics.
- Nominal credit: International Financial Statistics.
- Real private consumption: Organization for Economic Cooperation and Development.
- Real residential investment: Organization for Economic Cooperation and Development.
- Output: Organization for Economic Cooperation and Development.
- Current account: Organization for Economic Cooperation and Development.
- Consumer Price Index (CPI): Haver Analytics (Core PCE), OECD, and IFS.
- Quick ratio: Organization for Economic Cooperation and Development.
- Policy rates: Bloomberg Financial Markets, National Authorities, and Thomson Datastream.
- Real long-term rates: International Financial Statistics.

Data appendix notes (exact text):
- "Haver Analytics used to obtain nominal house price data for Austria, Greece and Portugal. BIS used to obtain nominal house price data for Belgium."
- "Real house price data on Austria, Greece, Portugal and Belgium obtained by taking the nominal series and deflating by CPI data taken from the OECD."

### Key tables and exact statistics
- Table 1. House Price and Stock Price Busts from 1970 to 2008 (mean values)
  - Total number of busts: House prices 47; Stock prices 98.
  - Subsample counts:
    - Before 1985: House prices 22; Stock prices 41.
    - 1985–2008: House prices 25; Stock prices 57.
  - Number of busts per country: House prices 2.76; Stock prices 4.67 (Full sample). Before 1985: 1.29 and 1.95. 1985–2008: 1.47 and 2.71.
  - Cumulative decline in prices (percent)1: House prices –17.71; Stock prices –37.38 (Full sample). Before 1985: –19.43 and –35.27. 1985–2008: –15.58 and –38.90.
  - Duration (quarters): House prices 10.02; Stock prices 6.98 (Full sample). Before 1985: 11.22 and 7.92. 1985–2008: 9.74 and 6.29.
  - Cumulative decline in output (percent relative to trend)2: House prices –4.27; Stock prices –1.31 (Full sample). Before 1985: –5.41 and –1.33. 1985–2008: –3.27 and –1.29.
  - Notes:
    - 1 Cumulative price decline is measured over the entire duration of the bust period.
    - 2 Cumulative decline in output is measured as the accumulated deviation from a one-sided Hodrick-Prescott filter with a smoothness parameter of 1600 for the first four quarters of a bust.

- Table 2. Classification of Observations Based on Variable Thresholds (alarm matrix)
  - Alarm raised / Asset Price Bust 1–3 Years Later: A
  - Alarm raised / No Asset Price Bust 1–3 Years Later: B
  - No alarm / Asset Price Bust 1–3 Years Later: C
  - No alarm / No Asset Price Bust 1–3 Years Later: D

- Table 3. Percentiles Used as Thresholds for Alarms (percentile of distribution where noise-to-signal ratio minimized; grid search limited to 70th, 75th, 80th and 90th percentiles)
  - Credit/GDP: House Price Bust 90; Stock Price Bust 90.
  - Current account/GDP: House Price Bust 90; Stock Price Bust 90.
  - Residential investment/GDP: House Price Bust 90; Stock Price Bust 90.
  - House price growth: House Price Bust 90; Stock Price Bust 70.
  - Stock price growth: House Price Bust 70; Stock Price Bust 70.
  - Growth: House Price Bust 75; Stock Price Bust 80.
  - Inflation: House Price Bust 90; Stock Price Bust 90.

- Table 4. Marginal Probabilities Based on Probit Regressions (marginal probabilities at mean values; z-statistics in parentheses; ***, **, * at 1, 5, 10 percent)
  - Full Sample (N = 1699; Pseudo R2 = 0.14)
    - Credit/GDP (house price bust): 0.241** (2.180)
    - Credit/GDP (stock price bust): 0.546*** (4.070)
    - Current account balance (house price bust): –3.910*** (–7.560)
    - Current account balance (stock price bust): 0.691 (1.200)
    - Residential investment/GDP (house price bust): 1.956 (1.520)
    - Residential investment/GDP (stock price bust): 6.392*** (5.280)
    - House price growth (house price bust): 0.798*** (5.240)
    - House price growth (stock price bust): 0.577*** (3.110)
    - Stock price growth (house price bust): 0.249*** (6.060)
    - Stock price growth (stock price bust): 0.337*** (5.250)
    - Output growth (house price bust): –0.413 (–0.810)
    - Output growth (stock price bust): 1.686** (2.540)
    - Inflation (house price bust): 2.511*** (7.180)
    - Inflation (stock price bust): 4.373*** (7.030)
  - Before 1985 (N = 435; Pseudo R2 = 0.16)
    - Credit/GDP (house price bust): –0.864* (–1.740)
    - Credit/GDP (stock price bust): 0.052 (0.130)
    - Current account balance (house price bust): –3.472*** (–3.640)
    - Current account balance (stock price bust): –2.851*** (–2.990)
    - Residential investment/GDP (house price bust): 4.621 (1.370)
    - Residential investment/GDP (stock price bust): 4.801 (1.550)
    - House price growth (house price bust): 2.147*** (5.140)
    - House price growth (stock price bust): 1.046*** (3.170)
    - Stock price growth (house price bust): 0.577*** (4.890)
    - Stock price growth (stock price bust): 0.323*** (2.660)
    - Output growth (house price bust): –0.916 (–0.940)
    - Output growth (stock price bust): 0.280 (0.290)
    - Inflation (house price bust): 3.786*** (5.460)
    - Inflation (stock price bust): 4.721*** (5.470)
  - 1985–2008 (N = 1264; Pseudo R2 = 0.15 for house price bust; 0.10 for stock price bust)
    - Credit/GDP (house price bust): 0.443*** (4.280)
    - Credit/GDP (stock price bust): 0.638*** (4.210)
    - Current account balance (house price bust): –3.191*** (–5.440)
    - Current account balance (stock price bust): 1.768** (2.510)
    - Residential investment/GDP (house price bust): 2.456* (1.930)
    - Residential investment/GDP (stock price bust): 7.327*** (5.290)
    - House price growth (house price bust): 0.455*** (2.910)
    - House price growth (stock price bust): 0.318 (1.340)
    - Stock price growth (house price bust): 0.111*** (2.680)
    - Stock price growth (stock price bust): 0.349*** (4.660)
    - Output growth (house price bust): –0.160 (–0.300)
    - Output growth (stock price bust): 2.428 (2.620)
    - Inflation (house price bust): 0.681 (1.640)
    - Inflation (stock price bust): 3.732*** (4.130)

  - Note (exact text): "Dependent variable takes a value of 1 if there is a bust between 12 and 4 quarters ahead and zero otherwise. Estimation is carried out using robust standard errors. Z-statistics are reported in parentheses. ***, **, and * refer to significance at the 1, 5, and 10 percent level, respectively. Marginal probabilities computed at the mean values of other variables are reported. Variables are measured as deviations from an eight-year moving average."

### Figures — captions and reported metrics (exact)
- Figure 2. Selected Macro Variables before and during House Price Busts (Median percent deviation from trailing eight-year moving average; asterisk indicates statistically significant difference of post-1985 deviation from zero; t = 1 denotes first quarter of bust). Source: IMF staff calculations.
- Figure 3. Selected Macroeconomic Variables before and during Stock Price Busts (Median percent deviation from trailing eight-year moving average; asterisk indicates statistically significant difference of post-1985 deviation from zero; t = 1 denotes first quarter of bust). Source: IMF staff calculations.
- Figure 4. Selected Macroeconomic Variables before and during High-Cost and Other House Price Busts, 1985–2008 (Median percent deviation from trailing eight-year moving average; asterisk indicates statistically significant difference between medians; t = 1 denotes first quarter of bust).
- Figure 5. The Probability of an Asset Price Bust (Percent of times a bust occurs 1–3 years after an alarm is raised relative to the unconditional probability of a bust). 1985–2008. Source: IMF staff calculations.
- Figure 6. The Failure of the Indicators to Predict an Asset Price Bust (Per cent of quarters the variables failed to raise an alarm 1–3 years before a bust). Source: IMF staff calculations.
- Figure 8. Warning Signs for Recent House Price Busts (Percent of countries with recent house price busts that raised alarms). Source: IMF staff calculations.
- Figure 9. Macroeconomic Patterns underlying Recent House Price Booms (various scatterplots; reported R  = 0.54 2 for Real House Prices and Residential Investment; R  = 0.24 2 for Real House Prices and Current Account; R  = 0.37 2 for Real House Prices and Growth Rate of Nominal Credit relative to GDP).
- Figure 10. Recent House Price Booms and Household Balance Sheets (reported R  = 0.60 2 for Growth Rate of Nominal Credit relative to GDP and Household Quick Ratio).
- Figure 11. Monetary Conditions Leading up to the Current Crisis (Real Policy Rates: United States, Euro area, Average of other advanced economies; Consensus one-year ahead inflation forecasts used to define real policy rates).
- Figure 12. Inflation and Output for Advanced Economies in Recent Years (Percent; Japan omitted).
- Figure 13. House Prices and Monetary Conditions (scatterplots; reported R  = 0.17 2, R  = 0.03 2, R  = 0.05 2, R  = 0.19 2 in various panels).
- Figure 14. Stock Prices and Monetary Conditions (reported R = 0.00 2, R = 0.03 2, R = 0.007 2 in various panels).
- Figure 15. Growth Rate of Nominal Credit Relative to GDP and Real Policy Rates (reported R = 0.02 2).
- Figure 16. Selected Macro Variables before and during House Price Busts (Median percent deviation from a rolling Hodrick-Prescott filter; asterisk indicates statistically significant difference of post-1985 deviation from zero; t = 1 denotes first quarter of bust). Source: IMF staff calculations.
- Figure 17. Selected Macroeconomic Variables before and during Stock Price Busts (Median percent deviation from a rolling Hodrick-Prescott filter; asterisk indicates statistically significant difference of post-1985 deviation from zero; t = 1 denotes first quarter of bust). Source: IMF staff calculations.

*Content unit: _wp09252 - REFERENCES (PDF chapter/section).*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2009/_wp09252.pdf_