## 15.  Back to Earth: Predicted and Actual House Price Changes

## Source details

**Canonical URL:** [15.  Back to Earth: Predicted and Actual House Price Changes](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12217.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12217.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12217.pdf.json)

---

### Introduction and headline findings
- Global housing boom 2000–2006:
  - "Never before had real house prices risen so fast, for so long, and in so many countries at the same time."
  - With the exception of Germany and Japan, real house prices in all OECD countries increased from 2000 to 2006.
  - "House prices rose 50 percent in real terms in the median advanced economy."
  - "House prices rose... almost 30 percent in the median developing country."
- Crisis timing and data cutoff:
  - Global housing boom turned into a housing bust during 2007 in almost all countries except the United States, where a housing correction has been under way since 2006.
  - Data cut-off for analysis: end-2009.
- Main findings preview:
  - Long-run price dynamics are mostly driven by local fundamentals such as income and population growth.
  - Globally connected factors such as interest rates have a weaker effect.
  - Credit market conditions can cause short-run deviations and, when corrections start, financial stability and overall economic activity suffer.
  - Structural characteristics of housing and mortgage markets influence severity of impacts.
- Magnitude and duration assessments:
  - If past is prologue, ongoing house price corrections could "average about 23 percent and be spread out over a period of 4-4.5 years from peak to trough."
  - Alternative assessment based on past cycle characteristics suggests falls "is likely to amount to half of the steep climb that took place since the mid-1990s, averaging slightly more than 60 percent around the globe."
- Data and research recommendation:
  - Major obstacle: lack of consistent, reliable housing and real estate financing data.
  - Recommendation: concentrate efforts on increasing data availability to enable closer study of macro-financial linkages.

### Data, measurement, and frequency
- Sources and coverage:
  - International sample combines national housing indicators and demographic factors with IMF International Financial Statistics and OECD macroeconomic variables.
  - House price indices used from OECD, BIS, and Global Property Guide (seasonally adjusted).
  - U.S. sub-national data: FHFA, BEA, BLS.
  - Formal econometric analysis constrained to 22 advanced countries because of data availability.
- Measurement challenges:
  - Heterogeneity of housing, infrequency/opacities of transactions, repairs/improvements changing property value over time.
  - Multiple series per country due to differences in coverage, methodology, and computation techniques (example: OFHEO/FHFA vs S&P Case-Shiller differences).
  - Aggregation at national level can mask substantial regional variation (examples: Northern Ireland vs London; China; Hong Kong SAR; India; Korea; Netherlands; United States).
- Frequency and seasonal adjustment:
  - Quarterly frequency used (1970–2010 where available); series seasonally adjusted using X-12 in E-Views.

### Descriptive housing-cycle characteristics and dating
- Nature of housing cycles:
  - Prone to boom-bust cycles due to rigid supply, infrequent trades, opaqueness, short-term construction finance vs long-term occupancy finance.
  - Prices and quantities tend to move in the same direction.
- Comovement evidence (examples):
  - U.S. new home sales and new home prices correlation: 0.61.
  - Hong Kong SAR prices and number of sales contracts correlation: 0.4.
  - Netherlands regional price changes and turnover correlation coefficient: 0.27.
- Dating methodology:
  - Seasonally-adjusted national house price series at quarterly frequency for 55 countries used to date cycles.
  - Long series: Harding (2003) dating with minimum duration six quarters.
  - Shorter series: downturn/upturn defined as four or more consecutive quarters of decline/rise.
- Typical cycle statistics (median sample values):
  - Upturns: duration 16 quarters; trough-to-peak increase 37 percent.
  - Downturns: duration 11 quarters; amplitude -17 percent.
- Observed extremes:
  - Upturn duration up to 99 quarters (Belgium).
  - Price increases more than quadrupling (Ukraine).
  - Downturns up to 76 quarters (Japan).
  - Largest price decline: 76 percent (Latvia).
- Convergence and synchrony:
  - Timing of phases has been converging; greater simultaneity of upturns and downturns across countries.
- Prices vs quantity robustness:
  - For 17 of 21 countries, correlation between house prices and residential investment exceeds 0.6.
  - Exceptions: correlation negative in Germany and Korea; positive but <0.5 for Sweden and Switzerland.
  - Conclusion: dating cycles by price series alone is unlikely to create major robustness issues.

### Nonparametric gap estimates and history-based inferences
- Practical approach:
  - Use characteristics of past housing cycles and valuation ratios to infer potential outcome of ongoing correction.
  - Nonparametric, history-dependent analysis examining cycle characteristics and valuation ratios, supplemented by formal regression where data permit.
- Past upturn characteristics:
  - The last upturn (mid-1990s onward) was "extraordinarily long" and of "extraordinarily large amplitude."
  - Last upturn lasted almost twice as long as average past upturn and prices increased more than double past increases.
- Duration-based projections:
  - Simple duration comparison implied that as of 2010Q1, the current downturn would continue another 3 quarters and subside by end-2010.
  - Longer upturns tend to be followed by longer downturns; downturn might last almost twice as long as average downturns and "house prices might continue their decline as far as 3 years down the road."
- Valuation yardstick summary assessments:
  - One assessment: ongoing corrections "could average about 23 percent and be spread out over a period of 4-4.5 years from peak to trough."
  - Alternative history-based assessment: corrections "is likely to amount to half of the steep climb that took place since the mid-1990s, averaging slightly more than 60 percent around the globe."
- Heterogeneity emphasized:
  - Large cross-country heterogeneity; outcomes depend on housing finance characteristics and country-specific cycle timing.

### Price-to-Income Ratio (PIR), Price-to-Rent Ratio (PRR), and user cost analysis
- Price-to-Income Ratio (PIR):
  - PIR interprets overvaluation when prices exceed historically observed ratios relative to income.
  - Figure 10: deviation of house prices to per capita income from historical average for 55 countries.
  - Out-of-sample lower-panel: deviations as of 2006Q4 versus realized house price changes since that date show a strong negative relationship.
  - Countries with fast appreciation and harsh downturns: Denmark, Estonia, Iceland, Ireland, Latvia, Lithuania, Spain, Ukraine, and the U.S.
  - Potential misalignment remains in Denmark and Spain and a handful of countries with milder downturns.
- Housing Affordability Index (HAI) — U.S. sub-national:
  - Definition: whether a typical (median-income) family could qualify for a traditional mortgage on a median-priced home.
  - Assumptions: down payment = 20 percent of home price; qualifying ratio = 25 percent of median family monthly income.
  - Interpretation: HAI = 100 means exact income to qualify; HAI > 100 means more than enough; HAI < 100 means cannot afford.
  - States with least affordable homes: Hawaii, District of Columbia, California, Massachusetts, New Jersey, Arizona, New York, Oregon, and Washington.
  - Affordability has deteriorated across the board except Northeast Corridor when benchmarked to 1990; however, these states also show deterioration relative to 2000.
  - Overall, affordability ratios remain above their historical levels (1960 to 2004) in 3 out of 5 states.
  - Policy implication: substantial within-country heterogeneity suggests blanket measures may produce unintended consequences.
- Price-to-Rent Ratio (PRR) and user cost (U.S.):
  - PRR declined from 23.17 in 2008Q2 to 19.97 in 2009Q4; latest values remain slightly above historical norms.
  - User cost framework (Poterba, 1992): PRR equals inverse of user cost expression (parameters τm, i, τp, β, d, m, π, P in text).
  - Parameter choices: β + m + d assumed constant at 4 percent; π proxied by an eight-year moving average of consumer price inflation.
  - Results:
    - PRR at 24.84 lies above the “equilibrium” level around 15.2 as of 2008Q2 (Figure 14).
    - Historical average excess of PRR over inverse user cost: 5.15.
    - Holding rent constant, reaching historical average discrepancy implies almost 26 percent drop in nominal house prices.
    - Assuming a 4 percent increase in rental rates, implied drop is around 21 percent.
  - Sensitivity: estimates particularly sensitive to expected capital gains (π); alternative expectation formation (e.g., 3-year basis with downward updating) brings PRR and inverse user cost closer.

### Parametric regression analysis of real house price changes
- Model specification:
  - Dependent variable: change in real house prices (quarterly).
  - Regressors include: change in real income per capita, change in working-age population, change in stock prices, change in bank credit to private sector, short-term and long-term interest rates, and lagged affordability (log price-to-income per capita).
- Estimation approach:
  - Ordinary least squares (OLS) used due to limited series length and weak Johansen cointegration evidence.
  - Country-by-country regressions for 22 advanced economies; pooled regressions also reported.
- Key country-level empirical findings (summary):
  - Affordability negatively related to change in prices in 82 percent of countries.
  - Income per capita change positive in all countries except the U.S. (U.S. coefficient negative but not statistically significant).
  - Population growth positive in 86 percent of cases; statistically significant about one-third of the time.
  - Bank credit to private sector positive in 91 percent of countries; 64 percent deliver correctly-signed statistically significant coefficients.
  - Interest rate coefficients inconsistent: correct (negative) sign in 50 to 64 percent of countries; correct and significant in 27 to 41 percent.
  - Stock price coefficients positive in 68 percent of cases; statistically significant in 36 percent.
  - Average R-squared across countries: 0.33; poor fit for Australia, Canada, and Italy; good fit for Norway and Spain (and to a lesser extent Finland, Japan, the U.K.).
- Pooled regression highlights (selected exact coefficients from Table 5):
  - Affordability, lagged: Column 1 -0.0130*** (0.0025); Column 2 -0.0335*** (0.0081); Column 3 -0.0269*** (0.0041); Column 4 -0.0268** (0.0042).
  - Income per capita, change: Column 1 0.415*** (0.0474); Column 2 0.408*** (0.0472); Column 3 0.429** (0.0684); Column 4 0.4397** (0.0948).
  - Working-age population, change: Column 1 0.694*** (0.1150); Column 4 1.6677*** (0.3066).
  - Stock prices, change: Column 1 0.0136*** (0.0021).
  - Credit, change: Column 1 0.0369*** (0.0057).
  - Short-term interest rate: Column 1 -0.0008** (0.0003); Column 4 -0.0016** (0.0006).
  - Column 1 observations: 27,182; R-squared 0.16; Number of countries 222.
  - Column 4 observations: 15,21; R-squared 0.27; Number of countries 22.
- Translating changes to price gaps:
  - Procedure: set equilibrium house price index = 100 at base dates (1997Q4, 1998Q4, 1999Q4, 2000Q4, 2001Q4), compute predicted changes forward, compare actual index to predicted to obtain estimated price gap.
  - Heat map (Table 6) reports average estimated gaps across selected dates.
  - Findings: overheating at end-2006 and 2007; misalignment unwinding in 2008 and 2009 but large price gaps remain in a handful of countries.
- Predictive performance:
  - Predicted correction as of 2007Q4 compared to actual change since that date shows strong positive correlation (Figure 15): Correlation coefficient 0.53; R² = 0.28.
  - Average predicted decline: 9 percent; realized decline: 5 percent — discrepancy possibly due to policy-induced rebound or an ongoing correction anticipating additional 4 percent decline on average.
  - Notable under- and over-predictions: under-predicted severity in Ireland, Denmark, and the U.S.; over-predicted downturn for Australia and Canada.

### Macroeconomic consequences — VAR analysis of housing demand shocks
- VAR setup and identification:
  - Baseline variables: real GDP, real private consumption, real private residential investment, CPI, nominal short-term interest rate, real house prices.
  - Identification: Cholesky decomposition ordering with house prices last; housing shocks labeled "housing demand shocks."
  - Sample period: 1986Q1 to 2010Q1.
  - Data transformation: log levels for variables except CPI (log first differences) and short-term interest rate (levels).
- Shock normalizations and average effects:
  - Normalizations: a) initial house price response as decline of 10 percent; b) decline equal to two standard deviations of quarterly house price change.
  - Average impact of a 10 percent house price decline on GDP: -1.87 percent.
  - Literature comparisons: Cardarelli et al. (2009) ~ -2 percent; Goodhart and Hofmann (2008) ~ -1 percent.
- Cross-country heterogeneity (examples):
  - Small-impact countries (GDP decline around or less than ½ percentage points): Austria, Italy, Netherlands, Norway, Switzerland.
  - Large-impact countries (GDP decline close to or more than 1 percentage point): Denmark, Greece, Ireland, Japan, New Zealand, Spain, United States.
  - Effects on consumption and residential investment: highly heterogeneous across countries.
- VAR maximum impacts (examples from Table 7, decline of 10 percent):
  - United States: GDP -1.89; Consumption -2.78; Residential investment -20.20.
  - Japan: GDP -5.49; Consumption -3.03; Residential investment -19.74.
  - United Kingdom: GDP -1.03; Consumption -1.40; Residential investment -12.24.
  - Germany: GDP -4.61; Consumption -5.74; Residential investment -38.45.
  - New Zealand: GDP -4.18; Consumption -5.70; Residential investment -28.76.
  - Notes: alternative normalizations produce smaller reported impacts (alternative columns in Table 7).

### Mortgage-market characteristics, banking exposure, and vulnerabilities
- Mortgage-market features related to vulnerability:
  - Prepayment/refinancing availability:
    - Penalty-fee prepayment and refinancing available: average GDP impact -2.77 (countries included: Australia, Denmark, Japan, United Kingdom, United States).
    - Penalty-fee prepayment and refinancing not available: average GDP impact -1.71 (Austria, Belgium, Canada, Finland, France, Germany, Greece, Italy, Netherlands, Norway, Spain, Sweden).
  - Equity withdrawal options:
    - When homeowners can withdraw lines of credit on home equity: average estimated GDP impact -2.27.
    - When not available: average estimated impact -1.61.
  - Loan-to-value (LTV) at origination:
    - Typical LTV ≤ 60 percent: average impact -0.06 (Austria, Italy).
    - Typical LTV between 60 and 79 percent: average impact -1.67 (Canada, Finland, France, Germany, Norway, Spain, United Kingdom).
    - Typical LTV ≥ 80 percent: average impact -2.21 (Australia, Belgium, Denmark, Ireland, Japan, Netherlands, Sweden, United States).
  - Mortgage-debt-to-GDP:
    - When outliers (Germany, Japan, Greece, Netherlands) excluded, vulnerability appears positively related to ratio of residential mortgage debt to GDP.
  - Weaker correlations:
    - Correlation with other mortgage-market characteristics and structural dimensions (labor intensity of construction, share of construction in value added) seems weaker.
- Banking sector exposure indicator:
  - Value of mortgage loans held by banks expressed as multiple of overall market capitalization:
    - About 4 in Denmark and Germany.
    - Less than 3 in Spain.
    - About 1.5 in Canada, Japan, and the United Kingdom.
    - Less than 1 in the United States.
  - Cross-country declines in residential investment during housing cycles have been higher in countries with greater banking sector exposure to mortgage lending, though effect weaker than with mortgage-debt-to-GDP ratio.
  - Current concern: bank balance sheets under renewed stress and low bank equity prices increase potential for adverse real economy impact from banking exposure to mortgage lending.

### Other factors affecting downturn amplitude and transmission
- Variable-rate mortgages:
  - Finland, Ireland, and Spain have mostly variable-rate mortgages.
  - Higher debt burden (interest payments relative to household disposable income) responds more to interest rate hikes when mortgage rates adjustable; historically associated with bigger declines in residential investment during busts.
- Legal provisions and recourse:
  - In many countries (France, Germany, Ireland, the Netherlands, Spain, the United Kingdom), debtors personally liable for full mortgage debt, reducing foreclosure incentives.
  - In the United States, foreclosure rules vary by state:
    - In six states lenders have recourse only to mortgaged property.
    - In other states debtors are personally liable, but practical recovery of deficiency judgments varies and lenders may often choose not to seek them.

### Conclusions and policy implications
- Historical correction magnitude and duration:
  - If past is prologue, corrections could average about 23 percent and be spread out over 4-4.5 years from peak to trough.
  - This assessment aligns with valuation-ratio inspections and econometric models.
- Cross-country differences:
  - Differences in macroeconomic impact depend on housing finance system characteristics, particularly ease of recent household access to mortgage credit.
  - Ease of access to mortgage credit correlated with larger investment declines during corrections and stronger dampening of consumption.
  - Pre-crisis financial system exposure to housing markets may affect extent of macroeconomic losses and shape/timing of recovery.
- Data and research recommendation:
  - Major obstacle: lack of data on housing and real estate financing markets.
  - Recommendation: increase availability of consistent, reliable data to enable closer study of macro-financial linkages in future research.

*Italicized source attribution: IMF Working Paper chapter "15.  Back to Earth: Predicted and Actual House Price Changes" (data cut-off end-2009) from the provided PDF content.*

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

### _wp12217 - References

### Tables
- 1.  House Price Cycles by Country.................................................... .................... 26
- 2.  Characteristics of House Price Cycles.................................................... .......... 29
- 3.  Price-to-Rent Ratio: Adjustment to Benchmark.................................................29
- 4.  Modeling House Price Changes: Country-by-Country Regressions...................... .... 30
- 5.  Modeling House Price Changes: Pooled Regressions............................... .............. 32
- 6.  Misalignment in House Prices: A Heat Map....................................................  33
- 7.  Macroeconomic Impact of House Price Corrections............... ..................................... 34
- 8.  Macroeconomic Impact of House Price Corrections and Country Characteristics...... ... 35

### Figures
- 1.  An Unprecedented Global Housing Boom.................................................... ... 36
- 2.  A Global Housing Boom Turning into a Global Bust.................. ............................... 37
- 3.  Measuring House Prices: Differences in U.S. Indices..................... ........................... 38
- 4.  National versus Sub-national Prices: Selected Indices in the U.K. .................. .......... 39
- 5.  Housing Markets: Price versus Quantity.................................................... ...... 40
- 6.  Regional Housing Markets: Prices versus Activity..................................... .......... 41
- 7.  Global Housing Cycles: Concurrence of Local Markets...................... ...................... 42
- 8.  Global Booms and Busts.................................................... ............................... 43
- 9.  House Prices and Residential Investment.................................................... ..... 44
- 10.  Price-to-Income Ratio by Country.................................................... .............. 45
- 11.  Housing Affordability in the U.S. .................................................... .............. 46
- 12.  Housing Affordability in the U.S.: Now and Then........................... ...................... 47
- 13.  Price-to-Rent Ratio Adjustment in the U.S..................................................... 48
- 14.  User Cost of Housing in the U.S. .................................................... ............... 49

*Source: _wp12217 - References (PDF chapter/section).*

### 15.  Back to Earth: Predicted and Actual House Price Changes......................... ............ 50

### 15.  Back to Earth: Predicted and Actual House Price Changes

### Introduction
- Global housing boom 2000–2006:
  - "Never before had real house prices risen so fast, for so long, and in so many countries at the same time."
  - With the exception of Germany and Japan, real house prices in all OECD countries increased from 2000 to 2006.
  - "House prices rose 50 percent in real terms in the median advanced economy" during this period.
  - "House prices rose... almost 30 percent in the median developing country."
- Crisis and timing:
  - The global housing boom turned into a housing bust during 2007 in almost all countries except the United States, where a housing correction has been under way since 2006.
  - Data cut-off for analysis: end-2009.
- Main findings preview:
  - Long-run price dynamics are mostly driven by local fundamentals such as income and population growth.
  - Globally connected factors such as interest rates have a weaker effect.
  - Credit market conditions can cause short-run deviations and, when corrections start, financial stability and overall economic activity suffer.
  - Structural characteristics of housing and mortgage markets influence severity of impacts.
- Forecasts and comparisons:
  - If past is prologue, the ongoing house price corrections could "average about 23 percent and be spread out over a period of 4-4.5 years from peak to trough."
  - Alternative assessments (based on past cycle characteristics) suggest the magnitude of falls "is likely to amount to half of the steep climb that took place since the mid-1990s, averaging slightly more than 60 percent around the globe."
- Data and research recommendations:
  - Major obstacle: lack of consistent, reliable housing and real estate financing data.
  - Recommendation: concentrate efforts on increasing data availability to enable closer study of macro-financial linkages.

### Data
- Sources and coverage:
  - International sample combines national housing indicators and demographic factors with IMF International Financial Statistics and OECD macroeconomic variables.
  - House price indices used from OECD, BIS, and Global Property Guide (seasonally adjusted).
  - Sub-national U.S. data: Federal Housing Finance Agency (FHFA), Bureau of Economic Analysis (BEA), Bureau of Labor Statistics (BLS).
  - Formal econometric analysis constrained to 22 advanced countries because of data availability.
- Measurement challenges:
  - Heterogeneity of housing, infrequency/opacities of transactions, repairs/improvements that change property value over time.
  - Multiple series per country arise from differences in coverage (geographical and property type), methodologies, and computation techniques.
  - Example (United States): OFHEO/FHFA and S&P Case-Shiller indices both use repeat-sales methodology for single-family homes but differ in geographical coverage, treatment of refinance appraisals, loan-type coverage (conforming vs jumbo/non-prime), and weighting schemes.
  - Aggregation at national level can mask significant regional variation (examples: Northern Ireland vs London; China; Hong Kong SAR; India; Korea; Netherlands; United States).
- Frequency and seasonal adjustment:
  - Quarterly frequency used (1970–2010 where available); series seasonally adjusted using X-12 in E-Views.
- Empirical analysis outline:
  - Three parts: (1) demonstrate cyclical movement; (2) infer extent of ongoing correction using past cycles and nonparametric gap estimation; (3) estimate regression model for long-term relationship between house prices and macro determinants; then analyze consequences on consumption, investment, financial distress, and GDP.

### Empirical analysis — Housing Cycles: descriptive characteristics and dating
- Nature of housing cycles:
  - Housing markets prone to boom-bust cycles due to rigid supply, infrequent trades, opaqueness, short-term construction finance vs long-term occupancy finance.
  - Prices and quantities tend to move in the same direction.
- Comovement evidence:
  - U.S. new home sales and new home prices correlation: 0.61 (based on data from the early 1960s to present in the sample).
  - Hong Kong SAR prices and number of sales contracts correlation: 0.4.
  - Netherlands: positive relation between regional house price changes and turnover; correlation coefficient: 0.27.
- Dating methodology:
  - Seasonally-adjusted national house price series at quarterly frequency for 55 countries used to date cycles.
  - For long series (start in 1986 or earlier) used Harding (2003) dating with a minimum duration of six quarters.
  - For shorter series, downturn/upturn defined as four or more consecutive quarters of decline/rise.
- Typical cycle statistics (median sample values):
  - Upturns: duration 16 quarters; trough-to-peak increase 37 percent.
  - Downturns: duration 11 quarters; amplitude -17 percent.
- Extremes observed:
  - Upturn duration as long as 99 quarters (Belgium).
  - Price increases more than quadrupling (Ukraine).
  - Downturns up to 76 quarters (Japan).
  - Largest price decline: 76 percent (Latvia).
- Convergence and global synchrony:
  - Timing of phases has been converging—upturns and downturns across countries overlap more now than previously.
  - Proportion of countries in upturn at any point shows greater simultaneity over time.
- Prices vs quantity robustness:
  - Private residential investment (OECD) broadly confirms price-quantity comovement.
  - For 17 of 21 countries with available data, the correlation between house prices and residential investment exceeds 0.6.
  - Exceptions: correlation negative in Germany and Korea; positive but <0.5 for Sweden and Switzerland.
  - Conclusion: dating cycles by price series alone is not likely to create major robustness issues.

### Modeling housing cycles and nonparametric estimates
- Modeling approaches in literature:
  - Empirical and theoretical models typically link house prices to supply rigidity, demand fundamentals, rent expectations, construction lags, capitalization of future rents, and interactive demand-supply systems.
- Practical approach in this paper:
  - Use characteristics of past housing cycles to infer potential outcome of ongoing correction.
  - Employ nonparametric, history-dependent analysis examining cycle characteristics and valuation ratios.
  - Adopt formal regression framework for long-term relationships where data permit (at cost of reducing country coverage).
- Past cycles versus present:
  - The last upturn (mid-1990s onward) was "extraordinarily long" and of "extraordinarily large amplitude" (Table 2).
  - On average, the last upturn lasted almost twice as long as the average past upturn and prices increased more than double past increases.
  - Simple duration comparison implied that as of 2010Q1, the current downturn would continue another 3 quarters and subside by the end of the year.
  - However, longer upturns tend to be followed by longer downturns; thus one could argue the downturn might last almost twice as long as average downturns and "house prices might continue their decline as far as 3 years down the road."
  - Cross-country variation substantial: e.g., markets such as Ireland or Ukraine (earlier sharp downturns) may be closer to bottom than Poland.
  - Stimulus plans implemented end-2008 and 2009, often including housing measures, may have temporarily ended downturns; preliminary assessments suggest these effects are likely to be short-lived.
- Valuation yardsticks and gap estimates:
  - Nonparametric and valuation-ratio assessments are used to evaluate misalignment of house prices vis-à-vis fundamentals.
  - One summary assessment: ongoing corrections "could average about 23 percent and be spread out over a period of 4-4.5 years from peak to trough."
  - Alternative history-based assessment suggests corrections "is likely to amount to half of the steep climb that took place since the mid-1990s, averaging slightly more than 60 percent around the globe."
  - The paper emphasizes large cross-country heterogeneity and dependence on housing finance characteristics.

### Consequences of house price corrections (overview)
- Economic impacts when corrections start:
  - Financial stability consequences: credit institutions coming under stress.
  - Real economy: slowing real economic activity, affecting consumption and investment.
- Severity determinants:
  - Structural characteristics of housing and mortgage markets, particularly ease of household access to mortgage credit in recent years, are crucial in cross-country differences in impact.

*Italicized source attribution: IMF Working Paper chapter "15.  Back to Earth: Predicted and Actual House Price Changes" (data cut-off end-2009) from the provided PDF content.*

### section into work is to construct ratios of prices to key determinants of house prices.

### _wp12217 - section into work is to construct ratios of prices to key determinants of house prices.

### Price-to-Income Ratio (PIR) analysis
- PIR interprets house price overvaluation when prices exceed historically observed ratios relative to income.
- Figure 10: deviation of house prices to per capita income from historical average for 55 countries.
- Out-of-sample lower-panel exercise: deviations as of 2006Q4 versus realized house price changes since that date show a strong negative relationship: the larger the overvaluation at end-2006, the larger the subsequent drop in house prices.
- Countries with fast house price appreciation and harsh downturns: Denmark, Estonia, Iceland, Ireland, Latvia, Lithuania, Spain, Ukraine, and the U.S.
- Potential misalignment still remains in Denmark and Spain and in a handful of countries with milder downturns.

### Housing Affordability Index (HAI) — U.S. sub-national analysis
- HAI definition: whether a typical (median-income) family could qualify for a traditional mortgage on a median-priced home.
- Data sources: Bureau of the Census (median family income, median home value), Federal Reserve Board (mortgage interest rate).
- Assumptions: down payment = 20 percent of home price; qualifying ratio = 25 percent of median family monthly income.
- Interpretation:
  - HAI = 100: median family has exactly enough income to qualify.
  - HAI > 100: median family has more than enough income (example: 130 means 30 percent more than necessary).
  - HAI < 100: family cannot afford mortgage on a median-priced home (example: 70).
- Figures:
  - Figure 11: HAI as of 2009Q4 across states.
  - Figure 12: comparison of HAI in 2009 to 1990.
- States with least affordable homes: Hawaii, District of Columbia, California, Massachusetts, New Jersey, Arizona, New York, Oregon, and Washington.
- Affordability trends:
  - Affordability has deteriorated across the board except Northeast Corridor (Connecticut, Maine, Massachusetts, New Hampshire, Rhode Island) when benchmarked to 1990; however, these states also show deterioration relative to 2000.
  - Overall, affordability ratios remain above their historical levels (1960 to 2004) in 3 out of 5 states.
- Policy implication: substantial within-country heterogeneity suggests blanket measures (e.g., monetary tightening) may produce unintended consequences.

### Price-to-Rent Ratio (PRR) and user cost of owner-occupied housing
- PRR rationale: asset-pricing analogy — price-to-rent should align with returns (market rent); when PRR is high relative to rents, renting is more attractive and should exert downward pressure on prices.
- U.S. sub-national PRR:
  - Data: median house prices from National Association of Realtors; market rental rates from Department of Housing and Urban Development in metropolitan statistical areas.
  - PRR evolution: decline from 23.17 in 2008Q2 to 19.97 in 2009Q4; latest values remain slightly above historical norms.
- User cost framework (Poterba, 1992):
  - User cost equation presented in text with parameters: τm (marginal tax rate), i (nominal mortgage interest rate), τp (property tax rate), β (risk premium), d (depreciation), m (maintenance), π (expected capital gains), P (house price index).
  - Relationship: price-to-rent ratio equals inverse of user cost expression (as shown in the rearranged formula).
- Data and parameter choices:
  - Tax rates and deductions from Internal Revenue Service and the NBER.
  - Nominal mortgage rates from Federal Reserve Board and Federal Housing Finance Board.
  - House prices from Office of Federal Housing Enterprise Oversight; rent from Bureau of Labor Statistics.
  - Sum β + m + d assumed constant at 4 percent.
  - π proxied by an eight-year moving average of consumer price inflation (Bureau of Labor Statistics).
- Results:
  - PRR at 24.84 lies above the “equilibrium” level around 15.2 as of 2008Q2 (Figure 14).
  - Historical average excess of PRR over inverse user cost: 5.15.
  - Holding rent constant, reaching historical average discrepancy implies almost 26 percent drop in nominal house prices.
  - Assuming a 4 percent increase in rental rates, the implied drop is around 21 percent.
- Sensitivity:
  - Estimates are particularly sensitive to expected capital gains (π) assumptions; alternative expectation formation (e.g., 3-year basis with downward updating) brings PRR and inverse user cost closer.

### Parametric estimates — regression model of real house price changes
- Model specification (change in real house prices as function of):
  - Change in real income per capita (over last quarter).
  - Change in working-age population (over past year).
  - Change in stock prices (over the year before last).
  - Change in bank credit to private sector (over past year).
  - Short-term and long-term interest rates.
  - Lagged affordability level (log ratio of house prices to income per capita).
- Estimation approach:
  - Ordinary least squares (OLS) due to limited country-specific time series length and weak Johansen cointegration test results (cointegration found for 1 out of 3 countries).
  - Sample constrained to 22 advanced economies with sufficient quarterly data.
  - Country-by-country regressions and pooled regressions performed (Tables 4 and 5).
- Key empirical findings:
  - Affordability negatively related to change in prices in 82 percent of countries.
  - Change in income per capita positive in all countries except the U.S. (U.S. coefficient negative but not statistically significant).
  - Population growth positive in 86 percent of cases but statistically significant about one-third of the time.
  - Bank credit to private sector positive in 91 percent of countries; 64 percent deliver correctly-signed statistically significant coefficients.
  - Interest rate coefficients inconsistent: correct (negative) sign in 50 to 64 percent of countries; correct and significant in 27 to 41 percent.
  - Stock price coefficients: positive in 68 percent of cases; statistically significant in 36 percent.
  - Average R-squared across countries: 0.33; poor fit for Australia, Canada, and Italy; good fit for Norway and Spain (and to a lesser extent Finland, Japan, the U.K.).
- Pooled regression extensions:
  - Including square of lagged affordability: coefficient has same sign as affordability, reinforcing stronger downward pressure when deviations are larger.
  - Including change in construction costs: higher construction costs linked to increases in house prices.
  - Including change in commodity (fuel and nonfuel) prices: captures global cyclical synchronization; coefficients on original variables remain virtually unchanged across extensions.
- Translating changes into price gaps:
  - Procedure: assume equilibrium house price index = 100 at chosen base dates, compute predicted changes forward, compare actual index to predicted to obtain estimated price gap.
  - Robustness: averages computed using five base dates (1997Q4, 1998Q4, 1999Q4, 2000Q4, 2001Q4).
  - Heat map (Table 6) used to report average estimated gaps (point estimates not disclosed).
  - Findings: overheating in housing markets at end-2006 and 2007; misalignment unwinding in 2008 and 2009 but large price gaps remain in a handful of countries.
- Predictive performance:
  - Predicted correction as of 2007Q4 (estimated price gap × -1) compared to actual change since that date.
  - Strong positive correlation between predicted and actual changes (Figure 15).
  - Estimates under-predicted severity in Ireland, Denmark, and the U.S.; over-predicted downturn for Australia and Canada.
  - Average predicted decline: 9 percent; realized decline: 5 percent — discrepancy possibly due to policy-induced rebound or ongoing correction anticipating additional 4 percent decline on average.

### Macroeconomic consequences — VAR analysis of housing demand shocks
- VAR setup:
  - Baseline variables: GDP, private consumption, residential investment, CPI (inflation), short-term interest rate, house prices.
  - Identification: Cholesky decomposition ordering — GDP, consumption, residential investment first; CPI and interest rate next; house prices last.
  - Housing shocks labeled “housing demand shocks” reflecting short-run demand-driven price movements with supply fixed.
  - Sample period: 1986Q1 to 2010Q1.
  - Data transformation: log levels for all variables except CPI (log first differences) and short-term interest rate (levels).
  - Lag length: 1 to 2 lags chosen per criteria.
- Shock normalization and results:
  - Two normalizations: initial house price response as decline of 10 percent; and decline equal to two standard deviations of quarterly house price change.
  - Average impact of a 10 percent house price decline on GDP: -1.87 percent.
  - Comparison to literature: Cardarelli et al. (2009) ~ -2 percent; Goodhart and Hofmann (2008) ~ -1 percent.
- Cross-country heterogeneity in GDP response to a 10 percent house price decline:
  - Small-impact countries (GDP decline around or less than ½ percentage points): Austria, Italy, Netherlands, Norway, Switzerland.
  - Large-impact countries (GDP decline close to or more than 1 percentage point): Denmark, Greece, Ireland, Japan, New Zealand, Spain, United States.
  - Other countries fall in the middle.
- Effects on consumption and residential investment: highly heterogeneous across countries.
- Mechanisms for cross-country differences:
  - Greater credit tightening response where financial systems are deeper or mortgage equity withdrawal is available → larger consumption and investment multipliers.
  - Larger or more labor-intensive construction sectors amplify negative employment and activity impacts.

### Relating vulnerabilities to mortgage-market characteristics
- Limited cross-country evidence suggests vulnerability (GDP impact magnitude) relates to:
  - (i) availability of penalty-free prepayment and refinancing,
  - (ii) equity withdrawal options on mortgage loans,
  - (iii) loan-to-value ratios.
- Example statistic (partial): in countries where borrowers can prepay/refinance with little or no penalty, estimated GDP impact of a 10 percent house price decline is -2.77 percent on average (text cuts off before completing contrast statistic).

*Source: _wp12217 - section into work is to construct ratios of prices to key determinants of house prices.*

### 1.71 percent in countries where this option is not available. Similarly, if homeowners are

### _wp12217 - 1.71 percent in countries where this option is not available. Similarly, if homeowners are

### Mortgage-market characteristics and vulnerability
- Average estimated impacts of access to home equity or lines of credit:
  - When homeowners are allowed to withdraw lines of credit on their home equity, the average estimated impact is -2.27 percent.
  - When such an option is not available, the average estimated impact is -1.61 percent.
- Loan-to-value (LTV) at origination:
  - Among countries where a typical mortgage loan carries a loan-to-value ratio of less than 80 percent at origination, the average estimated impact stands much lower than when the typical loan-to-value ratio is greater than 80 percent.
- Relation with mortgage-debt-to-GDP:
  - When outliers (Germany, Japan, Greece, and Netherlands) are excluded, vulnerability appears to be positively related to the ratio of residential mortgage debt to GDP.
- Weaker correlations:
  - Correlation with other characteristics of mortgage markets and structural dimensions such as labor intensity of construction and the share of construction sector in total value added seem to be weaker.
- Mechanism:
  - Ease of household access to mortgage credit amplifies vulnerability because residential real estate and mortgage debt constitute the largest chunk of household balance sheets; if mortgage markets provide opportunities to exploit increases in collateral values more easily, the financial accelerator effect is larger.

### Other factors affecting the amplitude of downturns after house price corrections
- Prevalence of variable-rate mortgages:
  - Finland, Ireland, and Spain have mostly variable rate mortgages.
  - Higher debt burden (interest payments relative to household disposable income) responds more to interest rate hikes if mortgage rates are adjustable and has been historically associated with bigger declines in residential investment during housing busts.
- Legal provisions and recourse:
  - In many countries (France, Germany, Ireland, the Netherlands, Spain, and the United Kingdom), debtors are personally liable for the full amount of mortgage debt, thus reducing incentives for foreclosure.
  - In the United States, mortgage foreclosure is regulated at the state level:
    - In six states, lenders have recourse only to the mortgaged property, which they may repossess and sell.
    - In the other states, debtors are also personally liable for the full amount of the debt, but there are differences in the extent to which lenders can recover the difference between the mortgage debt and the foreclosure sale price.
    - In practice, lenders may choose not to seek deficiency judgments mainly because of the time and cost involved.

### Banking sector exposure to housing
- Indicator described:
  - The value of mortgage loans held by banks, expressed as a multiple of their overall market capitalization, indicates their ability to withstand deterioration of real estate loan portfolios.
- Cross-country variation in the indicator:
  - Indicator varies from about 4 in Denmark and Germany.
  - Less than 3 in Spain.
  - About 1.5 in Canada, Japan, and the United Kingdom.
  - Less than 1 in the United States.
- Empirical implications:
  - Cross-country declines in residential investment during housing cycles have been higher in countries with greater banking sector exposure to mortgage lending, though the effect has not been as strong as that shown earlier with the mortgage-debt-to-GDP ratio.
  - At the current juncture, with bank balance sheets under renewed stress and bank equity prices low, the potential for an adverse impact on the real economy from banking system exposure to mortgage lending is perhaps greater than in the past.

### Conclusions and policy implications
- Historical correction magnitude and duration:
  - If past is prologue, these corrections could average about 23 percent and be spread out over a period of 4-4.5 years from peak to trough.
  - This assessment is in line with inspections of valuation ratios and econometric models of house prices.
- Cross-country differences:
  - Differences in the impact of corrections on the macroeconomy are likely to depend on characteristics of housing finance systems, particularly the ease with which households have been able to access mortgage credit in recent years.
  - Ease of access to mortgage credit is likely correlated with the extent of investment declines during house price corrections and could also have a dampening impact on consumption.
  - Pre-crisis exposure of the financial system to housing markets may affect the extent of macroeconomic losses and the shape and timing of the recovery.
- Data and research recommendation:
  - A major obstacle in conducting formal analysis of global housing cycles and their implications is the lack of data.
  - Efforts should be concentrated on increasing the availability of consistent and reliable data on housing and real estate financing markets.
  - Future research can aim to look into these macro-financial linkages more closely to better understand the consequences of housing boom-bust episodes.

*Source: IMF working paper text (excerpt).*

### References

### References

### Key empirical findings and stylized facts
- Global House Price Index (real house prices, seasonally-adjusted quarterly series, average for 55 countries):
  - Duration: 36 quarters
  - Amplitude: 41 percent
  - Additional sample durations/amplitudes reported: Duration: 22 quarters / Amplitude: 34 percent; Duration: 18 quarters / Amplitude: 17 percent.
- Stylized facts on duration and amplitude of house price cycles (average values across 55 countries):
  - Upturn — Average duration: 19.97; Amplitude: 74.77; Present cycle upturn Duration: 38; Amplitude: 12; Past cycles Duration: 48; Amplitude: 18–23.
  - Downturn — Average deviation from trend (percent): 23.11; Time to adjustment (years): 14.13; Drop necessary (percent): 514.13 (table presents other summary statistics including Median, Maximum, Minimum, Standard deviation with exact numeric values).
- Table 1 (House Price Cycles by Country) reports country-by-country turning points, durations (quarters), and amplitudes (percent). Examples (exact entries):
  - Argentina: Peak 1981Q4^ Trough 1984Q3 — Duration 11; Amplitude -76.43; Upturn duration 11; Amplitude 39.93.
  - Estonia: Peak 1997Q4 — Duration 39; Amplitude 419.77.
  - United States: Peak 2007Q3 — Duration 10; Amplitude -13.80; Past cycles include 1973Q4–1975Q3 Duration 7 Amplitude -4.38, 1979Q1–1982Q4 Duration 15 Amplitude -9.28, 1989Q4–1995Q1 Duration 21 Amplitude -5.54.
- Table 3 (Price-to-Rent Ratio: Adjustment to Benchmark) reports current values, deviation from trend (percent), time to adjustment (years), and drop necessary (percent). Exact summary statistics include:
  - Average: Current value 19.97; Deviation from trend (percent) 4.77; Time to adjustment (years) 23.11; Drop necessary (percent) 14.13; Deviation from trend (percent) 514.13.
  - Median: 18.69, 0.00, 20.00, 12.72, 312.72.
  - Maximum: 53.51, 147.80, 59.27, 81.06, 1562.49.
  - Minimum: 7.71, 0.00, 20.00, 0.00, 20.00.
  - Standard deviation: 7.04, 9.72, 13.25, 10.93, 10.93.
- Table 6 (Misalignment in House Prices: A Heat Map) reports point estimates of price gaps for selected dates. Examples (as of quarters shown):
  - Australia: as of 2006Q4 16.4; as of 2007Q4 18.6; as of 2008Q4 16.8; as of 2009Q4 19.1.
  - Ireland: as of 2006Q4 21.2; as of 2007Q4 19.8; as of 2008Q4 11.7; as of 2009Q4 -1.8.
  - United States: as of 2006Q4 11.4; as of 2007Q4 8.3; as of 2008Q4 5.2; as of 2009Q4 -1.9.

### Econometric results — country-by-country and pooled regressions
- Country-by-country regressions (dependent variable: log change in house price index over the last quarter). Coefficients vary by country; examples of reported point estimates and significance (standard errors in parentheses):
  - Canada (selected coefficients): Affordability, lagged -0.0244 (0.0207); Income per capita, change 0.584** (0.2950); Credit, change 0.0151 (0.0166); Short-term interest rate 0.0010 (0.0016); Observations 155; R-squared 0.12.
  - United Kingdom (selected coefficients): Affordability, lagged -0.0412*** (0.0100); Income per capita, change 0.539** (0.2690); Credit, change 0.0651*** (0.0167); Short-term interest rate 0.0014 (0.0011); Observations 155; R-squared 0.42.
  - United States (selected coefficients): Affordability, lagged -0.0071 (0.0082); Income per capita, change -0.0755 (0.1030); Credit, change 0.100** (0.0171); Short-term interest rate -0.00162*** (0.0005); Observations 155; R-squared 0.32.
- Pooled regressions (Table 5) — dependent variable: House price index, change. Selected coefficients (columns numbered 1–4; robust standard errors in parentheses):
  - Affordability, lagged:
    - Column 1: -0.0130*** (0.0025)
    - Column 2: -0.0335*** (0.0081)
    - Column 3: -0.0269*** (0.0041)
    - Column 4: -0.0268** (0.0042)
  - Income per capita, change:
    - Column 1: 0.415*** (0.0474)
    - Column 2: 0.408*** (0.0472)
    - Column 3: 0.429** (0.0684)
    - Column 4: 0.4397** (0.0948)
  - Working-age population, change:
    - Column 1: 0.694*** (0.1150)
    - Column 2: 0.641*** (0.1160)
    - Column 3: 0.991** (0.1980)
    - Column 4: 1.6677*** (0.3066)
  - Stock prices, change:
    - Column 1: 0.0136*** (0.0021)
    - Column 2: 0.0129*** (0.0021)
    - Column 3: 0.0048* (0.0027)
    - Column 4: 0.0087** (0.0031)
  - Credit, change:
    - Column 1: 0.0369*** (0.0057)
    - Column 2: 0.0367*** (0.0057)
    - Column 3: 0.0187*** (0.0052)
    - Column 4: 0.0174** (0.0079)
  - Short-term interest rate:
    - Column 1: -0.0008** (0.0003)
    - Column 2: -0.0007** (0.0003)
    - Column 3: -0.0009** (0.0004)
    - Column 4: -0.0016** (0.0006)
  - Long-term interest rate: coefficients close to zero across columns (e.g., -0.0004 in Column 1 (0.0004)).
  - Observations and model fit:
    - Column 1: Observations 27,182; R-squared 0.16; Number of countries 222.
    - Column 4: Observations 15,21; R-squared 0.27; Number of countries 22.
- Notes on regressions:
  - Estimation period in principle covers 1970Q1 to 2010Q1, varying by country based on data availability.
  - Affordability is defined as the log of the ratio of house prices to income per capita.
  - Log changes in working-age population and bank credit to the private sector are calculated as the year-on-year change in log levels.
  - All variables are in real terms except short-term and long-term interest rates. Robust standard errors reported. ***, **, * denote significance at the 1, 5, and 10 percent level.

### Macroeconomic impacts of house price corrections
- Table 7 (Maximum Impact of a Negative Shock to House Prices; VAR analysis, identification via Cholesky decomposition with ordering: real GDP, real private consumption, real private residential investment, CPI, nominal short-term interest rate, real house prices). Examples of reported maximum impacts (decline of 10 percent and decline of 2 standard deviations):
  - United States (sample entry): GDP -1.89; Consumption -2.78; Residential investment -20.20; Alternative column entries: GDP -0.42; Consumption -0.62; Residential investment -4.48.
  - Japan: GDP -5.49; Consumption -3.03; Residential investment -19.74; Alternative: GDP -1.21; Consumption -0.67; Residential investment -4.36.
  - United Kingdom: GDP -1.03; Consumption -1.40; Residential investment -12.24; Alternative: GDP -0.52; Consumption -0.71; Residential investment -6.19.
  - Germany: GDP -4.61; Consumption -5.74; Residential investment -38.45; Alternative: GDP -0.65; Consumption -0.81; Residential investment -5.42.
  - New Zealand: GDP -4.18; Consumption -5.70; Residential investment -28.76; Alternative: GDP -1.74; Consumption -2.37; Residential investment -11.95.
  - Notes: VARs estimated for period 1986q1 to 2010q1. For all countries except Ireland variables transformed into log levels (CPI as log first differences; short-term interest rates in levels). For Ireland, GDP and house prices first-differenced.

### Cross-country characteristics and average impacts (Table 8)
- Average estimated impact on GDP by VAR analysis (in percent), conditional on country characteristics:
  - Penalty-fee prepayment and refinancing available: -2.77 (countries included: Australia, Denmark, Japan, United Kingdom, United States).
  - Penalty-fee prepayment and refinancing not available: -1.71 (Austria, Belgium, Canada, Finland, France, Germany, Greece, Italy, Netherlands, Norway, Spain, Sweden).
  - Option to withdraw mortgage equity available: -2.27 (Australia, Canada, Denmark, Finland, Greece, Ireland, Japan, Netherlands, Norway, Sweden, United Kingdom, United States).
  - Option to withdraw mortgage equity not available: -1.61 (Austria, Belgium, France, Germany, Italy, Spain).
  - Typical loan-to-value ratio at origination:
    - less than or equal to 60 percent: -0.06 (Austria, Italy)
    - between 60 and 79 percent: -1.67 (Canada, Finland, France, Germany, Norway, Spain, United Kingdom)
    - more than or equal to 80 percent: -2.21 (Australia, Belgium, Denmark, Ireland, Japan, Netherlands, Sweden, United States)

### Measurement, regional dynamics, and additional figures
- Figures illustrate global concurrence of housing booms and busts, measurement differences across house price indices, and regional versus national dynamics:
  - Figure 2 displays distributions of real house price changes for 2000–07 and 2007–09 across countries (visual ordering of country tickers shown).
  - Figure 3 (U.S. indices) shows year-on-year national house price changes (examples: OFHEO series, Case-Shiller indices, state indices weighted by purchases/refinances).
  - Figure 6 (Local House Price Movements and Turnover in the Netherlands) reports a correlation coefficient of 0.27 between turnover ratio (number sold/available × 100) and lagged quarter-on-quarter house price changes (data for 40 regions from 1995 to 2009).
  - Figure 15 reports out-of-sample fit: Correlation coefficient 0.53; R² = 0.28 between predicted and actual house price changes for 2007-09 (predicted change = estimated price gap × -1; estimated price gap calculated as of 2007Q4).

### Appendix and data notes
- Details on variables used in the econometric analysis are in Appendix Tables 1 and 2. A brief discussion of the construction of the series in Figure 3 is included.
- Data coverage and transformation notes are provided alongside tables: estimation periods, variable definitions (affordability, income per capita, working-age population, credit, stock prices), and treatment of interest rates and inflation measures.

*Italic: Source PDF filename: _wp12217 - References*

### Appendix Table 1. List of Variables

### Appendix Table 1. List of Variables

### Variable list and data sources
- House price index — OECD, BIS, Global Property Guide (compilations of data from private and/or public national sources)
- Disposable income — OECD
- Working-age population — OECD
- Bank credit to the private sector — IMF International Financial Statistics
- Equity price index — IMF International Financial Statistics
- Short-term interest rate — IMF International Financial Statistics
- Long-term interest rate — IMF International Financial Statistics
- Gross domestic product — IMF International Financial Statistics
- Private consumption — OECD
- Residential investment — OECD
- CPI — IMF International Financial Statistics

### Data treatment notes
- All variables except population, credit, and interest rates are in real terms.

*Source: Appendix Table 1. List of Variables*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2012/_wp12217.pdf_