## _wp11277

## Source details

**Canonical URL:** [_wp11277](https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11277.pdf)

## Other formats

- [Markdown version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11277.pdf.md)
- [Structured JSON version](/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11277.pdf.json)

---

### I. Introduction and purpose
- Purpose: Use an econometric model of residential property prices in Hong Kong SAR to assess the effectiveness of alternative policies in curbing pressures on property prices.
- Fundamental drivers identified:
  - real GDP per capita
  - real domestic credit
  - construction costs
  - land supply
  - the real interest rate
- Policy instruments discussed:
  - land supply
  - interest rates
  - prudential and tax policy (loan-to-value (LTV) ratio; stamp duty on property transactions)
- Methodology overview:
  - Co-integration estimation methodology (two-stage): estimate long-run co-integrating relationship, then short-run dynamics with an error correction equation.
  - Stationary policy variables (LTV and stamp duty) included in the error correction equation to assess their impact.
- High-level conclusions (from model estimation and event study):
  - Real GDP per capita has the strongest long-run influence on property prices, followed by land supply (with a significant lag).
  - Real interest rate and construction cost index have weaker effects.
  - Real domestic credit has the weakest impact.
  - LTV ratio and stamp duty both have statistically significant effects on price; stamp duty effect is very small.
  - LTV and stamp duty effects are temporary and can slow but not fundamentally alter rising price trends.
  - Policy implication: focus on raising land supply to restrain property price increases over the long run, mindful of implementation lags; LTV and stamp duty dampen speculative activity and should be guided by long-run objectives of financial stability and counteracting speculation.

### II. Policies to restrain residential property prices in Hong Kong SAR
- Categories of measures adopted in response to the surge in housing prices since mid-2009:
  - (i) Prudential measures: reductions in LTV limits.
  - (ii) Anti-property market speculation measures: special stamp duty (SSD) on transactions.
  - (iii) Increases in land supply to boost number of apartments.
- Prudential measures — LTV:
  - LTV ratio long-used to limit credit risk.
  - At present, properties valued HK$10 million or above are subject to a 50 percent LTV ratio.
  - A cut in LTV reduces credit available for a given down payment, slowing mortgage lending and purchases.
  - LTV limits are tighter in the luxury market segment; effects may erode over time as borrowers accumulate assets.
- Stamp duty measures — SSD:
  - Government introduced a special stamp duty (SSD) in November 2010 on residential properties resold within 24 months, in addition to an existing stamp duty (of 4.25 percent).
  - SSD tax rate is 15 percent for properties resold within six months; declines in steps the longer a property is held, falling to zero after 24 months.
  - SSD raises the cost of flipping, encourages longer holding; reduces transaction volumes with ambiguous effect on price (possible short-lived sales price push before fading).
- Event-study evidence:
  - Policy change dates examined: October 2009, April 2010, August 2010, and November 2010.
  - November 2010 combined introduction of 15 percent SSD and a 10 percentage point LTV cut.
  - Empirical finding: residential property price growth slowed in the month after policy adjustments but then rebounded; transaction volumes fell for two months then rebounded — suggesting short-lived impact on price and transactions.
- Land supply:
  - Changes in supply affect property price with a substantial lag.
  - Prices fell early last decade after relatively large increase in supply, then rose rapidly in 2009-10 after global financial crisis curtailed new supply.
  - Lag arises because increase in land supply translates into a rise in completed apartments several years later.
  - Government adopted a more proactive land policy with regular auctions to reduce risk of sharp falloff in new supply.

### III. Econometric model specification
- Modeling approach:
  - Reduced-form model derived from supply and demand models.
  - Co-integration framework chosen because property prices and fundamentals are co-integrated.
  - With a single co-integrating vector, use Engle-Granger two-step (first estimate long-run co-integration relationship; second estimate error correction model of short-run dynamics).
  - Stationary policy variables (LTV and stamp duty) are added to the error correction equation to assess their short-run impacts.
- Supply-side variables:
  - Land supply (h): square meters sold at land auctions; expected negative correlation with price.
  - Building costs (c): index of material and labor costs; expected positive correlation with price.
- Demand-side variables:
  - GDP per capita (y): higher income raises housing demand; positive correlation with price.
  - Interest rate (r): the prime rate; higher rates reduce affordability; negative correlation with price.
  - Domestic credit (l): included rather than mortgage credit; captures broader liquidity conditions and outperforms mortgage credit in estimation.
- Specified long-run co-integrating relationship (equation preserved as in source):
  - p* = α0 + α1 y + α2 r + α3 l + α4 c + α5 h_{t-m}
  - h enters with an m-quarter lag to allow for long lag between land supply and completed flats.
- Error-correction short-run model (equation preserved as in source):
  - Δp_t = β0 + λ Π_{t-1} + Σ Δ(variables) + β_LTV Δ LTV_t + β_SDT Δ SDT_t + υ_t
  - Π_{t-1} is the lagged deviation of actual price from long-run equilibrium (error correction term); λ is the speed of adjustment.
  - LTV and stamp duty (SDT) are stationary policy variables included in the error-correction equation.

### IV. Main estimation results — long-run co-integration (DOLS)
- Data and sample:
  - Quarterly data from 1997 to 2011 Q2.
  - Nominal variables deflated by the CPI for overall and luxury markets.
  - Unit root tests: all variables in the co-integration equation have one unit root.
  - Johansen’s system co-integration test: one long-run co-integrating relationship.
  - Sample size: Obs. 56 (1997Q3-2011Q2).
- Estimation method:
  - Dynamic least squares (DOLS) with two lags; heteroskedasticity-autocorrelation consistent (HAC) standard errors.
- Co-integration equation estimation results (Dependent variable: Log(real property price))
  - Overall Market (Beta (t-stat)):
    - Constant: 5.342*** (3.085)
    - Real interest rate: -0.016** (-2.331)
    - Log(real GDP per capita): 1.492*** (6.510)
    - Log(real domestic credit): 0.245** (2.660)
    - Log(real construction cost index): 0.479*** (7.336)
    - Log(Land supply) with 5-quarter lag: -0.800*** (-7.336)
    - Adjusted-R2: 0.966
    - DW-Statistic: 1.577
    - Obs.: 56 (1997Q3-2011Q2)
  - Luxury market (Beta (t-stat)):
    - Constant: 6.872*** (4.055)
    - Real interest rate: -0.027*** (-3.627)
    - Log(real GDP per capita): 2.370*** (10.916)
    - Log(real domestic credit): -0.132 (-1.316)
    - Log(real construction cost index): 0.262*** (3.976)
    - Log(Land supply) with 5-quarter lag: -0.828*** (-8.142)
    - Adjusted-R2: 0.980
    - DW-Statistic: 1.538
    - Obs.: 56 (1997Q3-2011Q2)
- Elasticity interpretations (overall market, variables in natural logs except real interest rate):
  - A one percent rise in real GDP per capita raises real residential property price by around 1.5 percent.
  - A one percent rise in real domestic credit increases real residential property price by 0.2 percent.
  - A one percent increase in real building cost pushes up real residential property price by around 0.5 percent.
  - A one percent increase in land supply drives real residential property prices down by 0.8 percent (with a 5-quarter lag).
  - A one percentage point hike in real interest rate reduces real residential property price by around 1.5 percent.

### V. Main estimation results — error-correction (short run)
- Sample and estimation:
  - Observations: 55 (1997Q4-2011Q2) for both overall and luxury samples.
  - Relationships estimated using ordinary least squares (OLS) with heteroskedasticity-autocorrelation consistent (HAC) standard errors.
  - Significance notation: *** 1% significance, ** 5% significance, * 10% significance.
- Key estimated coefficients (Dependent variable: ∆ Log(real property price))
  - Constant:
    - Overall Market: -0.427*** (-3.780)
    - Luxury market: -0.446*** (-3.063)
  - Error correction term (1-quarter lag):
    - Overall Market: -0.662*** (-3.359)
    - Luxury market: -0.550* (-1.877)
  - ∆ Real interest rate:
    - Overall Market: -0.007 (-1.322)
    - Luxury market: -0.004 (-0.665)
  - ∆ Log(real GDP per capita):
    - Overall Market: 2.267*** (4.291)
    - Luxury market: 2.615*** (4.133)
  - ∆ Log(real domestic credit):
    - Overall Market: 0.143 (0.736)
    - Luxury market: 0.201 (0.944)
  - ∆ Log(real construction cost index):
    - Overall Market: 0.256* (2.009)
    - Luxury market: 0.215 (1.422)
  - ∆ Log (Land supply) with 5-quarter lag:
    - Overall Market: -0.342 (-0.811)
    - Luxury market: -0.364 (-0.817)
  - LTV:
    - Overall Market: 0.641*** (3.673)
    - Luxury market: 0.678*** (3.082)
  - SDT:
    - Overall Market: 0.006*** (3.312)
    - Luxury market: 0.005** (2.206)
  - D03 (SARS dummy):
    - Overall Market: -0.063* (-1.884)
    - Luxury market: -0.062* (-1.956)
  - D08 (Lehman/September 2008 dummy):
    - Overall Market: 0.069** (2.427)
    - Luxury market: 0.056* (1.887)
  - Model fit:
    - Adjusted-R2: 0.432 (Overall); 0.407 (Luxury)
    - AIC: -2.846 (Overall); -2.758 (Luxury)
- Short-run significance and interpretation:
  - Real interest rate, land supply, building costs and real domestic credit do not have statistically significant short run effects on property prices, potentially reflecting long lags and indirect transmission channels.
  - Both the LTV and the stamp duty (SDT) have statistically significant effects on real property prices, but the effect of the stamp duty is very small.
  - A 10 percentage point cut in the LTV temporarily slows inflation by 6.4 percentage points.
  - Stamp duty has an ambiguous effect on price: tax-incidence could initially push up prices while the fall in transactions would tend to depress prices; both effects should fade quickly as buyers hold property for longer to avoid the duty.
  - Evidence indicates the SSD sharply reduced residential property confirmor transactions—confirmor transactions declined by around 80% from 229 cases in November 2010 to 49 cases in May 2011.
  - Note: Stamp duty (SDT) appears not statistically significant on the sample of (1997Q4-2010Q4) for both overall and luxury residential property markets.
  - Footnote: Multiple lags for the LTV and stamp duty do not significantly alter the estimation results in Table 2.

### VI. Explaining the rapid rise in property prices
- The model finds the recent rapid rise in property prices is well explained by highly favorable macroeconomic fundamentals.
- A deviation of actual price from the long run equilibrium price that is statistically “large” is defined as two standard deviations and can be interpreted as evidence of a bubble; Figure 9 shows the actual price stays within the two standard deviation band.
- The error-correction term coefficient corrects around two-third of the deviation between the actual and long-run equilibrium property prices within one quarter.
- Fundamental drivers identified as highly supportive:
  - Negative real interest rates
  - Limited supply of new apartments
  - Rapid real GDP growth
  - Rapid domestic credit growth
- Policy implication from this finding: focus on restraining these fundamental drivers of property prices.

### VII. Conclusions and policy implications
- Long-run influences on property prices (ranked by strength):
  - Real GDP per capita: strongest long run influence.
  - Land supply: significant, works with a significant lag.
  - Real interest rate.
  - Real construction costs.
  - Real domestic credit: weakest impact.
- Policy recommendations:
  - Focus on raising land supply to restrain property price increases over the long run; however, because of the significant lag, exercise care to avoid exacerbating the property price cycle.
  - Recognize that a tightening of liquidity conditions from more restrictive macroprudential policies may raise the real interest rate, which also has a long run influence on prices.
  - Use LTV ratio and stamp duty to dampen speculative activity: these measures can slow the increase in the short run but are temporary and cannot fundamentally alter the long-run rising trend in property prices.
  - Guide LTV and stamp duty policies by long run objectives of financial stability and counteracting speculation, respectively.
- Overall assessment: the recent rapid rise in property prices is well explained by highly supportive macroeconomic fundamentals and the model’s “equilibrium” price tracks the actual price closely.

### VIII. Annex — key descriptive statistics and unit-root tests
- Annex Table 1 — Variables (means, medians, extremes, etc.):
  - Mean:
    - Real property price index: 99.586
    - Real interest rate (% pa): 5.575
    - Real GDP per capita (HK$1,000): 49.787
    - Real tender price index: 848.231
    - Accumulative land supply (1000 sq m): 2154.989
    - Real domestic credit (HK$ bn): 1972.386
  - Median:
    - Real property price index: 94.512
    - Real interest rate (% pa): 5.665
    - Real GDP per capita (HK$1,000): 47.456
    - Real tender price index: 808.217
    - Accumulative land supply (1000 sq m): 2417.630
    - Real domestic credit (HK$ bn): 1869.152
  - Maximum:
    - Real property price index: 157.631
    - Real interest rate (% pa): 14.366
    - Real GDP per capita (HK$1,000): 65.787
    - Real tender price index: 1286.750
    - Accumulative land supply (1000 sq m): 3026.203
    - Real domestic credit (HK$ bn): 3295.911
  - Minimum:
    - Real property price index: 59.300
    - Real interest rate (% pa): -0.795
    - Real GDP per capita (HK$1,000): 37.727
    - Real tender price index: 667.729
    - Accumulative land supply (1000 sq m): 143.853
    - Real domestic credit (HK$ bn): 1509.486
  - Std. Dev.:
    - Real property price index: 25.139
    - Real interest rate (% pa): 3.779
    - Real GDP per capita (HK$1,000): 7.959
    - Real tender price index: 148.995
    - Accumulative land supply (1000 sq m): 721.776
    - Real domestic credit (HK$ bn): 403.767
  - Skewness:
    - Real property price index: 0.589
    - Real interest rate (% pa): 0.467
    - Real GDP per capita (HK$1,000): 0.345
    - Real tender price index: 0.947
    - Accumulative land supply (1000 sq m): -1.332
    - Real domestic credit (HK$ bn): 1.877
  - Kurtosis:
    - Real property price index: 2.665
    - Real interest rate (% pa): 2.726
    - Real GDP per capita (HK$1,000): 1.717
    - Real tender price index: 3.104
    - Accumulative land supply (1000 sq m): 4.087
    - Real domestic credit (HK$ bn): 6.156
  - Jarque-Bera:
    - Real property price index: 4.125
    - Real interest rate (% pa): 2.609
    - Real GDP per capita (HK$1,000): 5.399
    - Real tender price index: 9.893
    - Accumulative land supply (1000 sq m): 22.771
    - Real domestic credit (HK$ bn): 62.129
  - Observations:
    - Real property price index: 66
    - Real interest rate (% pa): 66
    - Real GDP per capita (HK$1,000): 61
    - Real tender price index: 66
    - Accumulative land supply (1000 sq m): 66
    - Real domestic credit (HK$ bn): 62
  - Note: Nominal data are deflated by CPI at the quarter. Data sources: CEIC and Hong Kong Building Department.
- Annex Table 2 — Unit root (ADF) test results:
  - Log(real property price index): Level (c, t, 1) = -0.675; First difference = -5.664***; I(1)
  - Real interest rate: Level (c, t, 1) = -2.247; First difference = -5.425***; I(1)
  - Log(real GDP per capita): Level (c, t, 1) = -1.792; First difference = -5.826***; I(1)
  - Log(building work tender price index): Level (c, t, 1) = -1.698; First difference = -6.044***; I(1)
  - Log(land supply): Level (0, 0, 3) = 0.953; First difference = -31.673***; I(1)
  - Log(real domestic credit): Level (0, 0, 5) = 1.081; First difference = -2. 740***; I(1)
  - Notes:
    - Null hypothesis: variable has a unit root. Lags in the test are automatically selected.
    - *** 1% significance

*Content derived exclusively from the supplied IMF PDF content unit.*

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

### _wp11277 - References

### I. Introduction
- Purpose: Use an econometric model of residential property prices in Hong Kong SAR to assess the effectiveness of alternative policies in curbing pressures on property prices.
- Fundamental drivers of the real property price identified:
  - real GDP per capita
  - real domestic credit
  - construction costs
  - land supply
  - the real interest rate
- Policy instruments discussed:
  - land supply
  - interest rates
  - prudential and tax policy (loan-to-value (LTV) ratio; stamp duty on property transactions)
- Methodology overview:
  - Co-integration estimation methodology (two-stage): estimate long-run co-integrating relationship, then short-run dynamics with an error correction equation.
  - Stationary policy variables (LTV and stamp duty) included in the error correction equation to assess their impact.
- High-level conclusions (from model estimation and event study):
  - Real GDP per capita has the strongest long-run influence on property prices, followed by land supply (with a significant lag).
  - Real interest rate and construction cost index have weaker effects.
  - Real domestic credit has the weakest impact.
  - LTV ratio and stamp duty both have statistically significant effects on price; stamp duty effect is very small.
  - LTV and stamp duty effects are temporary and can slow but not fundamentally alter rising price trends.
  - Policy implication: focus on raising land supply to restrain property price increases over the long run, mindful of implementation lags; LTV and stamp duty dampen speculative activity and should be guided by long-run objectives of financial stability and counteracting speculation.

### II. Policies to restrain residential property prices in Hong Kong SAR
- Context: Surge in housing prices since mid-2009 prompted policy measures to contain risks.
- Three categories of measures:
  - (i) Prudential measures: reductions in LTV limits.
  - (ii) Anti-property market speculation measures: special stamp duty (SSD) on transactions.
  - (iii) Increases in land supply to boost number of apartments.
- Prudential measures — LTV:
  - LTV ratio long-used to limit credit risk (Wong et al., 2011).
  - At present, properties valued HK$10 million or above are subject to a 50 percent LTV ratio.
  - A cut in LTV reduces credit available for a given down payment, slowing mortgage lending and purchases.
  - LTV limits are tighter in the luxury market segment; effects may erode over time as borrowers accumulate assets.
- Stamp duty measures — SSD:
  - Government introduced a special stamp duty (SSD) in November 2010 on residential properties resold within 24 months, in addition to an existing stamp duty (of 4.25 percent).
  - SSD tax rate is 15 percent for properties resold within six months; declines in steps the longer a property is held, falling to zero after 24 months.
  - SSD raises the cost of flipping, encourages longer holding; reduces transaction volumes with ambiguous effect on price (possible short-lived sales price push before fading).
- Event-study evidence:
  - Policy change dates examined: October 2009, April 2010, August 2010, and November 2010.
  - November 2010 combined introduction of 15 percent SSD and a 10 percentage point LTV cut.
  - Empirical finding: residential property price growth slowed in the month after policy adjustments but then rebounded; transaction volumes fell for two months then rebounded — suggesting short-lived impact on price and transactions.
- Land supply:
  - Changes in supply affect property price with a substantial lag.
  - Prices fell early last decade after relatively large increase in supply, then rose rapidly in 2009-10 after global financial crisis curtailed new supply.
  - Lag arises because increase in land supply translates into a rise in completed apartments several years later.
  - Government adopted a more proactive land policy with regular auctions to reduce risk of sharp falloff in new supply.

### III. An econometric model of property prices in Hong Kong SAR
- Model purpose: control for fundamental macroeconomic variables driving property prices to assess policy impacts.
- Modeling approach:
  - Reduced-form model derived from supply and demand models.
  - Co-integration framework chosen because property prices and fundamentals are co-integrated.
  - With a single co-integrating vector, use Engle-Granger two-step (first estimate long-run co-integration relationship; second estimate error correction model of short-run dynamics).
  - Stationary policy variables (LTV and stamp duty) are added to the error correction equation to assess their short-run impacts.
- Supply-side variables:
  - Land supply (h): square meters sold at land auctions; expected negative correlation with price.
  - Building costs (c): index of material and labor costs; expected positive correlation with price.
- Demand-side variables:
  - GDP per capita (y): higher income raises housing demand; positive correlation with price.
  - Interest rate (r): the prime rate; higher rates reduce affordability; negative correlation with price.
  - Domestic credit (l): included rather than mortgage credit; captures broader liquidity conditions and outperforms mortgage credit in estimation.
- Specified long-run co-integrating relationship (equation (1) as given in source):
  - p* = α0 + α1 y + α2 r + α3 l + α4 c + α5 h_{t-m}  (formula notation preserved as in source)
  - h enters with an m-quarter lag to allow for long lag between land supply and completed flats.
- Error-correction short-run model (equation (2) as given in source):
  - Δp_t = β0 + λ Π_{t-1} + Σ Δ(variables) + β_LTV Δ LTV_t + β_SDT Δ SDT_t + υ_t  (formula notation preserved as in source)
  - Π_{t-1} is the lagged deviation of actual price from long-run equilibrium (error correction term); λ is the speed of adjustment.
  - LTV and stamp duty (SDT) are stationary policy variables included in the error-correction equation.

### IV. Model estimation results
- Data and sample:
  - Quarterly data from 1997 to 2011 Q2.
  - Nominal variables deflated by the CPI for overall and luxury markets.
  - Unit root tests: all variables in the co-integration equation have one unit root.
  - Johansen’s system co-integration test: one long-run co-integrating relationship.
  - Sample size: Obs. 56 (1997Q3-2011Q2).
- Estimation method:
  - Dynamic least squares (DOLS) with two lags; heteroskedasticity-autocorrelation consistent (HAC) standard errors.
- Co-integration equation estimation results (Table 1) — Dependent variable: Log(real property price)
  - Overall Market (Beta (t-stat)):
    - Constant: 5.342*** (3.085)
    - Real interest rate: -0.016** (-2.331)
    - Log(real GDP per capita): 1.492*** (6.510)
    - Log(real domestic credit): 0.245** (2.660)
    - Log(real construction cost index): 0.479*** (7.336)
    - Log(Land supply) with 5-quarter lag: -0.800*** (-7.336)
    - Adjusted-R2: 0.966
    - DW-Statistic: 1.577
    - Obs.: 56 (1997Q3-2011Q2)
  - Luxury market (Beta (t-stat)):
    - Constant: 6.872*** (4.055)
    - Real interest rate: -0.027*** (-3.627)
    - Log(real GDP per capita): 2.370*** (10.916)
    - Log(real domestic credit): -0.132 (-1.316)
    - Log(real construction cost index): 0.262*** (3.976)
    - Log(Land supply) with 5-quarter lag: -0.828*** (-8.142)
    - Adjusted-R2: 0.980
    - DW-Statistic: 1.538
    - Obs.: 56 (1997Q3-2011Q2)
- Elasticity interpretations (overall market, variables in natural logs except real interest rate):
  - A one percent rise in real GDP per capita raises real residential property price by around 1.5 percent.
  - A one percent rise in real domestic credit increases real residential property price by 0.2 percent.
  - A one percent increase in real building cost pushes up real residential property price by around 0.5 percent.
  - A one percent increase in land supply drives real residential property prices down by 0.8 percent (with a 5-quarter lag).
  - A one percentage point hike in real interest rate reduces real residential property price by around 1.5 percent.
- Error-correction equation:
  - Estimated results are shown for overall and luxury market segments (Table 2 referenced in source).
  - Dummy variables included for September 2008 (Lehman Brothers failure) and the SARS episode in Hong Kong SAR.
  - Short-run result noted: a one percentage point increase in the real GDP growth rate raises real property price inflation by (result text continues in source).

*Italic source attribution: Content derived exclusively from the supplied IMF PDF content unit.*

### 2.3 percentage points; but that the real interest rate, land supply, building costs and real domestic

### _wp11277 - 2.3 percentage points; but that the real interest rate, land supply, building costs and real domestic

### Key empirical findings
- Real interest rate, land supply, building costs and real domestic credit do not have statistically significant short run effects on property prices; this may reflect long lags and relatively indirect transmission channels (i.e. through the mortgage market) that limit their short run effects.
- Both the LTV and the stamp duty (SDT) have statistically significant effects on real property prices, but the effect of the stamp duty is very small.
- A 10 percentage point cut in the LTV temporarily slows inflation by 6.4 percentage points.
- Stamp duty has an ambiguous effect on price: tax-incidence could initially push up prices while the fall in transactions would tend to depress prices; both effects should fade quickly as buyers hold property for longer to avoid the duty.
- Evidence indicates the SSD sharply reduced residential property confirmor transactions—confirmor transactions declined by around 80% from 229 cases in November 2010 to 49 cases in May 2011.

### Estimation results (Error Correction Model for Residential Property Prices)
Dependent variable: ∆ Log(real property price)

- Constant
  - Overall Market: -0.427*** (-3.780)
  - Luxury market: -0.446*** (-3.063)
- Error correction term (1-quarter lag)
  - Overall Market: -0.662*** (-3.359)
  - Luxury market: -0.550* (-1.877)
- ∆ Real interest rate
  - Overall Market: -0.007 (-1.322)
  - Luxury market: -0.004 (-0.665)
- ∆ Log(real GDP per capita)
  - Overall Market: 2.267*** (4.291)
  - Luxury market: 2.615*** (4.133)
- ∆ Log(real domestic credit)
  - Overall Market: 0.143 (0.736)
  - Luxury market: 0.201 (0.944)
- ∆ Log(real construction cost index)
  - Overall Market: 0.256* (2.009)
  - Luxury market: 0.215 (1.422)
- ∆ Log (Land supply) with 5-quarter lag
  - Overall Market: -0.342 (-0.811)
  - Luxury market: -0.364 (-0.817)
- LTV
  - Overall Market: 0.641*** (3.673)
  - Luxury market: 0.678*** (3.082)
- SDT
  - Overall Market: 0.006*** (3.312)
  - Luxury market: 0.005** (2.206)
- D03
  - Overall Market: -0.063* (-1.884)
  - Luxury market: -0.062* (-1.956)
- D08
  - Overall Market: 0.069** (2.427)
  - Luxury market: 0.056* (1.887)
- Adjusted-R2: 0.432 (Overall); 0.407 (Luxury)
- AIC: -2.846 (Overall); -2.758 (Luxury)
- Observations: 55 (1997Q4-2011Q2) for both overall and luxury samples
- Notes:
  - Relationships estimated using ordinary least squares (OLS) with heteroskedasticity-autocorrelation consistent (HAC) standard errors.
  - Stamp duty (SDT) appears not statistically significant on the sample of (1997Q4-2010Q4) for both overall and luxury residential property markets.
  - Significance notation: *** 1% significance, ** 5% significance, * 10% significance.
- Footnotes:
  - Multiples lags for the LTV and stamp duty do not significantly alter the estimation results in Table 2.
  - Residential confirmor transactions declined by around 80% from 229 cases in November 2010 to 49 cases in May 2011.

### Explaining the rapid rise in property prices
- The model finds the recent rapid rise in property prices is well explained by highly favorable macroeconomic fundamentals.
- A deviation of actual price from the long run equilibrium price that is statistically “large” is defined as two standard deviations and can be interpreted as evidence of a bubble; Figure 9 shows the actual price stays within the two standard deviation band.
- The error-correction term coefficient corrects around two-third of the deviation between the actual and long-run equilibrium property prices within one quarter.
- Fundamental drivers identified as highly supportive:
  - Negative real interest rates
  - Limited supply of new apartments
  - Rapid real GDP growth
  - Rapid domestic credit growth
- Policy implication from this finding: focus on restraining these fundamental drivers of property prices.

### Conclusions and policy implications
- Long-run influences on property prices (ranked by strength):
  - Real GDP per capita: strongest long run influence.
  - Land supply: significant, works with a significant lag.
  - Real interest rate.
  - Real construction costs.
  - Real domestic credit: weakest impact.
- Policy recommendations:
  - Focus on raising land supply to restrain property price increases over the long run; however, because of the significant lag, exercise care to avoid exacerbating the property price cycle.
  - Recognize that a tightening of liquidity conditions from more restrictive macroprudential policies may raise the real interest rate, which also has a long run influence on prices.
  - Use LTV ratio and stamp duty to dampen speculative activity: these measures can slow the increase in the short run but are temporary and cannot fundamentally alter the long-run rising trend in property prices.
  - Guide LTV and stamp duty policies by long run objectives of financial stability and counteracting speculation, respectively.
- Overall assessment: the recent rapid rise in property prices is well explained by highly supportive macroeconomic fundamentals and the model’s “equilibrium” price tracks the actual price closely.

### Annex — Key descriptive statistics (Annex Table 1)
- Variables reported: Real property price index; Real interest rate (% pa); Real GDP per capita (HK$1,000); Real tender price index; Accumulative land supply (1000 sq m); Real domestic credit (HK$ bn).
- Mean:
  - Real property price index: 99.586
  - Real interest rate (% pa): 5.575
  - Real GDP per capita (HK$1,000): 49.787
  - Real tender price index: 848.231
  - Accumulative land supply (1000 sq m): 2154.989
  - Real domestic credit (HK$ bn): 1972.386
- Median:
  - Real property price index: 94.512
  - Real interest rate (% pa): 5.665
  - Real GDP per capita (HK$1,000): 47.456
  - Real tender price index: 808.217
  - Accumulative land supply (1000 sq m): 2417.630
  - Real domestic credit (HK$ bn): 1869.152
- Maximum:
  - Real property price index: 157.631
  - Real interest rate (% pa): 14.366
  - Real GDP per capita (HK$1,000): 65.787
  - Real tender price index: 1286.750
  - Accumulative land supply (1000 sq m): 3026.203
  - Real domestic credit (HK$ bn): 3295.911
- Minimum:
  - Real property price index: 59.300
  - Real interest rate (% pa): -0.795
  - Real GDP per capita (HK$1,000): 37.727
  - Real tender price index: 667.729
  - Accumulative land supply (1000 sq m): 143.853
  - Real domestic credit (HK$ bn): 1509.486
- Std. Dev.:
  - Real property price index: 25.139
  - Real interest rate (% pa): 3.779
  - Real GDP per capita (HK$1,000): 7.959
  - Real tender price index: 148.995
  - Accumulative land supply (1000 sq m): 721.776
  - Real domestic credit (HK$ bn): 403.767
- Skewness:
  - Real property price index: 0.589
  - Real interest rate (% pa): 0.467
  - Real GDP per capita (HK$1,000): 0.345
  - Real tender price index: 0.947
  - Accumulative land supply (1000 sq m): -1.332
  - Real domestic credit (HK$ bn): 1.877
- Kurtosis:
  - Real property price index: 2.665
  - Real interest rate (% pa): 2.726
  - Real GDP per capita (HK$1,000): 1.717
  - Real tender price index: 3.104
  - Accumulative land supply (1000 sq m): 4.087
  - Real domestic credit (HK$ bn): 6.156
- Jarque-Bera:
  - Real property price index: 4.125
  - Real interest rate (% pa): 2.609
  - Real GDP per capita (HK$1,000): 5.399
  - Real tender price index: 9.893
  - Accumulative land supply (1000 sq m): 22.771
  - Real domestic credit (HK$ bn): 62.129
- Observations:
  - Real property price index: 66
  - Real interest rate (% pa): 66
  - Real GDP per capita (HK$1,000): 61
  - Real tender price index: 66
  - Accumulative land supply (1000 sq m): 66
  - Real domestic credit (HK$ bn): 62
- Note: Nominal data are deflated by CPI at the quarter. Data sources: CEIC and Hong Kong Building Department.

### Annex — Unit root test results (ADF-test) (Annex Table 2)
- Log(real property price index): Level (c, t, 1) = -0.675; First difference = -5.664***; I(1)
- Real interest rate: Level (c, t, 1) = -2.247; First difference = -5.425***; I(1)
- Log(real GDP per capita): Level (c, t, 1) = -1.792; First difference = -5.826***; I(1)
- Log(building work tender price index): Level (c, t, 1) = -1.698; First difference = -6.044***; I(1)
- Log(land supply): Level (0, 0, 3) = 0.953; First difference = -31.673***; I(1)
- Log(real domestic credit): Level (0, 0, 5) = 1.081; First difference = -2. 740***; I(1)
- Notes:
  - Null hypothesis: variable has a unit root. Lags in the test are automatically selected.
  - *** 1% significance

*Source: IMF staff estimations and annex tables from the provided document.*

---


_Source: https://www.imf.org/-/media/websites/imf/imported-full-text-pdf/external/pubs/ft/wp/2011/_wp11277.pdf_