Article

People’s Republic of China—Hong Kong Special Administrative Region: Selected Issues

Author(s):
International Monetary Fund
Published Date:
January 2018
Share
  • ShareShare
Show Summary Details

House Prices and the Effectiveness of Housing Market Policies in Hong Kong SAR1

During the last decade, Hong Kong SAR has experienced a large increase in house prices and credit, prompting the authorities to respond with several rounds of tightening macro-prudential rules and increasing stamp duty taxes. This chapter analyzes the effectiveness of these measures, and finds that they have helped reduce house price appreciation. The estimated impact of a 10 percent LTV tightening is a reduction of house prices of 4.8 percent over the next year. The estimated impact of a 1 percent increase in the ad valorem stamp duty tax is a reduction of house prices of 1.2 percent over the next year. Without these policies, house prices would have been 12.5 percent higher, and the mortgage credit-GDP ratio 15 percent higher.

A. Background

1. House prices and credit have increased substantially in Hong Kong SAR during the last decade. Global low interest rates after the Great Recession and abundant liquidity have led to a credit and housing boom, as in other advanced open economies such as Australia, Canada, Denmark, Norway, Sweden, and Singapore. In addition, domestic factors, including land supply constraints, have contributed to house price increases. In real terms, house prices have increased by 145 percent since mid-2007, while the credit-to-GDP ratio has increased from 183 percent to 316 percent. In contrast, real sector indicators have increased at a slower rate: real GDP growth and CPI inflation have averaged 3 percent in the last decade. This divergence has raised questions about the sustainability of the current credit and housing booms, the potential risk of a disorderly housing market correction, and the implications for financial stability.

House Prices and Credit

Sources: Haver Analytics; BIS.

2. In response, the Hong Kong Monetary Authority (HKMA) has implemented eight rounds of macroprudential tightening since October 2009.2 The tightening measures have included reductions in the maximum loan-to-value (LTV) ratio and debt service-to-income ratio (DSR). Currently, the maximum LTV ratio is 60 percent for the mass market and 50 percent for the luxury market. The maximum DSR ratio is 50 percent. These ratios are tighter for borrowers: (i) whose main source of income is not Hong Kong-based, (ii) with multiple mortgages, (iii) who purchase properties for investment purposes (non-owner occupied properties), and (iv) who receive mortgages based on their net worth instead of their income. The HKMA has also applied stricter risk weights for residential mortgages on banks.

3. The authorities have also implemented several housing market-related tax measures to stabilize the housing market. The Hong Kong SAR government has implemented several increases in the Ad Valorem Stamp Duty (AVD) tax since 2010, with a current rate of 15 percent.3 In October 2012, the government introduced a 15 percent Buyer’s Stamp Duty (BSD) tax that targets certain types of investment demand, including by foreign buyers. In November 2010, a Special Stamp Duty (SSD) tax was implemented, targeting properties resold within 24 months or less with a highest rate of 15 percent. The restriction period for resale was extended to 36 months or less and the highest tax rate raised to 20 percent in October 2012. These measures appear to have had a stronger effect on the number of transactions than on prices. The impact of these measures on house prices, number of agreements and new mortgages appear to be short-lived (Figure 1). However, the econometric analysis in this chapter shows that housing market policies have had a moderating effect on prices.

Figure 1.The Role of Housing Market Policies

Sources: Haver Analytics; and HKMA.

Note: In November 2010 and February 2013, both macro-prudential and stamp duty tax was changed.

4. Housing market policies have helped slow down house price appreciation. The analysis in this chapter suggests that the increase in house prices can be explained by macroeconomic fundamentals, both in the short- and in the long-run.4 In particular, demand factors such as real GDP growth, credit growth, and interest rates, and supply factors such as rents, building costs, and land supply explain the evolution of house prices well. The analysis finds that housing market polices have had significant effects on house prices, but with some differences across market segments. Specifically:

  • The estimated impact of a 10-percentage point tightening in the loan-to-value (LTV) ratio is a reduction of aggregate house prices by 4.8 percent over the next four quarters.5 The analysis shows differences in the transmission mechanism across market classes, with LTV changes having a faster impact in the luxury market.
  • The estimated impact of a 1-percentage point increase in the AVSD tax is a reduction of aggregate house prices by 1.2 percent over the next four quarters.
  • There are differences in the impact of stamp duty taxes across property markets as well: the BSD tax and the SSD tax affect luxury market prices but not mass-market prices.
  • A simulation based on a Dynamic Stochastic General Equilibrium (DSGE) model suggests that without housing market policy measures, real house prices would have been 12.5 percent higher, and the mortgage credit-to-GDP ratio about 15 percent higher.

B. Time Series Empirical Analysis

5. An equilibrium model of the housing market suggests that both demand and supply factors affect house prices in the long run. Following a methodology similar to Craig and Hua (2011) and Leung, Chow and Han (2008), a two-step error correction model specification was used to analyze the long- and short-run determinants of Hong Kong’s housing prices. In the first stage, the level of real house prices is related to macroeconomic fundamentals in a cointegration framework. In addition to the aggregate housing market, the model was applied to all five residential property classes in Hong Kong SAR.6 In the long run, aggregate real house prices are cointegrated with real rents, real credit, real interest rates, the construction deflator, and land supply.7 In particular, as expected, house prices rise with increases in rents, credit and construction costs, but decline with increases in real interest rates and land supply (lagged 4 periods). In the aggregate, real GDP per capita does not affect house prices. (See Table 1).

Table 1.Hong Kong SAR: Long-Run Determinants of House Prices
AggregateClass AClass BClass CClass DClass E
LOG(RENTS)1/0.626

(4.957)***
0.532

(3.106)***
0.676

(5.381)***
0.835

(7.380)***
0.878

(8.058)***
0.935

(9.779)***
LOG(REAL_GDPPC)0.051

(0.468)
-0.148

(-1.396)
0.065

(0.587)
0.535

(3.776)***
0.670

(4.211)***
0.874

(5.374)***
LOG(AGG_CREDIT)0.342

(2.718)***
0.404

(3.248)***
0.289

(2.275)**
0.272

(1.727)*
0.268

(1.590)
0.251

(1.499)
REAL_RATE-0.010

(-2.041)**
-0.015

(-3.221)***
-0.007

(-1.538)
-0.004

(-0.595)
-0.004

(-0.602)
-0.004

(-0.655)
LOG(CONSTRUC_DEFLATOR)0.548

(3.448)***
0.702

(4.476)***
0.517

(3.234)***
0.235

(1.135)
0.176

(0.767)
0.124

(0.533)
LOG(LAND_SUPPLY(-4))-0.005

(-3.081)***
-0.004

(-2.737)***
-0.004

(-2.933)***
-0.006

(-3.247)***
-0.007

(-3.530)***
-0.008

(-3.881)***
C-6.017

(-3.332)***
-4.933

(-2.821)***
-5.494

(-3.054)***
-9.957

(-4.290)***
-11.305

(-4.412)***
-13.306

(-5.143)***
Observations:858585858585
R-squared:0.9790.9850.9770.9610.9510.954
Sample period:1996Q2-2017Q21996Q2-2017Q21996Q2-2017Q21996Q2-2017Q21996Q2-2017Q21996Q2-2017Q2
Note: 1/ Rent data varies based on type of property; 2/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance.
Note: 1/ Rent data varies based on type of property; 2/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance.

6. Long-run determinants of house prices differ across class types. The same model when applied to the five property classes reveals important differences across class types (See Table 1). First, while the coefficient on rents is significant for all property types, it progressively increases through classes A-E. This suggests that prices are more closely linked to rents in the luxury market, where more properties are likely to be used for investment purposes. In the mass market, strong preference for ownership and tax incentives could explain why prices respond less strongly to rents. Second, the estimated coefficients on real GDP per capita are close to zero and non-significant for classes A and B, but considerably larger and significant for classes C-E. This suggests greater pro-cyclicality of the mid and luxury segments of the market. Third, the coefficients on aggregate credit and the real interest rate are significant for class A properties, signifying the impact of availability and cost of credit on lower-end property prices. The construction deflator also has a sizeable impact on classes A and B showing that higher construction costs are associated with higher prices only in the mass market. Finally, the coefficients on land supply are very significant for classes A-E, although, the coefficients seem to be marginally higher for the luxury market.

7. In the short-run, luxury property prices exhibit stronger procyclical behavior compared to the market as a whole. In the second stage of the analysis, the growth rate of house prices is related to the growth rate of macroeconomic fundamentals, the lagged error-correction mechanism coming from the first stage, and policy variables. The second step analysis reveals that house prices, both aggregate and across property classes, are highly correlated with real GDP per capita in the short-run. However, the coefficients are substantially higher for the luxury market (1.22 for class D and 1.65 for class E) demonstrating stronger pro-cyclicality in the short-run. Additionally, rents and the construction deflator continue to be significant determinants of house prices while credit, interest rate and land supply cease to have an impact on prices in the short-run (Table 2).

Table 2.Hong Kong SAR: Short-run Determinants of House Prices and the Role of Macro-prudential Policies
AggregateClass AClass BClass CClass DClass E
C0.001

[0.151]
0.001

[0.147]
0.001

[0.195]
-0.001

[-0.112]
-0.003

[-0.635]
-0.004

[-0.704]
D(Log(Prop_Prices(-1))) 1/0.21

[1.977]*
0.261

[2.598]**
0.199

[2.048]**
0.175

[1.696]*
0.312

[2.730]***
0.075

[0.442]
D(Log(Rents)) 1/0.762

[3.722]**
0.81

[4.823]***
0.751

[4.093]***
0.654

[3.629]***
0.397

[1.650]
0.629

[2.349]**
D(Log(Real_GDPPC))0.57

[2.482]**
0.472

[2.217]**
0.491

[1.996]**
0.866

[3.608]***
1.193

[3.703]***
1.708

[4.781]***
D(Log(Construc_Deflator))0.358

[2.265]**
0.357

[2.585]**
0.338

[2.264]**
0.38

[1.906]
0.374

[1.535]
0.469

[1.928]*
Error_Correction(-1)-0.21

[-3.855]***
-0.252

[-4.881]***
-0.185

[-3.317]***
-0.16

[-2.987]***
-0.157

[-2.792]***
-0.179

[-2.707]***
Changes in LTV (1 Lag)0.0493

[1.029]
0.044

[0.995]
0.039

[0.754]
0.135

[2.300]**
0.200

[2.476]**
0.379

[2.398]**
Changes in LTV (2 Lag)0.104

[2.267]**
0.079

[1.883]*
0.079

[1.481]
0.098

[1.846]*
0.096

[0.747]
0.027

[0.128]
Changes in LTV (3 Lag)0.212

[0.063]**
0.187

[3.395]***
0.163

[2.278]**
0.27

[4.134]***
-0.026

[-0.215]
-0.228

[-1.103]
Changes in LTV (4 Lag)0.114

[1.874]*
0.202

[3.906]***
0.035

[0.485]
0.067

[0.963]
0.004

[0.022]
0.161

[0.599]
Sum of LTV coefficients 4/0.480.5120.2810.5030.2000.379
Wald Test p-value0.0149**0.0030***0.0872*0.0033***0.015**0.019**
Observations:848484848484
R-squared:0.7660.7720.7670.7130.6550.626
Sample period:1996Q3-2017Q21996Q3-2017Q21996Q3-2017Q21996Q3-2017Q21996Q3-2017Q21996Q3-2017Q2
Notes: 1/ Data for property price, rent varies based on type of property; 2/ Data for LTVs categorized as luxury (classes D,E) and non-luxury (classes A,B,C and average); 3/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance. 4/ Sum of LTV coefficents up to the last significant lag at the 10 percent level.
Notes: 1/ Data for property price, rent varies based on type of property; 2/ Data for LTVs categorized as luxury (classes D,E) and non-luxury (classes A,B,C and average); 3/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance. 4/ Sum of LTV coefficents up to the last significant lag at the 10 percent level.

8. Macro-prudential measures implemented by the HKMA have been effective in reducing housing prices, though the transmission is heterogeneous across different market segments. The second step analysis also introduces macro-prudential measures to the housing price equations, in first differences, with lags up to 4 quarters. This choice helps avoid reverse causality issues, and allows for lags in the transmission of policies. The results show changes in macro-prudential policy implemented by the HKMA have reduced housing prices. The estimated impact of a 10 percent tightening in the LTV ratio is a reduction of aggregate house prices by 4.8 percent over the next four quarters.8 The impact of changes in the LTV is reflected much faster in the luxury market as compared to the aggregate and mass markets. For example, the impact of changes in the LTV caps is visible in classes D and E within one quarter, while it takes between three to four quarters to fully materialize in the aggregate market and in classes A to C.

9. The impact of stamp duty taxes is more homogenous across class types, with the ad valorem stamp duty being the most effective tax. Of the three fiscal measures implemented, increases in the ad-valorem stamp duty tax have proven most effective in reducing house prices: a 1 percent increase in the ADV tax reduces aggregate prices by 1.2 percent. The effects of the ADV tax build over four quarters in all markets, similar to the effect of LTVs. Increases in the BSD have a more muted effect and only seem to reduce prices in the luxury market segment. The SSD has been successful in reducing house prices in all categories, but the effect is significant in the aggregate price level and in the luxury market. (Table 3).9

Table 3.Hong Kong SAR: The Effect of Stamp Duty Taxes
AggregateClass AClass BClass CClass DClass E
Ad-valorem Stamp Duty
Lag 1-0.001

[-0.006]
-0.014

[-0.187]
-0.031

[-0.335]
-0.022

[-0.212]
-0.255

[-1.084]
0.326

[1.245]
Lag 20.098

[2.746]***
0.100

[3.129]***
0.125

[3.352]***
-0.027

[-0.582]
-0.190

[-1.100]
-0.446

[-3.320]***
Lag 3-0.060

[-0.123]
-0.308

[-0.642]
-0.143

[-0.303]
-0.059

[-0.112]
0.010

[0.0535]
-0.350

[-0.681]
Lag 4-1.283

[-3.458]***
-1.328

[-4.001]***
-1.639

[-4.714]***
-1.442

[-2.828]***
-0.813

[-2.633]**
-1.047

[-2.524]**
Sum of Coefficients 4/-1.245-1.550-1.688-1.551-1.248-1.518
Wald Test p-value0.049**0.0157**0.0082***0.021**0.021**0.110
Buyer’s Stamp Duty
Lag 10.028

[0.793]
0.031

[0.906]
0.021

[0.607]
-0.004

[-0.102]
0.008

[0.107]
-0.152

[-1.761]*
Lag 2-0.037

[-0.704]
0.012

[0.266]
-0.057

[-1.076]
-0.124

[-2.375]**
-0.210

[-2.640]**
-0.002

[-0.019]
Sum of Coefficients 4/-0.0090.044-0.036-0.128-0.202-0.154
Wald Test p-value0.9150.5790.6630.1470.1860.446
Special Stamp Duty
Lag 1-0.106

[-0.862]
-0.003

[-0.0196]
-0.117

[-0.609]
-0.259

[-0.948]
-0.074

[-0.525]
-0.405

[-2.966]***
Lag 2-0.157

[-1.249]
-0.159

[-0.784]
-0.255

[-1.292]
-0.310

[-1.283]
-0.140

[-0.981]
-0.001

[-0.006]
Lag 3-0.195

[-1.682]*
-0.389

[-0.951]
-0.118

[-0.568]
-0.108

[-0.500]
-0.122

[-1.860]*
-0.229

[-4.491]***
Sum of Coefficients 4/-0.457-0.551-0.489-0.677-0.336-0.634
Wald Test p-value0.1340.3900.3530.2550.2590.022**
Notes: 1/ Data for property price, rent varies based on type of property; 2/ Data for property taxes categorized as luxury (classes D,E) using the highest rate and non-luxury (classes A,B,C and average) using the lowest rate; 3/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance. 4/ Sum of coefficents up to the last significant lag at the 10 percent level.
Notes: 1/ Data for property price, rent varies based on type of property; 2/ Data for property taxes categorized as luxury (classes D,E) using the highest rate and non-luxury (classes A,B,C and average) using the lowest rate; 3/ *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance. 4/ Sum of coefficents up to the last significant lag at the 10 percent level.

C. Counterfactual Scenario

10. A macroeconomic model with a housing sector and credit is employed to evaluate the effects of macroprudential and fiscal policies.10 A standard real business cycle open economy model that takes as given the world interest rate is extended with a housing sector. Both domestic households and foreign investors can purchase the domestic housing stock, which is in fixed supply.11 Household credit is subject to a loan-to-value constraint set by the regulatory authorities. Changes in the LTV affect new loans, but not the stock of existing loans. This assumption adds realism to the model and helps match the evidence that the effects of macroprudential policies on the housing market are subject to lags, as shown in the previous section. The model is estimated using four macroeconomic time series (real GDP, real house prices, the household credit-to-GDP ratio, and the 3-month LIBOR rate deflated by the CPI) and three policy measures (the average LTV ratio, the AVD tax and the BSD tax) for Hong Kong SAR since 1997, using Bayesian methods.

11. Counterfactual simulations suggest that real house prices would be 12.5 percent higher while household credit would be 15 percent higher without policy responses. Since the model is structural, counterfactual scenarios that are Lucas-critique free can be constructed.12 The counterfactual scenarios assume that the policy variables are kept at their pre-October 2009 levels, before any of the new macroprudential and stamp duty taxes were implemented. This assumption implies that the LTV is kept constant at 70 percent, the AVD at its average rate of 1.875 percent, and the BSD at 0 percent. The model confirms the results from the short-run regression analysis and attributes a larger effect to tax measures than macroprudential measures (Figure 2). The cumulative impact of LTV tightening measures dissipates overtime, and is estimated to have contributed about 3 percent to the reduction in house prices at the end of the sample period. This estimate is within the conventional confidence interval of the regression results. The model is also used to understand the effect of all policies on the household debt/GDP ratio, and suggests that the tightening of LTV policies have contributed to reducing household leverage. The contribution of LTV policies is to have reduced the household credit-to-GDP ratio by about 12 percent, and the remaining 3 percent is due to changes in stamp duty taxes This finding is in line with studies such as He (2014).

Figure 2.Counterfactual Scenarios

Sources: Haver Analytics; BIS; and IMF staff calculations.

D. Conclusion

12. Housing market measures have helped reduce house price appreciation and credit growth in Hong Kong SAR, but have not prevented the credit and housing boom. The Hong Kong SAR housing market faced strong demand pressures over the last recent years. Low global interest rates, high liquidity and the cyclical position have contributed to increased house prices, amid tight supply constraints. The macroprudential policy measures have had a significant, although quantitatively small impact on house prices, with some differences across markets, and have helped contain systemic risk by lowering household credit. Stamp duty taxes have also helped contain house price appreciation, with the AVD having been the most effective tax. The BSD and SSD have mostly had an impact on the luxury market.

References

    AhujaAshvin and MalharNalbar2011. Safeguarding Banks and Containing Property Booms: Cross-Country Evidence on Macroprudential Policies and Lessons from Hong Kong SAR. IMF Working Paper 11/284.

    • Search Google Scholar
    • Export Citation

    ChungKenneth2012. Determinants of Residential Property Prices in Hong Kong: A Cointegration Analysis. International Research Journal of Finance and Economics Issue 96 pp. 5562.

    • Search Google Scholar
    • Export Citation

    CraigSean and ChangchunHua2011. Determinants of Property Prices in Hong Kong SAR: Implications for Policy. IMF Working Paper 11/277.

    FunkeMichael and MichaelPaetz2013. Housing Prices and the Business Cycle: An Empirical Application to Hong Kong. Journal of Housing Economics Vol. 22 pp. 6276.

    • Crossref
    • Search Google Scholar
    • Export Citation

    GlindroEloisaSubhanijTientipSzetoJessica and HaibinZhu2008. Determinants of House Prices in Nine Asia Pacific Economies. BIS Working Paper 263.

    • Search Google Scholar
    • Export Citation

    HeDong2014. The Effects of Macroprudential Policies on Housing Market Risks: Evidence from Hong Kong. Banque De France Financial Stability Review No. 18April2014.

    • Search Google Scholar
    • Export Citation

    LeungFrankChowKevin and GaofengHan2008. Long-Term and Short-Term Determinants of House Prices in Hong Kong. HKMA Working Paper 15/2008.

    • Search Google Scholar
    • Export Citation

    WuTommyChengMichael and KenWong2017. Bayesian analysis of Hong Kong’s housing price dynamicsPacific Economic Review Vol. 22 pp.312331.

    • Crossref
    • Search Google Scholar
    • Export Citation
Appendix I. Data Sources and Unit Root Tests
Table A1.Hong Kong SAR: Data Description and Sources
IndicatorDescriptionSource
Property PricesHK: Property Market Price Index: Private Domestic: All Classes (SA, 1999=100)

HK: Property Market Price Index: Private Domestic: Class A (SA, 1999=100)

HK: Property Market Price Index: Private Domestic: Class B (SA, 1999=100)

HK: Property Market Price Index: Private Domestic: Class C (SA, 1999=100)

HK: Property Market Price Index: Private Domestic: Class D (SA, 1999=100)

HK: Property Market Price Index: Private Domestic: Class E (SA, 1999=100)
Haver Analytics
RentsHK: Property Market Rental Index: Pvt Domestic: All Classes (SA, 1999=100)

HK: Property Market: Rental Indices: Private Domestic: Class A (SA, 1999=100)

HK: Property Market Rental Index: Private Domestic: Class B (SA, 1999=100)

HK: Property Market Rental Index: Private Domestic: Class C (SA, 1999=100)

HK: Property Market Rental Index: Private Domestic: Class D (SA, 1999=100)

HK: Property Market Rental Index: Private Domestic: Class E (SA, 1999=100)
Haver Analytics
Real GDP per capitaHong Kong: Gross Domestic Product (SA, Mil.Chn.2015.HK$)

Hong Kong: Mid-Year Population (Thous)
Haver Analytics
CreditHong Kong: Loans and Advances to Customers: Total (EOP,NSA, Mil.HK$)Haver Analytics
InflationConsumer Prices, period averageIMF, World Economic Outlook
Interest RateMortgage Interest Rate (BLR until 2009; BLR & HIBOR Blended rate post 2009)HKMA
Construction DeflatorHong Kong: GDP: GDFCF: Building and Construction: Private Sector (NSA, Mil.HK$)

H.K.: GDP: GDFCF: Bldg & Construction: Private Sector (NSA, Mil.Chn.2015.HK$)
Haver Analytics
Land SupplyGovt. Land Auction Sales: AreaCEIC
Loan-to-valueChanges in LTV ratios for: Income derived in HK, for self use, applicants who have not borrowed or guaranteed other outstanding mortgage(s) at the time of making a mortgage application (DSR based)HKMA
Stamp duty taxesChanges in ad-valorem, special stamp duty, and buyer’s stamp duty taxesInland Revenue Department, Hong Kong
Table A2.Hong Kong SAR: Unit Root Tests
(c,t,L)Level(c,t,L)First Difference
LOG(AGGREGATE PRICES)(c,0,3)0.139(c,0,0)-4.176***
LOG(CLASS A PRICES)(c,0,3)0.418(c,0,0)-4.008***
LOG(CLASS B PRICES)(c,0,3)0.065(c,0,0)-4.318***
LOG(CLASS C PRICES)(c,0,1)-1.289(c,0,0)-4.454***
LOG(CLASS D PRICES)(c,0,1)-1.449(c,0,2)-5.748***
LOG(CLASS E PRICES)(c,0,1)-1.293(c,0,0)-5.469***
LOG(AGGREGATE RENTS)(c,0,3)-0.967(c,0,2)-6.031***
LOG(CLASS A RENTS)(c,0,2)-0.713(c,0,0)-4.111***
LOG(CLASS B RENTS)(c,0,1)-1.85(c,0,2)-6.164***
LOG(CLASS C RENTS)(0,0,3)-0.369(c,0,2)-5.709***
LOG(CLASS D RENTS)(c,0,2)-2.445(c,0,1)-5.543***
LOG(CLASS E RENTS)(c,0,2)-2.316(c,0,1)-5.431***
LOG(REAL_GDPPC)(c,0,2)-0.468(c,0,1)-4.562***
LOG(AGG_CREDIT)(c,0,4)-0.383(0,0,3)-2.089**
REAL_RATE(c,0,0)-1.461(c,0,0)-11.569***
LOG(CONSTRUC_DEFLATOR)(c,0,4)-0.15(c,0,3)-2.737*
LOG(LAND_SUPPLY](0,0,2)1.399(c,0,1)11.779***
RESIDUAL (AGGREGATE)(c,0,0)-3.684***
RESIDUAL (CLASS A)(c,0,0)-4.102***
RESIDUAL (CLASS B)(c,0,0)-3.473**
RESIDUAL (CLASS C)(c,0,0)-3.603***
RESIDUAL (CLASS D)(c,0,0)-3.866***
RESIDUAL (CLASS E)(c,0,0)-4.235***
Notes: Augmented Dickey-Fuller Unit Root Test; Lags in the test are automatically selected; c= constant, t=trend, L=lags; Null hypothesis: variable has a unit root; *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance.
Notes: Augmented Dickey-Fuller Unit Root Test; Lags in the test are automatically selected; c= constant, t=trend, L=lags; Null hypothesis: variable has a unit root; *** refers to 1% significance, ** refers to 5% significance, * refers to 10% significance.
1Prepared by Pau Rabanal (RES) and Ananya Shukla (APD).
2Appendix III in the 2017 IMF Hong Kong SAR Staff Report details all the measures implemented by the authorities between October 2009 and end-2017.
3The highest AVD rate of 15 percent does not apply to an agreement or conveyance for a residential property where the purchaser or transferee is a Hong Kong SAR permanent resident acting on his own behalf and who does not own any other residential property in Hong Kong SAR at the time of acquisition of the subject property. In this case, a lower rate of up to 4.25 percent, depending on the value of the property, applies.
4Chung (2012), Glindro et al. (2013), Wu, Chen, and Wong (2017), and Leung et al. (2008) have examined the determinants of house prices in Hong Kong. Craig and Hua (2011) and Ahuja and Nabar (2011) study the effect of macroprudential policies on house prices in Hong Kong.
5Ahuja and Nabar (2011) report long lags (up to 8 quarters) in the transmission of LTV changes to house prices.
6The residential property classes are as follows: Class A includes properties whose size is less than 40m2, Class B includes properties whose size is between 40 and 69.9m2, Class C includes properties whose size is between 70 and 99.9m2, Class D includes properties whose size is between 100 and 159.9m2, and Class E includes properties whose size is above 160m2. Classes D and E comprise the luxury market.
7Using aggregate domestic credit for use in Hong Kong SAR delivered better results than using mortgage credit or household debt. The Appendix describes the dataset (Table A1) and details the unit root tests for each time series, as well as the cointegration tests (Table A2).
8This estimate coincides with the sum of LTV over the previous 4 quarters because the effect of the first lag of the dependent variable and the effect of reversion to the error correction mechanism cancel out. In the aggregate regression, and for classes A-C, the LTV ratio for borrowers whose income is derived in Hong Kong SAR, who purchase a single property, who only have one mortgage, and who purchase a property with a value below the threshold for the luxury market is used. For classes D-E, the LTV ratio for a borrower of the same characteristics who purchase a property whose value is above the threshold for the luxury market is used.
9For the aggregate price equation, the average of the Scale 1 ad valorem tax rate is used. For classes A-C, the average of the Scale 2 ad valorem tax rate is used. For classes D-E, the highest ad valorem tax rate is used. In the case of the special stamp duty, the average rate is used in the aggregate equation, the lowest rate is used for classes A-C, and the highest rate is used for classes D-E.
10The analysis is based on the forthcoming IMF Working Paper “An Estimated DSGE Model to Analyze Housing Market Policies in Hong Kong SAR” by Rabanal (2018). Funke and Paetz (2013) evaluate the effect of macroprudential policy in Hong Kong using an estimated macroeconomic model.
11Higher housing demand is also driven by foreign investors (particularly from mainland China) who view the Hong Kong housing market as a safe asset.
12The ‘Lucas critique’ is a criticism of econometric policy evaluation procedures that fail to recognize that optimal decision rules of economic agents vary systematically with changes in policy. It criticizes using estimated statistical relationships from past data to forecast the effects of adopting a new policy, because the estimated regression may change along with agents’ decision rules in response to a new policy.

Other Resources Citing This Publication