Journal Issue
Share
Article

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

Author(s):
International Monetary Fund. Asia and Pacific Dept
Published Date:
January 2019
Share
  • ShareShare
Show Summary Details

Financial Conditions and Growth-at-Risk in Hong Kong SAR1

As a small, open economy and a regional financial center, Hong Kong SAR’s financial conditions are affected by domestic and external developments and can exert a sizable impact on growth. Financial conditions indicators are useful in detecting financial vulnerabilities and can serve as a predictor of downside risks to growth. Using a country-specific Financial Conditions Index for Hong Kong SAR, we find that asset market valuations play an important role in determining its financial conditions. While near-term risks to growth are limited, easy financing conditions pose downside risks to medium-term growth. Moreover, a sizable correction to equity and property markets could have a large negative impact on near-term growth.

1. Financial conditions refer to the ease with which households and corporates access funding. To measure funding conditions, including the costs of credit as well as the terms and conditions borrowers face, financial condition indexes (FCI) include information on the price of risks, credit aggregates and external conditions. Using methodologies introduced in the IMF GFSR (2017b), the FCI for Hong Kong SAR is estimated using principal component analysis (PCA) and incorporates information from 17 variables.2 FCI variables are grouped into separate (i.e. “partitioned”) categories to help better assess different financial conditions.

Table 1.Hong Kong SAR: Underlying FCI Variables by Category
Price-based indicators (Price of risk)Property indicatorsEquity indicatorsLeverage-based indicatorsFX
3-month Hibor change, yoy, bpsResidential property price index, log diff.Equity market capitalization, log diff.HKD domestic credit, log diff.HKD REER, Log diff
3-month Hibor-Libor spread, bpsOffice space price index, log diff.Hang Seng Index (HSI), log diff.HK credit gap (HKMA measure), level diff.
3-month Hibor-3-month EFB spread, bpsRealized HSI volatility, log diffHousehold loans, log diff.
2-year EFN – 3 month EFB spread, bpsHousehold loans/bank asset, level diff.
Property-related loans, log diff.
Property-related loans/bank asset, level diff.
Outstanding HKD debt instruments, excluding EFBN, log diff.
Source: Bloomberg, CEIC, HKMA, IMF staff estimates
Source: Bloomberg, CEIC, HKMA, IMF staff estimates

For example, given concerns regarding housing market overvaluation and to better understand the role asset prices play in Hong Kong SAR’s financial conditions, property and equity market indicators are grouped together to summarize asset price valuations. Other variables, such as price-based indicators, are grouped separately to measure the cost of borrowing.

2. While financial conditions in Hong Kong SAR remain accommodative, the degree of accommodation is declining. Since the end of 2017, interest rates, reflecting rising rates in the U.S., have increased. These developments, combined with sizable equity market losses have contributed to a tightening of financial conditions in Hong Kong SAR.

Hong Kong SAR Financial Conditions Index

Sources: Bloomberg LP, CEIC and IMF Staff Estimates.

3. Financial conditions in Hong Kong SAR are highly correlated with external conditions, reflecting its role as a financial center. Not surprisingly, given the presence of international banks and Hong Kong SAR’s role as a trade intermediary, including as a gateway to Mainland China, global risk sentiment affects local market volatility. Financial conditions have been mostly accommodative since the Great Financial Crisis. However, increased volatility and uncertainty during the European sovereign debt crisis in late 2011 as well as changes to the RMB fixing mechanism in 2015 tightened financial conditions in Hong Kong SAR.

4. Asset market performance – changes in property and equity market valuations in particular – plays a large role in the swings in Hong Kong SAR’s FCI. On average, property market valuation explains 27 percent of FCI movements in 2016–2018. Equity valuation and volatility captured another 13 percent. Price-based indicators, a proxy for funding costs, captured roughly 25 percent while leverage, proxied by credit supplied, played a relatively limited role, at 13 percent over the same period. Gains in property and equity markets as well as low borrowing costs in 2017 helped to ease overall financial conditions. However, recent losses in equities, combined with rising rates and reduced credit growth in the economy, have moved overall financial conditions closer towards neutral.

Underlying FCI Variables by Category

Source: Bloomberg, CEIC, Staff estimates

5. Quantile regressions, based on the partitioned FCI components, offer insights into the impact of financial conditions on growth and underscore the large role asset prices play in Hong Kong SAR’s financial conditions. These regressions show that different aspects of financial conditions – risks or asset valuations, for example – exert varying degrees of impact on growth and their impacts vary across different growth quantiles (Figure 1, Annex tables 1 & 2 and Technical Appendix). For example, while a one-unit increase3 in the price of risk is associated with declines in growth rates in the near term,4 higher risk prices have bigger impacts when growth rates are low (e.g. below the 50th percentile). Specifically, increases in the price of risks, such as higher funding costs or wider yield spreads by one standard deviation, for example, have limited impact on growth when growth rate is at the 90th percentile, but could slow growth by as much as 0.8 percentage points when growth rates are at the 25th or 10th percentile. By contrast, over the same horizon, increases in property market values are particularly supportive during low growth periods but have relatively muted impact when growth is high (e.g. above the 50th percentile). Meanwhile, while leverage and property prices tend to be positively correlated, increases in leverage are supportive of growth only in high-growth scenarios, as noted earlier, while property price increases provide bigger boosts to growth at lower growth levels. This divergence is likely related to the fact that house prices in Hong Kong SAR are relatively more detached from leverage (HKMA Quarterly Bulletin 2002).

6. Interestingly, while increases in leverage are generally correlated with higher growth, leverage can also detract from growth. An increase in leverage is associated with reduced growth at lower growth quantiles in the near term (though the results are not conclusive). This finding, taken together with insights from the impact of higher risk prices on growth distribution, suggests a negative feedback loop between higher debt, higher funding costs and growth. In particular, rising debt service payments, combined with a growing debt load, is particularly pernicious during low-growth periods, and could amplify the negative feedback loop between debt service capacity and investment and consumption.

7. Over the medium term, easier financing conditions tend to be associated with weaker growth, underscoring their risks to longer-term growth. Notably, property price gains are associated with lower growth over the medium term, in contrast to the boost these increases provide over the near term. Reflecting the large role asset prices play in Hong Kong SAR’s financial conditions, changes in property prices generally have larger impacts on growth relative to price and leverage, particularly at lower growth quantiles. This suggests that property prices can be a powerful amplifier of risks during lower-growth periods, providing sizable boosts to growth during upturns but also considerable drags during downturns.

Figure 1.Hong Kong SAR: Changes in Financial Conditions Affect Growth Risks

Note: The coefficients are standardized to show the impact of one standard deviation increase in the current quarter FCI on GDP 4 and 12 quarters ahead (also expressed as standard deviations). Solid blocks indicate statistical significance.

8. Given current financial conditions, the Growth-at-Risk analysis suggests a low probability of recession in the near and medium term. Financial conditions are currently still easy and are moving towards neutral. The distribution of risks to growth is thus largely stable over the next three years. The forecasted probability of a recession over the next year was around 10 percent, not much different from 2017. Over the medium term out to 2020, recession risks edge higher to about 18 percent. Using quantile regressions, the 5th percentile in the near term is -2.12 percent, and medium term is -3.36 percent5

Growth at Risk Density

(Probability Density as of 2018 Q2)

Sources: Bloomberg Finance L.P.; IMF, World Economic Out look database; and IMF staff estimates.

9. The large role asset valuations play in Hong Kong SAR’s financial conditions suggests that their correction – with some already underway – could negatively affect financial conditions and output growth. These potential changes in financial conditions matter as they can amplify adverse shocks to the economy through financial frictions. Buoyant asset markets lift risk appetite, encourage investment and reduce funding costs. However, as these conditions deteriorate, they can introduce sizable downside risks. Tighter financial conditions have historically been associated with a decline in output growth. As seen in the results from the quantile regression discussed above, higher funding costs, lower credit supplied and reduced wealth effects from lower asset prices could weigh on growth, introducing a negative feedback loop between tightening financial conditions and weakening consumption and investment. There is also a trade-off between growth over the near term and the medium term as suggested by the quantile regressions: to the extent that near-term growth slows due to tightened financial conditions, risks to medium-term growth is lower by comparison.

References

    AdrianT.N.Boyarchenko andD.GiannoneVulnerable GrowthAmerican Economic Review forthcoming.

    HeX. andL.Zhu2003A Lack-of-fit Test for Quantile RegressionJournal of the American Statistical Association98 no. 464: 10131022.

    • Search Google Scholar
    • Export Citation

    Hong Kong Monetary Authority (2002) Quarterly Bulletin August 2002.

    International Monetary Fund2017aAre Countries Losing Control of Domestic Financial Conditions?Global Financial Stability Report April 2017 Chapter 3Washington DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund2017bFinancial Conditions and Growth at RiskGlobal Financial Stability Report October 2017 Chapter 3Washington DC.

    • Search Google Scholar
    • Export Citation

    International Monetary Fund2018A Bumpy Road AheadGlobal Financial Stability Report April 2018 Chapter 1Washington DC.

    KoenkerR. andMachadoJ. A.1999. “Goodness of Fit and Related Inference Processes for Quantile RegressionJournal of the American Statistical Association94(448) pp.12961310.

    • Search Google Scholar
    • Export Citation

    SingerS. andNelderJ.2009. “Nelder-mead algorithmScholarpedia4(7) p.2928.

Appendix I. Estimation of the Conditional Quantiles

For the horizon h ∈{4,12}, where h represents the quarters ahead, quantile regressions of the dependent GDP growth variables yt+h are estimated conditional on macro-financial variables Xit, Q(yt+h,τ | {Xi,t}i∈p), for a given date t, based on the point estimates of the coefficients α^τandβ^iτ:

Where yt+h represents future growth h quarters ahead, Xit is the partition i (for instance price, property or leverage), βiτ is the coefficient of the τ quantile regression, ατ is the associated constant and ϵi,tτ the residual. The quantile regressions are estimated at different points of the distribution of yt+h, τ ∈ {0.1, 0.25, 0.5, 0.75 ,0.9}. Each beta coefficients represents the macrofinancial linkage between the variable Xi,t and future growth, at different points of the distribution of GDP growth (basically, the business cycle).

Using quantile regressions for estimating the conditional distribution has many advantages: first, under standard assumptions, quantile regressions provides the best unbiased linear estimator for the conditional quantile; second, quantile regressions are robust to outliers. Finally, the asymptotic properties of the quantile regression estimator are well-known and easy to derive.

For a more detailed discussion, please refer to Adrian et al.

Appendix Table I.Quantile Regressions Results 4 Quarters Ahead
QuantileEstimates (standardized)Standard Error95% ConfidenceLimitsP-Value
(intercept)0.1-1.070.26-1.33-0.810.00
Price0.1-0.740.25-0.98-0.490.00
Property0.10.800.360.441.160.00
Equity0.1-0.580.37-0.95-0.210.01
Leverage0.1-0.200.43-0.630.230.44
REER0.10.200.34-0.140.530.33
Real GDP CHN (yoy growth)0.10.060.30-0.240.360.73
(intercept)0.25-0.560.17-0.73-0.390.00
Price0.25-0.770.20-0.97-0.570.00
Property0.250.510.240.270.740.00
Equity0.25-0.240.23-0.47-0.010.09
Leverage0.25-0.060.28-0.340.230.73
REER0.250.140.23-0.090.380.32
Real GDP CHN (yoy growth)0.250.280.220.060.500.03
(intercept)0.50.120.120.000.230.10
Price0.5-0.360.17-0.53-0.190.00
Property0.50.450.160.300.610.00
Equity0.5-0.100.17-0.280.070.33
Leverage0.50.120.19-0.070.320.30
REER0.5-0.150.16-0.310.000.11
Real GDP CHN (yoy growth)0.50.240.160.070.400.02
(intercept)0.750.430.130.300.560.00
Price0.75-0.240.17-0.41-0.070.02
Property0.750.190.170.020.360.07
Equity0.750.000.18-0.180.180.99
Leverage0.750.210.220.000.430.10
REER0.75-0.280.18-0.46-0.100.01
Real GDP CHN (yoy growth)0.750.220.200.030.420.06
(intercept)0.90.960.180.781.140.00
Price1.9-0.030.27-0.300.240.85
Property2.90.020.29-0.270.310.92
Equity3.90.270.31-0.040.580.15
Leverage4.90.220.33-0.110.540.28
REER5.9-0.510.28-0.79-0.230.00
Real GDP CHN (yoy growth)6.90.170.28-0.110.440.32
Appendix Table II.Quantile Regressions Results – 12 Quarters Ahead
QuantileEstimates (standardized)Standard Error95% ConfidenceLimitsP-Value
(intercept)0.1-1.190.26-1.45-0.930.00
Price0.1-0.010.22-0.240.210.92
Property0.1-0.980.41-1.40-0.570.00
Equity0.10.420.240.180.660.00
Leverage0.10.620.290.320.910.00
REER0.1-0.530.34-0.87-0.190.01
Real GDP CHN (yoy growth)0.1-0.530.31-0.84-0.220.01
(intercept)0.25-0.570.20-0.76-0.370.00
Price0.250.050.27-0.210.320.74
Property0.25-0.510.26-0.77-0.250.00
Equity0.250.120.30-0.190.420.53
Leverage0.250.060.26-0.200.330.69
REER0.25-0.150.29-0.440.140.40
Real GDP CHN (yoy growth)0.25-0.460.28-0.74-0.180.01
(intercept)0.5-0.030.17-0.200.140.78
Price0.50.280.260.020.530.08
Property0.5-0.390.23-0.61-0.160.01
Equity0.50.040.26-0.220.300.80
Leverage0.5-0.160.29-0.440.130.36
REER0.5-0.150.23-0.380.070.26
Real GDP CHN (yoy growth)0.5-0.180.23-0.420.050.20
(intercept)0.750.670.200.470.880.00
Price0.750.020.35-0.330.370.92
Property0.75-0.190.30-0.490.110.29
Equity0.750.310.310.010.620.09
Leverage0.750.300.42-0.120.720.25
REER0.75-0.290.26-0.55-0.030.07
Real GDP CHN (yoy growth)0.750.220.31-0.090.520.24
(intercept)0.90.950.200.751.150.00
Price0.9-0.130.44-0.570.300.61
Property0.9-0.070.31-0.380.240.71
Equity0.90.110.30-0.180.410.53
Leverage0.90.370.46-0.090.820.18
REER0.9-0.150.26-0.410.110.35
Real GDP CHN (yoy growth)0.90.270.230.040.500.05
1Prepared by Sally Chen and Sheheryar Malik. We thank Romain Lafarguette, Alan Feng and Prasad Ananthakrishnan for their thoughtful comments and suggestions.
2The FCI is normalized to have a zero mean over the estimated period of January 1995 to June 2018.
3Units are measured in standard deviations to standardize measures across different variables.
4Near term refers to the 1 -year horizon; medium term refers to the 3-year horizon.
5The estimation of conditional quantiles for growth forecasts is based on Adrian et al (forthcoming) and IMF (2017). The estimation of t-skew distribution parameters is based on Singer and Nelder (2009) which provides the basis from which to calculate the associated growth at risk (GaR).

Other Resources Citing This Publication