January 2017

A Bayesian analysis of Hong Kong's housing price dynamics Tommy Wu, Michael Cheng and Ken Wong ‡ Research Department Hong Kong Monetary Authority

Abstract This paper uses a Bayesian Vector-Autoregressive (BVAR) model with sign restrictions to estimate the underlying drivers of Hong Kong’s housing price dynamics in the short run. While existing studies are useful in analyzing housing valuation, little attention has been paid in the short-run dynamics. In contrast, this paper identifies short-run drivers of housing prices using structural identification with theoretical underpinnings. We find that among the shocks that we have identified, bank lending shock and housing supply shock were the main factors affecting Hong Kong’s housing prices. Low mortgage rate was also another key factor that led to the significant increase in housing prices after the global financial crisis.

JEL Classification: C32, E51, E58, R21, R31 Keywords: House Prices, Bank Credit, Macro-prudential Policies, Bayesian VAR, Sign Restrictions



Authors’ E-Mail Addresses: [email protected]; [email protected]; [email protected]. The authors would like to thank Lillian Cheung, Frank Leung, Charles Leung, and the two anonymous referees for their useful comments. The authors have also benefitted from the discussions with Simon Kwan and other participants at the Conference on Real Estate and Financial Stability co-organised by Hong Kong Institute for Monetary Research and City University of Hong Kong, Department of Economics and Finance, Global Research Unit (GRU) for their comments. The views and analysis expressed in this paper are those of the authors, and do not necessarily represent the views of the Hong Kong Monetary Authority. All errors are our own.

1.

Introduction Many economies experienced housing boom in the post-global financial crisis

(post-GFC) period. Some prime examples are the real estate markets in London, Vancouver, Sydney, and Hong Kong. The surge in housing prices has attracted the attention and created concerns among academics, market participants, and policymakers on the overheating risks in the housing market.

As such, many economies, particularly those in Asia including

Singapore and Hong Kong, have introduced macro-prudential measures to safeguard banking stability against risks stemming from the housing market. As boom-bust cycles in housing prices would pose risks not only to financial stability but also to price stability of an economy, it is important to have a good grasp of housing price dynamics to facilitate pre-emptive policymaking. In this paper, we focus on the drivers of Hong Kong’s housing price dynamics. The case of Hong Kong is of particular interest due to the city’s free market economy and full capital account convertibility. Many studies have analyzed Hong Kong’s property market, in light of its salient features and importance to the economy (e.g. Leung and Tang (2015a), Chang et al. (2013), Hui et al. (2010)).1 Nevertheless, most of these studies focus either on the transmission of financial market fluctuations to the housing market, or on the micro features using transaction-level data. This paper aims to fill the void by focusing on the macro drivers of housing prices. It is common to analyse housing price dynamics using the error correction framework (e.g. Malpezzi (1999), Capozza et al. (2004), Leung (2014), Beenstock and Felsenstein (2010)).2 In the case of Hong Kong, several studies (e.g. Leung et al. (2008), Craig and Hua (2011), and Chung (2012)) have estimated error correction models (ECM) 1

Leung and Tang (2015a) find that the initial public offerings of Chinese firms in the Hong Kong stock market and Hong Kong’s housing market can improve the prediction of each other, pointing to the role of market sentiment as the driving force. Chang et al. (2013)’s regime-switching models indicate that unexpected shock of US stock returns had the most significant effect on HK asset returns and GDP. Hui et al. (2010) use a hierarchical Bayesian approach to value Hong Kong’s residential properties with reference to its micro features, and find that their model can outperform other valuation methods that are based on average price-per-squarefeet or expert assessments. 2 Malpezzi (1999) proposes an ECM featuring housing price-to-income ratio and finds that the model can match the state-level of US data well. Capozza et al. (2004) proposes an ECM featuring some long-run equilibrium housing prices and find support for their model from US data. Leung (2014) builds a DSGE model that can produce reduced-form dynamics consistent with the ECMs proposed by Malpezzi (1999) and Capozza et al. (2004). Beenstock and Felsenstein (2010) introduce the spatial element into ECM of regional housing prices in Israel. 2

with a focus on the long-run determinants of housing prices, such as demographics, housing supply and other fundamental variables. While useful in analyzing housing valuation, the literature has paid little attention in revealing the short-run drivers of housing prices. This paper adds to the previous studies by identifying the role of different factors in affecting housing prices in the short run. We use a Bayesian Vector Autoregression (BVAR) model with sign restrictions to study the Hong Kong housing market. The approach we take is similar to that in Towbin and Weber (2015), and we modify the model specifications to suit the case of Hong Kong. We find that among the shocks that we have identified, bank lending shock and housing supply shock were the main drivers of the short-run housing price dynamics in our sample period between 1996 and 2016. Low mortgage rate was also another key factor that led to the significant increase in housing prices after the GFC. The rest of the paper is organized as follows. Section two discusses Hong Kong’s housing market development since the Asian financial crisis. Section three presents the BVAR model and discusses the sign restrictions for shock identification.

Section four

presents the empirical results. Section five concludes and discusses policy implications. 2.

Hong Kong’s Housing Market Development After a downturn between the Asian financial crisis in 1998 and the SARS outbreak

in 2003, Hong Kong’s housing market has been on a long rally for more than twelve years. Hong Kong’s housing prices have increased nearly four-fold between its trough in 2003 and its peak in 2015, despite occasional declines (Figure 1). Several factors are likely to have contributed to the robust growth of housing prices in Hong Kong. On the demand side, the solid growth in household income over the past decade and the steady growth in the number of household have continued to drive housing need (Figures 2 and 3).

3

Figure 1: Housing price index and major market events Index (1999=100) 400 + Introduction of Macro-prudential measures by HKMA Oct 2009: 1st round Aug 2010: 2nd round Nov 2010: 3rd round Jun 2011: 4th round Sep 2012: 5th round Feb 2013: 6th round Feb 2015: 7th round

350

300

Oct 2012: Introduction of Buyers' Stamp Duty & extension of Special Stamp Duty

Dec 2015: First US Fed rate hike Feb 2013: Introduction of since 2008 Double Stamp Duty

250 Jul 1997: 85,000 Building Target announced

200

Nov 2010: Introduction of Special Stamp Duty Asian Financial Crisis

150

Sep 2012: Quantitative Easing 3

European Debt Crisis began

Nov 2010: Quantitative Easing 2

Global Financial Crisis Sept 11th Terrorists Attack

100

Nov 2008: Quantitative Easing 1

Nov 2002: Suspension of regular land auction

50

0 1993

1995

1997

1999

2001

Outbreak of SARS

2003

2005

2007

2009

Feb 2010: Resumption of Land Sale Programme

2011

Dec 2014: Promulgation of Long Term Housing Strategy

2015 2016 (Jun)

2013

Sources: Rating and Valuation Department of Hong Kong (R&VD) and Hong Kong Monetary Authority (HKMA).

Figure 2: Housing price index and median household income Index (1999=100) 350

HK$ 40,000 35,000

Housing price index (lhs)

300

Figure 3: Housing price index and number of households

Median household income (rhs)

30,000

250

Index (1999=100) 350 300

No. of household (mn) 3.0

Number of households (rhs)

2.7

Housing price index (lhs)

250

25,000 200

2.4

200 20,000

150

150

2.1

15,000 100

10,000

100 1.8

50

5,000

0 93

95

97

99

01

03

05

07

09

11

13

0 15 16 (Jun)

50 0 93

95

97

99

01

03

05

07

09

11

13

1.5 15 16 (Jun)

Sources: Census and Statistics Department of Hong Kong (C&SD), R&VD, and authors’ estimates.

Under the Linked Exchange Rate System (LERS), the movement of Hong Kong dollar interest rates hinges on the US monetary policy. The ultra-low US policy rates since late 2008 have driven Hong Kong mortgage rates to historical low levels, stimulating investors to search for yields (Figure 4). As such, the rent-price ratio, which represents the return on housing, has also declined to the lowest level in recent years (Figure 5). The Hong

4

Kong Government has introduced several demand-management measures since 2010 to curb short-term investment activities. Confirmor transactions and flipping-trade have declined to low levels since then. Tight housing supply is expected to be another key driver of the uptrend in housing prices, and changes in the expected housing supply could affect short-run housing price dynamics. Back in 2002 when housing prices had fallen by more than half since its peak in 1997, the Government once suspended regular land sales in 2003. In the subsequent years, housing commencement and new housing completion declined to low levels (Figure 6). Even though the Government resumed regular land sales in 2010, housing supply was still tight until more recently, due to the time lag between housing commencement and completion. Together with the pent-up housing demand and the low-interest rate environment after the GFC, a housing demand-supply imbalance emerged, though the gap had narrowed gradually. Meanwhile, housing vacancy rate fell in tandem with the decline in housing supply, and such trend is also seen in the housing stock per household living in private housing (Figure 7). Figure 4: Housing price index and average mortgage rate Index (1999=100) 350

Figure 5: Housing price index and rent-price ratio % 14

Index (1999=100) 350

% 350

12

300

300

250

10

250

200

8

200

200

150

6

150

150

100

4

100

100

50

2

50

50

0 15 16 (Jun)

0

Housing price index (lhs) Average mortgage rate (rhs)

300

0 93

95

97

99

01

03

05

07

09

11

13

Housing price index (lhs) Rent-price ratio (rhs)

93

95

97

99

01

03

05

250

07

09

11

13

0 15 16 (Jun)

Sources: R&VD, HKMA, and authors’ estimates.

5

Figure 6: Land supply and housing commencement

Figure 7: Vacancy rate and housing stock per household living in private housing

Gross floor area ('000 m2) 2,000

Unit ('000 ) 50

1,800

Land disposals for housing (lhs)

45

1,600

Housing commencement (rhs)

40

1,400

35

1,200

30

1,000

25

800

20

600

15

400

10

200

5

0

0 94

96

98

00

02

04

06

08

10

12

14

Sources: Buildings Department and Lands Department of Hong Kong, and authors’ estimates.

% 8

Unit 1.04

7

1.02

1

6

0.98

5

0.96 4 0.94 3 2

Vacancy rate (lhs)

0.92

Housing stock per household living in pirvate housing (rhs)

0.9

1

0.88

0

0.86 96 97 98 99 00 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15

Sources: Buildings Department, R&VD, C&SD and authors’ estimates.

Since the GFC in 2008, the banking sector has faced ample global liquidity due to the unconventional monetary easing in major advanced economies. To maintain banking stability and enhance banks' risk management, the Hong Kong Monetary Authority (HKMA) has introduced seven rounds of macro-prudential measures on new mortgages since 2009. In particular, the maximum loan-to-value (LTV) ratio was lowered according to the value of the houses and applicants’ financial conditions.

While the amount of new mortgage loans

approved have not dropped much due to the increase in housing prices since 2009 (Figure 8), the number of mortgage loan applications have declined quite visibly, and the average LTV ratio had decreased from about 64% before the macro-prudential measures to around 50% in the second quarter of 2016 (Figure 9). The debt servicing ratio, which measures the monthly debt repayment of the borrower as a percentage of monthly income, also fell from 40% in 2010 to 34%. On the effectiveness of macro-prudential measures, Wong et al. (2016) suggest there is a state-dependent effect in deploying LTV policy for dampening credit growth, as the LTV policy is expected to be more effective in the case of excess credit demand when credit supply is binding, and is less effective in the case of excess credit supply. The authors then showed that banks indeed saw excess credit demand than excess credit supply after the first introduction of LTV policy in October 2009 up to the end of sample period in 2012. As such, the macro-prudential measures have been effectively transmitted to the market and have strengthened the resilience of banks by reducing the leverage of mortgage borrowers.

6

Meanwhile, the performance of Hong Kong’s property market is widely believed to be influenced by market sentiment, given Hong Kong’s status as an international financial centre with free capital mobility. As stock market returns often reflect asset market sentiment, one can see that sharp falls in housing prices were often associated with plunges in the stock market (Figure 10). Figure 8: Loan-to-value ratio and the amount of new mortgage loans approved HKD bn

Figure 9: Loan-to-value ratio and the number of mortgage applications

% 80

Number of cases ('000)

80

80

% 80

70

70

70

70

60

60

60

60

50

50

40

40

30

30

30

30

20

20

20

20

10

10

10

10

0

0

50

Mortgage loan applications (lhs)

Mortgage loan approvals (lhs) 40

Loan-to-value ratio (rhs)

0 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 (Jun)

Loan-to-value ratio (rhs)

50 40

0 01 02 03 04 05 06 07 08 09 10 11 12 13 14 15 16 (Jun)

Source: HKMA.

Source: HKMA.

Figure 10: Housing price index and the Hang Seng Index Index (1999=100) 600

Index 35,000

Housing price index (lhs)

500

30,000

Heng Seng Index (rhs)

25,000

400

20,000

300 15,000 200

10,000

100

5,000

0 93

95

97

99

01

03

05

07

09

11

13

0 15 16 (Jun)

Sources: R&VD and Hong Kong Exchanges and Clearing Limited (HKEx).

Relating the stylized facts to the real estate literature, Leung et al. (2002) use transaction-level data between 1991 and 1998 and show that there was a strong, positive correlation between real housing price and transaction volume in Hong Kong. There are two competing theories to explain such a strong correlation which is also observed in other

7

housing markets such as the US market. Stein (1995) suggests the downpayment effect, in which a rise in housing price increases the wealth of homeowners and relaxes the downpayment constraint for buying new homes. This in turn increases trading volume. Alternatively, in Berkovec and Goodman (1996), a negative housing demand shock reduces the number of transactions, and in a search and matching environment sellers would have the incentive to speed up the selling process by reducing their selling prices. Hence, transaction volume and housing prices have strong positive correlation. To give us a sense as to whether the positive correlation between housing prices and transaction volume in Hong Kong continues to hold, we compute in Table 1 the correlations using frequency-filtered aggregate housing price and transaction volume data at the business cycle frequency between 2000 and 2016. The correlations are significant for the pre-GFC period. For the full sample period which includes the post-GFC period, the correlations remain significant, but the effect appears to have weakened. This suggests that the existing theories of price-transaction volume correlation may not have very well captured Hong Kong housing market as circumstances evolve over time.

As such, we need an

empirical model which considers multiple underlying factors or shocks to explain Hong Kong’s housing price dynamics in the short run, as well as including a time-varying component to capture structural changes. Table 1: Correlation between housing price and transaction Periods Sample before GFC (Jan 2000-Sep 2008) Full sample (Jan 2000-Jun 2016)

Correlation between real housing price and transactions in Baxter-King Filter Christiano-Fitzgerald Filter 0.7013 0.5710 (8.7447) (7.0597) 0.5293 0.2168 (7.6146) (3.1092)

Note: t-statistics are in parenthesis. The two columns show the results based on the Baxter-King filtered data and the Christiano-Fitzgerald filtered data, respectively. The filtered data come from the raw data filtered by the respective methods with periodicities between six and thirty-two quarters.

3.

Empirical Model, Data and Identification of Shocks To examine the importance of various factors or shocks in affecting housing price

dynamics in Hong Kong, we estimate a Bayesian VAR (BVAR) model with sign restrictions.

8

There are different approaches to study housing price dynamics, including the use of regimeswitching models (e.g. Chang el al. (2013)), models with time-varying coefficients (e.g. Guirguis et al. (2004)), and large-scale models with an extensive number of variables (e.g. Das et al. (2009) and Stadelmann (2010)).3 However, most of these studies focus either on the forecasting ability of the models, or from an asset pricing perspective. In contrast, the focus of this paper is to identify driving factors or shocks that can affect short-run housing price dynamics, using structural identification with theoretical underpinnings wherever plausible. The use of large-scale models may not be suitable for such purpose. Moreover, even though the above price-trading volume correlation analysis and the results in Leung et al. (2013) both suggest the possibility of a structural change in Hong Kong housing market after a financial crisis, the use of non-linear models such as regime-switching models would limit the number of variables to be included. While the use of time-varyingparameter VAR (TVP-VAR) may control for the structural change, it also suffers from criticism that the invariance of parameters cannot be rejected, such as the literature by Bernanke and Mihov (1998), Sims (2001) and Stock (2001). To circumvent this problem, we include a time-varying deterministic component into the BVAR model, following Clements and Hendry (1996). In our estimation, we consider the following BVAR model: 𝐲𝒕 = ∑𝐿𝑖=1 𝐀 𝒊 𝐲𝒕−𝒊 + 𝐞𝒕

with

𝐞𝑡 ~ 𝚴(𝟎, 𝚺) ∀ t = 1, …, T

(1)

where 𝒚𝒕 is a vector of explanatory variables at time t. L is the lag length and 𝐀 𝒊 denotes the coefficient matrix for the ith lag. 𝐞𝒕 is a reduced-form error term with variancecovariance decomposition 𝚺. The vector of explanatory variables is given by: 𝒚𝒕 = (Δ𝑃𝑡 𝐻𝑡 𝑉𝑡 𝑖𝑡 𝑀𝑡 Δ𝑌𝑡 𝑅𝑡 /𝑃𝑡 Δ𝑆𝑡 1)𝑇 .

(2)

3

Chang et al. (2013) compare a medium-size linear VAR with a small regime-switching VAR to study the relationship between US and Hong Kong’s financial variables with Hong Kong’s real GDP and asset returns, including housing returns. Guirguis et al. (2004) use a time-varying-parameter model to forecast US housing prices. Das et al. (2009) use small- and large-scale BVAR model and dynamic factor model to compare competing models’ forecasting abilities using South African data. Stadelmann (2010) uses Bayesian Model Averaging approach to select variables from a large variable set to select variables that can affect housing prices. 9

Δ𝑃𝑡 is the monthly growth of real housing prices as measured by the log difference of the housing price index of Hong Kong’s Rating and Valuation Department (R&VD), deflated by the underlying CCPI inflation rate. 𝐻𝑡 denotes housing commencement, which is the log of units of buildings commenced from the Buildings Department of Hong Kong. We take the 3-year moving average to reflect the timing of between housing commencement and completion.4 𝑉𝑡 denotes housing vacancy. Instead of using the R&VD vacancy data series which is available only in annual frequency, we proxy vacancy by the monthly housing stock per household living in private housing, given the close relationship between the two as shown in Figure 7. The data on housing stock is from the R&VD and the data on household living in private housing is from the General Household Survey data compiled by the Census and Statistics Department (C&SD) of Hong Kong. The real mortgage interest rate 𝑖𝑡 is the difference between HKMA’s internal estimate of the average mortgage rate and the inflation rate. 𝑀𝑡 is the log of value of new mortgage loans approved by banks in real terms. The data source of nominal mortgage rate and mortgage loans approved is the HKMA’s Residential Mortgage Survey. Δ𝑌𝑡 is the log difference of total household real income in Hong Kong, constructed using the data on median household income and the number of domestic household from the Census and Statistics Department, adjusted for inflation. 𝑅𝑡 /𝑃𝑡 is the rent-price ratio constructed using nominal housing rent and price data from the R&VD. Δ𝑆𝑡 is the monthly stock market return as measured by the log difference of the Hang Seng Index (HSI), adjusted for inflation. These variables in nominal terms have been graphed in Section 2. Finally, the unit vector captures the time-varying deterministic component, or the “noshock” path. The inclusion of housing commencement helps identify expected housing supply. Vacancy contains information that can help disentangle housing demand and supply, with an increase in housing supply increasing the vacancy rate, while an increase in housing demand would reduce it. Mortgage interest rate and mortgage approval by banks reflect bank lending condition on both price and quantity terms of bank loans. Total household income captures information on households’ budget constraint and housing needs. The inclusion of the rentprice ratio helps control for the long-run housing price dynamics given the expected cointegrating relationship between housing prices and rents. Finally, monthly stock return 4

Data on housing commencement only goes back to April 1995, which means that by taking 3-year moving average the series used for estimation would start only in April 1998. Since it takes about three years for housing commencement to completion, we use housing completion in the 3 years ahead to construct the data on 3-year moving average housing commencement for the months before April 1998. 10

captures information on financial market sentiment which affects the general housing market performance, particularly during turning points. The sample covers the period from January 1996 to June 2016. The BVAR is estimated with the Normal Wishart prior implemented using dummy observations, in which extra “data" are added to the sample in order to express prior beliefs about the parameters. We draw the posteriors using Gibbs sampling. The hyper-parameters are based on Bańbura et al. (2008). The hyper-parameter (λ) that governs the tightness of priors of the lag terms is set to 0.262 for a small-to-medium size BVAR model. The hyper-parameter (τ) that controls for the degree of shrinkage is 10 times λ. The hyper-parameter on the constant terms is normalized to 1. The model is estimated with a lag length of 3 months.5 Sign restrictions have been used to identify structural shocks following the work of Canova and De Nicolo (2002) and Uhlig (2005) and others in the literature. Comparing to parametric restrictions, sign restrictions could be less stringent (Lütkepohl and Netšunajev (2014)) and more appealing as variables are usually determined in a simultaneous manner (Fry and Pagan (2011)). In particular, Towbin and Weber (2015) applied sign restriction identification schemes to a BVAR model on US housing prices.

We modify the sign

restrictions in Towbin and Weber (2015) according to the selected variables in our model and the specific features of Hong Kong housing market. We estimate the model with 10,000 iterations and 5,000 burn-ins, in which we accept 5,000 models with impulse response functions being consistent with the specified sign restrictions for all shocks.6 We identify five shocks in our model, namely housing supply shock, housing demand shock, mortgage rate shock, market sentiment shock and bank lending shock. Table 2 summarises the corresponding sign restrictions, and all signs are restricted in the way that 5

Estimation using a lag length of 6 months or more does not alter the results significantly. That said, the sign restriction approach has received a number of criticisms. Kilian and Murphy (2012) and Lütkepohl and Netšunajev (2014) argue that sign restrictions do not identify shocks uniquely but allow a range of admissible shocks. Fisher et al. (2016) also note that sign restrictions may not be able to identify shocks as successfully as claimed. To address the issue, Kilian and Murphy (2012) suggest imposing additional bound on the impulse responses to narrow the range of admissible shocks, while Lütkepohl and Netšunajev (2014) use change in volatility to provide additional identifying information and check the validity of sign restrictions. As relevant empirical evidences are limited, we refrain from imposing any additional bound on the impulse response functions. Given the length of our sample period and the number of variables in our model, it would also be difficult to allow for Markov-switching in the error terms. On the other hand, Arias et al. (2014) criticize that the implementation of sign restrictions using the penalty function approach would introduce additional sign restrictions on variables that are seemingly unrestricted. Hence, they propose a new algorithm that can address such issue. We do not use the penalty function approach in this paper noting that there is no natural choice for such function. 6

11

the initial responses of real housing prices to the respective shocks are positive. Also, we assume that shocks are instantaneous, except mortgage rate shock and bank lending shock, which would affect housing commencement for 3 periods, the same as the lag length of the BVAR model. The latter assumption is made because it is unlikely that a shock that is only instantaneous would affect housing commencement, and hence we extend the period under impact to make the assumption more realistic. Table 2: Shock identifications Housing price

Housing commencment

Vacancy

Mortgage rate

New mortgage loans approved

Household income

Housing supply (-)

+

-

-

Housing demand (+)

+

-

+

-

+

Mortgage rate (-)

+

Market sentiment (+)

+

-

-

+

+

Bank lending (+)

+

Rent-price ratio

HSI

Shocks:

+ +

-

-

+

+

A negative housing supply shock represents a reduction in the expected new housing units available due to an unexpected decline in housing commencement, and this would pose upward pressures on housing prices. Housing supply shock reflects changes in the primary market supply are largely due to changes in land supply, land regulations, as well as property developers’ business strategies (e.g. Leung and Tang (2015b)). As changes in mortgage interest rate and bank loan quantity could also affect property developers’ decisions on housing commencement, these factors are separately identified as mortgage rate shocks or bank lending shocks which we will discuss below. A negative housing supply shock would also drive down vacancy as it would shift the housing supply curve inward.7 There is also an associated drop in total household income stemming from the decline in residential investment in the primary market, much like the income approach and the expenditure approach which are mirror images of each other in national income accounting. A positive housing demand shock comes from an increase in household formation or income growth, either of which would raise total household income and push up housing 7

A negative housing supply shock could also affect secondary-market activities by changing homeowners’ willingness to sell as they would raise their offer prices in a search and matching environment. For instance, in a search-theoretic housing model (such as Wheaton (1990) and Head et al. (2014)), one can imagine that an increase in new housing supply can affect housing market tightness and the search intensity of home sellers in the secondary market. Since we do not include any variable that can directly capture buyers and sellers behaviour in the secondary market, we focus on the primary-market channel when interpreting the impact of a housing supply shock. 12

prices. We also assume that vacancy would decline as the shock would shift the housing demand curve outward. A positive demand shock would also raise the real mortgage rate due to increased mortgage loan demand. In this regard, even though movements of Hong Kong dollar short-term interest rates largely follow those of the US interest rates under the LERS, banks in Hong Kong could still determine the risk premium and interest rate spreads charged on their mortgage loans. To help distinguish a shock to housing demand from an easing in bank credit, we follow Towbin and Weber (2015) and assume that new mortgage loans approved would fall in response to a positive housing demand shock. A negative mortgage rate shock reduces the real mortgage rate and increases housing prices. A mortgage rate shock can arise from a change in either the US monetary policy, risk premium due to banks’ pricing of risks, or the term structure of interest rate. Lower mortgage rate would increase housing demand and thus boost housing prices, and one can think of this as a movement along a downward-sloping housing demand curve. Although we assume housing commencement to be largely affected by land supply and policies, a lower mortgage rate, which could reflect lower interest rates in general, would reduce property developers’ costs of borrowing and hence increase housing commencement. The increase in both housing demand and supply means that the impact of such shock on vacancy is indeterminate and depends on the relative sensitivities to the mortgage rate. We also restrict a decline in the mortgage rate to be associated with a decrease in new mortgage loans approved in order to disentangle a mortgage rate shock from a bank lending shock. Meanwhile, the performance of Hong Kong’s property market is widely believed to be influenced by market sentiment, given Hong Kong’s status as an international financial centre with free capital mobility. Leung and Tang (2015a)’s analysis of Hong Kong’s IPO activities and housing market also point to the role of market sentiment as the driving force in the asset markets. Hence, we identified a positive market sentiment shock in the BVAR system, in which a positive shock pushes HSI and housing prices upward. The increase in housing prices would widen any housing demand-supply gap and increase vacancy. We also assume that the real mortgage rate would rise due to higher demand for mortgage credit, while the new mortgage loans approved would fall. The latter assumption helps distinguish bank lending shocks from banks’ perspectives on housing prices which should be reflected in the market sentiment shock instead. We do not assume that a market sentiment shock would

13

change housing commencement as it is mostly affected by Government policies and property developers’ decisions. A positive bank lending shock captures the non-mortgage-rate component of banks’ mortgage loan, and is mainly captured through a positive shock to new mortgage loans approved which could result from a mix of changes. Ideally, the shock should capture influences from the credit supply side alone, which include shocks to banks’ credit stance and the effects of introducing new bank regulations or macro-prudential measures. Theoretically, however, bank lending is affected by both the credit demand and credit supply side, and it is difficult to disentangle the two within the current BVAR model without modeling a structure similar to that in Wong et al. (2016). As such, bank lending shocks should be interpreted as shocks arising from either the credit supply and demand side, or both. As discussed above, the sign restrictions for housing demand, mortgage rate, and market sentiment shocks orthogonalize the effects of these shocks from that of bank lending shock, hence helping identify the latter shock.8 Generally speaking, a positive bank lending shock would support housing prices as more mortgage loans are being made. We also assume that the shock would boost bank lending to property developers which would increase housing commencement. On the other hand, macro-prudential policies would be captured by negative bank lending shocks. The role of housing as collateral and how it is related to bank loans have been extensively discussed in the literature.9 4.

Results We first present the impulse response functions of the identified shocks generated

from the estimated BVAR model. We then illustrate the breakdown of Hong Kong’s housing prices by the various shocks through forecast error variance decomposition and historical decompositions. Figure 10 shows the impulse response functions of the seven variables with respect to the five shocks that we have identified using sign restrictions described above. Instead of reporting the pointwise median which has been criticized in the literature (e.g. Fry and Pagan (2011) and Inoue and Kilian (2013)), we report the pointwise mean as the main 8

While the negative sign restriction of a positive housing demand shock on new mortgage loans approved may help separate part of the demand for credit from the supply of credit, it does not directly pin down credit supply as changes in mortgage approvals could be due to many factors. 9 The literature includes Kiyotaki and Moore (1997, 2002), Ortalo-Magné and Rady (2006), Chen and Wang (2007), Chen and Leung (2008), Jin et al. (2012), among others. 14

summary measure, and the pointwise 16th and 84th percentiles as the error bands. 10 The shocks are normalized to one standard deviation, and the directions of the impulse responses follow from the sign restrictions, except for the cases where no restrictions have been imposed. Figure 10: Impulse response functions

Note: The solid, blue lines are the pointwise mean of accepted impulse functions. The dotted, black lines are the 16th and 84th percentile error bands. Source: Authors’ estimates.

For the impulse response functions in which sign restrictions have been imposed, the initial responses followed the sign restrictions as described in Table 2 and are mostly significant. There are few results worth highlighting. First, both a negative housing supply 10

Pointwise median could mix the response from different models, as the median at each horizon doesn’t necessarily come from the same model. Ouliaris and Pagan (2016) and Baumeister and Hamilton (2015) also note that the distribution of impulse responses would depend on the way in which these quantities are generated, and hence quantities such as the median will change as the generation method changes. To address the shortcoming of pointwise median, Fry and Pagan (2011) propose the median target method which uses the impulse responses of a single model that minimize the distance to the median responses. However, Canova and Paustian (2011)’s simulation studies indicate that the pointwise median performs better than the Fry and Pagan (2011)’s median target method. Inoue and Kilian (2013) also argue that the median is not a defined concept in multivariate distributions. Acknowledging the caveats involved, we report the pointwise mean in this paper. 15

shock and a positive housing demand shock would increase real housing prices temporarily but would reduce vacancy by a similar magnitude. Second, a negative mortgage rate shock would have a significant medium-run impact on the rent-price ratio by reducing the user cost, even though the shock would only affect real housing prices at the very short run.11 Third, a positive market sentiment shock would increase vacancy. Finally, a positive bank lending shock would significantly reduce the rent-price ratio. Figure 11 shows the results on the forecast error variance decomposition of real housing prices, illustrating the relative importance of different shocks that have been driving Hong Kong’s housing price variations since 1996. These are also constructed using the pointwise mean for the full sample period. Bank lending shocks account for 19% of housing price variation at the four-year horizon (48 months). 12 What this result implies is that, for example, exogenous changes in banking liquidity or business strategy could affect bank lending through the bank credit supply channel.

The housing supply shock is another

important contributor to housing price variations, accounting for 13% at the four-year horizon.13 Given the tight, inelastic housing supply condition in Hong Kong, it is expected that housing supply shocks would account for a significant portion of housing price variation. It is also because of the inelastic housing supply that the price effect of a bank lending shock is stronger than would otherwise be. Shocks to mortgage rate account for 10% of housing price variation, while market sentiment shocks and housing demand shocks account for about 8-9% of the variation. The relatively small contribution from housing demand shock may be somewhat surprising at the first glance given the robust increase in the number of households in Hong Kong. However, short-run demand due to housing needs may not fluctuate a lot and hence not identified by our identification scheme, even though housing demand could be an important long-run driver of housing prices.

11

The concept of user cost comes from Poterba (1984), followed by more recent studies such as Diaz et al. (2008). Specifically, rental, which reflects the marginal value of services generated by housing, is equal to the user cost of housing in equilibrium. Chow and Wong (2003) find empirically a positive relationship between the user cost of housing capital and housing rents for the case of Hong Kong. 12 Identifying bank lending shock using the number instead of the value of new mortgage loans approved for yields similar results. We also used the loan-to-value (LTV) ratio for new mortgages for robustness check. While the results were similar, the analysis would not cover the upcycle prior to the Asian financial crisis given that the sample of LTV ratio starts in June 1998. 13 As a robustness check, identifying the housing supply shock using housing completion for the whole sample period instead of housing commencement yields similar results. 16

In sum, among the variations that we can explain which sum up to about 60% of total variations, bank credit conditions and housing supply are the major drivers of Hong Kong’s housing price variation over the two-decade period. This result arises naturally given that Hong Kong is a small open economy where capital flows freely and that Hong Kong’s housing supply is inelastic. Since the impacts of shocks could vary quite substantially across boom-bust cycles, we break down below the contributions of shocks to real housing prices by episodes. Figure 11: Forecast error variance decomposition of housing prices

Source: Authors’ estimates.

Figure 12 shows the historical decomposition of the monthly changes in real housing prices constructed using the pointwise mean.

The decomposition breaks down

housing price deviation from a deterministic path (i.e. without any shock) by the contribution of each identified shock. Similar to the variance decomposition analysis, bank lending shocks and housing supply shocks accounted for a lot of the swings in housing price growth throughout the two decades. Table 3 shows the average contribution of shocks to the level of real housing prices across boom-bust cycles, excluding the deterministic component estimated by the model. First, recognizing that the five shocks account for around 60% of real housing price movements, housing supply shock has always been one of the key contributors across different cycles among the identified shocks, with contributions staying at double-digit levels in most periods. Second, the contribution of housing demand shocks is around 10% on

17

average across different cycles. Even though housing demand is the least important factor in accounting for housing price variations as discussed above, it is the third most important factor among the identified shocks in supporting the level of housing prices. Third, the contribution of mortgage rate shock has risen from single-digit level to 13% after the GFC. Given that the shocks are “surprises” that are not part of a systematic response as implied by the BVAR model, mortgage rate shocks in this period can be interpreted as the decline in the interest rate spread and risk premium between the benchmark interest rates and the actual mortgage rate as a result of international spillovers of unconventional monetary policy (e.g. Bauer and Diez de los Rios (2012), Chen et al. (2012), Kandrac and Schlusche (2016)). Figure 12: Historical decomposition of housing prices (monthly changes) ppt 10

mom % 10

8

8

6

6

4

4

2

2

0

0

-2

-2

-4

-4

-6

-6

-8

-8 Others (lhs) Housing demand (lhs) Market sentiment (lhs) Housing price growth (rhs)

-10 -12 -14 1996

1998

2000

2002

2004

2006

2008

Housing supply (lhs) Mortgage rate (lhs) Bank lending (lhs)

-10 -12 -14

2010

2012

2014

2016 (Jun)

Source: Authors’ estimates.

Table 3: Average contribution of shocks across boom-bust cycles (%) – excluding deterministic component Housing supply Housing demand

Mortgage rate Market sentiment

Bank lending

Others

Contribution to boom: Apr 1996-Oct 1997

20.5

15.9

11.1

-3.7

29.3

27.0

Aug 2003-Jun 2008

6.8

13.9

5.6

12.3

21.7

39.7

12.7

9.1

12.5

9.7

15.4

40.6

Nov 1997-Jul 2003

14.0

11.6

9.1

5.9

18.5

40.9

Jul 2008-Dec 2008

17.4

1.9

3.2

9.8

23.6

44.0

Jan 2009-Jun 2016 (present) Contribution to bust:

Note: To calculate the average contributions for a specified time window, we first compute the cumulative contributions at each point in time with the same starting period, and then we take the average of the contributions of all the time periods within the specified time window.

18

Source: Authors’ estimates.

The contribution of market sentiment shock is sizeable in some periods. The contribution of bank lending shocks was on average around 23% before the GFC, but it had declined to a certain extent in the post-GFC period. While the decline was partly substituted by the rise in contributions from shocks to mortgage rate, the macro-prudential and demandmanagement measures introduced since 2009 might possibly have a role to play. Figure 13 illustrates the cumulative historical decompositions of housing prices in the Asian financial crisis–SARS episode, which is different from the average historical decompositions in Table 3. The cumulative view helps visualize the contribution of shocks within a boom-bust cycle, whereas the averages summarize shock contributions across boombust cycles. Among the identified shocks, shocks to bank lending, housing supply, and housing demand were the main contributors to housing price changes during the Asian financial crisis–SARS episode.

In particular, housing supply shock’s contribution had

changed from positive to negative after 1998. The contribution of housing demand shock had become more prominent in the down-cycle after 2000. Figure 13: Cumulative historical decomposition of housing prices (the Asian financial crisis – SARS episode) Contribution 40 20

Level in log terms Others (lhs) 40 Housing supply (lhs) Housing demand (lhs) Mortgage rate (lhs) 20 Market sentiment (lhs) Bank lending (lhs) Log housing price minus deterministic (rhs)

0

0

-20

-20

-40

-40

-60

-60

-80 1996 1997 1998 Source: Authors’ estimates.

-80 1999

2000

2001

2002

2003 (Jul)

19

Figure 14 shows the cumulative historical decomposition of housing prices in the post-GFC episode. Among the identified shocks, housing supply shock continued to be a key driver of housing price changes, while the contribution of mortgage rate shock was larger and persistent compared with that in the Asian financial crisis episode. The role of bank lending shock also remained significant, but as shown in Table 3, its impact had diminished compared with previous cycles. Figure 14: Cumulative historical decomposition of housing prices (the post-GFC episode) Contribution 120 100 80

Level in log terms 120

Others (lhs) Housing supply (lhs) Housing demand (lhs) Mortgage rate (lhs) Market sentiment (lhs) Bank lending (lhs) Log housing price minus deterministic (rhs)

100 80

60

60

40

40

20

20

0

0

-20 2009 2010 2011 Source: Authors’ estimates.

5.

2012

2013

2014

2015

-20 2016 (Jun)

Concluding Remarks Given the strong influences of housing prices on Hong Kong’s macro-financial

conditions, this paper dissects the short-run dynamics of housing prices by different factors. Doing so not only complements earlier studies which focus mainly on the long-run housing prices, but also helps provide a timely guide to housing market development. Using a Bayesian Vector Autoregression (BVAR) model with sign restrictions, we find that among the shocks that we have identified, bank lending shock and housing supply shock were the main drivers of housing prices in a sample period of 1996-2016. Low mortgage rate was also another key factor that led to the significant increase in housing prices after the global financial crisis.

20

While remained significant, the role of bank lending shock has diminished during the post-global financial crisis era.

The decline was partly substituted by the rise in

contributions from shocks to mortgage rate, while the macro-prudential measures introduced since 2009 might possibly have a role to play. While being anecdotal in nature, the latter offers a suggestion for future studies that, to assess more comprehensively the overall impact of macroprudential measures on housing prices, there may be a need to take into account the indirect impact through banks’ credit stance. This is indeed relevant given that the literature usually examines only the direct impact and finds mixed evidence on such issue. Given that housing supply shock seem to be one of the key drivers of housing price dynamics in Hong Kong, a key part of the solution to restrain housing price increase, as discussed in Craig and Hua (2011), should involve raising land supply. Even though policies on land supply or housing supply are long-term policies, they can affect short-run housing price dynamics through influencing the expectation of new housing units available in the primary market. Similar conclusion has also been drawn in studies of land supply policies and housing prices dynamics (e.g. Li et al. (2015)). Hence, our results shed some light on the importance of anchoring private sector expectation of land and housing policies, as well as highlighting the usefulness of monitoring and analyzing such expectation for research and policy analysis.

21

References Arias, Jonas, Juan Rubio-Ramirez, and Daniel Waggoner (2014), “Inference Based on SVARs Identified with Sign and Zero Restrictions: Theory and Applications”, Dynare working group. Bańbura, Marta, Domenico Giannone and Lucrezia Reichlin (2008), “Large Bayesian VARs”, European Central Bank Working Paper Series No. 966. Bauer, Gregory H. and Antonio Diez de los Rios (2012), “Global Risk Premiums and the Transmission of Monetary Policy”, Bank of Canada Review, Summer, 12-20. Baumeister, Christiane and James D. Hamilton (2015), “Sign Restrictions, Structural Vector Autoregressions, and Useful Prior Information”, Econometrica, 83, 5, 1963-1999. Beenstock, Michael and Daniel Felsenstein (2010), “Spatial Error Correction and Cointegration in Nonstationary Panel Data: Regional House Prices in Israel”, Journal of Geographical Systems, 12, 2, 189‐206. Berkovec, J. and J. Goodman (1996), “Turnover as a Measure of Demand for Existing Homes”, Real Estate Economics, 24, 421-440. Bernanke, Ben S. and Ilian Mihov (1998), “Measuring monetary policy”, Quarterly Journal of Economics, 113, 3, 869-902. Canova, Fabio and Gianni De Nicolo (2002), “Monetary Disturbances Matter for Business Fluctuations in the G-7”, Journal of Monetary Economics, 49, 6, 1131-1159. Canova, Fabio and Matthias Paustian (2011), “Business Cycle Measurement with Some Theory”, Journal of Monetary Economics, 58, 4, 345-361. Capozza, Dennis, Patric H. Hendershott, and Charlotte Mack (2004), “An Anatomy of Price Dynamics in Illiquid Markets: Analysis and Evidence from Local Housing Markets”, Real Estate Economics, 32, 1, 1‐32. Chang, Kuang-Liang, Nan-Kuang Chen, and Charles Ka Yui Leung (2013), “In the Shadow of the United States: The International Transmission Effect of Asset Returns”, Pacific Economic Review, 18, 1, 1‐40.

22

Chen, Qianying, Andrew Filardo, Dong He, and Feng Zhu (2012), “International Spillovers of Central Bank Balance Sheet Policies”, BIS Papers No 66. Chen, Nan-Kuang and Charles Ka Yui Leung (2008), “Asset Price Spillover, Collateral and Crises: with an Application to Property Market Policy”, Journal of Real Estate Finance and Economics, 37, 4, 351‐385. Chen, Nan-Kuang and Hung-Jen Wang (2007), “The Procyclical Leverage Effect of Collateral Value on Bank Loans—Evidence from the Transaction Data of Taiwan”, Economic Inquiry, 45, 2, 395-406. Chow, Ying-Foon and Nelson Wong (2003), “Property Value, User Cost, and Rent: An Investigation of the Residential Property Market in Hong Kong”, Journal of Business & Economics Research, 1, 9, 9-17. Chung, Kenneth Hoi-Ki (2012), “Determinants of Residential Property Prices in Hong Kong: A Cointegration Analysis”, International Research Journal of Finance and Economics, 96, 55-62. Clements, Michael P. and David F. Hendry (1996), “Intercept Corrections and Structural Change”, Journal of Applied Econometrics, 11, 475-494. Craig, R. Sean and Changchun Hua (2011), “Determinants of Property Prices in Hong Kong SAR: Implications for Policy”, IMF Working Paper WP/11/277. Das, Sonali, Rangan Gupta, and Alain Kabundi (2009), “Could we have predicted the recent downturn in the South African housing market?”, Journal of Housing Economics, 18, 4, 325‐335. Díaz, Antonia and Maria Jose Luengo-Prado (2008), “On the User Cost and Homeownership”, Review of Economic Dynamics, 11, 3, 584-613. Fisher, Lance, Hyeon-Seung Huh, and Adrian Pagan (2016), “Econometric Methods for Modelling Systems with a Mixture of I(0) and I(1) Variables”, Journal of Applied Econometrics, 31, 892‐911. Fry, Renée and Adrian Pagan (2011), “Sign Restrictions in Structural Vector Autoregressions: A Critical Review”, Journal of Economic Literature, 49, 4, 938-60.

23

Glaeser, Edward L., Joseph Gyourko and Albert Saiz (2008), “Housing Supply and Housing Bubbles”, Journal of Urban Economics, 64, 2, 198-217. Guirguis, Hany S., Christos I. Giannikos, and Randy I. Anderson (2004), “The US Housing Market: Asset Pricing Forecasts Using Time Varying Coefficients”, Journal of Real Estate Finance and Economics, 30, 1, 33‐53. Head, Allen, Lloyd-Ellis, Huw and Hongfei Sun (2014), “Search, Liquidity, and the Dynamics of House Prices and Construction”, American Economic Review, 104, 4, 1172-1210. Huang, Haifang and Yao Tang (2012), “Residential Land Use Regulation and the US Housing Price Cycle between 2000 and 2009”, Journal of Urban Economics, 71, 1, 9399. Hui, Sam K., Alvin Cheung, and Jimmy Pang (2010), “A Hierarchical Bayesian Approach for Residential Property Valuation: Application to Hong Kong Housing Market”, International Real Estate Review, 13, 1, 1-29. Inoue, Atsushi and Lutz Kilian (2013), “Inference on Impulse Response Functions in Structural VAR models”, Journal of Econometrics, 177, 1, 1-13. Jin, Yi, Charles Ka Yui Leung, and Zhixiong Zeng (2012), “Real Estate, the External Finance Premium and Business Investment: A Quantitative Dynamic General Equilibrium Analysis”, Real Estate Economics, 40, 1, 167‐195. Kandrac, John and Bernd Schlusche (2016), “Quantitative Easing and Bank Risk Taking: Evidence from Lending”, Federal Reserve Board manuscript. Kilian, Lutz and Daniel P. Murphy (2012), “Why Agnostic Sign Restrictions Are Not Enough: Understanding The Dynamics Of Oil Market VAR Models”, Journal of the European Economic Association, 10, 5, 1166‐1188. Kiyotaki, Nobuhiro and John Moore (1997), “Credit Cycles”, Journal of Political Economy, 105, 2, 211‐248. Kiyotaki, Nobuhiro and John Moore (2002), “Balance ‐ Sheet Contagion”, American Economic Review, 92, 2, 46‐50.

24

Leung, Charles Ka Yui (2014), “Error Correction Dynamics of House Prices: An Equilibrium Benchmark”, Journal of Housing Economics, 25, 75‐95. Leung, Charles Ka Yui, Patrick Wai Yin Cheung, and Edward Chi Ho Tang (2013), “Financial Crisis and the Co‐movements of Housing Sub‐markets: Do relationships change after a crisis?”, International Real Estate Review, 16, 1, 68‐118. Leung, Charles Ka Yui, Garion C.K. Lau, and Youngman C. F. Leong (2002), “Testing Alternative Theories of the Property Price‐Trading Volume Correlation”, Journal of Real Estate Research, 23, 3, 253‐264. Leung, Charles Ka Yui and Edward Chi Ho Tang (2015a), “Speculating China Economic Growth through Hong Kong? Evidence from Stock Market IPOs and Real Estate Markets”, International Real Estate Review, 18, 1, 45‐87. Leung, Charles Ka Yui and Edward Chi Ho Tang (2015b), “Availability, Affordability and Volatility: the case of Hong Kong Housing Market”, International Real Rstate Review, 18, 3, 383-428. Leung, Frank, Kevin Chow, and Gaofeng Han (2008), “Long-Term and Short-Term Determinants of Property Prices in Hong Kong,” HKMA Working Paper 15/2008. Li, Ling-Hin, Siu Kei Kelvin Wong, and Ka Shing Cheung (2015), “Land supply and housing prices in Hong Kong: The political economy of urban land policy”, Environment and Planning C: Government and Policy, 34, 5, 981-998. Lütkepohl, Helmut and Aleksei NetŠunajev (2014), “Disentangling Demand and Supply Shocks in the Crude Oil Market”, Journal of Applied Econometrics, 29, 3, 479-496. Malpezzi, Stephen (1999), “A Simple Error Correction Model of House Prices”, Journal of Housing Economics, 8, 1, 27‐62. Ortalo ‐ Magné, François and Sven Rady (2006), “Housing Market Dynamics: On the Contribution of Income Shocks and Credit Constraints”, Review of Economic Studies, 73, 2, 459-485.

25

Ouliaris, Sam and Adrian Pagan (2016), “A method for working with sign restrictions in structural equation modelling”, Oxford Bulletin of Economics and Statistics, 78, 5, 605‐ 622. Poterba, James M. (1984), “Tax Subsidies to Owner-Occupied Housing: An Asset-Market Approach”, Quarterly Journal of Economics, 99, 4, 729-752. Sims, Christopher A. (2001), “Comments on Cogley and Sargent’s ‘Evolving Post World War II U.S. Inflation Dynamics’”, NBER Macroeconomics Annual, 16, 373-379. Stadelmann, David (2010), “Which Factors Capitalize into House Prices? A Bayesian Averaging Approach”, Journal of Housing Economics, 19, 3, 180‐204. Stein, Jeremy C. (1995), “Prices and Trading Volume in the Housing Market: A Model with Downpayment Effects”, Quarterly Journal of Economics, 110, 379-405. Stock, James H. (2001), “Discussion of Cogley and Sargent ‘Evolving Post World War II U.S. Inflation Dynamics’”, NBER Macroeconomics Annual, 16, 379-387. Towbin, Pascal and Sebastian Weber (2015), “Price Expectations and the U.S. Housing Boom”, IMF Working Paper WP/15/182. Transport and Housing Bureau (2014), Long Term Housing Strategy, The Government of the Hong Kong Special Administrative Region, Hong Kong. Uhlig, Harald (2005), “What are the Effects of Monetary Policy on Output? Results from an Agnostic Identification Procedure”, Journal of Monetary Economics, 52, 2, 381-419. Wheaton, William C., (1990), “Vacancy, Search, and Prices in a Housing Market Matching Model”, Journal of Political Economy, 98, 6, 1270-1292. Wong, Eric, Andrew Tsang, and Steven Kong (2016), “How Does Loan-To-Value Policy Strengthen Resilience of Banks to Property Price Shocks –Evidence from Hong Kong”, International Real Estate Review, 19, 1, 120-149.

26

February 2012

competing theories to explain such a strong correlation which is also ...... Wheaton, William C., (1990), “Vacancy, Search, and Prices in a Housing Market ...

831KB Sizes 1 Downloads 443 Views

Recommend Documents

February board minutes 2012
Board Meeting. February 16, 2012. CANDO office. CALL TO ORDER. Chair Peters called the meet eters called the meeting to order at 6:39 p.m.. ROLL CALL.

February 2012.pdf
Dec 9, 2011 - We each have a part of his car. that we keep with us to help us remember. the good times,” Baltayan said. Throughout the discussion amongst.

February 2012 - Newsletter.pdf
29 Augustana Lutheran Church 6:00 pm Dinner/7:30pm Worship. 69-05 Ditmars Blvd., East Elmhurst. MARCH 7 First Reformed Church of Astoria 6:00 pm Dinner/7:30 pm Worship. 27-35 12 Street, Astoria. MARCH 14 Atonement Lutheran Church 6:00 pm Dinner/7:30

Wuxing Cards February 2012.pdf
Download. Connect more apps... Try one of the apps below to open or edit this item. Wuxing Cards February 2012.pdf. Wuxing Cards February 2012.pdf. Open.

Minutes February 17, 2012.pdf
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. Minutes ...

nuts magazine february 2012 pdf
There was a problem previewing this document. Retrying... Download. Connect more apps... Try one of the apps below to open or edit this item. nuts magazine ...

Entertainment-February-2012.pdf
Page 1 of 8. insideentertainment. Inside Time in association with Gema Records February 2012. Welcome to the Inside Time entertainment supplement, in association with Gema Records,. featuring the latest interviews, reviews and iconic images from the

Package 'TeachingSampling' February 14, 2012 Type Package Title ...
Feb 14, 2012 - Creates a matrix of domain indicator variables for every single unit in ... y Vector of the domain of interest containing the membership of each ...

Revised February 2012 US Hours and Productivity ...
hours worked is defined as total hours worked per population ages 16 to 64. ... The population covered by our analysis is the non-institutional population, ages ...

February 6, 2012 Meeting Minutes Page 1 of 6 School Board Niles ...
National College Advisor,. 205 days .... Niles North Automotive Program. ... Sophomore College Night had 600 people in attendance, largest participation yet.

Please register by the February 1, 2012 discount date ...
Uniforms - County players and Double Nickel teams will wear the TJAY uniform without exception. Jersey, shorts and socks are provided by TJAY Soccer (See ...

february newsletter.pdf
review a few of the most popular apps used by. SLPs and also ones that can be used at home. ***Note: There are MANY more apps available then.

February Newsletter
used some of your support money to get a Mongolian Bible for a girl named ... Thank you for being the best Support Team I could imagine having! A note to my ...

February 2017
Book Fair. 24. 25. Star of the Week. Odin ame Word Family. 26. Art. Media/Tech. Happy Birthday. William. 27 Music/P.E.. Happy Birthday. Odin. RV Spirit Wear/. Wear Blue Day. 28 P.E./Music. I love to read month! READ, READ, READ. Look at those Read-A-

February 15, 2015
Feb 15, 2015 - loved ones on the list. Reading Food ... Follow us and stay updated with social media! ... season we are all encouraged to participate in our.

February 15, 2015
Feb 15, 2015 - moral habitus, or what we would call vice, lack conformity to doing what is good and thereby put us ... PARISH CENTER (PAC) IMMEDIATELY FOLLOWING .... Licensed Nursing 24 hours a day • Recreational Therapy ...

February 2010
Feb 24, 2010 - This year it's all about new, exciting, and thrilling books to read! Sunday. Monday ... What teams play today? Compare their season ... books? 8. February is… Black History Month. Library Lovers Month. American Heart Month. Canned Fo

January/February
3:00 – 5:00 PM. Ames at Boone. 3:00 – 5:00 PM. Valley at Ames. 3:00 – 5:00 PM. Swim. 6:00 PM. Light Wts/Swim. 5:30 PM. Swim. Sunday. Monday. Tuesday. Wednesday. Thursday. Friday. Saturday. 17. 18. 19. 20. 21. 22. 23 wk. 8:00 – 10:00 AM. 6:00

Adds From February 1 - February 29, 2016.pdf
Feb 29, 2016 - There was a problem loading this page. Adds From February 1 - February 29, 2016.pdf. Adds From February 1 - February 29, 2016.pdf. Open.

1 February 13, Friday February 14, Saturday -
Feb 1, 2013 - 11.00am. An overview of Conference: District Trainer PDG Rtn.Babu Joseph ... Interlude: Service project: Roses & Wings of Love (RC Cochin.

February Newsletter.pdf
Page 1 of 1. Glennville Middle School ~ Parent Newsletter. 912-654-1467 Friday, February 2, 2018 www.gms.tattnallschools.org. January 2018. Students of the Month. 6th Grade: Aniyah Conyers &. Bode DeLoach. 7th Grade: Walker Robinson &. Zamya Smith. 8

february
Sunday. Monday. Tuesday. Wednesday. Thursday. Friday. Saturday. 1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12. 13. 14. 15. 16. 17. 18. 19. 20. 21. 22. 23. Crazy Hat Day. 24. 25. 26. 27. 28. Collin's. Birthday. Groundhog. Day. Valentine's. Day. President's. D