Short Run Housing Market Dynamics: An Application to Hong Kong Lok Sang Ho Department of Economics Lingnan University Tuen Mun, Hong Kong Phone: (852)26167178 Fax: (852) 28917940 Email:
[email protected] Donald R. Haurin Departments of Economics, Finance, and Public Policy Ohio State University 1010 Derby Hall, 154 N. Oval Mall Columbus OH 43210 Phone: 614-292-0482 Fax: 614-292-9530 Email:
[email protected] Gary Wong Centre for Public Policy Studies Lingnan University Tuen Mun, Hong Kong Phone: 852-26167182 Fax: 852-25910690 Email:
[email protected]
January 28, 2003 Keywords: Housing, Price Dynamics, Equity Effects, Public Housing, Privatization JCL Classification: E32, H30, R21, R31, Acknowledgements: The authors are grateful for the comments and suggestions received from Terence Chong, Daniel Cheung, Yue Ma, and Jimmy Ran. The work described in this paper was supported by a grant (LU3008/00H) from the Research Grants Council of Hong Kong Special Administrative Region, China.
1
Short Run Housing Market Dynamics: An Application to Hong Kong
Abstract
We develop hypotheses about the dynamic changes of house prices and the volume of transactions based on a synthesis of the housing filtering model and a model of the interactions of home equity with downpayment constraints. We predict that a shock to one quality segment of the housing market will create ripple effects in house prices and transactions throughout the housing quality continuum.
Using data from
Hong Kong, we find strong evidence that policy changes that impact the lower end of the quality continuum lead to changes in house price and transactions volume throughout the quality continuum, spreading from low quality units to the high end of the quality continuum. These results highlight the interconnectedness of the housing market. In particular, we show the important role that equity effects play in the dynamics of the housing market, especially when amplified by downpayment constraints.
2
There is wide recognition that housing market dynamics are complex and not well understood.
Theoretical approaches to housing dynamics include models of
neighborhood change, filtering, search, equity effects, urban growth, and housing chains. Empirical tests of these models are infrequent, perhaps due to the complexity of the models or the lack of data. However, given the importance of housing to urban areas and given its importance to policy issues and the growth of household wealth, there is a need for empirical studies of house price and transaction dynamics. Our study focuses on one aspect of housing market dynamics, a test of “ripple effects” across different quality levels of housing. The application is to Hong Kong. This locality is chosen because a form of policy experiment occurred there and data are readily available. We find that the housing market behaves in a manner consistent with the predictions of our model. One implication of the empirical results is that houses of all quality levels are closely tied together through dynamic processes. Another finding is that price effects move through the quality continuum of a housing market with relatively great rapidity. A third is that we offer additional confirmation of the positive correlation between the turnover rate of housing and house prices, this relationship found in other recent studies of the housing market.
Review of Theoretical Approaches to Housing Market Dynamics The dynamics of housing markets have been described using a variety of approaches. We first describe filtering and urban growth models, these highlighting the long run impact of policy changes on the stationary state of house prices and quantities. We then focus on models of neighborhood change and succession, housing chains, and models describing spatial ripples in house price changes. Finally, we review search and equity effect models, these focusing on the short run consequences of shocks to the housing market on price and turnover rates.
Each of these models plays a role in
guiding our theoretical approach to housing dynamics.1 The seminal contributions to the theoretical description of the filtering model are the papers by Sweeney (1974a, 1974b).
In filtering models, the housing market is
separated into distinct quality levels. Households differ in terms of income and other characteristics, and in equilibrium they are matched to housing of different qualities according to their income levels and willingnesses to pay. The durability of housing is a central part of the model and depreciation causes higher quality units to “filter down” to lower income households until eventually demolition becomes more economical than
3
maintenance. In Sweeney’s models, depreciation can be partially offset by landlord maintenance.2 The model predicts the long run equilibrium distributions of the quality of housing units and prices. Public policies can be evaluated using this model by finding the impact on new construction, the price and quality distributions, and household welfare. After a shock, equilibrium is reestablished when supply equals demand at each quality level and households have no incentive to move to some other quality of housing. Empirical tests of the filtering model have focused on its long run equilibrium implications. Muth (1969) used a reduced form test and found the expected negative relationship between household income and the age of properties. Bond and Coulson (1990) in a study of FHA data used Rosen’s (1974) two-step method to estimate households’ demand for dwelling age. In contrast with Muth, they found little support for the claim that households’ demand for new dwellings is affected by their income. A positive (but not causal) correlation between household income and the newness of dwellings is observed because newer houses are on average larger than older homes and demand for size rises with income. The short run properties of the filtering model are reasonably transparent. The housing market is a continuum linked through prices, and housing of one quality level is a substitute for housing of the next quality level. If a group’s income rises incrementally, its demand for higher quality properties increases. In particular, the demand for homes in the next higher quality level rises, causing price to rise in that quality level. The short run equilibrium between this level and the next higher level is disturbed, causing further substitution among higher levels of housing quality. This price change should quickly ripple up throughout the quality continuum. One limitation of Sweeney’s model is that spatial location is not included in the model.
A more important limitation of his model is that landlords do not face a
downpayment constraint when purchasing housing. Thus, the dynamics caused by the interaction of the downpayment constraint and household equity levels are excluded from the model. A third limitation is that the search and matching processes of buyers and sellers are not modeled. Many researchers have developed models of a growing urban area that incorporates durable housing of various quality levels. Early efforts include Braid (1979), Arnott (1980), Brueckner (1980a, 1980b), Akita and Fujita (1982), and Wheaton (1982a, 1982b). Models continue to be refined; for example, Arnott et al. (1999) incorporates space into a general equilibrium model of an urban area with durable capital and
4
endogenous maintenance. Their paper focuses on establishing the conditions needed for the equilibrium to be a stationary state characterized by the price of a unit of housing at a particular location being time invariant (except when quality changes). This focus reduces the relevance of the paper for our study of short run price dynamics. Arnott et al. note that restrictive assumptions are needed to generate a stationary state rather than rent cycles and they propose that future theoretical work focus on non-stationary state dynamic analysis. The literature on residential succession and neighborhood change is related to both filtering models and house price dynamics. One focus of residential succession has been on changes in the racial and ethnic composition of neighborhoods. A seminal contribution was Bailey’s “border model” (1959) that not only highlighted equilibrium price differentials in Black and White neighborhoods, but also described the path of house price dynamics that would lead to stable price differentials between areas. Again, the substitutability of housing in different locations is key to the model’s predictions. Megbolugbe, Hoek-Smit, and Linneman (1996) summarize Grigsby’s (1963, 1987) contributions to models of neighborhood change, noting that he viewed urban areas as aggregations of submarkets linked with each other through cross supply and demand elasticities. However, they note that there is little emphasis in this line of research on price dynamics, and only an occasional prediction of price changes as neighborhoods change. Linkages in the housing market at the micro level are the foundation of “housing chain” models.
Rosenthal (1997) starts a chain by assuming that a household
permanently exits from the housing market (e.g. out-migration) or a new housing unit is created.
She makes the very strong assumption that an existing homeowner who
wishes to move into this unit cannot complete the transaction until his or her household’s current home is sold. Potential first-time homeowners enter the market each period, searching for a dwelling.
Next, space is introduced into the model with the further
assumption that households search for housing only in their own and nearby submarkets. The result is a housing chain.
This model predicts that existing owner-occupiers’
duration of marketing time is longer than the search time of either first-time buyers or the time-on-market of the houses owned by sellers who are exiting the housing market. Rosenthal introduces house prices to the argument in an ad hoc manner by assuming that potential sellers reduce their reservation price if a transaction is not completed during the initial time period and by assuming that first-time buyers increase their offer
5
price if they must wait for a transaction to be completed. However, no formal model of this process is presented. She asserts that house price changes for owner-occupiers should temporally lag those of first-time buyers and exiting sellers, and similarly, the price changes for used homes should lag that of newly constructed homes. Rosenthal (1997) used the 5% Sample Survey of Building Society Mortgages and 425,000 observations to create constant-quality house price indexes for 11 United Kingdom regions from 1979 to 1992. She separated the series into four groups: firsttime owners, existing owners, new properties, and existing properties. Her first finding was the series were cointegrated. She then performed Granger causality tests on the pairs of price series and found fairly strong evidence that the price series for first-time owners Granger caused that for owner-occupiers (and not vice versa) and that for new units Granger caused that for used homes. A drawback of Rosenthal’s (1997) study is that the claims underlying the specific empirical tests are not well grounded in theory. An advantage is that it shows there are short run linkages of housing submarkets through changes in house prices. Similar empirical methods have been used to study “ripple effects” in regional house prices, particularly in the United Kingdom (MacDonald and Taylor 1993; Alexander and Barrow 1994, Meen 1999.) As Meen notes, the issue has both been viewed as a statistical problem with the goal of describing the patterns of changes in regional house prices, and as an economic problem with the goal of explaining why price ripples occur. The statistical properties of house price series in the regions of the U. K. are fairly well established for the period 1968-1997 and they suggest ripple effects from the southeastern region to the north. Meen suggests that ripple effects could be caused by four factors: migration, equity transfer, spatial arbitrage, and spatial differences in the regional determinants of house prices. Migration could cause house price ripples if households relocate in response to changes in the spatial distribution in house prices. House prices need not equalize among regions because there are long lasting differences in regional or metropolitan area fixed endowments (e.g. climate) or scale economies (Haurin 1980). However, an exogenous shock to a region may disrupt local house price levels, causing migration (Haurin and Haurin 1988). Migration spreads the impact of the shock throughout a region or country.
The observed outcome would be a spatial ripple of house price
changes. Meen’s second potential cause of house price ripples is the equity effects model
6
described by Stein (1995). We discuss this model in detail later, but the basic driving mechanism is that changes in house prices change homeowners’ equity. An increase in equity relaxes downpayment constraints, permitting additional mobility.
In contrast,
falling nominal house prices reduce equity and constrain mobility. This effect is very relevant for a metropolitan area such as Hong Kong. Meen’s third explanation is based on the spatial diffusion of house prices, a manifestation of arbitrage mitigated by search costs or by the diffusion of news throughout a region. Pollakowski and Ray (1997) tested whether house price changes in one region or PMSA predict price changes in other regions or PMSAs using a VAR model. Their work built on Tirtiroglu (1992) and Clapp and Tirtiroglu (1994) who found that excess returns to housing in a submarket diffused to other submarkets of the same MSA.
Pollakowski and Ray found statistically significant cross-price effects at the
regional level, but there was no sensible economic pattern to their results. They also found significant cross-PMSA effects in the New York CMSA, with a slight preponderance of effects being for contiguous areas.
However, the economic
interpretation of their mixed results is not clear. One problem with this purely spatial approach is that it implicitly argues that the transmission mechanism flows across space, not across economically similar housing submarkets. For example, one could argue that a shock to a locality with high house prices would most likely affect other localities with similar houses, and these houses need not be in the “next door” community. Our study addresses this issue by focusing on the transmission of price shocks across similar property types within a metropolitan area, rather than narrowly focusing on the spatial distance between communities. Meen’s fourth explanation is not based on regional spillovers, but on comparisons of the spatial pattern of the underlying independent price processes among regions.
He assumed that all regions react to shocks with different speeds. House
prices change first in the fastest reacting region, followed by price changes in slower reacting areas.
The apparent result is price ripples, however no transmission
mechanism is present. Meen (1999) developed an econometric test of this hypothesis using U. K. regional data. He found evidence supporting the claim that the reason for the observed price ripples is that short run regional behavioral responses to national shocks occur at different rates. In the long run, house prices tended to return to the same preshock relative values. Hort (2000) argued that changes in the turnover rate in housing are linked with
7
changes in house prices. The basis for this argument is a search model developed by Berkovec and Goodman (1996).
Hort’s goal was to test whether the turnover rate
changes before price changes are observed. The argument why this relationship should be observed is that sellers establish list prices based on their expectations of sales prices, these expectations influenced by recent market prices and the recent “ease of selling”. Buyers’ offers are influenced by recent prices but also are subject to demand shocks such as unexpected changes in unemployment, income, population growth, mortgage interest rates, and migration. Thus, the distribution of buyers’ offers moves before that of sellers’ reservation prices. The result is a rapid change in the turnover rate as the market quickly clears in up-markets and houses remain unsold in down-markets.3 Initially, there is little upward price movement in the up-market because sellers have previously set list prices based on the set of price expectations at the time of listing. There is little downward price movement initially in a down-market because even though houses remain unsold, those that sell do so near their list price (skimming off the upper tail of the distribution of buyers’ offers).
However, as sellers become aware of the
change in the expected marketing time for a home, they adjust their list prices, inducing a positive correlation between the turnover rate and house price changes. Hort (2000) found the dynamics of the Swedish housing market followed this model’s predictions. The last theoretical approach we consider is based on Stein’s (1995) model where equity effects are incorporated into the housing market.
Stein’s goal was to
explain both the large price swings observed in some housing markets and the positive correlation of transactions volume with house price changes. He notes that households who wish to own must make a downpayment and pay closing costs. If house prices are rising, current owners’ home equity rises, increasing their wealth, allowing them to make a larger downpayment on another, more expensive, home.
Thus, increasing house
prices facilitate trading-up and should increase transaction volume. Stein showed that this effect is enhanced if mortgage lender imposed minimum downpayment requirements constrain a large percentage of current owners’ choice of dwelling. Similarly, if house prices fall, a household’s equity falls, and this household’s ability to purchase another house is reduced, perhaps greatly. Transaction volume should fall at the same time that house prices are falling. Evidence of equity-constrained behavior includes purchasing a downsized home or opting for a lower downpayment but higher monthly payment (Hendershott, LaFayette, and Haurin 1997). Linneman and Wachter (1989), Zorn (1989), and Haurin, Hendershott,
8
and Wachter (1997) provide convincing evidence that a substantial portion of U.S. households face binding downpayment constraints. Downpayment requirements in the U.S. are among the least restrictive in the world, thus equity constraints are likely prevalent throughout developed countries. The Stein model is rich in predictions. Even though his formal predictions are based on a static process, the extensions to a dynamic setting are fairly transparent. Predictions about the dynamic consequences of equity constraints include: 1) a positive correlation of the trading volume of residential properties with changes in house prices, 2) the time-on-market for houses being negatively correlated with house price changes, and 3) house prices being more sensitive to shocks the greater the percentage of constrained homeowners in the area. Genesove and Mayer (1997) and Lamont and Stein (1999) tested Stein’s predictions. Genesove and Mayer used data from the Boston condominium market and found that potential sellers with high LTVs (thus little home equity) set relatively high list prices and have longer marketing times in periods when condominium prices have fallen. These results are broadly consistent with Stein’s predictions. Lamont and Stein found that real house prices are more sensitive to shocks to per capita income in cities where a relatively high percentage of homeowners have a high LTV, this finding consistent with Stein’s model.
Their results are robust to many specification checks, the only
insignificant finding occurring when they use a weak instrument for LTV. In summary, there are multiple theoretical models of house price dynamics and multiple empirical techniques for testing the models. The most relevant models for our study of Hong Kong are the search and equity effects models because we focus on short run changes in house prices and turnover rates. However, filtering and housing chain models also are relevant.
The Hong Kong Housing Market Hong Kong’s population was 5.7 million in 1990 and 6.7 million in 2000. In 2000, there were 2.1 million dwelling units. About 40 percent of the population lived in public rental housing during the 1980s. Public rental housing (PRH) accounts for about 70 percent of the total public housing program, the remaining 30 percent being in the form of “Home Ownership Scheme” housing (HOS). With average rent set at no more than 10 per cent of the median household income of tenants, the implicit subsidies in public rental housing were substantial.
Also, until quite recently Hong Kong protected the
9
tenure of public housing tenants. 4
Regardless of changes in financial conditions, a
family and its children could remain in the unit (Housing Authority 1993).
These
subsidies provided a significant incentive for households to remain in the public rental sector even though they had the resources to become homeowners. Homeownership in Hong Kong is tax-advantaged because homeowners do not pay a capital gains tax on their (owner-occupied) home’s appreciation in value. However, the Inland Revenue Department imposes a profits tax on the short-term sale of properties. Properties that are not principal homes are taxed at the profits tax rate (currently about 16.5%) if they are sold within two years of purchase. Properties that are vacant and sold beyond two years may still be subject to profits tax if the owner cannot give a satisfactory reason why it is not rented or occupied. In March 1998, the government initiated a new mortgage interest allowance up to a ceiling of HK$ 100,000 per year for five years (the exchange rate with U.S. currency is about 8 to 1). Prior to 1997, the typical loan-to-value ratio (LTV) was 0.70 for newly purchased houses priced less than HK$ 6 million. Luxury dwellings’ LTVs were typically 0.60 during this period.5 The implication is that mortgage lender constraints are stricter in Hong Kong than in the U.S. LTVs increased beginning in February 1999 when banks began to offer loans up to 85% of house value.
The cause was the Hong Kong Mortgage
Corporation began to provide mortgage insurance covering up to 15% of the value of the property. Beginning in April 1987, the Hong Kong Housing Authority undertook a series of policy changes that decreased the incentives for households to remain in public rental housing. 6
First, it required tenants who had been housed for over ten years in the
program and who had income exceeding twice that of the Waiting List Income Limit, (WLIL) to pay double rent.7 The Housing Authority relaxed this requirement slightly in April 1993, when households with income exceeding twice the WLIL were required to pay only 1.5 times the rent, while those with income exceeding three times the WLIL had to pay double the rent.8 In April 1996, public policy changed again. Tenants of public housing whose household income and net asset value exceeded prescribed limits had to pay market rent (Housing Authority 1995, 1996). Thus, beginning in 1987, actual and impending policy changes encouraged households living in public rental housing units to buy residential properties as a hedge or to exit and become homeowners. A survey conducted by the Housing Authority in 1992-1993 showed that 24 per cent of home purchases in the market were by public housing tenants and that over 13 per cent of
10
tenants owned one or more residential properties. This activity was clearly also a result of the sizable savings accumulated during the high growth and high inflation period in the 1980s and early 1990s (Watanabe, 1998).9 The final change in public policy that we consider was announced on December 8, 1997. The “Tenants Purchase Scheme” (TPS) allowed existing public rental housing tenants to purchase their own flats at up to an 88 percent discount from the estimated market price, provided that the sitting tenants committed to buy within a specified period. The units were priced from about HK$ 70,000 (less than US$ 10,000) to about HK$ 300,000. The scheme was implemented in phases, with each phase covering about 25,000 tenants, starting in 1998. This policy reversed the previous incentive of tenants in public rental housing to leave their dwellings and move into the ownership market. The announcement of the TPS program in 1997 greatly changed the incentive structure for existing tenants in the public rental program.
They no longer had an
incentive to leave the public sector and buy expensive HOS housing or private sector homes. Rather, their incentive was to remain in place and purchase the rental unit. The expected effect would be to greatly reduce the flow of these households into the private sector ownership market.
Figure 1 shows a general upwards trend in property
transactions (mainly residential, but also inclusive of commercial and industrial) from 1987 through 1997, but a dramatic decline following the announcement of the TPS in December 1997. These data generally correspond to our expectations, but they report only the aggregate number of transactions while our interest focuses on ripple effects within the quality continuum. [INSERT FIGURE 1]
A Model of the Hong Kong Housing Market The question we address is: what are the market-wide implications of the series of public policy changes by the Hong Kong Housing Authority? Our stylized model draws on components of the filtering, search, equity, and housing chain models described above. Assume there is a hierarchy of housing ranging in quality from low to high, similar to filtering models. The supply of units at each quality level is fixed in the short run. Households are homogenous in preferences and income, but they have differing endowments of current wealth. A downpayment is required for ownership, it being a fixed percentage of house price. The capital market is imperfect so that a household
11
cannot borrow against future income to procure the downpayment. If a household’s wealth is not sufficient to make the downpayment, it cannot purchase at that level of quality and must select a unit from lower in the quality distribution, similar to the Stein model. All households are equity-constrained except those at the highest quality level. Because the distribution of wealth in the population is continuous, there is a range of wealth within each quality level. Households with the least wealth in a particular level have just enough wealth to make the minimal required downpayment. Households with the greatest amount of wealth in the same quality level have a level of wealth just shy of the amount needed to make the downpayment required for the next higher quality level. All households in this quality level have sufficient income to purchase a higher quality unit, but they lack the needed wealth. The aggregate demand for a particular quality level depends on the number of households that have sufficient wealth to make a downpayment for that quality level and the level’s house price. We assume there is continual movement of first-time homeowners into the ownership market. Households exit from the ownership market due to deaths or outmigration from the area. Short run price is determined by equating supply and demand at each level. Short run equilibrium is characterized by a constant price at each quality level. In Hong Kong, the change in public housing policy beginning in 1987 increased the flow of households into the ownership sector. In general, these households had sufficient wealth to become owners, but only at relatively low quality levels.
The
implication is that the aggregate demand at low levels of quality increased, resulting in upwards pressure on house prices at this level. This price increase raised the home equity of existing owners of low quality housing. Those households with wealth levels nearly sufficient to move up in the quality continuum now would have sufficient wealth to make a downpayment on a higher quality house. Their exit from this quality level and movement into the next higher level results in price and equity increases in the higher level. At each level, exits to the next higher level are determined by the amount of increased wealth, this dependent on the price increase, which depends on the number of households demanding ownership at this level.
Clearly, a simultaneous solution is
needed to determine the short run equilibrium price increase and number of exits to higher quality levels. Exits from a quality level cannot cause house price in this level to decrease below the original price level because exits are created solely by increases in wealth.
Just as in a housing chain, the process will move upwards throughout the
12
quality continuum. The first predicted outcome of the change in Housing Authority public housing policy from 1987 up to December 1997 is that the transaction volume and house price at the lowest quality level of owner-occupied housing increase. We expect to then observe the next higher quality housing tier increase in price and quality, this continuously followed by increases throughout the housing quality continuum. The December 1997 announcement of the TPS policy suddenly reversed the policies of the prior decade. By offering tenants the opportunity to buy their own units at a below market price, including those tenants with relatively high wealth, the policy diverted the flow of potential first-time homeowners from the private ownership housing market to private ownership of public sector rental housing. According to the above model, there should be a dramatic reversal of the price changes and the turnover rate. The decline in demand for low quality owner-occupied housing should result in falling prices at this level.
The consequence is that owners’ equity would fall, making it
impossible for households to move up in the quality continuum. Cutting off the normal flow of households upwards through the quality continuum affects the next higher level of housing quality. The reduction in demand again reduces price, homeowners’ equity, and the turnover rate in the next higher quality level. These effects ripple upwards through the quality continuum. In all cases we expect to observe a positive correlation of changes in house prices and the turnover rate of housing. The primary hypotheses are that house prices and the turnover rate change first in low quality owner-occupied housing, followed by changes in the next highest quality level, and so on throughout the quality continuum. We test these hypotheses using data from Hong Kong in the period 1987 to 2000.
Data and Empirical Tests House price data for four classes of housing from the period 1987 to 2000 were obtained from the Hong Kong Rating and Valuation Department. The housing classes are defined according to the size of housing unit: Class A (under 40 m2), Class B (40-60 m2), Class C (70-99.9 m2), and Class D (100 m2 and above). 10
The Ratings and
Valuation Department house price indices “are designed to measure rental and price changes with quality kept at a constant” (Hong Kong Property Review 2001). 11 We expect that the above described changes in housing policy would first affect house prices in Class A units, followed by B, C, and D.
13
Our second data set is composed of monthly transaction volume data for 1995 to 2002. These data pertain to existing private housing transactions and were obtained from the Land Registry: Centaline Databank Management Department. We separate properties by price and observe the time series of transactions. House price categories are: Home Ownership Scheme housing (HOS), 1-2 million $HK, 2-3 million, 3-5 million, and 5-10 million.
We expect that the changes in housing policy would first affect
transaction volume in lower valued properties, followed sequentially by properties with higher prices.12 The first empirical test is a causality test of whether house price changes in smaller properties lead those in larger properties throughout the distribution of property sizes. Under the standard Granger test (1969), if X causes Y, then changes in X should precede changes in Y. That is, lagged values of X can help improve the prediction of current values of Y. First, we estimate the unrestricted model: k
k
i =1
i =1
Yt = µ t + ∑ α i Yt −1 + ∑ β i X t −1 + ε t
(1)
Next we estimate the restricted model: k
Yt = µ t + ∑ α i Yt −1 + ε t
(2)
i =1
The null hypothesis that X does not Granger-cause Y (Ho : β1= β2= ··· = βk =0) can be tested by the reported F-statistics. If the coefficients on the lagged values of X are jointly and significantly different from zero, the null hypothesis can be rejected. Y then is said to be Granger-caused by X.13 We begin the analysis by checking the stationarity of the price series using the Augmented Dickey Fuller (ADF) test. This check is necessary because the Granger Causality tests require all time series be stationary.
Table 1 shows that the null
hypothesis that the price series contain a unit root cannot be rejected for any series, indicating they are non-stationary. Thus, each variable must be transformed into a firstdifference form to achieve stationarity before conducting the test.
In implementing
equation (1), in order to avoid the ambiguity in choosing the lag lengths, we use the Akaike’s final prediction error criterion to determine the optimal lag specifications. [INSERT TABLE 1] Six sets of the Granger test results are reported in Table 2. Although we cannot reject the null hypothesis that the price movements of Class A do not cause those in Class B, we can reject the null that price movements in Class A do not cause those in
14
Class C (1% level) and Class D (10% level). Also, we can reject the null that price movements in Class B do not cause those in Class C (1% level) and, similarly, reject those the null that price movements in Class B do not cause those in Class D (1% level). All of these results support the hypotheses that prices changes in low quality properties lead those of higher quality properties. A ripple effect of prices upwards through the quality continuum is present. We also find that the Granger tests do not support reverse causality.
Price
changes in Class B, C, and D do not Granger cause those in Class A. Neither do price changes in Class C or D Granger cause those in Class B. Finally, there is no evidence that price changes in Class D Granger cause those in Class C. Thus, all of the tests support a single direction of the price ripple. [INSERT TABLE 2] The model also predicts transmission patterns in the number of transactions, again beginning with low quality housing and continuing upwards through the quality continuum. Our base level is Homeownership Scheme housing (HOS), this representing the publicly subsidized ownership program. Typically, the households demanding to become HOS owners were participants in the subsidized rental scheme or households from the general public who are eligible under income and asset level tests.
As
described above, after the December 1997 TPS program was instituted, public rental housing tenants’ demand for HOS housing fell dramatically because they could purchase their rental unit at about 12% of the market price. This drop in demand for HOS housing should reduce transaction volume, price, and equity among HOS owners. Our model predicts that following the impact on HOS owners, there should be ripple effects on transaction volume upwards through the quality continuum. [INSERT TABLE 3] Our results are reported in Table 3. The total number of transactions over the period from 1995 to 2002 by house value category ranges from 164,922 in the one to two million $HK category, to 90,497 for two to three million $HK, to 72,171 for three to five million $HK, to 40,640 for five to ten million $HK. We find that changes in the transaction volume of HOS housing Granger caused changes in transaction volume in all higher valued housing. This is convincing evidence that the lowest tier of housing quality was the leading submarket in terms of changing transaction volume. There is no evidence of reverse causality from higher to lower quality levels. Additional strong evidence of a ripple effect is that changes in the transaction
15
volume of private homes valued between one and two million Granger caused changes in transaction volume for all groups of houses with higher prices. Thus, booms or busts in the submarket for low quality housing are passed on with a lag to the next highest quality housing, and subsequently from this level to higher valued levels. The results become mixed only among the highest quality housing levels. There is no effect of two to three million $HK private housing on five to ten million $HK private housing and the causality between two to three million and three to five million is weakly the opposite of what we expected. However, as expected, changes in the transaction volume of three to five million $HK private housing Granger caused changes in five to ten million $HK private housing. As shown in table 3, the AIC criterion identifies 1 lag for cases (1), (3), (4) and (8); 4 lags for case (2) and (9); and 5 lags for case (5), (6), (7), and (10). Thus, the evidence suggests that the ripple effect on transaction has a time lag of one to five months. The above evidence characterizes the Hong Kong housing market in general form. Next, we present a specific test of the impact of the Tenants Purchase Scheme (TPS), which was announced on December 8, 1997.
This program allowed sitting
tenants in the public rental housing program to purchase their unit at a deep discount from market price.
Our dependent variable is the monthly number of purchases of
existing homes in the period surrounding the announcement of the TPS program. The first test is of the impact of the announcement of the TPS. We create a dummy variable that takes the value 1 during this period and is 0 before December 1997.
Our
expectation is that this variable’s coefficient will be negative. Confounding the analysis of the impact of the TPS was the advent of the Asian financial crisis (AFC), this beginning in the third quarter of 1997.
We expect that
increased uncertainty about the HK-US dollar link would discourage purchase of HK dollar denominated assets, particularly homes, and thus reduce turnover. We use the difference between the spot exchange rate and the one-year forward rate to measure the AFC. This measure does not require subjective judgment as to when the crisis began and when it phased out.
It also is highly sensitive to the swings of market
confidence during financial crises. We normalize this variable and constrain its value to be between zero and unity. We also include in the regression a measure of the rate of house price appreciation during the previous year.
This value is measured by the
average price appreciation during the past six months as compared to one year earlier. As
discussed
above,
greater
house
price
appreciation
relaxes
households’
16
downpayment constraints and thus increases the ability of homeowners to trade up the housing quality continuum. [INSERT TABLE 4] The results of the estimation are in Table 4.14 We find that the Tenants Purchase Scheme had a significant and large negative impact on the volume of residential property transactions.
Because our dependent variable measures the volume of
transactions in the secondary market (private and public housing), we are capturing all of the ripple effects described above. There is some evidence that the Asian Financial Crisis dampened transaction volume, and some evidence that high rates of house price appreciation increase transaction volume, but neither coefficient is statistically significant.
Conclusions This study makes two contributions. Our theoretical guidance comes from a synthesis of existing models of housing filtering and of the interactions of changes in house price and home equity with downpayment constraints. We focus on the short run dynamic changes of house prices and transaction volume. We predict that a shock to the housing market will create ripple effects in prices and transactions throughout the housing quality continuum. Specifically, the shock should spread from one quality level to the next and so on, especially in markets with a high proportion of downpayment constrained households.
Our empirical work studies the effects of shocks due to
changes in the Hong Kong Housing Authority’s policies. The first policy change was one that increased the flow of public housing renters to the private ownership market, the second being one that dramatically decreased this flow. We find strong evidence that these policy changes resulted in price and transactions volume changes throughout the quality continuum, spreading from low quality unit to high quality ones. These results are consistent with our model of short run housing market dynamics and they highlight the interconnectedness of the housing market within a locality.
In particular, the results
highlight the importance of equity effects in the housing market, especially when amplified by downpayment constraints. The second contribution of the paper derives from its focus on Hong Kong’s economy. According to Jao (2001), the sharp downturn of the Hong Kong economy in 1998 and the slow recovery posed a puzzle.
“There was no pervasive financial
mismanagement, no reckless borrowing internally or externally. Hong Kong’s banking system was one of the best supervised in the world… Hong Kong had no sovereign debt.
17
While the private sector undoubtedly had some external debt, collectively it was well covered by its external assets…” (Jao 2001, p.140). We show there was a link between the change in housing policy in December 1997 and the subsequent collapse in the turnover rate of housing.
The associated declines in house prices rapidly spread
throughout the quality continuum, reducing home equity at each quality level. A topic for further research is measurement of the impact of these changes on the Hong Kong economy.
18
References Akita, T. and M. Fujita. 1982. “Spatial Development Processes with Renewal in a Growing City,” Environment and Planning A 14: 205-223. Alexander, C. and M. Barrow. 1994. “Seasonality and Cointegration of Regional House Prices in the UK,” Urban Studies 31: 1667-1689. Arnott, Richard. J. 1980. “A Simple Urban Growth Model with Durable Housing,” Regional Science and Urban Economics 10: 53-76. Arnott, Richard J. and Ralph M. Braid. 1997. “A Filtering Model with Steady-State Housing,” Regional Science and Urban Economics 27: 515-546. Arnott, Richard. J., Ralph M. Braid, Russell Davidson, and David Pines. 1999. “A General Equilibrium Spatial Model of Housing Quality and Quantity,” Regional Science and Urban Economics 29: 283-316. Bailey, Martin J. 1959. “Note on the Economics of Residential Zoning and Urban Renewal,” Land Economics 35: 2988-292. Berkovec, James A. and John L. Goodman Jr. 1996. “Turnover as a Measure of Demand for Existing Homes,” Real Estate Economics 24: 421-440. Brueckner, Jan K. 1980a. “A Vintage Model of Urban Growth with Perfect Foresight and Multiple Income Groups,” Journal of Urban Economics 30: 385-407. Brueckner, Jan K. 1980b. “Residential Succession and Land-Use Dynamics in a Vintage Model of Urban Housing,” Regional Science and Urban Economics 10: 225-240. Braid, Ralph M. 1979. “Theoretical Models of Urban Housing as Durable Capital,” Ph.D. Dissertation, M.I.T. Cho, Man. 1996. “House Price Dynamics: A Survey of Theoretical and Empirical Issues,” Journal of Housing Research 7: 145-172. Clapp, John M. and Dogan Tirtiroglu. 1994. “Positive Feedback Trading and Diffusion of Asset Price Changes: Evidence from Housing Transactions,” Journal of Economic Behavior and Organization 24: 337-355. Coulson, N. Edward and Eric W. Bond. 1990. “A Hedonic Approach to Residential Succession,” Review of Economics and Statistics 72: 433-444. Gatzlaff, Dean H. and Dogan Tirtiroglu. 1995. “Real Estate Market Efficiency: Issues and Evidence,” Journal of Real Estate Literature 3: 157-192. Genesove, David and Christopher J. Mayer. 1997. “Equity and Time to Sale in the Real Estate Market,” American Economic Review 87: 255-269.
19
Granger, C. W. J. 1969. “Investigating Causal Relations by Econometric Models and Cross-Spectral Methods,” Econometrica 37: 424-438. Grigsby, William C. 1963. Housing Markets and Public Policy. University of Pennsylvania Press: Philadelphia. Grigsby, William C., M. Baratz, George Galster, and D. MacLennan. 1987. The Dynamics of Neighbourhood Change and Decline. Pergamon Press: Oxford. Haurin, Donald R. 1980. "The Regional Distribution of Population, Migration, and Climate," Quarterly Journal of Economics 95: 293-308. Haurin, Donald R. and R. Jean Haurin. 1988. "Net Migration, Unemployment, and the Business Cycle," Journal of Regional Science 28: 239-254. Haurin, Donald R., Patric H. Hendershott, and Susan Wachter. 1997. “Borrowing Constraints and the Tenure Choice of American Youth,” Journal of Housing Research 8: 137-154. Hendershott, Patric H., William C. LaFayette, and Donald R. Haurin. 1997. “Debt Usage and Mortgage Choice: The FHA-Conventional Decision,” Journal of Urban Economics 41: 202-217. Hort, Katinka. 2000. “Prices and Turnover in the Market for Owner-Occupied Homes,” Regional Science and Urban Economics 30: 99-119. Housing Authority. 1993. Medium Term Report of the Long Term Housing Strategy. Hong Kong. Housing Authority. 1995. Safeguarding Rational Allocation of Public Housing Resources: A Consultation Document. Hong Kong. Housing Authority. 1996. Safeguarding Rational Allocation of Public Housing Resources: Report on Final Recommendations. Hong Kong. Jao, Yu-ching. 2001. The Asian Financial Crisis and the Ordeal of Hong Kong. Quorum Books: Westport CT. Lamont, Owen and Jeremy C. Stein. 1999. “Leverage and House Price Dynamics in U. S. Cities,” Rand Journal of Economics 30: 498-514. Linneman, Peter and Susan Wachter. 1989. "The Impact of Borrowing Constraints on Homeownership,” AREUEA Journal 17: 389-402. MacDonald, R. and M. P. Taylor. 1993. “Regional House Prices in Britain,” Scottish Journal of Political Economy 40: 43-55. Meen, Geoffrey. 1999. “Regional House Prices and the Ripple Effect: A New Interpretation,” Housing Studies 14: 733-753.
20
Meen, Geoffrey. 2002. “The Time-Series Behavior of House Prices: A Transatlantic Divide?” Journal of Housing Economics 11: 1-23. Megbolugbe, Isaac. F., Hoek-Smit, M. C., and Peter D. Linneman. 1996. “Understanding Neighborhood Dynamics: A Review of the Contributions of William G. Grigsby,” Urban Studies 22: 1779-1795. Muth, Richard. 1969. Cities and Housing. University of Chicago Press: Chicago. Pollakowski, Henry O. and T. S. Ray. 1997. “Housing Price Diffusion Patterns at Different Aggregation Levels: An Examination of Housing Market Efficiency,” Journal of Housing Research 8: 107-124. Rating
and
Valuation
Department.
2001.
Hong
Kong
Property
Review.
Rosen, Sherwin S. 1974. “Hedonic Prices and Implicit Markets: Product Differentiation and Pure Competition,” Journal of Political Economy 82: 34-55. Rosenthal, Leslie. 1997. “Chain-Formation in the Owner-Occupied Housing Market,” The Economic Journal 107: 475-488. Stein, Jeremy C. 1995. “Prices and Trading Volume in the Housing Market: A Model with Down-payment Effects,” Quarterly Journal of Economics 110: 379-406. Sweeney, James L. 1974a. “A Commodity Hierarchy Model of the Rental Housing Market,” Journal of Urban Economics 1: 288-323. Sweeney, James L. 1974b. “Quality, Commodity Hierarchies, and Housing Markets,” Econometrica 42: 147-168. Tirtiroglu, Dogan. 1992. “Efficiency in Housing Markets: Temporal and Spatial Dimensions,” Journal of Housing Economics 2: 276-292. Watanabe, Mariko. 1998. “The Impact of the Public Housing Policy on Household Behaviour in Hong Kong,” M.Phil. Thesis, University of Hong Kong. Wheaton, William C. 1982a. “Urban Residential Growth Under Perfect Foresight,” Journal of Urban Economics 12: 1-21. Wheaton, William C. 1982b. “Urban Spatial Development with Durable but Replaceable Capital,” Journal of Urban Economics 12: 53-67.
21
Table 1. Augmented Dickey-Fuller Unit Root Test Property Type
Test on
No Trend
Trend
Conclusion
Class A
Level 1st difference
-2.179 -3.026**
0.461 -4.425***
I (1)
Class B
Level 1st difference
-2.077 -3.219**
0.185 -4.377***
I (1)
Class C
Level 1st difference
-2.191 -3.360**
0.193 -4.522***
I (1)
Class D
Level 1st difference
-1.835 -3.774***
-0.787 -4.521***
I (1)
Notes: 1) 95% critical values for the ADF statistic with and without the trend are –2.914 and -3.49 respectively. 2) ** Indicates significance at the 5 % level, *** indicates significance at the 1% level.
22
Table 2. Granger Causality Tests of House Price Changes: Period 1987Q1 to 2000Q4 Null hypothesis: CLASS B does not Granger Cause CLASS A (1) CLASS A does not Granger Cause CLASS B (2)
CLASS C does not Granger Cause CLASS A CLASS A does not Granger Cause CLASS C
(3)
F-Statistic 1.238 0.121 0.210 6.393
Probability 0.271 0.729
Causality ---
0.648 0.015***
AÆC
CLASS D&E does not Granger Cause CLASS A 0.140 CLASS A does not Granger Cause CLASS D&E 3.163
0.710 0.081*
AÆ D&E
(4)
CLASS C does not Granger Cause CLASS B CLASS B does not Granger Cause CLASS C
0.417 11.80
0.521 0.001***
BÆC
(5)
CLASS D&E does not Granger Cause CLASS B 0.001 CLASS B does not Granger Cause CLASS D&E 6.339
0.973 0.015***
B Æ D&E
(6)
CLASS D&E does not Granger Cause CLASS C 1.303 CLASS C does not Granger Cause CLASS D&E 0.456
0.28 0.71
---
Notes: 1) Akaike’s final prediction error criterion identifies 1 lag for cases (1) to (5) and 3 lags for case (6) 2) * Indicates 10% significance level *** indicates 1% significance level
23
Table 3. Granger Causality Tests of Transaction Volume Changes: July 1995 to July 2002
Null Hypothesis:
Probability
Causality
(1)
PR 1m
