Housing affordability, self-occupancy housing demand and housing price dynamics

Housing affordability, self-occupancy housing demand and housing price dynamics

Habitat International 40 (2013) 73e81 Contents lists available at SciVerse ScienceDirect Habitat International journal homepage: www.elsevier.com/lo...

389KB Sizes 3 Downloads 200 Views

Habitat International 40 (2013) 73e81

Contents lists available at SciVerse ScienceDirect

Habitat International journal homepage: www.elsevier.com/locate/habitatint

Housing affordability, self-occupancy housing demand and housing price dynamics I-Chun Tsai* Department of Finance, National University of Kaohsiung, No. 700, Kaohsiung University Rd., Nanzih District, 811 Kaohsiung, Taiwan

a b s t r a c t JEL classification: R21 R31 Keywords: Housing market cycle Serial correlation Mean reversion Self-correcting pattern Investment-motivated housing demand

Classic theory suggests that the real estate market cycle reflects the consequences of an inherent selfcorrecting pattern. Previous studies found evidence showing the existence of two stochastic processes, serial correlation and mean reversion, in housing price dynamics. The present study utilized data from the Taiwan housing market to observe whether the self-correction pattern driven by housing demand occurs and whether it can explain the housing dynamics. This paper hypothesizes that the demand side of the housing market causes a self-correcting mechanism of housing prices. The hypotheses are examined using panel data of five major cities in Taiwan. Empirical evidence reveals that when housing prices rise, housing affordability decreases, followed by reduction in self-occupancy housing demand. Furthermore, change in demand structure raises the risk of prices dropping because of an increase in investment-motivated housing demand, eventually resulting in lower housing prices. Ó 2013 Elsevier Ltd. All rights reserved.

Introduction Housing market demand declined in the United States, triggering a global financial crisis (i.e., subprime mortgage crisis). Considerable research has focused on factors that destabilized the housing market. Research topics, such as the housing market bubble, government control over the housing market, and irrational behavior of investors, have caught the attention of researchers. However, no sound mechanism in the housing market that is capable of preventing market irrationality appears to exist. Whether the housing market is highly inefficient and whether its cycle must consist of bubble and collapse are questions that need to be addressed. Previous research reveals that those questions can be answered using classical theories. Roulac (1996) documents that the classic real estate market cycle reflects the consequences of an inherent self-correcting pattern of expansion, slowdown, contraction, correction, recovery, re-expansion, and so on. In classical theory, a self-correcting mechanism is initiated mainly by the supply side of the real estate market. Researchers begin from the perspective of a production element in macroeconomics: land.1 When price of real estate rises, profit from invested land becomes higher compared to that from other investments (e.g., building factories for

* Tel.: þ886 7 5919767. E-mail address: [email protected] 1 Considerable literature discusses the issue of space. 0197-3975/$ e see front matter Ó 2013 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.habitatint.2013.02.006

manufacturing products). Therefore, though the supply of land and real estate increases, it is followed by a price decrease. In a more recent paper, Glaeser, Gyourko, and Saiz (2008) also discuss the housing market cycle from the perspective of adjustment by the supply side. They compare the housing bubbles in different US states with different supply elasticities, and find that if supply is more inelastic, the duration of the housing bubble will be longer. A number of researchers emphasize the influence of economic activities on the real estate market and discuss the relationship between the business cycle and housing market fluctuation (Baxter, 1996; Davis & Heathcote, 2005; Greenwood & Hercowitz, 1991; Jud & Winkler, 2002; Roulac & Volk, 1989). Moreover, Leamer (2007) claims that for the US, housing is a business cycle; he finds that developments in the housing sector actually lead to economic activity. Elbourne (2008) proposes a monetary transmission mechanism through the housing market and argues that monetary policies affect the economy through house prices. Hence, research related to the behaviors of housing prices is clearly important. Research likewise explains the relationship between real estate and other markets.2 Roulac (1996) integrates the relationships among variables and considers that the relationship among capital flow, space supply, property performance, and financial return can

2 Discussions in these bodies of literature focus on cash flow between the real estate market and stock market, which initiates the relevance between the two markets (wealth effect). For example, Quan and Titman (1999), Green (2002), Sim and Chang (2006) and Tsai, Lee, and Chiang (2012) examine the relation between the two markets to test whether a wealth effect occurs between them.

74

I-C. Tsai / Habitat International 40 (2013) 73e81

potentially contribute to the housing market cycle. Fisher (1997) proposes the co-movement between residential and nonresidential investments since complementarity exists between the household and business capital in the production of goods. According to the above literature, the relationship between the real estate market and other markets likewise explains real estate market fluctuation. The next question revolves around the reason why the housing market continues to adjust at a slow pace, resulting in market bubble and collapse if a self-correcting mechanism (Roulac, 1996) is present in the real estate market. Several researchers suggest that this is caused by inappropriate government intervention. For example, in the case of subprime mortgage, excessive subsidy policy (Shiller, 2009) and easing monetary policy caused the housing market imbalance in the US. Issues on the impact of macroeconomic policy on housing market dynamics have been addressed as well by Muellbauer and Murphy (2008) and Goodhart and Hofmann (2008), to name a few. Existing literature focuses on discussing the impact of monetary policy on housing price (e.g., Iacoviello, 2005; Mishkin, 2007; Muellbauer & Murphy, 2008; Vargas-Silva, 2007). Research likewise suggests that the market is not as efficient as assumed in theory (Case & Shiller, 1989; Shiller, 1993, 2005). This is because not all buyers behave rationally according to the hypothesis in the theoretical model, and their irrational behavior causes inefficiency in housing price.3 Riddel (2004) also proposes a disequilibrium housing market model that separates disequilibrium caused by supply-side disturbances from demand-side disturbances. Riddel applies the model to the US housing market for the period 1967e1998 and finds that inefficiencies impede market clearing; thus, the market is characterized by sustained periods of disequilibrium. Previous studies document that two kinds of dynamic behavior are present in housing price. The first is serial correlation (Abraham & Hendershott, 1993; Case & Shiller, 1989), which refers to the autocorrelation between different periods of housing price variation. The second is mean reversion (Abraham & Hendershott, 1993; Capozza & Seguin, 1996), which refers to the reversion of housing price to fundamental value. A number of researchers also observe that the dynamics of housing prices vary according to location (Abraham & Hendershott, 1993). Lamont and Stein (1999) suggest that this is due to different financial leverage of homeowners in different areas. Therefore, homeowners are sensitive to market impact to a different degree, reacting to the impact on housing price at different paces. Housing price index is not merely “another” macroeconomic variable; in the same vein, adjustment in housing price index does not merely involve the change of economic variables. According to the discussion of Skinner (1989, 1996), Case, Quigley, and Shiller (2001), and Campbell and Cocco (2004), Leung (2004) proposes that significant fluctuations in housing price would imply significant fluctuations in wealth and thus potentially significant household wealth effects. According to Lamont and Stein (1999), change of housing price is possibly shaped by the household characteristics of different areas. Different dynamics of housing prices likewise significantly affect different households. For example, if houses are purchased during a housing boom, the buyer bears an increased burden. When an economic downturn is present, the income of

3 Previous studies have sufficiently documented evidence showing that traders in the housing markets are irrational. For example, Genesove and Mayer (2001) examine trading data in the real estate markets of central Boston in the 1990s and confirm the presence of the “disposition effect,” since real estate sellers were unwilling to recognize capital losses.

buyers decreases, and they are forced to sell their homes as they are unable to afford their mortgage. If housing prices drop and hit bottom, buyers do not only lose their property but also accrue a massive debt due to capital loss in house trading. However, this situation is only reflected in the overall house price index and is merely the negative serial correlation of housing price or correction mechanism of returning to the mean reversion. If buyers purchase houses in a market with steadily increased price (variation of the housing price is in positive serial correlation), buyers continue to profit from house trading even if their salary decreases and their houses are sold. Therefore, the characteristics of housing price variation are essential to make the choice between renting and buying and for the strategic decision of property investment. Moreover, the discussion on whether the self-correction pattern occurs and whether the housing price dynamics can be explained by this pattern is crucial. This paper focuses on the housing market in Taiwan, which is a distinctly emerging market. According to Tsai and Peng (2011), buyers continue to purchase houses, regardless of whether the market is booming or experiencing a housing price bubble and whether investor behavior is seemingly highly irrational. Tsai and Peng (2011) also find that this behavior leads to increasing burden on the part of buyers, signifying that the Taiwan market is extremely inefficient. Consequently, one may wonder whether the price-correcting mechanism is absent in the market. A Chinese proverb reads: “Only land tenure contributes to wealth.” Buyers do not only purchase houses for self-occupancy; housing is one of their favorite modes of investment. Buyers may continue to purchase houses for investment, even when the burden of housing investment increases. However, an increase in investment-motivated housing demand may influence the housing market structure. For example, Marshall and Marsh (2007) find that elasticity of demand is different for consumers as opposed to investors. Hence, whether differences exist between the market where demand motivated by self-occupancy is high and the market where investment-motivated demand is high must be examined. This paper examines the self-correcting mechanism of housing price initiated by the demand side to investigate the previous question. This analysis employs the quantity of new housing supplied to control for factors from the supply side. The study proposes that increasing housing price lowers housing affordability, reduces consumer demand, and raises investor demand, eventually resulting in decreased housing price. Using Taiwanese data, the hypotheses are supported by empirical evidence, providing a theoretical basis for the government to control the housing market instability, which is created by increasing investment-motivated demand. Literature reviews and hypotheses Literature reviews Roulac (1996) proposes that five critical interdependent forces constitute the real estate market and move in cyclic patterns: economic structure, space demand, space supply, capital flow, and investment performance. Although the reasons or determinates of cyclic patterns in different markets may vary, previous research on the real estate market cycle focuses more on discussing the cycle emerging from macroeconomic factors (Drucker, 1993) and the supply side of the housing market. Leinberger (1993a, 1993b) specifies that the real estate market cycle consists of three general phases of market conditions: (1) upturn, lasting one to two years; (2) mature, lasting two to five years; and (3) downturn, lasting two to four years. However, the cyclic patterns in different markets during different periods are very different. Therefore, the real estate market cycle is difficult to forecast.

I-C. Tsai / Habitat International 40 (2013) 73e81

Since the boom-bust recovery cycle had previously occurred in many markets, Himmelberg, Mayer, and Sinai (2005) and Smith and Smith (2006) turned to employ the fundamental value model of a housing market, seeking to estimate directly the fundamental value of the asset and subsequently compare it with the observed value in the market. If a divergence from the fundamental value occurs in the market value, scholars will propose that a financial bubble exists in the market. A few studies directly measure boom and bust in the housing market, and several analyze the dynamic behaviors of housing price. Previous studies offer evidence that demonstrates the existence of the two stochastic processes, serial correlation and mean reversion, in housing price dynamics. Scholars determine the features of market by observing the behavior of housing price. For example, Case and Shiller (1989) and Shiller (2005) find that irrational behavior exists in housing markets in many cities because irregular serial correlations and slow mean-reverting corrections exist in the housing price. However, the results of Abraham and Hendershott (1996), Capozza and Seguin (1996), Malpezzi (1999), and Capozza, Hendershott, Mack, and Mayer (2002) imply that US housing buyers act rationally, since in their studies prices in the US market exhibit a behavior of mean-reverting to the fundamental values. Capozza et al. (2002) illustrate that the dynamic behavior of housing price can influence cyclic patterns in the housing market. They propose that as the serial correlation increases, amplitude and persistence of cycles increase, whereas as reversion increases, frequency and amplitude of the cycle increases. However, Capozza et al. (2002) do not provide an explanation for when serial correlation will increase and when reversion will increase. Additionally, if a self-correcting pattern exists within house prices, the reason why housing prices remain at a high level, which leads to a housing bubble, requires discussion. Previous studies often attribute this phenomenon of housing price inefficiency to housing supply. For example, Glaeser et al. (2008) find that real price appreciations, especially in the 1980s, were correlated with inelasticity in supply. They mention that cities with more inelastic supply experience a larger price boom. Therefore, housing price-generating processes are non-stationary in supply-inelastic cities; these processes are non-stationary to a lesser extent in supply-elastic cities. Limits on housing supply do not only affect variations in housing prices; they also influence the relationships between housing prices and other variables. For example, Himmelberg et al. (2005) have constructed a metric for user costs of owning houses for different US cities. They find that unit costs of owning in cities constrained by inelastic supply are relatively higher than renting. Moreover, in these cities with inelastic supply, housing prices are more sensitive to interest rate changes as well. Previous studies have explained how housing supply influences the characteristics and dynamics of housing prices as well as the relationships between housing prices and other variables. If inelastic supply can influence the dynamics of housing prices, can elasticity of demand have an impact on variations in housing prices? Previous studies assume that two types of economic agents exist (i.e., consumers and investors), who have different objectives and respond differently to changes (e.g., see Friedman, 1976; Glennon, 1989; Huang, 1969; Muth, 1960). Moreover, Marshall and Marsh (2007) discuss consumer and investment demand for manufactured housing units and provide substantial evidence that consumers and investors respond differently in the manufactured housing market. The current study employs the viewpoint of housing demand structure to study related phenomena among housing prices. The purpose is to investigate how factors of housing market demand

75

may cause self-correction in housing prices and use this theory to explain the dynamics of housing prices in the Taiwan market. Hypothesis This study infers that apart from the characteristics of housing supply, characteristics of housing demand may likewise influence the dynamics of housing prices. This study proposes three hypotheses to state how structural changes in housing market demand may result in self-correction in housing prices. H1: An increase in housing prices reduces housing affordability. Himmelberg et al. (2005) find that in regions with lower market supply elasticity, unit cost of buying a house exceeds rental cost. This is because when supply elasticity of the market is low, any factor that increases housing prices causes more substantial price increases. Naturally, an increase in housing prices causes the unit cost of buying a house to exceed rental cost, thus reducing housing affordability. Tsai and Peng (2011) also find that in housing bubble conditions, housing purchases place an increasingly heavy financial burden on buyers. Previous studies investigate the escalating housing affordability problem resulting from the bubble-like behavior of house prices (Kau, Chang, & Lin, 2000; Malpezzi, 1990; Memery, 2001; Yamada, 1999). Although different ways exist to address the issue of affordability of home ownership, Gan and Hill (2009) propose that affordability can be conceptualized in at least three different ways, namely, purchase, repayment, and income. Purchase affordability considers the ability of a household to borrow enough funds to purchase a house, while repayment affordability considers the ability of a household to repay the mortgage. Income affordability measures the ratio of house prices to income. Based on repayment affordability, this paper uses the ratio between mortgage payment and income to estimate housing affordability. Hulchanski (1995) states that the term “housing affordability” can summarize the difficulties individual households face in their access to decent or adequate housing. Hulchanski proposes that households encounter a housing affordability problem when they pay more than a certain proportion of their income to obtain suitable housing. Other things being equal, an increase in housing prices will raise mortgage payment and consequently reduce housing affordability. H2: Reduction of housing affordability reduces demand for the purchase of housing for residential purposes but increases demand for housing purchase as an investment. If purchasing houses under housing bubble conditions places an increasingly heavy financial burden on buyers, why do people continue to purchase houses, thus preventing prices from being reduced immediately? This phenomenon occurs because of the unique characteristics of housing, a product involved in both consumption and investment. Housing can be used as an expensive form of consumption to satisfy the residential needs of the public, but it can be used for investment as well. An increase in housing prices causes the unit cost of buying a house to exceed rental cost. Thus, consumers with demand for housing services will opt to rent a house (Coulson, 2002; Goodman, 1988; Haurin & Kamara, 1992). However, in terms of investment, housing can be either leased to earn rental revenue or sold to earn capital gains. Therefore, previous studies (Dusansky & Koc, 2007; Follian, 1982; Goodman, 1988; Grange & Pretorius, 2000) also find that expected high capital gains in the housing market will stimulate households to buy homes.

76

I-C. Tsai / Habitat International 40 (2013) 73e81

According to the abovementioned literature, higher housing price causes different buyer responses, since two types of traders with different objectives exist. This study infers that a reduction in housing affordability reduces self-occupancy housing demand but increases investment-motivated housing demand. H3: Structural changes in housing demand influence the reduction of housing prices. As mentioned above, housing is a special asset class that simultaneously possesses investment demand and consumption demand. These two types of demand differ greatly from one another. When ratio of the two demands varies in the market, the composed market characteristic may vary as well. For example, when investors consider housing as an investment product, they may trade real estate in the same way they trade ordinary securities. Majority of investment activities on the housing market consist of investment demands that resemble those of securities. Thus, how can we observe if investment demand of the housing market retains the characteristics of breakeven and defense of traditional housing markets? This paper proposes that the housing market that consists of more investment demands will contain more price concessions. This study infers from the three abovementioned hypotheses that even if housing supply conditions remain unchanged, housing prices will continue to self-correct through housing demand. Through the self-correcting pattern, this paper proposes that housing price changes will guide the market toward a new equilibrium. Fig. 1 illustrates these three hypotheses and the self-correction process of housing prices. Data and empirical tests Housing price dynamics This study uses panel data of five major cities in Taiwan (i.e., Taipei city, Taipei county, Taoyuan and Hsinchu area, Taichung city, and Kaohsiung city) over the period of 2003Q1 to 2009Q2. The sample size is 130. Housing price indices provided by Cathay Real Estate Development Co. are adopted to measure the performance of housing markets.

Information in the panel data is more adequate than both the time series and cross data. Particularly in recent years, several studies have been devoted to developing panel-based tests (e.g., Breitung, 2000; Im, Pesaran, & Shin, 2003; Levin, Lin, & Chu, 2002). These studies demonstrate that panel tests are more powerful than tests applied to an individual series because information in the time series is enhanced by those contained in cross-section data. Therefore, this paper uses panel data and panel-based models to draw more general conclusions. Before testing whether a self-correction pattern exists in the Taiwan housing market, this paper first needs to analyze the dynamic behavior in housing prices. Previous studies document that serial correlation and mean reversion are present in housing price dynamics. The panel-based unit root test is used to test the stationarity of housing price. Behavior of a stationary variable should be stable, and its mean and variance are constant over time. Therefore, if housing price is a stationary variable, then behavior of mean reversion should occur more often. This paper uses the test proposed by Im et al. (2003), and illustration of the test is described in the Appendix 1. Table 1 shows statistics on housing price indices in the five main cities in Taiwan (i.e., Taipei city, Taipei county, Taoyuan and Hsinchu area, Taichung city, and Kaohsiung city). Fig. 2 illustrates the time series of housing price indices in these five regions, while Table 2 shows the results of the panel-based unit root test. Fig. 2 demonstrates that within the data period, housing prices in the five areas exhibited a rising trend. For analysis, this study used data from the Taiwanese real estate market between 2003 and 2010. Taiwan is an emerging market, and both its economic growth and real estate market development have been rapid. The five main cities in Taiwan have always shown the most rapid economic development. Taipei City comprises the financial center of Taiwan and has the highest property values. Housing prices in Taipei City, in particular, have grown at an astounding rate. However, in some cities (i.e., Kaohsiung), housing prices only increased for a certain period of time and subsequently plummeted. Therefore, whether or not housing prices exhibited mean reversion is difficult to deduce from the figure. The results in Table 2 reveal that whether viewed from a model including a time trend or a model without a time trend, housing

Fig. 1. Self-correction pattern in housing price.

I-C. Tsai / Habitat International 40 (2013) 73e81 Table 1 Descriptive statistics of housing price indices.

77

Table 2 Unit root test for housing price.

Housing price indices

Taipei city

Taipei county

Taoyuan and Hsinchu area

Taichung city

Kaohsiung city

Mean Std. dev. Skewness Kurtosis

92.47 18.71 0.16 1.32

92.02 13.56 0.05 1.48

93.91 8.53 0.14 1.65

96.89 8.36 0.55 2.20

101.26 5.97 0.20 3.40

prices are non-stationary. These results are identical to those of research conducted in other regions, indicating that mean reversion in housing prices is not apparent. However, Table 2 shows that although statistics fail to reject the unit root hypothesis at the 1% significance level for the series in levels, the same tests deliver strong stationary evidence for the series in differences. These results suggest that the housing prices are I(1) series, that is, housing price return is stationary with the mean-reverting process. As mentioned above, this paper proposed that based on the selfcorrecting pattern driven by housing demand, housing price changes will guide the market toward a new equilibrium. The mean reversion behaviors in housing returns will verify these characteristics of price dynamics. As such, this study uses housing price return to observe its autocorrelation. Table 3 illustrates the autoregressive model of housing price return. Data reveal that housing price return was positively correlated with return on housing prices in the previous quarter and the fourth previous quarter (i.e., one year ago). Housing price return was negatively correlated with return on house prices in the third previous quarter and the eighth previous quarter (two years ago). In other words, autocorrelation of housing price return followed a sequence: positive, negative, positive, negative. This result shows that the dynamics of housing price return in Taiwan are similar to those in other markets, and exhibit serial correlation and mean reversion. Return on housing prices is positively auto-correlated in the short term and negatively auto-correlated over the long term. Results in Table 3 indicate the existence of an inherent selfcorrecting pattern in housing price return. This paper goes on to verify the three abovementioned hypotheses, and to test whether these can explain the housing dynamics shown in Table 3. Proxy variables and empirical results This study utilizes data from the Taiwan housing market to observe whether a self-correction pattern driven by housing demand side occurs and whether the housing dynamics can be explained by this pattern. In this section, this paper attempts to verify the three hypotheses and uses the results to describe the

Fig. 2. Time series of housing price indices in five regions.

Housing price Variable in level IPS test (with intercept) IPS test (with intercept and trend)

0.82 0.50

Variable in difference IPS test (with intercept) IPS test (with intercept and trend)

5.43*** 3.77***

Notes: IPS test denotes the test proposed by Im et al. (2003). ***Denotes statistical significance at the 1% level.

dynamics of housing price return in Taiwan. The following variables are used for estimation in the empirical research. First, this study uses debt burden ratio (the ratio between mortgage payment and income) as a proxy variable for housing affordability. As proposed by Gan and Hill (2009), debt burden ratio is a variable used to measure repayment affordability. Moreover, according to Hulchanski (1995), the variation of a certain proportion of household income to obtain suitable housing can measure the housing affordability problem. Rate of self-occupancy housing demand is used to measure changes in housing demand. This rate is derived from a questionnaire survey administered to citizens purchasing houses in the five main cities of Taiwan.4 In the questionnaire, respondents indicated whether their motivation for buying a house was self-occupancy or investment. When combined, the proportion of buyers who purchased housing for self-occupancy and the proportion of buyers who purchased houses for investment equaled 1. Therefore, the rate of self-occupancy housing demand demonstrates whether or not the proportion of housing demand for self-occupancy purpose exceeds housing demand for investment purpose. Variation in the rate of self-occupancy housing demand reveals a structural change in housing demand. Lastly, the trend of residential housing prices is used to measure the risk of a slide in housing prices. This is a statistic calculated from the score structure that indicates the perspectives of citizens in five major cities in Taiwan with regard to housing prices. The score structure is constructed according to the Consumer Confidence Index, with 100 as the reference point. A score exceeding 100 indicates that the respondent felt that housing prices have a higher possibility of rising; a score below 100 indicates that the respondent felt that housing prices have a higher possibility of falling. This paper also attempts to test whether the structural changes in housing market demand can cause self-correction in housing prices. Thus, it is more appropriate to add variables from the housing supply in the model to control effects from the supply side. Hence, the quantity supplied for new housing is used as a controlling variable, which stands for the supply of new housing by builders. The variable is an index constructed by Cathay Real Estate Development Co. and Taiwan Real Estate Research Center at National Chengchi University. This study uses data from five main cities in Taiwan for the abovementioned variables. The data period and frequency are identical to those of the housing price data used in the previous section (from 2003Q1 to 2009Q2). The data source is the database of the Taiwan Economic Journal. Table 4 shows simple data statistics and the results of unit root tests. Table 4 shows that on average, mortgage payments account for almost 30% of income. Regarding the intention behind buying a house, 80% of cases involve buyers who intended to live in the

4 The report of questionnaire survey is provided by Cathay Real Estate Development Co. and Taiwan Real Estate Research Center at National Chengchi University.

78

I-C. Tsai / Habitat International 40 (2013) 73e81

Table 3 P Results of autoregressive model (AR model). Model: RPi;t ¼ a0 þ ni¼ j aj RPi;tj þ 3i;t , where RP denotes the series of a housing return and n is the number of lag autoregressive terms. Variable

Coefficient

Std. error

t-Statistic

p-Value

Constant RPt1 RPt2 RPt3 RPt4 RPt5 RPt6 RPt7 RPt8

1.0603 0.3142 0.0196 0.2468 0.3550 0.2204 0.1221 0.1633 0.3032

0.4704 0.1187 0.1240 0.1239 0.1298 0.1330 0.1503 0.1486 0.1436

2.2538 2.6457 0.1579 1.9928 2.7362 1.6569 0.8119 1.0994 2.1113

0.0271 0.0099 0.8750 0.0499 0.0077 0.1017 0.4194 0.2751 0.0380

Schwarz Information Criterion

5.1781

Adjusted R-squared

0.1962

houses themselves. Housing prices are expected to rise by approximately 103 points, indicating that on average, people expect the performance of the housing market to improve in the future. The results of unit root tests reveal that these three variables are stationary. Therefore, the paper uses the original data of these three variables and housing prices after the first difference (i.e., return on housing prices or housing price variation) for estimation. Firstly, to verify the first hypothesis, Table 5 shows the estimated results of the relationship between debt burden ratio and housing price return. As demonstrated in the table, housing price return in the current and previous quarters did not significantly influence the burden of affordable housing. However, increases in housing price returns of the second previous quarter (six months before), third quarter (nine months before), and fourth quarter (one year before) significantly increased the burden of affordable housing. This lagged reaction of housing price return in the burden of affordable housing is rational because real estate transactions and debts develop over a longer period of time. Results in Table 5 verify the first hypothesis in this study; in particular, when housing prices rise, housing affordability (i.e., repayment affordability) decreases. Households are compelled to pay more of the proportion of their income to obtain suitable housing. The result is obtained by excluding the effect from the supply side. Next, this study estimates the relationship between housing demand for self-occupation purposes and the burden of affordable housing. Table 6 shows that an increase in the burden of affordable housing during the first and second previous quarters significantly

Table 4 Descriptive statistics and unit root test for the proxy variables. Descriptive statistics

DBR (ratio)

SOHD (ratio)

THP (index)

Qs (index)

Mean Std. dev. Skewness Kurtosis

28.95 5.51 0.83 3.35

81.88 4.86 0.45 3.20

103.81 19.88 1.19 4.18

83.18 32.64 0.19 2.86

Variable in level IPS test (with intercept) IPS test (with intercept and trend)

DBR 4.99*** 7.13***

SOHD 4.93*** 4.16***

THP 0.04 7.55***

Qs 3.29*** 0.20

12.55*** 9.96***

10.12*** 9.05***

10.19*** 8.86***

8.75*** 11.28***

Variable in difference IPS test (with intercept) IPS test (with intercept and trend)

Notes: IPS test denotes the test proposed by Im et al. (2003). DBR, SOHD, and THP denote the debt burden ratio, rate of self-occupancy housing demand, and trend of residential housing prices, respectively. Qs is the quantity supplied for new housing. ***Denotes statistical significance at the 1% level.

Table 5 P Housing price and housing affordability. Model: DBRi;t ¼ a0 þ nj¼ 0 aj RPi;tj þ b0 Qsi;t þ 3i;t , where DBR denotes the debt burden ratio, RP denotes the series of a housing return, Qs is the quantity supplied for new housing, and n is the optimal lag length. Variable

Coefficient

Std. error

t-Statistic

p-Value

Constant RPt RPt1 RPt2 RPt3 RPt4 Qst

27.8381 0.1864 0.3249 0.3897 0.4987 0.5082 0.0057

1.6417 0.1930 0.2002 0.2058 0.2170 0.2120 0.0181

16.9569 0.9653 1.6228 1.8941 2.2976 2.3969 0.3163

0.0000 0.3368 0.1078 0.0612 0.0237 0.0184 0.7525

Schwarz Information Criterion

6.2598

Adjusted R-squared

0.1686

Notes: debt burden ratio, or the ratio between mortgage payment and income, is a proxy variable for housing affordability. Optimal lag length is chosen by the Bayesian information criterion.

reduced the number of housing purchases for which selfoccupancy was the motivation. This shows that if buyers continue to purchase houses when the burden of affordable housing is heavy, they are likely to be motivated by investment. This result verifies the second hypothesis in this study. Lastly, structural changes in housing market demand have an influence on whether housing prices can be reduced. Table 7 illustrates the relationship between the rate of housing demand for self-occupancy and trend of housing prices. The results in Table 7 reveal that in the current and previous quarters, when self-occupancy housing demand was higher, people felt that the continuously rising trend of housing prices would reverse itself because the regression coefficient was significantly negative. Interestingly, when self-occupancy housing demand was higher in the previous third and fourth quarters, housing prices rose significantly. These results demonstrate that the relationship between self-occupancy housing demand and trend of housing prices differs over the short and long term. The two variables have a negative short-term relationship but a positive long-term relationship. These results are entirely rational because any increase in demand, whether for investment or self-occupancy, increases market prices. However, transaction factors of these two types of buyers are very different. If majority of buyers in the market purchase houses for self-occupancy, the related consumption characteristics will come into play, and the characteristics of housing prices will differ from those of other assets. Conversely, if most buyers purchase houses due to investment demand, performance of housing Table 6 Housing affordability and self-occupancy housing demand. Model: SOHDi;t ¼ a0 þ Pn j ¼ 0 aj DBRi;tj þ b0 Qsi;t þ 3i;t , where SOHD denotes the rate of self-occupancy housing demand, DBR denotes the debt burden ratio, Qs is the quantity supplied for new housing, and n is the optimal lag length. Variable

Coefficient

Std. error

t-Statistic

p-Value

Constant DBRt DBRt1 DBRt2 DBRt3 Qst

96.3785 0.0448 0.2140 0.3591 0.0074 0.0044

2.5231 0.1250 0.1270 0.1303 0.1001 0.0136

38.1984 0.3589 1.6848 2.7553 0.0735 0.3217

0.0000 0.7204 0.0949 0.0069 0.9416 0.7483

Schwarz Information Criterion

5.8522

Adjusted R-squared

0.2831

Notes: debt burden ratio, or the ratio between mortgage payment and income, is a proxy variable for housing affordability. Optimal lag length is chosen by the Bayesian information criterion.

I-C. Tsai / Habitat International 40 (2013) 73e81 Table 7 Self-occupancy housing demand and the trend of residential housing prices. Model: P THPi;t ¼ a0 þ nj¼ 0 aj SOHDi;tj þ b0 Qsi;t þ 3i;t , where THP denotes the trend of residential housing prices, which is used to measure the risk of a slide in housing prices; SOHD denotes the rate of self-occupancy housing demand; Qs is the quantity supplied for new housing; and n is the optimal lag length. Variable

Coefficient

Std. error

t-Statistic

p-Value

Constant SOHDt SOHDt1 SOHDt2 SOHDt3 SOHDt4 Qst

96.0211 1.2150 1.5409 0.2259 1.5420 1.5414 0.0386

39.8776 0.4112 0.4305 0.4603 0.4818 0.4339 0.0628

2.4079 2.9549 3.5797 0.4907 3.2002 3.5522 0.6141

0.0178 0.0039 0.0005 0.6247 0.0018 0.0006 0.5405

Schwarz Information Criterion

8.7520

Adjusted R-squared

0.2917

Notes: the optimal lag length is chosen by the Bayesian information criterion.

prices will match the fluctuations of prices of other regular assets. Buyers who purchase a home for self-occupancy may only purchase one house in their lifetime; thus, they are unlikely to engage in short-term transactions. However, buyers who purchase houses for investment are more likely to be influenced by short-term factors. Therefore, in the short term, if the proportion of housing demand for self-occupancy exceeds housing demand for investment purposes, investors who engage in short-term transactions will be less optimistic about the market, as they will feel that a greater possibility of a short-term slide exists in housing price. If, over the long term, the proportion of housing demand for self-occupancy exceeds housing demand for investment, fewer short-term investors and more stable buyers are in the market. Therefore, over the long term, these conditions make house prices less likely to decrease and more likely to increase. The results shown in Table 7 support the third hypothesis in this study: structural changes in housing demand influence the continuity of variation in house prices. Results likewise reveal that an increase in housing demand for self-occupancy does not have a particularly positive influence on housing prices. Therefore, according to Tables 5e7, the three hypotheses in this study are all verified. Based on these data, the cycle of variation in housing prices can be observed. Empirical evidence shows that when housing prices rise, housing affordability decreases, followed by a reduction in housing demand for self-occupancy. Furthermore, change in demand structure raises the risk of prices dropping because of an increase in investment-motivated housing demand, eventually resulting in lower housing prices. The results in Tables 5e7 show the causal effects between housing return and affordability, between housing affordability and self-occupancy housing demand, and between self-occupancy housing demand and housing price, respectively. Bidirectional Granger causality tests are also performed to serve as the robustness test for these causal relations, and the results are shown in Appendix 2. The results of the Granger causality tests are consistent with those in Tables 5e7, supporting the three hypotheses in this study.

79

correction in housing prices. The pattern can be described as follows: an increase in housing prices will increase mortgage payment, cause a reduction of housing affordability, and then decrease housing demand for self-occupancy while housing demand for investment increases. A change in demand structure causes housing prices to rise in the short term and to decrease in the long term. The hypotheses are examined using the panel data of five major cities in Taiwan. The hypotheses are supported by empirical evidence under controlling factors from the supply side. The empirical evidence can explain why housing price return is positively autocorrelated in the short run and negatively auto-correlated in the long run, demonstrating why the house price processes contain serial correlated and mean reversion behavior. This study proposes that the variables from the demand side can initiate the self-correcting mechanism of housing prices. The correlation of these variables demonstrates a lagged reaction, which indicates that while housing prices will self-correct, self-correction may be delayed for many quarters. The result is a positive autocorrelation of housing prices in the short term and a negative autocorrelation over the long term after self-correction becomes apparent. When housing price is high, the government subsidizes households to increase housing demand for self-occupancy. This policy, however, can allow the continuous increase of housing prices, extend the time period during which self-correction occurs, and increase the amplitude and persistence of cycles. Therefore, the implication of this paper is consistent with those of previous papers that proposed that the excessive subsidy policy can cause housing market inefficiency (Shiller, 2009). Moreover, the easing of monetary policy is the cause of the housing market imbalance in the US. This paper claims that if easing the monetary policy causes the proportion of housing demand for investment to increase, then the risk of housing prices dropping will increase. Investment-motivated housing demand increases market prices only in the short run; thus, traders who purchase houses for investment are likely to sell eventually when they realize the capital gains. Thus, investment demand will increase the fluctuations of housing prices and cause the housing market to become more unstable. The findings of this paper provide a theoretical basis for the government to control housing market instability, which is increased by rising investmentmotivated demand. In contrast to previous studies that discussed a self-correcting mechanism in the housing market resulting from the adjustment by the supply side, this study proposed the self-correction pattern driven by housing demand to explain the housing dynamics. A more comprehensive and precise analysis on the self-correcting mechanism of housing prices can thus be derived from this study and added on to the previous literature in housing price dynamics. Acknowledgments Funding from the National Science Council of Taiwan under Project No. NSC 100-2628-H-390-002 has enabled the continuation of this research and the dissemination of these results.

Conclusion Appendix 1. The illustration of the panel unit root tests Previous studies (Roulac, 1996) have mentioned a selfcorrecting mechanism in the housing market resulting from the adjustment by the supply side. This paper utilizes data from the Taiwanese housing market to determine whether a self-correction pattern driven by the housing demand side occurs as well. First, this study proposes three hypotheses to state how structural changes in housing market demand may result in self-

The panel unit root tests developed by Levin et al. (2002) and Breitung (2000) are similar to tests conducted on a single series because they all assume a common unit root process across relevant cross-sections. The LLC and Breitung tests employ a null hypothesis of a unit root using the following basic Augmented Dickey Fuller (ADF) specification:

80

I-C. Tsai / Habitat International 40 (2013) 73e81

yi;t ¼ ryi;t1 þ

X

fij Dyitj þ xi;t g þ vi;t

(A1)

where yi,t stands for the endogenous variable; xi,t indicates exogenous variables in the model, such as city fixed effects and individual time trends; and g is the corresponding vector of coefficients. The symbol vi,t refers to the error terms assumed to be mutually independent disturbances. Notably, r is assumed to be identical across the cross-sections, but the lag order for the difference terms across the four sectors is allowed to vary. By contrast, the test proposed by Im et al. (2003) allows r to vary across cross-sections. In particular, ADF regressions can be modified as follows:

yi;t ¼ ri yi;t1 þ

X

fij Dyitj þ xi;t g þ vi;t

(A2)

The less restrictive IPS statistic (Im et al., 2003) is based on averaging individual ADF unit root tests (ti) according to the following formula:

tIPS

pffiffiffiffi  N t  E½ti jri ¼ 0 pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi /Nð0; 1Þ ¼ var½ti jri ¼ 0

(A3)

P where t ¼ N 1 N i ¼ 1 ti . The IPS test is less restrictive and is therefore used in this research to determine the stationarity of data.

Appendix 2. Granger causality tests Null hypothesis:

F-statistic

Prob.

DBR does not Granger Cause RP RP does not Granger Cause DBR SOHD does not Granger Cause DBR DBR does not Granger Cause SOHD THP does not Granger Cause SOHD SOHD does not Granger Cause THP

0.7658 2.5400 13.9772 6.9052 4.6444 4.7091

0.5500 0.0445 0.0000 0.0015 0.0044 0.0041

Notes: DBR denotes the debt burden ratio, RP denotes the series of a housing return, SOHD denotes the rate of self-occupancy housing demand, THP denotes the trend of residential housing prices, which is used to measure the risk of a slide in housing prices.

References Abraham, J., & Hendershott, P. H. (1993). Patterns and determinants of metropolitan house prices, 1977e91. In Browne, & Rosengreen (Eds.), Real estate and the credit crunch, proceedings of the 25th annual Federal Reserve Bank of Boston conference (pp. 18e42). Abraham, J., & Hendershott, P. H. (1996). Bubbles in metropolitan housing markets. Journal of Housing Research, 7, 191e207. Baxter, M. (1996). Are consumer durables important for business cycles. Review of Economics and Statistics, 78, 147e155. Breitung, J. (2000). The local power of some unit root tests for panel data. In B. Baltagi (Ed.). Advances in econometrics, 15: Nonstationary panels, panel cointegration, and dynamic panels (pp. 161e178). Amsterdam: JAI. Campbell, J., & Cocco, J. (2004). How do house prices affect consumption? Evidence from micro data. Paper presented at 2004 SED meeting. Capozza, D. R., Hendershott, P. H., Mack, C., & Mayer, C. J. (2002). Determinants of real house price dynamics. NBER working paper series 9262. Capozza, D. R., & Seguin, P. J. (1996). Expectations, efficiency, and euphoria in the housing market. Regional Science and Urban Economics, 26, 369e386. Case, K., Quigley, J., & Shiller, R. (2001). Comparing wealth effects: The stock market versus the housing market. Cowles Foundation discussion paper 1335. Yale University. Case, K. E., & Shiller, R. J. (1989). The efficiency of the market for single family homes. The American Economic Review, 79, 125e137. Coulson, N. E. (2002). Regional and state variation in homeownership rates; or if California’s home prices were as low as Pennsylvania’s would its ownership rate be as high? Journal of Real Estate Finance and Economics, 24(3), 261e276. Davis, M., & Heathcote, J. (2005). Housing and the business cycle. International Economic Review, 46, 751e784. Drucker, P. E. (1993). Post-capitalist society. New York: Harper Business. Dusansky, R., & Koc, C. (2007). The capital gains effect in the demand for housing. Journal of Urban Economics, 61(2), 287e298.

Elbourne, A. (2008). The UK housing market and the monetary policy transmission mechanism: an SVAR approach. Journal of Housing Economics, 17(1), 65e87. Fisher, J. (1997). Relative prices, complementarities and comovement among components of aggregate expenditures. Journal of Monetary Economics, 39, 449e474. Follian, J. R. (1982). Does inflation affect real behavior: the case of housing. Southern Economic Journal, 48(3), 570e582. Friedman, M. (1976). Price theory. Chicago: Aldine Publishing Company. Gan, Q., & Hill, R. J. (2009). Measuring housing affordability: looking beyond the median. Journal of Housing Economics, 18, 115e125. Genesove, D., & Mayer, C. (2001). Loss aversion and seller behavior: evidence from the housing market. Quarterly Journal of Economics, 116, 1233e1260. Glaeser, E. L., Gyourko, J., & Saiz, A. (2008). Housing supply and housing bubbles. Journal of Urban Economics, 64, 198e217. Glennon, D. (1989). Estimating the income, price and interest elasticities of housing demand. Journal of Urban Economics, 25, 219e229. Goodhart, C., & Hofmann, B. (2008). House prices, money, credit, and the macroeconomy. Oxford Review of Economic Policy, 24(1), 180e205. Goodman, A. C. (1988). An econometric model of housing price, permanent income, tenure choice and housing demand. Journal of Urban Economics, 23(3), 327e353. Grange, A. L., & Pretorius, F. (2000). Ontology, policy and the market: trends to home-ownership in Hong Kong. Urban Studies, 37(9), 1561e1582. Green, R. K. (2002). Stock prices and house prices in California: new evidence of a wealth effect? Regional Science and Urban Economics, 32, 775e783. Greenwood, J., & Hercowitz, Z. (1991). The allocation of capital and time over the business cycle. Journal of Political Economy, 99, 1188e1214. Haurin, D. R., & Kamara, D. A. (1992). The homeownership decision of femaleheaded households. Journal of Housing Economics, 2(4), 293e309. Himmelberg, C., Mayer, C., & Sinai, T. (2005). Assessing high house prices: bubbles, fundamentals, and misperceptions. Journal of Economic Perspectives, 19(4), 67e92. Huang, D. S. (1969). Effect of different credit policies on housing demand. In I. Friend (Ed.), Study of the savings and loan industry (pp. 1211e1239). Washington, DC: Federal Loan Home Bank Board. Hulchanski, J. D. (1995). The concept of housing affordability: six contemporary uses of the housing expenditure-to-income ratio. Housing Studies, 10(4), 471e491. Iacoviello, M. (2005). House prices, borrowing constraints and monetary policy in the business cycle. American Economic Review, 95(3), 739e764. Im, K. S., Pesaran, M. H. P., & Shin, Y. (2003). Testing for unit roots in heterogeneous panels. Journal of Econometrics, 115, 53e74. Jud, D., & Winkler, D. (2002). The dynamics of metropolitan housing prices. Journal of Real Estate Research, 23(1e2), 29e45. Kau, K. W., Chang, C. O., & Lin, V. (2000). An analysis of reasonable housing prices in Taipei from demand side. In 9th Annual conference housing study (in Chinese). Lamont, O., & Stein, J. (1999). Leverage and house price dynamics in US cities. RAND Journal of Economics, 30, 498e514. Leamer, E. (2007). Housing is the business cycle. NBER working paper, 13428. Leinberger, C. B. (1993a). Strategic planning for real estate companies. Urban land, 15. Leinberger, C. B. (1993b). Strategies for real estate companies: Marketing, finance, organization. Washington, D.C.: Urban Land Institute and National Association of Industrial and Office Properties. Leung, C. K. Y. (2004). Macroeconomics and housing: a review of the literature. Journal of Housing Economics, 13, 249e267. Levin, A., Lin, C. F., & Chu, C. S. (2002). Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics, 108, 1e24. Malpezzi, S. (1990). Urban housing and financial markets: some international comparisons. Urban Studies, 27(6), 971e1022. Malpezzi, S. (1999). A simple error correction model of housing prices. Journal of Housing Economics, 8, 27e62. Marshall, M. I., & Marsh, T. L. (2007). Consumer and investment demand for manufactured housing units. Journal of Housing Economics, 16, 57e71. Memery, C. (2001). The housing system and the Celtic Tiger: the state response to a housing crisis of affordability and access. European Journal of Housing Policy, 1(1), 79e104. Mishkin, F. S. (2007). Housing and monetary transmission mechanism. Paper presented at: Housing finance and monetary policy symposium, Federal Re-serve Bank of Kansas City (pp. 359e413). Muellbauer, J., & Murphy, A. (2008). Housing markets and the economy: the assessment. Oxford Review of Economic Policy, 24(1), 1. Muth, R. F. (1960). The demand for non-farm housing. In A. C. Harberger (Ed.), The demand for durable goods (pp. 27e96). Chicago: University of Chicago Press. Quan, C. D., & Titman, S. (1999). Do real estate prices and stock prices move together? An international analysis. Real Estate Economics, 27, 183e207. Riddel, M. (2004). Housing-market disequilibrium: an examination of housingmarket price and stock dynamics 1967e1998. Journal of Housing Economics, 13(2), 120e135. Roulac, S. E. (1996). Real estate market cycles, transformation forces and structural change. Journal of Real Estate Portfolio Management, 2(1), 1e17. Roulac, S. E., & Volk, L. D. (1989). Deal-making strategies in the new era. Real Estate Finance, 5(4), 19e31. Shiller, R. J. (1993). Macro markets: Creating institutions for managing society’s largest economic risks. Oxford University Press. Shiller, R. J. (2005). Irrational exuberance (2nd ed.). Doubleday.

I-C. Tsai / Habitat International 40 (2013) 73e81 Shiller, R. J. (2009). The subprime solution. Princeton: Princeton University Press. Sim, S.-H., & Chang, B.-K. (2006). Stock and real estate markets in Korea: wealth or credit-price effect. Journal of Economic Research, 11, 99e122. Skinner, J. (1989). Housing wealth and aggregate saving. Regional Science and Urban Economics, 19(2), 305e324. Skinner, J. (1996). Is housing wealth a sideshow. In W. David (Ed.), Economics of aging). Chicago: Univesity of Chicago Press. Smith, M., & Smith, G. (2006). Bubble, bubble, where’s the housing bubble?. Draft for the Brookings Panel on Economic Activity.

81

Tsai, I-C., Lee, C. F., & Chiang, M.-C. (2012). The asymmetric wealth effect in the US housing and stock markets: evidence from the threshold cointegration model. Journal of Real Estate Finance and Economics, 45, 1005e1020. Tsai, I-C., & Peng, C.-W. (2011). Bubbles in the Taiwan housing market: the determinants and effects. Habitat International, 35, 379e390. Vargas-Silva, C. (2007). Monetary policy and the US housing market: a VAR analysis imposing sign restrictions. Journal of Macroeconomics, 30, 977e990. Yamada, Y. (1999). Affordability crises in housing in Britain and Japan. Housing Studies, 14(1), 99e110.