Long-run Relationship Between REITs and Factors

2.3 Long-run Relationship Between REITs and Factors

This study applies cointegration method in exploring the long-run equilibrium relationship between REITs and other factors. Variables with cointegration relation are adjusted by a vector error correction term, and variables without cointegration relation are interpreted based on VAR approach. Finally, all the variables will do the causality test. This section reviews the researches about cointegration and VAR models.

2.3.1 Cointegration

Other than using linear regression model to analyze the relation between REITs and other markets, many researches shift to using cointegration to discuss the long-run equilibrium relationship between REITs and other factors in recent years. This concept was proposed by Engle and Granger (1987), where they explore the long-run relationship by using a two-step approach between the two variables. Later, Johansen (1991) proposed the maximum likelihood method for verification of cointegration relationship between multi-variables.

Nasseh and Strauss (2000) indicate that the Johansen framework is a useful setting for analyzing stock market and macroeconomic activity because it incorporates dynamic co-movements or simultaneous interactions. Through this method we can study the channels through macroeconomic variables affect asset prices as well as their relative importance. They find a strong, integrating relationship between stock prices and domestic and international macroeconomic variables in Europe, and the long-term interest rates are shown to negatively influence stock prices; whereas, short-term interest rates are shown to be positively related to stock prices. In addition, stock prices are grounded in economic fundamentals, influenced by production, interest rates, business expectations and the CPI, and the domestic and international macroeconomic activity and cointegration methodology can be an important tool in explaining stock returns. McCue and Kling (1994), Okunev and Wilson (1997) also state that cointegration model is commonly used in the analysis of correlation between real estate market and stock market, focusing on the long-term effects of the economic variables.

Liu, Hartzell, Greig and Grissom (1990), exploring the relationship between real estate market and the stock market, pointed out that the commercial real estate market

‧

is segmented from the stock market as a result of indirect constraints. However, the equity REITs are integrated with the stock market even though the commercial real estate that underlies the equity REITs is segmented from the stock market, and indirect barriers such as the cost, amount, and quality of information are the major source of segmentation. The commercial real estate market appears to be cointegrated with the stock market when the tests of cointegration are performed. Li and Wang (1995) apply the method of two-factor asset pricing model and risk premium. The result lead to a similar conclusion without assuming the existence of a real estate factor premium in the economy, providing further evidence that REIT stocks are integrated with the general stock market.

Ling and Naranjo (1999) support the hypothesis that markets for exchange-traded real estate companies, like REITs, are integrated with the market for exchange-traded (non-real-estate) stocks, and the level of integration significantly increase after 1990s. However, if the evaluated price is replaced by the actual market price, it will have an opposite result from the perspective of real estate portfolio, which means the real estate market and stock market will not be integrated. The hypothesis is that the estimated returns can’t accurately represent the actual portfolio returns.

Quan and Timan (1999) study the global relation between stock markets and real estate markets in 17 countries. The results show that apart from Japan, the stock markets and real estate markets are not cointegrated in other countries. However, the stock returns, real estate price changes and rentals would have a significant relationship if all the national data are added up and examined under a longer period.

In the long-run, stock returns and real estate returns are still correlated. Chaudhry, Myer and Webb (1999) further use Johansen test and the results suggest that the stocks tend to have an inverse long-run relationship with the real estate.

Johnl, Lu and So (2000) believe that if REITs act more like real estate, there would be a cointegration between inflation and REITs, but they find that before 1992, REITs acted more like fixed-income instruments. The results show that REITs are cointegrated with the bond market before 1992 and that this cointegration relationship vanished afterwards. From 1992 to 1996, the cointegration are found between stocks and REITs, and after 1992, stocks and REITs share some common factors, and asset pricing structures and REITs behave more like stocks. Clayton and Mackinnon (2003)

‧

also find that the dramatic growth and maturation of the REIT sector since 1992 led to claims that the link between REIT prices and real estate market fundamentals had become much stronger as their returns become less dependent on major stock indices and more akin to real estate and sector effects, making REITs more like real estate and less like stock.

Westerheide (2006) states that real estate stocks provide a (weak) hedge against consumer price inflation in almost every country, and cointegration is indicated in the Johansen specification as well. This implies that an equilibrium relationship exists, but the general stock market adjusts to the real estate stock market instead of the reverse. In the study of the U.S., Australia and Japan, there is weak evidence for a long-run equilibrium between real estate stock indicators and the CPI, indicating that real estate stocks could basically serve as an inflation hedge. Adrangi, Chatrath and Raffiee (2004) support the hypothesis the market for REITs and equities is integrated and the real REIT returns are negatively correlated with the inflation rate, and the result is robust for the long-term.

The cointegration literatures discussed above are under a linear assumption.

However, Okunev and Wilson (1997) suggest that using a standard cointegration tests β to produce conclusive evidence could be a failure because the relationship between real estate and financial assets markets may be nonlinear rather than linear. The results suggest that the real estate markets are nonlinearly related to the stock market, so there is a relationship between the stock and real estate markets but the link is weak and nonlinear, and the movement of the real estate market towards the stock market is slow. In 1999, the authors further use non-linear method to explore the long-run equilibrium relationship between the real estate market and stock market of the United States, Britain and Australia. The results show that the real estate market and the stock market do not have a significant long-run equilibrium relationship in the U.S. and the U.K., and the result in Australia is not significant. However, if the cutoff point is in 1987, the real estate market in U.S. and the U.K. will be cointegrated.

There are also some domestic researches for cointegration between different markets. Zheng (2008) illustrates that cointegration test is often interpreted as “The economic variables have a long-run equilibrium relationship.” If the linear combination between variables has a cointegration relationship, even a short period of imbalance with a deviation from equilibrium occurred, but the error correction

‧

function will gradually reduce this deviation and eventually adjust to the long-run equilibrium level. If cointegration exists between variables, then the error correction term must be added to fix the short-run non-equilibrium and to explain the short-run changes of the series in inter-relations.

Zheng and Chang (2007) point out that the T-REITs index and stock price index are not cointegrated, neither is T-REITs and construction index, suggesting that the investors can achieve risk diversification and gain profits by joining T-REITs in portfolio investment. In addition, the T-REITs have low correlation coefficient with stock price index and construction index. In this case, investors can effectively reduce the investment risk if T-REITs are included in the portfolio.

Wang (2007) shows that industrial/office, special use, retail and warehousing, and personal use of REITs do not have the characteristics against increasing price. The housing-based REITs, however, have the characteristics against increasing price, they are more able to offset the decline in purchasing power when prices rise. In addition, using unemployment as a macroeconomic indicator, the study found that unemployment only has negative effects on industrial/office-based REITs.

Nie and Zheng (2000) use general linear analysis and non-linear concept proposed by Okunev and Wilson to explore the relationship between housing price index and stock price index in Taipei City, Taipei County, Kaohsiung City, Taichung and Taiwan from March 1991 to April 1999. The results show that the long-term equilibrium relationship does not exist between housing price and stock price except in Kaohsiung City. In addition, the study found that during the study period, the interaction between Taiwan housing price and stock prices are mostly negative, which is contrary to other studies.

Overview of domestic and foreign cointegration literatures and researches of REITs, in foreign researches, the macroeconomic variables such as stock price, interest rate and inflation rate are cointegrated with REITs, thus they have long-term equilibrium relationship. However, the researches in Taiwan find that T-REITs are not cointegrated with stock price index. The foreign discussions about the ability of REITs against inflation do not reach any unanimous conclusion, but some studies suggest that it should not be a portfolio with similar nature products like stocks. In Taiwan, the studies suggest that although T-REITs is not an inflation hedge, they can

‧

serve as an effective tool for risk diversification because of its low stock price index-related.

2.3.2 Vector Autoregression

VAR model can be used to explore the effects and direction of short-term changes among REITs and other variables, and to understand the degree of influence of lag periods. Lee (1992) uses the VAR approach to explore the causal relations among stock returns, interest rate, real activity, and inflation. The results show that the stock returns appear prior to Granger-Cause and help explain a substantial fraction of the variance in real activity, and it responds positively to shocks in stock returns.

When adding interest rate in VAR model, the interpretation ability of stock returns on inflation is reduced, but the interest rate is with most of the explanatory power of inflation instead. In addition, the inflation rate has a negative relationship with stock returns when testing with real interest rates. Finally, the inflation explains little variation in real activity.

Ling and Naranjo (2003) apply VAR approach to examine the interrelationships between short- and long-run dynamics among capital flows to the REIT sector and REIT returns, particularly whether REIT capital flows affect REIT prices and returns and whether the effect is temporary or permanent. They also use impulse response functions to provide the time path of the short-run dynamic relationships that result from a shock to the variables in the system. The result suggests there is positive momentum in REIT flows, but this momentum reverses after two quarters. It also indicates that current flows are highly significant in explaining current returns.

Ling and Naranjo (2006) further use VAR to examine the effects of weekly and monthly capital flows into the dedicated REIT mutual fund sector on aggregate REIT returns. They find consistent evidence that REIT mutual fund flows are significantly and positively related to prior industry-level returns, but prior fund flows do not significantly influence subsequent REIT returns. In addition, contemporaneous fund flows do have an initial positive effect on returns, which is partially reversed one period later. Interestingly, the unexpected REITs mutual fund flows have positive contemporaneous effects and the expected portion is insignificant.

Bredin, O’Reilly and Stevenson (2007) apply VAR model to analyze the relationship between REITs and interest rate. They conclude that in comparison to

‧ 國

立 政 治 大 學

‧

N a tio na

l C h engchi U ni ve rs it y

21

previous studies of REIT interest rate sensitivity, this study find significant responses in both returns and volatility to unanticipated rate changes. However, the effect of the shock is significant on both returns and volatility and there is no evidence of asymmetry between them.

Glascoca, Lu and So (2002) use VAR to explore the relationship between REITs returns, real activity, monetary policies, and inflation. The evidence suggest that neither expected nor unexpected inflation signal REIT returns, but the finding is consistent with previous findings that REIT returns are sensitive to interest rate changes. In addition, they conclude that REITs returns do not behave as perverse inflation hedges.

To sum up the above studies, foreign researches about the relationship between REITs and other markets are more extensive, and apply different models to explore the long-run and short-run impacts. However, the relative empirical studies in Taiwan are still less prominent. Although there are some studies using cointegration model to discuss the long-run relationship between T-REITs and other variables, most of them focus on relationship between T-REITs and stocks or T-REITs and construction index.

There are other economic variables that can still be discussed as regards T-REITs, and that is the research gap this study would like to explore.

‧