Property Type Focus for Public Equity Reits Performance Finance Essay

Real Estate Investment Trusts (REITs) allow all investors, as opposed to only the most affluent ones to invest in real estate. Apart from this, shares in REITs are more liquid than actual property. Finally, double taxation is avoided, because REITs are exempt from corporate taxes if they distribute 90% of their taxable income in form of dividends [1] (among other conditions). The rapid expansion of equity REITs started in early 1990s and peaked in 2006 with about $400bn market capitalisation. [2] Given the apparent popularity, it is increasingly important to better understand this asset class. The aim of this paper is to find whether there are any differences in performance across REITs with different property-type focus. If there are no differences, investors would be wasting resources by analysing each REIT type separately. However, if the reverse is true, excess returns could be earned by increasing the share of a particular type in the portfolio. It would also be interesting, because it would violate the efficient market hypothesis. This study contributes to the existing discussion by investigating whether differences between property types emerged (or got amplified) in the recent market collapse. There is a number or reasons why this could be the case. For example, decrease of home-ownership rate resulting from the mortgage-crisis could have induced people to move to the rental sector. For residential REITs this would provide a cushion in form of an influx of extra demand for rented accommodation. As it can be seen from Figures 2 and 3, homeownership in the US fell below 67% in 2011 from almost 69% before the crisis. Rental vacancy rates increased during the crisis but this increase was not unprecedented. Possibly, if it was not for the ex-home-owners, the recent increase could have been more dramatic.

Figure 2 [3]

This effect is likely to be much lower for commercial real estate, as firms often prefer to rent rather than own their premises even in the favourable economic conditions, as this prevents tying up substantial amounts of capital [4] . As a result, when the crisis starts, they already were renters. As it can be seen from Figure 4, vacancies in the office sector almost doubled between 2007 and 2009 and net absorption was negative for two years. Also, it is often commercial real estate is more correlated with the unemployment rate and more vulnerable to shocks to the economy. By having a weaker link to the unemployment and GDP, residential REITs could have suffered less.

Figure 4 [5]

Industrial real estate faces also extra risk due to lack of tenant diversification. Warehouses and industrial premises tend to be large and typically occupied by a small number of tenants. Industrial REITs could have suffered from this problem in a downturn. Another major category is healthcare REITs. With increasingly aging population this seems to be an attractive investment. Nevertheless, any effect seen in the total returns data could be simply due to the 2010 healthcare reform, which implies that more people will be covered by the insurance in the US and therefore investors could expect more demand for healthcare property. On the other hand, healthcare REITs even before the reform had one of the lowest debt/capitalisation ratios (28.8% in 2009 compared to the average of 44.76% for all REITs) and offered relatively high returns. [6] More leveraged sectors could have found themselves is more trouble when the crisis started. Some researchers find that the type of REIT indeed determines the returns to investor. For example, Gyourko and Nelling (1996) regress betas on the percentage of properties of each type and find that retail REITs had significantly higher market sensitivity than industrial and warehouse REITs in the period of 1988-1992. Kim et al (2002) use Jensen index and perform ANOVA and find that hotel REITs underperformed other types. Redman and Manakyan (1995) regress Sharpe ratios of REITs on their characteristics. They find that property type focus together with geographical location and financial parameters are significant determinants of REITs' risk adjusted returns. Finally, Benefield, Anderson, Zupano (2009) calculate Treynor measure and Jensen alphas and test their significance using both parametric and non-parametric tests. They find that REITs with diversified portfolio of properties (with respect to property type) perform better when the market conditions are good. In the downturn, there is however little evidence of superiority of diversified REITs. However, other strands of research suggest that property type does not matter for the return. For example, Young (2000) finds that randomly chosen portfolios that do not take property type into account perform in the same way as property-type specific portfolios. This analysis was applied to 1989-98 data. Chiang et al (n. d.) use three models - CAPM, Fama-French and Cahart and apply it to data in the period of 1994-2005 and they conclude that there is no statistically significant differential in the risk adjusted performance of various types of REITs.

DATA

Data on monthly returns on 130 US public equity REITs was obtained from SNL Financial. SNL REIT indices were used to group REITs into "types". Details of what exact companies each type includes are available in Appendix 1. The returns include both: capital gains and the dividends. The types of REITs considered are the following: healthcare, hotels, industrial, diversified/other, office, retail, residential and self-storage. SNL US Equity REIT is an index that includes all publicly traded REITs. Measure of market portfolio used in this study is S&P 500. Monthly total returns on this index also come from the SNL Financial. Risk free rate and Fama-French factors were obtained from Kenneth R. French's [7] website. The factors are difference in returns on big and small stocks (SMB) and difference in returns on stocks with high and low book-to-value ratio (HML). Data covers April 2003 - January 2011 The frequency of the data is monthly, to clean the noise that can be found in the daily data. Also, this frequency is mostly used in the literature, so the study will be more comparable to others. Finally, yearly data would produce only few observations, which could raise concerns that the results are not reliable. In terms of the reliability of data, SNL Financial seems to be a relatively safe source, widely used by numerous investment banks. The only problem with the SNL is that it does not cover the pre-2000 period. No comparison with the previous periods can be therefore made. The last issue is that of division of data into PRE-CRISIS and CRISIS periods. 30 April 2007 was chosen as the division-point. The decision was based on the fact that most REITs started to offer negative returns at this time. Finally, it could be claimed that CRISIS covers the recovery time (when the recession ended). This is simply a nomenclature issue and inclusion of the recovery time is the intention of the author, as it is assumed that the crisis could have some long-lasting effects on the ability of certain REIT types to outperform others. Descriptive statistics for the period of April 2003 - April 2007 (referred to as PRE-CRISIS from now on), period of May 2007 - January 2011 (referred to as CRISIS) are presented in Tables 1-2. The mean return for most REITs was higher than that of S&P 500. However S&P 500 also carried lower risk. Among REIT types, self-storage had the highest mean return of almost 2.48% per month, but the standard deviation of those returns was also relatively large. Healthcare REITs offered large returns as well. The lowest mean return was on residential type - 2.06% per month. What strikes here is that the means do not seem to differ dramatically across types. Simple comparison of returns is not an appropriate method of investigating the problem. Higher return could be just a reward for a higher risk. There are three most popular risk-adjusted measures of performance: Sharpe ratio, Treynor ratio and Jensen alpha. They all benefit from having an intuitive interpretation. Sharpe and Treynor ratios are similar - they are simply excess return divided by risk. Risk in the Sharpe ratio is the total risk, while Treynor ratio only takes into account market risk. Jensen alpha, on the other hand, informs of the extra return that cannot be explained by the market excess returns. In the downturn, Sharpe and Treynor ratio can be negative, which makes results difficult to interpret. Furthermore, Sharpe ratio does not control for size or any other characteristics. Jensen alpha is much more flexible in that respect, because different models can be used to estimate it. For example, Fama-French three factor model, which controls for size and book to equity values. The shortcoming of both Sharpe ratio and Jensen alpha is that they assume well diversified portfolios. However, Jensen alpha seems an attractive measure, because, as mentioned in Kim et al (2002), unlike other measures it can be tested for statistical significance. Finally, Jensen alpha is the most popular measure in the literature. Taking all the above into consideration, it seems most reasonable to rely on the Jensen alphas. However, the Sharpe ratios were also calculated to see whether they offer similar results. First, Sharpe ratios were computed for all types of REITs and for both periods, using the following formula (based on Sharpe (1964)): The higher the ratio, the better the performance, because a higher return was achieved at a lower risk. The main performance measure is Jensen alpha (Jensen (1968)). Positive (negative) Jensen alpha indicates a superior (inferior) performance relative to the market. Jensen alphas were obtained for each REIT type from two different models: standard CAPM and Fama-French. The latter is preferred, as it was found that it captures the nature of REITs more accurately (for example: Chiang et al (n. d.)). The constant term from each model is the Jensen alpha. The above discussed models are as follows: a) Standard CAPM: The notation is the same as before and is the return on market portfolio at time t b) Fama-French model: Differences in Jensen alphas need to be then tested for statistical significance. First single-factor ANOVA was calculated. After this, non-parametric Mood's Median was used to check for robustness (following Benefield et al 2009). The method comes from Brown and Mood (1948). First, Jensen alphas were calculated for each REIT individually and these alphas then entered the ANOVA and Mood's tests. Mood's test consists in simply calculating how many stocks within each type have a higher or lower/equal Jensen alpha than the overall median of the REIT market. Then a simple A”¡2 test with observed and expected values is calculated to test the null hypothesis that medians of Jensen alphas are the same across REIT types.

RESULTS

Results of the point estimates of Sharpe ratios (using type-portfolios) are presented in Table 3. Sharpe ratios fell dramatically for all REIT types in the CRISIS period, which was expected. Second, rankings for the two periods differ. For example, hotel REITs appear to be the best performing in the PRE-CRISIS period, but they are only 5th during the CRISIS. Nevertheless, no conclusion can be drawn, because no information is available about the statistical significance of these results. Before the crisis all alphas were positive, which indicates superior performance of REITs compared to the market. However only hotel and retail REITs have (marginally) significant alphas and that significance only persists for hotel REITs when Fama-French model is used instead of the CAPM. After/during the crisis all alphas (except the one for industrial REITs) were positive but insignificant at any conventional level. Furthermore, betas are higher for the CRISIS period, regardless of which model is used. In addition, results are similar to Sharpe ratios. The differences between point estimates across types need to be tested for statistical significance. Jensen alpha for each individual REIT were calculated. Then ANOVA and Mood's Median test were carried out. In case of ANOVA, the null hypothesis is that mean excess returns are equal across types. In case of Mood's test, the null hypothesis is that median excess returns are equal across types. Both tests give similar results - the hypothesis that there are no differences in performance across REIT types cannot be rejected for the PRE-CRISIS period at the conventional 5% level. However for the CRISIS period, both tests suggest that the null hypothesis of no differences in performance across types can be rejected at the 5% level. The above is true regardless of whether CAPM or Fama-French model is used. For the analysis to be complete, it is reasonable to check which REIT types are different. Mood's Median Test for each pair of REIT types in the CRISIS period was performed. The null hypothesis in each case is that there is no difference in median Jensen alphas (from the Fama-French model). The results can be found in Table 9. This study investigates whether there are any differences in the performance of different property-types of REITs. The contribution of this paper is to analyse the question in the light of the recent crisis and compare the results to the ones obtained for the preceding period. Two measures of performance were used: Sharpe ratio and Jensen alpha. In order to test the significance of the differences across types, ANOVA and Mood's Median Test methodologies were used. In addition, Jensen alpha was obtained from two models - CAPM and Fama-French. Sharpe ratios seem to indicate different rankings of different types of REITs before and during/after the crisis. Jensen alphas are mostly positive in both periods, but many of them are insignificant. The results suggest no differences in performance across property-types of REITs before the crisis. However during and after the crisis, healthcare and residential REITs outperform other types. Future research could focus on understanding the exact causes of these results. For example it could be resolved whether low levels of debt or the healthcare reform were the causes of healthcare REITs' performance. Moreover, types could be broken down into subtypes to better understand the nature of the problem.

Cite this page