Active Management in Real Estate Mutual Funds Viktoriya Lantushenko and Edward Nelling 1 September 4, 2017 1 Edward Nelling, Professor of Finance, Department of Finance, Drexel University, email: nelling@drexel.edu, phone: 215-895- 2117, address: 3220 Market Street, 11th Floor LeBow Hall, Philadelphia, PA 19104; Viktoriya Lantushenko, Assistant Professor of Finance, Department of Finance, Saint Joseph s University, email: vlantush@sju.edu, phone: 610-660-1668, address: 5600 City Avenue, Mandeville 256, Haub School of Business, Philadelphia, PA 19131. 1
Active Management in Real Estate Mutual Funds Abstract This paper examines active management in real estate mutual funds (REMFs). We find that the REMF industry has expanded as the underlying REIT industry has developed over time, but the number of REMFs experienced a sharp decline following the global financial crisis. Smaller funds and funds with higher expense ratios are more likely to exit from the industry. REMF fund managers have become significantly less active over time. We use four measures of active management: Active Share, Fund R-squared, Property-Type Concentration, and Return Gap. In contrast to the findings for more broadly diversified equity funds, these activeness measures do not explain the future performance of REMFs. The exception is the Fund R- squared measure, as it displays persistence and can predict future fund alpha over the subsequent three quarters. Overall, our findings shed light on the uniqueness of REMFs relative to diversified equity mutual funds. 2
1. Introduction This paper examines active management in real estate mutual funds (REMFs). Our goal is to determine whether the level of active management has changed over time, as the REIT industry has evolved and institutional investors have played an increasingly important role. We also explore the relation between active management and fund performance. Since the passage of the Revenue Reconciliation Act in 1993, the REIT industry has grown dramatically. REIT security offerings and market capitalizations have increased, as has the level of institutional investment. The growth in the REIT market has been accompanied by an increase in the number and size of mutual funds specializing in real estate investment. It is plausible that the REMF industry has become more competitive, causing fund managers to become more active as a result. To our knowledge, this is the first paper that examines the evolution of REMF active management and its effect on fund performance. An extensive literature on mutual funds exists. Beginning with Jensen (1968), the early focus was on the performance of mutual funds, and whether managers were able to earn excess returns after adjusting for risk. Over time, researchers expanded the measurement of risk to include multi-factor models (e.g., Fama and French (1993), Carhart (1997)). Much of the existing research on REMFs also focuses on performance. Kallberg, Liu and Trzcinka (2000) find that actively managed REMFs generate higher alphas than passively managed ones. Chiang, Kozhevnikov, Lee and Wisen (2008) find that REMFs do not exhibit abnormal performance. Derwall, Huij, Brounen and Marquering (2009) note the importance of controlling for momentum in REITs when measuring fund performance. Cici, Corgel and Gibson (2011) analyze fund holdings, and find that REMF managers display evidence of stockselection ability and generate positive alpha. 3
Researchers have shown that REITs differ from the broader equity market, and that REMFs are different from diversified equity funds. For instance, Anderson, Boney, and Guirguis (2012) document that REIT returns are more volatile in response to unexpected changes in monetary policy than are general equity markets. The authors find that monetary shocks have nearly twice as much of an impact on REITs as they do on the broader equity markets during high-variance regimes. Zhou and Anderson (2010) find that REITs are characterized by significantly higher levels of extreme risks compared to other stock markets. Ro and Gallimore (2013) examine herding in REMFs, and document that this behavior is significantly less pronounced in REITs than in other stocks. Our paper contributes to this strand of literature by shedding light on how REIT-focused funds differ from those funds that invest primarily in the broader stock market. In addition, our results may help investors choose REMFs that outperform others in the future. Researchers have developed various measures of the level of active management in mutual funds, referred to activeness, and examined how these measures relate to fund performance. Cremers and Petajisto (2009) develop a measure of activeness that is based on the share of portfolio holdings that differ from benchmark holdings, and show its predictability of fund performance. Amihud and Goyenko (2013) suggest that R 2, generated from a multi-factor benchmark regression, explains future fund performance. Kacperczyk, Sialm, and Zheng (2005) document that mutual funds with industry-concentrated portfolios perform better. In another paper, Kacperczyk, Sialm, and Zheng (2008) construct a measure of activeness based on the difference between the reported fund returns and the returns on a portfolio of previously disclosed holdings, and show that this difference (referred to as Return Gap ) predicts fund 4
performance. Overall, these studies conclude that more active funds exhibit significantly higher performance. Notably, most of these studies exclude specialized or sector funds. In this paper, we analyze real estate mutual funds and examine whether the relation between activeness and performance of REMFs is similar to that of equity funds with a broader investment focus. One might formulate a null hypothesis that the activeness-performance phenomenon discovered among non-specialized mutual funds is apparent for REMFs as well. Even though the REIT sector is specialized, there can be a number of opportunities for a REMF manager to outperform other funds within the real estate market. For example, REMFs may attempt to identify undervalued firms. In addition, fund managers follow a broader approach, and may focus on specific property types or geographic regions that they believe will generate abnormal performance. In contrast, there are reasons to expect that REMFs do not exhibit a relation between activeness and performance. The traditional characterization of fund managerial skill is related to market timing or stock selection abilities. In comparison with a mutual fund manager who invests in the general stock market, a REMF manager faces more constrained investment opportunities. For example, even though a REMF manager can rotate REIT property types within the portfolio, her investments are mostly (or entirely) in the real estate sector of the broader market. Similarly, a REMF manager is constrained with the choice of securities in the REIT sector, which provides less opportunity for diversification. Therefore, we alternatively hypothesize that the developed measures of activeness are not indicative of REMF managerial skill, and thus are not associated with fund performance. To the best of our knowledge, the only published study of active management in REMFs is Kallberg, Liu and Trzcinka (2000), and their characterization of active or passive management 5
is self-reported by each fund. In contrast, we construct the measures described above and examine funds with various degrees of activeness within the universe of actively managed REMFs. We apply the construction of the developed measures to REMFs, explore the dynamics of REMF activeness, and compare it with that of diversified actively managed mutual funds. Our four measures of fund activeness are equivalents to Fund R 2, proposed by Amihud and Goyenko (2013), Active Share, based on the methodology of Cremers and Petajisto (2009), a Propertytype Concentration measure, based on the Industry Concentration Index developed by Kacperczyk, Sialm, and Zheng (2005), and Return Gap, also developed by Kacperczyk, Sialm, and Zheng (2008). We find that fund managers have generally become less active over time, while the number of REMFs has grown during the pre-crisis period and has descended after the financial crisis. This is in contrast to our hypothesis that funds have become more active as the competition in the REMF industry has intensified. Figure 1 illustrates a striking difference in the evolution related to the various types of funds in the industry. The number of mutual funds and ETFs steadily grows for the period of 1996 through 2013. While this trend is generally apparent for REMFs up to 2008, the drop in the number of this type of funds has not yet been recovered since the most recent crisis. Jack Bogle, founder of The Vanguard Group, highlights the importance of mutual fund proliferation. 2 He suggests that around 7 percent of mutual funds died each year between 2001 and 2012. However, the rate at which new funds appear is even higher. This explains why the mutual fund industry has grown, even though so many funds have failed. This observation, however, does not apply to REMFs during the post-crisis period, as the 2 http://money.usnews.com/money/personal-finance/mutual-funds/articles/2013/06/10/are-there-too-many-mutualfunds 6
evidence in Figure 1 illustrates. The reason for this difference may be that new funds with a broader equity focus have a lot of flexibility related to the type of products they can offer to the public, while REMFs are specialized and concentrated in the industry that has largely suffered in 2007-2008. This finding alone contrasts REMFs with other mutual funds. Our analysis of the probability of REMF terminations indicates that smaller funds are more likely to close. Also, the probability of exit is significantly higher for REMFs that charge higher expense ratios. Collectively, these findings suggest that funds experiencing lower of economies of scale (e.g., smaller funds, and funds with higher expense ratios) are more likely to terminate. We do not find evidence that fund activeness is associated with the likelihood of REMF exits. We then examine the determinants of REMF activeness. Even though the four measures of active management are constructed differently, our results consistently suggest a strong positive association between expenses and fund activeness. This finding is intuitive. More active funds are likely to incur higher costs associated with collecting and processing information used for decision making, and pass these costs on to investors in the form of higher management fees. Finally, we examine whether REMF activeness predicts fund returns, and find that three of our four measures of activeness do not explain the future performance of REMFs, which stands in contrast to the findings for more broadly diversified equity funds. This exception is the Fund R-squared measure. We find that the Fund R-squared displays persistence over time, and is able to predict future alpha over the subsequent three quarters. We also document that this relation between Fund (1-R 2 ) and future REMF performance is more pronounced among larger funds. 7
In Section 2, we describe the measures of fund activeness. We then discuss in Section 3 the estimation of fund abnormal performance. Section 4 describes the construction of our sample of REMFs. We then present our results on changes in activeness over time, and its ability to predict future performance in Section 5. The final section concludes. 2. Measures of active management for REMFs We construct four measures of activeness in the spirit of Amihud and Goyenko (2013) (Fund R 2 ), Cremers and Petajisto (2009) (Active Share), Kacperczyk, Sialm, and Zheng (2005) (Industry Concentration Index), and Kacperczyk, Sialm, and Zheng (2008) (Return Gap). We then examine how these measures have varied over time, and the extent to which they are related to future fund performance. We construct the first measure of activeness, Fund R 2, following Amihud and Goyenko (2013). Fund R 2 is the R 2 value from regressions of fund returns on benchmark factor returns. For our main specification, we use the four-factor-mimicking portfolios proposed by Fama and French (1993) and Carhart (1997). Using an estimation period of 36 months, we regress monthly fund returns (in excess of the one-month risk-free rate) on the market excess return, as well as the size, value, and momentum factors on a rolling monthly basis. For each fund in each month, these regressions produce estimates of R 2, which is a proxy for fund selectivity. The interpretation of this measure is that selectivity (1 - R 2 ) is greater if a fund s total return volatility is driven more by idiosyncratic risk than by systematic risk. Hence, a lower R 2 is attributed to 8
stronger selectivity. If R 2 captures managerial selection skill, it is expected to be inversely associated with fund performance. 3 We use the four factors constructed from all stocks in CRSP to run these benchmark regressions. When we treat the National Association of Real Estate Investment Trusts (NAREIT) US Real Estate Index or Ziman REIT index as a market benchmark, R 2 s from these rolling regressions appear to be high and do not exhibit much variation. Peterson and Hsieh (1997) document that risk premiums on a market portfolio and the returns on mimicking portfolios based on size and book-to-market across common stocks are significantly related to risk premiums on equity REITs. Thus, we are comfortable with the choice of using the fourmimicking-factor portfolios based on the universe of stocks to generate Fund R 2 s. Our second measure of activeness, Active share, for REMF portfolios is constructed using the method of Cremers and Petajisto (2009), as follows: where and are the portfolio weights of REIT i in the REMF portfolio and in the Ziman index, respectively. If a fund does not hold a REIT that is in the Ziman index, we assign the portfolio weight of this asset to zero. We treat the Ziman index as a portfolio benchmark for each REMF. The basic interpretation of Active Share is that it represents the share of portfolio equity that is different from the index. More active funds have a higher value of Active Share in their portfolios. Our third measure of activeness is similar to that of Kacperczyk, Sialm, and Zheng (2005), who develop the Industry Concentration Index. We construct a property-type 3 For robustness, we also construct the logistic transformation of R 2, as in Amihud and Goyenko (2013). This is performed to improve the qualities of the R 2 -distribution. Our results are similar when using this version of the activeness measure. 9
concentration measure for each REMF. We classify each REIT into one of seven property types (Retail, Residential, Office and Industrial, Healthcare, Lodging and Resorts, Diversified, and Other) based on the nature of its assets. We then assess the property-type concentration of REMF portfolios relative to the property-type weights in the Ziman REIT index. In each quarter, the property-type concentration (PTC) in REMF portfolios is calculated as follows: where is the weight of the REMF portfolio holdings in property type j and is the weight of property type j in the Ziman REIT index. The intuition behind this measure of activeness is that REMF managers may hold more concentrated portfolios in certain property types if they believe that these property types will outperform others in the near future. It is also possible that managers may have superior information about profitable REITs in certain property types. If so, skilled managers would over-weight their portfolios in these property types, and one would expect a positive association between REMF property-type concentration and fund performance. For our fourth measure of activeness, in the spirit of Kacperczyk, Sialm, and Zheng (2008), we estimate the Return Gap measure for REMFs. Return Gap is a proxy for unobserved actions of managers and is computed as the difference between the reported fund return and the return on a portfolio based on previously disclosed holdings. Kacperczyk, Sialm, and Zheng (2008) find that some fund managers persistently create value (and exhibit a positive Return Gap), while others destroy value (and exhibit a negative Return Gap). Hence, this measure reflects directly the value added or subtracted by fund managers relative to the return on a hypothetical portfolio from previously disclosed holdings. If it captures skill, it is expected to be 10
positively related to future fund performance. Following their approach, we construct Return Gap as a measure of activeness for our sample of real estate mutual funds. 3. Fund performance measures We use four measures of REMF performance. For our first measure, α CAPM, we run timeseries regressions of excess monthly fund returns on excess market returns using the preceding 36 months of data. We then obtain abnormal fund performance by subtracting the fitted value of the return using the factor loadings on the market portfolio estimated over the prior 36 months from the fund return in excess of the risk-free rate. We estimate the four-factor abnormal performance, α 4-factor, similarly. That is, we repeat the procedure above, except that we regress monthly fund returns on four factor portfolios based on size, value, and momentum, in addition to excess market return. The quarterly alpha, α CAPM or α 4-factor, is obtained by compounding monthly fund alphas within each quarter. In addition, we estimate REMF abnormal performance treating the Ziman Index as the relevant universe of stocks. We obtain α CAPM Ziman and α 4-factor Ziman, following the method described above, except that we use the Ziman REIT index as the market portfolio and the size, book-to-market, and momentum factors constructed based on REITs only. 4 4 For robustness, we used the National Association of Real Estate Investment Trusts (NAREITs) index as a benchmark, and the results were similar. 11
To form size portfolios, in June of each year, we split all REITs into two groups: those with market capitalization above and below the median. We form two portfolios based on this sort. The size portfolio return is the value-weighted return on small capitalization REITs minus the value-weighted return on large capitalization REITs. To form book-to-market portfolios, in December of each year, we divide all REITs into three groups based on their book-to-market ratios: low (bottom 30%), medium (middle 40%), and high (top 30%). The book-to-market portfolio return is the difference between the valueweighted returns on high and low book-to-market REITs. We construct monthly momentum portfolios by sorting all REITS based on cumulative returns over months (t-2) to (t-12), as in Cici, Corgei, and Gibson (2011). The momentum portfolio return is the value-weighted return on high-momentum REITs (top 30%) and lowmomentum REITs (bottom 30%). To facilitate comparison, we compute the average returns on the REIT factors for the period of 1995-2006, and compare them with those in Table 2 in Cici, Corgei, and Gibson (2011). Our factor returns are similar to theirs. 4. Sample Construction 4.1. Real estate mutual funds We select our sample of REMFs from the summary file in the CRSP mutual fund database. We choose funds with the investment objective EDSR (equity domestic sector real estate) or with the strategic insights code RLE (equity USA real estate) from January 1995 to December 2015. We also screen for funds with name phrases Real Estate, REIT, and Realty. These screens result in a sample of 174 funds. We further eliminate index funds by 12
screening their names. After excluding 23 index funds, our final sample consists of 151 unique actively managed REMFs. We aggregate the characteristics of funds with multiple share classes following the procedures established in the literature. (e.g. Wermers (2000), Kacperczyk, Sialm, and Zheng (2008), Cremers and Petajisto (2009)). Total assets under management (TNA) for each fund are obtained by summing up all values of assets across all fund share classes. Fund age is the maximum age of its share classes. For other fund characteristics (expense ratio, turnover ratio, returns), we take the TNA-weighted average across all share classes of a fund. The characteristics of our sample of funds are comparable with those reported by Cici, Corgel, and Gibson (2011). An average real estate fund manages about $606 million in assets, has an expense ratio of 1.22%, and incurs a turnover ratio of 81%. Using the WRDS MFLINK table with fund identifiers, we obtain REMF holdings from Thomson/CDA. Based on this match, the Thomson dataset provides information on holdings for 146 out of 151 funds in our sample. On average, a REMF portfolio consists of about 40-50 holdings (Table 1). Monthly REIT returns are obtained from CRSP, and monthly REIT indices are from Ziman Real Estate Data Series. All datasets used in this study are free of survivorship bias. 4.2. Diversified actively managed equity funds To construct a sample of diversified actively managed equity funds, we obtain fund holdings from the Thomson database over the period January 1995 through December 2015 and focus on mutual funds with aggressive growth, growth, and growth and income investment styles. Similarly to the procedures described for the REMFs, we merge the holdings dataset with the CRSP mutual fund database and eliminate index, international, balanced, sector, and bond 13
funds by screening fund names and using investment objective codes. We use the same method of aggregating multiple share classes into one fund as described for the sample of REMFs. We apply several filters commonly used in the mutual fund literature. Since Elton, Gruber, and Blake (2001) suggest that the fund returns with TNA less than $15 million are biased upwards in the CRSP mutual fund database, we eliminate funds with TNA below $15 million as of the beginning of month. We remove fund observations with fewer than 11 stocks in a portfolio (e.g. Kacperczyk, Sialm, and Zheng (2008)). Also, as in Evans (2010), we exclude the first 18 months of fund returns to mitigate the impact of incubation bias. Overall, our sample consists of 1,679 diversified actively managed mutual funds. The descriptive statistics of this sample are generally consistent with those documented in the literature (e.g. Huang, Sialm, and Zheng (2011), Amihud and Goyenko (2013), and Jiang and Verardo (2013)). An average fund manages about $2.3 billion in assets, has an expense ratio of 1.18%, is 21 years old, and a turnover ratio of about 79%. 5. Results 5.1. The evolution of active management over time We first explore how the active management measures of REMFs evolve over time. Panel A of Table 1 reports the average of each activeness measure for each year, over the period 1995-2015. The number of REMFs increased from 15 in 1995 to 115 in 2007, which marked the beginning of the global financial crisis. By year-end 2008, the number of REMFs dropped to 87, and declined further to 62 by 2015. The REMF managers appeared to become less active from 1995-2007, as all four measures of activeness generally decreased over this period. The Fund (1-R 2 ) measure decreased significantly after the financial crisis, and then increased to an all-time 14
high of 0.771 in 2014. The Active Share, Property-Type Concentration, and Return Gap measures remained low and continued to decline after the crisis. The Return Gap measure was negative in six of the seven years over the period 2009-2015, indicating that active management in REMFs has been value-reducing after the financial crisis. Panel B of Table 1 reports the correlations between the activeness measures. The correlations are relatively low, with the highest (0.39) between the Active Share and Property- Type Concentration measures. Since the measures are constructed to capture different aspects of fund activeness, these low correlations are expected. 5.2. The probability of REMF terminations Since we observe that the trend of the number of REMFs is distinct from that of other equity funds, it is natural to ask what explains REMF termination, and whether activeness appears to be one of the determinants. We model the probability of REMF exiting the industry as a function of activeness, size, expense ratio, turnover, age, and performance. The dependent variable is set equal to one if a fund closes in the next quarter, and zero otherwise. All independent variables are measured as of the quarter prior to the exit indicator quarter. Table 2 presents the results. We find no significant association between REMF activeness and the likelihood of fund closing. Of the four activeness measures, only the property type concentration is significantly associated with the probability of exit. The inverse relation between the property type concentration and the probability of exit suggests that REMF portfolios consisting of more property sectors are less likely to close to new investors. The results also suggest that smaller funds and funds with higher expense ratios are more likely to exit from the industry. Since these types of funds are likely to experience little economies of 15
scale, this may be an explanation for their increased probability of exit. Also, the coefficients on portfolio turnover are negative and significant. A possible interpretation is that funds tend to trade less in the quarter(s) preceding their exit, assuming the information about closing is known in advance. The data show that the portfolio turnover decreases by about 7 basis points, on average, during the year preceding the exit, whereas this trend looks stable for funds remaining in the industry. 5.3. The determinants of active management Since the focus of this paper is on active REMFs, we explore what determines the level of activeness among these funds, and model it as a function of fund characteristics. The dependent variables are the active management measures (Fund (1-R 2 ), Active Share, Property Type Concentration, and Return Gap). The explanatory variables include size, expense ratio, portfolio turnover, age, and quarterly performance. All independent variables are measured as of the quarter proceeding the activeness measurement quarter. As a robustness check, we find that our results are consistent when we measure the dependent and explanatory variables in contemporaneous terms. Table 3 reports the results of estimating the determinants of REMF activeness. Even though all four measures are built differently and capture distinct aspects of active management, the results suggest that Expense ratio loads significantly in all specifications and is positively associated with activeness. It is likely that more active funds incur higher costs of managing their portfolios. Investors may be willing to pay a higher rate to more active funds, if it is harder to replicate the strategies that these funds use. 16
5.4. Active management and future REMF performance We next explore if various measures of active management can be used to predict performance of REMFs. To do so, we run pooled cross-sectional regressions of one-quarterforward risk-adjusted performance on measures of active management and other fund characteristics for both REMFs and diversified actively managed funds. Table 4 reports the results using Fund (1-R 2 ) as a measure of activeness. Consistent with Amihud and Goyenko (2013), we find a strong positive association between Fund (1-R 2 ) and future performance for the group of diversified funds (Columns 5 and 6 of Table 4). Similar to diversified funds, REMFs that exhibit a higher level of selectivity (Fund (1-R 2 )) earn a higher alpha, when risk-adjusted performance is estimated using a broader equity stock market (Columns 1 and 2 of Table 4). We do not find this significant link between Fund R-squared and subsequent performance, if alpha is benchmarked against REITs only (Columns 3 and 4 of Table 4). Investors can use these findings and select those REMFs which, as a part of their diversified portfolio, are likely to outperform others in the future. We examine how other measures of active management are related to future fund performance. Table 5 reports the results. Similar to Cremers and Petajisto (2009), Kacperczyk, Sialm, and Zheng (2005), and Kacperczyk, Sialm, and Zheng (2008), Active Share (Panel A), Property-Type/Industry Concentration Index (Panel B), and Return Gap (Panel C) are all positively associated with future fund performance for diversified actively managed funds (Columns 5 and 6 of Table 5). In contrast, this relation does not hold for the group of real estate mutual funds, as the coefficients on the measures of activeness in the first four columns in each panel of Table 5 are insignificant. These findings highlight how REMFs are distinct from 17
diversified funds. From a practical perspective, our evidence shows that these measures may not be reliable in predicting future REMF performance. Overall, our results suggest that the activeness measures used to predict mutual fund performance are generally not applicable in the REMF sector. One exception is Fund R-squared, and its predictive ability is contingent on the method of performance measurement. 5.5. The persistence in Fund R-squared We have shown that out of four measures of active management, Fund (1-R 2 ) is significantly associated with future REMF performance. We further explore the persistence in this predictability over time. Each quarter, we sort all REMFs in quartiles based on Fund (1-R 2 ), and compute the average alpha of funds in each group in the following four quarters. The performance (alpha) is estimated using the market portfolio, as well as the size, value, and momentum factors based on the universe of all CRSP stocks. Table 6 reports the results. The relation between Fund (1-R 2 ) and subsequent performance of REMFs is generally monotonic, as funds with higher selectivity (higher Fund (1- R 2 )) exhibit stronger risk-adjusted returns. The significant differences in performance of the highest and lowest portfolios suggest that based on Fund (1-R 2 ), investors can choose REMFs that significantly outperform others by 83 to 92 basis points per quarter in the subsequent three quarters. This finding can be useful to investors seeking to diversify their portfolios by investing in funds with a specialization in real estate. 18
5.6. The predictability of Fund (1-R 2 ) and REMF size Studies such as Berk and Green (2004) of diversified mutual funds generally report an inverse relation between fund size and future performance. Larger funds may have a greater temporary price impact on stock prices because they place large trades, which adversely affects portfolio performance. We explore if the relation between the Fund (1-R 2 ) and performance is stronger for REMFs based on fund size. We split our sample of REMFs into two groups, with total net assets under management above and below the median value, and rerun the regressions in Table 4. We report the results in Table 7. Surprisingly, we find that this relation between the Fund (1-R 2 ) and future abnormal performance is pronounced among larger REMFs. The coefficient on Fund (1-R 2 ) is positive and significant (3.90**) for the subgroup of REMFs with TNA above the median. This relation between Fund (1-R 2 ) and future fund performance is not pronounced for smaller funds. Also, in contrast to diversified funds, there is no significant relation between fund size and future performance of REMFs. In summary, if investors are to use R-squared to choose REMFs that outperform others in the future, this is likely to work only for larger funds. 6. Conclusion This study examines active management in real estate mutual funds (REMFs). Using four measures of activeness, we explore how the level of active management has changed over time and examine its effect on REMF performance. Although the four measures capture different aspects of activeness, we find that REMF managers have become less active over time. Lower levels of activeness are particularly 19
pronounced after the financial crisis. The number of REMFs increased during the pre-crisis period, but declined after 2007. These statistics suggests that the REMF industry has become less competitive, and fund managers have become less active after the crisis. Our results also indicate that smaller funds and funds with higher expense ratios are more likely to exit from the industry. In contrast to the findings reported for diversified actively managed equity funds, only one out of four measures, Fund (1-R 2 ), explains future REMF abnormal performance. This relation between Fund (1-R 2 ) and future risk-adjusted return is pronounced among larger REMFs. Overall, our findings provide additional insights regarding REMF performance, and highlight how REMFs differ from other funds. 20
References Amihud, Y., & Goyenko, R. (2013). Mutual fund s R 2 as predictor of performance. Review of Financial Studies, 27, 3133-3170. Anderson, R. I., Guiguis, H., & Bony, V. (2012). The impact of switching regimes and monetary shocks: An empirical analysis of REITs. Journal of Real Estate Research, 34, 157-181. Berk, J., & Green, R. (2004). Mutual fund flows and performance in rational markets. Journal of Political Economy, 112, 1269-1295. Carhart, M. (1997). On persistence in mutual fund performance. The Journal of Finance, 52, 57 82. Chan, S., Leung, W., & Wang, K. (1998). Institutional investment in REITs: Evidence and implications. Journal of Real Estate Research, 16, 357-374. Chen, H., Jegadeesh, N., & Wermers, R. (2000). The value of active mutual fund management: An examination of the stockholdings and trades of fund managers. Journal of Financial and Quantitative Analysis, 35, 343 368. Chiang, K., Kozhevnikov, K., Lee, M., & Wisen, C. (2008). Further evidence on the performance of fund of funds: the case of real estate mutual funds. Real Estate Economics, 36, 47 61. Cici, G., Corgel, J., & Gibson, S. (2011). Can fund managers select outperforming REITs? Examining fund holdings and trades. Real Estate Economics, 39, 455 486. Cremers, J., & Petajisto, A. (2009). How active is your fund manager? A new measure that predicts performance. Review of Financial Studies, 22, 3329-3365. Daniel, K., Grinblatt, M., Titman, S., & Wermers, R. (1997). Measuring mutual fund performance with characteristic-based benchmarks. The Journal of Finance, 52, 1035 1058. Derwall, J., Huij, J., Brounen, D., & Marquering, W. (2009). REIT momentum and the performance of real estate mutual funds. Financial Analysts Journal, 65, 24-34. Fama, E., & French, K. (1993). Common risk factors in the returns on stocks and bonds. Journal of Financial Economics,33, 3 56. Gallo, J., Lockwood, L., & Rutherford, R. (2000). Asset allocation and the performance of real estate mutual funds. Real Estate Economics, 28, 165 184. Hartzell, J., Muhlhofer, T., & Titman, S. (2010). Alternative benchmarks for evaluating mutual fund performance. Real Estate Economics, 38, 121 154. Huang, J., Sialm, C., & Zhang, H. (2011). Risk shifting and mutual fund performance. Review of Financial Studies, 24, 2575-2616. 21
Jensen, M. C. (1968). The performance of mutual funds in the period 1945-1964. Journal of Finance, 23, 389 416. Jiang, H., & Verardo, M. (2013). Does herding behavior reveal skill? An analysis of mutual fund performance. Working paper. Kallberg, J., Liu, C., and Trzcinka, C. (2000). The value added from investment managers: an examination of funds of REITs. Journal of Financial and Quantitative Analysis, 35, 387 408. Kacperczyk, M., Sialm, C., & Zheng, L. (2005). On the industry concentration of actively managed equity mutual funds. Journal of Finance, 62,485-528. Kacperczyk, M., Sialm, C., & Zheng, L. (2008). Unobserved actions of mutual funds. Review of Financial Studies 21: 2383-2416. Lin, C., & Yung, K. (2004). Real estate mutual funds: performance and persistence. Journal of Real Estate Research, 26, 69 93. Ling, D., & Naranjo, A. (2006). Dedicated REIT mutual fund flows and REIT performance. Journal of Real Estate Finance and Economics, 32, 409-433. O Neal, E., & Page, D. (2000). Real estate mutual funds: abnormal performance and fund characteristics. Journal of Real Estate Portfolio Management, 6, 239 247. Peterson, J., & Hsieh, C. H. (1997). Do common risk factors in the returns on stocks and bonds explain returns on REITs? Real Estate Economics, 25, 321-345. Ro, S., & Gallimore, P. (2014). Real estate mutual funds: herding, momentum trading and performance. Real Estate Economics, 42, 190-222. Rodriguez, J. (2007). A critical look at the forecasting ability of real estate mutual fund managers. Journal of Real Estate Portfolio Management, 13, 99 106. Young, M. (2000). REIT property-type sector integration. Journal of Real Estate Research, 19, 3-21. Zhou, J., & Anderson, R. (2013). An empirical investigation of herding behavior in the U.S. REIT market. Journal of Real Estate Finance and Economics, 47, 83-108. 22
23
Table 1: Summary Statistics for Real Estate Mutual Funds (REMFs) This table reports the descriptive statistics of 151 REMFs in our sample. The first three columns of Panel A report the number of unique funds, their total assets under management, and the average number of portfolio holdings by all funds in each year (as of year-end). The last four columns report median measures of activeness. Panel B reports the correlation matrix for the all variables used in the main analyses. Panel A: Sample by year Year # of unique funds Total net assets ($ Millions) Average number of portfolio holdings Median Fund (1-R 2 ) Median Active Share Median PTC Median Return Gap (%) 1995 15 1,889 45 0.438 0.557 0.029-0.030 1996 18 4,807 43 0.735 0.564 0.024 0.086 1997 33 12,104 45 0.651 0.607 0.024 0.069 1998 51 9,326 44 0.459 0.596 0.026 0.069 1999 68 8,450 49 0.347 0.570 0.026 0.024 2000 66 11,275 41 0.479 0.486 0.023-0.038 2001 65 12,876 42 0.615 0.456 0.023 0.021 2002 72 16,761 41 0.632 0.494 0.016 0.107 2003 79 28,852 47 0.527 0.489 0.018-0.057 2004 83 46,196 49 0.652 0.488 0.016-0.077 2005 83 53,697 50 0.756 0.461 0.015 0.017 2006 89 83,351 46 0.694 0.420 0.013 0.023 2007 115 74,151 41 0.474 0.409 0.013 0.011 2008 87 33,210 37 0.243 0.410 0.012 0.074 2009 83 44,912 43 0.171 0.387 0.011-0.042 2010 79 58,679 45 0.173 0.398 0.009-0.017 2011 69 72,466 43 0.167 0.347 0.008-0.029 2012 69 90,039 44 0.260 0.373 0.008-0.006 2013 68 97,329 46 0.433 0.379 0.009 0.007 2014 61 128,166 50 0.771 0.382 0.010-0.028 2015 62 128,845 46 0.730 0.381 0.009-0.035 24
Panel B: Correlation matrix Active share Fund (1-R 2 ) Property-type concentration Return gap Log(TNA) Expense ratio Turnover ratio Log(age) Active share 1 Fund (1-R 2 ) 0.12*** 1 Property-type concentration 0.39*** -0.03** 1 Return gap 0.002 0.004-0.002 1 Log(TNA) -0.044** 0.05*** 0.101*** 0.006 1 Expense ratio 0.29*** 0.02 0.079*** 0.057*** -0.35*** 1 Turnover ratio -0.17*** -0.10*** -0.12*** 0.042** -0.18*** 0.17*** 1 Log(age) -0.20*** -0.16*** 0.01-0.02 0.37*** -0.32*** -0.10*** 1 25
26
Table 2: Probability of real estate mutual fund termination This table reports the probit regressions coefficients of the probability of fund termination. The dependent variable is 1 if a fund closes in the following quarter, and 0 otherwise. Termination probability is modeled as a function of activeness and other control variables, on a fund-quarter basis. The four measures of fund activeness include Fund (1-R 2 ), Active Share, Property Type Concentration, and Return Gap. Quarterly return is obtained by compounding monthly net-of-expense raw returns. All independent variables are measured as of the quarter prior to the exit quarter. Standard errors are clustered at the fund level. T-statistics are provided in parentheses. Continuous variables are winsorized at the 1% level. Fund (1-R 2 ) Active share Property type concentration Return gap Active management measure -0.119-0.160-6.123** 3.439 (-0.14) (-0.24) (-2.16) (1.02) Log(TNA) -0.362*** -0.362*** -0.371*** -0.389*** (-4.86) (-5.23) (-4.91) (-3.92) Expense ratio 46.65* 50.11*** 63.77*** 57.77** (1.81) (2.64) (3.13) (2.25) Turnover -0.193** -0.198** -0.207** -0.199** (-2.00) (-2.30) (-2.45) (-2.16) Log(age) 0.225 0.381** 0.448*** 0.355** (1.04) (2.52) (2.78) (2.11) Quarterly return 1.479 0.317-0.384-0.951 (0.71) (0.14) (0.13) (-0.33) Intercept -4.084*** -4.285*** -4.508*** -4.548*** (-4.91) (-8.95) (-9.49) (-8.50) Quarter FEs Yes Yes Yes Yes R 2 0.04 0.03 0.03 0.04 # of observations 3,181 3,699 3,687 3,175 27
Table 3: Determinants of active management measures This table reports the determinants of activeness. The dependent variables include active management measures: Fund (1-R 2 ), Active Share, Property type concentration, and Return Gap. Quarterly return is obtained by compounding monthly net-of-expense raw returns. All independent variables are measured as of the quarter prior to the activeness measurement quarter. All regressions include quarter fixed effects. Standard errors are clustered at the fund level. T-statistics are provided in parentheses. Variables are winsorized at the 1% level. Fund (1-R 2 ) Dependent variable Property type Active share concentration Return gap Log(TNA) 0.007 0.006 0.005* -0.0004 (1.58) (0.90) (1.85) (-0.98) Expense ratio 5.216** 8.669*** 1.888* 0.821*** (2.30) (3.05) (1.86) (6.34) Turnover -0.002-0.026*** -0.001 0.001 (-0.55) (-3.31) (-0.32) (0.80) Log(age) 0.003 0.005 0.003 0.0001 (0.11) (0.23) (0.29) (0.10) Quarterly return -0.199-0.196* -0.069-0.046 (-1.12) (-1.82) (-0.68) (-1.27) Intercept -0.070 0.231*** -0.035-0.003 (-0.95) (3.52) (-1.20) (-0.77) Quarter FEs Yes Yes Yes Yes R 2 0.14 0.29 0.05 0.15 # of observations 3,265 3,724 3,711 3,166 28
Table 4: REMF performance and Fund R 2 This table reports the coefficients of quarterly pooled regressions for real estate mutual funds and diversified actively managed equity funds. The dependent variables include one-quarter-forward riskadjusted performance. The measure of selectivity, Fund (1-R 2 ), is constructed following Amihud and Goyenko (2013). It is estimated from regressions of monthly fund returns in excess of the risk-free rate on the market excess return, as well as the size, value, and momentum factors using 36 monthly return observations. All regressions include quarter fixed effects. Standard errors are clustered at the fund level. Dependent variable: 1Q forward riskadjusted performance Real estate mutual funds Diversified actively managed equity funds α CAPM α 4-factor α CAPM Ziman α 4-factor Ziman α CAPM α 4-factor Fund (1-R 2 ) 2.918** 2.85** 0.803 0.747 3.693*** 1.764*** (1.99) (2.19) (0.44) (0.58) (11.08) (6.55) Log(TNA) 0.030 0.007 0.035 0.038-0.028** -0.021** (1.15) (0.22) (1.16) (1.31) (-2.21) (-2.16) Expense ratio -12.31-19.79-27.15* -38.47** -22.46*** -29.39*** (-0.85) (-1.62) (-1.87) (-2.43) (-3.43) (-5.72) Turnover 0.025-0.029-0.079* -0.083** 0.045-0.114*** (0.37) (-0.61) (-1.77) (-2.16) (1.36) (-4.21) Log(age) 0.004-0.018-0.095-0.085-0.108*** -0.016 (0.06) (-0.19) (-0.95) (-0.89) (-3.17) (-0.64) Intercept 2.14 1.98* -0.373-0.138 2.635*** 1.668*** (1.53) (1.71) (-0.24) (-0.11) (7.19) (5.71) Quarter Fes Yes Yes Yes Yes Yes Yes R 2 0.92 0.89 0.23 0.22 0.09 0.09 # of observations 3,193 3,193 3,193 3,178 53,893 53,893 T-statistics are provided in parentheses. Variables are winsorized at the 1% level. 29
Table 5: REMF performance and other measures of active management This table reports the coefficients of quarterly pooled regressions for real estate mutual funds and diversified actively managed equity funds. The dependent variables include one-quarter-forward riskadjusted performance. Panel A reports the results with Active Share as a measure of fund activeness. The variable of interest in Panel B is Property-type (Industry) Concentration (PTC) of REMF (diversified) portfolios. Panel C reports the results with the Return Gap activeness measure. All regressions include quarter fixed effects. Standard errors are clustered at the fund level. T-statistics are provided in Panel A: Active share Dependent variable: 1Q forward riskadjusted performance Real estate mutual funds Diversified actively managed equity funds α CAPM α 4-factors α CAPM Ziman α 4-factor Ziman α CAPM α 4-factor Active share -0.072-0.85** 0.239-0.141 3.023*** 0.574*** (-0.16) (-2.03) (0.49) (-0.30) (16.40) (4.91) Log(TNA) 0.018 0.001 0.030 0.042-0.047*** -0.023* (0.76) (0.02) (1.14) (1.61) (-2.68) (-1.82) Expense ratio -20.91* -21.34* -31.71** -34.89*** -50.10*** -35.31*** (-1.69) (-1.71) (-2.44) (-2.76) (-5.93) (-5.32) Turnover 0.009-0.064-0.078* -0.082*** 0.101** -0.08** (0.12) (-1.29) (-1.87) (-2.17) (2.33) (-2.21) Log(age) -0.028-0.039-0.108-0.075-0.094** 0.014 (-0.38) (-0.40) (-1.06) (-0.82) (-2.23) (0.43) Intercept 4.61*** 4.57*** 0.249-0.757* -1.66*** -0.74*** (10.27) (10.33) (0.58) (-1.89) (-6.11) (-3.83) Quarter FEs Yes Yes Yes Yes Yes Yes parentheses. Variables are winsorized at the 1% level. 30
R 2 0.91 0.88 0.23 0.22 0.10 0.10 # of observations 3,193 3,193 3,193 3,178 30,968 30,968 Panel B: Property-type concentration (PTC) Dependent variable: 1Q forward risk-adjusted performance Property-type/industry concentration index Real estate mutual funds Diversified actively managed equity funds α CAPM α 4-factor α Ziman α 4-factor Ziman α CAPM α 4-factor -0.715-2.85** -0.476 0.188 3.575*** 1.677*** (-0.53) (-2.04) (-0.39) (0.16) (6.56) (3.91) Log(TNA) 0.022 0.013 0.034 0.041-0.037*** -0.025*** (0.91) (0.46) (1.23) (1.54) (-2.83) (-2.65) Expense ratio -19.97-22.54* -28.66** -36.54*** -22.054*** -29.72*** (-1.52) (-1.87) (-2.18) (-2.71) (-3.23) (-5.54) Turnover 0.011-0.043-0.083** -0.079** 0.078** -0.09*** (0.15) (-0.88) (-1.98) (-2.09) (2.25) (-3.14) Log(age) -0.025-0.031-0.101-0.078-0.101*** -0.001 (-0.35) (-0.35) (-1.02) (-0.85) (-2.91) (-0.03) Intercept 4.57*** 4.28*** 0.283-0.781** -0.796*** 0.002 (12.21) (11.14) (0.76) (-2.08) (-4.00) (0.02) Quarter FEs Yes Yes Yes Yes Yes Yes R 2 0.92 0.89 0.23 0.22 0.10 0.10 # of observations 3,193 3,193 3,193 3,178 50,340 50,340 31
Panel C: Return Gap Dependent variable: 1Q forward riskadjusted performance Real estate mutual funds Diversified actively managed equity funds α CAPM α 4-factor α CAPM Ziman α 4-factor Ziman α CAPM α 4-factor Return gap 0.450 0.344 0.294 0.535 34.35*** 15.45** (0.72) (0.42) (0.42) (0.74) (4.58) (2.44) Log(TNA) 0.018-0.010 0.031 0.041-0.035*** -0.023** (0.67) (-0.33) (1.08) (1.58) (-2.63) (-2.38) Expense ratio -13.44-22.33* -33.34** -41.44*** -12.72* -25.68*** (-1.35) (-1.72) (-2.21) (-2.60) (-1.90) (-4.94) Turnover 0.014-0.031-0.055-0.052 0.063* -0.107*** (0.22) (-0.57) (-1.41) (-1.31) (1.87) (-3.73) Log(age) 0.001 0.005-0.110-0.079-0.116*** -0.024 (0.01) (0.05) (-1.03) (-0.81) (-3.27) (-0.92) Intercept 4.49*** 4.16*** 0.284-0.82** -0.622*** 0.171 (10.78) (9.43) (0.66) (-2.03) (-3.13) (1.06) Quarter FEs Yes Yes Yes Yes Yes Yes R 2 0.92 0.89 0.24 0.24 0.10 0.09 # of observations 2,866 2,866 2,866 2,866 51,711 51,711 32
Table 6: REMF R 2 persistence This table examines the persistence in predictability of the fund abnormal returns based on the Fund R 2 as a measure of activeness. Each quarter (quarter t), we sort REMFs in quartiles based on Fund (1-R 2 ) and calculate the average risk-adjusted performance of funds in each group in the following four quarters. The abnormal performance, α 4-factor, is estimated using the market portfolio, as well as the size, value, and momentum factors constructed using the universe of all stocks in CRSP. T-statistics are based on Newey- West adjusted standard errors. Quartile α 4-factor (t+1) (%) α 4-factor (t+2) (%) α 4-factor (t+3) (%) α 4-factor (t+4) (%) Lowest -0.44-0.66-0.55-0.37 2-0.16-0.08-0.14-0.09 3 0.22 0.25 0.25 0.31 Highest 0.39 0.26 0.35 0.12 Highest-Lowest -0.83** -0.92** -0.90** -0.49 t-statistics (-2.36) (-2.25) (-2.34) (-1.06) 33
Table 7: REMF R 2 and size This table reports the coefficients of quarterly pooled regressions for real estate mutual funds. The first column reports the regression results for REMFs with total net assets under management above median ($208M), and the second column reports the results for small REMFs, those with TNA below median. The dependent variable is one-quarter-forward four-factor alpha, α 4-factor. The measure of selectivity, Fund (1-R 2 ), is constructed following Amihud and Goyenko (2013). It is estimated from regressions of monthly fund returns in excess of the risk-free rate on the market excess return, as well as the size, value, and momentum factors using 36 monthly return observations. All regressions include quarter fixed effects. Standard errors are clustered at the fund level. T-statistics are provided in parentheses. Variables are winsorized at the 1% level. Dependent variable: 1Q forward risk-adjusted performance TNA>=median ($208M) TNA<median ($208M) Fund (1- R 2 ) 3.90** 0.856 (2.46) (0.58) Log(TNA) 0.022-0.039 (0.28) (-0.66) Expense ratio -29.44-18.02 (-1.37) (-1.26) Turnover -0.212-0.027 (-0.92) (-0.57) Log(age) -0.129 0.004 (-0.59) (0.04) Intercept 5.543*** 4.35*** (6.13) (7.98) Quarter Fes Yes Yes R 2 0.88 0.90 # of observations 1,600 1,594 34
*Source: Investment Company Institute 2014 Fact Book (p. 20) Figure 1: Number of funds by year This figure represents the trend in the number of mutual funds, ETFs, and REMFs over the period of 1996 through 2013. 35