Creating the XLRE: Market Implications for REITs and the Real Estate Sector

Abstract

On October 7, 2015, State Street Global Advisors launched the first new sector ETF (SPDR) since the inception of these sector funds in 1998. This was just one of many market events spurred by Morgan Stanley Capital International (MSCI) and Standard & Poor’s Dow Jones (S&P DJ) decision to officially give real estate its own market classification. Since this was the first time an event of this kind had taken place, markets were uncertain about the effects such a change could have on the underlying real estate assets. This paper provides a comprehensive study of the real estate sector SPDR (XLRE) launch event and the impacts on the underlying real estate investment trusts (REITs). We find evidence of positive price effects, no significant liquidity effects, and a negative volatility effect. This is the first study combining these effects where others focus on the volatility effects of ETFs on underlying stocks. Since our study contradicts the results of previous volatility studies, we show that further comprehensive analysis of these events is warranted.

Keywords:

• REITs
• XLRE
• ETF
• SPDRs

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 Goodwin and Liu (2021) and Liu (2022) are among the few comprehensive examinations of the GICS reclassification event.

2 A possible event of contamination is the creation of the real estate sector in the Global Industry Classification Standard (GICS), an industry taxonomy developed in 1999 by MSCI and Standard & Poor's (S&P) for use by the global financial community, on September 16, 2016, which falls within the post-event period of the two-year sample period, October 7, 2014 (a Tuesday) to October 6, 2016 (a Thursday). Nonetheless, we conduct all tests using the expanded 2-year sample period and find qualitatively similar results.

3 Similarly, Sorescu (2000) uses only the listing dates, citing a lack of price effect of the announcements, as documented in Conrad (1989) and Detemple and Jorion (1990). In this case, we were even not able to pinpoint an announcement date. After all, it is the listing and trading of the ETF that might generate potential effects, if any.

4 Using an equal-weighted market index yields similar results.

5 Nevertheless, the results are robust to relocating the estimation windows of at least 100 trade days in the two trading years around the event, suggesting no material changes, if any, in the parameters within the pre-event or post-event period.

6 This is the “crude dependence adjustment” discussed in Brown and Warner (1980). Yet, we find qualitatively similar results using the cross-sectional standard deviation.

7 Similar results for cumulative price effects from windows extended to day +15 and +30 in Table 2 suggest permanency of the price effect, as in Conrad (1989), while Sorescu (2000) focuses exclusively on the 11-day window. As suggested by a referee, we also tried to extend the cumulation window up to 100 days after the event, while finding qualitatively similar results, but hesitated to attach much import owing to possible contaminations.

8 We obtain statistically similar results from the Fama–French three-factor model, as detailed in Fama and French (1992).

9 Following Kumar et al. (1995), we also tried to skip the first trading month around the event (21 trading days in both the estimation and event windows) but found no material difference. Using the whole prior year as the estimation period has no material impact on the results.

10 The results reported in Tables 4 and 5 do not change materially when market-adjusted returns are used in the estimation of the bid/ask spread.

11 Using the same estimation window leads to the same conclusions, so does extending the windows to one full year in both directions. Perhaps, the symmetrical pair-wise comparisons make more sense in light of the auto-covariance nature of the implied spread.

12 The conclusions hold when the windows extend to one full year in both directions. Likewise, the conclusions are robust to alternative measures of volatility: squared daily return used by Skinner (1989) and variance of daily return employed by Kumar et al. (1995).

13 While Kumar et al. (1995) also include the implied bid-ask spread as a third control, we decide against that for three reasons: a lack of strong theoretical basis, potential collinearity with trading volume, and small sample size caused by positive auto-covariance.