The impact of 9/11 on US REIT returns: fundamental or financial?

In summary, the impact of the 9/11 attacks was to severely increase the uncertainty of returns in financial markets in the US but was in the main short lived. However, the impact of 9/11 on real estate returns is potentially more damaging and longer lived, especially for trophy buildings across the US. The following section therefore tries to quantify the impact of 9/11 on real estate returns to sort out whether the effects were short or long lasting, i.e. financial or fundamental.

3. METHODOLOGY AND DATA

The methodology used to determine whether the effects of the 9/11 crisis were fundamental or purely financial is that suggested by French and Roll (1986). French and Roll (1986) focused on the sharp drop in the hourly volatility of returns when exchanges are closed. They noted that if hourly stock return variances were constant across trading and non-trading periods and if returns are independent and identically distributed (i.i.d.), the variance of weekend returns (i.e., Friday close to Monday close) would be tree times the variance of weekday returns (e.g., Tuesday close to Wednesday close). The observed ratio of weekend variances to weekday variances of only 1.107, rather than 3.0, suggests that prices are much more volatile when markets are open. French and Roll (1986) hypothesized that the higher volatility during trading hours results from either information (both public or private) that arrives primarily when markets are open or noise trading that also occurs only during market hours.

A major difference between information and noise trading, according to French and Roll, is that information-based changes in stock prices persist while noise trading effects (the result of trading errors, miss-pricing, and overreaction) tend to be short-lived and reversed in subsequent periods. Following Perry (1982), French and Roll discriminate between the effects of information and noise trading on volatility by comparing daily return variances with the daily variances implied by variances for longer holding periods.

Tuluca et al (2003) extend this concept and suggest that the effects of information and noise trading on volatility can be detected by comparing the actual return variances of any shorter period with the implied variances derived from longer holding period, and use this idea to test the impact of the Asian crisis on international stock markets by comparing the actual daily return volatility versus implied daily volatility derived from monthly data. We follow this approach in the current paper and test the impact of 9/11 on real estate returns by examining the differences of post- and pre-crisis variance ratios and interpret an increase in the difference as a fundamental (informational) effect and a decrease as a purely financial (noise trading) effect.

The study, therefore, needs daily data from the US. However, private real estate data is only available on a quarterly basis. In contrast, performance in the public market is easier to measure, since daily share prices are readily available from the National Association of Real Estate Investment Trusts (NAREIT). The data composed of an overall REIT index (ALLREITs) and three sub-indexes: (1) Equity REITs (EREITs), (2) Mortgage REITs (MREITs) and (3) Hybrid REITs (HREITs). Since the event is on Tuesday 11th September 2001 we chose the following pre- and post-crisis periods: pre- Monday 13 March 2000 - Monday 10 September 2001 (391 daily prices corresponding to 390 daily log returns), 18 monthly prices (2nd Monday of the month, corresponding to 18 log returns): post- Wednesday 12 September 2001 - Wednesday 14 March 2003 (391 daily prices, 390 daily log returns), 18 monthly prices (2nd Tuesday of the month, corresponding to 18 log returns).

However, in using public market data to represent the performance of real estate a key question to be faced is whether REITs are a stock or property? It can be argued that the performance of REITs is ultimately dependent upon the underlying private market in that the asset values of the companies depend upon the capital value of the real estate owned, the ability to pay dividends depends upon the NOI from the property and the ability to trade profitably depends on increases in capital values which, in turn, depend on rental change and expectations of future growth. As a result, a close link between the public and private real estate might be expected. Nonetheless, the return behaviour of REIT returns is, in many ways, more similar to that of other equities (particularly small cap stocks) than of the underlying private real estate. Certainly, a number of studies have shown that REIT returns have much closer contemporaneous correlations with the stock market than with the underlying real estate market. Typical coefficients range between 0.65-0.85. Although, more recent statistical evidence finds that the equity component seems to be consistently shrinking over time while the sensitivity of REIT returns to the private real estate showed a significant increase in the 1990s (Gordon and Canter (1999) and Clayton and MacKinnon (2001)). This implies that the behaviour of REIT returns and those of the private real estate has tended to converge since the early 1990s.

In addition, previous studies demonstrate that once the equity component from REIT returns is removed, the resulting residual series is much more closely related to private real estate; see Clayton and MacKinnon (2001), Liang and McIntosh (1998), Sanders (1998), Chiang and Lee (2002) among others. This correlation also rises as longer time periods are analyzed; see Campeau (1994), Glascock et al (2000), Li and Wang (1995), Grissom and Oppenheimer (1998), Liang and Naranjo (1999), Okunev et al (2000), Wilson and Okunev (1996) and Wilson et al. (1998) among others. Thus over longer periods of time, REITs tend to behave more like property, even allowing for the presence of a large equity component. These findings suggest that if we remove the pure equity market component from the REIT returns, the resulting residual series tends to be highly correlated with direct real estate returns. In other words, despite their legal standing as public securities, once allowances are made for the influences of broad stock market effects and differences in the underlying real estate, REIT returns are more closely tied to the private real estate than the "raw" data suggests.

Regressing REIT returns on the returns of competing assets the residuals from this model should represent the real estate component of REIT returns. In our case, we specify that REIT returns are a function of large cap stock returns, small cap stock returns, value, growth indexes and bond returns. We estimate this model using conventional linear regression techniques and then compute the regression residuals. This residual series contains the real estate component of returns, with the influences of the stock and bond markets removed. Once this is done we represent the results in terms of differences of post- and pre-crisis ratios and interpret an increase in the difference as a fundamental (informational) effect and a decrease as a purely financial (noise trading) effect. Consequently, we focus on actual daily return volatility versus implied daily volatility derived from monthly data for the REITs. In other words, if security price changes result primarily from information changes that persist and if daily returns are independent and identically distributed (i.i.d.), the implied daily variances for a longer holding period, such as a month, would approximate the actual daily variances within the period. On the other hand, if daily returns reflect noise trading effects that tend to be short lived and reversed in subsequent periods, the implied daily variances for the longer period would be smaller than daily variances within the period. We represent the results in terms of differences of post- and pre- crisis ratios and interpret an increase in the difference as a fundamental (informational) effect and a decrease as a purely financial (noise trading) effect.

4. RESULTS

Columns 1 and 2 of Table 1 present the pre- and post-9/11 crisis variances of daily returns in the four REIT indexes. Columns 3 and 4 show corresponding variances of monthly returns (implied daily variances). Column 5 (Post/Pre) shows the ratios of the post-crisis daily variance (column 2) of returns to the pre-crisis daily variance (column 1). A Post/Pre ratio of 1.76 for the ALLREIT market as a whole indicates that the variance of daily returns rose by 76% following the 9/11 crisis. Variances rose in two REIT types, EREITS (73%) and MREITs (45%), but fell by 7% for HREITS. For the 4 REIT types, the average percentage increase in the variance of daily returns was 47%. We use an F-test of variance equality to assess if variances are statistically different in the post-pre crisis periods, with all but HREITS showing significant increases at the 1% level. In other words, 9/11 had a significant impact or return uncertainty for most REIT security process except Hybrid REITs (HREITS).

Column 6 (Post/Pre) shows the ratios of post-9/11 to pre-9/11 variances using the monthly returns from columns 4 and 3, respectively. As with the daily variances, monthly (implied daily) variances increased following the attack on the World Trade Centre by more than 7% for ALLREITs and 11% for EREITS, but fell by 3% for MREITs and 45% for HREITS and by an average 8% overall. Using an F-test to assess if the monthly variances are statistically different in the post-pre crisis periods, none of the REIT indexes show a significant increase at the usual levels of significance. This suggests that the attack on the World Trade Center on 9/11 did not have a significant impact on the uncertainty on real estate securities in the US in the long term.

Tuluca et al (2003) argue that the increase in volatility in the post-9/11 period compared with the pre-9/11 period could result from either a fundamental change, due to information flows, or noise trading due to increased uncertainly which is transitory. Therefore, was the effect of the attack on the World Trade Center fundamental or transitory? In order to answer this question we compare the post-crisis increases in monthly (implied daily) variances (column 6) versus the post-crisis increases in daily variances (column 5), the results shown in column 7. As can be readily appreciated the increase in monthly (implied daily) market volatility for all the REITs markets was less than the actual daily variance, the largest decrease by HREITS (40%) and the least by MREITs (33%) with an average decline of 37%. Thus, we conclude that the effect of 9/11 on real estate in the US was financial and transitory rather than fundamental and persistent.

Print Get the Mag Weekly Updates