Saturday 29 June 2013

Canadian REITs and Interest Rates

The past 2 months have been very difficult for Canadian investors. First interest rate worries spooked the banks and REITs, then weak commodity markets torpedoed the resources sector. Gold, the precious metal that is the go-to investment in times of inflation and worry is now trading at a 2 year low. Most recently, Canadian telecoms got whacked on news that Verizon was thinking of moving into Canada. Real estate investment trusts (REITs) have been good investments for a long time. REITs pay bond size dividends and offer the opportunity for equity like price appreciation. REITs have been so good for so long, that many investors have a sizable portion of their investment portfolio in REITs. To find out more about what has been happening with REITs, I decided to analyze how sensitive one of Canada's biggest REIT ETFs, the iShares capped REIT index (XRE) is to movements in interest rates. Here is how XRE has performed since 2008.



The last recession was hard on XRE, but recovery came quickly.

Here is how Canadian 3 month T-bills  and 10 year government bonds have performed. It seems reasonable to expect that REITs are negatively correlated with T-bill or bond yields, since increases in fixed income yields offer competitive less risky alternatives to investing in REITs. These falling yields have helped push the price of XRE higher.


I collected monthly data on XRE, 90 day T-bill yields, and the yield on 10 year government of Canada bonds. I calculate the one month return on XRE and denote it as xre_r. I regress one month XRE returns on the yields from T-bills and bonds.

A regression of xre_r on the 90 day T-bill yield produces the following results.

xre_r = 1.613 -0.365 tbill

The estimated coefficient is negative indicating that a 1% increase the tbill yield reduces monthly returns by 0.365%. The sign of this  coefficient is as expected, negative, but the estimated coefficient is not statistically significant at conventional levels. The R squared for this regression is 0.0115. Not much going on here.

A regression of xre_r on the 10 year bond yield produces the following results.

xre_r = 1.239 -0.105 bond

As expected the estimated coefficient on the 10 year bond variable is negative. This estimated coefficient is not, however, statistically significant at conventional levels of significance.The R squared from this regression is 0.0005. This is even lower than in the T-bill regression.

On the face of it, there does not seem to be too much sensitivity of REITs to movements in interest rates. Another possibility is that the relationship between XRE and interest rates is time varying. The regression results reported above assume that the coefficient on the interest rate variable is constant over the sample period. This may not be the case, in which case, a time varying beta approach may be more informative. To investigate this I used a rolling window analysis to estimate the coefficient on the T-bill variable using a rolling window regression approach with a fixed window length of 60 observations.


Whoa! Now here is something interesting. Up until the beginning of 2012, the sensitivity of XRE to the T-bill yield was fairly constant. Starting in early 2012, however, the relationship changed with REITs becoming more sensitive to interest rates. The most recent  value of the estimated coefficient on the T-bill variable is -5.41. This means that a 1% increase in the T-bill yield decreases monthly returns on XRE by 5.41%. For most of the sample period, REIT investors were not too sensitive to movements in the 90 day T-bill rates. That has clearly changed over the past year. REIT investors have become much more concerned with rising interest rates. With falling REIT prices, the yields on REITs will eventually start to look good on a risk adjusted basis. Given the large sell off in REITs, however, this could take some time.