|
by Susan Hudson-Wilson and Bret R. Wilkerson
Illustration by James Yang
Modern portfolio theory has increasingly gained acceptance in the private market as the best method for asset allocation and diversification of pools of real estate assets. MPT is the set of quantitative techniques, first developed by Harry Markowitz in the 1950s, that identifies portfolios offering the lowest risk for a given level of expected return. The role of the market cycle, as contrasted with the uniqueness of individual properties, became apparent during the last decade to anyone with more than one investment in Houston, New York, Los Angeles, or anywhere else. Careful portfolio construction - in the sense of stepping back from the details of property selection and focusing on the risk and return of the whole portfolio - can make or break the performance of a pool of assets.
Portfolio optimization, the basic tool of MPT, depends on thoughtful estimates for three crucial inputs: property selection and focusing on the risk and return of the whole portfolio - can make or break the performance of a pool of assets.
Portfolio optimization, the basic tool of MPT, depends on thoughtful estimates for three crucial inputs: expected returns, the riskiness or volatility of each return, and the cross-correlations among the returns. Historical and forecast returns on a market-by-market, property-type-by-property-type basis are needed to calculate these inputs. It is essential that these data be consistent across time, geography, and property types.
We modeled the behavior that was consistent and was forecasted to occur in each city and property type. Our DMRs (derived market returns) are modeled estimates of the historical and forecast performance of real estate for 60 urban areas and four property types.
Creating this data is a necessary, but hardly a sufficient, condition for creating a portfolio model. Our model produces results for 240 urban area/property type pairs, but no optimizer can (or should) incorporate this many assets. The next step was to unearth the "dimensions of diversification." How many truly differentiable investment choices are available in the U.S. property markets? We needed to discover the "sectors" (to borrow a term from the stock and bond markets) of the U.S. property market. The math of cluster analysis provides the means to do this. We think there are collections of urban areas and property types that share patterns of behavior. In other words, some groups of markets have very similar expected returns and historical return profiles.
Returns are produced by the complex interrelationships among demand, supply, and capital market factors, such as inflation and investment opportunity costs. As a result, some clusters are composed of markets that are intuitively thought of as substitutes for one another, and some are not! For example, the San Antonio and Austin, Texas, apartment markets have moved nearly in lock step with each other over the past 20 years, while the Cincinnati and Columbus, Ohio, apartment markets have an extremely large negative correlation. Clearly, geography is not the correct guide with respect to diversification.
This analysis, then, provides a rigorous answer to the question of which markets are complements and which markets are substitutes in the context of a portfolio. If two markets are complements, they move differently from one another and the investor needs to buy both to achieve diversification. If two markets are substitutes, they tend to move together and the investor can choose one or the other. Markets don’t need to have a correlation of negative 1 to be helpful, nor does the correlation even have to be negative. Any correlation less than 1 can be helpful as well.
Now that we have a model of how the markets relate to one another, it is time to investigate the nature of the investment universe confronting us. To do this we construct a market efficient frontier. Efficient frontiers in the context of any asset type describe the highest attainable returns associated with each level of risk. Each point on the frontier corresponds to a particular portfolio. The frontier is called "efficient" because it provides the highest expected return for each level of risk or, equivalently, the least risk for each level of expected return.
The frontier (see graph below) assumes that the investor has perfect mobility and can instantly and costlessly reweight investments to create a portfolio right on the efficient frontier. Of course, this is unrealistic. Three kinds of constraints influence REIT investment behavior: constraints on what can be sold, constraints on how much capital can be invested, and constraints on what assets the company is willing to hold. By incorporating these constraints into the portfolio, one may construct the efficient frontier actually faced by the investor. This constrained frontier more realistically represents the investor’s opportunity set. Companies that choose to invest in only one property type automatically have an efficient frontier that is lower than the unconstrained frontier.
What About REIT Portfolios?
Using data from REIT portfolios as of the end of the second quarter of 1999, we calculated the portfolio risk and the expected return of each company in our coverage universe. Recall that the risk level for each portfolio is determined by the risk levels of the individual markets in which a company has invested and the cross-correlations between markets. The expected return for each company’s portfolio is not our forecast for the stock price, but rather is comprised of an appropriate weighting of each of our DMRs, which are an expression of the average total returns privately held assets should expect to earn in a market. We have selected a dozen companies in order to illustrate a few points. The data on each of the portfolios is shown in the table below, while the portfolio locations relative to the efficient frontier can be seen in the graph below.
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
The risk level of individual portfolios within a property type sector may be similar, but they also may be very different. Interestingly, the risk level of Highwoods’ portfolio is actually less than that of Equity Office. EOP has assets in a large number of office markets, and HIW primarily has assets in very volatile Southeastern office markets, so "tummy rubbing" suggests that EOP would have a less risky portfolio. That’s why we run the numbers! Some of the markets that HIW is invested in have lower correlations with the rest of its portfolio, and the company has a small (but apparently integral) amount of its assets in other property types that help to diversify the large, risky office investments. On the flip side, with HIW’s office investments in markets where we expect the real estate fundamentals to continue to erode, the expected return of the portfolio is also lower than that of EOP.
Within a sector there can be tremendous differences in expected returns. In the office sector, SL Green has the highest expected returns and is concentrated in a single market. Our forecast for the New York office market is one of the strongest of all the markets we track. The very high expected return and the high-risk level place its cluster as the "anchor" on the efficient frontier. When things are good, they’re great, but eventually they will turn down, and at that point the lack of diversification will definitely hurt.
The two retail companies on the chart, Bradley Real Estate and Western Properties Trust, have very similar risk levels even though they concentrate their portfolios in different regions of the country. However, there is an expected return spread of 225 basis points - with Bradley’s portfolio expected to outperform. Returns are driven by the interaction between supply and demand, and while many of the West Coast markets that WIR invests in have a great demand story, several of them are seeing a significant supply response. Some of the more "boring" Midwest retail markets have attracted less development capital and appear to have a healthy period ahead.
Is a bigger portfolio a lower risk portfolio? Not necessarily. In addition to the EOP and HIW example, the relative positions of AIMCO and Mid-America Apartment Communities demonstrate that a smaller and more selectively concentrated portfolio can be less risky. MAA, while having some exposure to volatile (for apartment) markets, has exposures to some low volatility markets that are not perfectly correlated. In fact, the larger companies, by virtue of having so much capital to invest, look more and more like the index for the property type, and therefore have portfolios that are inefficient by definition.
Diversification is not diversification is not diversification. Looking at the three REITs labeled by many as "diversified," Washington REIT invests in only one market, Pennsylvania REIT is invested primarily in two property types, and Transcontinental Realty Investors is exposed to more of the clusters (as described above) than any other REIT we track. Each has a very different risk and return profile.
A final way to differentiate between portfolios is to look at risk-adjusted returns. In this case, we examine the portfolios of two industrial companies, EastGroup Properties and Cabot Industrial Trust. While EGP’s portfolio has a much higher expected return, CTR’s portfolio is less risky. However, look at the final column in the table on page 37, the modified Sharpe ratios where expected return is divided by the risk of the portfolio. In this case, it appears that CTR’s portfolio is superior because the company is getting paid more per unit of risk.
Lessons
Each REIT portfolio is distinct, and the differences are more dramatic than has generally been reflected in stock prices. While much attention has been paid to management teams, strategies, balance sheet issues, etc., in our opinion there has not been enough attention paid to asset allocation decisions. The portfolio construction matters and should be reflected in the stock price, the companies’ cost of equity and debt, and the story that companies are telling their investors. We believe individual real estate markets will continue to experience very different cycles from one another, and that going forward the market will begin to differentiate more and more between individual REIT portfolios. As long as this is the case, there will be opportunities for smart investors in public real estate companies.
No matter how real estate is held or how often it is priced, the long-term driver of performance will be the underlying property and the market in which it is located. Because of structural and capital markets issues, REITs have less flexibility when it comes to managing tomorrow’s portfolio. This does not suggest that investors should ignore the cycles of the markets they are invested in, but rather that they need to pay extra attention to those markets and their respective cycles. In particular, investors need to pay attention to the interaction of those cycles, so that when one market suffers from overbuilding, there are other exposures in the portfolio that are in the "sweet spot."
As for the REITs themselves, the mantra here is "rebalance, rebalance, rebalance." Stock and bond portfolio managers may rebalance their portfolios with astounding frequency, in some cases daily. While it is clearly not feasible to rebalance real estate portfolios daily, it is important to keep abreast of shifts in the market and to adjust portfolios as needed. When the benefits of shifts in the allocation outweigh the costs of shifts - SHIFT!
This means that an empirical analytical framework must become part of the fabric of the portfolio management process. The value will be captured in greatly reduced transactions and due diligence costs, more productively used staff time, and - most important - more reliable and more understandable portfolio performance.