The NCREIF return index has done more to shape U.S. economists' perceptions of real estate performance than any other return index. The NCREIF return index is produced quarterly by the National Council of Real Estate Investment Fiduciaries. The index is compiled by totaling up the values and cash flows for all unlevered properties in the database. A value is reported to NCREIF for every property, every quarter. Most of these values are, however, appraised values, appraised as of one or more quarters previously. As a result, practitioners and academics alike have questioned the accuracy of the NCREIF return index. Several economists have attempted to amend the NCREIF return index, but these amendments have concentrated on adding significant positive autocorrelation into the return series. Others have simply abandoned the use of the NCREIF return index altogether in favor an equity Real Estate Investment Trust (REIT)-based return index. It is well established, however, that an investment in the stock of a REIT is not exactly the same as an outright purchase of real estate, and that the two series do indeed behave differently in some respects that are quite important from a portfolio allocation perspective. The current paper presents a new type of evidence about the performance of real estate that avoids the statistical problems discussed above. We begin by applying a simple filter to an equity REIT-based return index to correct for the fact that REIT values derive from a very different type of market structure than that in which the private property market operates. This constructed return series is very useful. First, it makes it possible to disentangle true private property price movements from pure common stock price movements. Second, it allows us to see what errors are imbedded in the NCREIF return index. By comparing the constructed real estate return index with the NCREIF return index, it is possible to analyze and quantify the systematic differences between the two. The ultimate purpose of this analysis is, however, to examine the volatility characteristics of this new series and its conditional covariance with contemporaneous returns on other assets, principally common stocks. We analyze institutional investors' preferences for real estate by measuring the conditional covariance of real estate with those common stock categories that are meaningful to institutional investors. Modern experience suggests that institutional investors, when compared with other investors, prefer stocks that have greater market capitalization, are more liquid, and have higher book-to-market ratios and lower returns for the previous year. Prior studies of real estate performance miss this effect entirely. They assume that institutional investors do not differentially invest in stocks that have these characteristics. They then proceed to construct a single measure of the covariance of real estate with common stocks. But this estimation technique leads to a resulting covariance measure that is quite different from the ones presented here, and almost surely much less accurate. Our new conditional covariance estimates are much higher, and change significantly for different types of stocks (e.g., large-capitalization versus small-capitalization stocks). The higher covariance means that institutional investors, in particular, should allocate a much smaller fraction of their portfolio to real estate than most observers seem to think they should.

This item is no longer available.