Questions of investment "timing" and the possibility that real estate returns may be somewhat predictable have recently moved to the forefront of the concerns facing institutional investors in real estate. This is largely a consequence of the experience of many investors who made substantial property investments during the mid 198Os, only to see these assets lose much of their value when the property markets experienced several years of substantial losses in value during the late '80s and early '9Os. Even "optimal" diversification (as exemplified by the diversity in the Russell-NCREIF Index) would not have saved investors from this historical loss of value.
This article, which is based on a more extensive research report, seeks to address this concern by presenting some tools from modern financial economics which can be useful in predicting real estate returns and applying this predictability in the property valuation and investment processes. These tools may also be used to analyze the nature of risk and return in the private (i.e., "unsecuritized") property market. In particular, the research on which this article is based accomplishes four major tasks:
* Develops a forecasting model for predicting real estate cash flows and returns;
* Applies the forecasting model to demonstrate its use in developing a simple "buy/sell" decision rule to aid in investment timing decisions;
* Demonstrates how the forecasting model can be imbedded in an "improved" present value model to account for the time-variability and predictability of real estate returns in the valuation of property; and
* Uses this improved present value model to develop a simulated historical value series which shows the relative importance of changes in return expectations versus changes in cash flow expectations in producing changes in real estate value.
The forecasting model
The first step is to develop a model for forecasting the total nominal return and the operating cash flow of privately traded commercial property. The return forecasts from this model will then comprise the "expected" returns, or "discount rates" applied in the present value model of property value. The cash flow forecasts will represent the "numerators," or cash flow expectations, in the present value model.
The analysis is conducted on annual returns and cash flow data from 1975 through 1992. The real estate returns are based on Russell-NCREIF returns (extended back to 1975 by splicing PRISA returns onto the series prior to 1978), corrected for appraisal smoothing. The cash flow series (NOI level) is derived from the income and appreciation return components of the Russell-NCREIF/PRISA data.
The forecasting model is a first-order vector autoregression (VAR) with five variables. In addition to the real estate returns and cash flows, we use REIT returns, appraisal-based returns, and appraisal-based yields. The model predicts the (unsmoothed) real estate market returns with an adjusted R(2) of 56 percent, and the cash flows with an adjusted R(2) of 94 percent.
Once developed, the forecasting model can then be used to create market timing rules. For example, the rule:
Buy, if model predicts two years of above average returns;
Sell, if model predicts two years of below average return.
results in the buy/sell signals indicated in Figure 1. (Figure 1 omitted) Figure 1 shows the historical market value profile of commercial property (unsmoothed), with the-buy/sell signals indicated. The model and rule gave buy signals in 1976-80 and sell signals in 1981 and 1984-90. Interestingly, a buy signal was given again in 1992 for the first time in 10 years.
Improving the present value model
In the present value model used to value commercial property in a discounted cash flow framework, the discount rate is traditionally assumed to be constant. This discount rate is supposed to represent investors' expected total returns to the property investment reflecting investors' risk perception and risk tolerance for the real estate asset. …