Hoyt Model Backward Testing

Executive Summary

The Homer Hoyt Institute started investing in REITs in 1992. At that time, it also started development of a model for investing in REITs. The Institute has used this evolving model, known as the Hoyt REIT Model (Model), in its REIT investment program since 1992. The Institute's cumulative return on its REIT portfolio for the first few years of investing was twice that of the NAREIT Equity REIT index, an accepted industry benchmark. Currently, the cumulative return on the portfolio is approximately one and a third times that of the benchmark NAREIT Equity REIT index.

The Model was originally conceived and developed as a valuation model. Subsequently, a risk analysis model was created and merged with the valuation model. Thus, the current Model is used for REIT selection on an individual basis and for constructing portfolios diversified by type of risk.

The Model started as a judgmentally-based model and has since been rigorously enhanced by a variety of statistical analyses, including analysis of variance, correlation analysis, cluster analysis, and multiple regression. Multiple regression techniques were used to develop a reduced-form regression model that confirms the findings of the judgmental model. The multiple regression model uses significantly less data and offers greater objectivity.

This report does several things. First, it reports on the performance of Homer Hoyt Institute investments in REITs from Model inception to its present stage of development. Then, it shows how that performance might have varied had the Model output been available throughout the period.

This report provides quarterly estimates of the returns that would have been earned had the current form of the Model and its outputs been available throughout the period. These estimates are calculated using decision rules based upon trigger points for buying and selling. The focus is on individual REIT selection.

The Institute has used multiple criteria for its portfolio construction. But, formalized decision rules have yet to be developed for an automatic process utilizing the Hoyt Model for REIT portfolio construction.

Finally, this report explores the use of regression models built using existing model inputs. This is done as an opening to a discussion of the use of the Model as a decision aide. That discussion focusses on the development of an interactive model format allowing alternative judgements and/or use of the Model's analyses in other paradigms for decision making.

Hoyt Investment Performance

The Hoyt REIT investment portfolio has been developed over a four-year period. It has grown from a relatively small investment to a $1.975 million portfolio, accounting for one-third of the Institute's endowment and one quarter of its net worth. Investment performance has been outstanding, as shown in Chart 1. This chart shows cumulative performance of the Hoyt portfolio as compared to the benchmark NAREIT Equity REIT Index and two prominent real estate mutual funds.

The investment decisions that resulted in this portfolio were made by senior Hoyt Group officers with substantial professional and academic real estate experience. In their decision making, they relied, in part, on Model outputs and analysis provided by the senior analyst. Their decisions reflect two investment goals, total return and diversification. Thus, the investment returns described above are the result of a combination of expertise and the Hoyt Model.

Chart 1

Hoyt Model Investment Performance

Hoyt Model investment performance is shown in Chart 2. These charts show unweighted cumulative return over the test period.

These returns are based on a backward test of the Model over a three-year period using selected decision rules. They reflect what the Model's performance would have been over the subject period had the Model been available. Buy and sell decisions are reflected in hypothetical transactions that occur on a quarterly basis. These decisions are based on a value investment strategy, with no consideration given to diversification.

Chart 2

Use of Model as Decision Aide

The Hoyt Group has also developed a regression model to test the judgmental model. The regression model was backward tested, with impressive results. (See Chart 3.) The regression model can be used, in concert with the judgmental model, to explore theoretical aspects dealing with components of the Model and better explain REIT market behavior. As such, the regression model can be used as an aide, along with the Hoyt Model as an aide, to decision making.

The Hoyt Group is investigating a variety of ways of applying these models to assist investors in their decision making, especially portfolio construction. The Hoyt Group is in the process of developing an on-line product to complement its investment advisory services in order to assist investors in their decision making and fund further REIT analytical research.

Chart 3