The Risk and Return of Commercial Real Estate: A Property Level Analysis

Abstract

I compare the performance of the index-based time series approach and the cross-sectional approach in estimating factor loadings of nontraded assets, and show that the latter likely provides less biased and more efficient estimates. I then use the cross-sectional approach to estimate the loadings of privately owned commercial real estate on the Fama and French (1993) factors, the Pastor and Stambaugh (2003) liquidity factor, and two bond market factors, using a sample of 14,115 properties in the 1977-2012 period. I find statistically significant loadings, of which the signs seem consistent across property types, but the magnitude varies. Using the time series approach on the same data, I find insignificant loadings on virtually all factors. To investigate the sources of the weak results from the time series approach, I conduct a Monte Carlo simulation in which both approaches are correctly specified and indices can be estimated perfectly. Simulation results suggest that the cross-sectional approach provides more accurate estimates under reasonable market conditions.

Risk and Returns of Income Producing Properties: Core versus Noncore

“…A standard three‐stage approach ( e.g ., Case and Shiller ) to address this is to estimate the model using ordinary least squares (OLS) as the first stage, regress squared OLS residuals against the duration in the second stage and then use the fitted values of squared residuals as weights to estimate the model again using weighted OLS in the third stage. However, we find that squared OLS residuals are not positively correlated to duration, which is consistent with Peng (, ) and Sagi (). Therefore, we estimate Equation with OLS and calculate and report White's heteroskedasticity‐consistent standard deviations.…”

“…All cash flow variables are on an unlevered basis. Previous research, such as Pivo and Fisher () and Peng (), has used earlier releases of the NCREIF database.…”

“…We overcome this challenge by using the approach proposed by Peng () in the framework of the holding‐period factor model, which allows us to test whether alpha is related to properties building expansion or initial occupancy rate without estimating alpha. The holding‐period model is first adopted by Cochrane () to estimate the beta of venture capital investments, and then by Korteweg and Sorensen (), Driessen, Lin and Phalippou () and Franzoni, Nowak and Phalippou () to estimate factor loadings of private equity, and by Peng () to estimate factor loadings of private income producing real estate. In the framework of this model, Peng () proposes a novel approach to test whether individual properties’ acquisition cap rates predict their ex post returns and alpha, which does not require the estimation of alpha.…”

“…Private income producing properties ( e.g ., apartments and commercial real estate) constitute a large portion of the total wealth in the United States’ economy, and provides valuable diversification benefits for investors (see, e.g ., Brueggeman, Chen and Thihodeau , Hartzell, Hekman and Miles , Geltner , Goetzmann and Ibbotson , Plazzi, Torous and Valkanov , Peng ). Income producing property investors generally follow three main strategies—core, value‐added and opportunistic (see, e.g ., Peyton ).…”

“…This is because properties are infrequently traded, and we do not have the time series of returns to fit the standard factor models to estimate alpha for each property. We overcome this challenge with a holding‐period factor model, which Cochrane (), Korteweg and Sorensen (), Driessen, Lin and Phalippou (), Franzoni, Nowak and Phalippou () and Peng () use to estimate the risk of nontraded assets, including venture capital, private equity and private income producing real estate. In the framework of this model, Peng () proposes a novel approach to test for possible relationships between alpha (or factor loadings) and property attributes, which does not require the estimation of alpha (or loadings).…”