Housing Demand with Random Group Effects

This paper examines the random group effect, which has usually not been considered in traditional housing demand studies. Frequently, group level variables are used in housing demand estimation due to the data constraint. For instance, the US Index of Housing Price per administrative area is often used to measure the housing price when estimating the US price elasticity of demand for housing, and the average household income is often used as a proxy for the individual income in Taiwan when estimating the income elasticity of demand for housing. Econometricians argue that the traditional OLS estimation, when the random group effect is ignored, has been considered to have a downward bias in the estimated standard error. By following Amemiya (1978) and Borjas and Sueyoshi (1994), we propose a two-stage estimation technique to estimate housing demand with the random group effect. Using Taiwan’s cross-sectional survey data, we found that the standard error of the estimated coefficient for the group level income variable is underestimated in the traditional unadjusted OLS specification. This finding suggests that there may be a danger of spurious regression in the traditional OLS housing demand estimation