A Simple Estimator for Short Panels with Common Factors

There is a substantial theoretical literature on the estimation of short panel data models with common factors nowadays. Nevertheless, such advances appear to have remained largely unnoticed by empirical practitioners. A major reason for this casual observation might be that existing approaches are computationally burdensome and difficult to program. This paper puts forward a simple methodology for estimating panels with multiple factors based on the method of moments approach. The underlying idea involves substituting the unobserved factors with time-specific weighted averages of the variables included in the model. The estimation procedure is easy to implement because unobserved variables are superseded with observed data. Furthermore, since the model is effectively parameterized in a more parsimonious way, the resulting estimator can be asymptotically more efficient than existing ones. Notably, our methodology can easily accommodate observed common factors and unbalanced panels, both of which are important empirical scenarios. We apply our approach to a data set involving a large panel of 4,500 households in New South Wales (Australia), and estimate the price elasticity of urban water demand.