Companies’ investment determinants: comparison of different panel data estimators

Abstract

In this study, Aivazian, Ge and Qiu’s (2005) analysis using static panel models is extended to using dynamic panel estimators, considering data for listed Portuguese companies. The results confirm Aivazian et al.’s (2005) conclusion that an ordinary Least Squares (OLS) regression is not the best way to estimate the investment/determinant relationship. Investment decisions are probably dynamic, so the most suitable way to estimate the investment/determinant(s) relationship is using dynamic panel estimators. Alternatively a fixed effect panel model can be used, consistent with a first order autocorrelation. In this way, firstly, it is possible to determine more accurately the positive impact of sales (Neo-classic theory) and cash flow (Free Cash Flow theory) on the investments of listed Portuguese companies. Secondly, the positive effect of growth opportunities (Agency theory) is not overestimated when it seems to be the consequence of a first order autocorrelation. Using dynamic panel estimators permits correct measurement of dynamism in company investment decisions by examining the relationship between investment in the previous and the current periods.