I want to examine the modern capital structure and the possible of reversal from performance to capital structure thereby test the efficiency-risk and franchise value hypotheses. Most of the techniques I have come across like the two stage least squares can not account for the dynamic heterogenous non stationary panel data.The same problem is applicable to GMM.These common estimators do not account for cross sectional dependence and the heterogenous nature of the cross sectional element. They assumed they are homogeneous although they account for endogeneity and simultaneity bias. But the nature of my samples are cross sections of firms that have different characteristics and they are of different age and size and they belong to different industry. The kind of series of the parameter is not likely to be stationary. I was thinking there could be causal panel method that can be use to estimate the causal relationship between capital structure and firm performance.
there are no simple ways for sure. using a standard regression model will give you all sorts of problems (endogeneity, omitted variable bias) when interpreting the results, and will tend to give you non-reliable answers. you need a natural experiment on capital structure (unlikely to exist) or some instrument that causes it to change, possibly a change in taxation. the change in taxation should be such that some firms are affected (i.e., the "treated") and some are not (the "control").
Thanks.I have tried to use the two steps generalized methods of moments.The results was significant when the robust and nomata was not included in the model.I started from the static panel and finally employed the dynamic twostep GMM estimator.
Hi,
Perhaps you may benefit from reading this paper:
J L Coles, M Lemmon, JF Meschke, "Structural models and endogeneity in corporate finance: The link between managerial ownership and corporate performance", Journal of Financial Economics, 2012, 103, p 149 -168.
Best regards.