Capital adjustment cost and bias in income based dynamic panel models with fixed effects

The fixed effects (FE) estimator of "conditional convergence" in income based dynamic panel models could be biased downward when capital adjustment cost is present. Such a capital adjustment cost means a rising marginal cost of investment which could slow down the convergence. The standard FE regression fails to take into account of this capital adjustment cost and thus it could overestimate the rate of convergence. Using a Ramsey model with long-run adjustment cost of capital, we characterize this bias that does not go away even for a longer time dimension. The size of the bias is greater in economies with a higher adjustment cost. The cross-country regression suggests that the size of this bias could be substantial.