How Does Unobserved Heterogeneity Affect Value Relevance?

Most value relevance (VR) studies consider an accounting item value relevant if the regression coefficient (RC) of that item is statistically significant. Unobservable heterogeneity leads to biased RCs, interpretation of which generates incorrect inferences. To obtain unbiased RCs, the effect of unobservable heterogeneity on RCs should be mitigated. As two dimensions of unobservable heterogeneity are at the firm level and time level, outcomes with the following unobservable heterogeneity concerns are discussed: i) no fixed-effects (FE); ii) firm FE; iii) time FE; and iv) two-way (firm and time) FE. By employing a sample of Turkish firms from 2005-2014, we report several findings. First, we find that regressions with firm (time) FE yield large (low) RCs vis-a-vis regressions with no FE, and regressions with two-way FE generate balanced RCs compared to the others. Second, we compare RCs with i and iv, and conclude that the book value of equity becomes more value relevant while net income does not after controlling for unobservable heterogeneity. Last, we arbitrarily divide the entire period into two to reveal how unobserved endogeneity affects the comparison of RCs belonging to different periods. Our outcomes robustly reveal that unobserved endogeneity leads to erroneous RC comparisons.