Print
E-MailAntonio Gledson de Carvalho is an Assistant Professor at the Fundação Getúlio Vargas School of Business at São Paulo, Humberto Gallucci Netto is a Professor at the Federal University of São Paulo, and Bernard S. Black is Nicholas D. Chabraja Professor at Northwestern University. This post is based on their recent paper forthcoming in the European Corporate Governance Institute’s Finance Research Paper Series. Related research from the Program on Corporate Governance includes The Elusive Quest for Global Governance Standards (discussed on the Forum here) by Lucian A. Bebchuk and Assaf Hamdani; What Matters in Corporate Governance? (discussed on the Forum here) by Lucian A. Bebchuk, Alma Cohen, and Allen Ferrell; and Learning and the Disappearing Association between Governance and Returns (discussed on the Forum here) by Lucian A. Bebchuk, Alma Cohen, and Charles C.Y. Wang.
Researchers know little about what factors influence firms’ choices about firm level corporate governance (FLCG or CG). Durnev and Kim’s (DK, 2005) theoretical model predicts that (Prediction 1) firms with better investment opportunities, more concentrated ownership, and greater need for external financing (EFN) will have better FLCG; (Prediction 2) firms that have better FLCG are valued higher; and (Prediction) 3 these relations are stronger in weak legal regimes.
Some evidence has been found for Prediction 1, in particular. However, Doidge, Karolyi and Stulz (DKS, 2007) challenge Propositions 1 and 3. They show that country, rather than observable firms’ characteristics (OFC), explain most of the variance in FLCG; and that OFC explain very little of the variation in FLCG in less-developed countries (where weak legal regimes prevail). Country case studies confirm the negative result of DKS. In our paper, we claim that there may be a design problem in these studies because they use a linear model to study FLCG.
There are several concerns with whether a linear model is adequate to study FLCG. First, FLCG involves firm rules and practices. Changing FLCG is costly and may require time. Thus, one would not expect that firms adjust CG immediately every time the need for external finance changes. Second, improving FLCG is easier than worsening it, because investors may interpret worsening CG as a signal that a firm is more likely to expropriate investors. Thus, one should expect that the left-hand side of the regression model will move less frequently than the predictors on the right-hand side, and will move up more often than down. Third, some investors may have a CG threshold that they require to buy a firm’s shares. Thus, firms either reach the threshold level and can raise capital from investors, or keep their CG at a low level. Changes in the threshold level that investors demand will lead firms that met the previous threshold to adjust their governance, even with no change in financial characteristics. Given such dynamics for CG, it may be convenient to study changes in CG rather than the CG level. We do so using both linear and probit specifications (for cardinal changes and improvements in FLCG, respectively).
Brazil constitutes a unique ground to test the DK Propositions. First, one can objectively track FLCG over an extensive time period (2010-2019). Second, a common challenge in testing Proposition 3 is how to meaningfully classify countries into high and low-quality legal regimes. Legal regime is a country-specific characteristic that usually changes slowly over time, and will be related to many other country-specific characteristics. Studying Brazil provides an opportunity to circumvent this problem. In Brazil, strong and weak legal regimes for CG coexist, against the background of other legal rules and cultural influences that apply to both CG regimes.
We report several interesting descriptive aspects of the evolution of FLCG, that prior research has not highlighted. First, the improvement of FLCG is uneven over time. In the first half of the sample period, there was progressive, substantial improvement. In contrast, in the second half, FLCG was mostly stable. Similar patterns, where governance changes rapidly in one time period, and more slowly in another period, have been reported in other countries, but not recognized as a pattern. Second, some firms do not adjust their FLCG (or do so minimally), even when their initial levels were low. Third, positive changes in FLCG are much more common than negative changes. Moreover, most negative changes are small; large negative changes are very rare. Fourth, for firms under the strong legal regime for CG, we find compression over time in FLCG; at some point, there is little further improvement left for firms to do.
Consistent with prior work, we find limited evidence that firm financial characteristics predict the CG level. However, when we shift from predicting levels to predicting governance changes, we find evidence that a multi-year measure of equity financing need (EFN) predicts CG improvements for firms under the weak legal regime for CG (as in DK Proposition 1). In contrast, neither EFN nor the other firm-level variables we study predict CG changes for firms under the strong legal regime. This asymmetry between weak and strong legal regimes is consistent with DK Proposition 3, and with investors requiring a CG minimum threshold to buy shares (given that firms in the strong CG regime already satisfy it).
A measure of asset tangibility predicts fewer CG improvements for firms under the weak legal regime. This is also consistent with theory. Investors may find it easier to understand and value firms with more tangible assets, so governance may be less critical for firms with more tangible assets.
We find predictive power only during the first half of the sample period, when governance is changing rapidly, and not the second half, when changes are much more limited. A methodological implication of our results: to study FLCG with two-way fixed effects, one needs a panel data covering a period of substantial change.
