【正文】 analysis, the dummy independence assumes a value of unity for the most independent banks and zero for all the others (banks with block shareholders). The agency cost hypothesis cannot be rejected if the coefficient associated with independence is positive and statistically significant. Since banks are regulated, and major regulatory shifts occurred during our reference period 2020– 09, it is necessary to control for their influence in our model. The degree of regulatory pressure should capture the differences in the dividend payouts across distinct degrees of capitalization and risk appetites. The assumption is that banks with (riskweighted) capital ratios below or close to the minimum requirements will be subject to closer monitoring by their supervisors (see Section 1). Previous studies captured the effect of regulatory pressure by deploying the ratio of equity to total assets. However, because regulators closely follow the regulatory definition of capital, we measured regulatory pressure (capitalization) as the average of the tier 1 leverage ratio (tier 1 capital to assets) during the reference period. Lower leverage (., higher values for capitalization) signals stronger financial health and is expected to be associated with higher dividend payouts. Therefore, a positive relationship is expected between capitalization and dividend payout. Additionally, we included a dummy variable for regulatory pressure based on the capital categories of the Federal Deposit Insurance Corporation Improvement Act (FDICIA).14 Section 131 of the FDICIA establishes a system of prompt corrective actions derived from a classification system that divides banks into five categories: “ well capitalized ” , “ adequately capitalized ” , “ undercapitalized ” , “ substantially undercapitalized” and “ critically undercapitalized.” Banks are classified according to thresholds using riskbased capital and leverage ratios as the basis. The majority of the banks in our sample were classified as “ well capitalized” . We consider that regulators increase their pressure on bankswhen banks are approaching the minimumlevels of capital and not only when those levels are breached. Therefore, the banks considered to be subject to increased regulatory pressure are those not classified as “ well capitalized” and those currently classified as “ well capitalized” but which may be downgraded (., banks that present leverage or riskweighted capital ratios close to the limits of adequate capitalization). For the purpose of this variable, the following thresholdswere considered: 8% instead of 6% for the tier 1 riskweighted capital ratio, and 7% instead of 5% for the 共 頁 ， 第 5 頁 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 裝 ┊ ┊ ┊ ┊ ┊ 訂 ┊ ┊ ┊ ┊ ┊ 線 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ tier 1 leverage ratio. To capture this effect, we included a dummy PCA in our model. This variable assumes a value of unity if a bank does not meet at least one of these thresholds. In total, the variable assumes a value of unity for 44 observations before the crisis and 73 observations during the crisis. Undercapitalized banks (banks subject to regulatory pressure) are expected to be associated with lower dividend , a negative relationship is expected between PCA and dividend payout. In order to test the hypothesis that undercapitalized banks faced greater regulatory pressure to plowback their earnings than wellcapitalized banks, we considered the interaction of PCA with profitability. With this variable, the relevance of profitability as a determinant of banks39。s test indicated the presence of heteroskedasticity in the data. Because the dependent variable (dividend payout) does not assume negative values, the distribution can be considered censored to the left, a situation in which OLS can produce inconsistent estimates (Wooldridge, 2020). While the issue in the sample is not severe, Eq. (1) was estimated with a TOBIT regression. We present the baseline findings of our model in Table 5, considering only capitalization to capture the effect of regulatory monitoring (column 1), and additionally PCA and its interaction with profitability (column 2) to capture the effect of regulatory pressure on undercapitalized banks. The findings indicate that the Fama and French (2020) characteristics of dividend 共 頁 ， 第 8 頁 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 裝 ┊ ┊ ┊ ┊ ┊ 訂 ┊ ┊ ┊ ┊ ┊ 線 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ payers can be applied to banks: size and profitability are positively and significantly related and historical growth is negatively related to dividend payout, . larger and more profitable banks paid larger dividend payouts, and banks with low historical growth opportunities also paid more dividends. These findings are consistentwith the findings of previous studies (Casey and Dickens, 2020。s initiatives to limit dividend payouts by undercapitalized banks. Because our findings provide robust empirical support for the signaling and agency cost hypotheses, the reforms may have an unintended impact on the use of dividends as both signaling and agency cost reduction mechanisms. Inability to use these governance mechanisms may reduce the potential to attract external financing, both debt and equity. The level up to which regulators may want to allow signaling and agency mechanisms to function is an issue that deserves serious attention from academics and regulators alike. Acknowledgments We are grateful to the Managing Editor, Jeffry Netter, and the anonymous reviewer, for their ments and suggestions. We acknowledge the financial support from “ Funda2