【正文】 (2020), occurred during the financial crisis. We consider that banks with levels of capital close to the minimum requirement or below that minimum faced additional regulatory pressure, which we define using addons on the PCA thresholds. To take into account the subjectivity of this measure, we redefined the variable PCA considering both the actual PCA thresholds and an addon of 1 percentage point on those thresholds. In both cases and for the 3 scenarios (full sample, period before the financial crisis, period during the financial crisis), the findings remained similar to the In order to further test the robustness of our findings, we considered an alternative specification for capitalization. We used the tier 1 riskweighted capital ratio (tier 1 capital ratio) instead of the tier 1 leverage ratio (capitalization). The tier 1 capital ratio keeps the tier 1 capital in the numerator and considers the riskadjusted assets (RWA) instead of the total assets in the denominator. We measure tier 1 as the average of the tier 1 capital ratio throughout the reference period, with higher values for tier 1 reflecting a better position in terms of riskadjusted capital. Therefore, a positive relationship between tier 1 and dividend payout is expected. For the period during the financial crisis, we also considered another specification for regulatory pressure. Using the data relating to the participation of banks in the . Troubled Assets Relief Program (TARP),19 we replaced the variable PCA with a dummy (TARP) that assumes a value of unity if the bank received public funds from the As for PCA, we also considered the interaction of TARP with profitability, expecting a negative relationship between both TARP and profitability ? TARP, and dividend payout. We present the findings in Table 7. For the 3 scenarios considered (full sample, period before the financial crisis, period during the financial crisis), the findings are consistent with the previous results when we consider tier 1 capital ratio to capture the effect of regulatory monitoring, and also when we consider PCA and its interaction with profitability. Interestingly, the evidence for the regulatory pressure hypothesis is stronger when using the variable TARP (Table 7, column 7) relatively to the baseline. Both TARP and profitability ? TARP are negatively and significantly related to dividend payout. In the previous definition of regulatory pressure, only the interaction variable profitability ? PCA was significant. This finding indicates the robustness of the support for the regulatory pressure hypothesis during the financial 共 頁 ， 第 11 頁 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 裝 ┊ ┊ ┊ ┊ ┊ 訂 ┊ ┊ ┊ ┊ ┊ 線 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ crisis to differing specifications of regulatory pressure. Table 7 Robustness tests. This table presents the findings for the robustness tests on the definition of regulatory pressure considering the full sample, the period before the financial crisis and the period during the financial crisis. The tier 1 riskweight capital ratio (tier 1 capital ratio) is used instead of the tier 1 leverage ratio (capitalization). In column (7), regulatory pressure is considered through the dummy variable TARP, which assumes a value of unity if the bank has received public funds from the . Troubled Assets Relief Program (TARP). The variables are the same as those defined in Table 2. The heteroskedasticityconsistent standard errors are presented in brackets. 5. Conclusions Researchers in corporate finance have often found interest in studying financial firms due to their amplified agency and governance problems, and their critical importance for the good functioning of the modern financial system. The 2020– 09 financial crisis has further enhanced the interest as a result of the unique macroeconomic setting and the regulatory shifts that occurred during this period. We construct a new dataset on 462 . bank holding panies to study the dividend policy in the context of the 2020– 09 financial crisis. We test the signaling and agency hypotheses alongside the Fama and French (2020) characteristics of dividend 共 頁 ， 第 12 頁 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ 裝 ┊ ┊ ┊ ┊ ┊ 訂 ┊ ┊ ┊ ┊ ┊ 線 ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ ┊ payers controlling for the regulatory shifts during the period under analysis. Our main findings indicate that dividend payouts depend on the macroeconomic context (before and during the financial crisis). The Fama and French (2020) characteristics – larger, more profitable and low growth banks tend to pay more dividends – hold for both periods. Interestingly, despite the presence of external regulators in the financial industry, the findings still support the agency argument that dividends pensate for the need for monitoring. Our findings also support the signaling argument that dividends work as a signal of future growth opportunities, although only during the financial crisis. A possible interpretation is that bank holding panies have little need to signal unless the whole industry is under strain and it is important to be identified as a better than average bank. The controlling regulatory pressure hypothesis that undercapitalized banks plowback earnings to recapitalize themselves only holds during the financial crisis, a period during in which regulators exerted more pressure on banks with low capital buffers. Governance problems are often considered to be more severe and plex in financial firms. The 2020– 09 financial crisis further exposed the governance issues in financial firms creating the pathwa