Dividend Tax Cuts and Regulated Firms Stock Prices

Dividend Tax Cuts and Regulated Firms Stock Prices Ross N. Dickens University of South Alabama Kenneth J. Hunsader University of South Alabama We examine the impact of the Jobs and Growth Tax Relief Reconciliation Act of 2003 on regulated firms stock prices. Using a sample consisting of utility and financial firms, we find companies with higher dividend payments have greater abnormal returns than those with lower/no dividend payouts. We find considerable support for the static tax-rate effect and general, but not consistent, support for agency theory s excess cash hypothesis. Result variations across regulated firms SIC codes provide support that researchers should control for or exclude regulated firms in empirical studies. Introduction The Jobs and Growth Tax Relief Reconciliation Act of 2003 (JAGTRRA) reduces the marginal tax rate for an individual s dividend income from 38.6 percent (at the highest level) to 15 percent (equal to the capital gains rate). This change provides a unique opportunity to investigate the impact on security prices and the relationship of any such changes to firms financial management practices. Gadarowski, Meric, Welsh, and Meric (GMWM) (2007) utilize the JAGTRRA event to examine dividend policy variables relationship to abnormal stock returns for firms in unregulated industries. Eliminating regulated firms is standard practice for most dividend policy studies (Fama and French, 2001). Such an omission, however, leaves a hole in financial knowledge that our paper seeks to address. The reason to omit regulated firms from dividend policy studies generally follows the explanation given by Gadarowski, Meric, Welsh, and Meric that they exclude utilities because their dividend policies may be affected by regulation, and financial firms because their financial ratios are not comparable to those of industrial firms (p. 93). While the omissions are understandable, they limit the chance to study market reactions for firms in industries where the majority of firms actually pay dividends. Specifically, only 22 percent of the unregulated firms in Gadarowski, Meric, Welsh, and Meric s study pay dividends compared to the 79 percent of the 53 1939-8123/08/1400/0053/$2.50 Copyright 2009 University of Nebraska Lincoln

54 Dickens and Hunsader regulated firms in our sample (details to follow). Given this difference in the percentage of firms paying dividends, we examine if stock reactions to JAGTRRA are similar for regulated firms as Gadarowski, Meric, Welsh, and Meric s findings for unregulated firms. In their work, Gadarowski, Meric, Welsh, and Meric (2007) begin by excluding regulated industries (specifically, utilities SIC code 4900-4949 and financial firms SIC code 6000-6999). Taking special note of regulated firms is empirically sound. Multiple studies, for instance, Hansen, Kumar, and Shome (1994) and Collins, Saxena, and Wansley (1996), find utilities to have a higher dividend payout ratio than other firms. Also, studies of financial firms (for example Black, Ketcham, and Schweitzer (1995) and Bessler and Nohel (1996)) find that while financial firms stock price reactions to changes in dividends are in line with those of unregulated firms, the reactions are generally amplified. As such, while eliminating regulated firms from an empirical study may not be a necessity, controlling for them should be. Motivation Our chief motivation is to investigate whether regulated firms stock reactions to JAGTRRA will be similar to Gadarowski, Meric, Welsh, and Meric s findings for unregulated firms in relation to dividend policy variables. Gadarowski, Meric, Welsh, and Meric (2007) find high-dividend paying stocks outperform low-dividend paying stocks for unregulated industries, but firms paying no dividends have positive abnormal returns relative to dividend paying stocks. Also, Gadarowski, Meric, Welsh, and Meric find firms with more cash, lower debt ratios, and lower Tobin s Q measures have higher abnormal returns. Our results for regulated firms are generally in agreement with Gadarowski, Meric, Welsh, and Meric s and reinforce Gadarowski, Meric, Welsh, and Meric s two main findings that higher-dividend paying firms have positive stock price reactions and non-dividend paying firms stock returns are greater. Our findings for financial ratio-related measures also generally agree with Gadarowski, Meric, Welsh, and Meric s, but not uniformly. The biggest area of disagreement is in regard to Tobin s Q variables which could be related to the fact that Tobin s Q values for the regulated firms in our sample are nominally different from the unregulated firms values in Gadarowski, Meric, Welsh, and Meric s study. Gadarowski, Meric, Welsh, and Meric focus on dividend taxation hypotheses and agency theory to generate potential explanatory factors of stock price changes. We follow, as nearly as possible, their methodology for our different sample. 1 We provide a brief discussion of the theory s implications below. 1 See Appendix A for a summary of theoretical expectations for variables along with results from Gadarowski, Meric, Welsh, and Meric (2007) and our paper to allow ready comparisons.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 55 Tax-based Issues Gadarowski, Meric, Welsh, and Meric present their methodology as testing two layers of dividend taxation hypotheses. 2 The first layer involves the marginal investor s reactions based on taxation motivation. The second layer considers how taxes could impact the marginal investor via the short-term trader hypothesis. The tax-clientele hypothesis holds that the marginal investor selects stocks based on the tax impact on dividend income. (See Elton and Gruber (1970) for the classic study.) An investor with a high dividend income marginal tax rate would choose a stock paying no or low dividends, all else equal. As such, firms that pay greater dividends would find a following with investors with low or no marginal tax rates. Evidence for this hypothesis is mixed; Elton and Gruber (1970) and Litzenberger and Ramaswamy (1979) find support, but Koski and Scruggs (1998) find none. Kalay (1982) discusses the short-term trader hypothesis. The basic idea is that investors facing high marginal tax rates, but low enough transaction costs, will sell an otherwise desirable dividend-paying stock just before the ex-dividend date and repurchase the stock after the ex-dividend date to avoid taxes on the dividends. Boyd and Jagannathan (1994) hypothesize that the relationship will more likely hold when the marginal tax rate on dividends is higher than the rate on capital gains. Both tax hypotheses would view JAGTRRA s reduction of the marginal tax rate on dividend income as favorable, leading to an expectation of a positive stock price reaction. This expected reaction should be higher for firms with greater dividend yields (which Gadarowski, Meric, Welsh, and Meric (2007) term the static tax-rate effect). Given that a higher percentage of regulated firms pay dividends (and, thus, have a positive dividend yield), we expect to find a positive stock price reaction to JAGTRRA for our sample and especially for utilities, given the findings of Hansen, Kumar, and Shome (1994) and Collins, Saxena, and Wansley (1996) that utilities have higher dividend payout ratios. Agency Theory Issues Along with the two layers of dividend taxation hypotheses, Gadarowski, Meric, Welsh, and Meric (2007) consider agency theory s relation to dividend policy. Basic tenets of agency theory (Miller and Rock, 1985) hold that dividend-paying firms should, all else equal, have better alignment of shareholder and management interests. Therefore, a reduction in taxes on dividend income should, all else equal, add pressure to firms to initiate or increase dividend payments. Firms that are already paying dividends, however, should have better shareholder/manager alignment. Thus, there would be a greater expectation of dividend initiations at firms currently paying no dividend than for increases at firms that already pay dividends. The above 2 As this study follows Gadarowski, Meric, Welsh, and Meric (2007), we readily acknowledge our use of their phrasing and terminology.

56 Dickens and Hunsader relationship is more uncertain for utilities. Hansen, Kumar, and Shome (1994) argue that utilities have greater dividend payouts as a method to decrease stockholder/regulator conflicts as opposed to stockholder/management conflicts. Thus, utilities that pay no dividend may have an even stronger reaction than non-dividendpaying unregulated firms. Financial firms also may differ from unregulated firms, given the multiple regulatory agencies overseeing banking firms (which form the majority of financial firms in our sample). 3 The regulatory agencies presence may decrease agency costs between stockholders and management (Filbeck and Mullineaux, 1993). As such, stock price reactions may differ for financial firms if dividends are a method to decrease stockholder/regulator conflicts, as with utilities. Another issue related to agency theory is the excess funds hypothesis. The implication is that investors should expect firms with greater excess funds to increase dividend payments more because holding excess funds can be a sign of several agency problems (such as underinvesting and managerial slack). Given that financial firms (especially the banks owned by bank holding companies) face added liquidity issues stemming from short-term funding sources, it is not clear that their reaction to JAGTRRA will follow the expected excess funds relationship. In general, the excess funds hypothesis holds that companies with excess funds will be in better position (or face more pressure from shareholders) to initiate or raise dividend payments. A company with greater excess funds should, all else equal, have higher cash on hand, greater free cash flow, lower relative debt level, and lower Tobin s Q ratio (the measures we use in accord with Gadarowski, Meric, Welsh, and Meric). 4 Sample and Methodology We draw our sample from U.S. domestic firms with available data on the Center for Research in Security Prices (CRSP) and Research Insight (Compustat) databases. We include only companies in the regulated industries with SIC codes between 4900-4949 (utilities) and 6000-6999 (financial firms) while excluding Real Estate Investment Trusts (REITs) with SIC code of 6798. We exclude REITs as the JAGTRRA omitted these companies from the lower dividend income tax rates. Thus, 3 We will use the term bank or banking firm for those firms in our sample with SIC codes of 6000-6199. Almost all these firms are bank holding companies and, as such, are under the regulation of the Federal Reserve. Added regulation for each holding company follows from the Federal Deposit Insurance Corporation (FDIC), the Office of the Comptroller of the Currency (OCC), and/or the Office of Thrift Supervision given the particular charter of the individual bank(s) owned by the holding company. 4 To control for variables potentially impacting banking firms stock reactions, we also incorporate variables suggested by Dickens, Casey, and Newman (2002). We present this discussion and results in Appendix B.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 57 we intentionally exclude all companies in Gadarowski, Meric, Welsh, and Meric (2007) and examine those firms not included in their study (except REITs). After we apply all our data screens, we have 941 firms with complete data of which 107 are utilities and 834 are financials (with 609 banking firms and 225 non-bank financial firms). To allow better comparisons, we utilize the same proposal (January 7, 2003) and passage (May 28, 2003) dates as Gadarowski, Meric, Welsh, and Meric. All firms must have daily stock returns available for the estimation period. Following Gadarowski, Meric, Welsh, and Meric, we estimate abnormal returns using the value-weighted CRSP index as the market proxy with the estimation period running 255 days starting 45 days prior to January 7, 2003 with the model estimation utilizing Scholes and Williams (1997) correction for nonsynchronous trading. We also utilize cumulative abnormal returns (CARs) over a four-day period and standardized CARs (SCARs) using the Mikkelson and Partch (1988a, b) correction for market model error and serial correlation in keeping with Gadarowski, Meric, Welsh, and Meric. The firms also need share price data for January 2, 2003 as well as the financial statement variables from Compustat for the most recent year prior to January 7, 2003. 5 MKTVAL is the market capitalization value of each firm s common equity as of January 2, 2003 measured in millions of dollars. For regressions, we use the natural log of MKTVAL noted as LMKTVAL, to control for the Size-in-January effect following Lie (2000) and Gadarowski, Meric, Welsh, and Meric (2007). DIVZERO is a binomial variable assigned the value of 1 if a firm paid no common dividend in the prior year s Compustat data, else the value is 0. The static tax-rate effect expects firms having higher dividend yields will have more favorable stock price reactions to JAGTRRA s event dates, but Gadarowski, Meric, Welsh, and Meric find unregulated firms paying no dividends had a more favorable reaction. DIVYIELD is the annual dividend from Compustat divided by the January 2, 2003 stock price. As stated above for DIVZERO, the static tax-rate effect expects a positive relationship between DIVYIELD and stock price reaction. Agency theory, however, leads one to expect firms with higher DIVYIELD to be pressured less to increase dividend payments and, therefore, have less favorable stock price reactions to the event dates (once one controls for those firms with no dividend payments). In keeping with Gadarowski, Meric, Welsh, and Meric, we use four proxy variables to consider a firm s excess funds position. CASH/TA is the dollar value of a firm s cash and short-term assets divided by the dollar value of its total assets. All else equal, the excess funds hypothesis holds that a firm with a higher percentage of readily available cash will face greater pressure from shareholders to initiate 5 While Gadarowski, Meric, Welsh, and Meric (2007) utilize quarterly or annual Compustat data, our inclusion of financial firms requires using only annual data given the need to calculate our free cash flow measure (discussed below).

58 Dickens and Hunsader (increase) dividend payments. Thus, a higher value for CASH/TA would be positively related to stock performance on the event dates. Results from Gadarowski, Meric, Welsh, and Meric for unregulated firms support this expectation. We will not be surprised if the relationship to CASH/TA for financials is not as strong as for utilities. Given short-term financing sources (such as checkable accounts for banks), most financial firms maintain relatively high levels of shortterm assets to manage the associated liquidity risks. Thus, the stock price reaction to holding excess funds as measured by CASH/TA may be muted for depository institutions and, therefore, the financial industry sample. FREE/TA is a firm s free cash flow measure divided by its total assets. While Gadarowski, Meric, Welsh, and Meric (2007) use Compustat s free cash flow measure, 6 that variable is not available for financial firms. Thus, we calculate free cash flows in the manner of Lehn and Paulson (1989) where: FREE = EBITDA - Interest Expense - Taxes - Preferred Dividends - Common Dividends. (1) For utilities, interest expense comes from Compustat s interest expense variable. For financials, we utilize Compustat s interest expense total financial services variable. The expectation based on the excess funds hypothesis is that firms with higher FREE/TA ratios will have more positive stock price reactions to the event dates; however, Gadarowski, Meric, Welsh, and Meric find a negative relationship for unregulated firms. DEBT/TA is the ratio of a firm s long-term debt to its total assets. The excess funds hypothesis holds that firms with low DEBT/TA ratios should have the ability to borrow more and increase dividend payouts if desired and Gadarowski, Meric, Welsh, and Meric s findings for unregulated firms support this expectation. The result for financials may differ. Bank holding companies regulatory capital requirements often allow such firms to use specific forms of long-term debt toward required capital positions. Firms with higher DEBT/TA ratios may be signaling a willingness to take on debt to increase dividend payments. If true, the relationship between DEBT/TA and stock returns around event dates would be positive for financial firms. The last variable motivated by agency theory is our Tobin s Q measure designated TOBIN Q: TOBINQ = (MKTVAL + DEBT + PREF)/TA (2) where MKTVAL, DEBT, and TA are as defined above and PREF is the book value of preferred stock. Following Gadarowski, Meric, Welsh, and Meric, we define 6 Compustat s Free Cash Flow = Operating Activities Net Cash Flow Cash Dividends Capital Expenditures. The operating activities net cash flow includes changes in operating assets and liabilities, but is not available for banks.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 59 LOWQ as equal to 1 if TOBINQ is less than 1.0 or equal to 0 if TOBINQ 1.0. Agency theory holds a low Tobin s Q value should be associated with higher agency costs from excess funds. Thus, LOWQ would be positively related to stock returns on event dates and is what Gadarowski, Meric, Welsh, and Meric find for unregulated firms. One additional note on the variables we employ. One reason for not using financial firms in most studies is that some financial firm ratios are not comparable to those of other industries. For the financial ratios of this study, we make only two minor concessions to the process described in Gadarowski, Meric, Welsh, and Meric (2007). First, we must rely on annual data not a mixture of annual and quarterly data. Second, we can not use Compustat s free cash flow measure, but must calculate one using a slightly different interest expense measure for financials than for utilities (or as one would for other unregulated firms). The ratios used in this study do not appear to have the comparability problems that would exist for many other ratios. If our findings are not similar to Gadarowski, Meric, Welsh, and Meric s for regulated firms, our use of slightly different variables is a potential reason for the difference. Table 1 contains descriptive statistics. To provide detail, we report values in three sub-groups: utilities (Panel A), banks (Panel B), and non-bank financials (Panel C). 7 We highlight that 79 percent (745 of 941) of the firms in our sample pay dividends (93 percent of utilities, 86 percent of banks, and 54 percent of non-bank financial firms). These percentages compare to the 22 percent of dividend-paying firms in Gadarowski, Meric, Welsh, and Meric s study. Perhaps regulated firms stock price reactions will differ from non-regulated firms simply because a higher percentage of regulated firms pay dividends. The mean MKTVAL for all regulated firms is $2,226 million, but with a median of $177 million. Similar differences between means and medians are present for almost all variables which lead us to winsorize the data at the first and 99th percentile levels for the regression tests that follow. 8 The dividend yield for forms paying a dividend (DIVYIELD>0) is 6 percent for utilities, 2 percent for banks and 3 percent non-bank financials. DIVZERO s mean is 0.07, 0.14, and 0.46 corresponding to the 93 percent, 86 percent, and 54 percent of the dividend-paying firms for utilities, banks, and non-bank financials, respectively. CASH/TA ranges from 3 percent for utilities (Panel A) to 19 percent for non-bank financials (Panel C) and TOBINQ ranges from 0.29 for banks (Panel B) to 2.47 for non-bank financials (Panel C). FREE/TA is -4 percent for non-bank financials (Panel C), -1 percent for banks (Panel B), but +3 percent for utilities (Panel A). DEBT/TA ranges from 13 percent for banks (Panel B) to 34 percent for utilities (Panel A). While many of the 7 Descriptive statistics for the combined sample are available upon request. 8 Gadarowski, Meric, Welsh, and Meric (2007) winsorize their data at the 2.5 percent and 97.5 percent levels.

60 Dickens and Hunsader Table 1 Descriptive Statistics of the Sample This table shows the descriptive statistics for the firm characteristics of all U.S. utility (SIC code 4900-4949), banking (SIC code 6000-6199), and non-bank financial (SIC 6200-6999) firms exclusive of REITs (SIC 6798) with valid data from CRSP and Compustat (941 firms with complete data and 745 firms with positive dividend yield). We compute all financial statement variables from annual Compustat data for the 2002 calendar year. MKTVAL is the market value of common equity in millions of dollars. DIVZERO is a dummy variable equal to one if the firm pays no ordinary dividend in 2002, and zero otherwise. DIVYIELD is the annual dividend per share divided by the closing stock price per CRSP as of January 2, 2003. We report those companies with dividend yields greater than 0 (DIVYIELD>0). CASH/TA is the ratio of cash and short-term investments to total assets. FREE/TA is the ratio of free cash flow calculated as (EBITDA - Interest Expense - Taxes - Preferred Dividends - Common Dividends) from 2002 annual data divided by total assets. DEBT/TA is the ratio of long-term debt to total assets. TOBINQ is the ratio of market value of common equity plus total long-term debt plus preferred stock to total assets. LOWQ is a dummy variable equal to one if TOBINQ is less than one, and zero otherwise Standard Deviation 90 th Percentile 10 th Percentile Minimum N Mean Maximum Median Panel A: Utilities (SIC 4900-4949) MKTVAL 107 2,519 3,720 20,380 6,495 1,178 102 21 DIVZERO 107 0.07 0.25 1.00 1.00 1.00 1.00 0.00 DIVYIELD>0 100 0.06 0.04 0.35 0.11 0.05 0.03 0.01 CASH/TA 107 0.03 0.09 0.96 0.06 0.01 0.00 0.00 FREE/TA 107 0.03 0.11 0.11 0.06 0.04 0.00-0.26 DEBT/TA 107 0.34 0.11 0.78 0.46 0.33 0.23 0.00 TOBINQ 107 0.79 0.24 1.67 1.10 0.79 0.47 0.11 LOWQ 107 0.82 0.38 1.00 1.00 1.00 0.00 0.00 Panel B: Banks (SIC 6000-6199) MKTVAL 609 1,931 10,676 183,814 1,908 115 22 2 DIVZERO 609 0.14 0.35 1.00 1.00 1.00 0.00 0.00 DIVYIELD>0 524 0.02 0.01 0.11 0.04 0.02 0.01 0.00 CASH/TA 609 0.07 0.06 0.63 0.12 0.05 0.02 0.00 FREE/TA 609-0.01 0.02 0.25 0.01-0.01-0.02-0.10 DEBT/TA 609 0.13 0.12 0.84 0.27 0.10 0.01 0.00 TOBINQ 609 0.29 0.19 3.39 0.42 0.26 0.15 0.07 LOWQ 609 0.99 0.10 1.00 1.00 1.00 1.00 0.00 Panel C: Non-Bank Financial Firms (SIC 6200-6999) MKTVAL 225 3,054 11,849 157,298 6,067 345 13 1 DIVZERO 225 0.46 0.50 1.00 1.00 1.00 0.00 0.00 DIVYIELD>0 121 0.03 0.05 0.49 0.08 0.02 0.00 0.00 CASH/TA 225 0.19 0.21 1.00 0.53 0.10 0.02 0.00 FREE/TA 225-0.04 0.66 0.79 0.12 0.00-0.04-8.52 DEBT/TA 225 0.28 2.44 36.67 0.40 0.05 0.00 0.00 TOBINQ 225 2.47 7.93 64.62 3.64 0.65 0.11 0.03 LOWQ 225 0.69 0.46 1.00 1.00 1.00 0.00 0.00 nominal values appear to be different from the nominal values for the unregulated firms in Gadarowski, Meric, Welsh, and Meric s sample, we find no statistically significant differences based on standard differences of means tests.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 61 Table 2 Univariate Correlations of the Sample This table shows the Pearson correlation coefficients for the firm characteristics of all U.S. utility (SIC code 4900-4949), banking (SIC code 6000-6199), and non-bank financial (SIC 6200-6999) firms exclusive of REITs (SIC 6798) with valid data from CRSP and Compustat (941 firms with complete data and 745 firms with positive dividend yield). We compute all financial statement variables from annual Compustat data for the 2002 calendar year. MKTVAL is the market value of common equity in millions of dollars. DIVZERO is a dummy variable equal to one if the firm pays no ordinary dividend in 2002, and zero otherwise. DIVYIELD is the annual dividend per share divided by the closing stock price per CRSP as of January 2, 2003. We utilize only those companies with dividend yields greater than 0 (DIVYIELD > 0) for correlations including DIVYIELD. CASH/TA is the ratio of cash and short-term investments to total assets. FREE/TA is the ratio of free cash flow calculated as (EBITDA - Interest Expense - Taxes - Preferred Dividends - Common Dividends) from 2002 annual data divided by total assets. DEBT/TA is the ratio of long-term debt to total assets. TOBINQ is the ratio of market value of common equity plus total long-term debt plus preferred stock to total assets. LOWQ is a dummy variable equal to one if TOBINQ is less than one, and zero otherwise. The p-values of the correlation coefficients appear in parentheses LMKTVAL DIVZERO DIVYIELD CASH/TA FREE/TA DEBT/TA TOBINQ LOWQ Panel A: Utilities LMKTVAL DIVZERO -0.004 (0.9635) DIVYIELD -0.095 (0.3288) CASH/TA -0.119 (0.2211) FREE/TA 0.160 (0.1002) DEBT/TA 0.128 (0.1886) TOBINQ -0.122 (0.2088) LOWQ 0.161 (0.0977) 0.449-0.091 (0.3501) -0.075 (0.4423) -0.203 (0.0361) 0.123 (0.2071) -0.043 (0.6600) -0.628 0.130 (0.1818) -0.195 (0.0444) 0.028 (0.7720) -0.151 (0.1196) -0.324 (0.0007) -0.052 (0.5972) 0.089 (0.3624) -0.166 (0.0881) 0.191 (0.0489) -0.059 (0.5459) 0.159 (0.1011) -0.020 (0.8410) -0.713 Panel B: Banks LMKTVAL DIVZERO -0.198 DIVYIELD -0.037 (0.3961) CASH/TA -0.293 (0.4702) 0.214-0.026 (0.5552) FREE/TA 0.240 0.216 (0.0001) -0.184 0.182 DEBT/TA 0.018 (0.6499) 0.071 (0.0813) 0.047 (0.2809) -0.079 (0.0521) -0.154 (0.0001) TOBINQ 0.138 (0.0006) 0.121 (0.0028) 0.083 (0.0590) 0.057 (0.1586) 0.256 0.633 LOWQ 0.029 (0.4695) -0.152 (0.0002) -0.239 0.023 (0.5759) -0.465-0.119 (0.0036) -0.634

62 Dickens and Hunsader Table 2 (cont.) Univariate Correlations of the Sample LMKTVAL DIVZERO DIVYIELD CASH/TA FREE/TA DEBT/TA TOBINQ LOWQ Panel C: Non-bank Financials LMKTVAL DIVZERO -0.485 DIVYIELD -0.368 CASH/TA -0.193 (0.0036) FREE/TA 0.223 (0.0007) DEBT/TA -0.153 (0.0214) TOBINQ -0.090 (0.1804) LOWQ 0.042 (0.5278) 0.278-0.099 (0.1391) 0.081 (0.2285) -0.105 (0.1168) -0.021 (0.7496) 0.041 (0.6547) -0.561-0.093 (0.3080) 0.201 (0.0274) -0.191 (0.0358) -0.069 (0.3018) 0.097 (0.1458) 0.373-0.231 (0.0005) -0.467-0.214 (0.0012) 0.099 (0.1404) 0.303-0.102 (0.1255) 0.338 Utilities have a higher nominal DIVYIELD, less cash on hand, a greater free cash flow, and a greater debt load. The agency theory/excess funds position behind the preceding variables leads to the expectation that utility stocks should have a worse reaction to JAGTRRA than financials based on all but the free cash flow variable. The static tax-rate effect leads to the expectation that utility firms will have a better reaction given the higher DIVYIELD. Table 2 reports the univariate correlations for the variables in our models. Panel A contains the values for utilities, while Panel B and Panel C report results for banks and non-bank financial firms, respectively. Again, for space reasons, we report the segment components and do not include the combined sample results. 9 Visual evidence would seem to indicate differences between the three groups. There are 12 (12) of the 27 possible combinations where a correlation s significance level in the utility subset, differs from that in the bank (non-bank financial) subset. By differs we mean that the correlation is significant in one subset, but not in the other. There is (are) only one (three) case(s) where a variable is significant in both the utility and bank (non-bank financial) subsets and of opposite signs. TOBINQ is a factor in all such cases. Comparing banks and non-bank financials, we find 14 instances with 9 If one compares the results for our whole sample of regulated firms with the equivalents in Gadarowski, Meric, Welsh, and Meric, one would note 25 of the 27 correlation values in Gadarowski, Meric, Welsh, and Meric are significant at the 0.05 level or better (and 24 of them at better than the 0.0001 level). Our sample has 21 of the same 27 relationship values significant at the 0.05 level or better, but only 12 are significant at the 0.0001 level or better.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 63 differing significance for given variables and two cases of opposite signs. 10 Both cases involve the TOBINQ variable. Finally, we note that when regulated firms correlation values have the opposite sign from Gadarowski, Meric, Welsh, and Meric s non-regulated firms results, all such cases are related to Tobin s Q or DEBT/TA. Taken together, it appears regulatory firms could react differently to JAGTRRA based on financial fundamentals. Table 3 Univariate Analysis of Event Returns This table shows the univariate analysis of the cumulative abnormal returns for the proposal (CAR[PROP]) and the passage of (CAR[PASS]) JAGTRRA. In each panel (A, B, and C), we use correlation analysis in part 1 and the average returns of the sorted groups in part 2. We compute all financial statement variables from annual Compustat data for the 2002 calendar year. DIVYIELD is the annual dividend per share divided by the closing stock price per CRSP as of January 2, 2003. CASH/TA is the ratio of cash and short-term investments to total assets. FREE/TA is the ratio of free cash flow calculated as (EBITDA - Interest Expense - Taxes - Preferred Dividends - Common Dividends) from 2002 annual data divided by total assets. DEBT/TA is the ratio of long-term debt to total assets. LOWQ is a dummy variable equal to one if TOBINQ is less than one, and zero otherwise where TOBINQ is the ratio of market value of common equity plus total long-term debt plus preferred stock to total assets. We perform our analysis using for the full sample for each group after winsorizing the CARs at the 1 percent and 99 percent levels except with DIVYIELD. We use only firms with DIVYIELD greater than zero. To construct the groups, we sort the firms into three equal-sized groups using the firm characteristics with the entire sample except as follows: for dummy variable LOWQ, the high (low) portfolio contains firms when the dummy variable is equal to one (zero). The table also shows the differences in the means between the high and low group averages (high-low) along with the p-values (part 1 of each panel) and the t-statistics (part 2 of each panel) for the two-tailed tests DIVYIELD CASH/TA FREE/TA DEBT/TA LOWQ Panel A: Utilities (N = 104) 1: Univariate Correlations CAR[PROP] 0.468 0.159 (0.1090) -0.168 (0.0890) 0.088 (0.3759) 0.169 (0.0881) CAR[SIGN] 0.290 (<0.0041) -0.110 (0.2642) -0.011 (0.9110) 0.074 (0.4573) 0.188 (0.0555) 2: Group Comparisons CAR[PROP] High 0.044 0.029 0.019 0.023 0.024 Low 0.007 0.012 0.034 0.022 0.008 High-Low 1 0.037 0.017-0.017 0.002 0.016 t-statistic 4.29*** 1.87* -1.75* 0.18 2.62** CAR[SIGN] High 0.037 0.025 0.022 0.023 0.026 Low 0.009 0.030 0.019 0.017 0.009 High-Low 1 0.028-0.005 0.003 0.005 0.017 t-statistic 3.33*** -0.62 0.34 0.71 1.97* 10 These differences show the necessity for separating banking and non-banking financial firms. We thank the editor for leading us to this finding based on his observations using descriptive statistics in an earlier version of this paper.

64 Dickens and Hunsader Empirical Results from Event Returns Table 3 reports our analysis of the stock returns on the January 7, 2003 proposal (PROP) and May 28, 2003 presidential signing (SIGN) dates for utilities (Panel A), banks (Panel B), and non-bank financials (Panel C). As with Gadarowski, Meric, Welsh, and Meric s non-regulated firms, the univariate correlation between DIVYIELD and the cumulative abnormal returns (CARs) around both event dates Table 3 (cont.) Univariate Analysis of Event Returns DIVYIELD CASH/TA FREE/TA DEBT/TA LOWQ Panel B: Banks (N = 597) 1: Univariate Correlations CAR[PROP] 0.064 (0.1460) 0.059 (0.1536) -0.064 (0.1163) 0.086 (0.0365) -0.054 (0.1885) CAR[SIGN] 0.129 (0.0034) -0.094 (0.0216) 0.013 (0.7527) 0.001 (0.9860) -0.029 (0.4830) 2: Group Comparisons CAR[PROP] High 0.001 0.003 0.001 0.003 0.001 Low -0.004-0.001 0.004-0.002 0.010 High-Low 1 0.005 0.004-0.004 0.005-0.009 t-statistic 1.76* 1.48-1.30 1.71* 0.73 CAR[SIGN] High 0.013 0.009 0.011 0.009 0.009 Low 0.006 0.011 0.009 0.008 0.025 High-Low 1 0.008-0.003 0.001 0.001 0.015 t-statistic 2.40** -0.78 0.43 0.33 1.01 Panel C: Non-Bank Financials (N = 220) 1: Univariate Correlations CAR[PROP] 0.301 (0.0008) 0.069 (0.3024) -0.136 (0.0436) -0.086 (0.2030) -0.114 (0.0901) CAR[SIGN] 0.023 (0.8003) 0.084 (0.2168) -0.006 (0.9241) -0.131 (0.0528) 0.054 (0.4238) 2: Group Comparisons CAR[PROP] High 0.020 0.014 0.003 0.003 0.006 Low -0.002 0.005 0.021 0.014 0.012 High-Low 1 0.022 0.009-0.018-0.011 0.006 t-statistic 3.12** 1.30-2.35** -1.46 0.79 CAR[SIGN] High 0.019 0.032 0.024 0.016 0.022 Low 0.016 0.013 0.026 0.033 0.020 High-Low 1 0.002 0.019-0.002-0.017 0.002 t-statistic 0.28 2.17** -0.25-2.05** 0.28 1 Rounding may lead to high-low differences seemingly at odds with the corresponding high and low values *** Significant at the 0.01 level ** Significant at the 0.05 level * Significant at the 0.10 level

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 65 are positive for the regulated firms in keeping with the static tax-rate effect. 11 The utility subset s correlations (Panel A) are significant on both dates while banks (Panel B) and non-bank financials (Panel C) are significant on the signing and proposal dates, respectively. Regulated firms correlations between CARs and the various excess funds hypothesis-driven variables show little support for the excess funds hypothesis. LOWQ is positively related to CAR[PROP] and CAR[SIGN] for utilities (Panel A) and DEBT/TA is negatively related to CAR[SIGN] for non-bank financials (Panel C) in keeping with the hypothesis expectations. CASH/TA is negatively related to CAR[SIGN] for the bank subset (Panel B), FREE/TA is negatively related to CAR[PROP] for utilities (Panel B), DEBT/TA is positively related to CAR[PROP] for banks (Panel B), and LOWQ is negatively related to CAR[PROP] for non-bank financials (Panel C). These results are contrary to the excess funds hypothesis. 12 Following Gadarowski, Meric, Welsh, and Meric (2007), we divide the samples into high and low groups (based on thirds) and compare the results for those observations in the high third with those in the low third. In every case where statistically significant, the high third has a higher stock market reaction for DIVYIELD. Thus, not only is DIVYIELD positively related to CARs, but firms in the highest third DIVYIELD group have significantly greater stock returns associated with JAGTRRA than firms in the lowest third for every case except non-bank financials on the signing date. We note that utilities have the highest significance level (Panel A) which is in keeping with Hansen, Kumar, and Shome (1994) and Collins, Saxena, and Wansley (1996). These results continue to support the static tax-rate effect that firms with higher dividend yields will have greater positive stock price reactions. Utilities high-low comparison for CASH/TA is positive for CAR[PROP] which is the same finding as Gadarowski, Meric, Welsh, and Meric s for non-regulated firms and non-bank financials comparison for CASH/TA is significant for CAR[SIGN]. These results are in keeping with the excess funds hypothesis. Utilities and non-banks high-low comparisons for FREE/TA for CAR[PROP] are negative in keeping with Gadarowski, Meric, Welsh, and Meric s finding for unregulated firms, but contrary to excess funds hypothesis expectations. The high-low comparison for DEBT/TA is negative for non-banks (CAR[SIGN]) in agreement with Gadarowski, Meric, Welsh, and Meric s findings for unregulated firms and in agreement with excess funds hypothesis expectations. Banks, however, show a positive result for CAR[PROP]. Finally, the high-low comparisons of LOWQ for both event dates are positive for utilities as in Gadarowski, Meric, Welsh, and Meric for unregulated firms, in keeping with agency costs and excess funds hypotheses. 11 Appendix A provides a ready comparison of Gadarowski, Meric, Welsh, and Meric (2007) results and ours as related to theoretical expectations. 12 When we combine all regulated firms together, the support for the excess funds hypothesis is stronger, but still mixed. (Full results available upon request.)

66 Dickens and Hunsader Table 4 Regression Analysis of Standardized Abnormal Returns This table shows the regression results from the following model for Model 1 through Model 5 which follow from GMWM (2007): SCAR[event]i = c0 + c1lmktvali + c2divzeroi + c3divyieldi + c4cash/tai (or FREE/TAi or DEBT/TAi) + c5lowqi +c6cash/tai*lowqi (or LMKTVALi*DIVZEROi, or CASH/TAi*DIVZEROi, or LOWQi*DIVZEROi)+ γi where event is either PROP or PASS. SCAR[PROP] is the standardized cumulative abnormal return (CAR) for the four-day return period associated with the proposal: January 3-January 8, 2003. SCAR[PASS] is the standardized CAR for the four-day return period associated with the passage of JAGTRRA: May 22- May 28, 2003. We compute all financial statement variables from annual Compustat data for the 2002 calendar year. LMKTVAL is the natural log of the market value of common equity in millions of dollars. DIVZERO is a dummy variable equal to one if the firm pays no ordinary dividend in 2002, and zero otherwise. DIVYIELD is the annual dividend per share divided by the closing stock price per CRSP as of January 2, 2003. CASH/TA is the ratio of cash and short-term investments to total assets. FREE/TA is the ratio of free cash flow calculated as (EBITDA - Interest Expense - Taxes - Preferred Dividends - Common Dividends) from 2002 annual data divided by total assets. DEBT/TA is the ratio of long-term debt to total assets. LOWQ is a dummy variable equal to one if TOBINQ is less than one, and zero otherwise where TOBINQ is the ratio of market value of common equity plus total long-term debt plus preferred stock to total assets. We perform our analysis on the full sample of each group after winsorizing the explanatory variables at the 1 percent and 99 percent levels. We test each model for heteroskedasticity following White (1980) and adjust significance tests as needed. (Parentheses report t-statistics.) *** Significant at the 0.01 level. ** Significant at the 0.05 level. * Significant at the 0.10 level Panel A: (N = 100) Utilities Proporal of JAGTRRA Signing of JAGTRRA Regressor Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5 Intercept -0.6386-0.5044-0.8777-0.5149-0.5683 1.1455 0.7482 0.4910 1.4436 1.1445 (-1.58) (-1.20) (-2.03)** (-1.10) (-1.41) (2.48)** (1.49) (0.97) (2.70)** (2.44)** LMKTVAL 0.0589 0.0701 0.0590 0.0507-0.3469-0.1435-0.1388-0.1434-0.1632 0.2943 (1.16) (1.38) (1.17) (0.95) (-0.88) (-2.48)** (-2.29)** (-2.43)** (-2.70)** (0.64) DIVZERO 0.3296 0.0901 0.2822 0.3294 2.0514-0.1391-0.5065-0.6442-0.1395-2.5531 (0.92) (0.27) (0.91) (0.92) (0.81) (-0.34) (-1.28) (-1.79)* (-0.34) (-0.86) DIVYIELD 5.2531 4.2769 5.7791 5.5642 5.7200-0.6539 1.3867-0.4917 0.0958-0.8739 (2.42)** (1.65) (2.45)** (2.46)** (2.67)*** (-0.26) (0.45) (-0.18) (0.04) (-0.35) CASH/TA 0.0970-6.0629 16.6765-5.1907-20.034-16.157 (0.03) (-0.50) (2.06)** (-1.45) (-1.44) (-1.71)* FREE/TA -4.3579 2.9385 (-1.43) (0.81) DEBT/TA 0.5406 1.4157 (0.74) (1.65)

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 67 Table 4(cont.) Regression Analysis of Standardized Abnormal Returns Panel A: (N = 100) Utilities Proporal of JAGTRRA Signing of JAGTRRA Regressor Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5 LOWQ 0.4315 0.4398 0.4673-0.3365 0.4726 0.7151 0.7619 0.7914 0.4861 0.6990 (2.23)** (2.35)** (2.47)** (-1.26) (2.48)** (3.25)*** (3.41)*** (3.59)*** (1.61) (3.15)*** CASH/TA* LOWQ 6.6207 15.9536 (0.52) (1.11) LMKTVAL* DIVZERO 0.4005-0.4398 (1.01) (-0.95) CASH/TA* DIVZERO -20.876 12.8243 (-2.35)** (1.24) Adj. R 2 0.0760 0.1104 0.0977 0.0688 0.1132 0.1099 F-value 2.63** 3.46*** 3.14** 2.22** 2.80** 3.44*** *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level 0.1101 0.1299 0.1120 0.1058 3.45*** 3.96*** 3.08** 2.67**

68 Dickens and Hunsader Table 4(cont.) Regression Analysis of Standardized Abnormal Returns Panel B: (N = 597) Banks Proporal of JAGTRRA Signing of JAGTRRA Regressor Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5 Intercept 0.5558 0.4231 0.5006 1.1569 0.3467 0.2608 1.0472 0.2511 0.1830 1.0612 (1.20) (0.48) (1.08) (1.73)* (0.40) (0.52) (1.08) (0.50) (0.25) (1.12) LMKTVAL -0.0702-0.0670-0.0750-0.0693-0.1056 0.0488 0.0493 0.0422 0.0487 0.1446 (-3.23)*** (-2.88)*** (-3.54)*** (-3.18)*** (-1.33) (2.06)** (1.91)* (1.82)* (2.06)** (1.68)* DIVZERO 0.4314 0.4120 0.4403 0.4284 0.8979-0.0038 0.0335 0.0203-0.0034-1.3281 (3.15)*** (3.01)*** (3.28)*** (3.13)*** (0.83) (-0.03) (0.22) (0.14) (-0.02) (-1.13) DIVYIELD 3.3519 3.1898 3.3517 3.3348 3.4555 6.4279 6.2553 6.9465 6.4301 6.1117 (0.97) (0.92) (0.99) (0.97) (1.00) (1.72)* (1.62) (1.87)* (1.72)* (1.63) CASH/TA 0.2586-5.6841-0.0520-0.8825-0.1134-2.7043 (0.30) (-1.18) (-0.03) (-0.95) (-0.02) (-1.50) FREE/TA -0.7606-1.0429 (-0.25) (-0.31) DEBT/TA 0.3485-0.2782 (1.00) (-0.73) LOWQ -0.3290-0.1992-0.2793-0.9459-0.4063-0.3628-1.2070-0.3536-0.2830-0.1028 (-0.76) (-0.23) (-0.65) (-1.45) (-0.81) (-0.78) (-1.26) (-0.75) (-0.40) (-0.19) CASH/TA* LOWQ 6.1340-0.7938 (1.25) (-0.15) LMKTVAL* DIVZERO 0.0384-0.1009 (0.47) (-1.13) CASH/TA* DIVZERO 0.4461 2.4173 (0.23) (1.15) LOWQ* DIVZERO 0.2607-1.0636 (0.26) (-0.99) Adj. R2 0.0390 0.0331 0.0447 0.0399 0.0345 0.0128 0.0096 0.0084 0.0112 0.0140 F-value 5.67*** 4.95*** 6.46*** 4.99*** 3.57*** 2.50** 2.12* 1.98* 2.08* 2.02** *** Significant at the 0.01 level; ** Significant at the 0.05 level; * Significant at the 0.10 level

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 69 Table 4(cont.) Regression Analysis of Standardized Abnormal Returns Panel C: (N = 213) Non-Bank Financials Proporal of JAGTRRA Signing of JAGTRRA Regressor Model 1 Model 2 Model 3 Model 4 Model 5 Model 1 Model 2 Model 3 Model 4 Model 5 Intercept 0.4159 0.3924 0.4612 0.4923 0.0199-0.0787-0.0414 0.0523 0.1092-0.0983 (1.45) (1.79)* (1.65)* (1.61) (0.05) (-0.26) (-0.14) (0.18) (0.35) (-0.24) LMKTVAL -0.0566-0.0465-0.0498-0.0569-0.1020 0.0475 0.0485 0.0440 0.0468 0.0556 (-1.87)* (-1.91)* (-1.65)* (-1.88)* (-2.50)** (1.51) (1.47) (1.42) (1.49) (1.30) DIVZERO -0.1834-0.1048-0.1077-0.1901 0.4383-0.0105 0.0214 0.0305-0.0269 0.1092 (-1.12) (-0.70) (-0.67) (-1.16) (0.90) (-0.06) (0.13) (0.18) (-0.16) (0.21) DIVYIELD 5.5328 6.2370 5.9736 5.4288 6.5225 1.6670 2.2114 1.9549 1.4110 1.6973 (2.03)** (2.96)** (2.23)** (1.98)** (2.26)** (0.59) (0.77) (0.71) (0.50) (0.56) CASH/TA 0.4248 0.1318 0.1888 0.3083-0.4129 0.0812 (1.36) (0.26) (0.52) (0.95) (-0.80) (0.21) FREE/TA -0.1404 0.0575 (-0.27) (0.09) DEBT/TA -0.2006-0.5205 (-0.57) (-1.43) LOWQ -0.0795-0.0900-0.1149-0.1760-0.0140 0.1869 0.1205 0.1096-0.1009 0.0191 (-0.62) (-0.67) (-0.92) (-0.96) (-0.08) (1.02) (0.87) (0.85) (-0.53) (0.10) CASH/TA* LOWQ 0.4593 1.1306 (0.75) (1.77)* LMKTVAL* DIVZERO 0.1025-0.0355 (1.67)* (-0.55) CASH/TA* DIVZERO 0.6453 0.9288 (0.91) (1.25) LOWQ* DIVZERO -0.1622 0.2834 (-0.63) (1.05) Adj. R2 0.0525 0.0526 0.0510 0.0505 0.0598-0.0013 F-value 3.35*** 3.34*** 3.31*** 2.88** 2.68*** 0.95 *** Significant at the 0.01 level; ** Significant at the 0.05 level; *Significant at the 0.10 level -0.0030 0.0052 0.0090-0.0036 0.88 1.22 1.32 0.90

70 Dickens and Hunsader Table 4 reports the results from the regressions analyzing the standardized CARs (per Mikkelson and Partch (1988a, b)) for the four days around both the proposal and signing dates. We test for robustness to heteroskedasticity following White (1980). Again, we divide the table into panels for utilities (Panel A), banks (Panel B), and non-bank financials (Panel C). 13 We estimate the same five model variations (Model 1 through Model 5) as in Gadarowski, Meric, Welsh, and Meric (2007). We discuss the results from the proposal date first. All models are significant for the proposal period for all three data sets. The models for the utilities set have nominally higher adjusted R 2 s as the models explain between 6.88 percent and 11.32 percent of the variation in SCARs around the proposal date as compared to 3.31 percent to 4.47 percent for banks and 5.05 percent to 5.98 percent for non-bank financials. Panel B shows DIVZERO is positively related to SCAR in four of the five models for banks as the excess funds hypothesis predicts (and in general agreement with Gadarowski, Meric, Welsh, and Meric s findings for unregulated firms). DIVYIELD is positively related to SCAR in almost all models for utilities (Panel A) and in all models for non-bank financials (Panel B). This finding is in keeping with the static tax-rate effect. The LOWQ coefficient is positive as the excess funds hypothesis expects for four of the five models for utilities (Panel A). LMKTVAL is negatively related to SCARs for the proposal period for banks (Panel B) and non-bank financials (Panel C) meaning larger market value firms have lower abnormal returns. This finding is in line with Gadarowski, Meric, Welsh, and Meric s for unregulated firms at the time of the proposal and in accordance with the size-in-january effect. Model 5, which includes interactive terms, finds LMKTVAL*DIVZERO to be positively related to SCARs for non-bank financials (Panel C). Thus, a firm that has a greater market value and paid no dividend had a greater SCAR at the proposal date. This finding is in agreement with Gadarowski, Meric, Welsh, and Meric s finding for unregulated firms at the signing date. CASH/TA*DIVZERO has a negative relationship for utilities (Panel A). Thus, utilities that did not pay a dividend and had higher cash balance experienced lower stock price reactions to the JAGTRRA proposal. This result is contrary to expectations. 14 The results for analysis of the SCARs around the signing date show all six models significant for utilities (Panel A) and banks (Panel B), but with no models being significant for non-bank financials (Panel C). DIVYIELD is no longer significant in any estimated model for utilities, but is positive in three regressions in the banking subset (Panel B). LMKTVAL is positive in all models for banks (Panel B). That variable, however, is negative in four utility firm models. Thus, utilities repeat the pattern from the proposal date that firms with higher market values react worse to the 13 Results using the entire regulated firms sample available upon request. 14 We do not include LOWQ*DIVZERO for utilities given a linear combination report from the statistical package for this subset.

Quarterly Journal of Finance and Accounting, Vol. 48, No. 4 71 JAGTRRA signing, while banks and non-bank financials with greater market values have a more positive reaction. Of course, given that the size-in-january effect should not be present in May, there is no longer an expectation that smaller firms would have higher returns and the results indicate the different reactions of the subsets. Utilities continue to show a positive relationship between LOWQ and SCARs in keeping with excess funds and agency costs hypotheses. CASH/TA*LOWQ is positive for non-bank financials (Panel C) which also supports the excess funds theory as investors seem to expect firms with low growth opportunities and higher relative cash amounts to pay higher dividends in the future. In keeping with the tests reported in Gadarowski, Meric, Welsh, and Meric (2007), we re-estimate the models reported in Table 4 using CARs instead of SCARs. Given space limitations we do not report these results here, but they are available upon request. In no case is there a coefficient that was significant in Table 4 that is significant with the opposite sign in the corresponding model in Table 5. For the most part, the only differences relate to coefficients that are marginally significant in one model but no longer significant in the corresponding re-estimated version (or vice versa). The overall results are generally consistent with Gadarowski, Meric, Welsh, and Meric s (2007) for regulated firms and provide strong evidence to back the static tax rate effect while providing mixed evidence as to excess cash flow hypothesis ideas. Summary and Concluding Remarks This study closely follows the footsteps of Gadarowski, Meric, Welsh, and Meric (2007) who investigate unregulated firms stock reactions to the JAGTRRA in 2003. Following that study s methodology as possible, we examine regulated firms stock price reactions to the same events. The regulated firms are utility or financial businesses which are routinely eliminated from corporate finance/dividend policy papers given the influence of regulators on dividend policy as well as problems with financial ratio comparability. Two factors lead us to believe that omitting regulated firms leaves a hole in our financial knowledge. First, 79 percent of the regulated firms in the current sample pay dividends as compared to only 22 percent of the firms in Gadarowski, Meric, Welsh, and Meric s sample. Second, empirical evidence indicates that utilities have higher dividend payout ratios than unregulated firms for those firms that pay dividends (Hansen, Kumar, and Shome, 1994; Collins, Saxena, and Wansley, 1996) and financial firms generally have greater stock price reactions to changes in dividends (Black, Ketcham, and Schweitzer, 1995; Bessler and Nohel, 1996). Thus, it seems worthwhile to examine the stock price reactions for industries where most firms pay dividends to gauge the reaction to a law that is changing taxation on dividend income. The mean and median market value (MKTVAL) and dividend yield for stocks paying a dividend (DIVYIELD > 0) are similar between the current sample of regu-