NEW EVIDENCE ABOUT THE IMPACT OF TAXINGCORPORATE-SOURCE INCOME

Dividend Taxes and Firm Valuation: New Evidence

By ALAN J. AUERBACH AND KEVIN A. HASSETT*

The Jobs and Growth Tax Relief Act of 2003(JGTRA03) reduced the tax rates on dividends,with the highest statutory tax rate of 35 percentfalling to 15 percent. An interesting twist on thedividend tax cut was its temporary nature; theprovision as passed was effective only through2008, and (as recent congressional deliberationshave illustrated) the extension its supportersenvisioned was by no means certain. This largedividend tax reduction, along with its sunsetprovision, offer an unusual natural research ex-periment on the effects of dividend taxation.

The theory of dividend taxation suggeststhree possible scenarios for the effects of thedividend tax reduction. Under the tax irrele-vance view, the marginal shareholder is a tax-free entity (or a taxable investor who ignores orcan offset incremental taxes), and the dividendtax reduction has no effect on equity values orfirm behavior. Under the traditional view, themarginal source of equity finance is new shareissuance, and the tax reduction feeds through tothe firms user cost and stimulates extra capitalformation. Share values rise in the short run but,after full adjustment, the higher capital levelreduces the marginal revenue product of capitalenough to offset the dividend tax reduction,leaving equity prices the same. Under the new

view, the marginal source of finance is retainedearnings and the dividend tax cut is capitalizedinto the share price of the firm but has noinvestment effect. Understanding the economicconsequences of the dividend tax reduction re-quires knowledge of the empirical relevance ofthese competing views.

Our previous paper (Auerbach and Hassett,2005) performed an event-study analysis of alarge panel of firms to determine how firm at-tributes affected the valuation response overeight key event dates leading up to the 2003legislation. Our results, taken together, rejectedoutright the tax irrelevance view and generallyconformed to the predictions of the new viewfor firms that had paid dividends. We alsofound, however, that additional theory was re-quired, particularly with respect to the manyfirms that do not pay dividends and hence arenot covered directly by the standard theories.Among our findings were:

1. For mature firmsthose that had previouslypaid a dividenda 1-percentage-point in-crease in dividend yield led to a 0.5- to 1.5-percent abnormal return over the event dates.

2. Within this same set of firms, those morelikely to issue new shares also benefited ab-normally from the tax cut, with a 1-percentincrease in the probability of new share is-suance associated with a 0.2-percent in-crease in abnormal return.

3. Immature firmsthose that never had paida dividendsignificantly outperformed ma-ture firms on our event dates, by excessreturns from 3.7 percent to 8.6 percent.

To enhance our ability to distinguish amongdifferent potential explanations for these find-ings, we studied the pattern of abnormal returns

Discussants: Gustavo Grullon, Rice University; RogerGordon, University of CaliforniaSan Diego; Michael De-vereux, Warwick University.

* Auerbach: Department of Economics, University ofCaliforniaBerkeley, 549 Evans Hall #3880, Berkeley, CA94720-3880, and NBER (e-mail: auerbach@econ.berkeley.edu); Hassett: American Enterprise Institute, 1150 17th St.NW, Washington, DC 20036 (e-mail: khassett@aei.org). Wethank our discussant, Roger Gordon, for comments, andJoe Rosenberg and Trent Magruder for excellent researchassistance.

119

leading up to another key event, the 2004 pres-idential election, using the daily closing pricesfrom the Iowa Electronic Markets U.S. Presi-dential Winner Takes All Market as measures ofeach candidates election probability. We foundthat a higher probability of George W. Bushbeing reelected (which we associated with ahigher probability of the dividend tax cut beingextended past 2008) had the following effects:1

1. It reduced the positive valuation effect of thedividend yield.

2. It enhanced the positive valuation effect ofbeing likely to issue new shares.

3. It enhanced the bonus to being immature.

We argued that result 1 contradicted the pre-diction of the traditional view of dividend tax-ation and that the positive valuation of high-yield firms in result 1 comes through a largerreduction in the cost of capital. Under that view,a longer period of lower dividend taxes shouldhave enhanced that positive valuation effect. Bycontrast, results 1 and 1 are consistent with thenew view of dividend taxation, which sees thecurrent dividend yield as a matter of the timingof dividend payments. Under this view, result 1reflects a bonus to a firm paying a larger shareof its dividends during the low-tax period be-tween 2003 and 2008, and result 1 reflects thedecreased importance of dividend timing whenthe probability of an extension increases.

Results 2 and 3 reflect the fact that, for firmslikely to issue new shares (either by explicitprediction or by inference from their immatu-rity), the present value of dividends exceeds thevalue of existing equity. Intuitively, if a firmwas planning to sell new shares tomorrow, andthe dividend tax is reduced permanently today,then the firm benefits twice, first seeing thepresent value of dividends on its existing sharesrise (because future dividends on those shareswill face a lower rate of tax), and second, be-cause it can sell the new shares tomorrow for ahigher price (as those shares, too, will benefitfrom lower dividend taxes). Thus, a permanentdividend tax cut should provide a larger bonus

for such firms. The reinforcing effects of greaterpermanence in results 2 and 3 conform to thestory. While issuing shares necessarily takesfirms beyond the realm of the simple newviewin which retained earnings always serveas the marginal source of equity fundstheimportance of new share issue probabilities inpredicting excess returns also further reducessupport for the traditional view of dividend tax-ation, which treats new share issues as a regularactivity of all firms and hence is not useful indistinguishing among them.

Taken together, these results suggested thatthe bonus accruing to high-yield firms was notassociated with a decline in the cost of capital.Hence, except for immature firms and firmslikely to issue new shares, the legislation mayhave done more to increase share values andperhaps also to promote other activities, such asdividend payments (as found by Raj Chettyand Emmanuel Saez, 2005), than to increaseinvestment.

In this paper, we extend our analysis in twoways. First, we consider the impact of the 2004presidential election on yet another aspect offirm valuationthe prices of outstanding op-tionsto gain further insight into and confir-mation of the mechanism through which the2003 legislation affected firm values. Second,we provide further evidence for the 2003 eventdates, exploring in more detail the determinantsof the immaturity premium noted in results 3and 3. In particular, and in contrast to claims ina recent paper by Gene Amromin et al. (2005),we find that the premium is associated with thelikelihood of new share issuance, as inferred,but not demonstrated, in our original analysis.

I. Option Prices and the 2004 Election: Theory

Option prices reflect uncertainty. A presi-dential electionparticularly one in which thecandidates have quite different views about cap-ital income taxationeliminates an importantsource of uncertainty for firms, and thereforeshould affect option prices. Indeed, using theBlack-Scholes formula, we can estimate thevolatility (the standard deviation of the under-lying stocks rate of return) implied by an op-tion price and calculate the decline in thisvolatility around the 2004 election. Our primaryinterest here is how this decline in volatility

1 Including in these equations dummy variables forgroups of industries likely to favor George W. Bush or JohnKerry based on other elements of their respective platformshad little impact on the results reported here.

120 AEA PAPERS AND PROCEEDINGS MAY 2006

varied across firms and whether the change involatility agrees with our previous analysis.

As argued earlier, the reelection of GeorgeW. Bush in 2004 should have extended theexpected duration of the 2003 tax cut. The find-ings outlined above suggest that this shouldhave been most important to the firms for whichdividends, on average, lie farther in the future.Hence, the uncertainty surrounding the electionshould have increased volatility more for thesefirms during the pre-election period, and theresolution of uncertainty with the electionshould have been the most significant for themas well. Thus, we test whether the decline involatility follows the same pattern as the effects1 through 3, increasing with immaturity andthe likelihood of issuing new shares, and de-creasing with the firms dividend yield.

II. Option Methodology, Data, and Results

We start with the universe of firms from ourprevious paperall firms for which we canobtain stock-price data from CRSP and balance-sheet and income-statement data from Compu-stat from 2002 onward. We consider all calloptions for this initial sample of firms that hada strike price/current price ratio on November 1,2004 (the day before Election Day) between0.85 and 1.15 and expired in November 2004,December 2004, or January 2005. The first re-striction eliminates options far in or out of themoney that have limited trading volume; thesecond enhances our ability to discern changes

in implied volatility, which represents an aver-age expected volatility over the options re-maining life. This filtering procedure leaves5,333 options in our sample, of which 2,945 arefor mature firms (firms that had paid a dividendas of the end of 2002), and the balance are forimmature firms. We calculate the change inimplied volatility based on the closing prices ofeach option and its underlying stock on Novem-ber 1 and November 3.2

Table 1 presents the results of cross-sectionregressions with this change in implied volatil-ity as the dependent variable. The tables firstthree columns are based on unweighted regres-sions. The last three columns present the sameequations, weighted by the daily dollar volumeof option trades. In addition to coefficient esti-mates, the table also presents standard errors,clustered by 3-digit industry. The tables firstcolumn, which relates the change in volatilityfor the full sample of options to a constant anda dummy variable for mature firms, suggests adecline in volatility for both mature and imma-ture firms, but the decline for immature firms islarger by a statistically significant difference.Indeed, for the weighted version of this regres-sion, the entire decline in volatility is accountedfor by immature firms, a difference that again is

2 Options data, including calculations of implied volatil-ity, come from the IvyDB OptionMetrics dataset availablethrough Wharton Research Data Services, which covers allU.S. listed equity options. The option pricing model used tocalculate volatility incorporates dividend payments.

TABLE 1ELECTION-DAY CHANGES IN IMPLIED VOLATILITY(Standard errors in parentheses)

Sample

Full Mature Mature Full Mature Mature

Unweighted Dollar volume weighted

Constant 0.018 0.011 0.014 0.025 0.008 0.017(0.002) (0.001) (0.002) (0.008) (0.009) (0.006)

Dividend yield 0.014 0.045 0.449 0.035(0.059) (0.055) (0.345) (0.131)

Issue probability 0.007 0.041(0.007) (0.018)

Repurchase probability 0.017 0.050(0.012) (0.049)

Mature dummy 0.008 0.025(0.002) (0.011)

R2 0.006 0.0001 0.003 0.050 0.066 0.042N 5,333 2,945 2,643 5,333 2,945 2,643

121VOL. 96 NO. 2 THE IMPACT OF TAXING CORPORATE-SOURCE INCOME

statistically significant. Thus, our predictionthat immature firms should have experienced alarger decline in volatility is borne out.3

The second and third specifications consider theimpact of dividend yield and other covariates, andso are limited to the options of mature firms. Thesecond specification has only the dividend yield asa covariate, while the third also includes the prob-ability of a firm issuing shares or repurchasingthem based on a bivariate probit model describedin more detail in our earlier paper.4 In both ver-sions of these two specifications, the constant hasthe expected negative sign and is generally signif-icant; but the other two variables for which weoffered predictions above, the dividend yield (pre-dicted sign positive) and the new share probability(predicted sign negative), are insignificant in fiveout of six cases and of the wrong sign in half of thecases, with the only significant coefficient being ofthe wrong sign. Thus, there is no clear impact ofour covariates within the mature sample. But, asexpected, there is a clear relative decline in vola-tility for the immature firms that would, in theory,have been most affected by uncertainty concern-ing the dividend tax after 2008.

III. Additional Evidence on the ImmaturityPremium

In our earlier paper, we did not focus in detailon explaining the immaturity premium. Am-romin et al. (2005) challenge our interpretationof this premium, as well as those of other find-ings. They replicate a subset of our event studyresults, but go on to make arguments that theevidence is more consistent with the tax irrele-vance view. They base their alternative inter-pretation on three additional observations:

A. U.S. shares did not outperform those, suchas European stocks, that should have beenunaffected by the U.S. tax change.

B. The statistical significance of the results di-minishes if the event window is extendedthrough July 2003, several weeks past thetax reduction.

C. Immature firms generally had good stockmarket performance during this time, andthat performance appears unrelated to shareissuance.

We find this analysis unconvincing. First,without clarification the authors ignored mostof our event dates, making comparison ofresults difficult. Second, result A is drawnfrom an approach that has no power to dis-tinguish between its null and alternative hy-potheses. James Poterba (2004), for example,estimates that the dividend tax could haveboosted U.S. equity prices by 6 percent. Thischange would have been spread out over theentire period through which the probability ofa dividend tax change evolved from zero toone. The authors report standard error boundsfor their approach that are at times as wide as13 percent, suggesting that even if the fulleffect predicted by theory occurred duringtheir limited event windows, their approachwould have little chance of uncovering a sig-nificant effect. It is incorrect to conclude, asthe authors do, that there was no effect inthese circumstances. The correct conclusionis that their approach, by construction, couldfind no effect, and that event study analysisusing micro data is likely to be a much morefruitful research approach.

As to result B, it is common for statisticalsignificance to decline as event windowswiden and signal-to-noise ratios decline. Thisreflects the increased volatility of share pricesover longer time frames, and not the absenceof an effect, as the authors argued.

Finally, in arriving at result C, the authorsuse share repurchases as a proxy for the like-lihood of new share issuance, reasoning thatfirms that have recently repurchased are un-likely to issue new shares, and find that im-mature firms grouped by repurchasing activitydo not differ significantly in their immaturitypremium. While repurchasing may be an in-dicator of a lower probability of share issu-

3 To test whether this difference between mature firmsmight be due to other important differences related to vol-atility, notably firm size, we ran this regression separatelyfor groups of firms broken into four quartiles by firm size.Doing so had little impact on the coefficients but increasedstandard errors.

4 Covariates in the bivariate probit include two laggedvalues of investment, cash flow, market value, and debt, allscaled by firm assets, as well as the firms bond rating, thenumber of analysts following the stock, and categoricaldummies for firm size and industry. The sample size dropssomewhat with the addition of these covariates because ofmissing values for additional variables used in estimatingthe probit equation. The errors in the two branches of thebivariate probit have an estimated correlation of 0.968.

122 AEA PAPERS AND PROCEEDINGS MAY 2006

ance, this univariate approach is very crude.A superior approach would have been to pre-dict the probability of future share issuanceusing a model such as the one outlined in ourpaper, and to explore whether the immaturitypremium is related to the probability of issu-ing new shares in the future. We perform thistask in Table 2, using the sample and event studyapproach from our earlier paper, this time con-sidering the probability of issuing and repur-chasing shares for immature firms as well asmature firms.5 Consistent with our earlier inter-pretation, we see clear evidence that the imma-ture firms excess returns are highly correlatedwith the probability that they will issue newshares in the future. This result is significant inboth the weighted and unweighted runs. Alsoconsistent with our earlier findings for maturefirms, we find mixed evidence concerning sharerepurchases, suggesting that they alone are apoor proxy for the factors that determine themarginal source of finance and, hence, dividendtax effects.

IV. Conclusions

This paper reports additional evidence thatthe 2003 dividend tax reductions significantlyaffected equity markets. Together with ourearlier study, this evidence rejects the viewthat taxes are irrelevant. Among firms thatpay dividends and rarely issue shares, themain impact of the tax reduction appears tohave been to boost share prices and encourageother activities, such as dividend payment,rather than to reduce the cost of capital. Forfirms that have yet to pay dividends and thoselikely to issue new shares, the effect is morecomplex. Firms expecting to issue new sharesmay well have experienced some reduction intheir cost of finance and hence an investmentstimulus, consistent with the traditional viewof dividend taxation. In future work, we planto explore whether investment responded tothe lower dividend taxes in a manner sug-gested by our analysis of financial variables,that is, whether investment responses werestronger among firms more likely to haveexperienced a reduction in the cost of capital.

REFERENCES

Amromin, Gene; Harrison, Paul and Sharpe,Steve. How Did the 2003 Dividend Tax CutAffect Stock Prices? U.S. Federal ReserveBoard, Finance and Economics DiscussionSeries: No. 2005-61, 2005.

Auerbach, Alan J. and Hassett, Kevin A. The2003 Dividend Tax Cuts and the Value of theFirm: An Event Study. National Bureau ofEconomic Research, Inc., NBER WorkingPapers: No. 11449, 2005.

Chetty, Raj and Saez, Emmanuel. Dividend Taxesand Corporate Behavior: Evidence from the2003 Dividend Tax Cut. Quarterly Journal ofEconomics, 2005, 120(3), pp. 791833.

Poterba, James. Taxation and Corporate PayoutPolicy. American Economic Review, 2004,94(2), pp. 17175.

5 The results in Table 2 for mature firms are repeatedfrom our earlier paper.

TABLE 2CUMULATIVE ABNORMAL RETURNS: 2003EVENT DATES

(Standard errors in parentheses)

Sample

Mature Immature Mature Immature

unweighted weighted

Constant 0.026 0.058 0.086 0.086(0.016) (0.026) (0.028) (0.060)

Dividend yield 0.507 1.495(0.333) (0.602)

Issueprobability

0.187 0.223 0.160 0.235(0.042) (0.043) (0.070) (0.114)

Repurchaseprobability

0.121 0.386 0.087 0.143(0.066) (0.111) (0.119) (0.075)

R2 0.002 0.001 0.003 0.005N 512,073 889,587 512,073 889,587

123VOL. 96 NO. 2 THE IMPACT OF TAXING CORPORATE-SOURCE INCOME

This article has been cited by:

1. Mihir A Desai, Dhammika Dharmapala. 2011. Dividend Taxes and International Portfolio Choice. Reviewof Economics and Statistics 93:1, 266-284. [CrossRef]

2. Spencer Usrey, Edward Schnee, Gary Taylor. 2011. The Mutual FundAn Institutional Investor thatActs Like an Individual. Journal of the American Taxation Association 33:1, 1. [CrossRef]

3. Annette Alstadster, Erik Fjrli. 2009. Neutral taxation of shareholder income? Corporateresponses toan announced dividend tax. International Tax and Public Finance 16:4, 571-604. [CrossRef]

4. Gene Amromin, Paul Harrison, Steven Sharpe. 2008. How Did the 2003 Dividend Tax Cut Affect StockPrices?. Financial Management 37:4, 625-646. [CrossRef]

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

c.000282806777212495_29172.pdfNEW EVIDENCE ABOUT THE IMPACT OF TAXING CORPORATE- SOURCE INCOMEDividend Taxes and Firm Valuation: New EvidenceI. Option Prices and the 2004 Election: TheoryII. Option Methodology, Data, and ResultsIII. Additional Evidence on the Immaturity PremiumIV. ConclusionsREFERENCES