JEL Code: H25, G18 Key words: Australian corporate tax, franking credits, effective corporate tax rate

Transcription

1 Are franking creds valuable to Australian shareholders? Richard Heaney School of Economics, Finance and Marketing RMIT Universy Changes 1. interaction wh fcb put back into the equation 2. exclude the non franking balance firms 3. taxable income use to proxy benefs 4. focus on top 200 wh franking balance 5. deal wh issue where pay out all dividends fcb=0 but share includes all of the fcb Abstract The growing level of undistributed franking creds held by Australian companies provides an important puzzle given the theoretical value of dividend imputation creds to resident Australian shareholders. If retained franking creds are valuable to investors then some care needs to be taken in the way that firms and the government treat these growing balances. If they are not valuable to shareholders then this finding has implications for equy market risk premium estimation, valuation of firms and determination of appropriate rates of return for regulated firms. Analysis based on 8209 firm year observations, spanning the period 2001 to 2006, provides some support for the value of franking creds to shareholders. Further, analysis based on the characteristics of firms that hold franking cred balances does not support the argument that non resident shareholder preference for dividends explains the growth in franking cred balances. JEL Code: H25, G18 Key words: Australian corporate tax, franking creds, effective corporate tax rate Contact Richard Heaney, School of Economics and Finance, Business, RMIT Universy, level 12, 239 Bourke Street, Melbourne, 3000, Phone: , Fax: , Acknowledgements I acknowledge the Australian Research Council for funding this project via ARC Discovery Grant, DP I acknowledge the support of Sinclair Davidson and detailed comments and suggestions provided by participants at the Finance Research Day held at the Universy of Newcastle in August My thanks go to the research assistants who have worked on the project, Solomon Yaw Adomako, John Fowler, Peter Le, Muktah Noah and Binesh Seetanah for their diligence in entry and verification of the data sets used in this paper. 1 30/09/2009

2 1. Introduction The introduction of the imputation tax system in Australia in 1987 heralded a major change in the taxation of Australian dividends but one unforeseen aspect of this change is the build up of franking cred balances at the firm level. While the value of franking creds more generally has proven to be an important question for regulators and consultants in the determination of appropriate rates of return for regulated industries (Gray & Hall 2008) there has been ltle research into the value of franking cred balances at the firm level and this the focus of the analysis that follows. The theoretical link between tax and the value of a firm is well understood (Modigliani & Miller 1958, 1963; Miller 1977) though empirical support for the impact of taxes on the value of the firm has been less forthcoming (Myers 1984, 2001). 1 The introduction of dividend imputation provides an opportuny to further explore the impact of the taxes on the value of the firm. Empirical research into the impact of the imputation system supports the existence of tax effects in corporate financing and share valuation (Walker & Partington 1999; Faff et al. 2001; Twe 2001; Cannavan et al. 2004; Pattenden 2006) though this research is generally based on data drawn from the period prior to 1 July 1997 when tax regulations were introduced to lim franking cred schemes. The limed empirical research based on post 1 July 1997 data suggests that non resident investors are non longer able to benef from franking creds (Gray & Hall 2008; Lally 2008; Truong & Partington 2008). Wh the introduction of the dividend imputation system, Australian resident investors are creded wh Australian corporate tax paid by the Australian companies they invest in. These creds, called franking creds, are attached to dividends paid by Australian companies to resident shareholders eher as fully or partially franked dividends. This change to the Australian tax system creates an integrated tax system for resident shareholders wh all income being taxed at the marginal income tax rate. There are two groups of investors specifically excluded under the 1 There have been some exceptions (MacKie-Mason 1990; Graham 2003). 2 30/09/2009

3 regulation. The first group consists of non-taxable resident individuals and instutions implicly excluded because they do not pay tax. The second group includes non-resident taxpayers who are explicly excluded from the benef of franking creds. The exclusion of non-resident taxpayers is particularly important given that foreign investors account for more than 40% of the equy listed on the Australian Stock exchange. 2 While there has been ltle evidence of non-taxed resident investors attempting to benef from the franking cred system there are examples of non-resident investors using various schemes to do this (Cannavan et al. 2004). A further exclusion, perhaps more by omission than design, affects all investors and this relates to undistributed franking creds. Firms rarely distribute all of their earnings as dividends and so Australian resident shareholders rarely gain the full benef from franking creds that are accumulated by the firm. Since 1987, the amount of undistributed franking creds held by Australian companies has grown considerably, resulting in a substantial notional tax benef whheld at the corporate level for Australian resident shareholders. 3 This paper uses a large sample of corporations listed on the Australian stock exchange, drawn from the period 2001 to 2006, to explore the question of whether undistributed franking creds held at the firm level are of value to shareholders. There is some support for the argument that over this period franking creds are valuable to shareholders. The lerature is reviewed in the next section, wh data description provided in Section 3. The results of the analysis are reported and discussed in Section 4 and summary and conclusions follow in Section Lerature review The impact of dividend imputation on the cost of capal has been discussed extensively in the lerature wh a focus on further developing the tradional one-period model (Monkhouse 1993; 2 Foreign ownership of Australian listed equy amounted to 42% as of the end of the December quarter 2008 (ASX 2009). 3 Corporations maintain a record of franking creds that are available for distribution to investors and this balance is recorded in their annual reports. 3 30/09/2009

4 Officer 1994; Boyle 1996; Monkhouse 1996; Wood 1997; Lally & van Zijl 2003; Dempsey & Partington 2008). Much of the research is based on the assumption that the Australian market is segregated and this gives rise to the argument that share price impounds the value of franking creds. In this scenario undistributed franking creds may constute a valuable asset at the firm level. Yet, the value of franking creds at the firm level is not so clear when Australia is viewed as a small market whin an integrated world financial market (Officer 1994; Wood 1997; Gray & Hall 2006, 2008). The identy of the marginal investor has proven to be an important consideration in this lerature given the restrictions placed on non-resident Australian shareholder access to franking creds. If the marginal investor cannot benef from franking creds then the price of the share will not impound this benef (Officer 1994). Indeed, the possibily of restricting non-resident Australian shareholder access to franking creds while allowing resident investors full access to the creds is specifically catered for in the lerature. In an integrated market there may be a tendency for resident investors and non-resident investors to form tax based investment groups that are attracted to particular firms on the basis of the level of fully franked dividends paid by the firm (Wood 1997). Wood predicts that the market value of franking creds in a small, integrated market characterised by these two classes of investor is essentially zero. Argument about whether franking creds are actually valued in the market has continued wh the observation that the models used by regulators may not be internally consistent. On the basis of the officer model, is suggested that dividend yields would need to be set at excessively high levels to justify the franking creds value proposed in the lerature (Gray & Hall 2006, 2008). Indeed, Gray and Hall arguer that setting the value of franking creds to zero appears to be more consistent wh observed dividend yields. The response to this proposion in the lerature is generally based on the assumption that franking creds are valuable at the firm level (Lally 2008; Truong & Partington 2008). This poses the obvious question. Are franking cred balances valuable in the hands of the marginal investor? 4 30/09/2009

5 If franking creds are valuable then this should have some impact on corporate finance decisions. There is evidence to support an imputation driven change in capal structure away from debt and towards equy wh the introduction of the imputation system (Twe 2001; Pattenden 2006) from the 1980s through to the late 1990s. There is also evidence of imputation effects in the price drop that occurs on the dividend ex date, supporting the existence of valuable imputation tax benefs. 4 Yet, virtually all of the period covered by these studies is marked by fairly lax enforcement of the non resident franking cred benef rules, wh a number of schemes providing non residents access to franking cred benefs specifically denied to them under the legislation. But, from 1 July 1997 a set of regulations, introduced by the Australian Government, made considerably more costly for non resident shareholders to access franking creds and has been shown that this reduced the benef of franking creds to non resident investors around this time (Cannavan et al. 2004). 5 There has been ltle analysis of this question from 2000 onwards. The impact of dividend imputation on the cost of capal has attracted considerable attention in the recent lerature but empirical research has been limed, wh the exception of analysis of capal structure choice and dividend ex date price movement which tended to focus on the 1980s and the 1990s. Other than the work of Cannavan, Finn and Gray, there is ltle empirical work that touches on the value of franking cred balances held by Australian firms, particularly following the tax regulation changes that took effect from 1 July Modelling and hypothesis development Officer models the cash flows accruing to shareholders in terms of perpetuy equivalents for prof to shareholders, X S, dividends paid to shareholders, D S, operating income of the firm, X O, interest paid on loans, X D, and Corporate tax rate, T (Officer 1994). The corporate tax paid by the 4 This study is based on data drawn from the period from January 1995 through to March 1997 (Walker & Partington 1999). 5 The 45 day rule was one part of an extensive set of changes to the regulation of those receiving franking creds under the imputation system 5 30/09/2009

6 firm is defined, CTP T ( X X ) S =, wh the proportion of franking creds that shareholders O D access in equilibrium being the franking cred proportion, γ. Essentially, firms can only pay franked dividends out of income that is subject to Australian corporate tax and so not all dividends need be franked. Further, firms may choose not to distribute all of the prof subject to Australia corporate tax and so undistributed franking creds may accumulate over time. The proportion of creds optimally distributed to shareholders will depend on the mix of resident and non resident shareholders in the firm and the impact of imputation creds on price will be determined by the marginal shareholder (Wood 1997). And so, the prof attributable to share i can be decomposed into two terms (Officer 1994). Officer assumes that not all after tax earnings need attract imputation franking creds consistent wh γ [ 0,1]. Thus, the benef to shareholders is the sum of dividends received and the proportion of franking creds attached to these dividends. X S = DS γ T ( X O X D ) = DS + γctps + (1) Discounting wh an appropriate rate ( k) gives the value of the shares in the firm( S i ) in terms of the present value of one or more perpetuy. S DS + γctps = (2) k The franking cred proportion could take on a range of values ( 0 γ 1) and this will vary across firms. The discount rate will also vary across firms wh variation in risk. The required rate of return could be specifically modelled using eher a one factor market model (Sharpe 1964; Lintner 1965) or a multifactor model like the three-factor model proposed by Fama and French (Fama & French 1993). Alternatively, a statistical technique like fixed effects panel data analysis could be relied upon to capture the impact of firm specific effects such as discount rate and franking cred proportion using dummy variables. If is assumed that the marginal investor can fully use distributed franking creds then γ γ = 1 and the share price fully reflects available franking creds wh > 0 k. If γ = 0 then franking 6 30/09/2009

7 creds are not valued in the market and the share price is a function of the dividend perpetuy alone. In this case franking creds have no value to the marginal shareholder. If 0 < γ < 1 then γ franking creds are partially impounded into share prices again giving > 0. To get a sense of the k impact that franking creds could have on share price, assume that a share generates an equivalent dividend perpetuy of $20.00 per share and an equivalent corporate tax payments perpetuy of $6.00 per share. The discount rate is 0.05 and this is applied to both dividends and franking creds. If we set = 0. 5 γ 0.5 γ, as is often assumed in the lerature, then = = 10 k If γ = 1 then this ratio would be equal to 20. Table 1 provides a summary of the value of the firm equy under each of the three scenarios. The greater the proportion of franking, the more valuable is the franked dividend paying share in the hands of a resident shareholder. Wh the introduction of franking cred refunds after 1 July 2000 resident investors may obtain the full benef of the franking creds regardless of their taxable income. If the shareholder s tax is insufficient to claim a full tax offset for the franking cred then they may apply for a cash rebate for the remainder. For a non-resident shareholder, the value of franking creds is reduced due to the cost of setting up a scheme to recoup the creds and so a fully franked share is assumed to be less valuable to this group of investors, at worst being equal in value to an unfranked dividend paying share. [Insert Table 1 about here] The reliance on equivalent perpetuies is important in these simple valuation models. The actual cash flows generated by the underlying company could differ considerably over the life of the company and is likely that the distribution of franked dividends will vary over time to reflect variation in investment opportunies and business condions. Thus, actual franking cred balances will build up in some periods and reduce in other periods, regardless of whether these creds are valuable to the marginal shareholder or not. Thus, if the marginal investor is a non resident 7 30/09/2009

8 shareholder and share price does not reflect franking creds then franking cred balances will be uncorrelated wh the value of the firm, all else held constant. If a resident shareholder is the marginal shareholder then franking creds and franking cred balances will be valuable to the shareholder and this value will be reflected in share price. This leads to the first hypothesis. Hypothesis 1 Null: The value of franking cred balances is uncorrelated wh the value of the firm. Alternate: The value of franking cred balances is posively correlated wh the value of the firm. Given the importance of foreign investors in the Australian equy market is expected that this group of investors will have an impact on pricing. The lerature dealing wh foreign investors in small open economies, like the Australian economy, suggests that foreign investors favour large, liquid companies. This is evident both in Taiwan (Lin & Shiu 2003) and in Sweden (Dahlquist & Robertsson 2001) though there is also evidence of a preference for firms wh an international focus and more diversified shareholding. Given the restriction on non resident shareholder access to franking cred benefs in Australia is likely that non resident investors will prefer capal gains to dividends. Thus firms wh a dominant non resident shareholder group will tend to restrict dividend payments in favour of capal gains resulting in steadily growing franking cred balances over time. This gives rise to the second hypothesis. Hypothesis 2 Null: Undistributed franking cred balances are independent of firm characteristics Alternate: Undistributed franking cred balances are greatest in large, liquid companies wh low shareholder concentration and high international focus. 8 30/09/2009

9 As indicated in equation (2) the share price can be wrten in terms of the sum of the present value of an equivalent perpetuy of dividends and the present value of an equivalent perpetuy of franking creds. Under the null in both hypotheses, franking cred balances have no value to the marginal investor in companyi at time t and so the value of equy is the present value of the dividend perpetuy: S = 1 k D S, (3) Taking natural logs results in the linear form: s = d S, k * (4) ( ) ( ) ( ) where s = ln S d = ln D S, k * = ln k And, taking first differences of the model detailed in equation (4) gives: s = d S, k * (5) where s = ln( S ) ln( S 1 ) d = ln( D S, ) ln D S, 1 k * = ln( k ) ln( k 1 ) ( ) While the market value of equy can be observed, is not possible to observe equivalent dividend perpetuy for each of the firms and so dividend paid in the current year is used as a proxy for the dividend payment perpetuy. This assumption does not appear too extreme given the fairly sticky nature of dividends over time (Fama & French 2002). Further, while we do not have a proxy for the franking cred perpetuy is expected that the franking cred perpetuy will be posively correlated wh the dividend perpetuy and so the estimated coefficient for dividends will tend to capture both effects wh the result that the dividend coefficient will tend to exceed one where franking creds are valuable. Further, if franking creds are valuable then the franking cred balance held at the firm level will also be valuable and thus correlated wh firm value. This balance will tend to capture timing effects arising from time varying distributable profs and sticky dividends. 9 30/09/2009

10 If franking creds are valuable to shareholder then franking cred balances, fcb, accumulated over time by Australian firms should be posively correlated wh the value of the firm, consistent wh equation (2) and the estimated coefficient on dividends should exceed one. Given equation (4) and addional control variables for size, shareholder concentration and international focus the model could be wrten as: s = α + β 1 d S, + β 2 fcb + θ X + ε (6) where X is the vector of control variables The first difference form of the model, using d to reflect first differences where appropriate, is wrten as: ds = α + β 1 dd S, + β 2 dfcb + θ X + ε (7) The control variables enter both equation (6) and equation (7) in levels. Hypothesis 2 suggests that the magnude of the franking cred balances will be correlated wh the characteristics of the firm as previous research shows that non resident shareholders favour large liquid companies that have an international focus and diversified shareholdings when investing in a small, open economy. If the non resident shareholders that invest in Australian shares are subject to a classical tax system then they will prefer lower dividends than would be supported under the imputation system, resulting in a steady build up of franking cred balances over time. In analysis of this question is important to control for the level of dividends and so this variable naturally enters into the analysis. This is modelled in levels as: fcb α θ 1 + X + ε (8) = + β ds, And in change in levels as: dfcb α θ 1 + X + ε (9) = + β dds, 10 30/09/2009

11 4. Data The inial data set used in analysis is drawn from the population of listed Australian companies covered by Aspect Huntley over the period 2001 through to Franking cred balances are manually collected from annual reports for each of the firms. Some que large firms like Rio Tinto do not report this information in their annual reports, though many Australian listed corporations choose to report franking cred balance in the notes to their financial accounts. Information on shareholder concentration, foreign revenue and dividend buyback is also manually collected from the notes to the financial accounts. 6 Total dividend paid, aggregate market value of firm equy, book value of assets and book value of shareholders equy are obtained from the Aspect Huntley wh gaps filled manually from pdf copies of the financial accounts. The value of equy is measured using the natural log of the aggregate market value of the firm, ME and the change in the natural log of the market value of equy over the year, dme. Dividend payments are captured by the natural log of the sum of dividends paid and buy backs undertaken by the firm over the year, DIV as well as the change in the natural log dividend values, ddiv. Both the natural log of franking cred balances, FCB, and the change in the natural log, dfcb, are included in analysis. The natural log of the market value of assets is used as a proxy for size, SIZE, wh the market value of assets being approximated by sum of the book value of total assets less the book value of equy plus the market value of equy. The proxy for shareholder concentration is the ratio of the shares held by the largest shareholder to the shares held by the top 20 shareholders, SHRCONC and international focus is captured by the ratio of foreign revenue to total revenue, INTFOCUS. 6 The pdf copies of the financial accounts were obtained eher from the Aspect Huntley web se or form the individual company web ses /09/2009

12 Descriptive statistics are reported in Panels A, B, C and D in Table 2.. The inial sample of firms available for analysis ranges from 1250 firms in 2001 through to 1536 firms in 2006 but this paper focuses on those firms that report franking cred balances (columns 2 through 5, Panel A). The number of firms wh franking cred balances (512 firms on average each year) exceeds the number of firms paying dividends (449 firms on average each year) and those undertaking share buybacks (31 firms on average each year) over the period of the study. 7 Total franking cred balances for the sample increase over the sample period, rising from a total of $5 billion in 2001 through to $17 billion in 2006 and the average franking cred balance for those firms wh franking cred balances is $21 million over sample period. 8 [Insert Table 2, Panels A, B, C and D about here] While not reported separately in the panels, the average market value of equy is $749 million wh a maximum of $5.171 billion over the 6-year period for all firms that report franking creds in the final sample. The average dividends paid across all franking cred reporting firms over the 6-year period is $45 million and the average firm year share buy back is $104 million. The shareholder concentration ratio over all firm years is 34%, indicating the relative importance of the shareholding of the largest shareholder relative to the largest 20 shareholders for the sample as a whole. This is expected given the large number of small firms in the sample. The ratio of foreign revenue to total revenue averages 9% across all firm-years available for the franking cred reporting firms. Correlation matrices are reported for franking cred reporting firms based on the levels (Panel C) and change in levels (Panel D). 7 Franking cred balances tend to accumulate over time, increasing wh taxable income and reducing wh franked dividend payments. Firm dividend payments and share buy backs tend to vary wh the financial circumstances of the firm. 8 The method of calculating franking cred balances was changed wh the Business Tax Reform s Simplified Imputation System, which became effective from 1 July This required that franking balances as at 30 June 2002 be converted to reflect the cred balances on a tax paid basis rather than on an after tax distributable profs basis. The tax paid balances basically represent the imputation tax creds that are available for distribution. A factor of 30%/70% is applied to the 2001 and 2002 franking cred balances to bring them into line wh later year balances /09/2009

13 5. Analysis Fixed effects panel data analysis is used in testing hypotheses 1 and 2 proposed in Section 2. The key questions addressed include whether firm value is a function of franking cred balances and whether these franking cred balances are a function of firm characteristics like size, shareholder concentration or international focus. Given recent empirical work, there is some evidence that franking creds are not valued by shareholders in large Australian companies (Cannavan et al. 2004) after the 1 July This argument would appear to be consistent wh the substantial non resident shareholding evident in Australian listed companies and the importance of the regulations applying from 1 July 1997 that restrict the development of franking cred schemes for non residents. Yet, there has been ltle empirical analysis of the importance of franking creds to the broader Australian corporate sector particularly since 1997 other than that reported by Cannavan, Finn and Gray which focused on a small number of large, liquid companies. 5.1 Value of franking creds The analysis described in this section focuses on the sensivy of equy value to franking cred balances. If franking creds are valuable then equy value will be correlated wh franking cred balances. The estimated model is based on equations (6) and (7) wh the levels model set out in equation (10) and the change in levels model set out in equation (11) below. These models are estimated using unbalanced fixed effects panel data analysis and so there is adjustment for both firm specific effects such as the firm s discount rate and the appropriate imputation proportion as well as for time effects across the whole sample. The natural log of market value of equy, sum of dividends and share buybacks, franking cred balance and size are included in the model along wh the measures of shareholder concentration and international focus. The levels model takes the form: 13 30/09/2009

14 LME = α 0i + α 0t + α 1 LDIV S, + α 2 LFCB + α 3 SIZE + α 4 SHRCONC + α 5 INTFOCUS + ε The change in levels model takes the form: (10) dlme = β 0i + β 0t + β 1 dldiv S, + β 2 dlfcb + β 3 SIZE + β 4 SHRCONC + β 5 INTFOCUS + ε (11) The analysis focuses on those firms that report franking cred balances in their annual reports and robust standard errors are used in all panel data analysis reported below. The results for levels model are reported in Panel A of Table 3 and the results for the change in levels model are reported in Panel B of Table 3. In each panel the model is estimated using all available observations, that part of the sample that falls whin the largest 100 companies on the Australian Stock Exchange (by book value of total assets), that part of the sample that is not in the largest 100 group and those firms where a dividend payment is reported. [Insert Table 3 about here] Each of the four estimated models based on equation (10) are statistically significant at the 5% level of significance, given the F-test statistics reported on the last line of the table. The estimated franking cred balance coefficients are statistically significantly different from zero for each of the models, consistent wh hypothesis one that franking creds are of value to shareholders. While non resident franking cred schemes were apparently restricted in the 1990s these results suggest, at least prima facie, that non-resident shareholders are benefing from franking cred schemes over the period of this study, 2001 through to The estimated dividend coefficients are posive and statistically significant at the 10% level for all but the large firm group. Thus, firm equy market value is posively correlated wh the sum of dividend and share buy back payments. This result is consistent wh the standard dividend 14 30/09/2009

15 perpetuy pricing model as wrten in equation (3) where is assumed that the dummy variables estimated as part of the fixed effects approach capture firm specific and time based variation in the discount rate and imputation factor. For the control variables, while the size coefficients are posive and statistically significant there are no statistically significant coefficients for eher shareholder concentration or international focus. Thus for the levels based analysis there is evidence of franking cred balance being of value to the firm for large firms. It is important to note that large liquid firms tend to attract non resident shareholder investment (Dahlquist & Robertsson 2001; Lin & Shiu 2003) and so this adds credence to the possibily that this effect is tax driven. Further, equy market value is posively related wh size and this effect is persistent across each of the four data sets. Larger firms tend to have greater franking cred balances. The coefficients estimated for equation (11), the differenced model, are reported in Panel B of Table 3. These results for this specification are somewhat different from those reported for the levels model. The estimated franking cred balance coefficients are posive statistically significant for all but the large firm group. Yet, the change in dividend for the year is not correlated wh the change in equy value over the same period. The size coefficients remain posive and are statistically significant in all cases. Finally, the change in international focus has more impact on the change in share value in this specification, wh increases in international focus being associated wh decreases in the value of equy over the period for all but the largest 100 firm sub sample. Shareholder concentration variable is important for the larger firms in the sample, wh a statistically significant posive coefficient for this group of firms. There is some variation in the results, particularly for dividends, shareholder concentration and international focus though both size and franking cred balances are posively correlated wh equy market value whether levels or change in levels specifications are chosen for analysis. Thus, the estimated coefficients for firm franking cred balance support the Hypothesis 1 proposion that 15 30/09/2009

16 franking creds balances are valuable and thus that franking creds themselves must also be of value in the hands of the marginal shareholder. 5.2 Determinants of franking cred balances The second hypothesis is concerned wh the determinants of the level of franking cred balances wh estimation of levels model using equation (12): LFCB = α 0i + α 0t + α 1 LDIV S, + α 2 SIZE + α 3 SHRCONC + α 4 INTFOCUS + ε (12) and the change in levels model using equation (13): dlfcb = β 0i + β 0t + β 1 dldiv S, + β 2 SIZE + β 3 SHRCONC + β 4 INTFOCUS + ε (13) It is expected that where non resident shareholders would prefer to reduce the level of dividend payments where they are unable to benef from franking cred schemes and are subject to a classical tax system. In this suation is expected that franking cred balances would be higher than would otherwise be the case. Given that there is evidence that non resident shareholders tend to favour large liquid firms wh a foreign focus and diversified shareholdings is expected that firm variables that capture these characteristics would be correlated wh the level of franking cred balances that firms accumulate. The results of analysis are reported in the Panels A and B of Table 4 and while there is no evidence of statistically significant international focus or shareholder diversification effects on the level of franking cred balances there is some evidence of size effects. The size coefficients are posive and statistically significant in the levels analysis while the being negative and less often 16 30/09/2009

17 statistically significant in the change in levels analysis. 9 The negative size coefficient is inconsistent wh hypothesis 2 because if the non resident shareholder were to control franking cred balances then this would be most evident in the larger firms, yet the larger the firms the smaller the proportional change in their franking cred balances. [Insert Table 4 about here] 6. Conclusions The question of whether franking creds are valuable is an important one as this affects regulation and valuation of firms. This paper provides some insight into the impact of franking on share prices through analysis of the relation that exists between equy value and reported franking cred balances. Since imputation was introduced in 1987 there has been a fairly steady increase in the level of franking cred balances held at the corporate level. The existence of these balances is an important question for Australian resident shareholders because they represent valuable creds that are denied them by the firms they invest in. While non resident shareholders may gain ltle of these benefs due to legislative restrictions and regulatory changes, particularly those operating since 1 July 1997, resident shareholders currently have considerable flexibily in accessing these creds even where they pay no tax. The question of whether franking creds generally and franking cred balances more specifically are impounded in share prices depends on the identification of the marginal shareholder. If the marginal shareholder is a non resident wh ltle access to franking cred benefs then prices will not reflect these creds and equy market value will not be correlated wh these reported balances. If the marginal shareholder is an Australian resident for tax purposes or 9 The analysis was repeated whout the size variable but there was ltle change in the estimated coefficient signs or in the lack of statistical significance evident in the remaining coefficients /09/2009

18 non-resident shareholders are able to capture the benef of these creds then prices will reflect the value of these creds. The results from analysis reported in the paper suggest that the franking cred balances are correlated wh firm value and thus franking creds are of value to the marginal shareholder. There is que strong support for the existence of valuable franking cred balances in the levels specification and in all but the large firm sub group for the change in levels specification. There is ltle support for the second hypothesis, that companies most attractive to nonresident shareholders will report greater levels of franking creds. The very firm characteristics that should be attractive to non resident shareholders do not explain the magnude of franking cred balances in a consistent manner. Even the sign of the size coefficient varies between levels and change in levels based specifications /09/2009

19 Reference ASX, Australian Cash Equy Market. In: ASX: An International Centre for Resources Capal, pp. 1-8 Boyle, G.W., Corporate investment and dividend tax imputation. The Financial Review 31, Cannavan, D., Finn, F., Gray, S., The value of dividend imputation tax creds in Australia. Journal of Financial Economics 73, Dahlquist, M., Robertsson, G., Direct foreign ownership, instutional investors, and firm charactersics. Journal of Financial Economics 59, Dempsey, M., Partington, G., Cost of capal equations under the Australian imputation tax system. Accounting and Finance 48, Faff, R., Hillier, D., Wood, J., Taxation and Black's zero beta strategy revised. Financial Analysts Journal 57, Fama, E.F., French, K.R., Common risk factors in the returns on stocks and bonds. Journal of Financial Economics 33, 3-56 Fama, E.F., French, K.R., Testing the trade-off and pecking order predictions about dividends and debt. Review of Financial Studies 15, 1-33 Graham, J., R., Taxes and corporate finance: A review. Review of Financial Studies 16, Gray, S., Hall, J., Relationship between franking creds and the market risk premium. Accounting and Finance 46, Gray, S., Hall, J., Relationship between franking creds and the market risk premium: A reply. Accounting and Finance 48, Lally, M., Relationship between franking creds and the market risk premium: A comment. Accounting and Finance 48, Lally, M., van Zijl, T., Capal gains tax and the capal asset pricing model. Accounting and Finance 43, Lin, C.H., Shiu, C.-Y., Foreign ownership in the Taiwan stock market - An empirical analysis. Journal of Multinational Financial Management 13, Lintner, J., The valuation of risk assets and the selection of risky investments in stock portfolios and capal budgets. Review of Economics and Statistics 47, MacKie-Mason, J.K., Do taxes affect corporate financing decisions? Journal of Finance 45, Miller, M., Debt and taxes. Journal of Finance 32, Modigliani, F., Miller, M., The cost of capal, corporation finance and the theory of investment. American Economic Review 48, Modigliani, F., Miller, M., Corporate income taxes and the cost of capal. American Economic Review 53, Monkhouse, P.H.L., The cost of equy under the Australian dividend Imputation tax system. Accounting and Finance 33, 1-18 Monkhouse, P.H.L., The valuation of projects under the dividend imputation tax system. Accounting and Finance 36, Myers, S.C., The capal structure puzzle. Journal of Finance 39, Myers, S.C., Capal structure. Journal of Economic Perspectives 15, Officer, R.R., The cost of capal of a company under an imputation tax system. Accounting and Finance 34, 1-17 Pattenden, K., Capal structure decisions under classical and imputation tax systems: A natural test for tax effects in Australia. Australian Journal of Management 31, /09/2009

20 Sharpe, W.F., Capal asset prices: A theory of market equilibrium under condions of risk. Journal of Finance 19, Truong, G., Partington, G., Relationship between franking creds and the market risk premium: A comment. Accounting and Finance 48, Twe, G., Capal structure choices and taxes: Evidence from the Australian dividend imputation tax system. International Review of Finance 2, Walker, S., Partington, G., The value of dividends: Evidence from cum-dividend trading in the ex-dividend period. Accounting and Finance 39, Wood, J., A simple model for pricing imputation tax creds under Australia's dividend imputation tax system. Pacific-Basin Finance Journal 5, /09/2009

21 Table 1, Example of the effect of franking creds on share value Unfranked Dividends Partially franked dividends Fully franked dividends Gamma Dividends Perpetuy amount Corporate tax paid perpetuy amount Discount rate Dividend component of price Franing cred component of price Final share price /09/2009

22 Table 2, Descriptive Statistics Panel A, Sample characteristics and franking cred balances Sample counts and franking cred balances are provided for those firms that report their franking cred balance details in their annual reports for the years, 2001 through to The number of companies in the total sample along wh the number of companies that report franking cred balances, that pay dividends, involved in share buy backs appear in columns 2 through 5 of Panel A. Total of the franking cred balances for the year ($millions) and average franking cred balance per firm ($millions) are reported by year as well as for the full period in the last two columns of panel A. Companies Wh franking cred balances Companies paying dividends Companies wh buybacks Total franking cred balance ($ Millions) Average franking cred balance ($ Millions) Number of Year Companies in inial sample Total Average Panel B, Panel data descriptive statistics Summary statistics for the pooled sample are reported in Panel B for franking cred balance reporting firms. This panel includes statistics for all firm-year observations for these firms. LME is the natural log of the aggregate market value of equy, dlme is the change ME, LDIV is the natural log of the sum of dividend and share buy back, dldiv is change in LDIV, LFCB is the ratio of franking cred balance to book value of total assets, dlfcb is the change in LFCB, SIZE is the natural log of the sum of ME and book value of total assets less book value of shareholders equy, SHRCONC is the ratio of the shareholding of the largest shareholder to the total shareholding of the top 20 shareholders, INTFOCUS is the ratio of the foreign revenue to total revenue. Average Standard deviation Maximum Minimum Count LME dlme LDIV dldiv LFCB dlfcb SIZE SHRCONC INTFOCUS /09/2009

23 Panel C, Panel data correlation coefficient estimates levels Correlation coefficients for the variables used in the levels and change in levels analysis are estimated for pooled sample, which is used in the analysis that follows. Variables are defined in the note to Panel B in this table. (N = 2172) LME LDIV LFCB SIZE SHRCONC LDIV 0.87 LFCB SIZE SHRCONC INTFOCUS Panel D, Panel data correlation coefficient estimates change in levels Correlation coefficients for the variables used in the levels and change in levels analysis are estimated for pooled sample, which is used in the analysis that follows. Variables are defined in the note to Panel B in this table. (N = 1490) DLME DLDIV DFCB SIZE SHRCONC DLDIV 0.10 DFCB SIZE SHRCONC INTFOCUS /09/2009

24 Table 3, Regression analysis of equy value effects Panel A, Regression using levels specification Fixed effects panel data analysis is used in the estimation of the relation between the natural log of the market value of equy (LME) and the set of explanatory variables, natural log of dividends paid and share buy backs (LDIV), natural log of franking cred balance (LFCB), the natural log of the sum of the total book value of assets less the book value of equy plus the market value of equy (SIZE), the ratio of the shareholding of the largest shareholder to the that of the largest 20 shareholders (SHRCONC) and the ratio of foreign revenue to total revenue (INTFOCUS). Book value of total assets is used to rank all available firms into the largest 100 firm group and the not largest 100 firm group and these codes are carried through to this analysis of firms that report franking creds. Robust standard errors are used in the calculation of t-statistics. * (+) is statistically significant at the 5% (10%) level. Full sample Largest 100 firms Not one of the largest 100 firms Firm paid dividend in the year LDIV (1.89) (1.14) (1.68) (1.89) LFCB * * * * (3.73) (2.18) (3.28) (3.73) SIZE * * * * (40.94) (18.77) (36.17) (40.94) SHRCONC (1.16) (0.12) (1.42) (1.16) INTFOCUS (1.02) (-0.09) (1.14) (1.02) Observations Groups F-test * * * * Panel B, Regression using change in levels specification Fixed effects panel data analysis is used in the estimation of the relation between the change in natural log of the market value of equy (dlme) and the set of explanatory variables, change in the natural log of the sum of the dividends paid and share buy backs (dldiv), change in the natural log of franking cred balance (dlfcb), the natural log of the sum of the total book value of assets less the book value of equy plus the market value of equy (SIZE), the ratio of the shareholding of the largest shareholder to the that of the largest 20 shareholders (SHRCONC) and the ratio of foreign revenue to total revenue (INTFOCUS). Book value of total assets is used to rank the all available firms into the largest 100 firm group and the not largest 100 firm group and these codes are carried through to this analysis of firms that report franking creds. Robust standard errors are used in the calculation of t-statistics. * (+) is statistically significant at the 5% (10%) level. Full sample Largest 100 firms Not one of the largest 100 firms Firm paid dividend in the year ddiv (0.10) (-0.70) (0.09) (0.10) dfcb * * * (3.72) (0.69) (3.49) (3.72) SIZE * * * * (8.35) (3.76) (7.73) (8.35) SHRCONC * (1.37) (2.74) (1.19) (1.37) INTFOCUS * * * (-2.32) (-0.18) (-2.60) (-2.32) Observations Groups F-test 18.33* 4.17* 16.44* 18.33* 24 30/09/2009

25 Table 4, Regression analysis of the franking cred balances Panel A, Regression using levels specification Fixed effects panel data analysis is used in the estimation of the relation between the natural log of the franking cred balance (LFCB) and the set of explanatory variables, natural log of the sum of dividends paid and share buy back (LDIV), the natural log of the sum of the total book value of assets less the book value of equy plus the market value of equy (SIZE), the ratio of the shareholding of the largest shareholder to the that of the largest 20 shareholders (SHRCONC) and the ratio of foreign revenue to total revenue (INTFOCUS). Book value of total assets is used to rank the all available firms into the largest 100 firm and not largest 100 firm subsets wh this ranking carried through to this sample of firms that report franking cred balances. Robust standard errors are used in the calculation of t-statistics. * (+) is statistically significant at the 5% (10%) level. Full sample Largest 100 firms Not one of the largest 100 firms Firm paid dividend in the year LDIV (1.60) (0.79) (1.07) (1.60) SIZE * * * * (11.50) (4.15) (10.78) (11.50) SHRCONC (0.22) (-0.07) (-0.34) (0.22) INTFOCUS (-0.07) (-0.28) (0.25) (-0.07) Observations Groups F-test 39.42* 5.07* 37.00* 39.42* Panel B, Regression using change in levels specification Fixed effects panel data analysis is used in the estimation of the relation between the natural log of the franking cred balance (dlfcb) and the set of explanatory variables, natural log of the sum of dividends paid and share buy back (dldiv), the natural log of the sum of the total book value of assets less the book value of equy plus the market value of equy (SIZE), the ratio of the shareholding of the largest shareholder to the that of the largest 20 shareholders (SHRCONC) and the ratio of foreign revenue to total revenue (INTFOCUS). Book value of total assets is used to rank the all available firms into the largest 100 firm and not largest 100 firm subsets wh this ranking carried through to this sample of firms that report franking cred balances. Robust standard errors are used in the calculation of t-statistics. * (+) is statistically significant at the 5% (10%) level. Full sample Largest 100 firms Not one of the largest 100 firms Firm paid dividend in the year dldiv (0.44) (0.85) (0.08) (0.44) SIZE * (-1.61) (-2.83) (-0.10) (-1.61) SHRCONC (-0.17) (-1.78) (1.19) (-0.17) INTFOCUS (-0.28) (0.23) (-0.42) (-0.28) Observations Groups F-test * /09/2009