Australian off-market buy-backs: Empirical evidence

AUSTRALIAN OFF-MARKET BUYBACKS: EMPIRICAL EVIDENCE Adi Bowo Kharisma, Balasingham Balachandran and Michael Skully* Monash University

Address for correspondence: Professor Michael Skully Department of Accounting & Finance Faculty of Business and Economics Monash University PO Box 197 Caulfield East Vic 3145 Australia Fax: Phone:

(61)3 9903 2407 (61)3 9903 1443

Email:

[email protected] [email protected] [email protected]

* The author who will attend the conference and present the paper

1

AUSTRALIAN OFF-MARKET BUY-BACKS: EMPIRICAL EVIDENCE

ABSTRACT

This study examines market reaction to the announcements of off-market buy-backs in Australia over the period from June 1989 to October 2003. Australian imputation system, unlike classical tax system applied in the US, provides incentives for resident shareholders to receive dividends in lieu of capital gains. Hence, buy-back with franking credit distribution should be seen as more attractive than one with merely capital gain distribution. Surprisingly, a negative relation was found between price reaction and franking credit distribution in the off-market buy-backs. In a cross-sectional analysis, the magnitude of price reaction to off-market buy-back announcements was positively related to firm’s price earnings ratio, and negatively related to the firm’s size, dividend yield and the directors’ interests. However, no significant relations existed for variable buy-back size. Similarly, a firm’s first off-market buy-back program does not convey stronger signal than the subsequent ones. Key words: price reaction, buy-back, off-market buy-back, Australia

2

1. INTRODUCTION Share buy-back or share repurchase1 is defined as the act of a company purchasing its own outstanding shares. As a cash disbursement method, share buy-back is distinctive since it allows shareholders to decide whether or not to accept the buy-back offer. If the offer were to be accepted, these repurchased shares are subsequently cancelled and hence all rights attached to those shares are extinguished.2 Until 1 November 1989, share buy-backs were not permitted in Australia, as the Companies Code required firms to protect the interests of their creditors.3 However, a new interest share buy-backs developed after 1987 (Wise & Wise, 1987). This was because of the US companies’ share prices proved less effected due to their buy-back activities during the 1987 share market crash (Renton, 1998). The subsequent finding of US Securities and Exchange Commission that buy-back companies outperformed the S&P index by about 80% further underlined their importance during the crash (Leo & Hoggett, 1998). Buy-backs in Australia have since become significant with the value of buy-back in 1999 (A$11.562 billion) far outstripping that of new floats (A$4.930 billion) (ABS, 1999). Their grouping popularity can be seen in the increase of their value and number from only A$18 million with 7 buy-backs in 1990 to A$3.54 billion and 49 buy-backs in 1997 (Izan, Mitchell, & Lim, 2002).4 The buy-back programs in Australia are also becoming larger. The buy-back program announced by Telstra in August 2003, for instance, was worth approximately A$1 billion.

1

The terms share buy-back, share repurchase and stock repurchase are used interchangeably throughout this report. 2 This is the standard practice in Australia (sec. 206PC). In contrast. Section 317(b) of the US Internal Revenue Code provides that stock might be cancelled, retired, or held as ‘treasury stock’. Common US practice is to retain the repurchased shares in the company’s treasury (as ‘treasury stocks’) until they are cancelled or reissued (Peirson, 1990). 3 The ability of a company to meet its commitments to the creditors might be prejudiced by the company’s insolvency following the buy-back. 4 Australian Corporations Law recognizes five types of share buy-backs: on-market, equal access scheme, selective, employee share scheme and minimum holding buy-backs (see Division 2 of Part 2J.1 of the Corporations Act 2001).

3

The extensive studies on share buy-backs in the US and, more recently Australia acknowledges of their importance as a capital management tool for enhancing shareholders’ wealth. Empirical results, in both countries have shown a significant positive market reaction to the buy-back announcements (see for example, Vermaelen (1981), Dann (1981), Comment & Jarrell (1991), Lamba & Ramsay (1999), Otchere & Ross (2002) and Balachandran, Faff, & Troiano (2002)). However, whilst prior Australian studies have documented positive price reaction to the share buy-back announcements. These have concentrated on on-market buy-backs and merely tested the ‘information signalling/undervaluation’ hypothesis (see for example, Otchere & Ross (2002), Balachandran et al. (2002) and Izan et al.(2002)). It might be true that companies undertake buy-backs to convey management’s private information to the market about the undervaluation of the companies’ share prices and hence market reacts accordingly. It might also be true that ‘information signalling’ hypothesis best explains market reaction to the buy-back announcements (see for example, Vermaelen (1981), Dann (1981) and Comment & Jarrell (1991)). However, as Otchere & Ross (2002) noted, US data is not as good as Australia to test the ‘information signalling/undervaluation’ hypothesis due to the lack of disclosure for a particular buy-back. Besides, it is possible that the market reacts to other management reasons such as leverage effect, tax effects as well as excess cash and lack of profitable investments. Australian data also has problems since it is impossible to ascertain whether or not management’s intentions are credible (Otchere & Ross, 2002). This will subsequently lead to a ‘gambling’ behaviour in the market whether or not to believe management’s reasons. These facts justify the market perspective approach of this study rather than the corporate approach used by other Australian researchers. One apparent market perspective is taxation. This is important due to the clear distinction between Australian and US tax legislation, which is evident with the presence of imputation in Australia. In fact, it is believed that imputation will reduce the tax benefits of share buy-backs vis-à-vis dividend payments (see Balachandran et al.(2002) and Rosa et al.(2000)). This is because it makes the effective tax rate on the dividends become

4

relatively low compared to the capital gains tax.5 Therefore this study, considers the following questions: (i) does the proportion of franked dividend component in the buyback price affect market reaction to the off-market buy-back announcements? (ii) what drives market reaction to the off-market buy-back announcements? and (iii) does the firm’s first off-market buy-back program convey stronger signal than the subsequent ones? These questions have not yet been addressed in the literature. 2. THE IMPUTATION CREDIT HYPOTHESIS Under a classical taxation system, corporate entities and their members are treated as distinct taxpayers, hence are taxed separately. As this resulted in double taxation on the same underlying corporate profits, once in the hands of the company and again in the hands of the shareholders. This “double taxation” of dividends was addressed on 1 July 1987 when an imputation system was introduced in Australia. It provides corporate shareholders with credits for tax paid by the corporate tax entity, so that the corporate after-tax profits distributed as dividends (D) are typically not taxed again.6 Simply put, shareholders are taxed on the ‘grossed up’ dividends (D/(1-tc)) and other income, but they are able to utilize the imputation/franking credits7 (D/(1-tc)*tc) to reduce their personal income tax liability.8 This imputation system, however, only applies when both the corporate tax entity and the shareholder satisfy the ‘residency requirements’ and fulfill specific ‘anti-avoidance rules’.9 In addition, a ‘45-day (ownership) rule’ must be satisfied to be entitled for the imputation credits. Finally, non-resident and tax-exempt investors are not entitled for franking credits.10 5

Unlike capital gains that require investors to pay the capital gains tax on their marginal tax rate (discounted after 30 September 1999), dividends are effectively ‘tax-free’ if the investors’ marginal tax rate is equal to the corporate tax rate (need only to pay Medicare levy of 1.5%). Even, if their marginal tax rate is lower than the corporate tax rate, the investors can use the ‘excess’ tax credits to reduce their tax on other income or obtain a tax rebate (this applies to resident individuals, complying superannuation funds and registered charities since 1 July 2000). 6 The only condition where resident investors have to pay additional tax on the dividends is when their marginal tax rate is greater than the corporate tax rate. 7 It should be noted that franked dividends must be paid out of after-tax company’s profits. 8 tc represents corporate tax rate. 9 Anti-avoidance rules are designed to deny imputation benefits in certain circumstances, for example where such benefits are streamed to certain members ahead of others. 10 Non-resident investors must pay withholding tax on the unfranked component of any dividends, but not on the franked component. However, they will usually be taxed in their home country for the before-tax component of their Australian income. Hence, it is often only a matter of timing that benefits the non-

5

The value of franking credits has been subject to many empirical studies, especially in the analysis of the dividend drop-off ratio phenomenon.11 Hathaway & Officer (1992), for example, employed regression analysis to test whether the ex-dividend drop-off (as a proportion of the dividend) reflects the dividend’s franking level. Their results suggest that there is positive valuation on franking credits; but that this value of such credits is ambiguous. Similarly, Bruckner, Dews, & White (1994) estimated the initial value of franking credits at 33.5 cents per dollar of face value and 68.5 cents in a later sub-period. Extending the sample period to June 199512 and partitioning the sample into small and large capitalization firms, Hathaway & Officer (1996) concluded that whilst unclear results were found for small firms, franking credits were worth approximately $0.49 per dollar of face value for large ones. These studies show that market places a positive value on franking credits. Further empirical evidence from the dividend drop-off ratio in Australia and franking credit value is outlined below. Brown & Walter (1986) found clear evidence that over the 1974 to 1985 period, the dividend drop-off ratio was less than one (approximately 0.75). However, this study was conducted when Australia still employed a classical tax system. Imputation system, in fact, should change the ex-dividend behaviour of Australian share prices. Brown & Clarke (1993) provided theoretical support for such an increase, but they did not find empirical support, as there was found a fall in the drop-off ratio. Later researchers, however found empirical support for the Brown & Clarke (1993) argument. Bellamy (1994), for example, found that the dividend drop-off ratio was significantly higher for

resident investors. In the case of tax-exempt investors (whose assessable income is less than the tax threshold), Bellamy (1994) stated that they are one of the potential losers in the imputation environment. He argued that it is the motivational shift from maximizing after-tax income to maximizing before-tax income that causes them to lose, as firms are no longer worried to pay tax as they could transfer franking credits to their shareholders. However, as pointed out by Officer (1988), there was possibility to sell franking credits for those who could not derive the maximum benefits from such tax credits, and hence creating opportunities for those investors to benefit from the imputation system. 11 Dividend drop-off ratio is the ratio between the decline in the share price on the ex-dividend day to the dividend per share. A company with share price and dividend per share of one dollar and five cents respectively, which experiences a drop in the share price to $0.94 in the ex-dividend day, for example, has a drop-off ratio of 1.2. 12 The period of their first study is only from 1 July 1987 to 30 June 1989.

6

dividends with franking credits than for ones without them. This further confirms that franking credits are valuable to the market. Another study on dividend drop-off ratio in Australia was conducted by Walker & Partington (1999). They observed the new phenomenon of shares trading cum-dividend during the ex-dividend period.13 This new methodology allowed the observation of shares trading contemporaneously with and without dividends, so that dividend value could be measured with less noise than in traditional ex-dividend studies, such as by Elton & Gruber (1970).14 In their observation of 1015 pairs of time-matched trades arising from 93 ex-dividend events from 1 January 1995 to 1 March 1997, determined that the average of the ‘instantaneous’ drop-off ratio is 1.23 across trades and 1.15 across ex-dividend events.15 This implies that the market value of $1 fully franked dividend is more than $1.16 Examining the value of Australian imputation credits to equity investors over the period from 1989 to 2000, Handley & Maheswaran (2003) estimated that the after company before personal tax (utilization) value of Australian imputation credits, expressed as a proportion of the amount received, is approximately 0.80 across all investors (γ=0.80). In their study, γ is also estimated to be approximately 0.90 for both resident individuals and resident funds and 0.13 for non-resident investors. These indicate that imputation credits are perceived positively by the market, especially by resident individuals and resident funds. Other support for the positive valuation on franking credits are provided by the following arguments or evidence. First, the fact that franking credit trading gained in prominence for a relatively brief period during the mid 1990s shows that franking credits were

13

The decline in the share price is calculated as the closing price on the last cum-dividend day minus the closing price on the ex-dividend day (Walker & Partington, 1999). 14 Walker & Partington (1999) argue that the ‘instantaneous’ drop-off ratio provides much cleaner results. 15 Unlike traditional drop-off ratio, the prices for the ‘instantaneous’ drop-off ratio are “matched to within one minute of each other when shares were traded cum-dividend during the ex-dividend period”, (Walker & Partington, 1999, p.291). 16 After tax effects and transaction costs.

7

valuable to those who could fully use them (Antioch, 2001).17 Second, employing his model for pricing imputation tax credits and assuming that the Australian market was segmented from world equity markets, Wood (1997) suggested that the market value of tax credit associated with securities held by all investors is approximately 60% of their face value.18 Third, study on Dividend Reinvestment Plans (DRPs) by Chan, McColough, & Skully (1993) found empirical evidence to support shareholders’ demand for immediate transfer of imputation credits as they found a significant abnormal return of 0.55% on the announcement date of DRPs.19 And finally, a distinct clustering of franking levels near 0% and 100%, as well as the trend toward these levels of franking, imply that managers believe that shareholders value imputation credits (Bellamy, 1994). All of the evidence above exhibits how market places a positive valuation on the tax credits available under the imputation system. Assuming that tax credits are valuable, then the greater franked dividend component as opposed to the capital component in the off-market buy-back price means that greater value after tax for the participating shareholders.20 This is because franking credits can reduce the effective tax rate on the dividend component and so reduce the effective tax rate on the overall consideration received by shareholders. Jeffrey (1991) in his calculations showed that the more deemed franked dividend, the less tax to be paid by a shareholder in the off-market buy-back. With a very high franked dividend component, participating shareholders might substantially reduce their tax and even obtain a tax rebate for the capital loss incurred if the capital component is lower than their cost of the shares21. 17

Antioch (2001) noted that franking credits were traded both on and off market, and were often arranged through a stockbroker or other financial intermediaries. Such activities were curtailed by Budget Paper No 2: Budget Measure 1997-98 (Antioch, 2001). 18 There are two classes of investors in his model, those who can use the tax credits and those who cannot. 19 Bellamy (1994) argues that delay in the payment of the imputation credits will cause shareholders to lose the time value of the tax deductions. 20 Chan et al. (1993) noted that whilst imputation has removed double taxation on dividend, the capital gains tax means double taxation on retained earnings (see Howard & Brown (1992)). 21 Since this research was completed the Australian Tax Office on 14 January 2004 issued a draft determination (TD 2004/D1) which indicates that the price set for off market buy backs should better reflect the share’s then market price. The market value would be a function of the volume weighted closing price of the last five days before the first announcement of the buy back adjusted by the change in ASX 200 Index from the first day of the announcement to the day on which the buy back closes.

8

This is consistent with the CAPM-based imputation models by Ashton (1989), Lally (1992), Monkhouse (1993) and Wood (1997), which suggest that higher dividends mean higher imputation benefits for resident investors. Moreover, examining market reaction to the announcement of a new foreign investor dividend tax credit system in New Zealand, Chay & Marsden (1996) found significant positive abnormal returns concentrated on firms with fully imputed high dividend yields and high corporate tax rates. As a consequence, the price reaction is expected to be stronger the greater the franked dividend component in the off-market buy-back price.

There is a special case for the shareholders at the highest marginal tax rate as their value after tax declines as the dividend component increase. Nevertheless, this will not distort our expectation, as there is another explanation for this, i.e. the ‘tax-induced clientele’ hypothesis.22 In addition, it is observed that around 80 per cent of all Australians now pay only a top marginal tax rate of 30 per cent or less (ATO, 2003). Besides, married couples typically invest in the share market in the name of the spouse with the lowest marginal tax rate and so maximizes the benefits of the imputation system. These suggest that most Australian individual investors are likely to benefit from franking credits in the Australian tax environment.23 Moreover, the fact that Australian superannuation (since 1 July 1988) are taxed at 15% means that they benefited significantly from the Australian imputation system.24 This, in turn, increases their demand for franked dividend in order to exploit the tax rebate available to them.25

22

‘Clientele effect’ refers to different groups of shareholders (clienteles) that have different preferences over the dividend policies. ‘Tax clientele’ hypothesis basically argues that an investor would prefer to receive capital gains or low dividend payouts if the tax payable on capital gains is lower than that on dividends. 23 Hence, it is expected that this will trigger a positive market reaction to the announcement of franked dividend component in the off-market buy-back price. 24 This is for complying superannuation funds. Previously, these funds’ earnings were tax exempt. Noncomplying funds are taxed at the highest marginal tax rate. 25 Allen, Bernado, & Welch (2000) suggested that firms paying dividends are attractive for institutional investors that are taxed at a lower rate than individuals. Hence, they argued that dividend induced ‘ownership clientele effects’.

9

3. RESEARCH DESIGN (i) Data and Sample Off-market buy-backs announced by Australian companies listed in ASX for the period from June 1989 to October 2003 were observed in this research.26 Initial sample of buyback announcements is obtained from Bloomberg (using corporate action calendar function), DatAnalysis and SIRCA (ASX).27 JBWere also provided a list of major Australian companies engaged in buy-backs. If an announcement date was determined to be too ambiguous, the observation was deleted from the sample. The final sample required that each observation meet the following criteria: a. No confounding events are reported in the five days around the announcement date (see Harris & Ramsay (1995), Lamba & Ramsay (1999) and Otchere & Ross (2002))28; b. The buy-back should be the case where the company repurchases its ordinary shares; and c. Daily share price data over the estimation (-260, -6)29 and examination periods (5, +5)30 are available from DataStream, IRESS and SIRCA database.

26

Off-market buy-backs consist of 4 categories, i.e. equal access schemes, selective buy-backs, employee access schemes and minority holding buy-backs. However, only equal access buy-backs are used here as the representative of the off-market buy-backs. This is because of their franking credit distributions. Remember that Australian ‘anti-streaming rules’ prohibit such distribution in other types of buy-backs unless ATO approval has been obtained. They also provide all shareholders with equal opportunity to participate in the buy-backs. Nevertheless, there are six selective buy-backs on equal access conditions included in the sample of the off-market buy-backs. 27 A rigorous search on DatAnalysis employing Signal G was undertaken for the buy-back announcement dates. It is crucial to ensure that the announcement is unexpected by the investors, otherwise it will already be reflected in the share price as suggested by Fama (1970). Hence, announcement date is determined as the earliest date where ‘sufficient’ information becomes publicly available, either through lodging with the ASX, letter to shareholders, notice of meeting, results of annual general meeting or extraordinary general meeting, half-yearly or preliminary final reports, or all other media releases in the ASX Signal G. Here, announcement date is specifically defined as the date where market first knows about the buy-back price and the composition of the buy-back price as well as the extent of the franking. Such condition is used because it is the primary purpose of this research to observe market reaction to the release of such information (if any). 28 This is based on the belief that a confounding event might have a significant effect on the company’s share price during the period of study and, therefore, might potentially distort the results. However, the contaminated events are also reported and the corresponding abnormal returns will also be analyzed to be compared with those of the uncontaminated events.

10

Based on these criteria, it was apparent that, of the initial sample, 20, 2 and 2 observations did not meet the first, second and third criteria respectively. These resulted in a final sample of 28 off-market buy-backs (see Table 1 for details). [Table 1 about here] (ii) Abnormal Return Generation Market Model (MM)31 and Mean Adjusted Return Model (MEAR)32 are employed to calculate the announcement period abnormal returns for each announcing company. Announcement period abnormal returns are examined for two event windows over days (-1,+1) and days (0,+1).33 In addition, short-term pre- and post- announcement period abnormal returns are also evaluated over days (-5,-2) and days (+2,+5) respectively. As Boehmer, Musumeci, & Poulsen (1991) suggested that the use of Standardized Cross Sectional t-tests should overcome such event-induced increase in variance by adjusting for cross sectional heteroskedasticity; hence, such tests are employed to observe the significance of the abnormal returns in this study (see for example, Balachandran et al.(1999)). Considering that we only have a small sample of off-market buy-backs and

29

For comparison, see for example, Brown & Warner (1985) used estimation period (-244,-6), Harris & Ramsay (1995) used estimation period (-250,-6), Lamba & Ramsay (1999) used estimation period (-270,31) and Otchere & Ross (2002) used estimation period (-100, -1). 30 Brown & Warner (1985), Harris & Ramsay (1995) and Otchere & Ross (2002) used examination period (-5,+5) for the daily returns. Even though Lamba & Ramsay (1999) used examination period (-20,+20), they found that abnormal returns are significant only in days (-5, +5). 31 Market Model is commonly used in event studies because it also incorporates market as a proxy to determine the abnormal returns resulted from an announcement or an event (see for example, Harris & Ramsay (1995) and Otchere & Ross (2002)). As it considers market-wide movement in the same time frame, the relative strength of this model under a wide variety of conditions is alleged (Brown & Warner, 1985). 32 Mean Adjusted Return Model (MEAR) is relatively simple compared to the Market Model. Nevertheless, the strength of this model is also evident as similar results to the more sophisticated Market Model are frequently observed (Brown & Warner (1985) and MacKinlay (1997)). 33 The inclusion of day -1 and day +1 is due to the possibility that the returns on the announcement date (day 0) do not capture the information effects of the announcements. Lamba & Ramsay (1999) and Otchere & Ross (2002) for example, found abnormal returns on day +1. This happened because some companies announced buy-backs near the market closing time, so that the announcement effects occurred on the day +1. On the other hand, day –1 is also included in the announcement period to accommodate leakages that might happen before the announcement (see for example, Lamba & Ramsay (1999) and Balachandran, Cadle, & Theobald (1999)). MacKinlay (1997) considers this expansion of the announcement date to be a common practice in a research.

11

that the returns might not be normally distributed, this study also employs non-parametric tests in addition to the parametric tests used.

(iii) Cross-Sectional Analysis A cross-sectional (multivariate) regression model is used to examine a range of variables that are (cross-sectionally) attributed to the degree of abnormal returns observed in the off-market buy-backs. Announcement period abnormal returns employing Market Model over day 0 to day +1 (MM01) are used as the explained/dependent variable. The set of the potential explanatory/independent and dummy variables as well as the anticipated signs (in parentheses) of their coefficients and the rationales for their inclusion in the regression are outlined below. FC (+) FC is the amount of franking credit distributed as a percentage of the total off-market buy-back price. The rationale for the inclusion of this explanatory variable is derived from the ‘imputation credit’ hypothesis which suggests that market does value franking credits positively. Hence, it is expected that the predicted sign of this variable’s coefficient would be positive. FRAC (-) FRAC is the amount of shares intended to be bought back in the off-market buy-back as a percentage of the number of shares outstanding. Large fraction of shares to be acquired could be interpreted as the inability of the company to manage its capital and also the unavailability of profitable investment opportunities for the company. Consequently, the predicted sign of this variable’s coefficient would be negative.34 LMV (-)

34

This variable is going to be used to test the ‘free cash flow’ hypothesis. Under the ‘free cash flow’ hypothesis, the bigger the size of the buy-back the lower the agency problems and hence the higher the announcement period abnormal returns.

12

LMV is the natural logarithm of the company’s market value four weeks prior to the announcement date. This variable is included in the regression model because higher returns received by shareholders in small companies undertaking buy-backs as opposed to those in large companies have been reported by many researchers, such as by Lakonishok & Vermaelen (1990) and Ikenberry, Lakonishok, & Vermaelen (1995). Consistent with ‘information signalling’ hypothesis, the predicted sign of this variable’s coefficient would be negative. DY(-) DY is dividend yields on the company’s shares at the balance sheet date immediately prior to the off-market buy-back announcement. Considering the tax profile of the Australians and the ‘dividend surprise’ argument by Dhillon, Raman, & Ramirez (2003)35, it is predicted that the sign of this variable’s coefficient would be negative. PE (+) PE is price earnings ratio of the company at the balance sheet date immediately prior to the off-market buy-back announcement. Since high PE means that it takes a longer time for the shareholders to ‘break-even’, hence it is expected that market reaction to buy-back announcement will be higher for companies with high PEs as they will be able to ‘breakeven’ sooner. Higher PEs could also be interpreted as higher growth; thus, consistent with the ‘information signalling’ hypothesis, higher reaction is also expected. Hence, the predicted sign of this variable’s coefficient would be positive. DIR (+) DIR is the percentage of the directors’ interests in the company at one balance sheet date immediately prior to the off-market buy-back announcement.36 Substantial management ownership within a company might induce management to distribute more benefits to 35

They argue that it is the ‘dividend surprises’ that result in stock price reactions. This is the relevant interests of the directors in the share capital of the company as at the date of the annual report, as notified by the directors to the ASX pursuant to the provision of section 205G(1) of the Corporations Law. This includes interests which are beneficially held in own name, or in the name of a trust, nominee company or private company. The inclusion of the later three is due to the fact that franked dividend received indirectly through resident trusts or resident partnerships are generally taxed in the same way as if they were received directly by the beneficiaries or partners (Handley & Maheswaran, 2003).

36

13

them by distributing more fully franked dividends as opposed to capital gains in the offmarket buy-backs.37 This argument is based on Vermaelen (1981)’s finding that there is positive relation between buy-back premiums offered with the managers’ ownership proportion in his US study. As more premiums are to be distributed thus higher market reaction is expected. Thus, consistent with the ‘personal taxation’ hypothesis, the predicted sign of this variable’s coefficient would be positive. DFIRST (+) DFIRST is a dummy variable to substitute DEXT, which takes a value of unity if the offmarket buy-back announcement is the firm’s first off-market buy-back program and zero otherwise. It is expected that firm’s first buy-back program conveys stronger information than the subsequent one. Hence, the predicted sign of this variable’s coefficient would be positive. DCONT (+/-) DCONT is a dummy variable, which takes a value of unity if the off-market buy-back announcement was contaminated with other market sensitive information within five days before to five days after the announcement and zero otherwise. As it is believed that confounding events might affect the results, this variable’s coefficient is expected not to be zero. 4. RESULTS (i) Abnormal Price Reactions Panel A of Table 2 reports the announcement period abnormal returns of the off-market buy-back announcements generated using Market Model (hereafter MM) and Mean Adjusted Return Model (hereafter MEAR). In general, we found significant positive abnormal returns were found and they are relatively similar for both MM and MEAR. 37

There is one obvious difference between Australian equal access and US tender-offer. Whilst there is common preclusion of management participation in the tender-offer terms to reduce the self-dealing behaviour, Australian equal access provisions are believed to be able to reduce such behaviour thus allowing the participation of the management (Rosa et al., 2000). Consequently, there is no warranty that there will be no self-dealing activities in Australia.

14

[Table 2 about here] As can be seen from Panel A of Table 2, the average abnormal returns for off-market buy-back announcements are 3.692% from day 0 to day 1 and 3.986% from day –1 to day 1 employing Market Model. All of the abnormal returns are significant at the 1% level using both the standardised cross sectional t-test (hereafter SCST) and the generalised sign test (hereafter GST). Mean Adjusted Return Model produces similar results. Panel A of Table 2 also provides separate results for both uncontaminated and contaminated events to allow control for the confounding events. For the uncontaminated events, the average abnormal returns are 3.798% from day 0 to day 1 and 3.631% from day –1 to day 1 employing Market Model, and all are significant at the 1% level using both SCST and GST. Meanwhile, for the contaminated events, the average abnormal returns are 2.157% from day 0 to day 1 and 3.149% from day –1 to day 1 employing Market Model, and all are significant at the 5% and 10% level using both SCST and GST.38 Even though t-test and Mann-Whitney test statistics show that there is no significant difference in the announcement period abnormal returns between uncontaminated and contaminated events39, it is believed that confounding events might distort our results. Hence, subsequent discussions will only focus on the uncontaminated events. In Panel A, we also found that there are no short-term pre- and postannouncement period abnormal returns over days (-5,-2) and (2,5) respectively following off-market buy-back announcements.40 Panel B of Table 2 shows that all of the announcement period abnormal returns for offmarket buy-backs with and without franking credit distribution are statistically significant 38

We excluded one particular outlier in the sample of the contaminated events due to its small sample of only 20 and that one outlier might drive the results into a different direction. Zingaro (1997, p.120), for example, stated that “the removal of outliers insures that the results are not solely the consequence of the presence of outliers.” Here, an outlier is defined as an observation whose value is outside the range of three standard deviation from the sample mean. Its exclusion should provide a more objective and clearer picture of the market reaction (see for example Partington & Walker (2001) for the need of filtering outliers). As a consequence, outliers will be excluded in the subsequent analyses. 39 Further check on the contamination issues will be discussed in the regression analysis and in the robustness check. 40 See uncontaminated events.

15

at the 1% level using SCST for both MM and MEAR, and most of them are statistically significant at the 1%, 5% and 10% level using GST.41 However, it is apparent that market reaction to off-market buy-back is greater for one without franking credit distribution as opposed to one with franking credits. Over day 0 to day 1 employing MM, for example, we found that the average abnormal returns for off-market buy-backs without franking credit distribution is 4.732%, significantly higher at the 10% level using t-test than the average abnormal returns of 2.553% for off-market buy-backs with franking credits.42 A similar result is found for MEAR day 0 to day 1. These results suggest that there is little evidence (if any) to support that there is negative relation between market reaction and the amount of franking credits distributed in the off-market buy-back. This is somewhat contradicts with the perceived importance of the value of franking credits as suggested by prior researchers and hence is not consistent with the hypothesis of higher market reaction for higher franking credit distribution in the off-market buy-backs. As a consequence, this issue will be further addressed in the cross-sectional analysis in subsection (ii). As can be seen from Panel C of Table 2, there is no significant difference between announcement period abnormal returns for firm’s first buy-back programs and any subsequent ones using both the parametric t-test and non-parametric Mann-Whitney test statistics. This result suggests that firm’s first off-market buy-back program does not convey stronger signals to the market than subsequent ones. The market will react positively to a buy-back announcement as long as investors believe that they can benefit. Further examination on market reaction to first and subsequent buy-back programs is conducted in the cross-sectional analysis in subsection (ii).

(ii) Cross-Sectional Analysis The results of the cross-sectional regression analysis of abnormal price reaction to offmarket buy-back announcements are presented in Table 3. There are 2 panels in this table: Panel A contains regression results for all events (both uncontaminated and 41 42

Except GST for MEAR day 0 to day 1 for off-market buy-backs without franking credit distribution. No significant difference is found using non-parametric Mann-Whitney test statistics.

16

contaminated events) to see the significance of the dummy variable DCONT; and Panel B contains regression results only for uncontaminated events.43 [Table 3 about here] There are three models in Panel A of Table 3.44 It seems that only PE (price earnings ratio) is statistically significant at the 1% level across all models in all events (both uncontaminated and contaminated events). All models are significant at the 5% level, with Adjusted R-square ranges from 15.28% to 18.30%. Other independent variables, such as FC (franking credit percentage), FRAC (fraction of shares to be acquired), LMV (the natural logarithm of firms’ market value), DY (dividend yields) and DIR (directors’ interests) were statistically insignificant. Interestingly, dummy variable DCONT is significant at the 10% level in model 1. This result, combined with the robustness check in subsection (iii), suggest that there is significance difference between uncontaminated and contaminated events in the off-market buy-back case. Therefore, analysis for all events might not be valid. As a consequence, this allows and justifies the use of only uncontaminated events in the subsequent regression analysis (see Panel B) and in the final discussion. Panel B of Table 3 consists of four models. From Panel B, it is apparent that all models are significant at least at the 5% level, with Adjusted R-square ranges from 25.12% to 43.76% across all models. The high Adjusted R-square shows that independent variables used in the regression models have high explanatory power for the announcement period abnormal returns observed in the off-market buy-backs. The explanations on how each independent variable contributes to the announcement period abnormal returns observed are outlined below. FC

43

These cross-sectional analyses for all and uncontaminated events in Panel A and B respectively are executed by removing observations that contain outlier from the regressions. See footnote 44 for details. 44 All models in this paper have been adjusted to the problem of high correlations among independent variables and hence we believe that they are appropriate to be used as the basis of our interpretation.

17

The amount of franking credit distributed as a percentage of the total off-market buyback price is found to be statistically significant at the 5% level in model 1. However, unlike what have been predicted, the coefficient of this variable is negative. This is consistent with results in subsection (i) of greater market reaction to off-market buybacks without franking credit distribution as opposed to ones with franking credit distribution. Further analysis by including only off-market buy-backs with franking credit distribution in the regression analysis also produces a similar result.45 Brown & Clarke (1993) contended that despite the fact that changes in the taxation of capital gains, dividends and superannuation funds increase the attractiveness of dividends relative to capital gains, it was apparent that Australian share market has continued to prefer returns in the form of capital gains to dividends. One argument for the unattractiveness of off-market buy-back with dividend distribution is related to the fact that irrespective whether or not shareholders acquire ‘actual’ profit on the overall transaction, they will still be liable to pay tax on the dividend component of the offmarket buy-back price.46 Besides, capital gains tax is considered to be lower than the tax that needs to be paid in the case of unfranked/very-low-franked dividends distributed in the off-market buy-back. These, combined with the introduction of capital gains tax reform effective 21 September 1999 (Review of Business Taxation (1999)) that diminishes the attractiveness of dividends relative to capital gains, might increase market preference toward capital gains. As a result, it could be argued that probably imputation credits are not the main driver for the announcement period abnormal returns observed in the off-market buy-back announcements. Support for this claim was provided by Balachandran & Nguyen (2003) when they found that franked special dividend announcements did not lead to a larger abnormal price reaction than unfranked special dividend announcements. For the three-day window from day –1 to day 1, they found that 45

With only nine observations and three independent variables FC, PE and DIR, the coefficients for FC, PE and DIR are –0.047275 [-4.561***], 0.000999 [2.901**] and –0.059924 [-3.935**] respectively. The tstatistics are reported in brackets [ ] and are based on the standard errors adjusted for heteroskedasticity using White’s (1980) heteroskedasticity-consistent covariance matrix. Note: *** significant at the 1% level, ** significant at the 5% level. Adjusted R-squared for the regression is 0.5952, whilst the F-statistic and its probability are 4.9211 and 0.0593 respectively. 46 Or, in the case of shareholder with marginal tax rate lower than the corporate tax rate, gain a refund via the excess credits.

18

the average abnormal returns employing Market Model is 3.68% for fully franked special dividend announcements and 4.01% for unfranked special dividend announcements, even though there is no statistically different reaction was found between the two groups. Plausible explanation for the phenomenon outlined above could also be related to the size of the companies announcing off-market buy-backs. It is apparent in the sample that the bigger the firm the higher the percentage of franking credit distributed.47 As market reaction is lower for big size firms with low information asymmetry (see explanation under LMV), then a negative relation between the percentages of franking credit distributed with the announcement period abnormal returns could be expected. Other plausible explanation is the structure of the announcing firm’s clienteles. Considering that the sample contains only 28 uncontaminated off-market buy-back announcements announced by 23 companies, it is possible that coincidentally a majority of these firms have more shareholders with high marginal tax rate and a preference for capital gains over dividend income, or their clienteles consist of shareholders with preference for future growth hence retain earnings. The composition of resident versus non-resident investors might also explain the results, as franked dividend component is likely to attract resident investors as opposed to non-resident investors. Small sample size also creates problem in the sense that the results might be driven by few observations. Hence, the regression results should be interpreted with cautious. Of particular importance in the observed results is that the ‘45-day rule’ needs to be satisfied by an individual in order to qualify for any franked dividend benefits distributed via the off-market buy-backs. The ‘45-day rule’ basically requires that ordinary shares be held for a period of 45 days at risk (excluding the days of acquisition and disposal) within a period beginning on the date those shares were acquired and ending 45 days after the

47

When natural logarithm of firm’s market value was regressed to the percentage of franking credit distributed, it is found that the coefficient of LMV is 0.071688 [6.093***]. The t-statistics are reported in brackets [ ] and are based on the standard errors adjusted for heteroskedasticity using White’s (1980) heteroskedasticity-consistent covariance matrix. Note: *** significant at the 1% level. Adjusted R-squared for the regression is 0.5329, whilst the F-statistic and its probability are 29.5216 and 0.000014 respectively.

19

shares become ex-dividend.48 The same requirement effectively applies to corporate shareholders. Shareholders with shorter holding periods lack the same opportunity to exploit imputation credit benefits. Consequently, this might affect market reaction to the off-market buy-back announcements and therefore could partly explain the results of this study. Another explantion of these results is the market learning process. Mitchell, Dharmawan, & Clarke (2001), in their survey, found that Australian managers believed that while they are familiar with all matters related to the buy-backs, their shareholders are not. This finding is accentuated by Bowerman (2003) when he said that Telstra buy-back offer is complex, complicated and foreign to most retail investors but not for institutions which are used to the tender process. The complexity of the offer is especially due to the fact that the benefits may differ under different tax positions, such as the marginal tax rate and also the availability of capital gains to effectively use the capital loss created by the buyback (Bowerman, 2003). Considering that only 48 off-market buy-backs have been undertaken since 1989, the Australian market, in general, may still not be familiar with the full tax ramifications of this practice. As a consequence, market might react to offmarket buy-back announcement even though no franking benefits attached in the buyback. Finally, the fact that there are several announcements regarding off-market buy-backs, especially given the inconsistency in the type, contents and order of the announcements, complicates the findings of this study. As this research in particular wants to observe market reaction to the distribution of franking credits in the off-market buy-back, announcement date in the off-market buy-back is specifically defined as the date where market first knows about the buy-back price and the composition of the buy-back price as well as the extent of the franking. However, there are problem with this date determination as market might also react to the off-market buy-back announcement even though the buy-back price or franking level information has not been released. Merrill

48

The ‘45-day rule’ works on a ‘last-in-first-out’ (LIFO) basis so that shareholders will be deemed to have disposed of their most recently acquired shares for the purpose of applying the ‘45-day rule’.

20

Lynch, for example, commented that even though full details of Telstra’s off-market buyback announced on 3 August 2003 had not been released, the announcement had been well received (Nicholas, 2003). As a consequence, market reaction to off-market buyback announcement with complete information and market reaction to several releases of information are expected to be different. This might also affect the results of this study. FRAC The fraction of shares intended to be bought back in the off-market buy-back is found to be statistically insignificant in model 2 and model 4. This indicates that there is not enough evidence to support that the size of the buy-back affects market reaction to the off-market buy-back announcements, or in other words, significant positive abnormal returns could be acquired by announcing companies irrespective of the size of the buyback.49 Plausible explanation for this result is that a large buy-back could be both interpreted as a good and a bad sign. As good sign, it might solve agency problems for a company with excess cash flows. As a bad sign, it might reflect the management’s inability to invest these funds efficiently into a profitable investment and so the funds were returned to the shareholders. Hence, market reaction will be different across companies with different management teams and different financial positions. Further elaboration on announcing firms’ characteristics will be addressed in the following explanations for other independent variables. LMV Model 3 shows that firm’s size affects market reaction to the announcement period abnormal returns of the off-market buy-back announcements. As can be seen in the above models, the coefficient of firm size, as measured by the natural logarithm of the company’s market value four weeks prior to the announcement date, is negative and statistically significant at least at the 5% level. This result indicates that off-market buyback announced by small company triggers higher market reaction than for a larger company, and is consistent with prior studies such as by Lakonishok & Vermaelen (1990) and Ikenberry et al.(1995). 49

Hence, no support was provided for the ‘free cash flow’ hypothesis.

21

An explanation for this phenomenon comes from the ‘information signalling’ hypothesis which argues that smaller firms convey stronger information because they release less information to the market and are not extensively analysed and reviewed by financial analysts. Hence, given higher information asymmetry for smaller companies, their price reaction is expected to be larger. Large companies, on the other hand, are very closely scrutinised and followed by analysts and investors generally. The market, therefore, can well anticipate their buy-back announcements, which subsequently leads to smaller price reaction. DY The coefficient of dividend yields on the company’s shares at the balance sheet date immediately prior to the off-market buy-back announcement is negative and statistically significant at the 10% level in model 2, 3 and 4. This is consistent with the prediction that firms with low dividend yields announcing dividend distribution in the off-market buyback will experience stronger reaction than ones with higher dividend yields. The explanations for this result are related to the ‘tax-induced clientele’ hypothesis. Under this hypothesis, it is expected that companies with high dividend yields have clienteles that comprise of shareholders with low marginal tax rates or who favour of high dividend distribution. As a result, they will not react too strong to dividend distribution in the offmarket buy-backs as they are used to it50, or neither will they react strong to the capital gains distribution in the off-market without franked dividend distribution. On the other hand, companies with low dividend yields possibly have more shareholders with higher marginal tax rate who would prefer capital gains than dividends. As 57% of the offmarket buy-back sample in this study includes no franking credit distribution, this means that it is more attractive for shareholders with high marginal tax rate to accept the offer. All of the above explanations lead to the observed results.

50

Recall ‘dividend surprise’ argument by Dhillon, Raman, & Ramirez (2003).

22

PE Price earnings ratio is found to be statistically significant at the 1% level across all models. The positive coefficient of this variable implies that the higher the price to earnings ratio of the announcing firms the higher the reaction of the market to the offmarket buy-back announcements. As a high PE means that it takes longer for shareholders to ‘break-even’, the market reaction to buy-back announcement should be higher for companies with high PEs as they will be able to ‘break-even’ sooner. Higher PE could also be interpreted as higher growth, thus higher reaction is also expected. InvestorWeb senior investment analyst George Galanopoulos, for instance, noted that share buy-backs usually only add shareholder value in companies with sound business fundamentals and sturdy growth prospect (Nicholas, 2003). As a consequent, the result for this variable could be well justified. DIR All models also show that the percentage of the directors’ shareholding in the company at one balance sheet date immediately prior to the off-market buy-back announcement is statistically significant at least at the 10% level. However, negative coefficient of this variable contradicts to the expectation that is based on the ‘personal taxation’ hypothesis. As previously argued that substantial management ownership within a company might induce management to distribute more fully franked dividends as opposed to capital gains in the off-market buy-backs, higher market reaction is expected for firms with higher management ownership. The fact that negative coefficient is observed therefore could not be explained with the ‘personal taxation’ hypothesis, but perhaps could be with an ‘agency costs’ rationalization. Under this theory, the higher the directors’ interests the higher the agency problems that might exist and market might have negative perception in the sense that directors may only want to increase their percentage of ownership in the companies by conducting buy-backs. Therefore, lower abnormal returns following offmarket buy-back announcements could be expected. The negative coefficient of DIR might also be explained with the ‘information signalling’ hypothesis, which suggests that the higher directors’ interests, the lower the information asymmetry between insiders and

23

shareholders. Consequently, lower abnormal returns following off-market buy-back announcements could also be expected using the last rationale. DFIRST Model 4 reveals that dummy variable DFIRST’s coefficient is not statistically significant. Hence, it could be inferred that there is no significant difference in the announcement period abnormal returns between firm’s first off-market buy-back program and the subsequent one. This finding is similar to Otchere & Ross (2002)’s finding when they found that there is no significant result to support that the first share buy-back announcement in the industry or in a buy-back program conveys more information than the subsequent announcements. Therefore, it is concluded that it is not the sequence of the buy-back that affects price reaction. Market will react positively to a buy-back announcement, irrespective of whether the buy-back is an initial or a subsequent one, as long as benefits from such scheme could be obtained.

(iii) Robustness Check The primary purpose of this task is to provide justification for the use of uncontaminated events only in the regression in subsection (ii). In this section, further robustness checks for the results reported in the previous section are executed by converting the crosssectional regression model into an extended cross-sectional regression model employing an interactive dummy specification. Specifically, one pair of dummy variables to form one set of interaction term (dummy x variable) in the extended cross-sectional regression model is defined, i.e. DCONT (for contaminated events) and DUNCONT (for uncontaminated events), where DUNCONT = 1 - DCONT. The results of the extended cross-sectional regression analysis of abnormal price reaction to off-market buy-back announcements are presented in Table 4. There are 2 panels in this table. Panel A contains the coefficients of the regression for the uncontaminated

24

events whilst Panel B reports the coefficients of the regression for the contaminated events.51 [Table 4 about here] There are four models in this robustness check. From Table 4, it could be seen that all models are significant at least at the 5% level, with Adjusted R-square ranges from 21.38% to 43.79%. In short, from model 1 to model 4 in Panel A of Table 4, it can be seen that, for DUNCONT group, variables FC, LMV, DY and DIR are statistically significant and negatively related to the announcement period abnormal returns. Whilst variable PE is found to be significantly and positively related to the announcement period abnormal returns, no statistically significant relations are found between announcement period abnormal returns and variable FRAC and between announcement period abnormal returns and dummy variable DFIRST. On the other hand, for DCONT group in Panel B, model 1 to model 4 show that variables LMV, DY, DIR and dummy variable DFIRST are statistically significant and positively related to the announcement period abnormal returns.52 Whilst no significances are found for variables FC and PE, there are mixed results for variable FRAC. Overall, evidence reported in this section strongly confirms that the presence of confounding events distort the results of the study. Hence, this further enhances the robustness of the results reported in subsection (ii). 5. CONCLUSION This study examines market reaction to the announcements of on-market and off-market buy-backs in Australia over the period from June 1989 to October 2003. In general, significant positive announcement period abnormal returns following off-market buyback announcements are documented in this study.

51

These cross-sectional analyses in Panel A and B respectively are executed by removing observations that contain outlier from the regressions. See footnote 44 for details. 52 Except variable DIR in model 1 is insignificant.

25

Surprisingly, we found that there is negative relation between price reaction and franking credit distribution in the off-market buy-backs. This is somewhat contradicts with the perceived importance of the value of franking credits and hence is not consistent with the hypothesis of higher market reaction for higher franking credit distribution in the offmarket buy-backs. Thus, there would appear no advantage for a company to conduct an off-market buy-back with high franked dividend component. In a cross-sectional analysis, the magnitude of price reaction to off-market buy-back announcements is found to be positively related to firm’s price earnings ratio, and negatively related to the firm’s size, dividend yield and the directors’ interests. However, we found no significant relations for variable buy-back size. There is also not enough evidence to support that the announcements of firms’ first off-market buy-back programs lead to a stronger price reaction than announcements of the subsequent ones. Finally, these results might be driven by the small sample size of off-market buy-back announcements; hence, the result of this study should be prudently interpreted.

26

Table 1: Summary of Off-Market Buy-Back Announcements 1989 – 2003 (Classified by Financial Year) Financial Year 1989 – June 1990 July 1990 – June 1991 July 1991 – June 1992 July 1992 – June 1993 July 1993 – June 1994 July 1994 – June 1995 July 1995 – June 1996 July 1996 – June 1997 July 1997 – June 1998 July 1998 – June 1999 July 1999 – June 2000 July 2000 – June 2001 July 2001 – June 2002 July 2002 – Oct 2003 Total

Number 2 0 0 0 0 0 1 4 4 4 2 6 3 2 28

27

Table 2: Abnormal Price Reaction to Off-Market Buy-Back Announcements This table reports mean and median abnormal returns, the Generalised Sign Test and the Standardised Cross-Sectional t-test employing Market Model and Mean Adjusted Return Model for (a) all, uncontaminated and contaminated off-market buy-back announcements for the announcement periods over day 0 to day 1 and over day –1 to day 1 as well as for the short-term pre- and post- announcement period abnormal returns over day -5 to day -2 and over day 2 to day 5 in Panel A, (b) uncontaminated off-market buy-backs with and without franking credit distribution for the announcement periods over day 0 to day 1 and over day -1 to day 1 in Panel B, and (c) first and subsequent announcements for the announcement periods over day 0 to day 1 and over day –1 to day 1 in Panel C. It also provides t-test and non-parametric Mann-Whitney test statistics for the difference in mean abnormal returns in each panel.

Panel A: Uncontaminated versus Contaminated Events Day 0 to day 1

MM

MEAR

Day -1 to day 1

MM

MEAR

Day -5 to day -2

MM

MEAR

Day 2 to day 5

MM

MEAR

Sample size

MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST

All 3.692 2.485 (5.87)*** (4.84)*** 3.481 2.666 (5.48)*** (3.99)*** 3.986 3.146 (5.62)*** (3.97)*** 3.894 2.985 (5.41)*** (3.99)*** 0.840 0.061 0.48 0.22 0.824 -0.120 0.51 -0.34 -1.107 -1.066 -1.76* -2.67** -1.069 -1.148 -1.65 -1.50

Uncont 3.798 2.466 (5.67)*** (4.12)*** 3.506 2.697 (5.01)*** (3.06)*** 3.631 2.750 (5.55)*** (3.36)*** 3.457 2.946 (5.02)*** (3.06)*** 1.058 0.081 0.26 0.33 1.076 -0.129 0.21 0.03 -0.538 -0.936 -0.69 -1.56 -0.538 -0.817 -0.66 -0.73

Cont 2.157 2.447 (2.53)** (2.35)** 2.059 2.609 (2.49)** (2.26)** 3.149 3.210 (2.61)** (1.89)* 3.177 2.751 (2.64)** (2.26)** 0.565 0.138 0.50 0.05 0.500 -0.135 0.63 -0.50 -1.648 -1.010 -2.16** -2.25** -1.552 -1.443 -1.96* -1.42

48

28

19

t-test 1.33

MW -0.54

1.17

-0.54

0.37

-0.11

0.22

-0.09

0.34

-0.13

0.40

-0.07

0.81

-0.52

0.74

-0.46

28

Panel B: With versus Without Franking Credit Distribution Day 0 to day 1

MM

MEAR

Day -1 to day 1

MM

MEAR

MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST

Sample Size

With 2.553 2.094 4.46*** 3.10*** 2.305 2.210 4.39*** 2.88** 2.588 2.390 4.14*** 2.52** 2.460 2.392 3.87*** 2.29** 12

Without 4.732 2.685 3.82*** 2.77** 4.407 2.963 3.30*** 1.58 4.413 3.543 3.78*** 2.27** 4.205 3.946 3.36*** 2.08* 16

t-test53 1.38*

MW -0.84

1.33*

-0.51

1.18

-0.88

1.15

-0.88

Panel C: First versus Subsequent Events Day 0 to day 1

MM

MEAR

Day -1 to day 1

MM

MEAR

Sample Size Note: *** ** * SCST GST MM MEAR MW

53

MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST MEAN (%) MEDIAN (%) SCST GST

First 4.093 2.523 4.77*** 3.49*** 3.754 2.806 4.09*** 2.41** 4.079 3.182 4.85*** 3.05*** 3.877 3.839 4.34*** 2.41** 21

Significantly different from zero at the 1% level Significantly different from zero at the 5% level Significantly different from zero at the 10% level Standardised Cross-Sectional t-test Generalised Sign Test Market Model Mean Adjusted Return Model Mann-Whitney test statistics

Subsequent 2.914 2.062 3.03*** 2.21** 2.763 2.547 3.22*** 2.04** 2.285 2.259 2.76*** 1.44 2.199 1.927 3.05*** 2.04** 7

All Uncont Cont With Without First Subsequent

t-test -0.64

MW 0.40

-0.53

0.29

-1.01

1.09

-0.96

0.98

All Events Uncontaminated Events Contaminated Events With franking credit distribution Without franking credit distribution Firms’ first off-market buy-backs Subsequent off-market buy-backs

Based on one-sided t-test.

29

Table 3: Cross-sectional Regression Analysis of Abnormal Price Reaction to OffMarket Buy-Back Announcements This table provides cross-sectional regression results explaining market response to off-market buy-back announcements. The dependent variable used in this regression is the two-day abnormal return from the announcement date to the day after the announcement date employing Market Model (MM01). Independent variables that are used consist of FC - the amount of franking credit distributed as a percentage of the buyback price, FRAC – fraction of shares intended to be bought back, LMV - the natural logarithm of the company’s market value four weeks prior to the announcement date, DY - dividend yields on the company’s shares at the balance sheet date immediately prior to the off-market buy-back announcement, PE - price earnings ratio of the company at the balance sheet date immediately prior to the off-market buyback announcement, DIR - the percentage of the directors’ interests in the company, dummy variable DFIRST which takes a value of unity if the off-market buy-back announcement is the firm’s first offmarket buy-back program and zero otherwise, and dummy variable DCONT which takes a value of unity if the off-market buy-back announcement was contaminated with other market sensitive information within five days before to five days after the announcement and zero otherwise. The t-statistics are reported in brackets [ ] and are based on the standard errors adjusted for heteroskedasticity using White’s (1980) heteroskedasticity-consistent covariance matrix. There are 2 panels in this section. Panel A contains regression results for all events (both uncontaminated and contaminated events), whilst Panel B reports regression results for uncontaminated events only.

Panel A: All Events – Dependent variables: MM01 Independent variables FC

Expected Sign +

FRAC

-

LMV

-

DY

-

PE

+

DIR

+

DCONT Constant Adjusted R2 F-statistic Prob (F-stat) Sample Size

+/-

Model 1 -0.006989 [-0.42]

Model 2

Model 3

0.021010 [1.49]

0.001286 [3.63]*** 0.008320 [0.17] -0.022066 [-1.99]* 0.021039 [1.75]* 0.1553 2.7930 0.0411 40

-0.192734 [-0.94] 0.001399 [3.78]*** 0.000572 [0.01] -0.016305 [-1.53] 0.016902 [1.19] 0.1830 2.8373 0.0293 42

-0.000743 [-0.36] -0.200064 [-0.97] 0.001321 [3.73]*** -0.005753 [-0.10] -0.018366 [-1.61] 0.030258 [1.30] 0.1528 2.4792 0.0499 42

30

Panel B: Uncontaminated Events – Dependent variables: MM01 Independent variables FC

Expected Sign +

FRAC

-

LMV

-

DY

-

PE

+

DIR

+

DFIRST

+

Constant Adjusted R2 F-statistic Prob (F-stat) Sample Size Note: *** ** *

Model 1

Model 2

Model 3

Model 4

-0.036797 [-2.29]** 0.023967 [1.56]

0.018335 [1.19]

0.001163 [3.59]*** -0.045246 [-1.84]*

-0.404261 [-1.84]* 0.001261 [5.34]*** -0.059149 [-2.85]**

-0.004262 [-2.72]** -0.372174 [-1.75]* 0.001391 [5.77]*** -0.076503 [-2.89]***

0.031924 [2.37]** 0.2512 3.3486 0.0422 22

0.029637 [2.68]** 0.4063 4.9354 0.0067 24

0.060405 [3.23]*** 0.4376 5.4741 0.0042 24

-0.472529 [-1.84]* 0.001241 [6.00]*** -0.072393 [-2.48]** 0.012931 [1.21] 0.025749 [2.93]*** 0.3989 4.0520 0.0122 24

Significantly different from zero at the 1% level Significantly different from zero at the 5% level Significantly different from zero at the 10% level

31

Table 4: Robustness Check - Uncontaminated versus Contaminated Events This table provides extended cross-sectional regression results explaining market response to the offmarket buy-back announcements by using interactive dummy variables DCONT (takes a value of unity if the equal buy-back announcement is contaminated and zero otherwise) and DUNCONT (DUNCONT=1DCONT). The dependent variable used in this regression is the two-day abnormal return from the announcement date to the day after the announcement date employing Market Model (MM01). Independent variables that are used consist of FC - the amount of franking credit distributed as a percentage of the buyback price, FRAC - fraction of shares intended to be bought back, LMV - the natural logarithm of the company’s market value four weeks prior to the announcement date, DY - dividend yields on the company’s shares at the balance sheet date immediately prior to the off-market buy-back announcement, PE - price earnings ratio of the company at the balance sheet date immediately prior to the off-market buyback announcement, DIR - the percentage of the directors’ interests in the company at one balance sheet date immediately prior to the off-market buy-back announcement, and dummy variable DFIRST which takes a value of unity if the off-market buy-back was the firm’s first off-market buy-back program and zero otherwise. The t-statistics are reported in brackets [ ] and are based on the standard errors adjusted for heteroskedasticity using White’s (1980) heteroskedasticity-consistent covariance matrix. There are 2 panels in this section. Panel A contains the coefficients of the regression for the uncontaminated events, whilst Panel B reports the coefficients of the regression for the contaminated events.

Panel A: Uncontaminated Events (DUNCONT=1-DCONT) Independent variables FC

Expt. Sign +

FRAC

-

LMV

-

DY

-

PE

+

DIR

+

DFIRST

+

Constant

Model 1 -0.036797 [-2.26]**

Model 2

Model 3

0.023967 [1.53]

Model 4 0.018335 [1.16]

0.001163 [3.55]*** -0.045246 [-1.82]*

-0.404261 [-1.80]* 0.001261 [5.24]*** -0.059149 [-2.80]***

-0.004262 [-2.67]** -0.372174 [-1.71]* 0.001391 [5.66]*** -0.076503 [-2.84]***

0.031924 [2.34]**

0.029637 [2.63]**

0.060405 [3.17]***

-0.472529 [-1.80]* 0.001241 [5.86]*** -0.072393 [-2.42]** 0.012931 [1.19] 0.025749 [2.86]***

32

Panel B: Contaminated Events (DCONT) Independent variables FC

Expt. Sign +

FRAC

-

LMV

-

DY

-

PE

+

DIR

+

DFIRST

+

Constant Adjusted R2 F-statistic Sample Size Note: *** ** *

Model 1 0.038857 [1.37]

Model 2

Model 3

-0.043297 [-1.35]

Model 4 -0.071928 [-2.43]**

0.000651 [0.58] 0.078471 [1.31]

0.836922 [2.14]** 0.001001 [0.99] 0.137061 [1.83]*

0.009824 [3.00]*** 0.652273 [2.30]** 0.000512 [1.05] 0.193601 [2.67]**

-0.007111 [-0.34] 0.2138 3.0262** 40

-0.036537 [-2.00]* 0.3286 3.7464*** 42

-0.100086 [-5.40]*** 0.4379 5.2780*** 42

0.789520 [2.08]** -3.90E-05 [-0.04] 0.150940 [2.89]*** 0.033737 [2.13]** -0.029831 [-1.85]* 0.3819 3.7549*** 42

Significantly different from zero at the 1% level Significantly different from zero at the 5% level Significantly different from zero at the 10% level

33

REFERENCES ABS (1999). Financial Accounts - Listed Equity: Sources and Methods. Australian Bureau of Statistics (December 1999). Allen, F., Bernado, A., & Welch, I. (2000). The Theory of Dividends based on Tax Clienteles. Journal of Finance, 55(6), 2499-2536. Antioch, G. (2001). Franking Credit Trading. Australian Tax Forum, 16, 323-344. Ashton, D. (1989). The Cost of Capital and the Imputation Tax System. Journal of Business Finance and Accounting, 16(1), 75-88. ATO (2003). Tax Reform and the Individual - Introduction. Australian Taxation Office. http://www.taxreform.ato.gov.au/SiteHome.asp?placement=TR/IN&from=TR/IN. Accessed (10 May 2003). Balachandran, B., Cadle, J., & Theobald, M. (1999). Analysis of Price Reactions to Interim Dividend Announcements: A Note. Applied Financial Economics, 9(4), 305-314. Balachandran, B., Faff, R., & Troiano, R. (2002). New Evidence on the Market Impact of Share Buy-Back Announcements (Working Paper). Melbourne. Monash University Balachandran, B., & Nguyen, T. A. (2003). Signalling Power of Special Dividends in An Imputation Environment (Working Paper). Melbourne. Monash University (Accounting and Finance - forthcoming). Bellamy, D. (1994). Evidence of Imputation Clienteles in the Australian Equity Market. Asia Pacific Journal of Management, 11(2), 275-287. Boehmer, E., Musumeci, J., & Poulsen, A. (1991). Event Study Methodology Under Conditions of Event Induced Variance. Journal of Financial Economics, 30(2), 253-272. Bowerman, R. (2003). Telstra Buyback Best for Investors in Low Tax Brackets. The Sunday Age, 26 October. Brown, P., & Clarke, A. (1993). The Ex-Dividend Day Behaviour of Australian Share Prices Before and After Imputation. Australian Journal of Management, 18(1), 140. Brown, P., & Walter, T. (1986). Ex-Dividend Day Behaviour of Australian Share Prices. Australian Journal of Management, 11(2), 139-152. Brown, S. J., & Warner, J. B. (1985). Using Daily Stock Returns: The Case of Event Studies. Journal of Financial Economics, 14(1), 3-31. Bruckner, K., Dews, N., & White, D. (1994). Capturing Value from Dividend Imputation. McKinsey and Company Chan, K., McColough, D., & Skully, M. (1993). Australian Tax Changes and Dividend Reinvestment Plan Announcement Effects: A Pre- and Post- Imputation Study. Australian Journal of Management, 18(1), 41-62. Chay, J., & Marsden, A. (1996). Market Reaction to the Introduction of a Foreign Investor Tax Credit Regime in New Zealand. Pacific-Basin Finance Journal, 4(2), 129-152. Comment, R., & Jarrell, G. (1991). The Relative Signalling Power of Dutch-Auction and Fixed-Price Self Tender Offers and Open-Market Share Repurchases. Journal of Finance, 46(4), 1243-1271. 34

Dann, L. (1981). Common Stock Repurchases: An Analysis of Returns to Bondholders and Stockholders. Journal of Financial Economics, 9(2), 113-138. Dhillon, U., Raman, K., & Ramirez, G. (2003). Analysts' Dividend Forecasts and Dividend Signaling (Working Paper). USA. Binghamton University, Bentley College and Kennesaw State University Elton, E., & Gruber, M. (1970). Marginal Stockholder Tax Rates and the Clientele Effect. Review of Economics and Statistics, 52(1), 68-74. Fama, E. (1970). Efficient Capital Markets: A Review of Theory and Empirical Work. Journal of Finance, 25(2), 383-417. Handley, J., & Maheswaran, K. (2003). The Value of Australian Imputation Credits (Working Paper). Melbourne. University of Melbourne Harris, T., & Ramsay, I. (1995). An Empirical Investigation of Australian Share Buybacks. Australian Journal of Corporate Law, 4(4), 393-416. Hathaway, N., & Officer, R. (1992). The Value of Imputation Tax Credits (Working Paper). Melbourne. Graduate School of Management, University of Melbourne Hathaway, N., & Officer, R. (1996). The Value of Imputation Tax Credits (Working Paper). Melbourne. Melbourne Business School Howard, P., & Brown, R. (1992). Dividend Policy and Capital Structure under the Imputation Tax System: Some Clarifying Comments. Accounting and Finance, 32(1), 51-61. Ikenberry, D., Lakonishok, J., & Vermaelen, T. (1995). Market Underreaction to Open Market Share Repurchases. Journal of Financial Economics, 39(2-3), 181-208. Izan, H., Mitchell, J., & Lim, R. (2002). On-Market Buy-backs: An Examination of Valuation Issues (Working Paper). Nedlands. University of Western Australia Jeffrey, I. M. (1991). Taxation Implications of Share Buy-Backs. Charter, 62(11), 34-37. Lakonishok, J., & Vermaelen, T. (1990). Anomalous Price Behavior Around Repurchase Tender Offers. Journal of Finance, 45(2), 455-477. Lally, M. (1992). The CAPM under Dividend Imputation. Pacific Accounting Review, 4, 31-44. Lamba, A., & Ramsay, I. (1999). Share Buyback in A Highly Regulated and Less Regulated Market Environment (Working Paper). Melbourne. University of Melbourne Leo, K., & Hoggett, J. (1998). Company Accounting in Australia (4th ed.). Brisbane: John Wiley and Sons Australia, Ltd. MacKinlay, A. C. (1997). Event Studies in Economics and Finance. Journal of Economic Literature, 35(1), 13-39. Mitchell, J., Dharmawan, G., & Clarke, A. (2001). Managements' Views on Share BuyBacks: An Australian Survey. Accounting and Finance, 41(1/2), 93-129. Monkhouse, P. (1993). The Cost of Equity under the Australian Dividend Imputation System. Accounting and Finance, 33(2), 1-18. Nicholas, K. (2003). Investors Just Love Getting Cash Back. Australian Financial Review, 29 September. Officer, R. (1988). A Note on the Cost of Capital and Investment Evaluation for Companies Under the Imputation Tax. Accounting and Finance, 28(2), 65-71.

35

Otchere, I., & Ross, M. (2002). Do Share Buy Back Announcements Convey FirmSpecific or Industry-Wide Information? A Test of the Undervaluation Hypothesis. International Review of Financial Analysis, 11(4), 511-531. Partington, G., & Walker, S. (2001). A Note on Transaction Costs and the Interpretation of Dividend Drop-Off Ratios. Accounting and Finance, 41(3), 229-241. Peirson, G. (1990). Share Buy-Backs. CPA: Accounting Communique, 20, 1-4. Review of Business Taxation: A Platform for Consolidation Building on a Strong Foundation. Canberra: Dept of Treasury, 1999. Renton, N. (1998). Understanding the Stock Exchange: The Essential Reference (3rd ed.). Melbourne: Information Australia. Rosa, R. d. S., Lee, P., Preda, S., & Walter, T. (2000). Market Reaction to Share Buybacks and Special Dividends under A Tax Imputation System (Working Paper). Sydney. The University of Sydney Vermaelen, T. (1981). Common Stock Repurchases and Market Signalling: An Empirical Study. Journal of Financial Economics, 9(2), 139-183. Walker, S., & Partington, G. (1999). The Value of Dividends: Evidence from Cumdividend Trading in the Ex-dividend Period. Accounting and Finance, 39(3), 275296. White, H. (1980). A Heteroskedasticity-Consistent Covariance Matrix Estimator and a Direct Test for Heteroskedasticity. Econometrica, 48(4), 817-838. Wise, T., & Wise, V. (1987). Share Buy Back: A Post-Crash Possibility. The Chartered Accountant in Australia, 58(6), 26-28. Wood, J. (1997). A Simple Model for Pricing Imputation Tax Credits under Australia's Dividend Imputation Tax System. Pacific-Basin Finance Journal, 5(4), 465-480. Zingaro, L. (1997). Systematic and Firm-Specific Factors that Influence Corporate Cash Distribution Methods: A Qualitative Choice Study. Unpublished Dissertation, University of North Carolina, Chapel Hill.

36

Appendix 1 - Taxation on the Off-Market and On-Market Buy-Backs The taxation consequences of a share buy-back are outlined in the Division 16K of Part III (sec.159GZZZJ to sec.159GZZZT) of the Income Assessment Act 1936 (ITAA 1936). There are no income tax or CGT consequences from the buyer’s (i.e. company’s) point of view (sec.159GZZZN). However, from the seller’s (i.e. participating shareholder’s) point of view, the tax implications will differ significantly depending on: (1) whether the buyback is an “on-market” or an “off-market” buy-back and (2) the purchase price. The differences between “on-market” and “off-market” buy-backs are explained in sec.159GZZZK.54 Where the purchase is made in the ordinary course of trading on the official stock exchange then the purchase is considered to be an on-market buy-back, whereas any other buy-back is considered to be off-market buy-back.55 A buyer might use both its profits and its capital in determining the consideration for the purchase price. In the case of the off-market buy-back, the purchase price consists of the capital component and the dividend component.56 The capital component refers to the part of the purchase price which is debited against the seller’s share capital account, whilst the difference between the purchase price and the capital component is deemed to be the dividend component paid out of profits derived by the seller on the day the buyback occurs (sec.159GZZZP).57 These capital component and dividend component will be taxed based on the capital gains tax and general income tax accordingly. In the on-market buy-back, no part of the purchase price is treated as a dividend, even though the buyer may have used its profits to fund the buy-back: sec.159GZZZR. 54

see also sec.159GZZZL However, there is a case where an on-market transaction is deemed to be an off-market transaction. This is when the on-market transaction could be described under the rules of the stock exchange as “special”, i.e. where the price has been mutually agreed upon between the parties. 56 A company might undertake an off-market buy-back because it has franking credits greater than that can be used for normal dividend distribution. 57 This deemed dividend is a frankable distribution except to the extent that the purchase price exceeds the market value of the shares: sec.202-40 and 202-45. Where the purchase price is less than the market value of the shares, the consideration that is deemed to have been received by the seller is the market value of the shares at the time of the buy-back: sec.159GZZZQ(2). See also sec.159GZZZQ(3),(4). 55

37

However, if this is the case, the buyer may still be required to debit its franking account according to sec.205-30.58 In the on-market buy-backs, only capital gains tax applied to the proceeds of the buy-back. Summary of the taxation on the off-market and on-market buy-backs Type

Source of fund

Treatment

Off-market

Share capital account Non-share capital account

Capital component Dividend component

Share capital account Non-share capital account

Capital component Capital component

On-market

Tax applied to Buyer’s seller account CGT -

franking

General income tax + franking credit (if franked) CGT

Debited as in the dividend distribution (if franked) -

CGT

Debited as in the offmarket buy-back

58

The debit arises on the day of the purchase and is equal to the debit that would have arisen if the purchase were off-market (at the buyer’s benchmark franking percentage or at 100% franking percentage).

38