The gift that keeps on giving: Stock returns around CEO stock gifts to family members. Jennifer L. Brown* Associate Professor

Transcription

1 The gift that keeps on giving: Stock returns around CEO stock gifts to family members Jennifer L. Brown* Associate Professor G. Ryan Huston Assistant Professor Brian S. Wenzel Doctoral Student Arizona State University *W. P. Carey School of Business School of Accountancy Arizona State University PO Box Tempe, AZ Phone: Fax: January 2017 We would like to thank workshop participants at the EIASM Conference on Current Research in Taxation, Arizona State University and Indiana University, as well as the Tax Readings Group at the University of Texas for their helpful comments. In addition, we greatly appreciate feedback from Shannon Chen, Alex Edwards, Leslie Hodder, Artur Hugon, Janet Huston, Sonja Rego, and Jaron Wilde. Brian Wenzel would personally like to thank the Accounting Doctoral Scholars Program for financial support.

2 Abstract We examine CEOs stock dispositions for family gift purposes. CEOs prefer to time stock sales at relative price maximums, but should prefer to make family gifts at relative stock price minimums for estate and gift tax planning purposes. We demonstrate that returns leading up to dispositions are significantly lower for gifts relative to sales, and post-disclosure abnormal returns for gift dispositions are significantly positive. In shortwindow tests, we show that the market appears to miss the positive signal associated with post-disposition stock performance in the case of family gifts. JEL classification: G14, H24, K22 Keywords: Insider trading, estate & gift taxes, securities regulations 1

3 Bona fide gifts made by an Insider should be exempt [from trading prohibition under the Sarbanes-Oxley Act of 2002], as they are under Section 16. Gifts are made without the expectation of profits, and thus the policy reason behind Section 306 of the Act - to prevent Insiders from profiting at a time when rank and file employees cannot is not violated by the making of a gift. American Bar Association, Introduction A host of papers examine whether, and in which contexts, insider trades provide information to the market (e.g. Lakonishok and Lee, 2001; Jagolinzer, Larcker, and Taylor, 2011; Cohen, Malloy, and Pomorski, 2012). Insiders are in a unique position to act on private information, and prior studies provide evidence that managers strategically time their sales (Cheng, Nagar, and Rajan, 2007; Jagolinzer 2009). Company-initiated blackout policies and insider trading rules like Securities and Exchange Commission (SEC) Rule 10b-5 are designed to limit improprieties associated with informed insider trades (Bettis, Coles, and Lemmon, 2000), and evidence suggests that recent regulatory action has reduced insiders incentives to sell ahead of bad news (Brochet, 2010). While restrictions related to open-market insider sales (and open-market purchases) are clear and these restrictions have become more stringent over time, these limits generally do not apply to bona fide gifts (Sulcoski, 1989; Yermack, 2009). 1 In the large literature on insider trades, dispositions for family gifts are essentially ignored. We argue gift dispositions are interesting to study because, relative to sale dispositions, CEOs incentives and opportunities to time gift dispositions are distinct. In this paper, we examine whether the timing and pattern of CEOs stock dispositions for family gifts are consistent with estate tax planning incentives and whether CEO gift dispositions 1 SEC Rule 16b-5(a) provides that both the acquisition and disposition of equity securities shall be exempt from the operations of section 16 (b) of the Act if they are bona fide gifts Yermack (2009) notes that in proposing the amendment, the SEC staff wrote, Bona fide gifts present less likelihood for opportunities for abuse [compared to open market sales]. (p. 111) 2

4 are informative to the market. In prior work, Yermack (2009) examines CEOs dispositions for charitable donations, which like family gifts meet the reporting exception available for bona fide gifts. Managers incentives to time dispositions for charitable contributions mimic those for sale dispositions; in both cases CEOs receive the largest benefit (i.e. the most cash for open market sales or the largest tax deduction for charitable donations) when the disposition occurs at the highest stock price point. In contrast, CEOs should prefer to make stock gifts to family members in periods when the stock price is at a relative minimum. Estate and gift tax planning drives this incentive. Because the unified estate and gift tax rate is relatively high when compared to income tax rates (ranging between 40 and 55 percent during our sample period), gifting shares when the stock price is low, but expected to increase significantly in the future, allows CEOs to freeze their estate values. 2 This estate planning strategy effectively shifts future stock price appreciation from being taxed at the higher estate tax rate upon the donor s death to being taxed at the lower long-term capital gain rate (between 15 and 23.8 percent during our sample period) upon disposition by the donee. To determine whether these estate and gift tax incentives predictably impact CEOs gift dispositions, we begin our empirical tests by examining raw stock returns across varying time horizons both leading up to and following the disposition. We begin with raw stock returns to measure the CEO s family wealth benefit associated with a freeze strategy. We expect and subsequently find that the returns leading up to the disposition are significantly lower for gifts relative to sales; in addition, we find that post-disposition returns are significantly greater among 2 Although the appeal of the estate freeze strategy extends to high net-worth individuals more generally (Forbes.com, 2014), we examine gift dispositions of public company CEOs because these transactions represent a strong setting for finding predicted effects and examining whether the market reacts to the disclosure of this unique insider transaction. Specifically, we believe that CEOs have both enough family wealth to necessitate estate and gift tax planning as well as enough information about the future outlook of the firm to time gifts to family members at relative minimums. 3

5 gifts relative to sales. Specifically, mean raw returns for the year (two years) following gift dispositions are over 21 (36) percent, relative to 10 (25) percent for sales. Taken together, our results are generally in line with the incentives of an estate freezing strategy. After demonstrating the personal benefits associated with CEOs ability to time family gifts, we examine whether there is information content to the market associated with these dispositions. Specifically, we examine whether there are long-run positive market-adjusted returns following the disclosure of CEOs gift dispositions. We show that market-adjusted returns are statistically greater than zero in the one- and two-year periods following the SEC report disclosing gift dispositions. We find similar results analyzing buy-and-hold abnormal returns (BHARET). Finally, we calculate long-term annual returns using a Fama and French (1993) three-factor model with momentum (Carhart, 1997) and find positive alphas in both oneyear (0.43 percent) and two-year (0.36 percent) windows. Together these results suggest buying stocks when CEO gift transactions are disclosed can generate abnormal profits. To better understand whether gift transactions are associated with informed trading, we regress aggregated CEO dispositions at the firm-quarter level on whether EPS in the quarter of the disposition and the following two quarters is greater than the analyst-forecasted amount, similar to the work of Brochet (2010). This methodology allows us to infer whether CEOs are acting on knowledge that is not reflected in analysts forecasts. Additionally, for gift dispositions our findings demonstrate that firms are less likely to meet or beat consensus analyst forecasts in the quarter of a gift disposition, whereas they are more likely to meet or beat earnings targets in the period of a sale disposition. These results suggest CEOs strategically time the gifting of stock using information the market has not yet impounded into price. 4

6 After determining that positive returns can be garnered by buying stocks following CEO gift dispositions, we next examine the short-window market reaction to SEC filings disclosing gift dispositions. Our results demonstrate that the market treats family gift dispositions as a nonevent based on returns that are insignificantly different from zero. The market appears to distinguish between sales and gift dispositions, reacting negatively to sales, but appears to miss the longer-term positive signal associated with the estate freeze tax planning strategies employed by CEOs in these transactions. Long reporting lags, the period between the gift disposition and SEC disclosure, represent a potential alternative explanation for finding an onaverage insignificant short-term market reaction. To the extent that disclosures following a long reporting lag no longer represent relevant information to the market, we may be more likely to find a statistically insignificant reaction to gift disclosures. When we re-examine short-window returns in the context of reporting lag, we continue to find no difference in market reaction for low versus high reporting lag, bolstering our conclusion that the market is missing a positive signal associated with gift dispositions. In supplemental tests, we re-examine our results in the context of regulation designed to lessen the impact of inside information on insider trades, SEC Rule 10b5-1 and the Sarbanes- Oxley Act of 2002 (SOX). Rule 10b5-1, enacted in October 2000, adds context to the original Rule 10b-5 by providing a legal defense for insiders who pre-plan stock trades to create uninformed diversification. Further, effective August 29, 2002, as part of SOX, the Securities and Exchange Commission (SEC) issued new regulations requiring corporate insiders to report their sale dispositions (but not gift dispositions) by filing Form 4 within two business days of completed transactions. Brochet (2010) finds that restrictions on insider stock transactions associated with SOX have lowered the likelihood of insiders selling shares ahead of privately 5

7 known negative news. Our results corroborate Brochet s (2010) findings with respect to stock sales; returns leading up to sales are significantly less positive in the post-regulation period. However, we do not observe a similar shift in the timing of CEO family gifts across the pre- and post-regulation periods, as these regulations do not apply to gift dispositions. While accelerated filing requirements are not mandatory for gift dispositions, our final set of results suggest that the market would benefit if accelerated filing for gifts were mandatory. We compare stock performance during the period of the reporting lag, which is captured by the CEO, to stock performance in the year following SEC disclosure, which is available to the overall market. We find that when reporting lag is greater than the sample median, CEOs make a much greater personal return during the reporting lag period, to the detriment of the overall market. Because we find very significant raw returns that are generated following gift dispositions and prior literature has demonstrated that backdating is inherent in a number of CEO transactions (Heron and Lie, 2007; Dhaliwal, Erickson, and Heitzman, 2009; Yermack, 2009), we attempt to determine the extent to which our results are driven by backdating. We follow the Dhaliwal, Erickson, and Heitzman (2009) methodology examining the likelihood that gift dispositions occur on the day with the lowest closing price during the month of the gift. Our results suggest that approximately six percent of gift transactions occur on the lowest price day (referred to as suspects). In comparison, the chance likelihood that a transaction would occur on the lowest trading day of the month is one in 21 (4.75%). This does not by itself provide evidence of backdating. However, when we re-examine our reporting lag returns comparing suspect transactions to non-suspects, we find that suspect CEOs reap significantly greater benefits in the reporting lag period. These results indicate that backdating is at least partially associated with the large raw returns generated by CEOs. In sensitivity analyses, we re-run our raw and market 6

8 returns analyses separately for suspect and non-suspect dispositions and find that our takeaways with respect to market timing and the viability of a trading strategy are not altered based on the impact of backdating. Finally, we examine the impact of backdating before and after the enactment of SOX given its ability to curtail backdating activity in prior literature, finding that our results are unchanged following the enactment of SOX. Combined with our results examining the impact of reporting lag, these results continue to support the notion that it would be appropriate for regulations to apply to gift transactions. While a great deal of literature in accounting and finance has examined the information content and managerial incentives associated with insiders sales (e.g., Lakonishok and Lee, 2001; Jagolizer, Larcker, and Taylor, 2011; Cohen, Malloy, and Pomorski, 2012) little work has been done to further our understanding of gifts to family members. Gifts are important transactions for market participants to understand because the incentives and restrictions CEOs face when making gift dispositions are unique and differ from those associated with other types of stock dispositions. Specifically, market participants might believe that insider-trading regulations protect them from CEO malfeasance associated with inside information, but given that these rules do not explicitly apply to family gift transactions, it is unclear whether there is inside information being used in these transactions. 3 Our study is the first to demonstrate the differences in both incentives and outcomes for family gift transactions, as well as the opportunity to garner abnormal trading profits following disclosure of these transactions, giving us a broader understanding of CEO stock dispositions. Our findings suggest that estate and gift 3 The definitional section of the Securities Act of 1934 states that the [t]he terms sale and sell each include any contract to sell or otherwise dispose of. While this language seems broad enough that gifts could be considered 10b-5 sales, a review of case law suggests courts view exchange of value as an essential element of 10b-5 sales (Sulcoski, 1989). Moreover, the SEC has never pursued litigation related to donation of stock in a federal court (Yermack, 2009). 7

9 tax avoidance guides CEOs timing of gifts to family members, providing a very different signal to market participants for these transactions. Our results demonstrate an opportunity to manipulate CEO gift dispositions in spite of regulators expectations to the contrary. We also provide evidence that the quantity and value of family gift transactions has increased over time. Taken together our findings suggest CEO gift transactions are more important than ever to market participants and to regulators. The paper proceeds as follows: Section 2 provides background on prior literature, reporting requirements and tax planning considerations; section 3 outlines our sample selection and describes our data. Section 4 discusses our methodology and primary results. Section 5 provides additional analyses and sensitivity tests, and Section 6 concludes. 2. Background Prior research in both accounting and finance examines incentives associated with insiders stock sales and generally concludes that absent controls, insiders strategically time their sales immediately prior to negative stock returns that are often predicated by missing analyst forecasts or earnings manipulations (Cheng and Warfield, 2005; Ke, Huddart, and Petroni 2003; McVay, Nagar, and Tang 2006; Bergstresser and Philippon 2006). However, more recent research suggests that new restrictions on insider stock transactions mandated under SEC Rule 10b5-1 and SOX have lowered the likelihood of such informed sales (Brochet 2010). While there has been significant research examining insider open-market sales, studies examining other forms of insider stock dispositions are far more limited. Johnson and Moorman (2005) demonstrate that CEOs time charitable donations around stock price maximums to achieve the largest possible tax deductions. Yermack (2009) finds that CEOs donations to their own family foundations follow a similar pattern, occurring at stock price maximums followed by 8

10 large share price declines. Yermack s (2009) study indicates this pattern is driven by non- December transactions, and he suggests that transactions made in December are driven by lastminute tax planning rather than exploitation of inside information. While not specifically examined in these papers, it is possible that CEOs are better able to opportunistically time charitable donations than stock sales because restrictions on insider trades do not specifically apply to bona fide gifts, including charitable donations and family gifts. While CEOs incentives are generally similar for stock sales and charitable donations, namely to time dispositions around the highest share price, incentives related to gifts of stock to family members are very different. We expect estate and gift tax incentives to drive CEOs to time family gifts around relative stock minimums, consistent with the freezing estate tax planning strategy. 4 When a CEO makes a taxable gift to a family member of stock that is at a depressed price, the depressed price becomes the value included in the taxable estate rather than the expected higher price in the future. Appendix 1 provides a numerical example for the estate freeze strategy. This example assumes a CEO s holdings include company stock that has fair market value (FMV) of $1 million today but is expected to increase in value over the CEO s remaining life; for simplicity, this example assumes an annualized six percent rate over a twenty-year period. 5 Thus, if the CEO continues to hold the stock until death, the amount included in the estate tax base will be the FMV of the stock at the CEO s death (approximately $3.2 million). Alternatively, if the CEO makes a taxable gift of stock to her child today, the value of the gift included in the tax base will be frozen at a FMV of $1 million, as the sum of a CEO s lifetime 4 Studying family gift transactions of Korean controlling shareholders, Jung and Park (2009) find evidence that gifts occur in periods when the frequency of good news is lower and the frequency of bad news is higher relative to other periods, but to our knowledge we are the first to examine whether CEOs time family gifts in a U.S. context. 5 Given mean age of CEOs in our sample (56 years old) and typical life expectancy rates (81.2 years for females and 76.4 years for males in 2014), a twenty-year horizon seems reasonable. Further, the mean raw return in the first year (two years) following a gift transaction in our sample is 21.2% (36.3%), suggesting estimates of tax savings constructed using the annualized return rate for the S&P 500 over the previous twenty years (6%) are conservative. 9

11 taxable gifts must be included in her taxable estate under the unified estate and gift tax system. The trade-off in this example is that, upon receiving the gift, the child becomes liable for any capital gains taxes associated with future appreciation in the stock price. Since the long-term capital gains tax rate is significantly lower than the estate tax rate, this planning strategy saves the family taxes overall. The potential estate tax savings and net tax savings are economically significant. In this example, where the horizon is 20 years, the growth rate is six percent, the estate tax rate is 40%, and the long-term capital gains tax rate is assumed to be the maximum 23.8%, the strategy generates approximately 88 cents of estate tax savings and 36 cents of net tax savings for every frozen dollar gifted. 6 While the estate and gift tax rate has steadily decreased over time, from a high of 55 percent down to its current 40 percent rate, it has been and remains significantly higher than the long-term capital gain rates Sample and Descriptive Data 3.1 Sample Corporate directors, officers, and large shareholders must file public reports for all acquisitions and dispositions of company stock, including family gifts. Most of these insider transactions, notably open-market sales and purchases, must be reported on Form 4 to the SEC. Following the Sarbanes-Oxley Act of 2002 (SOX), the filing deadline for Form 4 was reduced from the tenth day of the month following the transaction to two business days following the transaction. Reporting for certain other insider transactions, however, including bona fide gifts made by insiders is exempt from Form 4 requirements (SEC Rule 16b-5). These exempt 6 In the appendix, we assume that the basis in gifted stock is equal to the $1 million fair market value. However, even if the basis in the gifted stock was $0, the strategy would generate approximately 30 cents of estate tax savings and 12 cents of net tax savings for every frozen dollar gifted. The difference would be based on increased capital gains taxes paid at the eventual sale ($1 Million * 23.8 percent). 7 It is interesting to note that while the difference between the estate and gift tax rate and the long-term capital gain tax rate has decreased during our sample period, we see an increase in the number of gift transactions during our sample period. 10

12 transactions must be reported on Form 5. 8 Notably, the filing deadline for Form 5 is 45 days following the disposition year s end, and rules exempting gifts from Form 4 requirements did not change under SOX. As a result, insiders have always had more discretion in the timing of their reporting for gifts relative to other transactions such as sales or purchases, and this difference is more acute post-sox. Thomson Reuters Insider Trading database provides machine-readable information regarding Form 4 reports and Form 5 reports. 9 Appendix 2 provides the Form 4 Filing of Starbucks CEO Howard Schultz on September 13, 2010 for reference. We examine all Form 4 and Form 5 observations with transaction code G (bona fide gifts) for CEOs during the period 1997 through Consistent with the methodology of Yermack (2009), we distinguish among these observations, which include both family gifts and charitable donations, based on whether there is an immediate indirect ( I in Column 6 of Table 1 on the SEC filing) acquisition ( A in Column 4 of Table 1 on the SEC filing) of the same number of shares. 10 When there is an indirect acquisition of the same number of shares as in the disposition, we label the transaction as a family gift; when there is no indirect acquisition, we label the transaction as a charitable donation. 11 For sale transactions, we examine all observations with transaction code S over the 8 The exception for bona fide gifts applies to family gifts as well as charitable deductions. Insiders may voluntarily report transactions that meet an exception on Form 4, and dispositions that have previously been reported on Form 4 are not required to be reported on Form 5. Approximately 84% of our gift observations come from Form 4 data and the remaining 16% come from Form 5 data. 9 Throughout this paper, we use the term SEC filing to refer to data from both Form 4 and Form 5. In sensitivity analyses, we re-estimate our models controlling for whether the report is made on Form 4 versus Form 5. Our results are not sensitive to whether the filing is made on Form 4 or Form It is important to note that the indirect acquisition could be made by the CEO s spouse. Because of the marital deduction available to spouses, an estate freezing strategy would not apply to spousal gifts. To the extent that spousal gifts are in our sample, these observations should bias us against finding the expected results for timing of gift dispositions. An acquisition by a family trust would likely be for the benefit of both the CEO s spouse and children, and thus the estate freezing strategy would be applicable. 11 In untabulated sensitivity analyses, we also examine charitable donations, finding that the timing of charitable donations looks nearly identical to stock sales with positive returns leading up to the disposition, followed by negative returns. 11

13 same period. 12 We exclude observations of different disposition types reported on the same SEC filing to avoid confounding differential market responses to differing types of dispositions. 13 Panel A of Table 1 details the sample construction. We begin with all observations from the Thomson Reuters Insider Filing database for CEO stock transactions between 1997 and 2013 with required data to identify the firm, the type of transaction, and the transaction value. 14 We then aggregate all transactions on the same disclosure of the same transaction type that occurred on the same transaction date into one observation. For instance, if the disclosure reports two sales on the same date with the same acquisition/disposition and ownership identifiers, we combine these two observations into a single aggregated observation. After removing observations of different transaction types that are reported on the same SEC disclosure (and observations missing other necessary variables for our multivariate regressions) we are left with 1,383 usable gift and 60,677 usable sale dispositions for our long-window, raw return analyses. Further, since our long-window, market-adjusted and buy-and-hold analyses are measured from the date of the SEC disclosure, we further aggregate all remaining transactions on the same disclosure into one observation. For instance, if the disclosure reports two separate aggregated sales, we combine these two observations into a single aggregated observation. This last combination provides 1,306 usable gift and 44,700 usable sale dispositions for our long-window, 12 For example, on Mr. Schultz s Form 4 in Appendix 2, we classify the direct disposition on December 30, 2009 as a charitable donation, since it is reported with a transaction code G and there is no corresponding indirect acquisition of any amount reported on that date. Further, we classify the direct disposition on January 15, 2010 as a family gift, since it is reported with a transaction code G and there is a corresponding indirect acquisition of the same amount reported on that date. Finally, we classify the three direct dispositions on September 13, 14 and 15 as separate open market sales, since they each are reported with a transaction code S and occur on separate days. 13 For this reason, Mr. Schultz s dispositions reported on Form 4 in Appendix A are not included in our sample. 14 CEO is identified by any of the four rolecode variables. To identify the firm, we require either the firm s CUSIP (cusip6) or stock symbol (ticker). To identify the transaction type, we require non-missing acquisition/disposition flags (acqdisp), direct/indirect ownership type (ownership), and the transaction code (trancode). To identify the transaction value, we require non-missing values of the shares in the transaction (shares), and the stock price at the transaction (non-missing tprice for trancode = S and non-missing price from CRSP for trancode = G ). Further, we need the date of the transaction (trandate) and the date of the disclosure with the SEC (secdate). 12

14 market-adjusted and buy-and-hold analyses. Panel B of Table 1 presents the number of sample gift and sale observations by year, along with the difference between the unified estate and gift tax rate and the long-term capital gain rate during our sample period. It is interesting to note that in spite of the decrease over time between these rates, the trend of gift transactions is increasing, as we demonstrate with the negative correlation between the rate difference and number of gifts (-0.633). 3.2 Descriptive Statistics Table 2 presents descriptive statistics for all control and test variables, with separate columns for gift and sale dispositions. Bolded statistics indicate that the mean (median) differs at the 5% level between the two samples. The mean (median) transaction size for gifts is approximately $2 million ($249 thousand), but there is wide variation associated with transaction size. Although transaction size is statistically different between gifts and sales, the transaction size in each of these two types of dispositions appears to be economically similar. If CEOs dispose of stock equally across months, we would expect the percentage of dispositions falling in each month to be around 8 percent (i.e. 1/12 =.083). Notably, a high percentage of family gift transactions occur in December (mean = 0.265), while sales are more in line with this average expectation (mean = 0.087). Finally, we find the reporting lag to be of interest given the differences between gifts and sales. Reporting regulations following SOX required a two-day maximum reporting lag for sale transactions, and the mean and median reflect this change in regulation; however, because these regulations do not explicitly apply to gifts, we see significantly longer lags. For this reason, we further examine the impact of reporting lags on gifts in later analyses. For observations with matching data available from Execucomp (S&P1500 firms), we provide comparative demographic information on CEOs making sales and gift dispositions. We 13

15 do this to give the reader some perspective on the makeup of our sample and to determine whether there are significant differences in CEO characteristics that might be driving our results. We see that age and tenure, while statistically different, are virtually identical for CEOs in our sale and gift samples (i.e. mean and median differences for age and tenure between the gift and sale sample are only around one year). There are subtle differences in compensation for those in the sale versus gift sample, but we do not believe that these differences are large enough to drive our results. Given sample differences are small and requiring Execucomp data would significantly limit our sample, we do not use demographic characteristics as control variables in our multivariate models. Looking at the return patterns in Table 2, we see that raw returns leading up to both sales and gifts are very positive, but the amount is significantly greater for sales relative to gifts. The positive raw return in the pre-period for gifts is inconsistent with our expectations based on an estate freezing strategy and suggests that executives are not timing transactions at the lowest point. 15 Moreover, in the post-disposition period, raw returns are much greater for gifts relative to sales in both the one-year and two-year windows. With respect to post-disclosure marketadjusted returns, our univariate results demonstrate a small positive return for gifts in the oneyear window, but a negative return for sale transactions. Finally, the univariate results show that the three-day cumulative abnormal return (CAR) around the disclosure of CEO sale dispositions is negative. While the reaction to gift disclosures is significantly greater than the reaction to sales, it is not significantly different from zero. Figure 1 presents a graphical depiction of raw 15 In sensitivity analyses, we examine a number of other windows leading up to these transactions, including onemonth, two-months, and three-months preceding the transaction. Our results demonstrate a great deal of consistency across specifications; however, as the window shortens returns are less positive. Based on the positive one-year returns, CEOs do not appear perfectly time transactions to minimize estate and gift tax. The shorter window returns, while positive, do not allow us to rule out backdating in the days leading up to the transaction. We further explore this in Section

16 and market-adjusted returns around gift (Panel A) and sales dispositions (Panel B). Panel A depicts a dip in the market-adjusted returns in the quarter just before gift dispositions consistent with CEOs incentives to time gifts around relative stock price lows. Further, Panel A shows that both raw and market-adjusted returns continue to increase post-gifts. In contrast, Panel B demonstrates that the returns leading up to sales dispositions climb steadily to a kink and then decline in the post-disposition period (flatten in the case of raw returns). Below we expand upon these univariate results with a series of multivariate tests examining both CEO behavior and market reaction around gift dispositions. 4. Methodology and Results 4.1 CEO Benefits Associated with Timing of Transactions We begin by examining gift transactions from the CEO s perspective. Based on their average compensation relative to the lifetime estate and gift tax exemption, we assume CEOs commonly use estate tax planning strategies, like the estate freeze strategy, to preserve family wealth. 16 To provide evidence that CEOs time their stock gift dispositions consistent with estate tax incentives we investigate the pattern of raw returns leading up to and following gift dispositions and compare these return patterns to those around sale dispositions. 17 Specifically, we regress raw stock market returns in the period leading up to (following) sales and gift dispositions to determine whether the timing of these transactions matches CEOs personal and family wealth maximization incentives using the following model: 16 Based on Execucomp data for our sample of CEOs making stock gifts, the mean (median) annual compensation (TDC1) is over $6 Million ($4.5 Million) per year and the mean (median) tenure in the sample is 13.5 (11) years. While not all CEO compensation will be reflected in total wealth at death, CEOs have a number of years to accumulate wealth, giving us comfort that sample CEOs have enough wealth to exceed the lifetime estate and gift tax exemption ($5.45 Million for individuals; $10.9 Million for married couples), making estate planning necessary. 17 In our main tests, we compare gifts to sales but in untabulated results we use purchases as an alternative benchmark. Purchases, like gifts, are optimally timed at relative price minimums. Overall our findings related to gifts are not sensitive to the choice of benchmark. Our supplemental analysis section provides further details regarding the differences between gifts and purchases. 15

17 RAWRET = α + β 1 GIFT + β 2 TRANSIZE + β 3 MVE + β 4 ROE + β 5 DECEMBER + ε (1) where: RAWRET = pre- (post-) disposition raw stock market returns, where pre-disposition returns are measured in the year (-252, -3) or quarter (-63, -3) leading up to the disposition and post-disposition returns are measured one year (+3, +252) or two years (+3, +504) following the disposition date; GIFT = 1 if the disposition is classified using transaction code G by Thomson Reuters and we observe total direct dispositions equal to total indirect acquisitions on the same filing, zero otherwise; TRANSIZE = the natural logarithm of the total dollar value of the transaction identified by Thomson Reuters; MVE = the natural logarithm of the four-quarter average of the size of the firm based on its market value of equity (Compustat CSHO multiplied by the CRSP stock price), ending the quarter before the start of the return window; ROE = the firm s four-quarter average pre-tax return on equity (Compustat PI divided by AT LT), ending the quarter before the start of the return window; and DECEMBER = 1 if the transaction occurs in the month of December, and zero otherwise. RAWRET captures the raw share price performance in the period leading up to the disposition or following the disposition. We use separate specifications for RAWRET in the pre- and postdisposition periods. In the pre-disposition period, we use a one-year (-252, -3) window, and in the post-disposition period, we use both a one-year window (+3, +252) and a two-year window (+3, +504). We use a longer window in the post-transaction period because to make an estate freezing strategy work, we expect CEOs and their heirs to have longer time horizons associated with gift transactions. Figure 2 presents a timeline that depicts all of the different measurement windows used in our various tests. The model intercept represents open-market sale dispositions, and GIFT represents family gift dispositions. Consistent with the tenets of the estate freeze strategy, we expect predisposition raw stock returns to be significantly lower for family gifts than stock sales. Thus, in our model of returns leading up to the disposition, we expect a negative coefficient on GIFT, 16

18 indicating that returns leading up to gifts are lower than returns leading up to sales (intercept). Conversely, given that CEOs face incentives to time stock sales at relative price maxima, we expect post-disposition raw stock returns to be significantly greater for family gifts than stock sales. Thus, in our model of returns following the disposition, we expect a positive coefficient for GIFT, indicating returns following gifts are higher than returns following sales. In addition to our test variables, we also include a number of control variables expected to influence returns around these transactions, noting that all continuous control variables are meanadjusted in all multivariate analyses to make inferences on our test variables more appropriate. We include the natural log of the size of the transaction (TRANSIZE), expecting that the larger the size of the transaction, the greater the CEO s incentive to time such transactions. 18 We also include controls for the size of the firm (MVE) and the firm s return on equity (ROE) leading up to each particular return window. Finally, we include an indicator variable for December transactions (DECEMBER). We expect a great deal of tax planning with respect to sales and gifts to happen in the month of December. CEOs, who generally operate on a calendar year basis for tax purposes, may engage in transactions in December to achieve certain capital gains or meet annual gift tax exclusions, and their actions will depend on their individual tax circumstances. Because CEOs engaged in last-minute December transactions are focused on ensuring that their transactions happen within the taxable year, we do not expect December dispositions to be as well timed for wealth maximization (i.e. maximum sale price or minimum price for gifts). In addition, anecdotal evidence indicates that in 2010 and 2012 individuals rushed to make family gifts on the expectation that tax cuts made as part of the Economic Growth and Tax Relief 18 In sensitivity analyses, we parse our samples across larger and smaller transactions, using both a dollar threshold of $1 Million, similar to Yermack (2009), as well as a sample median threshold. Our inferences continue to hold across these iterations. 17

19 Reconciliation Act of 2001 (EGTRRA) were going to be eliminated and estate tax exemptions 19, 20 would be lowered (Jacobs 2011; Sullivan 2013). Table 3 presents the results of estimating two specifications of Model 1. First we examine whether share price performance leading up to gifts varies predictably with our expectations about CEOs abilities and incentives to optimally time these dispositions. Column 1 presents results where RAWRET is the pre-disposition return measured in the year (-252, -3) leading up to the disposition. The variable of interest is GIFT, with the price performance of SALE dispositions represented in the intercept. We find a negative and significant coefficient for GIFT in column 1 (β 1 = , t = -4.19). Consistent with our expectations, we find that the intercept (sales) is positive and significant in column 1 (α= , t = 61.97). An F-test demonstrates that the returns leading up to gift dispositions are significantly positive for the year. This finding is consistent with our univariate findings and demonstrates that executives do not appear to time gift transactions at the lowest point. 21 Our control for the size of the transaction (TRANSIZE) is positive and significant in both specifications. Given that such a large percentage of our dispositions represent sales, it is not surprising that returns leading up to the disposition are much higher for larger transactions. 22 Our control for firm size (MVE) loads negatively and is statistically significant. In addition, we find 19 In sensitivity analyses, we eliminate December dispositions, finding results that are qualitatively unaltered from our main analyses. 20 These tax cuts and exemption amounts were subject to sunset on January 1, 2011, but were extended through 2012 by the Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of In sensitivity analyses, we have dropped gift observations around the expected sunsets for these provisions. The exclusion of these observations strengthens our findings. 21 In sensitivity analyses, we explore both raw and market-adjusted returns across shorter windows leading up to gift transactions. While we find generally similar results, we find that market-adjusted returns in the one-month pretransaction period (-21, -3) are negative for gifts indicating these stocks relatively underperform the market in the short-term. These negative returns could be consistent with backdating which we explore further in Section In sensitivity analyses, we interact transaction size with GIFT, finding that transaction size amplifies the findings for stock sales; however, we find that the returns are more negative for larger gift transactions. 18

20 that for December transactions (DEC), the pre-disposition returns are lower, suggesting that December transactions are likely more of a function of ensuring that the disposition happens by year-end rather than attempting to time the disposition at a relative stock price peak. 23 Columns 2 and 3 of Table 3 present our tests examining the share price performance following each disposition; RAWRET is the post-disposition raw return, where post-disposition returns are measured one-year (+3, +252) and two-year (+3, +504) from the disposition date, columns 2 and 3 respectively. We find that the coefficient on GIFT is positive and significant in both column 2 (β 1 = , t = 5.25) and column 3 (β 1 = , t = 4.16). F-tests also demonstrate that share price performance following the disposition for family gifts is statistically greater than zero. Overall the results in Table 3 provide evidence that the timing of CEO stock gifts is distinctly different from insider sales transactions and is generally consistent with estate tax incentives. 4.2 Information Content of Gift Disclosures After demonstrating how CEOs time gift dispositions consistent with their incentives, we next examine whether gift dispositions are informative to the broader market. We believe it is important to determine not only whether CEOs making gifts can profit from these dispositions but also to determine whether a trading strategy can be derived from buying shares following the public release of regulatory disclosures announcing family gift dispositions. To do this, we regress market-adjusted stock returns in the period following reported disclosure using the following model: MKTRET = α + β 1 GIFT + β 2 TRANSIZE + β 3 MVE + β 4 ROE + β 5 DECEMBER + ε (2) 23 In sensitivity analyses comparing December versus non-december dispositions, we find no differences in our long-window raw results, but that our long-window market-adjusted results (Table 4) are suppressed for December gift dispositions. 19

21 where: MKTRET = post-disclosure cumulative market-adjusted returns, measured one-year (+3, +252) and two-years (+3, +504) from the date the disclosure filing is received by the SEC. All other variables are as defined in Model (1). Model (2) is similar to Model (1); however, in Model (2) we focus on returns following the disclosure of gift dispositions. As in Model (1), the benchmark group, open-market sales, is represented by the model intercept, and GIFT represents family gift dispositions. We expect a negative coefficient for the intercept in the post-disclosure window consistent with prior research. We also expect a positive coefficient for GIFT, indicating that long-window marketadjusted returns for gifts are significantly greater than those for sales. Moreover, consistent with a viable trading strategy, we expect the sum of the intercept and the coefficient for GIFT to be significantly greater than zero. Panel A of Table 4 presents the results of our long-window market-adjusted return tests. Columns 1 and 2 present results where MKTRET is the post-disclosure market-adjusted return, measured one-year and two-years from the report date respectively. The coefficient on the intercept (sales) indicates that post-disclosure market returns are positive for sale transactions two-years out (column 2). The coefficient for GIFT is positive both in column 1 (0.0622; t = 4.43) and column 2 (0.0637; t = 3.30), indicating that market returns are significantly greater for gifts than for sales. In both specifications, significant F-tests suggest that market returns following the disclosure of gift dispositions are significantly different from zero, supporting a trading strategy. In addition to market-adjusted returns, we re-estimate Model (2) using buy-andhold abnormal returns (BHARET), and these results are presented in Panel B of Table 4. Our BHARET tests yield similar inferences to those of Panel A, demonstrating that long-term 20

22 abnormal profits can be earned through purchasing stock at the disclosure of a family gift disposition. Finally, in Panel C, we examine long-term annual returns using ta Fama and French (1993) three-factor model with momentum (Carhart, 1997). It is important to note that we do not include sales observations in these analyses. We report positive alphas in both one-year (0.43 percent) and two-year (0.36 percent) windows, and we find the coefficient estimates for each of the risk factors is generally consistent with those reported in prior literature. These results corroborate our other market analyses and suggest that an abnormal return can be generated from a trading strategy based on buying stocks following the disclosure of executive gifts to family members. While Panel C presents value-weighted returns, our results are not sensitive to using equalweighting or to eliminating the momentum factor. 4.3 Tests of Subsequent News Our long-window market return tests in Table 4 indicate disclosures of CEO gift dispositions are informative. Disclosure of these transactions should have signaling value to the extent CEOs gift dispositions reflect private information. To test this, we follow a model similar to Brochet (2010) and regress CEOs gift and sale dispositions at the firm-quarter level on three indicator variables that equal 1 if EPS at the end of quarters 1, q+1, and q+2 is greater than or equal to the consensus analyst forecast measured at the beginning of quarter q, and zero otherwise. This methodology allows us to determine if CEOs are acting on foreknowledge that is not reflected in analyst forecasts at the beginning of the transaction period. In our study, we expect a negative association between CEO sales and subsequent positive earnings surprises 24 ; however, we expect 24 This methodology is similar in nature to Brochet (2010); however, because of the unique nature of gift dispositions we make two deviations. First, because of the reporting lag associated with gift dispositions, we use the disposition date as opposed to the report date. Second, we do not consider changes pre- versus post-sox since this 21

23 a positive association with subsequent earnings surprises for gift dispositions. In addition to future earnings surprises, we also include returns for the quarter leading up to the disposition, returns in the disposition period, and aggregated returns for the two subsequent quarters. TRANSIZE = α + β 1 MBE q + β 2 MBE q+1 + β 3 MBE q+2 + β 4 RET q-1 +β 5 RET q + where: β 6 RET q+1,q+2 +γ(controls) + ε (3) MBE q+n = an indicator variable equal to 1 if actual EPS for quarter q+n is at least as high as the median analyst forecast at the beginning of quarter q, zero otherwise; RET q-1 = the 63 trading day cumulative, raw return from the quarter prior to the quarter of the disposition; RET q = the 63 trading day cumulative, raw return from the quarter of the disposition; and RET q+1, q+2 = the 126 trading day cumulative, raw return for the two quarters after the quarter of the disposition. Table 5 presents the results for estimating Model (3). Column 1 presents results for gift dispositions, and Column 2 presents results for sale dispositions. For sales, our results demonstrate that firms are more likely to meet earnings expectations in the quarter of the disposition. While our results here are somewhat weaker than those of Brochet (2010), we believe that our inferences are similar. Both RET q and RET q-1 are significantly positive for sale dispositions, suggesting that CEOs sell shares following a significant share price run-up. As expected, our results for gifts stand in contrast with our results for sales but mimic Brochet (2010) s findings for purchases. 25 Firms are less likely to meet expectations in the quarter of the gift disposition. Regarding returns, we see negative coefficients for both RET q and regulatory change primarily impacted reporting for sales rather than gifts (see Additional Analysis for more details on the impact of regulatory changes). Given the consistency of our sales results to those of Brochet (2010), we are confident that these changes do not materially change our inferences. 25 In untabulated tests we examine purchases. Our results for purchases are similar to Brochet (2010) s findings for purchases and similar to our findings for gifts. 22