Agency Costs, Charitable Trusts, and Corporate Control: Evidence from Hershey's Kiss-Off

In July 2002 the trustees of the Milton Hershey School Trust announced a plan to diversify the Trust's investment portfolio by selling the Trust's controlling interest in the Hershey Company. The Company's stock jumped from $62.50 to $78.30 on news of the proposed sale. But the Pennsylvania Attorney General, who was then running for governor, opposed the sale on the ground that it would harm the local community. Shortly after the Attorney General obtained a preliminary injunction, the trustees abandoned the sale and the Company's stock dropped to $65.00. Using standard event study methodology, we find that the sale announcement was associated with a positive abnormal return of over 25% and that canceling the sale was followed by a negative abnormal return of nearly 12%. Our findings imply that instead of improving the welfare of the needy children who are the Trust's main beneficiaries, the Attorney General's intervention preserved charitable trust agency costs of roughly $850 million and foreclosed salutary portfolio diversification. Furthermore, blocking the sale destroyed roughly $2.7 billion in shareholder wealth, reducing aggregate social welfare by preserving a suboptimal ownership structure of the Company. Our analysis contributes to the literature of trust law by supplying the first empirical analysis of agency costs in the charitable trust form and by highlighting shortcomings in supervision of charities by the state attorneys general. We also contribute to the literature of corporate governance by measuring the change in the Company's market value when the Trust exposed the Company to the market for corporate control.


INTRODUCTION
Domestic tax-exempt charitable organizations hold roughly $2.4 trillion in assets and have roughly $1.2 trillion in annual revenues. 1 These impressive figures speak to the American impulse toward philanthropy. 2 But the ugly secret about charitable organizations is that their lack of clearly defined owners invites deadweight losses arising from agency costs. True, the state attorneys general have formal authority to ensure that the managers of such organizations efficiently pursue a bona fide charitable purpose. 3 In reality, however, the typical state attorney general is an elected political official for whom the supervision of charitable organizations offers little political payoff. Accordingly, scholars tend to assume the existence of significant agency costs in charitable trusts and other charitable organizations. 4 A separate agency problem arises in the context of the public corporation. The concern here is that the interests of the company's managers will diverge from those of the shareholders. To align managers' incentives with the interests of shareholders, the law gives shareholders various rights, including the right to sell their shares, which invites a takeover bid if the company's current management yields inferior returns. An alternative mechanism for minimizing corporate agency costs arises when the firm has a controlling shareholder or a large blockholder. 5 A controlling shareholder or a large blockholder is more likely than an ordinary shareholder to have a concrete financial incentive to monitor actively and, if necessary, to impose value-maximizing operations on the company's managers. Indeed, a host of scholars and policymakers have come to embrace the utility of monitoring by blockholders, 6 and the incidence of such control blocks is increasing among public U.S. firms and is quite common among public companies in Europe. In 2002 the foregoing agency problems collided in a salient real world event: the attempted sale of the Hershey Company by its controlling shareholder, the Milton Hershey School Trust, a Pennsylvania charitable trust. This paper uses the 2002 Hershey incident as a natural quasiexperiment to investigate empirically: (1) the agency costs that are inherent in the charitable trust form; (2) the shortcomings in supervision of charitable entities by the state attorneys general; and (3) the comparative merits of takeovers (i.e., the market for corporate control) versus monitoring by controlling shareholders (i.e., the market for partial corporate control) in minimizing agency costs in corporate governance. 8 In so doing, we make fresh contributions to the literatures of trust law and corporate law and to the technical literature on event study methodology.

Background
The largest confectionary outfit in North America, the publicly traded Hershey Company (the "Company") is the maker of such familiar goodies as Almond Joy, Jolly Ranchers, Kit Kats, Milk Duds, Reese's Pieces, Twizzlers, York Peppermint Patties, and, of course, Hershey's Chocolate Bars, Chocolate Kisses, and Chocolate Syrup. 9 The Company is based in Hershey, Pennsylvania, a town that fancies itself "The Sweetest Place On Earth." 10 In Hershey, Cocoa Avenue intersects with Chocolate agers and the Hershey Trust Company's board collectively as the "trustees"). 18 The Trust's heavy investment in the Company is not healthy for the Trust or the Company. For the Trust, holding an undiversified portfolio makes the welfare of the School and the needy children it serves vulnerable to swings in the performance of a single company. The Trust, in other words, is exposed to what financial economists call uncompensated risk-risk for which there is no offsetting improvement in expected return and that could be avoided costlessly by diversifying the Trust's investment portfolio. 19 Indeed, the modern law of trust investment has come to regard lack of diversification as close to a per se breach of fiduciary duty. 20 For the Company, the Trust's controlling interest has two distorting effects. First, no matter how poorly the Company's managers run the Company, so long as the Trust maintains its controlling interest and the trustees remain listless, the Company's managers need not worry about being displaced in a takeover. Second, fear that taking on normal business risks might impair the welfare of the needy children who depend on a steady flow of dividends from the Company to the Trust has caused the Company's managers to forgo potentially lucrative business opportunities. For example, Milton Hershey's immediate successor was so worried about the Trust's exposure to volatility in the cocoa bean market that, instead of expanding the business to meet exploding consumer demand in the economic boom that followed World War II, he rationed the Company's products, failed to increase manufacturing capacity, and ordered the construction of silos to store an enormous reserve of cocoa beans. 21 In part as a result of this missed opportunity, the Company still lacks an international presence comparable to that of competitors such as Mars, Nestlé, and Cadbury. 22 The Trust's concentration in the Company has also created an unfortunate expectation in the town of Hershey that the trustees will retain control over the Company indefinitely. Indeed, after the trustees abandoned the 2002 effort to diversify the Trust's holdings by selling the Trust's controlling interest in the Company, a local lawyer who organized protests against the sale told the New York Times that "Our cash cow is safe; we're feeling really great." 23 This sentiment, however, obscures the sad reality of the Trust's exposure to uncompensated risk and the growing disconnect between the Trust's value and its mission. Only 1,700 children benefit from the Trust's $8.8 billion in assets. 24 The quid pro quo for charitable trust status, which the common law rewards with perpetual existence and the public subsidizes through generous state and federal tax exemptions, 25 is that the trust must benefit society by properly advancing a bona fide charitable purpose. It is hardly obvious that a trust with so large a corpus and so small a mission warrants the tax subsidy.

The Natural Quasi-Experiment
Having recognized the Trust's dangerous dependence on the Company, the trustees developed a plan in 2002 to diversify the Trust's holdings by selling the Trust's controlling interest in the Company. The Wall Street Journal broke the news of the trustees' plan on July 25, 2002. 26 The Company's stock, which is traded on the New York Stock Exchange, jumped to $78.30 from the previous day's closing price of $62.50, 27 implying that the capital markets saw vast potential in the Company that had not been realized while the Company was under the control of the Trust. However, the Pennsylvania Attorney General, who was then running for governor and whose office supervises Pennsylvania charitable trusts, brought suit to stop the sale on the grounds that it would lead to layoffs and plant closings that would harm the central Pennsylvania community. In September 2002, after the Attorney General obtained a preliminary 21. See Joël Glenn Brenner, The Emperors of Chocolate: Inside the Secret World of Hershey and Mars 198-200 (1999 injunction, the trustees abandoned the sale. 28 The Company's stock dropped to $65.00 on news of the trustees' surrender. We use a standard event study econometric methodology to isolate price movements in Hershey's stock related to news of the sale from movements owing to ordinary market volatility. 29 We find that news of the proposed sale was associated with a statistically and economically significant increase in the Company's value (in event study jargon, a "positive abnormal return") of over 25%. 30 News of the sale's cancellation was associated with a statistically and economically significant decrease in the Company's value (a "negative abnormal return") of nearly 12%. 31 The magnitude of these abnormal returns is astonishing; "positive abnormal returns of even 1 percent [are] considerable for competitive capital markets." 32 In dollars, our findings imply that blocking the sale destroyed roughly $2.7 billion in shareholder wealth, 33 suggesting a social welfare loss owing to a suboptimal ownership structure of the Company and its underlying assets. 34 Moreover, instead of improving the welfare of the needy children who are the principal objects of Milton Hershey's beneficence, the Attorney General's intervention preserved charitable trust agency costs on the order of roughly $850 million (about 15% of the Trust's 2002 value) and continued the Trust's exposure to uncompensated risk. 35

Trust Law
Regarding trust law, this paper joins the small but growing empirical literature by providing the first quantitative empirical analysis of agency costs in the charitable trust form, 36 37 If the trustees had been managing the Trust's assets efficiently, the sale announcement should not have had a very large positive effect on the value of Hershey shares not held by the Trust. Effective trustees would already have used their controlling vote block to impose a value-maximizing strategy on the Company's managers. 38 Such efficiency-inducing efforts would have accrued to the benefit of all shareholders in the form of a higher stock price.
Yet we find that publicly traded shares in the Company experienced a large and statistically significant positive abnormal return when news of the trustees' plan to sell broke, and a large and statistically significant negative abnormal return when the trustees abandoned the sale. Regardless of the mechanism for the increase in the Company's value on news of the sale-for example, expectation of better management from or synergies with a prospective buyer, overpayment by an empire-building buyer, or release from the worry that aggressive pursuit of profit through higher risk/return opportunities might imperil the needy children who depend on the Trust-the implication of our results is the same. The trustees not only exposed the Trust to uncompensated risk, but they also left money on the table that could have been realized for the benefit of the Trust's charitable purpose.
Our findings therefore provide the first quantitative empirical validation, albeit in a single case study, of prior theoretical claims of agency costs in charitable trusts and inefficient management of charitable trust assets. 39 Our findings also highlight an additional but less often dis- trustees with control over corporation "shall elect or appoint directors or other managers who will manage the corporation or enterprise in the best interests of the beneficiaries").
39. See infra notes 219-222 and accompanying text. Although there has been no significant empirical investigation of charitable trusts, see supra note 36, there is an empirical literature on not-for-profit firms. For example, in John E. Core, Wayne R. Guay cussed concern. Even if the Attorney General does intervene in the administration of a charitable trust, the intervention might be designed to promote the Attorney General's political interests rather than the trust's charitable purpose. 40 Indeed, the Attorney General's intervention in this instance preserved agency costs within the Trust on the order of roughly $850 million. 41 Accordingly, we provide an empirical grounding for policy analysis of new developments in the supervision of charitable entities. 42 For example, some have argued that the IRS should be more active in policing charities, 43 almost half the states now give the donor standing concurrent with the attorney general to enforce a charitable trust, 44  shey and the Company's workers, our findings also bear on the propriety of social investing by trustees and other fiduciaries. 46 The $850 million cost to the Trust of protecting the local "cash cow" by abandoning the sale equates roughly to $67,000 per resident of Hershey, or $62,000 per worker at the Company. 47

Corporate Law
Regarding corporate law, our findings add to the intertwined literatures relating to takeovers and controlling shareholders. Specifically, the aborted 2002 sale provides a natural quasi-experiment that allows us to measure the difference in the Company's value when the Company's managers were exposed to the market for corporate control in comparison to when the Company's managers were under the thumb of the trustees. We find a large, statistically and economically significant positive abnormal return on news of the Trust's probable divestment and a large, statistically and economically significant negative abnormal return when the sale was abandoned.
Accordingly, the evidence indicates that the concentration of vote power in the hands of the trustees did not improve the value of the Company relative to the value of the Company in the shadow of the takeover market. As such, our findings highlight the potential inefficacy of disciplining by a controlling shareholder when that shareholder itself has distorted incentives stemming from its own internal agency problems. On this view, a hedge fund is likely to be a more effective monitor than a pension or mutual fund. 48 Our findings also lend modest empirical support for the view that the takeover market reduces corporate agency costs.

Financial Econometrics
Because our study is by necessity limited to a single firm, a concern might arise about the statistical power of our results. To ameliorate this concern, we undertake a host of robustness checks, several designed specifically to address the single-firm problem. In so doing, we contribute to the literature of financial econometrics by offering the first demonstraneeded for the particular charitable purpose to such an extent that the continued expenditure of all of the funds for that purpose . . . would be wasteful." tion of the use of randomization inference as a robustness check for statistical inference in a single-firm event study. 49 * * * The remainder of this Article is organized as follows. Building on the foregoing statement of background, Part I frames the inquiry with more details on Milton Hershey, the Company, the School, and the Trust. Part II motivates the empirical analysis by examining the aborted sale in light of current understandings of the supervision of charitable trusts and corporate governance. In so doing, Part II situates our analysis in the existing literatures of trust law and corporate law, particularly agency cost analyses of those subjects. Part III presents our empirical analysis, which includes a graphical analysis of the data, a more formal event study analysis, and a nontechnical summary of our main findings. Mindful of the need for caution in generalizing from what is in effect a case study, Part IV assesses the positive and normative implications of our findings for the policy debates in corporate law and trust law identified above. After a short conclusion, three substantive appendices follow. Philadelphia, Chicago, and New York, but each failed. 52 Hershey then returned home to Pennsylvania and opened a caramel business. Here Hershey's luck improved. Thanks to a large, serendipitous order from an Englishman who happened upon Hershey's caramel while passing through town, a bridge loan approved against bank policy by a junior loan officer, and a better taste owing to Hershey's use in his caramel formula of milk instead of the more common paraffin, Hershey's caramel business prospered. 53 The business proved so successful that in 1900 Hershey was able to sell it for $1 million. Crucially, however, Hershey retained the Company's fledgling chocolate division. 54 Although disinterested in day-to-day business management, Hershey had "a natural flair for experimentation" and a feel for "the intimate relationship between timing, temperature and taste." 55 For Hershey, therefore, the lure of chocolate was irresistible. Chocolate-the rich, sensual, and satisfying food derived from the cocoa bean-had long been a pleasure of the upper class. 56 But because chocolate was so difficult to mass produce and ship in an edible and economically viable form, few others had ready access to it. 57 Milk chocolate, a solid form of chocolate with a pleasing taste that could be mass produced and widely distributed at bearable cost, was thus the holy grail for confectioners. The trick was to induce the water-based milk to combine with the fatty, oil-based cocoa. 58 Having earlier succeeded with milk in his caramel formula, Hershey began construction of the Hershey Chocolate Company's main factory and the surrounding town of Hershey in the heart of Pennsylvania's dairy country even before perfecting a viable milk chocolate recipe. 59 The timing was tight, but after much experimentation Hershey eventually solved the problem of combining milk and cocoa. Hershey's solution was to boil the milk to the brink of souring, which explains the hallmark bitter, harsh taste of Hershey's milk chocolate. 60  substance for many Americans. 61 Hershey's chocolate thus came "to define the taste of chocolate for" the American palate 62 -a process helped along by Hershey's production of "Field Ration D" during World War II, a high-calorie, high-melting point chocolate bar developed for the Army that garnered the Company an exemption from wartime rationing, 63 and by fawning press coverage of the corporate "utopia" that was the town of Hershey. 64 By the end of World War II, the Company dominated the U.S. chocolate market. 65 As with the caramel business, Hershey left the day-to-day operations of the chocolate company to others and turned his attention elsewhere, particularly to sponsorship of the Milton Hershey School, 66 which he and his wife Catherine founded for the benefit of orphan boys in 1909. 67 Indeed, because Milton and Catherine never had children of their own, the boys enrolled in the School became their surrogate children. "This was especially true for Milton, who reviewed every application and helped to choose the youngsters who got rooms at the homestead." 68 As Milton later explained, "Well, I have no heirs; so I decided to make the orphan boys of the United States my heirs." 69 The original 1909 Deed of Trust that created the School was quite detailed. The School-then known as the Hershey Industrial Schoolwas to "be permanently located in Derry Township," the political subdivision of Pennsylvania in which the town of Hershey is situated, on farm land conveyed to the Trust by Milton Hershey. 70 The trust instrument prohibited the use of Trust funds for any other purpose. 71 The School was to admit "poor, healthy, white, male orphans" between the ages of four and eight. 72 The Deed of Trust defined the term "orphan" to mean "a child whose father is deceased." 73 Preference was given first to orphans from the three counties closest to the town of Hershey, then to those from the rest of Pennsylvania, and finally to those from the remainder of the United States. 74 The purpose of the School was to produce productive members of society. The Hersheys instructed that: [T]he main object in view is to train young men to useful trades and occupations, so that they can earn their own livelihood. Each and every scholar shall be required to learn, and to be thoroughly instructed in some occupation or mechanical trade, so that when he leaves the School . . . he may be able to support himself. 75 Once admitted, an orphan could stay at the school until the age of eighteen. 76 Students who went on to college were eligible for tuition assistance. 77 Graduates who did not plan on further education were given "$100, a new wardrobe, and help in finding a job." 78 The School was also to feed, clothe, and board the orphans. "All orphans admitted to the School shall be fed with plain, wholesome food; plainly, neatly, and comfortably clothed, without distinctive dress, and fitly lodged." 79 From its founding and to the present day, the School has provided room and board by placing each student in a home with several other students supervised by a married couple who serve as houseparents. 80 In 1918, three years after Catherine died, Milton Hershey transferred substantially all of his assets-including his stock in the Company, then worth $60 million-to the Trust. 81 The transfer was not publicly disclosed until the New York Times ran a story about it five years later. 82 But once the School's dependence on the Company was known, the Company's managers took notice, becoming cautious and resisting new methods in production and marketing out of worry that change could imperil the steady flow of dividends to the Trust. 83 Eventually, after falling behind its leading domestic competitor the Mars Company in the 1970s, the Company was forced to embrace modernity. 84 The Company's managers now speak of being "more conscious of 74 our stock price and what we have to do to improve the stock price, to give value to our shareholders." 85 Thanks to better management, improved marketing, and several successful acquisitions, the Company is again number one in the North American confectionary market. 86 In 2007, the Company reported net sales of $5 billion and at year end had a market capitalization of $8.9 billion. 87 The Mars Company remains a close second, however, and in 2007 the Hershey Company announced a plan to streamline operations that will include domestic plant closings and layoffs and the construction of a new factory in Mexico. 88 Just as the Company has modernized, so have the Trust and the School. Since 1909, the original Deed of Trust has been modified to expand the age range for admission to four through fourteen (1933) and then to four through sixteen (1970); to redefine "orphan" to include children whose mothers are deceased (1933) and then to replace "orphan" with "children" who do not receive "adequate care from one of his or her natural parents" (1976); to drop the word "Industrial" from the School's name (1951); and, as is now typical for discriminatory trusts, to remove the racial (1970) and gender limitations (1976). 89 The School, which in the 1980s was called "terrible" by accrediting officials, has evolved from an agrarian vocational institution to more of a college preparatory school. 90  acres of rolling countryside" and boasts state-of-the-art facilities (including the world's second largest marble rotunda). 91 The School clothes, feeds, and provides health care for 1,700 students whose average family income is under $14,000. 92 The School also gives scholarships for college tuition. 93 All of the School's operations are funded by the $8.8 billion Trust. Table 1 summarizes the Trust's financial position and dependence on the Company. The Trust's size is striking, particularly given the School's small enrollment. In spite of the Trust's mushrooming corpus, the School "served no more children at the start of 2005 than it did in 1963." 95 The Trust's endowment is also remarkable for its breathtaking lack of diversification. Of the Trust's $8.8 billion corpus, $4.7 billion (54%) is in Company stock (representing 31% of all outstanding shares of the Company). 96 Close to half of the Trust's income from interest and dividends comes from its Hershey shares. What is more, these figures understate the Trust's diversification problem because the Trust's other assets include, among other local holdings, the Hershey Entertainment & Resorts Company. 97 The problem with an undiversified portfolio is that it entails uncompensated risk-risk for which there is no offsetting improvement in expected return and that could be avoided costlessly by diversifying. 98 To understand why, imagine ten companies, each in a different industry and each with different managers. If all ten have the same expected risk and return profile, then a portfolio of all ten stocks would have the same expected return as a portfolio of just one. But there will be less variance in the actual return of the ten-stock portfolio than in the portfolio of one, because in the larger portfolio losers will be offset by winners. 99  that, in certain circumstances, optimal portfolio for small investors may be very concentrated), that notion does not bear on analysis of the Hershey Trust. The trustees did not claim any benefits from the Trust's concentrated holding. On the contrary, as we detail in Part II.A, the trustees wanted to diversify.
99. Assuming returns in the ten firms are independent, the ten-firm portfolio would have 1/10 the variance of a one-firm portfolio. This is a simple application of the calculation of variance, which holds that if you have a nonrandom factor, b, multiplying a random outcome that has variance s 2 , the variance of the product will be b 2 s 2 . Furthermore, the variance of the sum of c independent random outcomes, each with single-stock portfolio presents more risk, as measured by variance, without an offsetting increase in expected return.
To be sure, an investor cannot diversify away general market risk (in the jargon of financial economics, "systematic risk"). But it is possible to avoid most firm risk ("unsystematic risk"). 100 The key to the risk-avoiding magic of diversification, which can be achieved with investment in as few as twenty stocks, 101 is that unsystematic risk is often inversely correlated across firms. A breakthrough in solar power, for example, would negatively affect an oil company, but would positively affect an energy-dependent manufacturing concern. In the case of the Hershey Company, and so the Hershey Trust, firm-specific risk factors include volatility in the cocoa and sugar markets. 102 According to published accounts, the bursting of the technology bubble and the Enron and Worldcom scandals sensitized the trustees to the perils of an undiversified portfolio. 103 In addition, the Pennsylvania Attorney General's office reportedly urged the trustees to diversify in meetings held in December 2001. 104 Nothing in the Deed of Trust requires the trustees to retain the Trust's stock in the Company. On the contrary, the trust instrument gives the trustees "full power and authority" over the Trust's portfolio management. 105 As a matter of hornbook law, therefore, the trustees have the authority to diversify the Trust's investment portfolio if they determine that diversifying is in the best interests of the Trust. 106 variance s 2 , will equal cs 2 . Thus, in the present example, the contribution of each individual firm with return variance s 2 to the portfolio will be (1/10) 2 s 2 = (1/100) s 2 leading to a portfolio variance of 10 * (1/100) s 2 = (1/10)s 2 . For the first application of this principle in the finance literature, see Harry  News of the trustees' plan to sell the Trust's controlling interest in the Company first became public on Thursday, July 25, 2002, in a frontpage story with a two-column headline in the Wall Street Journal. 108 Both the Trust and the Company confirmed the Journal's story that day. The subsequent media coverage, which credited the Journal with breaking the story, reported that securities analysts were surprised by the announcement. 109 In the ensuing litigation, the trustees and the Attorney General also credited the Wall Street Journal with making the sale plan public. 110 According to the July 25 Journal story, the trustees were "keen to diversify" the Trust's portfolio and had rejected alternatives to a sale "such as a buyback or major recapitalization" in which the Company would buy some of the Trust's shares at a premium. 111 The Journal also reported that "People familiar with the matter say that Pennsylvania Attorney General Mike Fisher's office, while not advocating a sale of the company, members of the H School governing board determine that it would be prudent for the charity to diversify its investments, and therefore to dispose of the stock in H Foods, and that not to do so would entail substantial financial risk that would far outweigh the benefit of community good will. Because the H school does not have as a charitable purpose maintaining a controlling interest in H Foods for the benefit of the town of H, the governing board may prudently diversify. Id. 107. Our narrative of the aborted sale is presented in day-by-day detail to allow the reader to evaluate our choices of event dates in the subsequent empirical analysis. To ensure that we did not miss important events, we examined every story in the Wall Street Journal, New York Times, and Pittsburgh Post-Gazette that mentioned " Hershey . The only possible exception to the otherwise uniform view that news of the sale broke on July 25 is a cryptic reference in the original July 25 Wall Street Journal story to "rumblings in the food industry six to eight months ago that the trust was looking to diversify its holdings and had hired an investment bank to advise it on a possible sale." Branch et al., Sweet Deal, supra note 26. These rumblings, however, do not appear to have lessened the surprise of the July 25 story. Further, if these rumblings did affect the price of the Company's stock prior to July 25, such an effect would be positive and so would work against our finding of a large, positive abnormal return on July 25. We take up the possibility of leakage in Appendix B.
has urged the Hershey trust to diversify." 112 The Journal story predicted a "grim" reception for the trustees' plan among Hershey residents and the Company's workers. 113 Hershey residents and workers received the news grimly indeed. With "tears in his eyes" the Company CEO Richard Lenny broke the news to the Company's employees at a 9 A.M. meeting. 114 Lenny also recorded a statement, which was broadcast on the internet every half hour, to the effect that he "disagreed with the trust's actions" and that he had offered the Trust alternative means to diversify. 115 The next day, Friday, July 26, a local newspaper reported that "Hershey residents questioned why the trust would need to sell Hershey Foods, wondered how the sale would affect the village, and asked if a sale would be consistent with Milton Hershey's vision of social responsibility." 116 The chairman of the local Township's Board of Supervisors warned that "[t]he economics of the region could be impacted severely" from loss of jobs. 117 On Sunday, July 28, a local paper quoted the president of the School's alumni association, which had previously accused the trustees of conflicts of interest and other forms of malfeasance, 118 as saying that the trustees' plan was "despicable." 119 Investors took a more favorable view. On Thursday, July 25, the day of the Wall Street Journal's first story, the Company's common stock (ticker symbol HSY) closed at $78.30, up $15.80 (more than 25%) from the prior day's closing price of $62.50. A little over 19 million HSY shares traded on July 25, about ten times as many as the day before. 120 On Friday, July 26, the press ran stories speculating about potential buyers and possible prices. 121  By Monday, July 29, local opposition to the sale began to organize. The School's alumni association asked D. Michael Fisher, the Pennsylvania Attorney General and the Republican candidate for governor in the following November gubernatorial election, "to remove the trust's leadership for 'fiscal waste and child-care incompetence.'" 122 On Wednesday, July 31, six former trustees (including the former Company CEO Richard Zimmerman) made public statements in opposition to the sale. 123 Both Fisher and Ed Rendell, the ultimately victorious Democratic gubernatorial candidate, openly opposed the sale. 124 As part of a grassroots "Derail the Sale" campaign, a public rally against the sale was held on Friday, August 2. 125 Residents began displaying yard signs with slogans such as "The Hershey Trust-An Oxymoron" and "Don't Shut Down Chocolate Town." 126 John Dunn, a former Hershey marketing executive, crafted the opposition's message to ensure favorable media coverage. 127 By Monday, August 5, the editorial page of the Lancaster New Era opined that "[i]t almost seems that the current members of the trust are about the only people to think the sale is a good idea." 128  On Tuesday, August 6, Fisher met with several of the trustees to urge them to reconsider their sale plan in favor of alternative modes of diversification. 129 On Wednesday, August 7, the trustees held a special day-long meeting to consider Fisher's proposals and the merits of continuing the sale. That evening, after the stock market closed, the trustees issued a release stating that they would continue to pursue the sale because they believed it to be "in keeping with [their] fiduciary responsibility to protect and preserve the trust." 130 In the next day's trading (Thursday, August 8) HSY closed at $74.09, up $2.20 (3%) from the prior day's closing price of $71.89.
Fisher then announced that he would turn to the courts. 131 On Monday, August 12, Fisher asked the local court with jurisdiction over the Trust-ironically known in Pennsylvania as the Orphans' Court-to issue an order to the trustees to show cause why any sale of the Company should not be subject to judicial approval. 132 The court issued the order to show cause the following Monday, August 19. 133 Said a spokesman for Fisher, "Ultimately, we hope that the court will require that any offer for Hershey Foods be revealed to our office and the court and that any proposed sale be subject to court approval." 134 HSY closed at $74.56 on August 12 and $74.71 on August 19.
Later that week, on Friday, August 23, Fisher asked the court to enjoin the trustees from any sale pending the court's resolution of the pro- priety of the trustees' sale plan. 135 HSY closed at $75.03. Fisher's office also announced that it was drafting legislation that, if passed, would require the trustees of a charitable trust to weigh the effect on the community in making investment decisions. 136 The trustees filed their papers in opposition to the injunction request on the following Monday, August 26. 137 HSY closed at $76.80. The Orphans' Court scheduled a hearing for Tuesday, September 3. 138 At the September 3 hearing, with protestors outside the courthouse "carrying signs and singing a Hershey's chocolate jingle," 139 the main witness for the Attorney General was former Hershey CEO Richard Zimmerman. Zimmerman predicted that an acquiring company would seek to cut costs first by eliminating jobs made redundant by the merger and second by shutting down less efficient manufacturing plants. 140 "And I suspect," testified Zimmerman, "that one would start with the [main Hershey] plant that's nearly a hundred years old." 141 In Zimmerman's view, although these actions might improve the profitability of the combined company, "there are very many more things in life more important than money." 142 The trustees argued that the trust instrument authorized them to divest the Trust's holdings in the Company and that prudent investment practice required diversification. 143 Indeed, the trustees' investment advisor testified that the Trust's heavy investment in the Company made its portfolio "twice as risky as the typical college or university or independent school portfolio." 144 The investment advisor also admitted, however, that 135. See Brief of Appellee, supra note 110, at 5-6; Brief of Appellant, supra note 132, at 12; see also Ellison, Sale of Hershey, supra note 104 (noting that Fisher had "asked the court to delay a sale until October" 139. D'Antonio, supra note 8, at 263. On Monday, September 2 (Labor Day), the day before the hearing, a local paper ran a story about Milton Hershey's "dream of building a 'model town'" and the reliance of the community on the Company, which "is considered a good corporate citizen and has an estimated annual budget of several million dollars for charities." Hershey: Candy  the Trust's portfolio had always produced enough income for the School in spite of its concentration in Hershey stock. 145 The next day, Wednesday, September 4, the trial judge issued the preliminary injunction with an opinion to follow. 146 Both the timing and the result of the judge's decision appear to have been a surprise. The Wall Street Journal's Wednesday story about the Tuesday hearing quoted the judge as having said that he would rule by "week's end" and reported that "[m]any legal observers said the judge appeared unlikely to issue an injunction." 147 In heavy trading (10 million shares, more than four times the prior day's trading volume), HSY closed on Wednesday at $72.51, down $3.09 (4%) from Tuesday's closing price of $75.60.
On Thursday, September 5, the trustees appealed and the appellate court scheduled a hearing for Wednesday, September 11. 148 In their Thursday coverage of the trial court's ruling, both the Wall Street Journal and the New York Times reported that "legal experts" predicted that the trial court would be reversed on appeal. 149 HSY closed at $74.30 on Thursday, up $1.79 (about 2.5%) from Wednesday's closing price. Meanwhile, Fisher praised the trial court's ruling, saying that the injunction would allow him "to represent the public's interest" and would allow "the court to determine how a sale could hurt this community." 150 On Friday, September 6, HSY closed at $73.85.
On Monday, September 9, the parties filed their appellate briefs; 151 HSY closed at $73.53. The Attorney General conceded that the Trust was "imprudently" undiversified "and that it would be 'desirable' for the Trust to diversify its holdings," but he also argued that "there was no testi- 145 mony that it needs to do so immediately, within the next few days or weeks." 152 By contrast, "the current employees of Hershey Foods would be worse off under an acquisition than they are now," and the sale of the Company "would seriously impair, if not destroy, the symbiotic relationship which has existed for many decades among the company, the School and its Trust, and the other institutions which together carry on Milton Hershey's unique vision." 153 The Attorney General argued that the balance of equities therefore favored upholding the preliminary injunction. 154 "At its core," the Attorney General explained, "this case raises the question of whether the administrators of a charitable trust can operate the trust to inflict injury and harm upon the public at large, which is the ultimate beneficiary and real party in interest of all charitable trusts." 155 In response, the trustees argued that "'the public' benefit is not the personal economic or social benefit that 'the public' derives from the assets that the charitable trust holds-rather, 'the public' benefits from the achievement of the specific charitable purpose for which the trust is established." 156 Moreover, according to the trustees, nothing in the Deed of Trust indicated that Milton and Catherine Hershey wanted the Trust to maintain its ownership of the Company, or to undertake responsibility for the economic health of the local community, or to provide continuing employment for the Company's workers. 157 On the contrary, the only interest named in the Deed of Trust was that of the School. 158 The Deed of Trust also gave the trustees "full power and authority to invest" the trust assets. 159 Accordingly, because "portfolio diversification is the hallmark of any prudent investor," 160 because the Trust's concentration of investment in the Company made the Trust "twice as risky as the typical college or university portfolio," 161 and because nothing in the trust instrument made the Trust's investment in the Company "sacrosanct," 162 152. Brief of Appellee, supra note 110, at 11-12. 153 160. Id. at 34. Elsewhere in their brief the trustees argued that "if recent events involving the collapse of companies that were pillars of the communities in which they were headquartered . . . reveal anything, they reveal that no company, no matter how special and unique . . . , is immune from market forces and potential adverse economic, business, or financial developments." Id. at 39 (emphasis omitted). 161. Id. at 36 (emphasis omitted) (internal quotation marks omitted). 162. Id. at 21. the trustees argued that it "was improper for the trial court to enjoin" the sale process, which was both "lawful and prudent." 163 On Tuesday, September 10, the trial judge rendered a written opinion explaining the issuance of the September 4 injunction; HSY closed at $74.10. Calling Zimmerman's testimony "persuasive" and noting the "symbiotic relationship among the School, the community, and the Company," the judge found that a sale of the Trust's "controlling interest in the Hershey Foods Corporation creates a likelihood that there will be reduction in the work force and that relocations of plant operations and closing of duplicate facilities will be matters of probable immediate consideration by the acquiring company." 164 The injunction was therefore necessary to prevent "the adverse economic and social impact" that would result from a sale, "particularly in its effect on employees of the [Company] and the community of Derry Township." 165 Put more bluntly (and again quoting the trial judge), after paying a premium for the Company an acquirer would probably "introduce management efficiencies in order to cut costs to achieve an acceptable return . . . . The likelihood is great that these efficiencies will result in reduced work forces with a potential for plant location changes." 166 The judge noted two other factors favoring the injunction's issuance. First, the Trust's income had always exceeded the School's operating expenses. 167 The judge therefore found that the proposed sale "appears to be excessive and unnecessary for any foreseeable needs of the Trust." 168 Second, in the teeth of the volatility experienced by HSY on September 4 and 5, the judge rejected as "not . . . compelling" the argument that an injunction would "cause significant fluctuations" in the Company's stock "that [would] result in possible loss to the School fund." 169 The Attorney General made a similar argument in his appellate brief, adding also that "nothing suggests that the stock will be priced differently even if a sale does not occur." 170 The appellate court heard argument the next day, Wednesday, September 11. "Who in the courtroom," the trustees' lawyer asked, "has not read in the paper what happens in today's economy when you invest too heavily in a single stock?" 171 The Deputy Attorney General arguing the case replied that "[t]here is absolutely no support, and in fact the 163 According to subsequent media reports, the trustees' Tuesday session was "emotional," "rancorous," and "sometimes teary," with the trustees feeling "embittered" by what they perceived to be "pressure" from the Attorney General's office to diversify followed by the Attorney General's heated opposition. 177 Having become the objects of calumny, the trustees felt "overwhelmed by the outcry of protest from the community." 178 The chairman of the board of trustees, who had received death threats, had been living with an armed guard assigned to his home. 179 The trustees deliberated for ten hours. William Wrigley delivered a "moving" speech in which he discussed the significance of adding Hershey to his eponymous company's name and promised a commitment to the Hershey community. 180 Just before midnight, however, the trustees an- nounced that they had voted ten to seven to reject Wrigley's bid and all the other bids too. 181 The trustees' decision was met by Hershey residents and workers with a mix of relief and joy. "All I can say is hooray," one resident told the Associated Press. 182 "I still want this company to be around for my grandchildren, so they can work here when they're old enough," said another. 183 Kathy Taylor, a former town supervisor who helped spearhead the "Derail the Sale" movement, offered a more blunt assessment to the New York Times: "Our cash cow is safe; we're feeling really great." 184 In something of an anticlimax, also on Wednesday the appellate court upheld the trial judge's injunction by a four to one vote. 185 Investors took a dim view of the sale's cancellation. On Wednesday, September 18, HSY closed at $65.00, down $8.81 (almost 12%) from Tuesday's closing price of $73.81. Some 20 million HSY shares traded on Wednesday, ten times the number that traded on Tuesday.
The trustees did not fare well in the aftermath. On October 16, the Orphans' Court dissolved the injunction but ordered the trustees to give the Attorney General's office "prompt written notice" of any future intention to sell the Trust's controlling interest in the Company, 186 formalizing what the trustees had already promised in writing to do. 187 The judge also criticized the trust board as being too "large" and too "distant and disconnected from the charitable interests they serve." 188 On November 14, Fisher announced that the seven trustees who had voted in favor of continuing the sale, as well as three others who had opposed the sale, would be stepping down in favor of four new board members, all hailing from central Pennsylvania. 189 Fisher, too, faced disappointment-at least initially. His announcement of the change in the trust board came a week after he lost the gubernatorial election by almost 10%. 190 2003. 192 The aborted sale also affected Pennsylvania law, which already deviated from the national norm by exempting trustees of existing trusts from an explicit statutory duty to diversify. 193 On November 6, 2002, the Governor signed an amendment to the Pennsylvania prudent investor statute. The amendment requires the trustees of a charitable trust "in making investment and management decisions" to consider "the special relationship of [a trust asset] and its economic impact as a principal business enterprise on the community." 194 It also bars the trustee of a charitable trust "holding a controlling interest in a publicly traded business corporation received as an asset from the settlor" from selling that controlling interest without first notifying the Attorney General and the Pennsylvania employees of the business. 195 If the Attorney General challenges the sale, the amendment puts the burden on the trustee to "prove by clear and convincing evidence" that the sale "is necessary to maintain the economic viability of the corporation and [to] prevent a significant diminution of trust assets or to avoid an impairment of the charitable purpose of the trust." 196 The amendment constitutes a radical departure from orthodox principles of prudent trust investment law. 197  See, e.g., Fremont-Smith, supra note 3, at 193 ("[The amendment] represents an unusual degree of interference with trustee discretion, not to say a perversion of the intent of many donors."); Sidel, Struggle for Hershey, supra note 8, at 44-48 (describing amendment as "position seemingly taken by no other state"); Dale, supra note 8, at 18-19 (describing skeptical commentary regarding amendment as "an understatement"); see also infra note 332 and accompanying text (discussing orthodox hostility, as embodied in Uniform Prudent Investor Act, to consideration by trustees of considerations other than beneficiary's interests).
As detailed in Table 1 above, in spite of occasional share repurchases by the Company, 199 the Trust remains overwhelmingly invested in the Company and, hence, manifestly undiversified.

B. Charitable Trusts
A private trust is a fiduciary relationship in which the trustee holds legal title to specified property, entrusted to him by the settlor, and manages that property for the benefit of one or more beneficiaries. 200 Hence, the trust presents the standard agency problem that arises when risk-bearing (the beneficiaries) and management (the trustee) are separated. 201 To safeguard the beneficiary from mismanagement or misappropriation by the trustee, trust law supplies a set of default terms known as fiduciary duties that prescribe the trustee's level of care (the duty of prudence) and proscribe misappropriation (the duty of loyalty). 202 A beneficiary who believes that the trustee acted disloyally or imprudently may sue the trustee for breach of trust. Moreover, because trust default law makes it difficult for the beneficiary to remove the trustee, and because the beneficiary's interest is typically inalienable (i.e., there is no market for trust control), the threat of fiduciary litigation is the primary force for aligning the interests of the trustee and the beneficiary-that is, for minimizing agency costs in the modern private trust. 203 199. See, e.g., 2008 Annual Report, supra note 9, at 33 (describing volume of shares repurchased by Company from Trust over preceding several years The paramount role of fiduciary law in minimizing agency costs in trust governance explains the traditional rule that a private trust must be for the benefit of an ascertainable beneficiary. 204 Requiring an ascertainable beneficiary ensures that there is someone with an economic incentive to enforce the trustee's fiduciary duties. 205 Unlike a private trust, however, a charitable trust must be for the benefit of a charitable purpose such as advancing education or the relief of poverty (hence the validity of the Trust)-not for a specific beneficiary. 206 "Even if individuals receive direct benefits from a charitable trust, such as scholarship grants, money, food, clothing, or any other direct assistance, they are considered the 'conduits of the social benefits to the public and are not in reality the beneficiaries of the trust.'" 207 Hence, for a charitable trust there is no identifiable beneficiary with an economic incentive and legal standing to ensure "that the trustee acts in accord with the settlor's charitable purpose and refrains from abuse or breach of fiduciary obligation." 208 Evelyn Brody aptly frames the resulting agency problem: "In the case of an entity having no owners and established for the benefit of indefinite beneficiaries, who is the principal on whom the law can rely to monitor the agents and enforce the charitable purposes?" 209 The traditional answer to the problem of agency costs in charitable trusts is to vest the state attorney general, as parens patriae, with standing  to enforce such trusts. 210 Indeed, many states have broadened the attorney general's common law enforcement power to include the power to investigate the operation of charitable entities. 211 Most states also require charitable trusts and other charitable entities to make regular reports to the attorney general's office (a disclosure requirement backstopped by federal law applicable to tax-exempt entities). 212 Furthermore, in many instances, the state attorney general is a necessary party in litigation involving a charitable trust or other charitable entity. 213 The state attorney general, however, is a political official, typically elected, with neither a personal financial stake nor, in the usual case, a political stake in the operation of a charitable trust. Most state attorneys general assign few (if any) lawyers to supervision of charities. 214 Unless an alleged breach of trust obtains enough media attention to achieve political salience, actual scrutiny of a charitable trust by the attorney general is unlikely. As a result, it is the politically salient, egregious cases that "trigger investigations," not "reviews of annual reports." 215 In the usual case there simply is not enough of a political payoff to the attorney general to warrant the diversion of resources from other initiatives. 216 The mirror-image worry, recently developed nicely by Evelyn Brody, is that when the attorney general does intervene in response to political pressure, he or she will be tempted to promote his or her political interests at 210 the expense of the trust's charitable purpose. 217 Brody provides the apt summation: "Political cynics believe that 'A.G.' stands not for 'attorney general' but for 'aspiring governor.'" 218 Accordingly, a diverse array of scholars have theorized that supervision of charitable trusts by the attorneys general is either lackadaisical, in which case the trustees will lack an incentive to manage the trust's assets in an efficient manner, 219 or perverse, entailing imposition of local political preference irrespective of whether those preferences are "congruent with the donor's conception of the beneficiary class." 220 The prevailing scholarly view, in other words, is that agency costs are rampant in charitable trust governance. 221 217. See Brody, Parochialism, supra note 8, at 968-79. In Brody's words: "Occasionally, though, we find the reverse problem: a board trying to do the right thing, but thwarted by an overreaching regulator." Id Although there is anecdotal evidence that is consistent with the widely held assumption of agency costs in charitable trusts, 222 there exists no quantitative empirical study of the issue. The Trust's 2002 aborted sale of the Company provides an opportunity to undertake such a study. Specifically, daily stock price data from before, during, and after the aborted sale provide a measurement of the value of a prominent charitable trust's principal asset (i.e., a controlling interest in the Company) while in the hands of the trustees as opposed to when that asset was for sale on the open market. These price data also allow for assessment of the broader welfare effects of an intervention by a state attorney general in a publicly salient case.
We posit that effective trustees would already have realized for the Trust the benefits of holding a controlling interest in the Company by using their controlling vote block to impose a value-maximizing strategy on the Company's managers. Such efficiency-inducing actions would inure to the benefit of all the shareholders in the form of a higher stock price. These assumptions yield a testable hypothesis: After controlling for general market trends, the sale announcement should not have had a positive effect on the value of Hershey shares not held by the Trust. On the contrary, a sale would end the ability of the other shareholders to free ride on the trustees' value-inducing efforts. Thus, if the trustees had been successfully imposing a value-maximizing strategy on the Company's managers, we would expect if not a negative then at least a neutral price effect from the sale announcement. By contrast, a positive effect would imply that the takeover threat associated with the market for corporate control better disciplined the Company's managers than the trustees in spite of the trustees' controlling vote block.
We must acknowledge, however, two important limitations in the foregoing research design: (1) a positive result might be consistent with an alternative hypothesis, and (2) our study is by necessity limited to a single charitable trust. We consider each limitation in turn.
1. Alternative Hypotheses. -A positive finding is arguably consistent not only with the charitable trust agency cost hypothesis that we state above, but also with two alternative hypotheses: (a) escape from the trust- ; see also Fishman, Faithless, supra note 96, at 3 n.1 (collecting "contemporary scandals involving charities"). In a related vein, there have been occasional studies detailing the insufficiency of resources allocated to charity enforcement, see supra note 214, and studies of agency problems in not-for-profit firms, see, e.g., Core et al., supra note 39, at 32-33 (finding agency costs to be primary force driving accumulation of excess endowments). ees' non-profit-maximizing objectives, and (b) the control premium phenomenon. 223 a. Non-Profit-Maximizing Objectives. -The first alternative hypothesis is that the trustees' controlling interest depressed the stock price of the Company not because the trustees were lax in monitoring the Company's managers, but rather because the trustees imposed policies that favored the interests of the Trust to the detriment of the other shareholders. On this account, which comes in strong and weak forms, the sale announcement would lead to an increase in the price of the Company's stock by freeing the Company's managers to pursue a profit-maximizing strategy.
The stronger but less plausible form is that the trustees extracted unequal distributions or other payments to the Trust not shared pro rata with the other shareholders. To have affected the stock price positively on news of the sale, however, the fact of inequitable treatment must have been publicly available. Yet our review of the public record for the relevant timeframe uncovered no evidence of such disproportionate extraction. Further, an accepted principle of corporate fiduciary law holds that a controlling shareholder may not secure a pecuniary or other benefit from the company that is not shared proportionally with the company's other shareholders. 224 Had other market participants known of such private benefits, litigation likely would have ensued.
The weaker but more plausible version of this alternative hypothesis is that the trustees imposed a cautious, risk-averse style of management on the Company that favored stable dividends and low risk projects at the expense of the sort of higher risk (but positive net expected value) projects that in the long run are more likely to maximize the return on the firm's underlying assets. For example, suppose the Company could pursue project A or project B, but not both. Project A offers a certain return of $100 while project B offers a 60% chance of a $200 return (hence an expected return of $120). Diversified shareholders would prefer project B because in the aggregate such shareholders will win often enough to offset the less frequent occasions that they lose. 225 However, 223. Given the Company's dual class stock arrangement, see supra note 17 and accompanying text, it might be thought that changes in the relative voting and dividend rules occasioned by the sale might affect the Company's stock price. However, the trustees instructed the Company to solicit bids for a takeover that would be implemented by a merger. In such a merger, all the Company's shares would be exchanged for cash or for cash and stock in the acquiring company. Because all Hershey shares would disappear at the same time, there would be no change in relative rights among the Hershey stockholders.
224 because the Trust is not well diversified, the trustees might prefer the certainty of project A in spite of its lower expected value. 226 When viewed in this manner, however, this alternative hypothesis is revealed to be merely a reformulation of ours. The trustees failed to maximize the value of the Trust's most significant asset, its shares in the Company, by sacrificing value for a reduced level of risk that could have been achieved in an alternative fashion without any loss in value. 227 If the trustees judged the risk associated with a profit-maximizing Company to be too great for the Trust to bear, then they could have sold the Trust's shares, realizing the increased price from the market's expectation of the Company's switch to profit-maximization once free of the Trust's control. The trustees could then have invested the resulting larger Trust corpus in a diversified portfolio with the desired balance of risk and return.
The key point is that regardless of the mechanism for the depressed price of the Company's stock-lax monitoring of the Company's managers or imposition of something other than a profit-maximizing strategythe implication of a positive finding is the same. The trustees unnecessarily sacrificed value, reducing the assets available for the Trust's charitable purpose while at the same time exposing the Trust to uncompensated risk.
b. The Control Premium Phenomenon. -The second alternative hypothesis posits that a positive result could reflect the control premium phenomenon. A controlling interest in a publicly-traded company sometimes has a control premium attached to it independent of any beliefs about the firm's underlying value. On this account, the trustees might have been active, efficient monitors of the Company's managers, but owing to a potential buyer's managers' desire to build a larger empire, to become too big to be a takeover target, or to acquire the Company for some other reason, a buyer might nonetheless value a control block in the Company more highly than the sum of the market's valuation of the individual shares. 228 A purchaser also might be willing 226. See D'Antonio, supra note 8, at 246 (suggesting that trustees were "interested in minimizing risk and pressed the company for consistent performance rather than quick profits"). 227. A closely related scenario involves the trustees failing to diversify because they anticipated political opposition and judged the costs of a diversification effort to outweigh its probable benefits. On this account, for which we hasten to add that there is little supporting evidence, the problem was not agency costs in the Trust's management, but rather defects in the political economy of the supervision of charitable trusts. Either way, the social welfare loss is the same and traces to structural weaknesses in the charitable trust form.
228. In this account, the control premium stems from agency costs within the acquiring firm. See to pay a premium in anticipation of the synergies created by the merger. 229 However, the control premium possibility bears only on the mechanism for, not the existence of, agency costs in the Trust's management. Because the trustees already had a controlling interest in the Company, they should have exploited the existence of a potential control premium for the Trust's charitable purpose, if not directly by imposing superior management on the Company (our original hypothesis), then indirectly by selling the Company to a buyer that for whatever reason values control more highly (this alternative hypothesis). Under either scenario, the trustees left money on the table that could have been realized for the benefit of the Trust's charitable purpose.
2. A Single Charitable Trust. -The second limitation in our research design is that our study encompasses only a single charitable trust. Here the worry relates to our econometric methodology (i.e., internal validity) and to the generalizability of our results (i.e., external validity).
With respect to internal validity, the concern is that, with only a single trust, the noise to signal ratio might be such that we ascribe statistical significance to what is in fact a random fluctuation in the data. To address this concern, we undertake a host of robustness checks, several of which are designed specifically to address the problem of noise in a sample of one. 230 We may also take comfort from the fact that the magnitudes of some of our findings are so large that, as a mathematical matter, they cannot be artifacts of random fluctuation in the data. 231 Regarding external validity, the concern is that our findings with respect to the Trust might not be representative of other charitable trusts more generally. Here the tension is between the quality of the natural quasi-experiment (the more unusual the circumstances, the better the experiment) and the generalizability of the results. Given the unique shock of the trustees' exposure of the Company to the market for corporate control, we must guard against reading our results more aggressively than our research design can justify.
Accordingly, in our view a positive result would indicate agency costs in the Trust structure and would tend to support the large body of theoretical literature that supposes agency costs in charitable trusts more generally. We cannot, however, generalize about the magnitude of agency costs in other charitable trusts from the magnitude of our specific Hershey finding, as each charitable trust operates in a different context. Still, a very large positive finding might imply that the problem is worse than previously believed, as the Trust has a high profile in Pennsylvania. A negative result, by contrast, would indicate that this charitable trust was well managed. But given the Trust's political salience in Pennsylvania, a negative result would not negate the claim in the theoretical literature that lower profile charitable trusts are often poorly managed.

C. Corporate Governance
Minimizing agency costs is also a central concern of corporate law. In the corporate context, agency problems arise from what Berle and Means famously dubbed the corporate form's "separation of ownership and control"-that is, the separation of the risk-bearing residual claim, which is held by the shareholders, from managerial authority, which is held by the directors and senior management. 232 To align the incentives of managers with the interests of shareholders, or at least to constrain the extent to which managers can advance their own interests at the expense of shareholders, the law gives shareholders information rights, the right to sue, the right to vote on the composition of the board and to approve certain critical transactions, and the right to sell their shares. 233 With respect to the large, publicly traded corporation, the prevailing scholarly view, which is supported by abundant empirical evidence, is that shareholder litigation and other such control mechanisms are weak devices for minimizing corporate agency costs when compared to the market for corporate control (i.e., the threat of a takeover). 234 Collective action problems plague diffuse shareholders, leaving each individual shareholder hesitant to engage in litigation, which is in effect a public good. A litigating shareholder will bear most of the litigation costs, but because the typical shareholder owns a tiny percentage of the company, the resulting gains from reducing corporate agency costs will accrue 232 mainly to the other shareholders. 235 The same intuition applies to other internal control mechanisms such as the right to vote. No individual shareholder has a real incentive to fight for controls that would benefit all shareholders or to engage in the sort of active monitoring that leads to informed voting. 236 The analysis is different when one considers the potential disciplining force of the threat of a takeover. If managers do not maximize the return on a firm's assets, the firm's stock price will be depressed. This depressed stock price represents a profit opportunity for an investor or group of investors that acquires a controlling interest in the firm and replaces the existing management team with one that maximizes the return on the firm's assets. The payoff is the resulting appreciation in the firm's stock price. Accordingly, by threatening managers with the loss of their jobs in the event that they shirk or underperform, the market for corporate control minimizes corporate agency costs. 237 On this account the right to vote and the right to sell are essential to good corporate governance not in their own right, but because they facilitate the operation of the market for corporate control.
An alternative disciplining mechanism arises when the firm has a controlling shareholder or a minority blockholder. Unlike a typical shareholder, for whom diversification plus passivity is a rational strategy, a controlling shareholder might have a sufficient financial incentive to monitor actively and, if necessary, impose new management or new managerial strategies on a company. 238 Thus, some scholars and policymakers have argued that increased activism by controlling shareholders will improve corporate governance. 239  discipline provided by a controlling shareholder will depend on the controlling shareholder's incentives and, if the controlling shareholder is itself an entity, the quality of the controlling shareholder's internal governance scheme. On this view, an individual or a well-managed, profitmaximizing hedge fund 240 might be a more effective monitor than a pension fund 241 or mutual fund run by managers with an attenuated stake 242 in the ultimate performance of the fund.
The aborted 2002 sale of the Company by the Trust provides a fortuitous natural quasi-experiment on the comparative virtues of a controlling shareholder versus the market for corporate control as alternative mechanisms for minimizing corporate agency costs. If the two mechanisms are equally effective in disciplining management, we should find no difference in the value of the Company during the sale window and during the periods immediately before and after that window. By contrast, if the market for corporate control is superior to the control provided by concentrated ownership, whether because blockholders are generally ineffective, because the managers of this particular blockholder (i.e., the trustees) have poor incentives owing to the agency problems that are inherent to the charitable trust form, or some combination of the two, we should find that Hershey's market value improves during the sale window. If the market for corporate control is inferior to control by the Trust in disciplining management, we should find that Hershey is worth less during the sale window.
As in our analysis of charitable trust agency costs above, we must again acknowledge potential limitations in our research design concerning: (1) alternative hypotheses and (2) reliance on observations of a single firm.
impractical because of requirements of risk-increasing concentration of investment and large initial expenditure); see also 1. Alternative Hypotheses. -We cannot disentangle the role of the Trust's particular governance structure from the role of switching from a controlling shareholder to the market for corporate control as alternative explanations for the Company's price dynamics in and around the sale window. Thus, if we have a positive finding-that is, if we find that the Company is more valuable when subject to the takeover market than when under the thumb of the trustees-we will know only that the market rated the threat of a takeover as superior to control by this controlling shareholder. Although such a result would highlight the problem of agency costs within a controlling shareholder's organizational form and would lend modest support to the takeover model, it would not resolve whether the takeover market is preferable to another controlling shareholder with better internal governance (for example, a hedge fund).
Further, to the extent that a positive finding is traceable to agency costs in this controlling shareholder, per our discussion in the prior section of alternate hypotheses relating to charitable trust agency costs, we cannot determine the specific form of the Trust's agency costs (e.g., lax oversight by the trustees, imposition of something other than a profitmaximizing strategy by the trustees, etc.).
In sum, our analysis will add to the stock of empirical knowledge on controlling shareholders and the takeover market as alternative devices for minimizing corporate agency costs, but the interpretation of our findings must be sensitive to our inability to disentangle how much of our findings are attributable to general versus specific features of the Hershey incident.
2. A Single Firm. -As before, we must consider the possibility that we might ascribe statistical significance to what is in fact random variation in the data because we examine only a single firm. 243

III. EMPIRICAL ANALYSIS
We examine the Company's share prices before, during, and after the period in which the trustees announced their intention to sell the Trust's controlling interest in the Company. Our analysis proceeds in four steps. First, in Part III.A, we describe the nature and sources of our stock market data. Second, in Part III.B, we undertake a graphical analysis of the data for the Company by itself and in comparison to its leading competitors and to the stock market as a whole. The stock price effects of the trustees' plan to sell and abandonment of the sale are so profound that simple graphical depictions are highly suggestive. Third, in Part III.C, we present the results of a more formal event study econometric analysis that measures stock price movements traceable to sale-related events by isolating them from movements related to ordinary market volatility. Finally, in Part III.D, we offer a nontechnical summary of our main 243. See supra text accompanying notes 230-231. empirical findings. Readers interested in our results, but not the formal methodology, will find Parts III.B and III.D of principal interest.

A. Data
We examine daily closing price data for the Company's common stock (ticker symbol HSY), the common stocks of Hershey's closest competitors that are traded on the New York Stock Exchange (NYSE), and a variety of market indexes, including the Standard and Poor's 500 (S&P 500) and the Dow Jones Industrial Average (DJIA). Our timeframe includes the period during which the sale was considered (that is, between July 25, 2002, and September 18, 2002), two years preceding the sale window, and two years following the sale window. We include two years before and after the sale window to put the observed price movements during the window into historical perspective.
We obtained our stock price data from the Center for Research in Security Prices (CRSP) at the University of Chicago Graduate School of Business. 244 We accessed these data through Wharton Research Data Services, 245 an interface maintained for academic use at the University of Pennsylvania.
A problem with raw stock price data spanning several years is that price movements may reflect nothing more than a stock split or a dividend payment, neither of which affect shareholder wealth. 246 To avoid distortion from stock splits and dividends in our examination of stock prices over time, in the graphical and event study analyses that follow we use adjusted closing prices provided by CRSP that account for stock splits and dividends.

B. Graphical Analysis
On July 25, 2002, the day the Wall Street Journal broke the news of the trustees' plan to sell, HSY closed at an unadjusted price of $78.30, an increase of $15.80 (more than 25%) from the prior day's closing price of While these price changes seem striking, it is possible that they are only coincidentally related to news of the sale and its abandonment. For example, the increase at the beginning of the sale window might reflect a preexisting upward trend in HSY, and the decline at the end of the sale window might be picking up the beginning of a downward trend. Another possibility is that these price movements, large though they may be, were the result of normal random variation in the price of HSY-that is, perhaps HSY is a volatile stock and this volatility manifested itself positively on July 25 and negatively on September 18. 247 Still another possibility is that these price changes could reflect general market movement common to all similar stocks on those days.
To rule out the possibility that the identified price movements are the result of existing trends or normal variation, we examine in Figure 1 the adjusted closing price of HSY during the sale window and for fifty trading days before and after the sale window. For a broader perspective, Appendix Figure A1 presents 500 trading days before and after the sale window, a practice we repeat for all the graphs in this section.   Figure 1 demonstrates that the price increase at the beginning and the decrease at the end of the sale window were not artifacts of preexist-247. As seen in the graphs that follow, as well as the subsequent regressions, HSY returns are not particularly volatile, which suggests that this possibility is unlikely. ing trends. On the contrary, the price of HSY was trending downward before July 25, 2002, and began an upward trend not long after the trustees abandoned the sale. 248 Another possible explanation for the observed HSY price movements is coincidental change in the chocolate market. To examine this possibility, Figure 2 and Appendix Figure A2 add contemporaneous price movements of three firms that compete with Hershey in the chocolate market: Cadbury Schweppes (CSG), Tootsie Roll (TR), and Rocky Mountain Chocolate Factory (RMCF). 249 These firms 250 represent Hershey's principal NYSE-traded competitors with important chocolate products. 251 We consider the potential endogeneity problems with using the performance of a competitor as a control in connection with our formal event study analysis in Part III.C below.
None of Hershey's chocolate market competitors experienced a comparable stock price jump on July 25, 2002, and none dropped as precipitously on September 18, 2002. 252 Hence, we infer that the observed HSY price movements on those dates were not related to peculiar coincidental factors in Hershey's primary product market.
The Company's business is broader, however, than just chocolate. In 2002 Hershey "also manufacture[d] and/or market[ed] grocery products in the baking, beverage, peanut butter and toppings categories." 253 Thus, in Figure 3 (and Appendix Figure A3) we compare HSY price 248. Appendix Figure A1 shows that the price of HSY returned to the sale window level about a year later, but per Appendix Figure A2, the subsequent upward trend in HSY resembles the upward trend experienced contemporaneously by other chocolate companies.
249. The RMCF stock price is plotted against the right hand axis in Figure 2 because its price range during the entire period was substantially lower than that of the other stocks.
250. A similar picture would emerge if we were to create a single portfolio of Hershey's competitors and plot that portfolio's return against the HSY return. 251. We took our roster of Hershey's chocolate rivals from Google Finance, The Hershey Company: Related Companies, at http://finance.google.com/finance?q=hsy (last visited Mar. 3, 2008) (on file with the Columbia Law Review), and Yahoo! Finance, Hershey Co.: Competitors, at http://finance.yahoo.com/q/co?s=HSY (last visited Mar. 3, 2008) (on file with the Columbia Law Review), which list these firms as Hershey's primary publicly traded competitors in the chocolate market. The other obvious contenders are Mars and Nestlé, but Mars is a privately held firm, so it has no public stock price data, and Nestlé is not traded on the New York Stock Exchange. While each of these competitors produces more than just chocolate, we categorize them as chocolate competitors because they do not reach the scale or scope of production of the firms that we categorize as food competitors in the analysis that follows in Figure 3.
252. Although TR does appreciate on July 25, 2002, the increase is small relative to the normal variation in TR's price. Likewise, RMCF appears to experience potentially interesting price movement in the middle of the sale window, though this movement also is small relative to the normal variation in RMCF's price. Our econometric analysis in Part III.C yields results consistent with these interpretations.  As in the chocolate market comparisons, we find no indication that trends in the food industry drove the observed HSY price movements during the sale window. For each of the comparison stocks, prices are largely flat throughout the sale window. Accordingly, we infer that the observed HSY price movements around the sale window were not related to peculiar coincidental factors in the food industry.
Finally, we examine whether general market trends might explain the observed HSY price dynamics. In Figure 4a and Appendix Figure  A4a, we plot HSY against the Standard and Poor's 500 market index (S&P 500 plotted against the right-hand axis). In Figure 4b and Appendix Figure A4b, we do the same for the Dow Jones Industrial Average (DJIA plotted against the right-hand axis). 255 254. As in the chocolate market, we took our list of rival food firms from Google Finance and Yahoo! Finance. See supra note 251. As before, some companies that would seem like good comparisons (e.g., Frito Lay) are not suitable because they are not publicly traded.
255. A similar picture would emerge if we plotted HSY against CRSP's value-weighted market index, which is often used in the finance literature as a representative market portfolio. See  The differences between HSY's price dynamics and those of the market indexes are not as striking as in the comparisons in Figures 2 and 3. Both the S&P 500 and the DJIA experienced price appreciation toward the beginning of the sale window and depreciation toward the end of the sale window. These market movements do not correspond precisely with that of HSY, however, as the market indexes began their rise after July 25, 2002, and started their downward trend before September 18, 2002. Further, in neither case is the relative price change as great as that experienced by HSY. Nonetheless, the observed comovement of the market indexes and HSY casts some doubt on the causal relationship between the Trust's proposed sale and the movement of Hershey's stock price. 256 To sort through these possibilities, in the next section we employ a more formal event study econometric analysis that separates the effect of salerelated news from general market trends.

C. Event Study Analysis
1. Introduction. -The comparisons of HSY and the market indexes give us pause in attributing the HSY price dynamics during the sale window to sale-related news. To untangle the effect of general market trends from sale-related news, we make use of an event study methodology that is well accepted in the literature of financial economics. 256. Although Hershey is included in the S&P 500, see Yahoo! Finance, Components for S&P 500 Index, at http://finance.yahoo.com/q/cp?s=%5EGSPC&alpha=H (last visited Apr. 5, 2008) (on file with the Columbia Law Review), it accounts for less than 0.2% of the S&P 500 index. The movement of the DJIA, which does not include HSY, follows the same general pattern.
Two ideas underpin the market model event study methodology that we employ below. 257 The first is the efficient capital markets hypothesis (ECMH). In its semi-strong version, the ECMH posits that the price of a publicly traded security reflects all public information on the present value of the future cash flow associated with ownership of the security. 258 Hence, in a semi-strong efficient market, only nonpublic information such as an unanticipated event will affect securities prices, and traders cannot exploit publicly available information for gain. Until recently, the nearly universal view among financial economists was that the domestic stock market is semi-strong efficient; and while some have begun to question this view in the last few years, semi-strong efficiency remains the prevailing view. 259 The second idea is that in the short run the relationship between an individual security and the market as a whole is relatively stable. 260 On this assumption, it is possible to assess in retrospect whether on a given day a particular stock experienced an "abnormal" price movement, meaning a change in price that deviates in a significant way from the stock's typical relationship with the market as a whole. Importantly, we do not claim that it is possible to predict a stock's future price movements based on past performance. Rather, the claim is that on Tuesday we can judge whether a particular stock's price movement on Monday was "normal" or "abnormal" for that stock in relation to the known movement of the market as a whole on that same Monday.
Taken together, these two ideas imply that it is possible to assess empirically the effect of an unanticipated event on a company's stock price. Consider a grossly simplified, stylized example. Suppose that for 100 days the price of Company A's stock increased or decreased in perfect concert with the market as a whole. Then on day 101 A's managers announce a new business plan and the stock goes up 12%. If the market as a whole went up 5% on that day, so that a "normal" movement of A's stock would 257 have been a 5% increase, we could attribute the "abnormal" difference of 7% to the unanticipated news of the new business plan.
2. HSY Event Study. -Implementation of the foregoing intuitions in a market model event study may be broken down into five steps: 261 (1) define the event and identify the date or dates on which information of the event became public; (2) measure the affected security's actual return on the dates of interest; (3) estimate the security's expected return on the dates of interest using historical data on the relationship of the affected security to the market as a whole; (4) compute the abnormal return by subtracting the expected return from the actual return; and (5) assess the statistical significance of the abnormal return. Once these five steps are completed, it is possible to evaluate the economic significance, if any, of the abnormal return on the dates of interest. We take each step in turn.
(1) Our event is the aborted sale of the Trust's controlling interest in the Company. Hence, we focus on the thirty-eight trading day sale window that began on (a) The impact of the July 25 sale announcement tests our hypotheses regarding agency costs in charitable trust governance and the comparative virtues of a controlling shareholder versus the market for corporate control in minimizing agency costs in corporate governance. For the reasons explained earlier, we predict a large, positive abnormal return on that date.
(b) The impact of the September 4 injunction provides a test of the welfare effects of the Attorney General's intervention. The September 4 injunction also provides a robustness check on our analysis of the July 25 sale announcement. Because the injunction reduced the probability of a sale, but did not extinguish the possibility of one, we predict a negative 261. Our five-step statement of the mechanics of an event study draws on the four-step statement in Bhagat & Romano, supra note 29, at 948-51, except that for expositional clarity we have broken their step four into two separate steps.
262. See supra notes 108, 146, 181 and accompanying text. We do not test the effect of the subsequent amendment to the Pennsylvania prudent investor rule, see supra text accompanying notes 193-198, for two reasons. First, prior to the amendment, the trustees committed in writing not to sell the Company without first notifying the Attorney General's office, and the trial court included this requirement in its October 16 order dissolving the preliminary injunction. See supra notes 186-187 and accompanying text. Second, it is hard to isolate a clean event date for legislative enactments without introducing bias. See Bhagat & Romano, supra note 29, at 949 (describing potential effect of various announcements throughout legislative process). Here the Attorney General made various public announcements about supporting the legislation, which then passed each house of the Pennsylvania legislature on different dates, and then was signed by the Governor on still another date. See Sidel, Struggle for Hershey, supra note 8, at 39-44. abnormal return that is smaller in absolute magnitude than July 25's abnormal return.
(c) The impact of the trustees' September 18 abandonment of the sale provides a second test of our agency cost hypotheses, both as a robustness check on our finding for July 25 and to exclude the confounding possibility that the fact of the sale announcement itself supplied new information about the Company. 263 We predict a negative abnormal return that, given the prior injunction, falls in absolute magnitude between that of the September 4 and July 25 dates. 264 Table 2 summarizes the key dates of interest and our predicted impacts of the events of those dates. (2) The next step, which is relatively simple, is to measure the actual return of HSY during the sale window. Actual return is routinely calculated as the price of the stock at time t + 1 minus the price at time t divided by the price at time t. Using adjusted daily closing prices, in the third column of Table 3 we present HSY's actual return for each trading day within the sale window.
(3) The third step, estimating the security's expected return, is more complicated. We use a statistical market model that, as is customary, assumes that HSY's expected return may be estimated as a linear function of the return of the market as a whole. 265 This relationship is expressed as follows: 263. See infra notes 357, 359 and accompanying text. 264. In theory, if there were no other indications during the sale window that the trustees might abandon the sale, we would have expected the September 18 abnormal return to equal the July 25 abnormal return, except oppositely signed. However, the Attorney General's actions during the window, as well as those of the courts and members of the local community, decreased the probability of a sale. Inasmuch as this expectation would have been capitalized (probabilistically) into the HSY price, we expect a large negative abnormal return on September 18, but one that is smaller in magnitude to that observed on July 25.
265. There are three notable alternatives to the statistical market model. Two are economic models (the capital asset pricing model and the arbitrage pricing theory) and the third is another statistical model (the constant expected returns model Where E (HSY t ) is the expected return for HSY during time period t, l HSY is an HSY-specific constant, f HSY is an HSY-specific coefficient (i.e., a measure of how the HSY return has varied historically in relationship to the market return), and Market t is the market return over timeframe t. To compute the expected return for HSY during the sale window, we multiply the market return during the window by f HSY and then add l HSY .
To estimate l HSY and f HSY , we perform an ordinary least squares regression on actual HSY returns for some period before the sale window (the "estimation period") using the returns for a market portfolio as a predictor variable. For our market portfolio, we use the S&P 500. 266 For our estimation period, we use 100 trading days prior to the event (i.e., the sale window), which is a common estimation period in event studies that use a statistical model. 267 In the fourth column of Table 3, we present our estimated expected return for HSY during each of the trading days during the sale window. In unreported regressions we also used an estimation period of 200 days before the sale window with similar results.
A potential criticism of using data prior to the sale window to estimate expected returns during the sale window is that the sale window might have coincided with a structural change in the relationship between HSY and the market portfolio. If so, the historical relationship between HSY returns and the market return would no longer hold. As a check on this possibility, we also computed expected returns for HSY during the sale window using estimation periods of 100 days after the sale window, 200 days after the sale window, and a pooled model that included 100 days before and 100 days after the sale window. Each of these alternate approaches yielded results similar to the 100-days-prior results that we report below.
Another potential concern is that during the 100 trading days prior to the sale window, Hershey factory workers went on strike (which lasted from April 26, 2002, to June 8, 2002). 268 To determine whether the strike introduced bias into our analysis, we investigated a 100 day estimation period that excluded the strike period as well as an estimation model that included a control variable for days during the strike. In both cases our results were little changed, 269 which is consistent with Hershey's own evaluation of the effect of the strike. In its quarterly filing with the SEC for the period that included the strike, the Company reported that "[t]he Fin. Econ. 3, 25-26 (1985). The choice of expected return model did not affect the results, however, as our findings were largely unchanged when we used a constant expected returns model or one of the economic models.
266. The principal alternatives to the S&P 500 are the DJIA and the CRSP index. The results were largely unchanged, however, when we used these alternatives.
267 work stoppage did not have a material impact on the Corporation's results of operations for the second quarter." 270 (4) The next step is computing the abnormal return for HSY during each trading day in the sale window. The abnormal return is computed by subtracting the expected return from the actual return. The fifth column of Table 3 presents our abnormal return estimates.
(5) The final step is to assess the statistical significance of the abnormal return. By statistical significance we mean a measure of the likelihood that the observed abnormal return reflects random variation rather than the influence of the event under study. A typical significance threshold is 5%, with 2.5% of abnormal returns falling below a negative critical value and 2.5% of abnormal returns falling above a positive critical value that is symmetric to the negative critical value. That is, if the HSY return were to follow its historical relationship with the market return and not be affected by an unanticipated occurrence, then on any given day its observed abnormal return would fall between our positive and negative critical values 95% of the time. The alternative possibility that the day in question experienced an abnormal return owing to random variation would be only 5% likely. Accordingly, if a given day's abnormal return falls outside of these critical values, we may infer that it is very unlikely that the day's abnormal return comes from the same distribution as the abnormal returns from the estimation period. Critical values may also be computed for significance thresholds other than 5%, such as 1% or 0.1% (implying 99% and 99.9% likelihood respectively).
In the sixth column of Table 3 we present the standardized abnormal return for HSY during each day in the sale window. The standardized abnormal return is the ratio between the abnormal return and the standard deviation of the abnormal returns during the estimation period. 271 For standardized abnormal returns that exceed 1.96 or are less than -1.96, there is less than a 5% chance that the observed abnormal return reflects random variation. That percentage falls to 1% for standardized abnormal returns above 2.575 or below −2.575, and to 0.1% for those above 3.277 or below −3.277. 272 In Table 3  271. There is concern in the event study literature that the standard deviation of abnormal returns during the estimation period is not a useful measure of variation because of increased volatility around potentially important events. See, e.g., Bhagat & Romano, supra note 29, at 951 (discussing problem). Hence, we examined a number of alternative measures of variation including the standard deviation of abnormal returns during the sale window. We also estimated expected returns using a GARCH model that uses a time series technique that allows for a more general variance structure. See James D. Hamilton, Time Series Analysis 665-71 (1994). Our results are not affected by these different measures of variation.
272. These significance levels assume that the abnormal returns will be distributed normally with a mean of zero. But in a one-firm event study the abnormal returns might that follow, we denote statistical significance at the 5, 1, and 0.1% levels with *, **, and *** notations respectively.
Our primary date of interest is July 25, 2002, the day in which the Wall Street Journal broke the news of the trustees' plan to sell. As we have seen, on July 25 the price of HSY jumped to an unadjusted closing price of $78.30 from the prior day's closing price of $62.50. This increase translates to an adjusted actual return of 0.253 on a day that, 273 given the movement of the market as a whole, HSY would have been expected to experience a return of −0.002. So the abnormal return for HSY on July 25 was 0.255, which translates to a standardized abnormal return of 21.25. Accordingly, HSY experienced a large, abnormal positive return on the date news of the trustees' plan to sell became public, and it is extremely unlikely that this finding reflects random variation instead of news of the sale plan (the finding is statistically significant at less than the 0.1% level). Indeed, excluding September 4 and September 18 (the dates on which the sale was enjoined and abandoned), the standardized abnormal return for July 25 was 8.5 times larger than the second largest and more than twelve times larger than the third largest standardized abnormal returns for any other day in the sale window.
Our second date of interest is September 4, 2002, the day that the trial court enjoined the sale. HSY dropped to $72.51 from the prior day's closing price of $75.60, which translates to an adjusted actual return of −0.041. The expected return for HSY was 0.006. Hence, on September 4 HSY experienced an abnormal return of −0.047. This result is statistically significant at less than the 0.1% level (the standardized abnormal return was −3.917).
Interestingly, the next day HSY experienced a positive abnormal return of 0.03 that is statistically significant at the 5% level. Although we did not hypothesize that there would be a large abnormal return on that day, this finding could suggest a number of possibilities. Perhaps the September 5 return represents one of the rare instances where the abnormal return falls outside of the critical values wholly by chance. Or maybe the market overreacted to the court's decision, believing that it represented a larger setback to the trustees than it actually was. 274 Perhaps investors believed that the trustees would abandon the sale after the issunot be well approximated by the normal distribution. We address this concern in Part III.C.4 and in Appendix C.
273. Recall that in the graphs and econometric analyses we use adjusted closing prices, not actual closing prices. Hence, we report actual return in relation to the adjusted price data.
274. There is an ongoing debate in the finance literature about whether investors systematically overreact to bad news and if so whether sophisticated investors can profit from this tendency by engaging in a contrary investment strategy (e.g., purchasing shares in firms that depreciate after bad news). The empirical evidence of this phenomenon is mixed. For one study finding such an effect, see H. Nejat Seyhun, Overreaction or  Note: ***p<0.001; **p<0.01; *p<0.05 (i.e., abnormal return is statistically significant at the 0.1%, 1%, and 5% levels respectively). ance of the injunction, but reevaluated those beliefs when the trustees appealed immediately after the court decision. In their September 5 coverage of the trial court's decision, both the Wall Street Journal and the New [Vol. 108:749 York Times stated that "legal experts" predicted that the injunction would be lifted on appeal. 275 In all events, the sum of the abnormal returns for September 4 and 5 is negative and so signed as we predicted for September 4. These results, moreover, undermine the trial court's conclusion that issuing the injunction would not "cause significant fluctuations" in the Company's stock. 276 Our third date of interest is September 18, 2002, the first trading day after the trustees abandoned the sale close to midnight on September 17. On September 18, HSY dropped to $65.00 from the prior day's closing price of $73.81, which translates to an adjusted actual return of −0.119. The expected return for HSY was −0.002. Thus, on September 18 HSY experienced a negative abnormal return of −0.118. 277 This result, which is statistically significant at less than the 0.1% level (the standardized abnormal return was −9.833), belies the Attorney General's argument in his appellate brief that "nothing suggests that the stock will be priced differently even if a sale does not occur." 278 Accordingly, for all three principal dates of interest, we find highly statistically significant abnormal returns in the direction (that is, positive or negative) predicted by our agency cost hypotheses. The magnitude of the abnormal returns associated with the sale announcement and then the sale abandonment are striking: 25.5% and −11.8% respectively. Apart from the three primary dates of interest and the day after the injunction, there are no other dates during the sale window with abnormal returns that are significant at the 5% level or better. Further, the five abnormal returns that are significant at the 10% level (critical value ±1.65) are signed as one would expect in view of the events on and around those days. 279 To summarize and give greater context to our findings, in Figure 5a we present Hershey's cumulative abnormal returns (i.e., the sum of the abnormal returns; "CAR") for the sale window plus twenty-five trading days before and after the sale window. 278. See supra note 170 and accompanying text. 279. On both July 30 and August 1, as local opposition to the sale began to organize, HSY experienced negative abnormal returns of 2.1%. On August 7, while the trustees met to consider Fisher's proposed alternative modes of diversification, HSY experienced a negative abnormal return of 2.2%. That evening, after the stock market closed, the trustees announced that they would press ahead with the sale. The next day, August 8, HSY experienced a positive abnormal return of 2%. Finally, on August 26, the day the trustees filed their papers in opposition to the Attorney General's petition for a preliminary injunction, HSY experienced a positive abnormal return of 2.1%.
280. Because in this Figure (and in Figures 5b and 5c below) we include abnormal returns for twenty-five trading days before the sale window, in our expected returns  Figure 5a demonstrates that HSY's CAR during the sale window was neither the continuation of a preexisting trend nor the start of a new trend. Consistent with our agency cost hypotheses, the company was simply more valuable to investors during the period when the market expected the trustees to sell the Trust's controlling interest.
Interestingly, the CAR for Hershey remains above zero for some time after the end of the sale window. Assuming this finding is significant, 281 the data do not allow us to isolate an explanation, except to say that the increased Hershey CAR is probably not an artifact of secular trends in the chocolate or food industries, as Figures 5b and 5c indicate that Hershey's competitors do not likewise exhibit systematically higher CARs after the sale window. 282 Early empirical work on failed mergers suggested that regression we use as our estimation period the 100 trading days before the first trading day plotted. 281. The CAR over the event window is statistically significant with a t statistic of 2.59 (i.e., the CAR divided by a measure of the CAR's variance is 2.59, which exceeds the critical value of 2.575 for significance at the 1% level). However, when calculated from the beginning of the event window through the end point in the graph above (October 22, 2002), the CAR drops below the threshold for statistical significance (t = 0.54). Thus, at some point during the month after the Hershey sale was abandoned, cumulative abnormal returns dropped to the point where they were no longer statistically distinguishable from zero. As compared to a single day's abnormal returns, CAR over a longer period is a less volatile measure of the effect of the event, albeit with a corresponding loss in clean identification of the event.
282. That Hershey's competitors did not likewise experience increased CARs runs counter to the empirical regularity that rival firms' returns tend to go up after a firm in the industry is involved in a takeover activity. See B. Espen Eckbo, Horizontal Mergers, Collusion, and Stockholder Wealth, 11 J. Fin. Econ. 241, 257-59 (1983); B. Espen Eckbo, Mergers and the Market Concentration Doctrine: Evidence from the Capital Market, 58 J. target firms often exhibit increased CARs after a failed merger. 283 Later work, however, discovered that these gains tend to dissipate over a longer time horizon. 284 The literature suggests two possible explanations for these results: (1) a "kick-in-the-pants" idea, in which a target firm's management is inspired to implement a more efficient business strategy after the failed takeover attempt, and (2) the market's expectation that another acquiring firm will attempt a takeover in the near future. 285 We conjecture that the former pertains here, but we can neither establish nor reject our conjecture empirically.
3. Competitor-Based Controls. -As a check to rule out the possibility that factors unrelated to the Hershey sale might explain the observed abnormal returns on the three primary days of interest, we performed a similar abnormal return analysis for Hershey's principal NYSE-traded competitors identified in the earlier graphical analysis. 286 The results are presented in Table 4.
None of Hershey's competitors experienced an unusually large abnormal return on September 4 or September 18, and there is no consistent relationship between the abnormal returns of Hershey and its competitors on September 5. However, on July 25, the day the Hershey sale was announced, three of Hershey's competitors (CSG, TR, and CAG) experienced large, positive abnormal returns that were unlikely to have come about by chance. These results are amenable to at least two different interpretations.
Bus. 325, 338 (1985). Although early studies did not find evidence consistent with any particular causal mechanism for this phenomenon, more recent work suggests that the increase in competitors' CARs stems from an expectation that other firms in the industry will also become takeover targets. See Moon H. Song & Ralph A. Walkling, Abnormal Returns to Rivals of Acquisition Targets: A Test of the "Acquisition Probability Hypothesis," 55 J. Fin. Econ. 143, 170 (2000). Our finding that Hershey's rivals did not experience the typical increase in returns is consistent with this explanation, as the Hershey takeover activity was prompted by the trustees, not by an external buyer that, after failing to acquire the Company, might be expected to try to acquire one of Hershey's rivals.
283. See, e.g., Michael Bradley, Interfirm Tender Offers and the Market for Corporate Control, 53 J. Bus. 345, 368-72 (1980)   Note: ***p<0.001; **p<0.01; *p<0.05 (i.e., abnormal return is statistically significant at the 0.1%, 1%, and 5% levels respectively). Table 4 is that news of the sale positively affected not only the Company's value, but also the value of some of its competitors. There is, in other words, a potential endogeneity problem in using HSY's competitors as controls because changes in HSY might affect the value of the competitors, and vice versa. For example, perhaps the Company's stock price was depressed because the Company was overproducing above the efficient level to protect the jobs of its employees. With new management, the Company might reduce its production to the profit-maximizing level, thereby improving its value. A reduction in the Company's output would also benefit at least some of its competitors, as they would have less competition from the Company for individual sales. 287 A further endogeneity problem arises from the possibility that one of these competitors might bid on HSY.

One possible interpretation of
Although the foregoing interpretations are consistent with our agency cost hypotheses (particularly the overproduction story), we do not find them persuasive. If CSG, TR, and CAG experienced positive abnormal returns on July 25 because the market anticipated a decline in output by Hershey, then those companies should have experienced negative abnormal returns on September 18, when the sale was cancelled, and possibly also on September 4, when the sale was enjoined. But only two of those six abnormal returns were negative, both were relatively small, and neither was statistically significant. Unlike HSY, in other words, CSG, TR, and CAG did not experience parallel negative abnormal returns on news that HSY would not be sold. The same evidence militates against the pos- sibility that investors were pricing the possibility of acquiring HSY into the competitors' share prices.
The alternative, and in our view more likely, interpretation of Table  4 is that there were other market peculiarities on July 25, 2002. For example, there might have been an exogenous shock, unrelated to the news of the sale, which affected the value of some chocolate and food companies. On this account, the abnormal return experienced by Hershey on July 25 might in part reflect the influence of the exogenous shock. We think that this is a more plausible interpretation because it is consistent with the lack of statistically significant abnormal returns among Hershey's competitors on September 4 or September 18. We hasten to add, however, that Hershey's abnormal return on July 25 is far larger than those of its competitors, and Hershey also experienced large, significant, abnormal returns in the predicted directions on September 4 and September 18. Hence, we remain confident that most of the observed HSY abnormal return on July 25 is attributable to news of the sale.
To get a sense of how much of the HSY abnormal return on July 25 should be attributed to news of the sale, we estimated a market model in which we regressed the HSY return on the S&P 500 return, as before, as well as the returns for CSG, TR, and CAG. This augmented estimation model therefore accounts for correlation among these firms. 288 The HSY abnormal return for July 25, 2002, using this augmented estimation model, is presented in Table 5. Note: ***p<0.001; **p<0.01; *p<0.05 (i.e., abnormal return is statistically significant at the 0.1%, 1%, and 5% levels respectively).
In Table 5, the standardized abnormal return for Hershey on July 25, 2002 is slightly smaller than in Table 3, but it is still quite large and highly statistically significant. 289 Accordingly, we conclude that news of the sale was the primary explanation for the run up in HSY on July 25.
As in our analysis of HSY alone, another way to view the results is in a graphical depiction of cumulative abnormal returns. Figures 5b and 5c plot the cumulative abnormal returns for HSY in comparison to its choco-288. We used the 100 trading days before the Hershey sale window for our estimation period, although, as before, the results are virtually unchanged if we use alternative estimation periods.
289. The results of the augmented model for the other key dates of interest are likewise consistent with those presented in Table 3. late market competitors and its food manufacturing competitors respectively, each for the sale window plus twenty-five trading days before and after the window. 290 HSY with those of CSG, TR, and CAG at the beginning of the sale window, for CSG and TR this correlation does not continue through the rest of the window, and there is no appreciable comovement in the periods before or after the sale window. With respect to CAG, although the relationship appears to persist even at the end of the sale window and beyond, CAG's decline at the end of the sale window is not as pronounced as that observed for HSY. Moreover, HSY's cumulative abnormal return exceeds CAG's during the sale window but trails CAG's before and after the window. Accordingly, these graphs provide further evidence that news related to the proposed sale, not generic chocolate or food market forces, drove HSY's abnormal returns during the sale window.
4. The Problem of a Single-Firm Event Study. -Our study is limited by necessity to a single firm. Although one-firm event studies appear in the academic literature 291 and are used routinely in securities litigation, 292 they nonetheless pose special problems owing to the greater volatility experienced by a portfolio of one security than a portfolio of many securities. In Appendix Table B1, for example, we find that six of the ten days prior to the sale announcement experienced a statistically significant abnormal return. Here the concern is that we might attribute statistical significance to an abnormal return that in fact reflects nothing more than random variation if the abnormal returns of the single firm are not well approximated by the normal distribution. In other words, our worry is that the normal critical values used to determine statistical significance (i.e., ±1.96, ±2.575, and ±3.277 for .05, .01, and .001 respectively) may not be appropriate given the weakened signal to noise ratio in a single-firm event study.
An initial answer to this concern is that the magnitudes of the abnormal returns for July 25 (the sale announcement) and September 18 (the sale cancellation) are so large that as a mathematical matter they are statistically significant in all possible abnormal return distributions. This is 291. See an application of Chebyshev's inequality that we detail more formally in Appendix C.
However, reliance on Chebyshev's inequality assumes that we have correctly estimated the HSY abnormal returns. That assumption is problematic if the observed volatility of HSY in the 100 days prior to the event is not a reliable indicator of the volatility that HSY would have experienced during the event period but for the event. Here the limitation to a single firm is worrisome if the firm's volatility (that is, the variance in HSY's abnormal returns) could be changing over time. Accordingly, we must consider the possibility that the variance we estimated in the presale period was coincidentally low relative to other periods.
As an initial check against this worry, as detailed above, we also computed HSY's abnormal returns with reference to the observed volatility of HSY in a variety of alternative estimation windows. Some of those alternatives used only days after the sale window. Others were pooled models in which we used days before and after the sale window. 293 As a further cautionary measure, we undertook a randomization inference robustness check by performing event studies (i.e., performed steps 2 and 3 as described in Part III.C.2 above) on HSY's returns for 600 days randomly chosen during the three years before and after the sale window. 294 We then calculated the standardized abnormal returns for each of those dates (i.e., performed step 4 as described in Part III.C.2 above) to determine how frequently various standardized abnormal return levels occurred in the random sample. Next, we used these values to generate a probability density function. This randomization inference approach has been used in other applications, 295 but we believe that we are the first to use it in an event study in the law and economics literature.
The estimated probability distribution, superimposed over a standard normal distribution, is presented in Figure 6a. Although the HSY-specific probability distribution appears roughly to approximate the normal distribution, the two are not perfectly congruent. Using the estimated probability distribution, we computed a set of HSY-specific critical values or significance thresholds. These HSY-specific critical values are not perfect, of course, because some of the 600 randomly chosen days might in fact have experienced an abnormal return owing to unanticipated events on those days. Nonetheless, these HSYspecific critical values provide a useful check on whether our findings are artifacts of normal volatility rather than news of the sale.
Using our Figure 6a HSY-specific critical values, the observed HSY abnormal return on July 25, the date the sale plan was made public, was still highly statistically significant (less than 0.1% level). The observed HSY abnormal return on September 18, the first trading day after the sale was abandoned, also remained strongly significant, though somewhat less so (0.2% level, still highly significant). By contrast, the significance level of the observed HSY abnormal return on September 4 (the date of the injunction) fell to the 9% level. Accordingly, assessing statistical significance in relation to our estimated HSY-specific probability function largely confirms our earlier results using the normal distribution. Although the September 4 date falls below the conventional threshold of 5% for statistical significance, the results for the date the sale was announced and the date the sale was abandoned remain strongly significant.
We performed a similar analysis using the abnormal returns from event studies on 100 randomly chosen days for each of Hershey's six competitors identified above during the three years before and after the sale window. 296 The results of this analysis provide us with another benchmark against which to compare HSY's abnormal returns. For example, perhaps during this general time period abnormal returns for companies in this general industry are not well approximated by the normal distribution because of industry-wide increased volatility or structural changes in the market. While the distribution in Figure 6a addresses this worry, a second comparison is useful to put our major results into additional context. Figure 6b presents the estimated probability distribution from our examination of Hershey's competitors superimposed over a standard normal distribution. Using the critical values from the distribution depicted in Figure 6b, the statistical significance of the abnormal returns presented in Table 3 associated with news of the trustees' plan to sell and the abandonment of the sale are literally off the chart. No competitor in the sample had a 296. As before, we used the Eventus software and the CRSP value weighted index as our market portfolio. For the output from this analysis, see Jonathan Klick & Robert H. Sitkoff, Eventus Competitor Output (Aug. 23, 2007), at http://mailer.fsu.edu/~jklick/ COMProbust.pdf (on file with the Columbia Law Review). Note that this analysis does not actually include 600 data points as would be expected with six firms and 100 randomly chosen days. During a few of the randomly chosen days early in the sample, RMCF and KFT were not traded on the NYSE, leaving us with 556 observations on which we could base this analysis. RMCF was newly listed during our sample period and KFT was private during the early part of our period. While it would have been possible to restrict date selection to only those days during which all firms traded, doing so has the potential to introduce a bias into our analysis, so we opted to allow for the selection of days where some of the firms were not trading. standardized abnormal return greater than 8.9, but HSY's standardized abnormal returns on July 25 and September 18 (the sale announcement and termination dates) were 21.25 and 9.83 respectively. Further, the statistical significance of the abnormal return on the day of the injunction goes from 9% in the prior analysis (using the distribution of Figure 6a) to 1.3% using the distribution of Figure 6b. Hence, under this approach, we find no reason to suspect that HSY's abnormal returns on the days of interest were the result of mere random variation. 297

D. Summary of Empirical Findings
The evidence presented above is strongly consistent with our hypothesis of agency costs in charitable trust governance. The evidence is also strongly consistent with the widely held belief that supervision of charitable trusts by state attorneys general is deficient. Indeed, in this case the Attorney General's intervention was counterproductive, both reducing aggregate social welfare and imposing unnecessary agency costs on the Trust. Finally, the evidence indicates that the capital markets rated the threat of a takeover as superior to the Trust's controlling position for minimizing agency costs in the Company. We summarize the evidence and our specific findings here.
Graphical examination of Hershey's stock price, both by itself and in comparison to that of its main NYSE-traded competitors, strongly suggests that news of the trustees' plan to sell the Trust's controlling interest in the Company led to an increase in the Company's stock price. Graphical analysis likewise strongly implies that abandonment of the sale led to a decrease. However, comparison of Hershey's stock price dynamics with those of broad market indexes reveals some comovement of Hershey's stock with the market as a whole, leading us to apply more rigorous methodology to isolate the effects of the proposed sale from other effects.
Using a standard event study econometric analysis that isolates stock price movement attributable to sale-related news from general market trends, we find that Hershey's stock experienced a large and statistically significant positive abnormal return on news of the sale and a large and statistically significant negative abnormal return on the trustees' abandonment of the sale. These findings endure across a variety of specifications and numerous robustness checks. 298 In most specifications we also 297. Our approach of benchmarking abnormal returns to a distribution generated from HSY during periods outside the sale window and to a distribution generated from HSY's competitors bears some similarity to the bootstrapping methods introduced in John D. Lyon, Brad M. Barber & Chih-Ling Tsai, Improved Methods for Tests of Long-Run Abnormal Stock Returns, 54 J. Fin. 165 (1999). We thank Dan Rubinfeld and Max Schanzenbach for stimulating our thinking on alternative benchmarks and bootstrapping.
298. Indeed, our findings regarding the effect of the sale announcement and sale abandonment were significant in all specifications at better than the 99.8% level and in all but one at the 99.9% level.
find that the trial court's issuance of the preliminary injunction against the sale led to a statistically significant negative abnormal return.
Put into numerical terms, we find that news of the trustees' plan to sell the Trust's controlling interest in the Company was associated with a statistically significant positive abnormal return of 25.5%. This finding implies that the Company was worth $2.7 billion more on the open market than when under the control of the trustees. 299 Accordingly, in this case the existence of a controlling shareholder was manifestly inferior to the market for corporate control as an alternative mechanism for minimizing corporate agency costs.
Regarding trust law, our findings imply agency costs arising from the Trust's charitable trust form on the order of $850 million, 300 about 15% of the 2002 value of the Trust. 301 Although the trustees controlled more than three-quarters of the shareholder votes in the Company, they failed to impose a value-maximizing strategy on the Company's managers. As we have seen, the market judged the Company as being $2.7 billion (or 25.5%) more valuable when the Company's managers were expected to be subject to the market for corporate control instead of supervision by the trustees.
Moreover, instead of reducing the agency costs associated with the Trust's charitable trust form, the Attorney General's intervention made those agency costs permanent. Without any offsetting financial benefit to the Trust, the Attorney General forced the Trust to retain an asset that was worth $850 million more on the open market than in the hands of the trustees. While the sale's detractors argued that the sale would hurt 299. The $2.7 billion figure is the product of a back-of-the-envelope calculation that reflects 25.5% (our point estimate of Hershey's abnormal return) of Hershey's market capitalization of $10.7 billion on July 25, 2002. The 95% confidence interval implies that the loss lies within the range of $2.4 billion and $3 billion. We computed Hershey's market capitalization by multiplying the number of outstanding shares by Hershey's July 25, 2002, closing price of $78.30. Our share volume data, which reflects the number of shares outstanding on July 26, 2002, comes from Aug. 2002 Quarterly Report, supra note 270, at 1.
An alternate measure of the social cost might be possible using the negative abnormal return that occurred when the sale was abandoned. We hesitate to use this measure, however, because at various dates in the sale window the abandonment of the sale was probabilistically capitalized into the HSY price. This can be seen most clearly on the day the court issued a preliminary injunction against the sale. Thus, using this alternate measure requires subjective judgments about what total abnormal return we can attribute to expectations that the sale would be abandoned. The first public disclosure of the proposed sale is a cleaner and better identified shock to Hershey's exposure to the market for corporate control.
300. As with our $2.7 billion figure, this $850 million figure reflects a back-of-theenvelope calculation that extrapolates from our point estimates. Specifically, $850 million is the rounded-down product of multiplying the percent of outstanding HSY stock held by the Trust in 2002 as noted in supra Table 1  Given the analysis in Appendix Table B1 regarding potential leakage, the foregoing estimates may well understate the true measure of agency costs and hence the social welfare loss. If we included the abnormal returns for the three-day potential leakage period before the announcement, our estimates would be roughly 8% higher. 303

A. Charitable Trusts
We have shown that the Trust's principal asset, its stock in the Company, was worth $850 million more (about 15% of the Trust's 2002 value) on the open market than when in the hands of the trustees-a damning indictment of the trustees' inability to use their voting control of the Company to maximize value. We have also shown that the Attorney General's intervention to block the sale of the Company forced the Trust to maintain an undiversified and smaller portfolio. Our analysis therefore provides the first quantitative empirical evidence of agency costs in the charitable trust form and the inadequacy of supervision of charitable trusts by state attorneys general. Further, our findings highlight the related but less often remarked upon concern that, even if the attorney general does intervene in the administration of a charitable trust, such intervention might be designed to promote the attorney general's political interests at the expense of the trust's charitable purpose. 303. Still another alternate measure of the loss to shareholders (including the Trust) from abandoning the sale might be derived from Wrigley's $12.5 billion bid. That bid translates to $89 per share, which is $26.50 or 42.4% more than HSY's closing price of $62.50 on July 24, 2002, the day before news of the sale broke. Using these figures (see supra notes 299-300 for discussion of methodology), the total loss to shareholders amounts to roughly $5.1 billion, of which the Trust's share is $1.6 billion. However, we are skeptical of using the Wrigley bid for this purpose. While Wrigley's $12.5 billion bid might represent the value of the Hershey Company's assets under the control of Wrigley's management, a portion of that sum might also reflect agency costs in the Wrigley corporate structure. For example, the bid might include a control premium that Wrigley's management was willing to pay for the purposes of empire building or takeover protection. See supra Part II.B.1.b. As such, the Wrigley bid is a less reliable benchmark for assessing the welfare effects of terminating the sale than our estimate of the Hershey Company's abnormal return on July 25. The same analysis applies to the use of other bids.
As such, our findings provide an empirical grounding for at least three policy debates in the law of trusts: (1) the utility of alternative modes of charitable trust supervision; (2) the authorization of cy pres of charitable trusts on the ground of "wastefulness"; and (3) the propriety of social investing by trustees and other fiduciaries.
1. Alternative Modes of Supervision. -Although it is widely believed that the supervision of charitable trusts by the state attorneys general is inadequate, there is disagreement over the reasons and best remedy for that inadequacy. According to one view, supervision is inadequate because of insufficient resources. 304 Proponents of this view contend that increasing the budget for charitable supervision and perhaps also increasing the mandatory disclosure obligations of charitable organizations will improve oversight. 305 Another view holds that state attorneys general are lackadaisical monitors of charities because active supervision has little political payoff outside of isolated incidents where the charity's public salience makes imposing local preferences politically expedient. 306 On this view, increased funding and disclosure will have minimal effect because the problem is structural. Instead, alternative modes of supervision by people or entities with better incentives are required.
Our findings tend to support the latter view. As we have seen, even if the attorney general does intervene, the attorney general's political interests might not be congruent with the best interests of the trust or social welfare-here to the tune of $850 million and $2.7 billion respectively. More broadly, no amount of additional funding or increased disclosure will ameliorate the underlying structural problem that the attorney general is typically a political officer whose ambition toward higher office provides either little incentive to supervise charitable trusts or, as in the case of the Hershey Trust, perverse incentives to impose local political preferences. 307 In either scenario, monitoring by the state attorneys gen-304. See, e.g., Johnson, supra note 221, at 389 ("I attribute the failure of the attorneys general not to lack of interest or improper motive, but to the lack of funds devoted specifically to this purpose."); see also Brody, Parochialism, supra note 8, at 951 & n.50 (collecting studies on few resources allocated to charity enforcement).
305. See Gary, supra note 212, at 639-41, 646 (urging increased disclosure and proposing increased funding); Johnson, supra note 221, at 390 (proposing that charitable trusts pay fee proportional to size of corpus to fund attorney general's supervision); see also Chester, Improving Enforcement, supra note 221, at 475 (arguing that increased disclosure will not increase private enforcement absent relaxation of standing requirements); Fishman, Improving, supra note 211, at 270-72 (discussing recent federal disclosure requirements); Reiser, supra note 221, at 580-606 (analyzing legislative proposals for increased disclosure requirements for charitable trusts).
306. See, e.g., Brody, Parochialism, supra note 8, at 947-49 ("The incentives of this nearly universally elective office impel the incumbent to ignore cases that are politically dangerous and to jump into matters that are politically irresistible but implicate only 'business' decisions of charity managers." (footnote omitted)).
307 eral is inadequate to ensure that resources held by tax-exempt and, therefore, publicly-subsidized charitable trusts are deployed in efficient pursuit of a bona fide charitable purpose for the public good.
To be sure, we do not claim that all charitable trusts host the same magnitude of agency costs as the Hershey Trust. Nor do we claim that all interventions by state attorneys general are as maladroit as the intervention here. Instead our claim is that our results tend to validate the assumption in the theoretical literature, heretofore based chiefly on qualitative anecdotal evidence, of the prevalence of agency costs in charitable trusts and the failure by state attorneys general adequately to superintend such trusts. 308 Further, by showing that the Attorney General's intervention here reduced the value of the Trust and continued its exposure to uncompensated risk, our findings tend to support the view that the incentives of the state attorneys general may well be misaligned with the best interests of the charitable organizations that they superintend. Our findings, therefore, offer tentative support for alternative modes of supervision by agents with better incentives.
Legislators and academic commentators have proposed a variety of alternative forms of monitoring. For example, some have contended that the role of the IRS in supervising the operation of charitable entities should be expanded. 309 Others have argued for an independent federal agency, perhaps on the model of the Securities and Exchange Commission. 310 Almost half the states have given donors standing con- ("My strong recommendation, therefore, is that the first important action to increase foundation accountability be the establishment by Congress of such an NASD-like private, nonprofit organization . . . with sufficient resources to oversee the entire U.S. civic sector . . . ."); Fleishman, Public Trust, supra note 2, at 188-91 ("If we were starting from scratch, I would favor the creation of an independent agency . . . and would empower the agency with all aspects of the regulation of not-for-profit organizations . . . ."). But see Karst, supra note 219, at 481-83 (noting possibility of eventual federal regulation, but arguing that increased state regulation is superior). However, such an agency might be vulnerable to capture by charitable managers. Cf current with the attorney general to enforce a charitable trust. 311 And a handful of states allow relator standing to enforce charitable trusts-that is, standing for individuals in cases where the attorney general elects not to pursue the matter. 312 Perhaps most intriguing (albeit still theoretical), Geoffrey Manne has urged "the creation of private, for-profit monitoring companies" that would contract with charitable organizations "to monitor both the financial and charitable aspects of the nonprofit's operation." 313 Each of these proposals presents pros and cons in the particulars that we pass over here. Analyzing the specifics of the various reform proposals is beyond the scope of this paper; hence the qualifier that our findings offer "tentative" support. For present purposes, it suffices to note that the theme that ties each of these reforms together-and what makes each potentially attractive in view of our findings-is that they are more sensitive than the status quo to the incentives of the agents charged with supervision. Indeed, several, particularly donor standing, are designed specifically to harness those incentives.
2. Cy Pres for Wastefulness. -Under the traditional understanding of the cy pres doctrine, a court may direct the trustees of a charitable trust to apply the trust property to another charitable purpose that approximates the donor's original purpose if the original purpose becomes impossible, impracticable, or illegal. 314 The principal justification for the doctrine is the risk that, because a charitable trust may have a perpetual existence, changed circumstances will render the trust's original purpose obsolete. 315 As Bruce Mann has explained, "the law of charitable trusts has never regarded a donor's wishes as inviolate, regardless of what anyone, donors included, might think or want. All the law promises is that it will do its best to honor those wishes-literally if possible, by approximation if not." 316 The leading casebook offers the example of a "nineteenth-century trust to care for old horses retired from pulling fire wagons and streetcars" as a good candidate for cy pres. 317 Recently, however, there has been a movement to broaden the grounds for cy pres to include "wasteful" allocation of resources. The idea is to account for "the increasingly frequent problem of trust funds that have become excessive for their stated charitable purposes." 318 Under this reformed conception of cy pres, if the trust property so "exceeds what is needed for the particular charitable purpose . . . that the continued expenditure of all of the funds for that purpose . . . would be wasteful," then a court "might broaden the purposes of the trust," for example by "direct[ing] application of the surplus funds to a like purpose in a different community." 319 The 2000 Uniform Trust Code, the 2003 Restatement (Third) of Trusts, and the most recent preliminary draft of the Principles of Nonprofit Organizations all adopt "wasteful" as a basis for cy pres. 320 Whatever the other merits of broadening the grounds for cy pres, our findings suggest that cy pres in the case of an excess endowment may have the salutary effect of minimizing agency costs. Given the Trust's excess endowment, another $850 million would not affect the Trust's ability to fund the day-to-day operation of the School. As a result, there was little pressure on the trustees to maximize value and little risk for the Attorney General that blocking the sale would immediately imperil the School. Indeed, opponents of the trustees' diversification plan argued can foresee the future, a rational donor knows that his intentions might eventually be thwarted by unpredictable circumstances . . . ."). that the Trust had more money than it needed for its specific purpose and hence the Trust did not need to sell its interest in the Company. 321 By contrast, if the trustees were under pressure to use the Trust's excess endowment to fund similar schools in other communities, the social welfare loss from failing to maximize the value of the Trust would have been more apparent. Consider that the trust instrument gives preference to local orphans but also allows for admission of orphans from across the country. 322 Yet even though the Trust's corpus has grown dramatically, the School "served no more children at the start of 2005 than it did in 1963." 323 What is more, the Trust's current plans for expansion of the School are both modest and limited to the single existing campus. 324 A further benefit of "waste" as a basis for cy pres is that it would draw attention to considerations of scale versus scope. Before seeking to expand a charitable trust's scope, the trustees should exhaust possible extensions in scale. In the case of the Trust, by contrast, there has been an implicit expansion of scope to include the interests of the local community rather than an explicit expansion of scale (i.e., more students and additional schools). 325 Put more broadly, our findings tend to confirm the theoretical intuitions that an excess endowment provides a cushion for managerial laxity 326 and invites imposition of local political preferences by state offi-Under orthodox trust fiduciary law, "[n]o form of so-called 'social investing' is consistent with the duty of loyalty if the investment activity entails sacrificing the interests of trust beneficiaries . . . in favor of the interests of the persons supposedly benefited by pursuing the particular social cause." 332 Because the Trust's charitable purpose is to board and educate needy children from across the country, not to subsidize life in Hershey, Pennsylvania, the Trustees could not permissibly maintain the Trust's controlling interest in the Company for the purpose of promoting the interests of the town and the Company's workers.
Opponents of the 2002 sale argued that Milton Hershey would have wanted the trustees to consider the interests of the community. They pointed to the provision in the trust instrument that privileged applications to the School from local orphans over orphans from the rest of the country, 333 arguing that it evidenced an intent to give the community something of a beneficial interest in the Trust. 334 But an equally plausible interpretation is that the Hersheys' primary aim was to care for orphans above all other purposes. If, after admitting all suitable local orphans, the School could accommodate more, then the School was to look outside the immediate community for additional students. Consistent with this interpretation, the trust instrument provided explicitly that the Trust's income was to be "exclusively devoted" to the School. 335 More generally, there are two inherent weaknesses in the what-thefounder-would-have-wanted mode of reasoning. First, charitable trusts have perpetual existence. Eventually, after the inevitable change of circumstances, assessments of what the founder would have wanted devolve into rank conjecture. Second, even if we could know with certainty what the founder would have wanted, it is not obvious that we should defer to those wishes if they have undesirable social welfare consequences. Although American law is solicitous of the dead hand, it does not blindly enforce all donor wishes-for example, a parent's gift to a child conditioned on the child's marrying someone of a particular religious faith is management's duty to shareholders should not be relaxed in favor of public interest as agency and information costs will overwhelm any social gain enforced only if judged reasonable. 336 Indeed, the case for disregarding the donor's preference in the face of a contrary public policy is stronger in a charitable trust than in other donative contexts such as a private trust because a charitable trust is in effect subsidized by state and federal tax exemptions. To be sure, the presumption should be in favor of the donor's intent. But in the case of a charitable trust, donative freedom is constrained by the necessity of a bona fide charitable purpose that confers a sufficient social benefit to justify the quid pro quo of perpetual existence and exemption from taxation. 337 But suppose that there was no doubt that the Hersheys would have wanted to favor the community in the event of surplus funds, and suppose further that the law allowed the trustees to take into account the interests of the community in such a scenario. 338 Even then, it is hardly obvious that maintaining control of the Company would be an efficient way to benefit the community or subsidize life in Hershey. For example, the Trust could have sold its interest in the Company, thereby realizing the $850 million gain and diversifying its portfolio, and then it could have paid out every dollar of that $850 million gain to the local community, either in cash or through a program of community enrichment. In this scenario, the community would have received an actual $850 million transfer from the Trust, the Company would thereafter be subject to the pressures of the takeover market, and the Trust would have at least achieved salutary portfolio diversification.
Instead, the Trust was forced to maintain its interest in the Company (and so an undiversified portfolio), the community did not receive a cash transfer, and the Company remained subject to the less efficient monitoring of the trustees. The bill for this inefficiency came due in early 2007 when, after further erosion of the Company's market share, the Company announced that it would close one-third of its assembly lines, layoff 1,500 workers (12% of its workforce), and open a new factory in Mexico. 339 The Trust's controlling interest can shield the Company from the takeover market, but it provides no defense against the vicissitudes of the products market.
True, in late 2007 the Trust forced a shakeup in the Company's top leadership and expressed dissatisfaction with the Company's "unsatisfactory performance." 340 Indeed, in a letter addressed to the Company's Board of Directors, the chair of the board of trustees complained that since January 2002, the price of the Company's stock grew "at half the rate of the S&P 500 Index." 341 Yet the Trust also pledged publicly to maintain its controlling interest in the Company. 342 That announcement put a formal end to informal merger talks started in January 2007 between Cadbury and the Hershey Company. 343 A merger with Cadbury would have given the Company the international presence that it currently lacks. 344 So long as the Trust remains unwilling to sell, the Company's suboptimal control structure will continue to hinder its development. 345 A similar analysis pertains to shareholder welfare. To make the abandonment of the sale welfare neutral (i.e., Kaldor-Hicks efficient) with respect to shareholders, each of Hershey's 13,700 workers on average would have to value blocking the sale at more than $199,000, or, alternately, each of Hershey's 12,771 residents on average would have to value blocking the sale at $214,000. 346 True, the public shareholders knew or should have known that the Trust had voting control over the Company and those shareholders bought their shares at a discount reflecting the Trust's continuing control. Nonetheless, from an aggregate social welfare perspective, canceling the sale preserved a suboptimal capital structure on the order of $2.7 billion.

B. Corporate Governance
Our analysis bears on two policy debates in corporate governance: (1) the role of controlling shareholders or large blockholders in minimizing corporate agency costs, and (2) the role of the market for corporate control (i.e., the takeover market) in minimizing corporate agency costs.
1. Controlling Shareholders in Corporate Governance. -The incidence of controlling shareholders and minority blockholders is increasing among public U.S. firms and is even more common among public companies in Europe. 347 We have shown, however, that the markets valued the Company as being $2.7 billion more valuable (an increase of more than 25%) when its managers were subject to market discipline instead of supervision by the trustees as agents for the Trust. This finding suggests that a controlling shareholder, at least one whose agents are poorly motivated, provides less discipline against corporate agency costs than the takeover market. Indeed, the existence of a listless controlling shareholder or minority blockholder might have perverse governance consequences if it insulates the managers by foreclosing the possibility of a takeover.
To be sure, our results do not imply that a controlling shareholder can never effectively monitor agency costs within the firm. We cannot disentangle the role of the agency costs owing to the Trust's weak governance structure from the role of switching from a controlling shareholder to the market for corporate control as alternative explanations for our findings. But our findings do suggest that corporate agency costs will not be controlled if the individuals who make decisions for the controlling shareholder have little to gain from better performance. Although this 346. These figures were computed using the same employment and population figures as before, see supra note 302, but substituting $2.7 billion for $850 million. The Kaldor-Hicks efficiency criterion holds that an action is efficient if those who gain from the action benefit, in the aggregate, more than those who lose from the action, in the aggregate. As a result, it would be possible for the winners to compensate the losers, leaving everyone better off after the action is taken. In this case, if the workers (or residents) valued blocking the sale, on average, at an amount less than those provided above, it would not be possible for them to compensate shareholders for the wealth loss from blocking the sale. For a useful introductory discussion of efficiency measures, see point has been made in general terms before, 348 our empirical analysis puts the problem into sharp relief, suggesting a pair of related implications-one concerning the future empirical research agenda and the other concerning policy. First, our analysis suggests that a blockholder's organizational form should be included as a variable in future empirical work on whether the existence of the blockholder has a positive effect on firm value. Indeed, the failure to control systematically for agency problems within the blockholder's form may explain why existing studies generally do not arrive at robust conclusions. 349 By contrast, because our study is by necessity limited to a single blockholder whose governance structure is evident, we can speculate that it was the agency costs inherent to the Trust's charitable trust form that caused it to be a deficient monitor. With only one firm, however, we are not able to identify this incentive effect separately from Hershey's other idiosyncrasies. Accordingly, our hypothesis regarding the importance of the blockholder's organizational form calls out to be tested in future studies.
Second, our analysis tends to validate the existing, generalized concern about agency costs within a controlling shareholder's internal structure. As we have seen, the Trust was a poor monitor of the Company, but the fact of the Trust's deficient supervision does not negate the possibility of efficacious supervision by a blockholder that is not impaired by its own agency problems. Thus, perhaps hedge fund activism, as championed by Marcel Kahan and Edward Rock, 350 will be more effective and occur more frequently 351 than activism on the part of mutual funds and pen-348. See, e.g., Black, Agents, supra note 5, at 873 (discussing factors influencing monitoring by money managers); Coffee, Liquidity, supra note 242, at 1326 ("[A]gency problems at the institutional level can frustrate efforts to correct agency cost problems at the corporate level, even if institutional shareholders own sufficiently large blocks to be able to resolve their collective action problems.").
349. See Holderness, supra note 7, at 60 ("First, it has not been definitely established whether the impact of blockholders on firm value is positive or negative. Second, there is little evidence that the impact of blockholders on firm value-whatever that impact may be-is pronounced."  352 Unlike mutual fund and pension fund managers, who receive management fees on the order of 0.5% to 3%, 353 hedge fund managers receive performance rewards in the neighborhood of an additional 15%. 354 Further, because hedge funds tend to have fewer investors, the investors themselves may have more of an incentive to monitor actively. 355 2. The Market for Corporate Control. -Our results make a modest contribution to the empirical literature that investigates the takeover model. Consistent with previous studies of the effect of a takeover bid on share prices, 356 our finding of a large, significant positive abnormal return on exposing the Company's managers to the market for corporate control suggests that the disciplining effect of the takeover threat improves shareholder value. Indeed, we improve on those studies in that we are better able to exclude the alternative hypotheses that the observed increase in the target's price reflects (a) new information provided by the fact of the takeover attempt itself, 357 or (b) the accumulation of shares by a bidder in advance of announcing the bid. 358 We can exclude (a) because the Hershey sale was not instigated by an outside party and the observed price appreciation largely disappeared once the sale was abandoned for exogenous reasons. 359 We can exclude (b) because the attempted takeperiod, a third study also finds that hedge funds appear to play relatively activist roles in target firms and that they exhibit a high level of success with respect to their proposals. See  Further, relative to studies that examine changes in state takeover law, our results do not suffer from the potential omitted variables bias inherent in such studies. Changes in state takeover laws may themselves be the result of broader changes in the state's business climate.
359. Although others have found that takeover premiums tend to disappear after a takeover is abandoned, see Bradley, Desai & Kim, Rationale, supra note 284, at 194, it is difficult to rule out the possibility that a takeover is abandoned for reasons that are germane to firm value such as a downward revision of expected future profit streams or the implementation of a new takeover defense. In our case, this endogeneity problem may be ruled out because the sale was abandoned for reasons unrelated to the Hershey Company's underlying business prospects. over of the Company was invited by the Trust, before a prospective outside bidder would have begun buying shares to obtain a toehold in the Company. 360 Like other existing studies of share price movement on news of a takeover, we cannot rule out the possibility that the observed appreciation was driven by a capitalized control premium or expected synergies between the Company and the eventual acquirer. However, compared to the existing studies, these effects should be attenuated in our study because the takeover announcement came from the target before a buyer had been identified. The existence of a bidder-specific control premium or synergy was therefore uncertain when news of the trustees' plan to sell broke.
Although we improve on the existing literature in the foregoing ways, we nonetheless characterize our contribution to the takeover literature as modest for two reasons. First, given the possibility that the Trust might have imposed something other than profit maximization on the Company's managers, we cannot exclude the possibility that the observed price movement represents the market's expectation of a shift in management objective rather than the containment of corporate agency costs. In other words, our results may reflect only charitable trust agency costs. Second, our study is by necessity limited to a single firm. Given these limitations, the existing literature, which includes a host of multiple-firm takeover studies, supplies crucial interpretive context for our findings. When multiple studies using different research designs with different strengths and weaknesses produce similar results, confidence in the general validity of those results increases. 361 CONCLUSION This Article uses the aborted 2002 sale of the Hershey Company by the Hershey Trust as a quasi-natural experiment to investigate theoretical claims in the literatures of trust law and corporate law. In so doing, this Article makes fresh contributions to both literatures, and it also contributes to the technical literature on event study methodology. [Vol. 108:749

Trust Law
Our analysis provides the first quantitative empirical assessment of the prevailing scholarly view that agency costs are rampant in charitable trusts. Because a charitable trust must be for the benefit of a charitable purpose, not an identifiable beneficiary, there is no one with both an economic interest and legal standing to ensure that the trustee efficiently pursues the trust's charitable purpose. State attorneys general have formal authority to enforce charitable trusts, but the attorney general is typically a political official for whom supervision of such trusts has little political payoff. Further, when a charitable trust does achieve enough political salience to attract the attention of the attorney general, the attorney general's political interests often militate toward imposing local political preferences, which may not be congruent with the charitable purpose or the interests of society more broadly.
Our findings imply agency costs in the high-profile, politically salient Hershey Trust on the order of $850 million (about 15% of the 2002 value of the Trust). Moreover, instead of reducing the agency costs associated with the Trust's charitable trust form, the Attorney General's intervention to block the Trust's sale of the Company made those agency costs permanent. The question thus arises, can these results be generalized? The answer is yes and no. We do not claim that all charitable trusts host the same magnitude of agency costs as the Hershey Trust or that all interventions by state attorneys general are as maladroit as the intervention here. Instead, we suggest that our results tend to validate the assumption in the theoretical literature, heretofore based chiefly on qualitative anecdotal evidence, of the prevalence of agency costs in charitable trusts and the shortcomings in supervision of charitable entities by the state attorneys general. Accordingly, our analysis throws light on current policy debates in trust law such as (1) alternative modes of supervision for charitable trusts, (2) expanding the cy pres doctrine to address excess endowments, and (3) the propriety of social investing by trustees.

Corporate Law
Agency costs are the central concern of the study and practice of corporate law. The difficult task is to align the incentives of managers with the interests of shareholders, or at least to constrain the extent to which managers can advance their own interests at the expense of shareholders. A growing empirical literature suggests that the takeover threat posed by exposure to the market for corporate control and monitoring by controlling shareholders or large blockholders are among the most powerful levers available to induce managers to act in the best interests of shareholders. Our analysis contributes to this literature by suggesting that, at least in the case of the Hershey Trust and Hershey Company, the takeover threat provided more managerial discipline than the presence of a controlling shareholder. 362 Although we cannot disentangle the role of the Trust's weak internal governance structure from the role of switching from a controlling shareholder to the takeover market as alternative explanations for this finding, our results nonetheless highlight the need to consider the incentives of a blockholder's agents when analyzing the efficacy of relying on the blockholder to police managerial agency costs. The $2.7 billion increase in the value of the Company on exposure to the takeover market is a damning indictment of the trustee's failure as agents of the Trust to maximize value.

Financial Econometrics
Although the event study technique is a well-accepted tool of financial econometrics, and single-firm event studies appear in the academic literature and are routinely employed in securities litigation, 363 such a study poses special problems on account of the increased volatility associated with a portfolio of one. The concern is that a single firm's observed abnormal returns might not be normally distributed, and if so, that we might attribute statistical significance to an abnormal return that in fact reflects only random variation. To address this worry, we develop and present a method of assessing statistical significance that uses an empirical distribution of abnormal returns based on randomization inference as a substitute for using the critical values for statistical significance from the standard normal distribution. This randomization inference technique has been used in other applications outside of financial economics, and it bears a similarity to recently developed bootstrapping methods, but we believe that we are the first to use it in an event study in the law and economics literature. Because the technique can easily be implemented in most single-firm event studies, and because it improves confidence in assessments of statistical significance, we commend its use in future single-firm event studies. 362. We consider the potentially conflating influences of a capitalized control premium or expected synergies in the text following supra note 360.