Are empowerment and education enough? Underdiversification in 401(k) plans.

Choi, James J. ; Laibson, David ; Madrian, Brigitte C. 等

AT THE END of 2000, current and former employees of the energy trading company Enron Corporation held $2.1 billion in the firm's 401(k) retirement savings plan. Sixty-two percent of that money was invested in Enron stock, then trading at $83 a share. In October 2001 Enron's finances began to unravel as its accounting improprieties came to light. Enron stock plummeted over the next several weeks, and on December 2, 2001, the company declared bankruptcy, rendering its shares worthless. Thousands of Enron employees lost their jobs and a large fraction of their retirement wealth simultaneously.

We thank Hewitt Associates for their help in providing the data. We are particularly grateful to Lori Lucas and Yah Xu, two of our many contacts at Hewitt, for their help with the data and their useful feedback on this project. We appreciate the research assistance of Ananya Chakravarti, Keith Ericson, Jean Lee, Hongyi Li, Shih En Lu, Dina Mishra, and Chris Nosko. We are also grateful for comments from Christopher Carroll, William Gale, Nellie Liang, and other participants on the Brookings Panel as well as at seminars at the American Economics Association, the National Bureau of Economic Research, the University of Chicago, the University of Michigan, and Harvard University. James Choi acknowledges financial support from the Mustard Seed Foundation. All three authors acknowledge individual and collective financial support from the National Institute on Aging (grants R01-AG-021650, P30-AG012810, and T32-AG00186). David Laibson also acknowledges financial support from the Sloan Foundation. The opinions and conclusions expressed are solely those of the authors and do not necessarily represent the opinions or policy of the National Institute on Aging, any other agency of the federal government, Hewitt Associates, or the National Bureau of Economic Research.

Although the Enron 401(k) debacle was highly publicized, Enron was neither the first nor the last company whose collapse decimated its workers' 401(k) accounts. Over the past few years a similar fate has befallen employees of WorldCom, Global Crossing, Polaroid, Kmart, Lucent, and Providian, among others. In response, many bills have been proposed in Congress that would regulate employer stock holdings within 401(k) plans. Compared with existing law on defined-benefit pension plans, which strictly prohibits plans from holding more than 10 percent of their assets in employer securities, most of the bills proposed for regulating defined-contribution plans appear mild. Common themes in these proposals are empowerment and education rather than prohibition: give employees the right to sell the employer stock in their 401(k), and inform them about the risks of not doing so.

For example, one of only two bills that have so far come up for a vote in either house of Congress, the Pension Security Act (see table A-1 in the appendix), has two key provisions relating to employer stock. First, it would prohibit employers from requiring employees to invest their own 401(k) contributions in employer stock. Second, it would require that employers allow plan participants to diversify any matching funds contributed by the employer three years after receiving that match. Other proposed legislation would require that plan participants be notified if the fraction of their assets invested in employer stock exceeds a certain threshold (such as 20 percent), that companies offer a certain number of alternatives to employer stock if it is made an investment option, or that companies educate plan participants about the risks of not diversifying their assets.

This paper assesses how effective the "empower and educate" regulatory approach might be at reducing 401(k) employer stock holdings. We begin by studying five natural experiments in which employees experienced a discrete change in the restrictions on employer stock holdings. In these examples the restrictions changed for one of two reasons: either employees crossed an age or tenure threshold above which they were allowed to diversify their holdings, or the company changed its rules to enable all employees to diversify their investments. We find only a modest employee response to either type of change. Merely allowing diversification does not cause it to happen.

We then consider whether educational efforts might motivate employees to diversify out of employer stock. Although many studies have concluded that financial education does affect employees' choices, the subset of studies that randomly assign education and measure subsequent actions have found small effects. (1) These studies still leave open the possibility that other kinds of education might yield larger behavioral changes.

Here we evaluate a different form of education: witnessing the real-life experience of others. Economists since Armen Alchian in the 1950s have argued that the imitation of successful strategies (and, conversely, the avoidance of unsuccessful strategies) is an important force pushing economic actors toward optimal behavior. (2) We test this hypothesis in the context of the media coverage surrounding the Enron, WorldCom, and Global Crossing bankruptcies. Specifically, we investigate how much workers at other companies reduced their employer stock holdings in response to the blizzard of media stories early in this decade illustrating the dangers of putting all of one' s retirement savings in employer stock.

We chose Enron, WorldCom, and Global Crossing because a large percentage of their employees' 401(k) assets was held in employer stock, and because their bankruptcies, the associated accounting scandals, and their decimated 401(k) plans received so much attention from so many media outlets. For example, the New York Times ran 1,364 stories mentioning Enron during the last quarter of 2001 and the first quarter of 2002, of which 112 ran on the front page.

We find that this media barrage had a surprisingly modest impact on employer stock holdings in other 401(k) plans, reducing the fraction of assets held in employer stock by no more than 2 percentage points from an initial 36 percent of balances. We present evidence that this small reaction is not due to restrictions on diversification. In addition, we show that workers in Texas, who were likely to have been disproportionately exposed to Enron-related news, did not reduce their investment in employer stock any more than did workers outside of Texas. Even in Houston--Enron's headquarters--where the Houston Chronicle ran 1,122 stories mentioning Enron in the six months surrounding the firm's collapse, employees did not show evidence of learning the lesson of Enron.

This paper raises broader questions about retirement savings policy. It adds to the convergent body of evidence that many employees do not make optimal financial decisions. Households typically behave passively, following the path of least resistance. Such inertia often translates into household acceptance of the investment choices made automatically by their company on their behalf, even when those default choices are suboptimal for them. (3) For example, many firms automatically allocate 401(k) matching funds to employer stock, where the money typically stays even if employees are permitted to subsequently rebalance their portfolio.

Two policy solutions present themselves. First, society could give firms incentives to adopt socially optimal default choices. (4) However, it is not always obvious what such optimal defaults would be. Moreover, one default is rarely right for every employee, since employees face different economic circumstances. Alternatively, society could adopt default-free systems, which do not confer an advantage on one choice over another, by forcing employees to explicitly state their preference. (5) However, default-free systems will work only if employees are likely to make good decisions when forced to do so.

We begin with a brief summary of the current regulation of employer stock in 401(k) plans. We then summarize previous research on employer stockholding in 401(k) plans and how 401(k) outcomes are affected by plan features. Next we describe the employee-level 401(k) data we have for the seven companies examined in our analysis. Our empirical analysis begins with the results from five natural experiments on relaxing diversification restrictions. We then turn to examining the impact of the Enron, WorldCom, and Global Crossing crises on employer stock holdings for a large sample of employees at other firms. We conclude by discussing alternative legislative approaches that are likely to decrease employer stock holdings substantially, and implications for savings policies more generally.

Regulation of Employer Stockholding in 401(k) Plans

Like defined-benefit pension plans, 401(k) plans are primarily regulated under the Employee Retirement Income Security Act of 1974 (ERISA). (6)

Under ERISA, plan fiduciaries have four responsibilities: to act for the exclusive benefit of plan participants and their beneficiaries (the "exclusive purpose rule"); to act with the care, skill, prudence, and diligence that a prudent person acting in a similar capacity would use (the "prudent man rule"); to diversify plan assets across different types of investments, geographic areas, industrial sectors, and dates of maturity so as to reduce the chance of large losses (the "diversification rule"); and to act in accordance with plan documents.

At the time ERISA was passed, defined-benefit pension plans were the primary employer-sponsored mechanism for providing income to persons in retirement. To help safeguard the assets of these plans, ERISA explicitly caps the holdings of employer stock at 10 percent of total assets. Defined-contribution plans, however, face no such limit. The most common type of defined-contribution plan today, the 401(k) plan, did not exist when ERISA was enacted. Those defined-contribution plans that did exist consisted mainly of profit-sharing plans, to which employers made variable contributions based on company earnings, and employee stock ownership plans (ESOPs), which were explicitly designed to encourage ownership of employer stock. ERISA exempts these plans from the diversification requirements for employer securities.

When 401(k) plans were first authorized, in 1978, it was not anticipated that they would supplant defined-benefit pension plans as the primary source of employee retirement income. Hence Congress did not extend the 10 percent limit on employer stock holdings to 401(k) plans. Employers thus have much latitude in determining how much employer stock employees may hold in their 401(k).

A common way of holding employer stock in a 401(k) is through an ESOP. This combination of a 401(k) and an ESOP is sometimes referred to as a KSOP. Many companies, however, offer employer stock in their 401(k) plan without an ESOP, and many companies also operate an ESOP that is separate from the 401(k) plan. A KSOP is a savings plan intended to benefit employees, but it is also a corporate finance mechanism that encourages employee ownership. The KSOP's dual purposes can create a conflict of interest between plan beneficiaries and the employer. As a result, employer stock regulations for 401(k) plans with an ESOP differ from those for plans without one.

A 401(k) plan without an ESOP may not require that more than 10 percent of the employee's own contributions be held in employer stock. This requirement does not hold for a KSOP. Neither plan, with or without an ESOP, is limited in the amount of matching contributions by the employer that may be directed into employer stock. All assets within an ESOP, however, are subject to a set of explicit diversification requirements. Employees with ten years of tenure must be allowed to diversify 25 percent of their employer stock holdings once they reach age fifty-five, and 50 percent once they reach age sixty. (These diversification requirements do not apply to employer stock holdings outside of an ESOP.) Companies can and do, however, adopt diversification policies that are more generous than those mandated by ERISA. Finally, should a lawsuit arise, the fiduciary standards of prudence and exclusive purpose against which the company's behavior is judged are lower for a KSOP than for a 401(k) plan without an ESOP, because of the dual purposes of the former. (7)

Previous Research on Employer Stock in 401(k) Plans

Olivia Mitchell and Stephen Utkus report that, averaging across all 401(k) plans--including those without employer stock as an investment option--19 percent of plan assets are held in employer stock. (8) This statistic understates the diversification problem, since most participants do not have employer stock as an investment option--their employers do not offer it or are not publicly held firms. Only about 10 percent of companies offer employer stock in their 401(k) investment menu. Because these companies tend to be larger firms, 35 percent of all 401(k) participants are in plans that do include employer stock as an option. (9) These employees often have 401(k) portfolios that are heavily concentrated in employer stock. William Even and David Macpherson calculate that 50 percent of assets in plans offering employer stock were held in employer stock in 1998. (10) Plans offering employer stock can be further divided into those in which the employee must choose an investment allocation for the employer match and those in which the employer match is directed into employer stock by default. In 2001 two-thirds of those plans directing the match into employer stock allowed participants to immediately trade out of employer stock, whereas the rest imposed a holding requirement. (11) Sarah Holden and Jack VanDerhei report that, in plans that offer employer stock as an investment option but do not direct the match into employer stock, 22 percent of assets are held in employer stock; in plans in which the employer match is directed into employer stock, a much larger fraction of assets, 53 percent, is held in employer stock. (12)

The Enron, WorldCom, and Global Crossing bankruptcies highlighted two major risks associated with concentrated investments in employer stock. First, investing in any single security is riskier than investing in a diversified portfolio such as a mutual fund. Several studies have estimated that, on a risk-adjusted basis, a single-stock portfolio is worth less than half the equivalent amount invested in a diversified portfolio. (13) Second, the value of employer stock may be positively correlated with employees' labor income; as noted at the outset, many Enron employees simultaneously lost their jobs and their retirement savings.

Shlomo Benartzi and coauthors identify some benefits to employees from holding employer stock in their 401(k) plan, but they note that these benefits are small relative to the substantial costs of nondiversification. (14) Why then do employees invest so much of their 401(k) portfolios in employer stock? A number of psychological factors may be important.

Employees may underestimate the risk of employer stock because of "familiarity bias." The John Hancock Financial Services Defined Contribution Plan Survey finds that participants on average rate employer stock as less risky than an equity mutual fund. (15) Similarly, Benartzi and others find that only 33 percent of participants believe that their employer stock is riskier than a diversified stock fund, whereas 39 percent believe it is equally risky and 25 percent believe it is safer. Furthermore, 20 percent of respondents said they would prefer $1,000 in employer stock that they could not diversify until age fifty over $1,000 that they could invest at their own discretion. (16)

Lauren Cohen argues that loyalty to one' s employer motivates workers to hold employer stock. (17) He finds that the share of employer stock held in 401(k) plans by employees of standalone firms is 10 percentage points greater than that of employees of conglomerates. Furthermore, when a division is spun off from a conglomerate, the average employee's holding of employer stock increases, whereas it decreases following a merger. Cohen finds no evidence that employees have superior information about future employer stock returns.

Benartzi, Choi and others, and Gail Huberman and Paul Sengmueller find that current contributions to employer stock are increasing in the stock's past returns. (18) If employees think past returns predict future returns, then the high volatility of employer stock relative to mutual funds and the inability to sell securities short in a 401(k) will jointly lead to over-weighting of employer stock in the average 401 (k) portfolio. Intuitively, employer stock is likely to be in either the upper or the lower tail of the 401(k) asset return distribution in any given period. Upper-tail outcomes have a greater impact on asset allocation than lower-tail outcomes because of the constraint on short sales. Like Cohen, Benartzi finds no correlation between employee allocations to employer stock and the stock' s subsequent returns.

Benartzi and Richard Thaler suggest that naive strategies for diversifying across investment options cause many investors to allocate part of their contributions to employer stock simply because it is available in the 401(k) menu. (19) Nellie Liang and Scott Weisbenner report evidence consistent with this hypothesis specifically for employer stock. (20) Employees may also perceive the presence of employer stock in the 401(k) fund menu as an endorsement of the stock by the employer. Benartzi, Holden, and VanDerhei, and Jeffrey Brown, Liang, and Weisbenner, find that discretionary contributions to employer stock are higher in firms where the employer directs matching contributions into employer stock than in firms where employer stock is simply available as another investment option. (21)

These investment menu effects are consistent with a growing body of literature that finds that 401(k) savings outcomes are strongly affected by the features of the plan, even when those features do not explicitly restrict employee choices. Madrian and Dennis Shea, as well as Choi and coauthors, document the tendency of participants to passively accept 401(k) enrollment, contribution, and investment fund defaults. (22) Switching from an opt-in to an opt-out enrollment mechanism raises 401(k) participation rates six months after hire by over 50 percentage points at some firms. The vast majority (65 to 87 percent) of newly hired employees retain the automatic enrollment default contribution rate and asset allocation upon enrollment in the 401(k) plan. Although the fraction of employees at the default declines over time, the inertia is very strong: three years after hire, nearly half of participants are still at the defaults. Generalizing from these results, one might expect that employer matching contributions made in employer stock will stay in employer stock, even if participants are able to immediately diversify out.

Choi and others show that requiring employees to explicitly choose within a month of hire whether or not to participate in a 401(k) plan (thus removing the option of passively accepting a default outcome) raises the initial participation rate by 28 percentage points, and the average contribution rate by 1.2 percentage points, relative to a standard opt-in enrollment procedure. (23) Such an "active decision" mechanism may also be an effective way of encouraging employees to diversify their assets. Employees could be told to make an explicit asset allocation choice for their employer matching funds, instead of the current system in which employers may direct matching funds into employer stock by default. (24)

Employee-Level Data on 401(k) Participation

The employee-level data that we use to examine the holding of employer stock in 401(k) plans come from Hewitt Associates, a large benefits administration and consulting firm. We use two different types of administrative data. The first is a series of repeated year-end cross sections containing all persons employed at the company at the time of the data snapshot. These cross sections include demographic information such as birth date, hire date, and compensation. They also include information on the worker's 401(k) participation at the time of the cross section, such as date of first participation, total balances in the plan, and allocation of assets among the available funds. The second type of data we have is a history of every transaction made in the plan since Hewitt assumed its administration. These longitudinal data are available for two of the companies analyzed in this paper (companies D and F). (25) For all of these firms, our analysis is restricted to active employees. Tables 1 and 2 summarize the demographic and 401(k) plan characteristics of the seven firms studied in this paper.

Empowerment: The Effect of Relaxing Diversification Restrictions

As already noted, current 401(k) regulations allow companies to severely restrict their employees' diversification out of employer stock. Eighteen of the twenty-one congressional proposals in appendix table A-1 to reform 401(k) employer stock regulations would relax these restrictions, permitting employees to trade out of employer stock in their 401(k) sooner.

To assess whether relaxing holding requirements would significantly reduce employer stock holdings, we study two types of natural experiments at five companies, all of which direct the employer match into employer stock. The first type results from changes in the diversification restrictions that employees face either as they age or as they increase their tenure at the company. We examine how much employees divest themselves of employer stock when they clear the applicable age or tenure hurdles. The second type of natural experiment arises from company-wide rule changes that eliminate or relax holding requirements for all employees. We measure how much employer stock holdings fall after such regime shifts.

Allowing Older Employees to Diversify out of Employer Stock

We begin with the firm we designate as company D, whose 401(k) employer stock is in an ESOP and thus subject to the ESOP diversification rules described above (for employees with ten years of tenure, 25 percent diversification is allowed at age fifty-five, and 50 percent at age sixty). The middle right panel in figure 1 shows the effect on employee portfolios of crossing these age and tenure thresholds. It compares the actual average fraction of employer matching balances held by individual participants of a given age in employer stock, as well as the actual dollar-weighted average fraction (that is, total employer stock balances divided by total balances for all employees of a given age), with the fraction that employees with at least ten years of tenure would hold if they all diversified the maximum amount allowed. The figure shows that actual diversification is minimal. Across all ages, the average participant holds at least 90 percent of match balances in employer stock, even as the maximum diversification allowed increases to 50 percent. There is no consistent divergence between the dollar-weighted and the unweighted series, suggesting that wealthier employees are not more likely to divest than poorer employees. This absence of a wealth gradient in diversification holds for many of the companies we study in this paper.

[FIGURE 1 OMITTED]

As previously noted, some companies have employer stock diversification policies that are less restrictive than what is legally required. (26) One of these is company A, which also has an ESOP within its 401(k) plan. Although this firm is subject to the same ESOP diversification requirements as company D, it voluntarily adopted less stringent diversification rules in 2002, allowing employees to diversify up to 25 percent of the employer stock in their match account at age forty-five, 50 percent at age fifty, and 100 percent at age fifty-five.

The top left panel of figure 1 plots the relationship at company A between age and the fraction of match balances in employer stock at the end of 2003, approximately eighteen months after diversification restrictions were relaxed. Among employees between the ages of forty-five and fifty, the average (non-dollar-weighted) fraction held in employer stock fell by 7 percentage points. This is less than a third of the diversification available to these employees. Among employees at age fifty-five, employer stock holdings fell further, to 85 percent of total match balances, also about one-third of the actual diversification available. Among employees at age fifty-six, the fraction of match balances held in employer stock fell to 76 percent, about one-quarter of the actual available diversification, and the proportion does not drop below 64 percent through age sixty-five. In this company, dollar-weighted averages fell slightly more, indicating that richer employees are more eager to divest themselves of employer stock, but even the dollar-weighted share is far from the maximum allowed diversification.

Two other companies that have age- or tenure-based diversification rules--companies B and C--have life cycle diversification patterns similar to those of company A. Company B allows participants with at least ten years of service to diversify 25 percent of their match balance holdings starting at age fifty, 50 percent starting at age fifty-five, and 75 percent starting at age sixty. The top right panel of figure 1 shows the fraction of employer match balances held in employer stock at company B at the end of 2003 for employees with at least ten years at the company. The simple average seldom dips below 90 percent for workers between fifty and sixty-five years of age. Dollar-weighting the averages yields similar results.

Company C has more complicated diversification rules. Employees younger than fifty-five can diversify up to 50 percent of the employer stock balances in their match accounts if they have ten years of plan participation. Employees aged fifty-five to fifty-nine can also diversify up to 50 percent but need have only five years of plan participation to do so. Employees between the ages of 60 and 70 1/2 can diversify 100 percent of their employer stock balances if they have five years of plan participation. Finally, employees over age 70 1/2 can sell all their employer stock regardless of their time in the plan.

Figure 2 plots average employer stock holdings in match balances at company C at the end of 2002 against tenure in the plan for employees less than fifty-five years of age. The middle left panel in figure 1 plots average holdings against age for employees at the same company who are between fifty-five and seventy and have at least five years of tenure in the plan. It is apparent that, despite greatly relaxed holding constraints, most employees at this company continued to hold more than 90 percent of their match balances in employer stock years after they became eligible to diversify.

[FIGURE 2 OMITTED]

Rule Changes That Relax Diversification Restrictions for All Employees

Figures 1 and 2 suggest that allowing employees to diversify their 401(k) holdings has only a modest impact on actual diversification. However, the complexity of the diversification eligibility schedules may have confused employees, reducing diversification. Companies that completely eliminate diversification restrictions for all employees may see greater divestiture of employer stock. Here we study two companies that implemented such a change.

The first, company D, adhered to the standard ESOP diversification requirements (with the results analyzed in figure 1) until early 2002, when it eliminated all diversification restrictions. The company's match, however, continued to be directed to employer stock following this plan change. Employees who wanted to maintain a portfolio free of employer stock had to initiate trades on an ongoing basis as new matching contributions were made.

Figure 3 shows the fraction of the employer match account held in employer stock at company D from January 1998 through November 2003. Before the plan change, almost all employer match balances stayed in employer stock. This is not surprising, given that only 6.6 percent of employees met the standard ESOP age and tenure requirements for any diversification. After the 2002 rule change, the allocation to employer stock fell, but very slowly. At the end of 2003, almost two years after the diversification restrictions were lifted, 84 percent of match balances were still in employer stock.

[FIGURE 3 OMITTED]

Company E eliminated all diversification restrictions in mid-2002. It, too, continued to direct employer match contributions to employer stock. Unfortunately, lack of appropriate data prevents us from producing the analogue of figure 3 for company E. Instead the bottom left panel of figure 1 shows diversification by age at the end of 2003, approximately eighteen months after the diversification restrictions were lifted. At this time the average participant under age fifty still had 90 percent of his or her employer match balances in employer stock, and older employees continued to hold over 80 percent of their match balances in employer stock. There is more diversification on a dollar-weighted basis, especially among older participants.

The five case studies presented here paint a consistent picture: relaxing diversification requirements decreases employer stockholding, but the magnitude of that reduction is modest, especially for younger employees.

Education: The Effect of Enron, WorldCom, and Global Crossing

If giving employees the ability to diversify does not prompt them to do so, perhaps educating them on the perils of underdiversification will spur them to action. Six of the twenty-one proposed bills listed in appendix table A-1 include an educational provision. The studies reviewed in the introduction found that financial education alone generates only small changes in 401(k) outcomes. However, no one to our knowledge has formally studied the effect on plan participants of witnessing the consequences of others' financial follies.

Here we examine whether the media coverage of the plight of employees at Enron, WorldCom, and Global Crossing--three companies frequently mentioned together in news stories on the perils of holding too much employer stock in one's 401(k) account--caused employees at other companies to reduce their employer stock holdings. On December 31, 2000, before any of these firms entered serious financial distress, 62, 32, and 29 percent of 401(k) assets were held in employer stock at Enron, WorldCom, and Global Crossing, respectively, according to the companies' Form 5500 filings and their 11-K filings with the Securities and Exchange Commission. Figure 4 gives a timeline of the key events leading up to the demise of these companies. For the purposes of our analysis, the first "Enron event" was the $1.2 billion reduction in Enron shareholder equity that occurred on October 16, 2001.

[FIGURE 4 OMITTED]

We construct a series of variables to capture the amount of media coverage surrounding the financial collapses of Enron, WorldCom, and Global Crossing. These variables are based on the number of news stories mentioning the three companies published each day starting on October 16, 2001. We employ four different counts. The first is the number of stories published in five major newspapers: the New York Times, the Wall Street Journal, the Washington Post, the Los Angeles Times, and the Chicago Tribune. We used the Lexis/Nexis and Factiva databases to search these publications, and we excluded republished news, recurring pricing and market data, obituaries, sports news, and calendars. We examined the resulting list of stories by hand and deleted results that appeared to be duplicates as well as stories fewer than 100 words long (mostly one-paragraph teasers pointing to stories elsewhere in the paper). The second count includes both these newspaper stories and stories found in Lexis/ Nexis transcripts of television news programs on three major broadcast networks: ABC, CBS, and NBC. (27) There were 12,047 relevant newspaper stories and 1,927 relevant television stories from October 16, 2001, to December 31, 2003. The third and fourth counts exclude from the first and second, respectively, all stories that do not include the word "401(k)." This reduces the total number of stories to 761 for the newspapers-only count and 905 for the television-plus-newspapers count. Table 3 provides a breakdown of these totals by individual source. Our constructed variables serve only as proxies for the millions of items on Enron, WorldCom, and Global Crossing that appeared in local and national newspapers and magazines and on websites, radio, and television.

Using these four news story counts, we created four geometrically time-discounted news impact variables to capture the fact that reaction to news does not occur immediately. The impact of each story is assumed to decay by a factor [delta] for each day after its publication or broadcast. We multiply the product by (1 - [delta]) so that the mean value of the news impact variables is invariant to the value of [delta]. Thus, for example, a story published on January 2 would receive a weight of (1 - [delta]) on that day, a weight of [delta](1 - [delta]) on January 3, a weight of [[delta].sup.2] (1 - [delta]) on January 4, and so on. The impact of all stories on day t (where t = 0 on October 16, 2001) is the sum of the geometrically discounted impact of all previous stories:

news [impact.sub.t] = (1 - [delta])[t.summation over p=0][[delta].sup.t-p][N.sub.p],

where [N.sub.p] is the number of stories published on day p. Because we have little theoretical or empirical guidance on the appropriate value of 5, we try a range of values.

Figure 5 shows the number of relevant newspaper stories published on each day, as well as the value of the news impact variable constructed from those data using a value of [delta] = 0.9. The most active news coverage occurred from mid-December 2001 through the end of February 2002, when Enron and Global Crossing filed for bankruptcy. There is another peak in June and July 2002, when WorldCom disclosed accounting irregularities and shortly thereafter filed for bankruptcy.

[FIGURE 5 OMITTED]

Having created a proxy for media attention, we estimate its association with 401(k) employer stockholding in two domains. First we study its effect on aggregate 401(k) employer stockholding, using data on 401(k) participants in a number of large firms. However, some of these employees may have been required to hold employer stock by their companies or may not have had employer stock available as a 401(k) investment option. Therefore we also analyze two specific companies that offered employer stock but had no diversification restrictions. We then examine whether workers in Texas--and, even more specifically, Houstonians--reacted more to the Enron bankruptcy than did workers outside of Texas. Because Enron is headquartered in Houston, workers in Houston and Texas were exposed to more Enron media coverage and were more likely to have known people directly affected by the bankruptcy.

The Effect of News about Enron, WorldCom, and Global Crossing on Aggregate Employer Stock Holdings

Our aggregate employer stock data come from the Hewitt 401(k) Index, which measures daily trading activity in a large number of firms that collectively employ 1.5 million participants with nearly $90 billion in 401(k) assets. (28) Our Hewitt Index data run from August 4, 1997, through November 4, 2003 (seventy-five months). Because the set of firms represented in the index changes over time, we restrict our attention to the subset that are in the index for the entire sample period. These firms hold approximately half the 401(k) assets of all firms in the index.

For these firms figure 6 shows both the fraction of total 401(k) assets and the fraction of equities held in employer stock. The former ranges from 26 to 38 percent, which is higher than the 19 percent calculated by Holden and VanDerhei using a much larger database. (29) This discrepancy probably results from the fact that firms in the Hewitt Index tend to be very large, and large firms are more likely to offer employer stock as an investment option. The fraction of equities in employer stock ranges from 36 to 51 percent over the same period.

[FIGURE 6 OMITTED]

The share of all assets in employer stock shows a declining trend from August 1998 through March 2000, an increasing trend from March 2000 through May 2001, relative stability from May through November 2001, and a declining trend from November 2001 through November 2003. The beginning of that last declining trend coincides with the financial collapse of Enron. On October 15,2001, the day before Enron's shareholder equity fell by $1.2 billion, the fraction of total balances in employer stock at Hewitt Index firms was 35.9 percent. By April 2003 it had reached a low of 25.8 percent. However, a broader stock market decline began around the same time, and this would have mechanically reduced the fraction of 401(k) assets held as employer stock even if employees were completely passive.

If the decline in employer stock holdings were completely mechanically driven by a general decline in stock prices, then employer stock as a fraction of equities should not be greatly affected, because both the numerator and the denominator in this fraction would decrease by roughly the same proportion. Figure 6 shows, however, that this series is not flat over the post-Enron period, but the decline is much less pronounced than that for employer stock as a fraction of all balances. Indeed, it is striking that employer stock as a fraction of equity holdings barely moved during the time that the Enron scandal was unfolding. However, this stable ratio need not be due to a lack of trading in employer stock; plan participants could have aggressively sold both employer stock and other equities. With such concerns in mind, we turn to an analysis of trading activity in employer stock.

Our dependent variable is the value of aggregate net transfers into or out of employer stock, normalized by dividing by the preceding day's aggregate employer stock balance. (30) This variable will be positive when the total dollars transferred into employer stock exceeds the total dollars transferred out, and zero or negative otherwise. Figure 7 plots normalized net transfers over our sample period. (31) Casual visual inspection indicates that net trading volatility seems to have increased following Enron's collapse, but the mean direction of trade does not obviously shift.

[FIGURE 7 OMITTED]

To isolate the impact of the financial collapse of Enron, WorldCom, and Global Crossing on employer stock trading, we regressed normalized net employer stock transfers on the news impact variable, current and lagged market returns on the Standard and Poor's 500 stock index (calculated so that the returns are nonoverlapping), (32) day-of-the-week dummies, and a cubic polynomial time trend (in annualized trading days). Table 4 reports least-squares regression results using the four different versions of the news impact variable and [delta] values of 0.25 and 0.9.

The coefficients on the news impact variable are negative and statistically significant across all specifications, indicating that trading out of employer stock occurred in response to media attention to the Enron, WorldCom, and Global Crossing episodes. We calculate the total media-induced decline in employer stock holdings as a fraction of 401(k) balances using the formula

(1) [T-1.summation over t=0][beta] x news impact x employer stock [percent.sub.t-1],

where [beta] is the regression coefficient on the news impact variable, employer stock [percent.sub.t-1] is the fraction of total balances held in employer stock on day t - 1, and T- 1 is the total number of post-Enron trading days in our sample. The last row of table 4 shows the calculated economic magnitudes for each specification. (33) No matter how we defined a news story as relevant, and regardless of the [delta] value chosen, none of the specifications show that more than a 2.4-percentage-point drop in employer stock holdings as a fraction of the total portfolio can be attributed to news about Enron, WorldCom, and Global Crossing. (34) The coefficients become more negative when all stories rather than just 401(k) stories are included, but they attenuate when we broaden our media sources to include television. There is no consistent pattern with respect to [delta] values. The small economic magnitudes also hold in regressions (results not reported) that use other definitions of the story variable (for example, New York Times stories only, or stories about Enron only), that use other values of [delta], that exclude the trend variables, that use data aggregated to the monthly level, or that use a distributed lag specification for the impact of news rather than a geometrically decaying specification. (35)

If a large fraction of Hewitt Index firms do not offer employer stock in their 401(k), or if many employees in those firms are unable to diversify because of holding restrictions, then the aggregate results in table 4 will significantly understate the true effect of media education on employer stock holdings. Unfortunately, we do not know anything about the prevalence of employer stock availability or holding restrictions among the companies in the Hewitt Index. However, a 2003 Hewitt Associates survey that collected data on various aspects of 401(k) plan design from approximately 450 large firms reports that 49 percent of these firms offered employer stock as an investment option. (36) Of those that offered employer stock, 78 percent either had no diversification restrictions for matching contributions directed into employer stock or did not match in employer stock. If the Hewitt Index companies are similar to this sample of firms, and if we assume counterfactually that none of the employer stock reduction occurred in companies with holding restrictions, and if the specification that yields the most negative news impact coefficient is correct, then the media coverage of Enron, WorldCom, and Global Crossing caused employer stock holdings to drop by at most 2.35/0.49 x 0.78 = 6.15 percentage points in companies without holding restrictions, still a small effect. In reality, most employees at firms with holding restrictions are not constrained at the margin by the restrictions, because they hold so much employer stock in the accounts over which they have full discretion. Holden and VanDerhei report that employees with holding requirements invest 33 percent of their own contributions in employer stock, whereas employees without holding requirements invest 22 percent of their entire 401(k) portfolio in employer stock. (37) Therefore 6.15 percentage points is an upper bound on the true effect. The magnitude of this effect is consistent with other estimates of financial education we have reported elsewhere, and with the finding by Esther Duflo and Emmanuel Saez that financial education motivates only small changes in 401(k) saving behavior. (38)

The Effect of News on Companies Offering Employer Stock without Diversification Restrictions

To further investigate the hypothesis that holding requirements prevented a dramatic aggregate fall in employer stock holdings following the Enron, WorldCom, and Global Crossing scandals, we study two large firms offering employer stock in their 401(k)s for which we have daily trading data. The first, company D, had employer stock holding requirements in the employer match account for most of the sample period (see above). For this reason we examine only the employee contribution account for company D, where there were no such restrictions. The second, company F, had no holding requirements anywhere in its 401(k) plan. In addition, employer matching contributions at company F were not directed into employer stock; rather, their allocation mirrored that which employees chose for their own contributions.

Figure 8 plots the fraction of employee contribution balances held in employer stock from 1997 through 2003 for company D. Employer stock holdings declined in the early post-Enron period (October 2001 through January 2002) but increased for most of the period after that, including the time of greatest Enron-related media attention (January through March 2002). These patterns track the company's stock price. Taken by itself, this figure does not suggest that the employees of company D responded to news about Enron, WorldCom, and Global Crossing.

[FIGURE 8 OMITTED]

Figure 9 plots for company D the average fraction of employee contribution flows that went to employer stock on each date, with and without weighting by dollar value. We examine these series because employees who have decided to permanently decrease their employer stock holdings may begin by stanching the flow into employer stock that occurs at each payroll date. In fact, the allocations steadily increased over time and show only slight declines in the immediate wake of the Enron scandal before continuing their march upward. There is thus little evidence of significant adjustment on the contribution flow margin.

[FIGURE 9 OMITTED]

Figure 10 plots daily net transfers of employer stock at company D as a fraction of the previous day' s employer stock balance. We begin the sample in November 1997 because that is when the plan first allowed trading on a daily rather than a monthly basis. The magnitude of normalized net trade flows is larger than that observed in the Hewitt Index, but this is to be expected. An individual company's net trade flow will have a substantial idiosyncratic component, which will be averaged away when aggregated with many other companies in the Hewitt Index.

[FIGURE 10 OMITTED]

Table 5 shows the results of regressing this series on the news impact variable and several control variables, including those used in the previous subsection. (39) We find that, across all specifications, the largest news-induced drop in employer stock holdings is a statistically insignificant 1.21 percent of employee contribution balances. (40) When [delta] is set to 0.9, the regression results suggest that news of the major bankruptcies caused company D employees to trade into employer stock. Since we are studying accounts with no diversification restrictions, we can conclude that employees did not trade out of employer stock even in accounts in which they had every opportunity to do so. (41)

Figure 11 plots the fraction of total 401(k) account balances held in employer stock over time at company F, which had no holding restrictions on either the employee contribution or employer match balances. There are two complications with the company F data. First, company F allows its participants to put their 401(k) assets in a self-directed brokerage window, where they can trade securities not offered in the company's 401(k) investment menu. We do not observe daily balances in the self-directed window and thus cannot include them in the daily total balances we calculate. (We do, however, observe the self-directed window balances at the end of each year.) Therefore figure 11 overstates the fraction of all 401(k) balances allocated to employer stock. Because we observe total employer stock balances each day, our trading regressions will not be affected by this data issue. Second, our company F sample does not start until January 2002, which means that we do not observe pre-Enron behavior at this company. Therefore our estimates of the news effect at company F may be less reliable.

[FIGURE 11 OMITTED]

Figure 12 shows that, over the entire sample period, the fraction of contribution flows that participants at company F directed to employer stock slowly increased, even during the post-Enron period. This corresponds to the pattern in figure 11, which showed an increasing share of balances held in employer stock over time. As at company D, there is no evidence of Enron-induced adjustment in the contribution flow allocations.

[FIGURE 12 OMITTED]

Figure 13 plots company F's daily net transfers of employer stock as a fraction of the preceding day' s employer stock balance. Table 6 shows the news impact regression results. In calculating the economic effect, we modify equation 1 to reflect the fact that we do not observe total 401 (k) balances each day, but only at the end of each year. Our alternative formula is

(2) [[T.sub.0].summation over t = [[tau].sub.0]][beta] x news [impact.sub.t] x employer stock [balances.sub.t-1]/ total year-end 2003 401(k) balances,

where [[tau].sub.0] is January 1, 2002, and [T.sub.0] is the number of trading days for which we have transactions data for company F. (42) We find that the largest economic effect of media coverage across all the specifications is a statistically insignificant 5.2-percentage-point decrease. The average estimate is a 1.8-percentage point-decrease.

[FIGURE 13 OMITTED]

Overall, our analysis of employer stock trading at companies D and F shows no statistically or economically significant effect of the Enron, WorldCom, and Global Crossing debacles on employees' willingness to hold employer stock, even in accounts not subject to diversification restrictions.

Impact on Texas Employees Versus Non-Texas Employees

The Enron scandal is more likely to have been salient to workers living in Texas than those living outside Texas. Not only was local news coverage likely to have been more active, but residents of Texas are more likely to have known Enron employees or friends of Enron employees. About a quarter of Enron' s employees worked at the Houston headquarters, and more than half of the company's 8,000 job cuts in 2002 were in Houston. Here we explore whether Texan employees at companies C and G reduced their employer stock holdings more than did non-Texan employees at the same companies.

Companies C and G are large retail firms chosen for this study because they have a substantial number of employees both in and outside of Texas. Company G imposed no diversification restrictions on its employees, although employer matching contributions had to be initially invested in employer stock. Employees who wished to diversify had to initiate a trade periodically, but they were free to do so at any time. We do not have daily transactions data for company G, and so we rely on a different identification strategy than in the previous subsection. We have cross-sectional snapshots of portfolio allocations at the end of each year from 1998 to 2002. We use these snapshots, exploiting the fact that employees rarely change their 401 (k) elections after enrolling. (43) Therefore any year-end differences in employer stock holdings between Texans and non-Texans who enrolled at the same time are likely to be due to asset allocation decisions made at the time of enrollment. We investigate whether Texans who enrolled in the post-Enron period show an increasing tendency to deviate from non-Texans' employer stock allocations.

Figure 14 graphs the fraction of total balances held in employer stock against enrollment date for three groups of employees at company G: those living in the greater Houston metropolitan area (Harris County), those living outside of greater Houston but in Texas, and those living outside of Texas. The portfolio allocation is taken from the first year-end snapshot after the employee's enrollment date. This figure shows that the asset allocations of all three groups track each other fairly closely. There is no greater tendency for either Houstonians or Texans to reduce their employer stock holdings in the wake of the Enron scandal.

[FIGURE 14 OMITTED]

Company C imposes age- and tenure-based diversification constraints on its employees as described previously. Therefore we examine employer stock holdings in only those accounts over which the employees have full discretion, employing the same identification method as for company G. Figure 15 plots the discretionary holdings in employer stock against initial enrollment date. As in figure 14, the allocation is measured at the first year-end snapshot after enrollment. As with company G, there is no significant reduction in the employer stock holdings of Texan employees relative to non-Texans. There are not enough Houston-based employees at company C to analyze them separately.

[FIGURE 15 OMITTED]

The evidence presented here does not seem to support the hypothesis that proximity to financial disaster leads to wiser decisions about employer stock.

Conclusions

Our analysis suggests that empowering employees to trade out of employer stock and educating them about the risks of employer stock will have only a small effect on 401(k) employer stock holdings. Even after diversification restrictions were relaxed at five companies, it was typical for over 90 percent of employer match balances to remain in employer stock. Two years of headlines about decimated 401(k) accounts in the early part of this decade did little to drive investors out of their employer's stock. Our estimates suggest that the media publicity surrounding the Enron, WorldCom, and Global Crossing bankruptcies reduced the fraction of aggregate 401(k) assets held in employer stock by at most 2 percentage points, from about 36 percent to about 34 percent. These results support the conclusions of earlier studies that consumers are often passive, (44) and that educational interventions yield remarkably small changes in saving behavior. (45)

These conclusions pose a problem for policy, because education and empowerment are the key components of almost all of the proposals for 401(k) employer stock reform listed in appendix table A-1 (for example, bills S.9, S. 1992, S. 2032, and S. 2190). None of these legislative proposals are nearly as stringent as those currently applied to defined-benefit plans, which cap employer stock holdings at 10 percent of account balances.

The persistent appeal of employer stock leaves economists in a policymaking quandary. On the one hand, the well-known benefits of diversification argue that employer stock should not make up a large fraction of a worker's retirement portfolio. On the other hand, economists dislike paternalism and do not want to enact coercive policies such as banning employer stock in retirement accounts. More generally, household saving decisions are fraught with other economically significant mistakes, which only the most paternalistic would consider forbidding. (46)

Fortunately, some reasonable policy options are available that are likely to reduce employer stock holdings without denying investors freedom of choice. (47) Under today's status quo, many firms invest matching contributions in employer stock. The natural experiments reported in this paper show that such matching contributions tend to stick where they initially land, regardless of any subsequent diversification opportunities. Hence employer stock holdings would be significantly reduced if the default asset allocation for employer matching contributions were changed to provide greater diversification. For example, default portfolios could be required to meet the same fiduciary standards that apply to defined-benefit pension plans.

With diversified defaults, investor inertia would be rendered harmless, because only those investors who actively opt out of the default could pile into employer stock. As noted above, in 401(k) plans that offer employer stock as an investment option but lack an employer stock default for matching funds, only 22 percent of balances end up in employer stock. By contrast, in 401(k) plans with an employer stock default for matching funds, well over 50 percent of balances end up in employer stock. (48)

Strict libertarians may object even to restrictions on default portfolios. If requiring diversified defaults is deemed too paternalistic, employees could be asked to explicitly choose their own asset allocation instead of being defaulted into an undiversified allocation. (49) Alternatively, legislation could remove the perverse tax benefits that give firms an incentive to choose employer stock as a default asset allocation. (50)

The common theme is that defaults stick and bad defaults are not inevitable. Bad defaults could be replaced with either good (diversified) defaults or with default-free choice. Banning bad defaults does not coerce employees, since they are free to opt out of a good default and adopt any asset allocation they choose. However, banning bad defaults does coerce firms, since they are no longer free to use defaults to induce their employees to become shareholders. Although we see this as a good thing for employees, there are some potential countervailing effects, such as a potential decline in the willingness of firms to provide a 401(k) match, or, in the extreme, to offer a savings plan at all.

Requiring (or at least encouraging) good defaults is a plausible solution to many weaknesses in the nation's retirement savings system. Defaults can be used to encourage participation in retirement plans, raise retirement saving rates among at-risk groups, and diversify 401(k) portfolios. In every case, defaults need not be coercive to work. However, defaults are a two-edged sword, since poorly chosen defaults can wreak havoc on financially unsophisticated, inertia-bound consumers.</p> <pre> APPENDIX Table A-1. Congressional Proposals to Regulate Employer Stock in 401(k) Plans, 2000-05 Bill no. Title, sponsor, and date introduced Bills under consideration in the 109th Congress (2005-06) H.R.1960

Pension Preservation and Savings Expansion Act of 2005 Rep. Rob Portman (R-OH) April 28, 2005 H.R.1961 Pension Preservation and Savings Expansion Act of 2005 Rep. Ben Cardin (D-MD) April 28, 2005 S.219 National Employee Savings and Trust Equity Guarantee Act of 2005 Sen. Charles Grassley (R-IA)

January 31, 2005 Bills under consideration in the 108th Congress (2003-04) H.R.1000 Pension Security Act Rep. John Boehner (R-OH) February 27, 2003 H.R.1776 Pension Preservation and Savings Expansion Act of 2003

Rep. Rob Portman (R-OH) April 11, 2003 H.R.2101

Pension Fairness Act of 2003 Rep. George Miller (D-CA)

May 14, 2003 H.R.5397 Retirement Enhancement Act

of 2004 Rep. Robert Andrews (D-NJ)

November 19, 2004 H.R.5398 Retirement Enhancement Revenue

Act of 2004 Rep. Robert Andrews (D-NJ)

November 19, 2004 S.9 Pension Protection and

Expansion Act Sen. Tom Daschle (D-SD)

January 7, 2003 S.1892 Sen. Evan Bayh (D-IN)

November 19, 2003 S.2424 National Employee Savings and

Trust Equity Guarantee Act Sen. Charles Grassley (R-IA) May 14, 2004 Bills under consideration in the 107th Congress (2001-02) H.R.3640 Pension Protection and Diversification Act of 2002 Rep. William Pascrell (D-NJ)

January 29, 2002 H.R.3692 Pension Protection and

Diversification Act of 2002 Rep. Sheila Jackson-Lee (D-TX) February 7, 2002 H.R.3762 Pension Security Act of 2002 Rep. John Boehner (R-OH) April 15, 2002 H.R.5110 Omnibus Corporate Reform and Restoration Act of 2002 Rep. Sheila Jackson-Lee (D-TX)

July 12, 2002 S.1838 Pension Protection and Diversification Act Sen. Barbara Boxer (D-CA) December 18, 2001 S.1919 Retirement Security Protection Act of 2002 Sen. Paul Wellstone (D-MN) February 7, 2002 5.1971 National Employee Savings and Trust Equity Guarantee Act (NESTEG) Sen. Charles Grassley (R-IA) February 27, 2002 5.1992 Protecting America's Pensions Act of 2002 Sen. Edward Kennedy (D-MA) March 6, 2002 5.2032 Investor-Employees Need Financial Facts and Options for Responsible

Retirement Plan Management Act of 2002 (INFORM)

Sen. Richard Durbin (D-IL) March 19, 2002 S.2190 Worker Investment and Retirement Education Act of 2002 (WIRE) Sen. John Kerry (D-MA) April 17, 2002 Bill no. Provisions relating to employer stock in 401 (k) plans Bills under consideration in the 109th Congress (2005-06) H.R.1960 Employees may sell employer stock derived from their

own contributions immediately. Employees may sell

employer stock derived from employer contributions

after three years of service or three years after receiving

the employer contribution. Plans that offer employer

stock must offer at least three other investment options. H.R.1961 Employees may sell employer stock derived from their

own contributions immediately. Employees may sell

employer stock derived from employer contributions

after three years of service or three years after receiving

the employer contribution. Plans that offer employer

stock must offer at least three other investment options. S.219 Employees may sell employer stock derived from their

own contributions immediately. Employees may sell

employer stock derived from employer contributions

after three years of service. Plans that offer employer

stock must offer at least three other investment options. Bills under consideration in the 108th Congress (2003-04) H.R.1000 Employees may sell employer stock three years after receiving it. Companies may not require employees to invest their own contributions in employer stock. H.R.1776 Employees may sell employer stock derived from their own contributions immediately. Employees may sell employer stock derived from matching contributions after three years of service, and employer stock derived from all employer contributions after five years of service. Plans that offer employer stock must offer at least three other investment options. H.R.2101

Employers must notify participants and inform them of

the risks of not diversifying assets if employer stock

exceeds 10 percent of balances. Participants must be

informed of large employer stock sales by a corporate

insider or plan fiduciary. Employees may sell employer

stock after three years of service. H.R.5397 Companies may not require participants to invest their own contributions in employer stock. Employees may sell employer stock derived from employer contributions after three years of service. Plans that offer employer stock must offer at least three other investment options. H.R.5398 Employees may sell employer stock derived from employer contributions after three years of service. Plans that offer employer stock must offer at least three other investment options. S.9

Employers must notify participants if employer stock

exceeds 20 percent of 401(k) balances. Companies must

certify that employer stock is a prudent investment.

Firms may match in employer stock or offer it for employee contributions, but not both. S.1892 Plan administrators must notify participants that they may be overinvested in employer securities and real property if such assets' value exceeds 50 percent of 401(k) balances. S.2424

Employees may sell employer stock three years after

receiving it. Plans that offer employer stock must offer

three other investment options. Bills under consideration in the 107th Congress (2001-02) H.R.3640 Employer stock is capped at 20 percent of 401(k) assets. Employees may sell employer stock ninety days after receiving it. In ESOPs, employees may sell employer stock after five years of participation or at age thirty- five. H.R.3692 Employer stock is capped at 20 percent of 401 (k) assets. Employees may sell employer stock ninety days after receiving it. In ESOPs, employees may sell employer stock after five years of participation or at age thirty- five. H.R.3762 Companies may not require participants to invest their own contributions in employer stock. Employees may sell employer stock after three years in the plan. Plans that offer employer stock must offer at least three other investment options. H.R.5110 Employer stock is capped at 20 percent of 401 (k) assets. In ESOPs, employees may sell employer stock after five years of participation or at age thirty-five. S.1838 Employer stock is capped at 20 percent of 401(k) assets.

Employees may sell employer matching contributions

made in employer stock ninety days after receiving it. S.1919

Employer stock is capped at 20 percent of 401(k) assets.

Companies may not require participants to invest their

own contributions in employer stock. Employees may

sell employer contributions made in employer stock after

one year of service. 5.1971 Employees may sell employer stock three years after receiving it. Plans that offer employer stock must offer at least three other investment options. 5.1992 Employees may sell employer stock three years after receiving it. Companies may not require participants to invest their own contributions in employer stock. Plans

must provide information regarding diversification

rights and the importance of diversifying assets. Plans

that offer employer stock must offer at least three other

investment options. 5.2032 Employers are liable for excessive losses on employer securities during lockdowns on plan assets. Employers must notify participants if employer stock exceeds 30 percent of balances, and further employer contributions to employer stock are prohibited until the participant signs a form acknowledging the risks of not diversifying assets. S.2190 Plans must provide investment guidelines to participants. Companies may not require participants to invest their own contributions in employer stock. Plans that offer employer stock must offer at least three other investment options. Employees may begin selling employer contributions of employer securities after five years in the plan, with full divestment permitted after seven years. Employees may begin selling employer matching contributions of employer securities after three years in the plan, with full divestment permitted after five years. Bill no. Status as of August 19, 2005 Bills under consideration in the 109th Congress (2005-06) H.R.1960 In House subcommittee H.R.1961 In House subcommittee S.219 In Senate committee Bills under consideration in the 108th Congress (2003-04) H.R.1000

Passed in House; no vote taken in Senate H.R.1776

No vote taken H.R.2101 No vote taken H.R.5397 No vote taken H.R.5398 No vote taken S.9 No vote taken S.1892 No vote taken S.2424 No vote taken Bills under consideration in the 107th Congress (2001-02) H.R.3640 No vote taken H.R.3692

No vote taken H.R.3762 Passed in House; no vote

taken in Senate H.R.5110 No vote taken S.1838 No vote taken S.1919 No vote taken 5.1971 No vote taken 5.1992 No vote taken 5.2032 No vote taken S.2190 No vote taken Source: Library of Congress, THOMAS data and authors' analysis. </pre> <p>(1.) Duflo and Saez (2003); Choi and others (2002); Choi, Laibson, and Madrian (2005a).

(2.) Alchian (1950).

(3.) Madrian and Shea (2001); Choi and others (2002); Choi, Laibson, and Madrian (2005a).

(4.) Choi and others (2003, 2005).

(5.) Choi and others (2005); Choi, Laibson, and Madrian (2005b).

(6.) This section draws heavily on Ng (2003) and Reish and Faucher (2002).

(7.) Ng (2003); Reish and Faucher (2002).

(8.) Mitchell and Utkus (2003), citing Holden and VanDerhei (2001) and VanDerhei (2002).

(9.) Even and Macpherson (2004).

(10.) Even and Macpherson (2004).

(11.) Fidelity Investments (2002).

(12.) Holden and VanDerhei (2001) and VanDerhei (2002); these authors use administrative data to calculate the fraction of assets held in employer stock, whereas Even and Macpherson (2004) use IRS Form 5500 (tax reporting) data. The numbers in the two sources are not directly comparable.

(13.) Poterba (2003); Meulbroek (2005); Brennan and Torous (1999).

(14.) Benartzi and others (2004). See also Brown, Liang, and Weisbenner (forthcoming) for a discussion of the costs and benefits to firms and workers of holding employer stock in the 401(k) plan.

(15.) John Hancock Financial Services (2002). Interestingly, the perceived risk of employer stock in the John Hancock survey has not changed much over time, even following the Enron scandal.

(16.) Benartzi and others (2004).

(17.) Cohen (2005).

(18.) Benartzi (2001); Choi and others (2004b); Huberman and Sengmueller (2004).

(19.) Benartzi and Thaler (2001).

(20.) Liang and Weisbenner (2002).

(21.) Benartzi (2001); Holden and VanDerhei (2001); Brown, Liang, and Weisbenner (forthcoming).

(22.) Madrian and Shea (2001); Choi and others (2002, 2004a).

(23.) Choi and others (2005).

(24.) We are exploring this possibility in an additional natural experiment not discussed here.

(25.) The names of the companies are withheld for reasons of confidentiality.

(26.) Many firms have voluntarily relaxed their diversification restrictions following the collapse of Enron (Fidelity Investments, 2002), probably because of the negative publicity surrounding employer stock diversification restrictions and the liability risks of maintaining such restrictions.

(27.) The transcripts were limited to national news programs (such as the CBS Evening News, Nightline, and 20/20); transcripts of local news broadcasts were not available.

(28.) Hewitt Associates (undated-a, undated-b). Historical values for the index can be found at was4.hewitt.com/hewitt/services/401k/index.htm. We do not use the index itself but only certain inputs into the index.

(29.) Holden and VanDerhei (2001); VanDerhei (2002). They use the EBRI/ICI database compiled by the Employee Benefit Research Institute and the Investment Company Institute.

(30.) Dollars traded includes only balances that are actively transferred from one investment option to another. It does not include payroll contributions that are invested in employer stock, nor does it include withdrawals out of the plan entirely.

(31.) Occasionally, large changes in employer stock balances were observed because data for certain component companies were missing on those days, generating outliers in the calculated normalized trading activity. We eliminated such outlier days from our data.

(32.) The S&P return variables are the returns today, yesterday, last month, the last six months, and the last one, two, three, and four years. The short-horizon return variables allow us to control for transfers due to day traders.

(33.) Note that the coefficients across regressions where the story counts differ are not directly comparable, because the average magnitude of the news impact variable differs. The economic magnitudes, not the coefficients themselves, should be used to compare effects.

(34.) We also ran regressions including a post-Enron dummy variable to see if there was a discrete drop in employer stock holdings once the Enron scandal broke. We did not find any statistically significant or economically large coefficients, and the coefficient signs often indicated that the Enron scandal caused a small movement into employer stock.

(35.) When lags of [N.sub.p] are included in the regression, only the preceding day's lag is statistically significant. As so often in empirical exercises, we cannot rule out the existence of omitted variable bias, despite the robustness of our results to various specifications.

(36.) Hewitt Associates (2003).

(37.) Holden and VanDerhei (2001).

(38.) Choi and others (2002); Choi, Laibson, and Madrian (2005a); Duflo and Saez (2003).

(39.) We control for employer stock returns in excess of the S&P 500. Like Choi and others (2004b), we find that positive excess employer stock returns cause trading out of employer stock. We also add a dummy variable for the date when diversification restrictions on the match account were lifted at the company.

(40.) The coefficients in this regression and in the company F regression below are not directly comparable to those in the Hewitt Index regression, because the total economic effect depends upon the initial fraction of the portfolio held in employer stock, which differs across the three samples. The time trends should also not be expected to be similar across the three sets of regressions, because of firm-specific trends that are averaged away in the aggregate series.

(41.) The economic effect of one of the control variables in these regressions is anomalously large. The coefficients on the linear time trend imply a 60-percentage-point drop in employer stock holding due to the time trend. Since this would result in a negative average employer stock position in the absence of other forces, we do not interpret this magnitude literally but consider it an artifact of a nonlinearity (see figure 8) that is not well fit by a cubic polynomial.

(42.) We have data on employer stock balances for company F as of December 31, 2001, from the cross-sectional data.

(43.) Samuelson and Zeckhauser (1988).

(44.) Madrian and Shea (2001); Choi and others (2002).

(45.) Duflo and Saez (2003); Choi and others (2002); Choi, Laibson, and Madrian (2005a).

(46.) See Choi, Laibson, and Madrian (2005a).

(47.) See Sunstein and Thaler (forthcoming) for a general discussion of noncoercive prescriptive policies, which they call "libertarian paternalism."

(48.) Although employees in these plans are required to hold a significant fraction of this employer stock, the evidence in this paper and in others shows that these requirements are not binding at the margin for most employees, since voluntary employer stockholding is so high in these plans.

(49.) Choi and others (2005).

(50.) However, some perverse incentives would not be easy to remove. For example, stock held by employees is often viewed as a protection against hostile takeover (Rauh, forthcoming).

JAMES J. CHOI

Yale University

DAVID LAIBSON

Harvard University

BRIGITTE C. MADRIAN

University of Pennsylvania

Table 1. Workforce Characteristics of the Firms in This Study, End of
2003

 Average age of
Company employees (years) Percent male Percent married

A 44.9 76 n.a.
B 39.7 54 n.a.
C 39.0 35 n.a.
D 41.2 40 62
E 34.9 52 n.a.
F 44.8 28 n.a.
G (a) 37.3 64 n.a.
Memorandum: 39.7 53 59
U.S. workforce

 Average annual Median annual
Company salary salary Total employees

A n.a. n.a. ~20,000
B n.a. n.a. ~10,000
C n.a. n.a. 100,000+
D $57,422 $46,910 ~40,000
E $51,991 $37,040 ~10,000
F $66,253 $61,975 ~15,000
G (a) n.a. n.a. 100,000+
Memorandum: $36,522 $27,000 n.a.
U.S. workforce

Sources: Authors' calculations and March 2004 Current Population
Survey.

(a.) As of the end of 2002.

n.a. Not available.

Table 2. Features of the 401(k) Plans at the Firms in This Study, End
of 2003

 Percent of Employer match
 No. of assets in
 investment employer Percent of assets
Company options stock in match account

A 9 58 45
B 9 65 49
C (a) 14 69 50
D 7 28 25
E 8 21 24
F 11 46 36
G (a) 8 77 21

 Employer match

 Allowable employer
 stock
Company Match structure Vesting diversification

A 55 to 125 percent 100 percent Old: 25 percent at
 match, depending on immediate age fifty-five and
 profitability, up to 50 percent at age
 6 percent of pay sixty with ten
 contributed; directed years of tenure
 into ESOP New: 25 percent at
 age forty-five, 50
 percent at age
 fifty, 100 percent
 at age fifty-five
B 50 percent match up Five-year 25 percent at age
 to 6 percent of pay graduated fifty, 50 percent
 contributed; directed at age fifty-five,
 into ESOP 75 percent at age
 sixty with ten
 years of tenure
C (a) 100 percent match up Three-year 50 percent after
 to 5 percent of pay graduated ten years in plan
 contributed; directed or age fifty-five
 into ESOP with five years in
 plan; 100 percent
 at age 70 1/2, or
 age sixty with five
 years in plan
D 50 percent match up 100 percent Old: 25 percent at
 to 5 percent of pay immediate age fifty-five and
 contributed; 50 percent at age
 additional profit- sixty with ten
 contingent match years of tenure
 possible; directed New: 100 percent
 into ESOP immediately
E 100 percent match up Graduated 100 percent
 to 4 percent of pay (100 percent immediately
 contributed; directed at 5,000
 into employer stock hours)
F 75 percent match up Five-year 100 percent
 to 6 percent of pay graduated immediately
 contributed
G (a) 150 percent on first Three-year 100 percent
 1 percent of pay cliff immediately
 contributed; 50 per-
 cent on next 4 per-
 cent of pay contri-
 buted; directed into
 employer stock

Source: Plan documents. All companies offer employer stock as one of
the investment options.

(a.) Plan information as of end of 2002.

Table 3. News Stories on Enron, WorldCom, and Global Crossing, October
16, 2001, to December 31, 2003

 All stories

 Days with Total story
Medium stories count

Newspapers
New York Times 639 3,134
Wall Street Journal 534 3,294
Washington Post 681 2,374
Los Angeles Times 712 2,567
Chicago Tribune 327 678
Television
ABC News 182 700
NBC News 163 459
CBS News 246 768
All newspapers 810 12,047
All television 322 1,927
All newspapers and television 810 13,974

 Stories mentioning
 "401(k)" only

 Days with Total story
Medium stories count

Newspapers
New York Times 125 184
Wall Street Journal 115 157
Washington Post 134 176
Los Angeles Times 112 133
Chicago Tribune 86 111
Television
ABC News 26 37
NBC News 37 55
CBS News 43 52
All newspapers 332 761
All television 79 144
All newspapers and television 344 905

Source: Lexis/Nexis and Factiva databases.

Table 4. Effect of News about Enron, WorldCom, and Global Crossing on
Net Employer Stock Transfers, August 4, 1997, to November 4, 2003 (a)

 All stories

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0008 ** -0.0007 **
 (0.0003) (0.0002)
Return on S&P500 (c)
Same day -0.0149 ** -0.0150 **
 (0.0026) (0.0026)
Preceding day -0.0028 -0.0029
 (0.0023) (0.0023)
Preceding month -0.0013 -0.0012
 (0.0007) (0.0007)

Time trend (d)
Linear 0.0150 0.0140
 (0.0450) (0.0450)
Quadratic -0.0100 -0.0096
 (0.0173) (0.0173)
Cubic 0.0011 0.0011
 (0.0014) (0.0014)
Intercept 0.0370 0.0341
 (0.0502) (0.0504)
Adjusted [R.sup.2] 0.0821 0.0825
Total news effect (e) -2.35 -2.22

 All stories

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0008 * -0.0006 *
 (0.0004) (0.0003)
Return on S&P500 (c)
Same day -0.0148 ** -0.0148 *
 (0.0027) (0.0027)
Preceding day -0.0027 -0.0027
 (0.0024) (0.0024)
Preceding month -0.0012 -0.0012
 (0.0007) (0.0007)

Time trend (d)
Linear 0.0205 0.0209
 (0.0418) (0.0417)
Quadratic -0.0106 -0.0106
 (0.0168) (0.0167)
Cubic 0.0009 0.0009
 (0.0015) (0.0015)
Intercept 0.0589 0.0593
 (0.0402) (0.0399)
Adjusted [R.sup.2] 0.0814 0.0818
Total news effect (e) -2.00 -1.90

 Stories mentioning 401(k) only

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0075 -0.0057 *
 (0.0035) (0.0024)
Return on S&P500 (c)
Same day -0.0148 ** -0.0149 **
 (0.0027) (0.0027)
Preceding day -0.0027 -0.0027
 (0.0023) (0.0023)
Preceding month -0.0012 -0.0012
 (0.0007) (0.0007)

Time trend (d)
Linear 0.0153 0.0151
 (0.0449) (0.0449)
Quadratic -0.0097 -0.0096
 (0.0173) (0.0173)
Cubic 0.0011 0.0011
 (0.0014) (0.0014)
Intercept 0.0262 0.0254
 (0.0509) (0.0509)
Adjusted [R.sup.2] 0.0826 0.0826
Total news effect (e) -1.30 -1.20

 Stories mentioning 401(k) only

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0107 ** -0.0087 **
 (0.0041) (0.0032)
Return on S&P500 (c)
Same day -0.0148 ** -0.0148 **
 (0.0026) (0.0026)
Preceding day -0.0027 -0.0027
 (0.0023) (0.0023)
Preceding month -0.0012 -0.0012
 (0.0007) (0.0007)

Time trend (d)
Linear 0.0156 0.0160
 (0.0317) (0.0315)
Quadratic -0.0100 -0.0101
 (0.0140) (0.0139)
Cubic 0.0011 0.0011
 (0.0014) (0.0013)
Intercept 0.0302 0.0300
 (0.0316) (0.0316)
Adjusted [R.sup.2] 0.0830 0.0834
Total news effect (e) -1.79 -1.74

Source: Authors' calculations.

(a.) The dependent variable is aggregate net transfers of employer
stock in 401(k) plans, as a percentage of the preceding day's employer
stock balance. The unit of observation is a trading day. The sample
size in all regressions is 1,534. All regressions include
day-of-the-week dummies, month-of-the-year dummies, and additional
stock market return controls extending back four years. Number
parentheses are Newey-West (1987) standard errors with five lags.
* denotes statistical significance at the 5 percent level.

(b.) See the text for definition.

(c.) In percentage points.

(d.) Time is measured in years.

(e.) Calibrated percentage-point reduction in the fraction of the
total portfolio allocated to employer stock due to the aggregate
effect of the news impact variable over the sample period.

Table 5. Effect of News about Enron, WorldCom, and Global Crossing on
Net Employer Stock Transfers at Company D, December 1, 1997, to
December 31, 2003 (a)

 All stories

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0017 -0.0012
 (0.0025) (0.0017)
Company D excess return (c)
Today -0.0758 ** -0.0758 **
 (0.0229) (0.0229)
Previous day -0.0193 * -0.0193 *
 (0.0086) (0.0086)
Previous month -0.0051 * -0.0051 *
 (0.0024) (0.0024)
Return on S&P500 (d)
Today -0.0970 ** -0.0971 **
 (0.0190) (0.0190)
Previous day -0.0414 ** -0.0415 **
 (0.0137) (0.0137)
Previous month -0.0032 -0.0032
 (0.0055) (0.0055)

Time trend (e)
Linear -0.8247 ** -0.8258 **
 (0.3133) (0.3134)
Quadratic 0.2511 0.2514
 (0.1661) (0.1661)
Cubic -0.0195 -0.0195
 (0.0185) (0.0185)
Diversification dummy (f) 0.2484 0.2370
 (0.1338) (0.1288)
Intercept 0.3867 0.3873
 (0.7065) (0.7065)
Adjusted [R.sup.2] 0.1766 0.1766
Total news effect (g) -1.21 -0.99

 All stories

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) 0.0021 0.0013
 (0.0047) (0.0032)
Company D excess return (c)
Today -0.0761 ** -0.0761 **
 (0.0228) (0.0228)
Previous day -0.0194 * -0.0194 *
 (0.0085) (0.00085)
Previous month -0.0052 * -0.0052 *
 (0.0024) (0.0024)
Return on S&P500 (d)
Today -0.0967 ** -0.0966 **
 (0.0190) (0.0190)
Previous day -0.0410 ** -0.0409 **
 (0.0137) (0.0137)
Previous month -0.0030 -0.0030
 (0.0053) (0.0054)

Time trend (e)
Linear -0.8292 ** -0.8277 **
 (0.3055) (0.3053)
Quadratic 0.2653 0.2650
 (0.1662) (0.1621)
Cubic -0.0234 -0.0234
 (0.0184) (0.0184)
Diversification dummy (f) 0.1945 0.2103
 (0.1689) (0.1531)
Intercept 0.5650 0.5467
 (0.5215) (0.5229)
Adjusted [R.sup.2] 0.1767 0.1766
Total news effect (g) 1.35 0.99

 Stories mentioning 401(k)" only

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0037 -0.0141
 (0.0241) (0.0183)
Company D excess return (c)
Today -0.0759 ** -0.0758 **
 (0.0230) (0.0229)
Previous day -0.0194 * -0.0193 *
 (0.0086) (0.0086)
Previous month -0.0051 * -0.0051 *
 (0.0024) (0.0024)
Return on S&P500 (d)
Today -0.0969 ** -0.0970 **
 (0.0189) (0.0189)
Previous day -0.0412 ** -0.0412 **
 (0.0137) (0.0137)
Previous month -0.0029 -0.0032
 (0.0054) (0.0054)

Time trend (e)
Linear -0.8281 ** -0.8134 **
 (0.3136) (0.3144)
Quadratic 0.2530 0.2457
 (0.1661) (0.1665)
Cubic -0.0198 -0.0188
 (0.0185) (0.0185)
Diversification dummy (f) 0.2272 0.2251
 (0.1302) (0.1240)
Intercept 0.3839 0.3825
 (0.7077) (0.7058)
Adjusted [R.sup.2] 0.1765 0.1765
Total news effect (g) -0.15 -0.68

 Stories mentioning 401(k)" only

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) 0.0336 0.0139
 (0.0421) (0.0324)
Company D excess return (c)
Today -0.0762 ** -0.0761 **
 (0.0229) (0.0228)
Previous day -0.0195 * -0.0194 *
 (0.0085) (0.0085)
Previous month -0.0054 * -0.0052 *
 (0.0024) (0.0024)
Return on S&P500 (d)
Today -0.0970 ** -0.0967 **
 (0.0190) (0.0190)
Previous day -0.0413 ** -0.0411 **
 (0.0137) (0.0137)
Previous month -0.0031 -0.0031
 (0.0053) (0.0053)

Time trend (e)
Linear -0.8664 ** -0.8369 **
 (0.3070) (0.3074)
Quadratic 0.2824 0.2687
 (0.1631) (0.1632)
Cubic -0.0251 -0.0238
 (0.0185) (0.0185)
Diversification dummy (f) 0.1864 0.2209
 (0.1528) (0.1443)
Intercept 0.5488 0.5659
 (0.5222) (0.5249)
Adjusted [R.sup.2] 0.1771 0.1767
Total news effect (g) 1.33 0.65

Source: Authors' calculations.

(a.) The dependent variable is aggregate net employer stock transfers
in the employee contribution account, as a percentage of the preceding
day's employer stock balance. The unit of observation is a trading day.
The sample size in all regressions is 1,520. All regressions include
day-of-the-week dummies, month-of-the-year dummies, and additional
stock market return controls extending back four years. Numbers in
parentheses are Newey-West standard errors with five lags. * denotes
significance at the 5 percent level, and ** at the 1 percent level.

(b.) See the text for definition.

(c.) Annual rate of return on company stock in excess of the return on
the S&P 500, in percentage points.

(d.) In percentage points.

(e.) Time is measured in years.

(f.) Equals 1 if the date of transfer is after the employer began to
allow for full diversification of matching contributions for all
employees.

(g.) Calibrated percentage-point change in the fraction of the total
portfolio allocated to employer stock due to the aggregate effect of
the news impact variable over the sample period.

Table 6. Effect of News about Enron, WorldCom, and Global Crossing on
Net Employer Stock Transfers at Company F, January 1, 2002, to December
31, 2003 (a)

 All stories

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0003 -0.0013
 (0.0041) (0.0026)
Company F excess return (c)
Today -0.2038 ** -0.2041 **
 (0.0397) (0.0397)
Previous day 0.0407 * 0.0406 *
 (0.0204) (0.0204)
Previous month 0.0019 0.0020
 (0.0061) (0.0062)
Return on S&P500 (d)
Today -0.1632 ** -0.1632 **
 (0.0496) (0.0497)
Previous day 0.1188 ** 0.1189 **
 (0.0369) (0.0370)
Previous month -0.0127 -0.0125
 (0.0108) (0.0108)

Time trend (e)
Linear -1.2238 * -1.2660 *
 (0.5983) (0.5740)
Quadratic 0.9555 0.9848
 (0.6023) (0.6029)
Cubic 1.4361 * 1.4536 *
 (0.6105) (0.5924)
Intercept -1.1622 -1.1222
 (0.7088) (0.7101)
Adjusted [R.sup.2] 0.2333 0.2334
Total news effect (f) -0.48 -2.87

 All stories

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0031 -0.0017
 (0.0055) (0.0036)
Company F excess return (c)
Today -0.2038 ** -0.2039 **
 (0.0397) (0.0397)
Previous day 0.0406 * 0.0406 *
 (0.0204) (0.0204)
Previous month 0.0020 0.0020
 (0.0062) (0.0062)
Return on S&P500 (d)
Today -0.1628 ** -0.1630 **
 (0.0496) (0.0496)
Previous day 0.1192 ** 0.1190 **
 (0.0372) (0.0371)
Previous month -0.0124 -0.0125
 (0.0109) (0.0108)

Time trend (e)
Linear -1.3465 * -1.2988 *
 (0.6503) (0.6159)
Quadratic 1.0106 0.9977
 (0.6078) (0.6088)
Cubic 1.5269 * 1.4883 *
 (0.6453) (0.6228)
Intercept -1.1483 -1.1479
 (0.7020) (0.7020)
Adjusted [R.sup.2] 0.2334 0.2334
Total news effect (f) -5.21 -3.32

 Stories mentioning
 "401(k)" only

 [delta] = 0.25

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0048 -0.0140
 (0.0353) (0.0223)
Company F excess return (c)
Today -0.2036 ** -0.2032 **
 (0.0396) (0.0396)
Previous day 0.0409 * 0.0412 *
 (0.0203) (0.0203)
Previous month 0.0019 0.0019
 (0.0061) (0.0061)
Return on S&P500 (d)
Today -0.1631 ** -0.1631 **
 (0.0497) (0.0498)
Previous day 0.1189 ** 0.1191 **
 (0.0370) (0.0370)
Previous month -0.0128 -0.0132
 (0.0108) (0.0108)

Time trend (e)
Linear -1.2258 * -1.2397 *
 (0.5635) (0.5548)
Quadratic 0.9656 1.0081
 (0.6054) (0.6036)
Cubic 1.4348 * 1.4264 *
 (0.5894) (0.5843)
Intercept -1.1631 -1.1496
 (0.6996) (0.7004)
Adjusted [R.sup.2] 0.2333 0.2335
Total news effect (f) -0.51 -1.74

 Stories mentioning
 "401(k)" only

 [delta] = 0.90

 Newspapers Newspapers
Independent variable only and TV

News impact (b) -0.0002 -0.0017
 (0.0513) (0.0304)
Company F excess return (c)
Today -0.1632 ** -0.1632 **
 (0.0496) (0.0497)
Previous day 0.1188 ** 0.1189 **
 (0.0371) (0.0370)
Previous month -0.0127 -0.0129
 (0.0108) (0.0108)
Return on S&P500 (d)
Today -0.2038 ** -0.2037 **
 (0.0396) (0.0396)
Previous day 0.0408 ** 0.0408 **
 (0.0204) (0.0204)
Previous month 0.0019 0.0019
 (0.0061) (0.0061)

Time trend (e)
Linear -1.2140 * -1.2182 *
 (0.6000) (0.5823)
Quadratic 0.9507 0.9562
 (0.6128) (0.6129)
Cubic 1.4300 * 1.4315 *
 (0.6079) (0.5971)
Intercept -1.1670 -1.1665
 (0.6980) (0.6985)
Adjusted [R.sup.2] 0.2333 0.2333
Total news effect (f) -0.02 -0.21

Source: Authors' calculations.

(a.) The dependent variable is aggregate net employer stock transfers,
as a percentage of the preceding day's employer stock balance. The unit
of observation is a trading day. The sample size in all regressions is
504. All regressions include day-of-the-week dummies,
month-of-the-year dummies, and additional stock market return controls
extending back four years. Numbers in parentheses are Newey-West
standard errors with five lags. * denotes significance at the 5 percent
level, and ** at the 1 percent level.

(b.) See the text for definition.

(c.) Annual rate of return on company stock in excess of the return on
the S&P 500.

(d.) In percentage points.

(e.) In years.

(f.) Calibrated percentage-point reduction in the fraction of the total
portfolio allocated to employer stock due to the aggregate effect of
the news impact variable over the sample period.