Inefficient Choices in 401(k) Plans: Evidence from. Individual Level Data

Transcription

1 8/16/2002 Inefficient Choices in 401(k) Plans: Evidence from Individual Level Data Julie Agnew* The College of William and Mary *The College of William and Mary, School of Business Administration, P.O. Box 8795, Williamsburg, Virginia Tel. (757) The author thanks CitiStreet for providing the 401(k) plan data and Stefan Bokor for his immense help in organizing the data. The author thanks her dissertation committee, Pierluigi Balduzzi (Chair), Alicia Munnell, Eric Jacquier and Peter Gottschalk, for their careful comments and guidance. In addition, the author is grateful to Shlomo Benartzi for his comments and insight. Finally, the author thanks conference participants from the Retirement Research Consortium Fourth Annual Conference and the Frank Batten Young Scholars Conference. She gratefully acknowledges financial support from a dissertation fellowship from the Center for Retirement Research. Any errors are my own. The research reported herein was performed pursuant to a grant from the U.S. Social Security Administration (SSA) funded as part of the Retirement Research Consortium. The opinions and conclusions are solely those of the author and should not be construed as representing the opinions or policy of SSA or any agency of the Federal Government. 1

2 Inefficient Choices in 401(k) Plans: Evidence from Individual Level Data Abstract: This paper investigates how individual characteristics, such as age, salary, job tenure, and gender, influence an individual s decision to over-invest in company stock and follow naïve diversification rules. Using a new and unique data set from one 401(k) plan with over 73,000 eligible employees, the results suggest that individual characteristics do influence company stock holdings. Ordered probit regression results indicate that the probability of over-investing in company stock for the average participant is greater for males, decreases with salary, increases with past company stock performance and is related to the participant s division of employment. In addition, the paper investigates if individuals tend to follow a simple diversification rule, the 1/n heuristic. According to this rule, some 401(k) investors will choose to divide their contributions evenly among the n options available regardless of the type of investment vehicles offered. While the percentage of individuals who follow the 1/n heuristic in this study is lower than that found in previous studies, it still represents 5% of the sample. Probit regression analysis results suggest that the probability of following the 1/n heuristic decreases with increases in job tenure and salary. Finally, consistent with the mental accounting literature, participants demonstrate a tendency to treat company stock as a separate account. The evidence from this study indicates that over-investing in company stock and practicing naïve diversification strategies may be occurring frequently in 401(k) plans and that certain individuals are more prone to make these decisions than others. Given the pivotal role 401(k) savings play in individuals retirements, this research may help plan sponsors design plans and train participants to make better-informed decisions. 2

3 Introduction 401(k) holdings are projected to become the largest asset an individual owns, with the possible exception of his/her home (Sahadi (2001)). As a result, the financial security of most individuals retirements will depend on how well their 401(k) portfolio has performed. Thus, the importance of an individual s asset allocation choices cannot be overstated. As the debate over Social Security moves towards private accounts, these allocation decisions take on even greater significance. Unfortunately, empirical research focusing on the quality of asset allocation choices has been difficult to conduct in the past because of a lack of detailed individual level data. This paper overcomes this obstacle by taking advantage of a newly available database from one 401(k) plan with over 73,000 eligible participants. The fine level of detail in these data provides a unique opportunity to analyze two potentially inefficient and commonly made retirement choices in 401(k) plans: over-investing in company stock and following naïve diversification strategies. While the influence of plan design on these decisions has already been documented, this research contributes to the literature by highlighting how individual characteristics, such as age, salary, job tenure and gender, relate to the efficiency of allocation decisions. It is the first study to examine these two issues in one paper. The perils of over-investing in company stock are well known as a result of several recent high profile cases, including Enron, Lucent and WorldCom, where many individuals lost most of their 401(k) nest eggs by following this strategy. Given the benefits of diversification, it has long been a puzzle why so many individuals invest a substantial amount of their retirement funds in the stock of their own company. Not only are they concentrating their assets in a single stock, which is more risky than a well- 3

4 diversified portfolio, but they are investing in a security that is highly correlated with their own human capital. The risks of allocating a large portion of an individual s portfolio to company stock or even more simply to just one stock are so great that legislation limits such investments in defined benefit plans and mutual funds. 1 Yet despite these laws and evidence that individuals are prone to over-invest in company stock, the majority of 401(k) plans (80 percent) that offer company stock place no restrictions at all on company stock allocations (Dugas (2000)). Given the absence of restrictions, understanding what factors might lead an investor to over-invest in company stock is important. The influence of past company stock performance and plan design on company stock holding is already well documented in the literature. Benartzi (2001) finds that company stock holdings are higher for companies with relatively strong long-run stock performance and for companies with employer stock only matches. However, until now, no one has tested for an additional link between individual characteristics and company stock allocations. This paper fills this gap. It is the first paper to jointly test the significance of past company stock performance and demographic factors on company stock allocations. In addition to investigating the determinants of company stock holdings, this paper also examines the practice of naïve diversification strategies. The psychology literature predicts that naïve diversification strategies will result when individuals are faced with complicated decisions. The complexities of these decisions cause them to fall back on simple rules of thumb. This paper investigates one particular strategy, the 1/n heuristic, studied by Bernatzi and Thaler (2001). According to this strategy, some 401(k) investors overwhelmed by their investment choices will choose to simply divide their contributions 4

5 evenly among the n investment options offered. They do so regardless of the type of investment options they are given. While it can be argued that this rule of thumb will often lead to a diversified portfolio, Benartzi and Thaler (2001) show that it can also lead to large ex ante welfare losses when the portfolio chosen does not correspond to the individual s risk preferences. 2 This paper investigates the extent to which individuals follow this rule in this plan and if individual characteristics matter in this decision. Finally, this paper investigates whether individuals treat company stock as a separate asset class from other equities. Benartzi and Thaler (2001) find evidence supporting this practice using aggregate 401(k) plan data. As a result when company stock is an option in a 401(k) plan, Benartzi and Thaler (2001) predict that participants will choose to divide their non-company stock contributions evenly among the non-company stock options. In this paper, I call this the modified 1/n heuristic. Benartzi and Thaler (2001) test their prediction using a sample of (k) plans offering company stock and 67 plans that do not offer company stock. They find that the mean allocations of the asset balances for the plans not offering company stock are split approximately 50/50 between equities and fixed income. In contrast, they find that the mean allocation to equities is over 71 percent when company stock is an option. Looking closer, they find an average allocation of 42 percent to company stock. The remainder is divided almost evenly between non-company stock equities (29 percent) and fixed income (28 percent). This supports the modified 1/n heuristic. Treating company stock as a separate account does have theoretical foundations. This investment behavior is consistent with Shefrin and Statman s (2000) multi-account behavioral portfolio theory. 3 According to their theory, individuals have difficulty 5

6 processing covariances and other properties related to joint probability distributions so they simplify their decision making by separating their assets into different mental accounts. They ignore any correlation between these accounts when making their portfolio decisions. In the case of 401(k) plans with company stock, it appears that individuals are ignoring the correlation between their company stock account and their other financial investment account. The detail of the individual level data in this paper allows for a stronger test of the modified 1/n heuristic. In addition, the data in this study avoid problems associated with allocation drift of asset balances because the allocations are based on contribution allocations. Three main results emerge from the analysis of company stock investments. First, the company stock investment patterns of individuals tend to be multimodal, with 78 percent of the sample centered on allocations of 0, 25, 50, 75 and 100 percent. Second, over 75 percent of the individuals invest more in company stock than the maximum limit of 10 percent permitted by law in defined benefit plans. Third, results from ordered probit regressions that control for past company stock performance suggest that the probability of over-investing in company stock for the average participant is greater for males, decreases with salary and is lower for corporate division workers. The results of the analysis of the 1/n heuristic indicate that some individuals do appear to follow the 1/n heuristic. However, the percentage is not as large as that found in Benartzi and Thaler s (2001) survey results. Kernel densities suggest that individuals are treating company stock investment as a separate investment from other equities confirming Benartzi and Thaler s work. Finally, regression analysis results suggest that 6

7 the probability of following the 1/n heuristic decreases with increases in job tenure and salary/compensation status. Interestingly, individuals earning relatively higher salaries tend to make more efficient decision related to both their company stock allocations and diversification strategies than others. The remainder of the paper is organized as follows. Section I summarizes the data set. Section II describes the plan design and asset allocation choices. Section III discusses the participation level in this plan and provides an overview of the possible reasons why it is relatively low. Section IV summarizes the demographic and employment characteristics of the actively contributing participants. Section V and VI present the empirical results associated with company stock holdings and naïve diversification, respectively. Section VII concludes. I. Data This paper uses a detailed database supplied by CitiStreet. The cross-sectional data are from one large 401(k) plan with over 73,000 eligible employees. 4 The plan is sponsored by a global consumer product company. During the first two weeks of August 1998, 28,809 of the eligible participants made contributions. For the purpose of this study, these participants are considered active participants. The data set includes each active participant s contribution allocations and, for most participants, the actual date this allocation was chosen. CitiStreet took over administration of this 401(k) plan in As a result, some data are missing for participants who entered the plan prior to the administration change. In particular, the actual date the participant made their allocation decision is unavailable for 5,814 participants who were employed prior to 1992 and did not change their 7

8 contribution allocations during State Street s tenure. An indicator variable identifies these individuals so that the later analysis can control for the possible effects of administration style on asset allocation decisions. For these individuals, the date they chose their contribution allocation is estimated as either the date of employment or the date the plan began in 1983, whichever is later. Since these individuals did not change their allocations between , it seems reasonable to assume that they did not actively change their allocations prior to Therefore, their current allocations are most likely unchanged from the initial allocation decision they made when they first entered the plan. For each allocation decision date, the past company stock returns over several buy and hold periods are calculated. One of the important features of these data is that detailed demographic information is available for each eligible participant in the plan including each individual s participation status, salary, birthdate, date of employment, compensation status, and gender. For the empirical analysis, age and time employed are calculated based on the allocation decision date. This data set has three noteworthy features. First, the asset allocations of the contributions are broken down at the individual level. Aggregate contribution plan data can blur the results if high contributing participants invest differently than low contributing participants. The effect of large contribution levels is analogous to the influence of large market capitalization stocks on a value-weighted index. Second, the data are from one plan. While multiple plan data are appropriate for studying across-plan variation, they can create a potential for omitted variable bias related to plan design or plan educational efforts. Analyzing one plan eliminates this concern. A final advantage of 8

9 the data is that allocations are based on contributions not asset balances. Asset performance can move asset allocations based on asset balances away from the participant s intended allocation. Contribution allocations do not suffer from this potential bias. A disadvantage of these data is that information regarding participants assets outside of the plan is not available. However, the evidence suggests that investors tend to invest their retirement savings in the same fashion as their non-retirement assets (Uccello(2000)). Another drawback of the data set is that it is missing some variables that have been shown to impact asset allocation decisions: such as marital status, education and financial literacy (eg. Agnew, Balduzzi and Sunden (2001), Sunden and Surrette (1998), Dwyer, Gilkeson and List (2000)). II. Plan Design and Asset Choices In this plan, each participant may allocate his/her retirement fund contributions among four different investment vehicles: an equity income fund, an S&P 500 index fund, a guaranteed income contract fund (GIC), and company stock. Participants have the option to change their contribution allocations daily. The company offers no financial incentive for investing in company stock nor do they offer an employer match. The absence of an employer match is an advantage because it eliminates any confounding effects caused by the match design. III. Plan Participation Of the 73,721 eligible participants, 39 percent made at least one contribution during the first two weeks of August This participation rate is low compared to other studies. 5 The low participation rate observed in this plan might be a result of several 9

10 factors shown to decrease participation in the academic literature. One of the main reasons might be that the plan does not offer an employer match (Munnell, Sunden, and Taylor (2000), Papke and Poterba (1995)). Another is that the company offers a defined benefit plan. Many studies (e.g. Andrews (1992), Bernheim and Garrett (1996)) have shown that employees tend to participate less in their 401(k) plan if their company offers a pension plan. Another possible factor is the relatively large size of this plan. This plan is considered part of the large plan market (over 25,000 participants) and Clark and Schieber (1998) find that the probability of participation decreases with the size of the company. On the other hand, the plan sponsor did offer educational services, including seminars and literature. These forms of corporate communication and education have been shown to increase levels of participation (Munnell, Sunden and Taylor (2000), Clark and Schieber (1998), Bernheim and Garrett (1996)). It is unclear how these educational efforts affected participation in this plan. Finally, the definition of active participant in this study is fairly restrictive because it limits participants to those who made a contribution during the first two weeks of August Other studies use different definitions. For example, Clark and Schieber (1998) define an active participant as a person who made at least one contribution in the year IV. Demographic and Employment Characteristics of Active Participants Panel A of Table I describes the demographic and employment characteristics of the active participants. Age and time employed are measured as of August 1998, while salary is the 1997 annual salary. Individuals in this data sample are predominately male (78 percent) with an average age of 39 years old. It is noteworthy that the participants have relatively long average job tenures (10 years), which may indicate strong company 10

11 loyalty. Interestingly, the median time employed is approximately eight and half years and is over double the 1996 national median of nearly four years (CPS (1997)). In the plan, nine percent of the sample are considered highly compensated individuals. This is a legal designation based on several factors including salary. This status affects how much a participant can contribute but does not restrict their allocation decisions. Participants in the company work in one of four different divisions with the majority of the participants (99 percent) working in two large consumer product manufacturing divisions, Division 1 and Division 2. The Corporate Division employs 1% of the 401(k) participants and under 100 employees work for the Other Division. Participants earned mean 1997 salaries of approximately $46,000. Panel B of Table I compares the plan s median salary by age group to the median salary of the U.S. population. The table shows that participants in this plan earn more than the general population. However, the relationship between salary and age is similar between the two groups with the exception of the 65+ age group. The discrepancy in the 65+ group may be due to the limited number of participants (29) in this age group in this data set. Table II describes the demographic characteristics by division. The main difference between the four divisions appears to be the salary distributions. Employees of the two smallest divisions make significantly higher salaries than the other divisions. The Corporate Division s mean salary is approximately $107,000, while the Other Division s mean salary is nearly $140,000. These salaries compare to approximately $48,000 and $44,000 earned in Division 1 and 2, respectively. Employees in the two small divisions also earn significantly more in the tenth and ninetieth percentiles of their sample and they 11

12 are more likely to be highly compensated individuals. Except for the Corporate Division, the divisions are predominately male. The groups do not differ significantly in terms of average age or time employed. V. Company Stock Allocations A. General Findings Consistent with anecdotal evidence, participants in this 401(k) plan show a tendency to over-invest in company stock. The overall mean allocation to company stock holdings in this plan is quite high (49 percent) compared to the 10 percent legal maximum defined benefit plans may hold. The large average allocation might be partially explained by the above normal price performance of the plan s company stock. In this study, the company stock had an annualized stock price return of 20.6 percent over the 10-year period ending on December 31, 1997, compared to a S&P 500 s annual return of 14.7 percent over the same time period. Benartzi (2001) shows that firms with relatively high returns over the previous ten years have higher company stock allocations than poor performing firms. This finding motivates why past company stock performance is controlled for in the later regression analysis. The general patterns of company stock allocations also deserve mention. One interesting feature of the data is that despite the absence of restrictions on the participants allocations, 78 percent of the allocations are clustered within one percentage point of zero, 25, 50, 75 and 100 percent. Furthermore, there is a clear tendency for many of the participants (52 percent) to invest either all or none of their contributions to company stock. 12

13 B. Nonparametric Analysis Some interesting trends in company stock allocations emerge when these allocations are summarized based on the participant s demographic characteristics. This is not surprising because demographic characteristics can proxy for several factors that can affect a participant s company stock holdings including financial knowledge, nonretirement company stock holdings, risk preferences, company loyalty, and perceived influence and knowledge of the company. This section will describe in detail why these demographic characteristics could proxy for these factors. It presents a nonparametric analysis of the data that will complement the regression analysis to follow. Table III reports the company stock allocations based on demographic characteristics. The non-normal distribution of the company stock holdings makes standard summary statistics, such as means and standard deviations, less meaningful descriptors of the data. Therefore, in addition to these statistics, Table III reports the proportion of each demographic category that invests in six different investment ranges: zero percent, 1-25 percent, percent, percent, percent, and 100 percent. A simple test of proportions within each demographic category and investment range is used to test whether a statistically significant difference exists. If demographic characteristics do not matter, then a statistically significant difference in proportions should not be found. For example, under the null hypothesis gender does not matter. Therefore, the proportion of women investing 100 percent of their contributions to company stock should not be statistically different than the proportion of men investing 100 percent of their contribution to company stock. Notice that within each demographic category in Table III the top row is bolded. This row is considered the base category. For each 13

14 demographic group this base category is used in each test of proportions. Table III reports the results of the test of proportions. Two (one) stars beside the proportions denote a statistically significant difference from the base category at the one (five) percent level. The first demographic category tested is gender. Empirical evidence suggests that gender may proxy for financial education or risk tolerance. For example, research shows that when a measure of financial education is not available, gender may serve as an effective proxy for it. Dwyer, Gilkeson and List (2000) find that women typically have less financial knowledge than men and that the educational disparities can substantially explain the gender differences they find in risky mutual fund allocations. Indeed, there is broad evidence suggesting that individuals overall lack a general understanding of the risks associated with company stock investment and that education may explain much of the variation in financial aptitude. A recent John Hancock Financial Services survey highlights how individuals misread the risks of the market. In the survey, respondents on average thought that a diversified stock fund was more risky than investment in company stock. Benartzi (2001) reports an equally disturbing result that 84% of respondents to a Morningstar survey believe that the overall stock market is riskier than company stock. When this sample is limited to individuals with high school education or less, this number increases to 94%. Therefore if gender proxies for differences in education, men might be expected to invest less in company stock than women. On the other hand, empirical research has found that men are more likely to invest in riskier assets than women leading to the opposite conclusion (for example, Sunden and Surette (1998), Hinz, McCarthy and Turner (1997), Agnew, Balduzzi and Sunden (2001), Barber and Odean (2001)). 14

15 The tests of proportions support the latter. In all but the 51-75% range, there is a statistically significant (albeit economically small) difference in the proportion of men investing in each investment range than women. The most pronounced difference between the proportion of women and men investing in company stock is at the 100 percent investment range. Observe that 29 percent of the men allocate their entire contribution to company stock compared to 26 percent of the women and that this difference is significant at the one percent level. Furthermore, the mean allocation to company stock by men is 50% compared to 47% for women. Interestingly, this gender difference is smaller than that found by Clark, Goodfellow, Shieber and Warwick (2000). In their study of several 401(k) plans, men invested an average 41% to company stock compared to 27% for women. However, these gender differences could be a result of different plan designs or varied long run company stock performance across plans. The next two sections of the table demonstrate the influence of compensation level, either salary or compensation status, on company stock investment. Compensation level is another potential proxy for financial knowledge. Generally, financial knowledge is considered positively related to compensation. This leads to the hypothesis that employees who earn relatively high salaries or are considered highly compensated should hold less in company stock. Alternatively, compensation may be proxying for an employee s opportunities for stock based compensation. Generally, greater opportunities exist for higher salaried employees to receive stock based compensation than for their lower wage counterparts. This is the case in this plan where three stock option plans are offered. One plan is open to all full-time employees and the number of options available is based on earnings. The 15

16 second and third plans are targeted at middle and senior management. The options in these plans are based on reaching performance goals. Thus, higher salaried and middle and upper management employees have more opportunities to earn stock options than lower salaried employees. Research shows that highly paid executives are concerned about diversifying their company stock holdings but are often reluctant to sell their stock based compensation. As a result they are finding sophisticated ways to hedge their holdings. Results from one recent paper suggest that executives diversify their company stock holdings through the use of zero cost collars and equity swaps (Ofek and Yermack (2000)). Additional research shows that executives with high stock ownership negate much of the impact from their stock compensation by selling previously owned shares (Bettis, Bizjak and Lemmon (2000)). Given the demonstrated lengths that these employees go to diversify their holdings, one might expect that these employees will hold smaller amounts of company stock in their 401(k) accounts. The results support both theories. Table III shows a decrease from the lowest wage category to the highest wage category in the proportion of individuals allocating their entire contribution to company stock that is significant at the one percent level. Twenty eight percent of the under $25,000 category invest their whole contribution to company stock compared to 24 percent of the $100,000 plus category. The reverse trend is observed in the proportion of individuals who invest nothing in company stock. Here, 23 percent of the under $25,000 category invest nothing in company stock compared to 36 percent of the over $100,000 group. This difference in proportions is significant at the one percent level. This supports results from Goodfellow and Schieber s (1997) study of 24 different plans where low-wage earners were more likely to hold company stock than 16

17 high-wage earners. Table III also shows fewer of the highly compensated individuals allocating all of their contributions to company stock and more of these individuals allocating none of their contributions to company stock. Similar to gender, age may proxy for risk tolerance. Many life cycle theories predict that individuals will hold less risk in their financial portfolio as they age. Jagannathan and Kocherlachota (1996) suggest that young investors have a long stream of future income. As individuals age, this stream of future income shortens diminishing the value of their human capital. Therefore, they suggest that individuals should offset this decline in the value of their human capital by reducing the risk of their financial portfolio. Bodie, Merton, and Samuelson s (1992) model leads to a similar prediction. In their model, individuals can respond to low realized asset returns by increasing their supply of labor. However, labor flexibility generally declines with age. Therefore, like the previous model, older individuals are expected to hold more conservative investments in their financial portfolios. Table III is consistent with the stated life cycle hypotheses. Note that in Table III, age is measured at the time the allocation decision is made. The 65 plus age category will not be discussed because it includes only 6 participants. Notice the downward trend as individuals age in the proportion of participants investing their entire contribution to company stock. On the extreme ends, 19 percent of those between years old invest their entire contribution to company stock compared to 31 percent of the participants under 35 years old. The difference in proportions is significant at the one percent level. This trend is reversed and significant in the proportions investing nothing in company stock. 17

18 Time employed may also proxy for risk tolerance. One hypothesis is that a participant s human capital is more stable the longer the individual is employed. Therefore, employees with longer job tenures may choose to invest more in company stock. An alternative hypothesis is that individuals with long job tenures may feel that they have more influence on the company s performance than new employees do so they invest more. Finally, the length of an employee s tenure may be a sign of company loyalty or familiarity with the company. This also leads to the prediction of a positive relationship. The final hypothesis is based on Huberman s (1998) work. Huberman coined the phrase familiarity bias to describe the tendency of investors to invest heavily in what they know. He suggests that this is a reason for high company stock holdings in 401(k) plans. However, Benartzi (2001) presents conflicting empirical evidence that shows that the impact of familiarity on company stock holdings is insignificant when past company stock performance is included in the regression analysis. Table III results report a negative relationship between time employed and company stock holdings which is contrary to the predictions of all three theories. Here the proportion of participants investing their entire contribution to company stock has a marked decline with time employed. Thirty four percent of those with less than two years experience invest their entire contribution to company stock compared to 22 percent of those with greater than 26 years experience. It is not surprising that these results are similar to the age findings because these variables are most likely highly correlated. 18

19 The results show that the employee s company division also explains some variation in company stock holdings. One possible explanation for this is that there might be a significant difference in the predominate occupation of the employees in each division. Among other things, occupation type might proxy for the probability of earning stock based compensation. For example, a corporate division may be more heavily concentrated with executives who earn greater stock based compensation than employees of a division predominately comprised of factory workers. Thus, the expected average allocation to company stock would be relatively lower in the corporate division compared to the other division. The occupation type may also provide additional information about the employee s education level beyond that obtained from salary information. It seems reasonable to assume that a corporate division may be more heavily comprised of executives with college degrees, while a factory division may have a high percentage of blue-collar workers who are predominately high school graduates. On the other hand, the division variables may also proxy for many other unobservables so care must be taken not to over interpret these results. In this study, the predominate occupation does differ between divisions. A discussion with CitiStreet indicates that the Corporate Division consists mainly of executives, while the employees of Division 1 and Division 2 tend to be factory workers. As predicted, Table III shows the Corporate Division has the lowest proportion of individuals investing their entire contribution to company stock and the highest proportion of individuals who invest nothing in company stock. These results support the theory that either the executives in the Corporate Division are limiting their company stock holdings to 19

20 compensate for stock based compensation or they are doing so because they have a relatively better understanding of the inherent risks of company stock investment. Finally, an indicator variable highlights the participants that enrolled in the plan prior to CitiStreet s administration and did not make allocations during CitiStreet s tenure. This variable controls for the possible influence of plan administration style. Observe that the difference in the proportions is striking and statistically significant at every investment range between the two groups. For example, 50 percent of those who made no allocation changes during CitiStreet s tenure invested their entire contribution to company stock compared to 23 percent of those who did make a change or entered the plan during CitiStreet s tenure. This finding strongly supports the use of this variable as a control variable in the following regression analysis. C. Econometric Analysis of Company Stock Holdings The nonparametric evidence suggests that there are relationships between the demographic variables and company stock holdings. This section will econometrically test for the joint effects of these factors on company stock allocations. In addition, it will control for the effects of past company stock performance. Generally, a two-limit censored regression model is used in studies of asset allocations. For example, Agnew, Balduzzi and Sunden (2001) use this model to study the relationship between demographic characteristics and equity allocations in one 401(k) plan. However, given the prevalence of company stock allocations clustered at 0, 25, 50, 75 and 100 percent, an ordered probit regression seems more appropriate for these data. Therefore for the econometric analysis, the company stock allocations are grouped into 20

21 six categories (0%, 1-25%, 26-50%, 51%-75%, 76%-99%, 100%) and an ordered probit regression is used to study the effects of the individual characteristics on company stock allocations. 6 Table IV reports the marginal effects from the ordered probit regression for each asset allocation range. For the average participant, the marginal effects show the change in probability of staying in that investment range given a small change in the independent variable. In the case of indicator variables, the marginal effect is for a discrete change of the indicator variable from zero to one. It is clear that most of the economically significant variation is at the extreme investment ranges (0% and 100%). The intermediate investment ranges marginal effects while statistically significant are not economically meaningful. Therefore, the discussion will concentrate on the 100 percent range. The results suggest that men are 3.3 percent more likely to invest their entire contributions to company stock than women, supporting the theory that men tend to make more risky asset allocation choices. Salary is also significantly related to company stock holdings. The results suggest that an average employee earning $100,000 is 3.7 percent less likely to invest his/her entire contribution to company stock than an average employee earning $40,000. In this case, salary may be a proxy for financial education or the amount of stock based compensation. Given the nonparametric results, it is not surprising that individuals who made their allocation decision during the previous administrator s tenure are 20 percent more likely to invest everything in company stock. The division of employment also has a significant role in company stock holdings. Relative to the Corporate Division, participants in Division 1 are 4 percent more likely to 21

22 invest in company stock. Similarly, participants in Division 2 are 10 percent more likely. The results support the hypothesis that either the executives in the Corporate Division are limiting their company stock holdings to compensate for stock based compensation or they are limiting their company stock holdings because they have a relatively better understanding of the inherent risks in company stock investment. Interestingly, age and time employed are not significantly related to company stock holdings. Finally, the results support Benartzi s (2001) findings that past raw buy and hold returns are positively related to company stock holdings. 6 The reported regression includes one year buy and hold company stock returns measured prior to each individual s allocation decision date. The sample returns range from a minimum of negative 26 percent to a maximum of 86 percent. The average return is 26 percent, with a standard deviation of 21 percent. The regression predicts that a 10 percent increase in the one year return of the company stock relevant to its sample average increases the probability of investing the entire contribution to company stock by 1.6 percent. VI. Naïve Diversification A. General Findings The analysis now shifts from a study of the factors related to a participant overinvesting in company stock, to the factors related to the practice of naïve diversification strategies. One of the main findings is that the frequency of participants following the 1/n heuristic is less than the percentage found in survey tests conducted by Benartzi and Thaler (2001). In their study, several different surveys were mailed to University of California employees. The surveys asked the employees how they would allocate their investments to a selection of funds. The results vary with the investment choices offered 22

23 but, in general, over 20 percent of each surveyed group chose the 1/n option. In this study less than four percent follow the 1/n heuristic and five percent follow the modified 1/n heuristic. In fact, most participants (41 percent) allocate their entire contribution to only one fund. Looking closer, the majority (70 percent) of those participants that invest in only one fund invest their entire contribution to company stock. The percent of the sample decreases with the number of funds held. A little under twelve percent hold all four funds. One possible explanation for why these results differ from previous empirical work is that the participant s choice may be influenced by the manner in which the allocations are selected. In Benartzi and Thaler s (2001) survey, participants selected their allocations by filling out a form with the possible investment choices listed. In this study s plan, individuals chose allocations over the phone using an automated system. Interestingly, the company stock option was the last option described on the phone. The ordering of the company stock option might have contributed to its popularity. There is some psychology literature that suggests that the ordering of choices influences decision making, with either the options listed first (primacy effects) or listed last (recency effects) having the largest influence. Understanding how these two approaches to allocation selection ultimately influence the participant s final choice is an interesting area for future research. 23

24 B. The Modified 1/n Heuristic Using aggregate data, Benartzi and Thaler (2001) find evidence that individuals treat company stock as a separate asset class from other 401(k) investments. As a result, the participants tend to split their non-company stock investment evenly among the noncompany stock options. This paper refers to this practice as the modified 1/n heuristic. This section investigates whether similar behavior applies to this plan. This analysis complements previous work because it provides a stronger test of this practice. It is a stronger test for two reasons. First, the tests in this paper are based on contribution allocations rather than asset balance allocations. Therefore, the influence of fund performance on allocations is not a concern. Second, the individual level data allows for the calculation of the allocation of non-company stock holdings by individual rather than by plan. This permits the examination of the distribution of company stock holdings across individuals. The analysis begins with an examination of the mean and median allocations to each fund in Table V. The first two columns of Table V list the mean and median allocations to each fund and the last two columns list the adjusted mean and median allocations to each fund. The adjusted allocations are simply the percent allocated to the particular noncompany stock investment vehicle divided by the total invested in non-company stock investment vehicles. The first subsample includes all the participants that invest in all four funds and comprises roughly eleven percent of the sample. Notice that the adjusted allocations are very close to 33.3%, which equates to evenly splitting the non-company stock contributions among the non-company stock assets. The same exercise is repeated 24

25 for subsamples of investors that hold three funds (including company stock). The results again support the modified 1/n with the adjusted allocations close to 50 percent. Kernel densities further support the findings. Panel A of Figure 1 displays a kernel density representing company stock holdings and adjusted and unadjusted kernel densities for the holdings of the three non-company stock investment vehicles. These densities are drawn from the sample of participants who invest in all four funds. As expected, the company stock density shows a multimodal distribution. The spikes at zero and one are no longer observed because the sample is restricted to those who invest in all four funds. Therefore, a zero or 100 percent allocation to company stock, or any investment vehicle for that matter, is not possible. Looking at the three other investment vehicle kernel densities, the difference is striking between the unadjusted (denoted by a line with triangles) and the adjusted (denoted by a line with squares) kernel densities. The unadjusted kernel densities for the non-company stock funds have several probable allocations. However, the adjusted distributions are strongly centered at 33%. This percentage equals the 1/n allocation an investor would choose if allocating between three funds. This implies that after adjusting for company stock holdings, these individuals tend to allocate their remaining assets evenly among the other funds. This finding strongly supports Benartzi and Thaler s (2001) assertion that some individuals treat company stock as a separate asset class and as a result slightly modify how they follow the 1/n rule. Panels B-D repeat the exercise for subsamples of participants holding company stock and two other funds. The results for Panel B and D are very similar to the earlier findings. As expected, the adjusted kernel densities are centered at 50 percent. 25

26 C. Econometric Analysis The final empirical question is what type of person is most likely to follow the 1/n heuristic? To test this a dummy variable that equals one if the individual follows the 1/n heuristic and zero if not is constructed. An additional dummy variable is created to take into account the impact of the company stock option on the 1/n rule. This additional dummy variable is constructed based on the modified 1/n heuristic. Tables VI displays the results of probit analysis using the two different 1/n dummy variables. The marginal effects of salary and employment tenure are significant and negative for both regressions. This suggests that high salary individuals and participants with longer job tenure are less likely to follow the 1/n rule. In both regressions, the average employee earning $100,000 is nearly three percent less likely to follow the 1/n rule than an average employee earning $40,000. One theory is that the higher salaried individuals are more educated and thus less likely to rely on simple rules for investing. It is not clear why the employees with longer job tenures are less likely to follow the 1/n rule. VII. Conclusion This paper focuses on two potentially inefficient and commonly made allocation choices in 401(k) plans: over-investing in company stock and following naïve diversification strategies. Given the pivotal role of 401(k) savings on an individual s retirement, a better understanding of who may make these uninformed allocation decisions is important. This research contributes to the literature by highlighting what types of individuals are most likely to make these inefficient choices. The results can help 26

27 plan sponsors target high risk individuals and improve plan design, as well as inform the current Social Security debate. Three important results emerge from the company stock analysis. First, employees tend to cluster their allocations at zero, 25, 50, 75 and 100 percent. Empirical researchers should bear this in mind before applying standard econometric techniques to allocation data. Second, in this plan the tendency to invest in more company stock than the recommended limit is high. Third, demographic and employment characteristics affect company stock allocations. Results of regressions that control for past company stock performance suggest that the probability of over-investing in company stock for the average participant is greater for males, decreases with salary and is lower for corporate division workers. The investigation of whether employees follow the 1/n heuristic confirms that investors tend to treat company stock as a separate account. While the percent of the individuals who follow the 1/n heuristic is lower than that found in previous studies, it still represents 5% of the participants in this study. Regression results suggest that highly compensated individuals are less likely to invest follow the 1/n heuristic, as well as employees with relatively long job tenures. Interestingly, the results show that individuals earning relatively higher salaries tend to make more efficient decisions related to both their company stock allocations and diversification strategies than others. Policy makers and plan sponsors should bear this in mind. After all, even though the Social Security system provides higher replacement ratios for individuals with relatively lower incomes, lower income participants will most 27

28 likely depend the most on their 401(k) savings during retirement. This research shows that this group tends to make the most inefficient choices. 28

29 References Agnew, Julie, Pierluigi Balduzzi and Annika Sunden, 2001, Portfolio Choice, Trading, and Returns in a Large 401(k) Plan, Center for Retirement Research, Working paper , Boston College. Andrews, Emily S., 1992, The Growth and Distribution of 401(k) Plans. in John Turner and Daniel Beller, eds.: Trends in Pensions 1992, Washington, D.C.: U.S. Department of Labor. Barber, Brad M., and Terrance Odean, 2001, Boys will be Boys: Gender, Overconfidence, and Common Stock Investment, Quarterly Journal of Economics 116 (1), Bassett, William F., Michael J. Fleming, and Anthony P. Rodrigues, 1998, How Workers Use 401(k) Plans: The Participation, Contribution, and Withdrawal Decisions. Federal Reserve Bank of New York Staff Report, No. 38 (March). Benartzi, Shlomo, 2001, Excessive Extrapolation and the Allocation of 401(k) Accounts to Company Stock? Journal of Finance 56, Benartzi, Shlomo, and Richard Thaler, 2001, Naive Diversification Strategies in Retirement Saving Plans, American Economic Review 91(1), Bernheim, B. Douglas and Garrett, Daniel M., 1996, The Determinants and Consequences of Financial Education in the Workplace: Evidence from a Survey of Households, National Bureau of Economic Research, Working paper Bettis, J. Carr, John Bizjak and Michael Lemmon, 2001, Managerial Ownership, Incentive Contracting, and the Use of Zero-Cost Collars and Equity Swaps by Corporate Insiders, Journal of Financial and Quantitative Analysis 36 (3), Bodie, Zvi, Robert C. Merton, and William F. Samuelson, 1992, Labor Supply Flexibility and Portfolio Choice in a Life Cycle Model, Journal of Economic Dynamics and Control 16, Clark, Robert L. and Sylvester Schieber, 1998, Factors Affecting Participation Levels in 401(k) Plans, in Olivia S. Mitchell and Sylvester J. Schieber, eds.: Living with Defined Contribution Plans, (University of Pennsylvania Press, Philadelphia, PA) 29