Pension plan choice among university faculty.

1. Introduction

There has been a large amount of economic research concerning the effects of pension plans on both a worker's mobility and retirement timing. (1) Defined benefit plans tend to be less portable but limit the financial market risk faced by the employee. Defined contribution plans are more portable but expose the employee to a greater financial risk from a market downturn. Because of these distinct characteristics between the two types of pension plans, we can expect that workers would not be indifferent between them. Instead, workers will prefer a particular type of pension plan depending on their tastes and preferences as well as their aversion to various types of risk. Most private-sector employers do not offer their employees a choice of pension plan type for their primary retirement plan. (2) It is generally assumed that workers make the choice of an employer and a pension plan at the same time. As a result, there is little empirical research on the preferences of workers concerning the type of pension they receive. (3)

Unlike private-sector employees, many university faculty members do receive a choice of pension plan after they are hired. The results of a recent survey of university administrators by the American Association of University Professors (AAUP) are shown in Table 1. Almost 36% of the institutions surveyed offered a choice to their faculty between a defined benefit and a defined contribution plan. Over half of the public colleges and universities surveyed offered a choice of pension plan types, whereas fewer than 1% of the private schools did. Among four-year institutions, approximately 40% of research and doctoral universities offered a choice compared to only 14% of the baccalaureate institutions. For two-year institutions, 69% of those with faculty ranks gave their faculty a choice, but only 30% of the two-year schools without ranks did so.

Because of the differing effects of these two pension plans on mobility and retirement timing, understanding the pension choices of faculty members can help university administrators and other employers plan for possible quits and retirements, allowing them to better manage faculty flows. (4) Participants in defined benefit plans have lower turnover rates and higher age-specific retirement rates that include large spikes in the probability of retiring at the early and normal retirement ages specified in the plan. Faculty who select to enroll in either the defined benefit or the defined contribution plan are locked into the plan for their entire career. Senior administrators who know the distribution of faculty in the two plan types will be able to better predict retirement patterns and other turnover rates. Employers that engage in labor force planning benefit from better information on age-specific rates of termination and retirement. In particular circumstances, presidents and provosts may seek to change available plan options or to alter the incentives embedded in the existing retirement plans to better achieve the desired composition and size of their faculty (Clark and Ma 2005). In response to delayed retirement by senior faculty covered by defined contributions, some research institutions have attempted to modify their retirement policies to provide greater retirement incentives. Recognizing patterns of plan choice can help employers develop and refine their overall retirement policies.

This study examines the pension choices of faculty in the University of North Carolina system (UNC) using employment records from 1983 to 2001. Estimates indicate that faculty choices conform to predictions based on economic models. The findings from this analysis are examined for their implications on the continued evolution of employer pension plans and the current Social Security debate.

2. Pension Plan Choice in the University of North Carolina System

Since 1971, all newly hired faculty members in the University of North Carolina system have had a choice between enrolling in the Teachers' and State Employees' Retirement System (the state plan) or in one of several Optional Retirement Plans (ORPs) as their primary retirement plan. The state retirement plan is a final-pay, defined benefit plan. Faculty members who are enrolled in the state plan are required to contribute 6.0% of earnings to this pension plan. The benefit formula for the state plan is 1.82% of final average earnings multiplied by years of service, where average earnings are based on the employee's four highest consecutive years of earnings. Participants are fully vested after 5 years of service in the state plan.

The state pension plan provides significant monetary incentives to retire for faculty members who meet the age and service requirements for early and normal retirement. The normal retirement age is 65 with 5 years of creditable service; however, the plan also provides unreduced benefits with 30 years of service regardless of age or at age 60 with 25 years of creditable service. Early retirement with reduced benefits is available at age 50 with 20 years of creditable service or at age 60 with 5 years of service.

To see how the defined benefit plan includes monetary incentives to retire once eligibility is achieved, consider the case of a male faculty member hired at age 25, whose career earnings follow the typical path of the faculty in our sample (as estimated by an age--earnings profile). During his 30th year of employment, his pension wealth rises by 7.5% because he becomes eligible for full retirement benefits. By contrast, his pension wealth rises only 2.1% as he moves through his 31st year of employment. Thus, the defined benefit plan provides significant incentive to remain with the university prior to eligibility, but this incentive declines substantially once eligibility is achieved.

The ORPs are defined contribution plans. Currently, faculty choosing to enroll in an ORP can select one of four providers: TIAA-CREF, Lincoln National, Fidelity, and VALIC. Faculty members are required to contribute 6 percent of salary, and the state contributes 6.84% of salary to the individual's retirement account. The level of benefits in the ORP at retirement depends on lifetime contributions and accrued rates of return. Every faculty member must participate in either the state plan or one of the ORPs. If a faculty member does not choose between the plan types, he or she is automatically enrolled in the state plan; however, the number of faculty placed in the state plan by default appears to be very small. (5) Employee contributions are immediately vested in both plans. In addition to their employer-provided pension plan, all UNC faculty are covered by Social Security.

The data used in this study are employment records for the 15 campuses in the UNC system awarding tenure. (6) Each campus is required to report its annual faculty census to the General Administration in September of every year. The census reports general characteristics and employment information about each faculty member. These include gender, race, age, salary, rank, and tenure status as well as which type of pension plan the individual member selected. Unfortunately, no data are available on the size of the ORP accounts and the allocation of these accounts between stocks and bonds. These data files do not include any information on other household wealth, participation in supplemental retirement plans, marital status, or employment of spouse. The data are available from 1982 through the academic year 2001. During this time, 7035 new tenure-track faculty were hired across the 15 campuses. The hiring ages range between 22 years and 75 years with a median age at employment of 37 years.

3. An Economic Model of Pension Choice

Within 30 days after employment, all newly hired tenure-track faculty at any of the 15 tenuregranting campuses in the UNC system must choose to enroll in either the defined benefit state plan or one of the defined contribution ORPs. This is a one-time election by a new faculty member. Faculty members selecting one of the ORPs must remain in an ORP for the remainder of their careers at UNC, but they are free to move among the ORPs offered by their institution. How does a new faculty member decide whether to enroll in the state plan or participate in one of the ORPs?

There are several studies that have modeled the choice between a defined benefit and a defined contribution pension. Bodie, Marcus, and Merton (1988) compare the benefits of these two types of plans. They show that defined benefit plans provide a stable replacement rate of final income to workers and thus offer some degree of insurance against real wage risk. Their model also indicates that the advantages of defined contribution plans are greatest during periods of high inflation uncertainty and include the predictability of pension wealth, the ability to invest in inflation-hedged portfolios, and the fully funded promise that the defined contribution plan provides. McCarthy (2003) presents a three-period model that shows that defined benefit plans are more valuable to older employees and are more attractive when equity market returns are lower than expected or are highly variable. In addition, defined benefit plans provide more security when annuity markets are inefficient.

Because workers can change pensions by changing employers, Childs et al. (2002) use a real options framework to examine the tradeoffs between a defined benefit and a defined contribution pension. They find that defined benefit plans are generally preferred by those who plan to remain with their current employers, older employees, and those for whom the annual defined benefit accrual rate is high. Defined contribution plans are preferred by those for whom the option to switch employers is more valuable, younger workers, and those who work for employers with relatively high defined contribution rates. Surprisingly, they also find that more risk-averse employees should prefer a defined contribution plan to a defined benefit plan.

The basic characteristics of defined benefit and defined contribution pension plans are shown in Table 2. Because of the differences in these plan provisions, the expected value of retirement benefits will vary across individuals depending on their personal characteristics, their risk preferences, and their career patterns. In a deterministic world without risks, a new employee could calculate the value of participating in the defined benefit plan by selecting the length of the working career and the age to begin receiving retirement benefits. Using this information plus projected earnings over the working career and the state plan benefit formula, the individual could determine a future annual retirement benefit and its present value. Similarly, the value of participating in the defined contribution plan could be calculated by estimating future contributions as a function of projected earnings and annual rates of return on pension investments. These calculations would produce an account balance at various ages. The expected account balance produced by the defined contribution plan could be compared to the present value of retirement benefits under the defined benefit plan. The individual would then select the pension that yielded the greatest value for the expected career pattern while an employee of the UNC system.

To illustrate the relative generosity of the two plans in a risk-free world, we estimated an age-earnings profile by gender. Real lifetime earnings are estimated using employment records from 1982 until 2001. Using plan characteristics, we calculated the present value of pension benefits at age 65 conditional on the person remaining at his or her institution for each year between age 25 and age 65. All present value calculations are based on life tables for the 1940 birth cohort developed by the Social Security administration (Bell and Miller 2002). (7) For participants in an ORP, we assumed that they allocated 50 percent of their pension funds to stocks and 50 percent to bonds. These investments are assumed to have certain real rates of return of 3 percent for bonds and 6 percent for equities. (8) For men, pension wealth under the ORP was always greater than the pension wealth from the state plan. If a male faculty member hired at age 35 remained with the university until age 65, ORP pension wealth would be $362,781, whereas pension wealth in the state plan would be $351,425.

In this risk-free world, several predictions can be made concerning how faculty would choose which plan to enroll in. First, the value of participation in the defined benefit plan as a percentage of compensation increases with age (Kotlikoff and Wise 1989a), whereas annual contributions to defined contribution plans are not typically a function of age. Therefore, faculty members hired at older ages who plan to complete their careers in the UNC system should be more likely to select the defined benefit plan. For example, male faculty members hired at age 45 who remain until age 65 will accumulate 20% more pension wealth under the state plan than if they had chosen an ORP. In fact, male faculty members hired at age 45 who remain with the institution until age 60 or older will do better by selecting the defined benefit plan. The advantage is even greater if the newly hired faculty member is even older.

Second, because defined benefit plans are required by law to provide the same annual benefits to men and women, the value of these plans will be greater for women because, other things being equal, they have a greater life expectancy. For example, pension wealth for women hired at age 35, conditional on remaining with the institution until age 65, will be higher in the state plan than in an ORP, $306,590 in the state plan compared to $287,832 in the ORP. Female faculty hired at age 45 who remain at the university until age 65 will accumulate 34 percent greater pension wealth in the state plan. All women hired at age 45 who remain at least 10 years with the university will have greater pension benefits if they select the state plan. These differences in relative pension wealth from participating in the two plans between men and women arise from the greater life expectancy of women, whereas the greater dollar values of the pension wealth of men are the result of their higher lifetime earnings. Thus, for this reason, women should be significantly more likely to enroll in the defined benefit state plan.

However, the future is not known, and various types of risk have different effects on expected retirement benefits under defined benefit and defined contribution plans. Sources of risk that affect future retirement benefits include those associated with changing jobs, financial market fluctuations, uncertainty surrounding the retirement date, variation in the growth rate of real earnings, and inflation before and after retirement. Participants in defined benefit plans who change jobs suffer significant losses in the expected present value of retirement benefits. Thus, the risk associated with job mobility is borne by the employee in these plans (Clark and McDermed 1988). Employees in defined contribution plans, however, can usually change employers without the loss of any retirement benefits after vesting requirements are met. Retention in institutions of higher education is centered on the renewal of initial appointments (usually after three or four years) and the granting of permanent tenure and promotion after six or seven years. Thus, tenure standards and the probability of being retained by the institution will play a major role in the selection of a pension plan among those faculty hired without tenure.

Each newly hired faculty member must assess the probability of remaining with the university until retirement age versus the likelihood that he or she will change employers at certain ages during the career. Newly hired professors who expect to remain in the UNC system until retirement should have a greater tendency to select the state pension plan. Faculty members who do not expect to complete their careers in the UNC system will be more likely to enroll in one of the defined contribution plans. In the academic labor market, mobility expectations will depend on personal preferences, potential opportunities, and anticipated performance rewards in terms of tenure, promotions, and salary growth. Turnover risks may also be associated with gender, race, and type of institution. In addition, participants in defined contribution plans in the nonacademic sector are typically not offered an annuitization option in the plan. Thus, retirees who want the certainty of a lifetime retirement benefit must purchase an annuity in the market, often at disadvantageous rates.

Several other types of risks also confront employees in making their choice of a pension. Defined benefit plans promise a lifetime annuity at a specified level. Thus, these plans protect the worker against unanticipated increases in life expectancy. In contrast, participants in defined contribution plans will receive a lower annual payout from any given level of pension wealth if life expectancy has increased.

As mentioned above, a unique factor of the academic labor market is its promotion and tenure process. In the UNC, as at most universities, new assistant professors are hired on an initial four-year contract. They are reviewed in the third year of this contract. If teaching and research performance are adequate, the individual is renewed for a second contract of three years; however, persons whose productivity is deemed to be inadequate are terminated. Faculty who receive a second contract are then reviewed in their sixth year. Those who meet institutional standards are promoted and awarded permanent tenure, whereas those whose productivity is below acceptable standards are terminated. This evaluation and career path results creates spikes in involuntary turnover for faculty after three and six years of service. Newly hired faculty must make an assessment concerning the likelihood that they will be high performers and remain at the university or that their future productivity may be insufficient to merit promotion and tenure. Although this formal tenure process is unique to universities, many organizations have probationary periods and some limited protection of long service workers.

Table 3 reports the ratio of pension wealth under the two pension plans for men and women, and Table 4 presents the rates of return to assets in the defined contribution plan necessary to provide the same present value of pension wealth generated by the state defined benefit plan. The ratios are reported separately for individuals hired at ages 25, 35, 45, and 55, and indicate the relative values of the two plans for individuals staying with the university for alternative numbers of years. These values clearly show the importance of assessing turnover probabilities (both voluntary moves and those that result from being denied tenure) in the choice of a pension plan.

The entries in Table 3 show that, for faculty who are hired at young ages and leave the university before reaching the early or normal retirement age, participation in the defined benefit plan results in much lower pension wealth. For example, the present value of the state plan for a man hired at age 35 who leaves the university after 10 years of service is only 57% of the value of the defined contribution plan with a mixed asset allocation (or 89% of the defined contribution value if all assets are invested in bonds). For men hired at age 35 who stay at the university until retirement at age 65, the state plan provides benefits that are almost as good as those provided by the defined contribution plan. The table clearly shows that women have greater benefits under the state plan, as do workers who are hired at older ages.

Table 4 presents the internal rates of return for investments in the defined contribution plan necessary to achieve parity with the benefits provided in the defined benefit plan. The internal rates of return for the defined contribution plan must be higher for those employed at older ages, for women, and for those that expect to remain with the university for longer periods of time. The larger rates of return to defined contribution assets necessary to equate pension wealth in the ORP to that in the state plan illustrate why faculty with certain characteristics are more likely to enroll in defined benefit plans.

Most defined benefit plans calculate benefits based on final average earnings over the last three to five years of employment. Therefore, the rate of salary growth will have a significant effect on the retirement benefit from a defined benefit plan, whereas the influence of salary growth should not be as important for defined contribution plans. As a result, with other things constant, employees whose salaries are expected to rise rapidly throughout their careers will be inclined to select a defined benefit plan. In recent years, university salaries have been rising relatively slowly, thus reducing the general tendency for new faculty to select the state plan; however, variation across individuals remains an important consideration in the choice of a pension plan.

Other risks facing pension participants include variations in rates of return on pension investments, the potential decline in real after-retirement benefits with inflation, and changes in nominal earnings before retirement in response to inflation. Participants in defined benefit plans do not bear investment risk, whereas those in defined contribution plans do. Thus, employees with a high degree of risk aversion toward financial investments will tend to favor defined benefit plans. Real retirement benefits decline with inflation unless nominal retirement benefits rise. Many government pension plans (including the state plan in the UNC system) provide for regular increases in retirement benefits to offset declines in purchasing power caused by inflation. (9) In defined contribution plans, the extent of this risk depends on how the funds in the retirement account are invested during the retirement period.

The turnover and retirement behavior of faculty are affected by their choice of a pension plan. In other papers, we have estimated age-specific retirement rates for faculty in the state plan compared to those in the defined contribution plans (Ghent, Allen, and Clark 2001; Clark and Ghent 2004). Our estimates indicated that faculty who select defined benefit plans have higher retirement probabilities than those who enroll in a defined contribution plan; thus, the expected ages of retirement differ depending on plan choice. Although the comparisons in Tables 3 and 4 are for faculty who leave at the same ages, we should recognize that years of service are, in fact, different between participants in the two types of pension plans.

One method of illustrating the effect of financial market risk on the choice of a pension plan is to assume that the entire equity premium is merely to compensate for the added risk. Similarly, we could assume that ORP participants who are averse to financial market risk would select to place all of their pension funds into bond accounts. The impact of this change in investment is shown in the bottom half of Table 3. The lower real risk-adjusted return on pension investments sharply reduces the pension wealth of both men and women of all ages who are enrolled in an ORP while leaving pension wealth in the state plan unaffected. If risk is accounted for in this manner, men hired at age 35 can now expect greater pension wealth in the state plan provided that they expect to remain at the institution for just over 15 years; women need to expect to remain only 10 years before they accumulate greater wealth in the state plan. Financial market risk preferences may also differ by gender and race. (10)

Following Clark and Pitts (1999), we assume that the choice between the two types of plans depends on the relative risk-free value of participation in the defined benefit plan compared to the defined contribution plan, the existence of certain types of risk, and the individual's aversion to risk. Let the expected value of enrolling in a defined benefit plan be

E[V(db)] = f[CV(db), r(m), r(f), r(w)],

where CV(db) is the certainty value of enrolling in the defined benefit plan based on the plan formula, age at employment, desired retirement age, and assumed rates of wage growth and inflation. Plan choice is also influenced by mobility risk, r(m); the potential for a decline in the real value of benefits as a result of inflation, r(f); and variation in the rate of wage growth, r(w).

Similarly, let the expected value of enrolling in the defined contribution plan be

E[V(dc)] = f[CV(dc), r(I), r(f), r(w)],

where CV(dc) is the certainty value of participating in the defined contribution plan with known contribution rates and expected rates of return. Other factors influencing the value of the defined contribution plan include the rate of return risk on pension investments, r(I); the responsiveness of the retirement portfolio to fluctuations in inflation, r(f); and the rate of real wage growth, r(w).

The pension choices observed in the data are the outcomes of each individual's comparison of the expected value of participation in the state defined benefit plan, E[V(db)], with that of a defined contribution plan, E[V(dc)]. The result depends on the certainty value of the two plans, the levels of job mobility risk and investment risk each faces, and each one's aversion to each type of risk. We define this expected value comparison to be

[y.sup.*.sub.i] = E[V(db)] - E(V(dc)] = [beta]'[x.sub.i] + [[epsilon].sub.i],

where [x.sub.i] = a vector of individual and university characteristics, [beta] = the influence of x on the expected value comparison, and [[epsilon].sub.i] = unobserved random factors.

Variables included in [x.sub.i] are gender; ethnic origin; year hired; age, rank, and salary at hire; and type of institution. The individual chooses to enroll in the defined benefit plan if E[V(db)] > E[V(dc)] or [y.sup.*.sub.i] > 0. The comparison value, [y.sup.*.sub.i], is not directly observed in the data. We observe only the outcome of the decision,</p>

<pre> [y.sub.i] = {0 if the individual chooses to enroll in a

defined contribution plan, {1 if the individual chooses to enroll in the state defined benefit plan. </pre> <p>We assume that the unobserved random factors are normally distributed, N(0, [sigma]), and use a probit model to estimate the parameters of the pension choice model.

4. Pension Choices for New Hires, 1983-2001

Enrollment rates for the state defined benefit plan shown in Figures 1 through 4 illustrate how pension choices have varied across time as well as by individual and institutional characteristics. The percentage of new hires enrolling in the state defined benefit plan declined steadily during the sample period. In 1983, 23% of newly hired faculty chose to enroll in the state plan. This number fell by over 11 percentage points during the next 18 years. The percentage of new faculty choosing the state plan differed greatly by gender, race, age at hire, and type of institution. As predicted, a larger percentage of women chose the state retirement plan than did men.

[FIGURE 1 OMITTED]

Figure 2 shows the percentage of new hires who chose the state defined benefit plan by race. In 1983, 18.2% of white new faculty members selected the defined benefit plan compared to 60.6% of new black faculty. By 2001, these rates had fallen to 9.3% for new white faculty and 32.8% for new black faculty. Figure 3 shows that as age at hire increased, the percentage of new hires choosing the state plan rose as predicted. Figure 4 indicates that, in general, new faculty in masters-granting and baccalaureate institutions had a higher enrollment rate in the state pension plan than new hires at research and doctoral-granting institutions.

[FIGURES 2-4 OMITTED]

Summary statistics are presented by pension type in Table 5. They further highlight the differences between new faculty choosing the defined benefit plan and those choosing one of the defined contribution plans. Masters-granting institutions account for over 36% of the new hires, but they account for over 57% of faculty selecting the defined benefit plan. Women make up 35% of new hires and over 41% of those enrolling in the state pension plan. Fewer than 12% of new hires are black, but they represent almost one third of persons choosing the defined benefit plan. New hires in the defined benefit plan earn less than the sample average, $33,552 (in 1983 dollars), and new hires in the defined contribution plan earn more, $38,995 (in 1983 dollars).

The estimates for the pension choice model for new hires from 1983 to 2001 are shown in Table 6. They describe how the individual and institutional characteristics interact to determine which pension plan the faculty member chooses upon hire. Significant individual characteristics are gender, ethnicity, and age. Newly hired female and nonwhite faculty members are significantly more like to enroll in the state defined benefit plan than are their male and white counterparts. (11) The estimates indicate that women are more than 2.2 percentage points more likely to choose to enroll in the state plan compared to men with similar characteristics. The greater likelihood of selecting the state plan may be related to greater life expectancy of women, greater financial risk aversion, or different turnover possibilities. Given the data limitations of the employment records, we are unable to disentangle these potential determinants of the gender effect on plan choice.

Other things being constant, enrollment rates in the state plan among newly hired black faculty members are estimated to be almost 22 percentage points higher than those for newly hired white faculty. New hires of other races are estimated to be 5.4 percentage points more likely to choose the defined benefit plan. These differences across race may reflect differences in the career patterns of black faculty members. Almost two-thirds of the newly hired black faculty members in our sample are employed by historically black universities in the UNC system.

To examine this result further, we reestimated the pension choice decision for the subsample of newly-hired black faculty members employed by the UNC. Those faculty members employed at the five historically black universities (HBUs) are estimated to be 45 percentage points more likely to select the state plan than their counterparts at other campuses. The higher incidence of enrollment in the state plan could be caused by peer group effects associated with employment at one of the HBUs.

Alternatively, these differences might reflect the labor market alternatives facing HBU faculty members compared to black faculty at the other UNC campuses. (12) Lacking data on professional reputation, publications, and other variables that impact employment opportunities in academia, we are unable to further examine these differences in plan choice among black faculty.

Defined benefit enrollments for new faculty rise with the age of the faculty member. Enrollment rates for new faculty who are 36 years to 40 years of age are estimated to be more than 3.5 percentage points higher than for those aged 31 years to 35 years, and the enrollment rates for new hires who are aged 41 to 45 are estimated to be more than 7 points higher. Enrollments in the defined benefit plan for new faculty members over the age of 45 are estimated to be 15 percentage points higher. These age effects reflect the greater economic value of participation in the state defined benefit plan to older faculty and are consistent with the economic model presented earlier.

Enrollments also vary significantly across university classifications. Newly hired faculty members at masters-granting and baccalaureate institutions are more likely to choose the state defined benefit plan than are comparable faculty at research institutions. Enrollments in the state plan are estimated to be almost 24 percentage points higher in baccalaureate schools and 12.5 percentage points higher for masters-granting institutions. However, enrollments at doctoral institutions are estimated to be 3 percentage points lower than in research universities. Faculty hired at the doctoral and research institutions typically face different employment and labor market options. Research output is more important in their ability to earn tenure, and publications provide a clear market signal of quality to other institutions. Thus, these faculty likely have significantly different expectations concerning the probability of remaining at a specific institution for their entire career. As a result, they will be more concerned about the lack of portability of benefits in the state plan and be more likely to enroll in an ORP.

Across the United States, there has been a noticeable trend toward increased coverage by defined contribution plans and a decline in the prevalence of defined benefit plans. Studies examining this trend (including Clark and McDermed 1990; Gustman and Steinmeier 1992; Ippolito 1995) attribute it to the introduction of 401(k) plans, the rising cost of administering defined benefit pension plans, and changes in the industrial distribution of the U.S. economy. Although these studies examine the changing preferences of employers away from defined benefit plans and toward defined contribution plans, they do not discuss how employee preferences for each type of pension may have changed over time as well. In studying the pension choice decisions of new faculty hired by North Carolina State University, Clark and Pitts (1999) found that faculty hired more recently are significantly less likely to choose the state defined benefit pension plan than those hired earlier. They concluded that this was a result of changing faculty preferences for defined contribution plans relative to the defined benefit plan because of mobility considerations.

Table 7 reports the predicted probability of a new hire choosing the state defined benefit pension for each year of the data. Table 7 shows that, after accounting for individual and institutional characteristics, enrollments in the state defined benefit plan were lower than 1983 enrollments for all years examined in this study. From 1989 through 2001, new faculty were significantly less likely to choose the defined benefit plan than were those hired in 1983. Beginning in the early 1990s, predicted enrollments declined by 7 percentage points over those in 1983, and in the late 1990s, enrollments declined another 3 percentage points to 10 percentage points below the 1983 enrollments.

The declining probability of enrolling in the state defined benefit plan has occurred despite increases in life expectancy, which have made participation in the state plan more valuable. There has been no increase in mandatory employee contributions during this time, so the value of being in the state plan relative to one of the ORPs has increased. Despite this change, more and more newly hired faculty are opting to participate in the defined contribution plans. This would suggest that other characteristics of the ORPs have become more valuable to faculty; the most likely candidate is the portability of these plans. (13)

5. Conclusion

Pensions are an important component of lifetime earnings, and thus the decision between a defined benefit pension and a defined contribution pension is an important one. On the one hand, a defined benefit plan limits the amount of market risk faced by the employee and provides a predictable level of benefits. On the other, defined contribution pension plans are more portable and thus desirable to workers who want to remain mobile.

Economic theory suggests that different types of workers will prefer either a defined benefit pension or a defined contribution pension. Older workers will likely find the defined benefit more attractive because of the relative increase in the certainty value of the defined benefit plan compared to the certainty value of the defined contribution plan. Also, persons hired at older ages might have a lower relative desire for mobility than younger hires. Women may be more likely than men to prefer defined benefit plans if pension wealth is increased as a result of higher life expectancy; however, the relative importance of financial market risk and mobility risk may also be gender specific. The results of this study indicate the pension plan choices of newly hired faculty in the UNC system are consistent with economic theory. New employees who are older or female are significantly more likely to enroll in the state defined benefit plan. In addition, there has been a declining trend in participation in the state defined benefit plan.

A majority of public institutions offer their faculty the choice of enrolling in a defined benefit pension plan or a defined contribution plan. Academic leaders, along with legislators, have determined that many faculty want a choice of pension plans, and when given the choice, faculty overwhelming select the defined contribution plans. Virtually all private universities offer only defined contribution plans. Leaders of these universities have assessed the value of providing the two types of plans and have chosen defined contribution plans. The most likely explanation for this decision is the preference of faculty for the greater portability of defined contribution plans. Employers in the nonacademic sector also must decide whether to offer a pension plan and, if so, what type of plan to provide. Over time, more and more firms have decided to offer defined contribution plans. Employers seek to provide their workers the greatest value for each dollar spent on labor compensation. Understanding employee preferences is a key to this decision. The results of this paper provide evidence on how employees value pension plans. This knowledge will be useful to employers as they seek to develop their retirement policies, appeal to specific types of workers, and to influence future turnover and retirement rates.

It is tempting to speculate on the implications of these findings for the proposal for individual accounts in Social Security. At first blush, one might conclude that these results indicate that workers are increasingly demanding that their retirement plans have the basic characteristics of defined contribution pension plans (see Table 2). However, one should remember that Social Security is a national plan covering almost all jobs, so these benefits are portable across employers, and there is no loss in Social Security wealth from changing jobs. Because portability is one of the main reasons workers prefer defined contribution plans, we should not attempt to directly transfer our findings from employer pensions to preferences for individual accounts in Social Security.

However, some of our findings may be helpful in considering what types of workers would select voluntary individual accounts within Social Security. Concern has been expressed by policy makers that some types of workers may select individual accounts while others remain in the traditional plan. Reviewing the results presented in Table 6 provides some support for the conclusion that women and blacks might be less likely to choose voluntary individual Social Security accounts. One should remember that the choice of an employer plan in our data is made by faculty who are already covered by the current defined benefit Social Security system. It would he interesting to repeat this analysis for public employees who are not covered by Social Security.

Received September 2004; accepted June 2005.

References

Allen, Steven, Robert Clark, and Ann McDermed. 1993. Pension bonding and lifetime jobs. Journal of Human Resources 28:463-81.

Ameriks, John. 2002. Recent trends in the selection of retirement income streams among TIAA-CREF participants. Research Dialogue 74, December.

Bell, Felicitie C., and Michael L. Miller. 2002. Life Tables for the United States Social Security Area 1900-210. Actuarial Study No. 116, August.

Bodie, Zvi, Alan J. Marcus, and Robert C. Merton. 1988. Defined benefit versus defined contribution pension plans: What are the real trade-offs? In Pensions in the U.S. economy, edited by Zvi Bodie, John Shoven, and David Wise. Chicago, IL: University of Chicago Press, pp. 139-62.

Brown, Jeffrey R. 2003. Redistribution and insurance: Mandatory annuitization with mortality heterogeneity. The Journal of Risk and Insurance 70:17-41.

Burkhauser, Richard. 1979. The pension acceptance decision of older workers. Journal of Human Resources 14:63-75.

Childs, Paul D., Douglas Fore, Steven H. Ott, and Claude C. Lilly. 2002. Defined benefit vs. defined contribution? Determining the optimal benefit plan choice using a real options framework. TIAA-CREF Institute Working Paper Series No. 9-060102.

Clark, Robert L., and Linda S. Ghent. 2004. Did the elimination of mandatory retirement affect faculty retirement? Comment. Unpublished manuscript, North Carolina State University.

Clark, Robert L., Linda S. Ghent, and Juanita Kreps. 2000. Faculty retirement and the impact of the elimination of mandatory retirement at three North Carolina universities. In To retire or not? Retirement policy and practice in higher education, edited by Robert L. Clark and P. Brett Hammond. Philadelphia: University of Pennsylvania Press, pp. 21-38.

Clark, Robert L, John J. Haley, and Sylvester J. Schieber. 2001. Adopting hybrid pension plans: financial and communication issues. Benefits Quarterly First Quarter:7-17.

Clark, Robert L. and Jennifer Ma. (Eds.). 2005. Recruitment, retention, and retirement in higher education: Building and managing the faculty of the future. Northampton, MA: Edward Elgar Publishing.

Clark, Robert, and Ann McDermed. 1988. Pension wealth and job changes: The effects of vesting, portability, and lump-sum distributions. The Gerontologist 28:524-32.

Clark, Robert, and Ann McDermed. 1990. The choice of pension plans in a changing regulatory environment. Washington, DC: American Enterprise Institute.

Clark, Robert L., and M. Melinda Pitts. 1999. Faculty choice of a pension plan: Defined benefit vs. defined contribution. Industrial Relations 38:18-45.

Committee on Retirement, American Association of University Professors. 2000. Survey of changes in faculty retirement policies. Washington, DC: American Association of University Professors.

Ghent, Linda S., Steven G. Allen, and Robert L. Clark. 2001. The impact of a new phased retirement option on faculty retirement decisions. Research on Aging 23:671-93.

Gustman, Alan, Olivia Mitchell, and Thomas Steinmeier. 1994. The role of pensions in the labor market: A survey of the literature. Industrial and Labor Relations Review 47:417-38.

Gustman, Alan, and Thomas Steinmeier. 1992. The stampede toward defined contribution pension plans: Fact or fiction? Industrial Relations 31:361-9.

Halek, Martin, and Joseph G. Eisenhauer. 2001. Demography of risk aversion. Journal of Risk and Insurance 68:1-24.

Ippolito, Richard. 1987. Why federal workers don't quit. Journal of Human Resources 22:281-99.

Ippolito, Richard. 1995. Towards explaining the growth of defined contribution plans. Industrial Relations 34:1-20.

Kotlikoff, Laurence, and David Wise. 1989a. The wage carrot and the pension stick. Kalamazoo, MI: Upjohn Institute for Employment Research.

Kotlikoff, Laurence, and David Wise. 1989b. Employee retirement and a firm's pension plan. In The economics of aging, edited by David A. Wise. Chicago: University of Chicago Press, pp. 279-330.

McCarthy, David. 2003. A life-cycle analysis of defined benefit pension plans. Journal of Pension Economics and Finance 2:99-126.

Mitchell, Olivia. 1982. Fringe benefits and labor mobility. Journal of Human Resources 17:286-98.

Papke, Leslie E. 1998. How are participants investing their accounts in participant-directed individual account pension plans? American Economic Review 88:212-6.

Papke, Leslie E. 2004. Pension plan choice in the public sector: The case of Michigan state employees. National Tax Journal 57:329-39.

Samwick, Andrew. 1998. New evidence on pensions, Social Security, and the timing of retirement. Journal of Public Economics 70:207-36.

Schubert, Renate, Martin Brown, Matthias Gysler, and Hans Wolfgang Brachinger. 1999. Financial decision-making: Are women really more risk averse? American Economic Review 89:381-5.

Sunden, Annika E., and Brian J. Surette. 1998. Gender differences in the allocation of assets in retirement savings plans. American Economic Review 88:207-11.

Yang, Tongxuan. 2005. Understanding the defined benefit versus defined contribution choice. Pension Research Council Working Paper No. PRC WP-2005-4.

(1) Allen, Clark, and McDermed (1993), Gustman, Mitchell, and Steinmeier (1994), Ippolito (1987), and Mitchell (1982) examine the effects of pension coverage on worker mobility. See Burkhauser (1979), Kotlikoff and Wise (1989b), and Samwick (1998) for discussion of the effects of pensions on retirement timing.

(2) In addition to their primary plan, many workers are also covered by a supplemental retirement plan (e.g., 401[k] or 403[b] plans) that allows them to make voluntary tax-deferred contributions.

(3) Papke (2004) examines the choices of Michigan corrections workers who were given the opportunity to switch from a defined benefit plan to a defined contribution plan. Yang (2005) analyzes a similar choice for workers in a nonprofit firm.

(4) For example, a study by Clark, Ghent, and Kreps (2000) estimates age-specific retirement probabilities for faculty at Duke University, University of North Carolina--Chapel Hill, and North Carolina State University. In their analysis, participants in the state retirement defined benefit plan are found to have significantly higher retirement probabilities than comparable individuals in one of the defined contribution plans.

(5) We are unable to identify whether any faculty member has failed to choose a pension plan and thus has been defaulted into the stale defined benefit plan. Kitty McCollum, Associate Vice President for Human Resources and University Benefits Officer for the UNC, stated that in her assessment the number of faculty defaulted into the state plan was very small, and the default option was rarely if ever used. Newly hired faculty are frequently engaged by human resource staff until they make their selection of a pension plan.

(6) The North Carolina School of the Arts is excluded because it does not award tenure.

(7) To control for mortality differences in the population as a result of differences in socioeconomic status, Brown (2003) uses mortality estimates that are differentiated by age, gender, educational attainment, race, and ethnicity. Given that our sample has a relatively small amount of variation in socioeconomic status and that this table is used only as a demonstration of the relative values of defined benefit and defined contribution plans, we have chosen to use unadjusted mortality estimates for the 1940 birth cohort.

(8) Some analysts argue that risk premiums for equities should not be considered in these types of calculations. For that reason, we also provide similar calculations based on all assets earning a 3% annual rate of return. Calculations using a lower real rate of return were also performed. As expected, a drop in the real rate of return on stocks and bonds increased the relative value of the DB plan for both men and women.

(9) For the state defined benefit plan, postretirement adjustments are made annually by the North Carolina legislature and are dependent on changes in the price level and the state budget status. In the calculations for Tables 3 and 4, it is assumed that these postretirement adjustments fully correct for the impact of inflation.

(10) Previous empirical studies indicate mixed results concerning differences in risk aversion across demographic groups. See, for example, Papke (1998); Schubert et al. (1999): Sunden and Surette (1998); and Halek and Eisenhauer (2001).

(11) If discrimination is present in the academic labor market, one might expect that nonwhite and female faculty members will have lower probabilities of receiving tenure and promotions. This would reduce the relative attractiveness of the DB plan for these individuals. Thus, our estimated effects of gender and race on the probability of choosing the DB plan may biased downward.

(12) In addition, if black faculty members are less likely to encounter discrimination at HBUs, the probability of receiving tenure and hence the relative value of the DB plan both increase.

(13) Interestingly, there seems to be a national trend among faculty in defined contribution plans to select lump sum distributions or partial annuities. Ameriks (2002) examines the decisions of TIAA-CREF participants and finds a striking trend away from the choice of life annuities. Thus, the movement away from defined benefit plans may also be reflecting a desire for greater access to retirement funds at retirement.

Robert L. Clark, * College of Management, North Carolina State University, Raleigh, NC 27695, USA; E-mail [e-mail].

Linda S. Ghent, Department of Economics, Eastern Illinois University, Charleston, IL 61920, USA; E-mail [e-mail]; corresponding author.

Ann A. McDermed, Department of Business Management, North Carolina State University, Raleigh, NC 27695, USA; E-mail [e-mail].

Table 1. Distribution of Retirement Plans Offered by Institutions:
AAUP Survey Data

                                  Pension Plan Type Offered

                                Defined           Defined
                              Benefit Only   Contribution Only

Total                             15.3              41.1
Public                            20.9              12.7
Private                            5.1              93.0

AAUP school classification:
  4-year institutions:
    Doctoral I                    11.4              35.0
    Masters                       12.0              41.8
    Baccalaureate                  6.5              75.2
  2-year institutions:
    w/ faculty ranks              15.5               8.5
    w/o faculty ranks             46.8              11.7

                                  Pension Plan Type Offered

                              Combination    Choice of Defined
                                  Plan       Benefit or Defined    N
                                                Contribution

Total                              7.6              35.9          608
Public                            10.9              55.2          393
Private                            1.4               0.5          215

AAUP school classification:
  4-year institutions:
    Doctoral I                    11.4              42.3          123
    Masters                        6.5              39.1          184
    Baccalaureate                  3.9              14.4          153
  2-year institutions:
    w/ faculty ranks               7.0              69.0           71
    w/o faculty ranks             11.7              29.9           77

Source: Committee on Retirement, American Association of University
Professors (2000).

Table 2. Features of Alternative Employer-Sponsored Plans

Plan Feature               Defined Benefit Plan       Defined
                                                      Contribution
                                                      Plan

Employer contributes       Virtually always           Sometimes
Employee contributes       Very rarely                Virtually always
Participation              Automatic                  Employee choice
Contribution level         Automatic                  Employee choice
PBGC Insurance             Yes, but capped            Not needed
Early departure penalty    Yes                        No
Benefits easily portable   No                         Yes
Annual communication       Benefit at end of career   Current balance
Retirement incentives      Occur at specific ages     Neutral
Accrual of benefits        Loaded to career end       Level over career
Financial market risks     Employer bears             Employee bears
Longevity insurance        Typically yes              Typically yes

Source: Clark, Haley, and Schieber (2001).

Table 3. Relative Value of Defined Benefit Plan to Defined
Contribution Plan at Age 65 (a)

                             Men

                         Age Hired At

                   25     35     45     55

Defined Contribution Plan with 50% Stocks and 50% Bonds

  Stay 5 years    0.30   0.50   0.81   1.28
  Stay 10 years   0.35   0.57   0.93   1.46
  Stay 15 years   0.41   0.66   1.06    --
  Stay 20 years   0.47   0.75   1.20    --
  Stay 25 years   0.53   0.86    --     --
  Stay 30 years   0.61   0.97    --     --
  Stay 35 years   0.69    --     --     --
  Stay 40 years   0.77    --     --     --

Defined Contribution Plan with 100% Bonds

  Stay 5 years    0.60   0.80   1.08   1.45
  Stay 10 years   0.66   0.89   1.19   1.59
  Stay 15 years   0.74   0.98   1.31    --
  Stay 20 years   0.81   1.08   1.44    --
  Stay 25 years   0.89   1.18    --     --
  Stay 30 years   0.98   1.30    --     --
  Stay 35 years   1.07    --     --     --
  Stay 40 years   1.17    --     --     --

                            Women

                        Age Hired At

                   25     35     45     55

Defined Contribution Plan with 50% Stocks and 50% Bonds

  Stay 5 years    0.34   0.56   0.92   1.46
  Stay 10 years   0.39   0.65   1.05   1.65
  Stay 15 years   0.45   0.74   1.19    --
  Stay 20 years   0.52   0.84   1.34    --
  Stay 25 years   0.58   0.92    --     --
  Stay 30 years   0.66   1.07    --     --
  Stay 35 years   0.74    --     --     --
  Stay 40 years   0.83    --     --     --

Defined Contribution Plan with 100% Bonds

  Stay 5 years    0.68   0.91   1.23   1.65
  Stay 10 years   0.75   1.00   1.35   1.80
  Stay 15 years   0.82   1.10   1.47    --
  Stay 20 years   0.90   1.20   1.61    --
  Stay 25 years   0.98   1.27    --     --
  Stay 30 years   1.07   1.43    --     --
  Stay 35 years   1.16    --     --     --
  Stay 40 years   1.26    --     --     --

(a) It is assumed that the individual begins pension benefits
at age 65. Life expectancy at age 65 is assumed to be 16.79
years for men and 19.96 years for women.

Table 4. Internal Rate of Return for Defined Benefit
and Defined Contribution Plans (a)

                           Men

                       Age Hired At

                 25     35     45     55

Stay 5 years    1.91   2.45   3.43   5.79
Stay 10 years   2.09   2.72   3.92   7.14
Stay 15 years   2.30   3.03   4.51    --
Stay 20 years   2.59   3.21   5.21    --
Stay 25 years   2.76   3.78    --     --
Stay 30 years   2.95   4.22    --     --
Stay 35 years   3.32    --     --     --
Stay 40 years   3.63    --     --     --

                          Women

                       Age Hired At

                 25     35     45     55

Stay 5 years    2.22   2.83   3.92   6.46
Stay 10 years   2.39   3.10   4.41   7.78
Stay 15 years   2.59   3.39   4.98    --
Stay 20 years   2.80   3.74   5.66    --
Stay 25 years   3.02   4.12    --     --
Stay 30 years   3.28   4.53    --     --
Stay 35 years   3.55    --     --     --
Stay 40 years   3.85    --     --     --

(a) It is assumed that the individual begins pension benefits
at age 65. Life expectancy at age 65 is assumed to be 16.79
years for men and 19.96 years for women.

Table 5. Summary Statistics by Pension Type: New Hires

                             Full Sample   Defined      Defined
                                           Benefit    Contribution

Year of hire (%)
  1983                           4.9           7.0        4.5
  1984                           4.9           8.0        4.3
  1985                           4.4           5.9        4.1
  1986                           5.3           7.6        4.9
  1987                           4.4           6.3        4.0
  1988                           5.3           7.7        4.8
  1989                           6.4           6.7        6.4
  1990                           5.5           4.4        5.7
  1991                           4.1           3.3        4.3
  1992                           5.4           4.8        5.5
  1993                           5.3           4.6        5.4
  1994                           4.9           4.9        4.9
  1995                           5.3           4.0        5.6
  1996                           4.7           6.3        4.4
  1997                           4.5           2.2        4.9
  1998                           5.0           2.9        5.4
  1999                           5.6           3.2        6.1
  2000                           5.8           4.2        6.1
  2001                           8.2           6.0        8.6

Camegie classification (%)
  Research                      32.3          16.3       35.4
  Doctoral                      25.0          11.1       27.7
  Masters                       36.5          57.1       32.5
  Baccalaureate                  6.2          15.4        4.4
  Female (%)                    35.3          41.1       34.2

Ethnic origin (%)
  White                         80.9          60.3       84.9
  Black                         11.4          32.2        7.4
  Other (a)                      7.6           7.5        7.7

Age at hire (%)
  30 or younger                 12.5           8.6       13.2
  31-35                         27.4          18.9       29.0
  36-40                         23.4          22.3       23.7
  41-45                         16.5          19.3       16.0
  Over 45                       20.2          30.9       18.1
  Salary at hire (1983$)     38111.94      33551.80   38994.73

Rank at hire (%)
  Full professor                12.9          13.8       12.7
  Associate professor           16.1          21.8       15.0
  Assistant professor           68.0          59.9       69.5
  Other (b)                      3.0           4.4        2.7
  Tenured at hire (%)           11.6           8.7       12.2

Sample size                   7035          1141       5894

(a) Native American, Asian, and Hispanic.

(b) Instructors, Lecturers, and others.

Table 6. Probit Analysis of Pension Choice: New Hires (a)

                                                           Estimated
                                  Parameter    Standard     Marginal
Variable                          Estimate      Error      Effect (b)

Constant                          -1.1717 *   0.1055          --

Female                             0.1083 *   0.0427        0.0223

Ethnic Origin: (Omitted Category = White)
  Black                            0.7926 *   0.0545        0.2167
  Other (c)                        0.2402 *   0.0742        0.0544

Age at hire (omitted category ages 31-35)
  30 or younger                   -0.1013     0.0746       -0.0195
  36-40                            0.1724 *   0.0595        0.0367
  41-45                            0.3186 *   0.0660        0.0725
  Over 45                          0.6321 *   0.0681        0.1561
  Salary at hire/10,000 (1983$)   -0.0144     0.0135       -0.0029

Rank at hire (omitted category assistant professor)
  Full professor                  -0.0065     0.1000       -0.0013
  Associate professor              0.0015     0.0582        0.0003
  Other (d)                        0.5876 *   0.1092        0.1572
  Hired with tenure               -0.1305     0.1015       -0.0248

Carnegie classification (omitted category research)
  Doctoral                        -0.1622 *   0.0640       -0.0312
  Masters                          0.5650 *   0.0548        0.1247
  Baccalaureate                    0.8367 *   0.0813        0.2397
N                                               7035
                                              -2380.1445
Log-likelihood                                -2550.0503

                                  * [alpha] < 0.05; ** [alpha] < 0.10

(a) The dependent variable is 1 if the individual chooses the defined
benefit pension plan and 0 if the individual chooses the defined
contribution plan. The analysis includes dummy variables for each
year (omitted year 1983).

(b) The estimated marginal effect is evaluated at the means of all
continuous variables and for the discrete change from 0 to 1 for all
dummy variables.

(c) Native American, Asian, and Hispanic.

(d) Instructors, Lecturers, and others.

Table 7. Predicted Probabilities of Defined Benefit Pension
Choice over Time: New Hires (a)

Year of Hire            Predicted Probability of
                            Choosing DB Plan

  1983                           11.00
  1984                           10.60
  1985                            7.53 **
  1986                            8.80
  1987                            8.73
  1988                            9.18 *
  1989                            4.81 *
  1990                            3.11 *
  1991                            3.05 *
  1992                            3.27 *
  1993                            3.64 *
  1994                            3.55 *
  1995                            2.48 *
  1996                            5.85 *
  1997                            1.12 *
  1998                            1.48 *
  1999                            1.68 *
  2000                            2.36 *
  2001                            2.44 *

* [alpha] < 0.05; ** [alpha] < 0.10

(a) The predicted probabilities are determined from a probit
equation of pension choice where the dependent variable is 1
if the individual chooses the defined benefit pension plan and
0 if the individual chooses the defined contribution plan. The
probabilities are estimated for a base case, which is a white
male faculty member, hired as an assistant professor without
tenure at a research university at a salary of $38,112. All
statistical significance indicators are relative to the omitted
year in the estimation (1983).