skip navigational linksDOL Seal - Link to DOL Home Page
Photos representing the workforce - Digital Imagery© copyright 2001 PhotoDisc, Inc.
www.dol.gov/ebsa
November 8, 2004    DOL > EBSA > Publications > Pension Report   

Does 401(k) Introduction Affect Defined Benefit Plans?

Table of Contents

  • Abstract

  • Introduction & Summary

  • An Overview of 401(k) Plans

    • Relevant Literature

  • Sample Construction

    • The 1985 Sample

    • The 1992 Sample

    • Sample Attrition

  • Econometric Estimates

    • Sponsor-level Estimates of Plan Substitution

    • Plan-level Estimates of Plan Substitution

    • Changes in DB Participation

    • Changes in the DB Funding Ratio

  • Data Appendix

    • Measures of Participation and Eligibility

    • Calculation of the Funding Ratio

      • Calculation of the 1985 funding ratio

      • Calculation of the 1992 funding ratio

  • Tables

    • Table 1: Single Employer Pension Plans with 100 or More Participants

    • Table 2: Summary Statistics for Sample of Primary Defined Benefit Plans

    • Table 3: Summary Statistics for Accompanying 401(k) Plans PY 1985

    • Table 4: Summary Statistics for Sample of Primary Defined Benefit Plans, PY 1992

    • Table 5: Summary Statistics for Accompanying 401(k) Plans PY 1992

    • Table 6: Linear Probability Model Dependent Variable: Firm Attrition in 1992

    • Table 7: Linear Probability Models Dependent Variable: Plan Attrition Due to Firm Attrition in 1992

    • Table 8: Summary Statistics for Variables in Linear Probability Attrition Models PY 1985

    • Table 9: OLS Regression Dependent Variable: Change in the Number of DB Plans Offered by the Firm Between 1985 and 1992

    • Table 10: OLS Regression Dependent Variable: Change in the Number of DB Plans Offered by the Firm Between 1985 and 1992

    • Table 11: Linear Probability Model Dependent Variable: Is the DB Plan Still Offered in 1992?

    • Table 12: Linear Probability Dependent Variable: Is the DB Plan Still Offered in 1992?

    • Table 13: OLS Regression Dependent Variable: Change in DB Participation Between 1985 and 1992

    • Table 14: OLS Regression Dependent Variable: Change in DB Funding Ratio Between 1985 and 1992

  • References

  • Endnotes

 Back To Top

Abstract

This paper examines whether sponsors of traditional defined benefit (DB) plans are gradually replacing them with 401(k) or other defined contribution (DC) plans. I compare pension plan offerings by sponsors of a DB plan in 1985 with their offerings in 1992 using Form 5500 filings from those two years. I find that 401(k) and other DC plans are substituting for terminated DB plans and that offering a DC-type plan increases the probability of a DB termination.

Using several specifications, a sponsor that starts with no 401(k) or other DC plan and adds a 401(k) is estimated to reduce the number of DB plans offered by at least 0.3. That is, the estimates imply that one sponsor terminates a DB plan for about every three sponsors that offer at least one new 401(k) plan. The addition of a non-401(k) DC is also estimated to reduce DB plan offerings.

The point estimates indicate that if a 401(k) plan is added by a sponsor, the DB termination probability increases by about 18 percentage points to 35 percent. The addition of a non-401(k) plan similarly increases the probability that a DB plan will be terminated.

The opinions expressed in this paper are the sole responsibility of the author and do not represent the views of the U.S. Department of Labor.

 Back To Top

Introduction and Summary

The 401(k) plan has become the leading form of tax-deferred individual retirement saving. Recent research has focused on whether individual contributions represent net new saving (Engen, Gale, and Scholz (1994), and Poterba, Venti, and Wise (1995)). Since employer-sponsored 401(k) plans are typically offered along with a defined benefit plan (Andrews (1992), and Papke (1995)), substitution between pension plan types is possible at the employer level as well. Indeed, recent articles in the popular press often raise the issue of defined benefit (DB) plan phase out in favor of defined contribution (DC) plans (Newsweek, 1993). Employers may view the 401(k) plan as an opportunity to reduce their pension costs since participation in 401(k) plans is voluntary. The purpose of this paper is to examine the effect of 401(k) offerings on existing defined benefit plans. The influence of non-401(k) defined contribution offerings will also be examined.

The principle findings of this study are:

  • The 1985 sample consists of 16,597 DB plans and 11,950 distinct sponsors. Eighty-three percent of employers offer one DB plan, and over thirteen percent offer one 401(k) plan in addition to their DB plan in 1985. A non-401(k) defined contribution plan is offered by 17.21 percent of these employers. By 1992, 6,974 of these sponsors are still ongoing. Of these, 20.78 percent (1,449) offer no DB plan. Thus, over twenty percent of the employers filing Form 5500s in 1992 drop their 1985 DB plan, but retain or add a DC or 401(k) plan.

  • DB offerings are reduced when a sponsor introduces either a 401(k) or other DC plan. The addition of a 401(k) plan is estimated to reduce the number of DB plans by about 0.3 relative to a sponsor that offers no 401(k) or other DC plan. That is, a sponsor terminates a DB plan for about every three sponsors that offer at least one new 401(k) plan. In the sample, this effect accounts for almost 18 percent of the DB plans lost to ongoing sponsors in the panel and the DB loss affects four to five percent of all 1985 DB participants. Maintaining the 401(k) offered in 1985 reduces the number of DB plans by about 0.2.

  • Non-401(k) DC plans are also estimated to reduce DB plan offerings, and the estimated effects are larger than those for 401(k) plans. The addition of a DC plan reduces the number of DB plans offered by about 0.4. Maintaining the existing DC leads to a termination of 0.26 of a DB plan.

  • The point estimates indicate that if a 401(k) plan is added by a sponsor, the DB termination probability increases by about 18 percentage points to 35 percent. The addition of a non-401(k) plan similarly increases the probability that a DB plan will be terminated.

  • There is some evidence that adding a 401(k) or other DC plan reduces participation in the accompanying DB plan, but overall, not much variation in DB participation is explained by changes in 401(k) or DC plan status. Changes in 401(k) or other DC plan offerings do not affect DB funding ratios.

 Back To Top

An Overview of 401(k) Plans

Employer-sponsored pre-tax saving plans, often called 401(k) plans, differ from traditional employer-sponsored retirement plans in that eligible employees are permitted to make pre-tax contributions to the plan. Most employers match the contributions (typically with a match rate of 50 percent up to six percent of compensation). With the matching provision, a higher contribution limit, and no adjusted gross income phase-out of deductibility, 401(k) plans have a greater potential for pre-tax saving than do Individual Retirement Accounts. The percentage of employees with an employer who sponsors a 401(k) plan increased from 26.9 percent in 1988 to 36.8 percent in 1993. The fraction of participating workers rose from 57.0 percent to 64.6 percent.(1)

Employers may find 401(k) plans a cost-effective way to offer pension benefits to workers with different preferences. The employer need only make a contribution (if they match) for workers who have already made a contribution. Even and Macpherson (1993) find that the probability of an employee participating in a pension plan offered by a firm is lower if the primary plan is a 401(k).

Because the saving is tax-favored, the Internal Revenue Service imposes contribution limits, and restricts participant access to the funds. The plan must pass certain nondiscrimination tests to remain tax-qualified.(2)

 Back To Top

Relevant Literature

There are two strands of the literature that address, either directly or indirectly, substitution between types of pension plans. The first documents broad economy-wide shifts in the shares of employees covered by types of plans from the mid-1970s to mid-1980s.

Gustman and Steinmeier (1992) analyze the role of unions, firm size, and industry composition using repeated cross sections from Form 5500 filings. They find that these factors could account for at most half of the decline in primary DB coverage from 1977 to 1985 (from 89.7 percent in 1977 to 79.3 percent in 1985).

Clark and McDermed (1990) reexamine the issue using Form 5500 data from 1979 and 1983. They conclude that only 3.1 of the 15 percentage point decline in the proportion of DB coverage can be explained by changes in the size distribution of firms and industrial composition. They supplement this plan level analysis with the 1979 and 1983 Current Population Survey supplements on pension coverage, and the 1983 Survey of Consumer Finances.

They conclude that demographic, employer size, and broad industry and occupation variables explain none of the coverage decline. They speculate that pension regulation could be responsible for the increased use of DC plans.

Petersen (1992) examines DB terminations without directly addressing the shift to DC plans. He finds that financial factors play a strong role in the decision to terminate overfunded DB plans (greater needs for cash and higher costs in obtaining external capital increase the risk of termination and a reversion of assets).

The issue of plan substitution is addressed directly in a recent survey by Papke, Petersen, and Poterba (1996). Their evidence suggests that a majority of sponsors do not terminate their defined benefit plans as a result of 401(k) introduction. They find that only one (albeit large) sponsor in their small sample of 43 plans reports termination as a result of 401(k) plan introduction between 1986 and 1990. Forty-five percent of responding firms, representing 37 percent of 401(k) participants in the survey, indicated that another pension plan was converted to the 401(k) (converted from thrift or profit-sharing). These survey responses indicate that 401(k) plans typically supplement a preexisting DB plan (63 percent) or DC plan (19 percent). The 401(k) was the only retirement plan for covered workers at 19 percent of plans, representing about 10 percent of participants.

The second strand of literature examines substitution between DB and DC plans with plan level tabulations of participation and assets for the entire pension sector. Kruse (1995) matches sponsors using Form 5500 data from 1980 to 1986 to study adoptions and terminations of different types of plans, and estimates the number of employees involved. For each type of plan, he tabulates participant counts for companies that adopt, maintain, or terminate a plan (plan categories are DB, All DC, Profit sharing, and ESOPs) over this period. His key conclusions for the purposes of this report are the following: (1) Very little of the growth in DC participants came from companies that terminated DB plans, with most growth coming from companies with no change in DB status; (2) The largest declines in DB participants came from companies that had no change in DB or DC status; (3) ESOPs tend to supplement rather than replace DB plans.

Kruse documents a total of 7.2 million 401(k) participants as of 1986 (out of a total of almost 33 million DC participants), but 401(k) data are not available for 1980 and he is unable to separate out the effect of the adoption of these plans on other plan offerings or participant counts.

The aggregate plan and participant counts over the first half of the 1980s -- the period Kruse examines -- is broadly characterized by a stable number of net DB plans, and a net increase in the number of and participation in DC plans (see Table 1). While the number of large single employer DC plans (with 100 or more participants) rises from 12,978 to 25,713 over this period, net DB plans are virtually unchanged (22,424 in 1980 and 22,426 in 1986), though Kruse documents adoptions and terminations.

However, aggregate data from the latter half of the 1980s and early 1990s suggest a need to re-examine the issue of substitution of DC for DB plans. Two changes in particular indicate an increase in substitution possibilities. First, the net number of DB plans falls dramatically. From 1985 to 1992, the number of large DB plans (over 100 participants) falls 25 percent (from 22,619 to 17,019) while the number of DC plans increases about 64 percent (from 23,279 to 38,283). If single employer plans of all sizes are considered, the DB count falls from 167,911 in 1985 to 86,797 in 1992, and the number of single employer DC plans increases from 461,158 to 618,429 in 1992 (see Table 1). Changes in the number of active participants mirror these movements.(3)

Second, as Table 1 illustrates, adoption of 401(k) plans escalates over this period. The availability of 401(k) plans rises from 29,869 in 1985 to 139,704 in 1992. The number of active participants (who may also participate in other plans) more than doubles (from 10.3 million in 1985 to 22.4 million in 1992). Assets rise from $143.9 billion to $552.9 billion over this period (See U.S. Department of Labor, 1996, Table E23.)

Papke (1994) uses some preliminary Form 5500 data from the second half of the 1980s to estimate the effect of 401(k) introduction on the characteristics of the companion defined benefit plan. Using a difference-of differences methodology, she compares changes in defined benefit participation, funding, and intention to terminate the existing DB plan, between 1985 and 1991 for plans whose sponsors introduced a 401(k) plan over this period and for plans whose sponsors did not.

Papke matches single DB plans in 1985 to their filings in the preliminary 1991 Form 5500 file along with any 401(k) plan that was added over that period to create a panel of 3,406 matched ongoing defined benefit plans. She compares differences in intended termination probabilities, participation rates, and funding ratios in the ongoing defined benefit plans between 1985 and 1991 for sponsors that introduced a 401(k) and those that did not.

Papke finds no evidence that the introduction of a 401(k) plan affects defined benefit plans offered by larger sponsors (those with over 200 employees). However, it appears that the smaller sponsors do reduce participation and funding, and are more likely to indicate they intend to terminate their defined benefit plan, once a 401(k) plan is introduced.

This paper extends Papke's work in a number of ways. First, multiple DB plans with at least 100 participants are included in the initial 1985 sample of plans - not just single DB plans. I will examine the effect of 401(k) introduction on the number of DB plans offered by a sponsor. Second, the earlier analysis is limited to DB plans that are maintained over the period so it misses DB plans that terminate and stop filing Form 5500s between 1985 and 1991. In contrast, this analysis tracks sponsors of DB plans functioning in 1985. The initial DB plan(s) and all related plans are tracked between 1985 and 1992. I note termination of the original DB plan over the period as well as the reported intention to terminate in the 1992 filing. Third, this paper examines both 401(k) introduction and the effect of changes in the number of 401(k) and non-401(k) DC offerings on the sponsor's DB plan.

However, this paper does not address substitution that occurs when a firm closes and terminates its DB plan, and start-up firms adopt DC rather than the traditional DB pension form. Kruse's (1995) work concerning the first half of the 1980s remains the most recent on that issue.

 Back To Top

Sample Construction

This study addresses two types of questions. How do the sponsor's 401(k) plan offerings affect DB plan offerings? In particular, does the introduction of a new 401(k) plan affect DB offerings by that sponsor? At the plan level, how does a 401(k) plan affect termination probabilities, participation, and funding of the accompanying DB plan? The main focus of this study is to compare the characteristics of DB plan offerings after a change in the sponsor's 401(k) offerings, but the effects of non-401(k) DC plan offerings are also measured.

I use data from the 1985 and 1992 Form 5500 filings available from the Pension and Welfare Benefits Administration of the U.S. Department of Labor. The Form 5500 is filed annually with the Internal Revenue Service by all sponsors of pension plans with 100 or more participants. These reports include information on plan eligibility, participation, employment, administrative cost, distributions, and contributions.

The 1985 sample of DB plans is determined in the following way. All single employer defined benefit (DB) plans with no more than one associated 401(k) are selected. Using the Employer Identification Number (EIN) of the sponsor to identify companion plans, I summarize information on accompanying 401(k) and other DC plans for those sponsors. This summary information includes the count of companion 401(k) and DC plans and their type (ESOP or profit sharing, for example), and participation, eligibility, assets, and coverage.

The Form 5500 includes plan participation, but it is not possible to determine how many employees participate in more than one plan offered by the sponsor. It is generally understood that employees participate in only one DB plan, but participation in an accompanying DC plan, or in more than one DC plan is possible. Therefore, when appropriate, I sum characteristics across all accompanying 401(k) or other DC plans to create the characteristics of the accompanying plans.

The 1985 list of DB plans is matched by EIN to plan reports in the 1992 Form 5500 data, the most recent year for which the Form 5500 data are available. All plan offerings by the 1985 list of sponsors are selected. The original DB plan is matched to its report if present, and I summarize the information on the sponsor's accompanying 401(k) and other DC plans.

Some of the 1985 DB plans do not appear in the 1992 filings, and it is important to distinguish the two types of sample attrition that account for their disappearance. First, the plan's sponsor (EIN) may disappear entirely. Second, the DB plan may disappear, but the sponsor has filed a Form 5500 file for another plan. In this second case the DB plan is considered to be terminated.

The absence of a DB plan from the 1992 Form 5500 data when the sponsor has filed for other plans likely indicates a plan termination, but it could also be the result of a plan merger that changes the plan number, (about three percent of plans indicate an intention to merge in 1985), or a filing error. These possibilities are not explored here.

To summarize the construction of the panel, a DB plan is matched to its 1992 record if available and accompanying 401(k) and DC plan information is kept. The DB plan is considered to be ongoing, unless the sponsor indicates the plan has been terminated or a resolution to terminate the plan has been filed. In these cases, the plan is considered to be terminated. If the 1985 DB plan has no match in 1992, but the sponsor reports other non-DB plans, the original 1985 DB plan is presumed to have been terminated over the period. Its accompanying 401(k) and DC plan information is kept. If no plans for the sponsor of the original 1985 DB plan appear in 1992, the 1985 DB plan is dropped from the panel data analysis since either the sponsor has ceased operation over this period or there has been a filing error.

 Back To Top

The 1985 Sample

The 1985 sample consists of 16,597 DB plans, 11,950 (72 percent) belong to distinct sponsors (that is, the plan's sponsors have different EINs -- the terms sponsor and employer will be used interchangeably here). Eighty-three percent of employers offer one DB plan, 10.43 percent offer two DB plans, and 2.86 percent offer three DB plans. The remaining 3.55 percent of employers in the sample offer more than three DB plans.

Of the 11,950 employers, 1,586 (13.27 percent) offer one 401(k) plan in addition to their DB plan in 1985. A non-401(k) defined contribution plan (hereafter, a DC plan) is offered by 17.21 percent of these employers; most of these (82.63 percent) offered only one DC plan.

Table 2 displays the summary statistics for the 1985 sample of DB plans. Both means and medians are presented since the largest plans dominate the means of some characteristics. Two measures of participation are reported. The first, a response to question 7a4 on the 1985 Form 5500 ("Total number of active participants at the end of the plan year") indicates an average active participant count of 587, and a median of 211. The second measure reported in response to question 17h ("Information about employees of employer at end of the plan year: Employees participating") indicates more participants on average, 772, but a similar median, 216. (See the Data Appendix for a discussion of the various measures of plan participation.)

The participation rate (participation divided by eligible employees) for DB plans is virtually 100 percent, with an average coverage rate (number of eligible employees divided by total employees) of 57 percent (a median coverage rate of 65 percent). Seventeen percent of DB plans in the sample are accompanied by a 401(k) plan. Almost nine percent of DB plans filing in 1985 have been terminated or indicate on their Form 5500 that a resolution to terminate has been filed. The average funding ratio is 1.63 (See the Data Appendix for a discussion of the funding ratio calculation.). The average age of the DB plans is 18 years, and median plan assets are $2.33 million (average assets are $12.4 million).

Table 3 reports summary statistics for the 401(k) plans that accompany these DB plans in 1985. Both measures of participation indicate that the average 401(k) plan has about 1000 participants, although the median plan is much smaller - about 270. The participation rate is 83.3 percent (using question 7a4), and the average coverage rate 63.7 percent. Most 401(k) plans are relatively new -- the average age is 2.84 years, median assets are $1.88 million (average assets are $11.3 million), and the calculated average match rate offered is 0.64.(4)

 Back To Top

The 1992 Sample

For the purposes of comparison over time with the 1985 sample of plans, I create a 1992 matched pair sample. In the event that the sponsor has no DB or other plan filing in 1992, for the purposes of describing sponsor attrition in this section of the paper, I create a phantom or place-holder plan. These plans are not included in the panel data analysis to follow.

Of the 16,597 DB plans in the 1985 sample, 52 are dropped from the 1992 sample due to a change in entity type or an incorrectly reported plan type. There are 16,545 DB plans in the 1992 sample, 7,017 plans (42.41 percent) are true matches with a 1985 record, and 9,528 (57.29 percent) are constructed place-holder DB plans that do not appear in 1992. Note that the sponsors of these phantom DB plans may have exited completely, or may continue to sponsor 401(k) or DC plans. That is, the employer filed a Form 5500 for a 401(k) or DC plan, but not for the DB plan it reported in 1985. Including phantom plans, there are 11,916 distinct employers in the 1992 sample.

Of the 11,916 employers in the 1992 sample, 58.53 percent (6,974) filed at least one Form 5500 -- either for a DB plan, a DC plan, a 401(k) plan, or some combination of these. These are ongoing firms. The remaining 41.46 percent (4,940) of employers are phantoms -- their 1985 DB plan does not appear in the 1992 sample and they have no accompanying DC or 401(k) plan in 1992. Thus, we cannot be sure if the absence of their DB plan indicates a true termination or an administrative change or error. However they exited, about 41 percent of employers left between 1985 and 1992.

Of the 6,974 real employers who appear in the 1992 sample, 20.78 percent (1,449) offer no DB plan. Thus, over twenty percent of the employers still reporting in the 1992 sample dropped their 1985 DB plan, but retained or added a DC or 401(k) plan. The majority, 68.30 percent of these employers (4,763), offer one DB plan, 7.26 percent (506) offer two DB plans, 1.82 percent (127) offer three DB plans. The remaining 1.88 percent of employers offer more than three DB plans.

Of the 6,974 ongoing employers in 1992, 61.40 percent (4,282) offer no 401(k) plan, 33.68 percent (2,349) offer one, and 3.87 percent (270) offer two. The remaining 1.08 percent offer more than two 401(k) plans. Most of the employers offer no accompanying DC plan (71.36 percent), 23.27 percent offer one, 4.24 percent offer two, and the remaining 1.13 percent offer three or more DC plans. By 1992, more employers offer an accompanying 401(k) than a DC plan.

Table 4 reports the summary statistics for the 1992 sample of ongoing DB plans. Two measures of participation are reported. The first, reported in response to question 7a4 (as in 1985, "Total number of active participants as of the end of the plan year") indicates an increase in the average size of DB plans from 587 participants on average in 1985 to 843. The median has also increased to 257 (from 211).The second measure of participation is taken from question 21, which is the most comparable to question 17h from 1985. However, the set of questions concerning employees of the employer have been extensively expanded to include questions related to anti-discrimination requirements. Many fewer records contain answers to the subparts of question 21 than to question 17 in 1985. The participation subpart, question 21m, ("Enter the number of non-excludable employees who benefit under the plan") indicates that on average 475 employees benefit under the plan(s).(5)

Average participation rates in DB plans have fallen relative to 1985 using either measure (to 80 or 88 percent of eligibles). (Note that the sample size for the rate calculation is halved due to missing values to question 21j ("Number of non-excludable employees"). The DB coverage rate has increased to 71 percent. About five percent of DB plans in 1992 report a termination or an intention to terminate. The average funding ratio has fallen to 1.01. (The funding ratio is expected to fall over this period due to the funding limits in the Omnibus Reconciliation Act of 1987 (OBRA87) regardless of sponsor activity.)

Table 5 reports the summary statistics for the 401(k) plans that accompany an ongoing DB in 1992 or that are reported by an ongoing firm that sponsored a DB in 1985. Recall that these may be sums over multiple 401(k) plans if more than one 401(k) plan is offered by the employer. Average participation in the 401(k) plan(s) has risen since 1985 from 1,067 to 1,592. The participation rate (the sum of all plan participants divided by the sum of eligibles in 401(k)s) is about the same, 81.9 percent, and the coverage rate is up from 63.7 percent in 1985 to 70.3 in 1992. The average match rate has fallen to 0.40 (although the median rate is about the same, 0.33), and the average maximum match rate across plans offered by the same sponsor is 0.42. Average assets in 401(k) plans have increased from $11.3 million in 1985 to $41.7 in 1992, and the 1992 median is $4.71 million.

 Back To Top

Sample Attrition

As mentioned above, a 1985 DB plan may leave the 1992 sample for one of two reasons. Either its sponsor's EIN vanishes entirely, or the plan number vanishes but the EIN continues to file Form 5500s for its other plan(s). The second case is considered a plan termination not sample attrition. This section discusses plan absence due to sponsor attrition.

Table 6 presents estimates of a linear probability model that explains EIN attrition in 1992 based on the 1985 sponsor-level characteristics that are available on the Form 5500. The dependent variable equals one if the sponsor is absent in 1992, and zero otherwise. The estimates in the second column of Table 6 indicate that sponsors that offer more plans are less likely to vanish in 1992. The probability of sponsor attrition falls by .019 or is almost two percentage points less for each DB plan offered, 13.2 percentage points less for each 401(k) plan, and 8.7 percentage points less for each DC plan. These point estimates are statistically significant at well under the five percent level.

The addition of the log of total firm employment (column three) and 1-digit SIC industry dummies (column four) does not change these estimates appreciably. The most complete model (column four) indicates that larger firms are more likely to vanish (a 10 percent increase in firm employment increases the probability of attrition by .0134 or 1.34 percentage points). Although highly statistically significant, the economic effect is small.

Estimates from a second linear probability model of plan attrition are reported in Table 7. The dependent variable equals one if the plan is absent from the 1992 sample because its sponsor's EIN is absent, and zero otherwise. Plan attributes from 1985 are included as independent variables, along with the sponsor-level variables used in the previous model.

The most complete model (column six) indicates that the probability of plan and it's sponsor vanishing falls by .024 for each DB plan sponsored. They are 9.1 percentage points less likely to vanish for each 401(k) plan, and 8.1 percentage points less for each DC plan. A 10 percent increase in firm employment increases the probability of attrition by .004 or .4 percentage points, and a 10 percent increase in plan assets reduces the probability of attrition by the same amount. All of these point estimates are statistically significant at the five percent level or below.

A higher plan participation ratio in 1985 reduces the probability of attrition. (Participation is measured in these regressions with the responses to question 7a4.) The point estimate indicates that an increase in the ratio of 0.1 reduces the probability by .01 or one percentage point. This estimate is not quite statistically significant at the five percent level. Plans with more assets are less likely to vanish but the effect is small -- a 10 percent increase in plan assets reduces the probability of attrition by .39 percentage points.

Older plans are less likely to vanish (the probability drops by 0.2 percentage points for each year of age), and if an intention to terminate the plan was indicated in 1985, the plan is 18.3 percentage points more likely to vanish by 1992. These estimates are statistically significant at the five percent level.

While administrative costs per participant appear to play no role in plan attrition, the composition of participants matters. The higher the ratio of retired or separated participants receiving or entitled to receive benefits to active participants (questions (7b+7c)/7a4) the higher the likelihood of attrition. For a 0.1 increase in this ratio, the probability of attrition increases by 0.62 percentage points, holding plan age fixed. This effect is statistically significant at the five percent level.

Summary statistics for the samples used to estimate the linear probability models are presented in Table 8.

 Back To Top

Econometric Estimates

This study addresses two types of questions. How do the sponsor's 401(k) plan offerings affect DB plan offerings? In particular, does the introduction of a new 401(k) plan affect DB offerings by that sponsor? At the plan level, how does a 401(k) plan affect termination probabilities, participation, and funding of the accompanying DB plan? The first pair of tables discussed in this section present estimates of the effects of 401(k) and DC offerings on DB plans at the sponsor level.

 Back To Top

Sponsor-Level Estimates of Plan Substitution

How does a sponsor's DB offerings change with a change in 401(k) or other DC offerings? Are these DC plans substituting for the traditional DB plan? Table 9 presents estimates of the effect of the change in the number of 401(k) or other DC plan offerings on the change in the number of DB plans offered by the sponsor. Specifying the regression equation in changes over the two years holds constant any characteristics of the sponsor that do not change over the period. The (log of) the level of firm employment in 1985 is included rather than the change in firm employment, because firm employment is missing for many plans in the 1992 data (see the Data Appendix for a discussion).

Estimates of the most complete model are presented in column four and indicate that both 401(k) and DC plans are substitutes for DB plans although the effect is not one for one. The addition of a 401(k) plan over this period results in a reduction of 0.184 of a DB plan. In other words, a DB plan is predicted to be terminated for about every five 401(k) plans added. An additional non-401(k) DC plan is estimated to result in a reduction of 0.152 of a DB plan, after controlling for total firm employment in 1985 and industry dummies.

In the sample, 66.12 percent (4,611 out of 6,974) of employers do not change the number of 401(k) plans offered, and 28.95 percent (2,019) add at least one. The number of net 401(k) plans added is 2,071. Thus, the point estimate indicates that in this sample, these additional 2,071 401(k) plans are associated with a loss of about 381 DB plans, 10.81 percent of all DB plans lost to ongoing sponsors (3,526 DB plans).

In the sample, 72.71 percent (5,033 out of 6,922) of employers do not change their DC offerings, and 17.77 percent (1,239) add at least one. The net change in DC plans is 700. The point estimate indicates that these 700 additional DC plans are associated with a loss of about 106 DB plans.

The estimates in Table 9 require that the effect of adding a 401(k) plan is the same (though of opposite sign) as subtracting a 401(k) plan. The estimates in Table 10 relax this assumption and allow for different 401(k) and DC plan status. Three plan status dummy variables are included for each plan type. The categories are: (1) the firm offers at least one new 401(k) plan over this period; (2) the firm maintains the 401(k) it had in 1985 and did not add another; (3) the firm loses the 401(k) it had in 1985. Similar dummy categories are defined for non-401(k) DC plans, adjusted for the fact that there may be multiple DC plans in 1985. The coefficient estimates on these dummy variables are interpreted relative to the base or omitted category: the firm offers no 401(k) or DC plan in 1985 or 1992.

The effects of 401(k) and DC plan status reported in Table 10 are fairly consistent across the models and again indicate less than one for one substitution across the plan types. The r-squared of the regression indicates that almost 12 percent of the variation in DB plan offerings by sponsors is explained by the status of accompanying 401(k) or other DC plans (column 2).

The most complete model is presented in column four and includes the log of firm employment in 1985 and industry dummies. These estimates indicate that the addition of at least one 401(k) plan reduces the number of DB plans by 0.312 relative to a sponsor that offers no 401(k) or other DC plan. (This is essentially the same as adding one 401(k) plan since 85 percent of sponsors that add at least one add only one.) This estimate implies that a sponsor terminates a DB plan for about every three sponsors that offer at least one new 401(k) plan. Since 2,019 sponsors added at least one 401(k), this estimate predicts a loss of 630 DB plans or almost 18 percent of the DB plans lost to ongoing sponsors in the panel.

Maintaining the 401(k) offered in 1985 reduces the number of DB plans by 0.213. In the sample, 673 sponsors maintained the 401(k) they had in 1985, so the estimate predicts a termination of 143 DBs. The t-statistics for these two effects are extremely high. Substitution of a DB for a terminated 401(k) plan is indicated as well, but the coefficient is small and statistically insignificant.

Using 1985 DB average plan characteristics from Table 2, it is possible to estimate a range in the number of participants who lose their DB plan when a 401(k) is added. Multiplying the number of predicted DB terminations (630) by one of the two measures of participation -- active participants or employees participating -- indicates that between 364,172 and 486,562 participants are involved. These participants represent 4.04 percent of all active participants or 4.91 percent of all employees participating in a DB plan in 1985.

Non-401(k) DC plans are also estimated to be substitutes for terminated DB plans, and the estimated effects are larger than those for 401(k) plans. The addition of a DC plan reduces the number of DB plans offered by 0.410. If 2.4 sponsors offer at least one new DC plan, then one DB plan is predicted to be terminated. Since 1,239 DCs were added, the estimate predicts a drop of 508 DB plans for ongoing sponsors. Maintaining the existing DC leads to a termination of 0.264 of a DB plan. For DC plans, the substitution does not appear to go in the other direction. That is, if a sponsor loses a DC plan, it is estimated to also terminate 0.134 of a DB plan. These estimated effects are also highly statistically significant.

 Back To Top

Plan-Level Estimates of Plan Substitution

Tables 11 and 12 present evidence that the adoption of a 401(k) or DC plan leads to an increased probability of DB plan termination. The dependent variable is one if the DB plan is still present in 1992. The dependent variable equals zero if the Form 5500 indicates that the plan has been terminated recently, or if a resolution to terminate the plan has been passed, or if the plan is missing from the 1992 data but the sponsor has filed Form 5500s for other plans (presumed terminated).

The linear probability estimates in the fourth column of Table 11 indicate that the addition of a 401(k) plan over this period reduces the probability of the DB continuing by .044 or by 4.4 percentage points. The addition of a DC plan similarly increases the DB termination probability by 4.2 percentage points. The effects of 1985 firm employment and plan assets continue to be small -- an increase of 10 percent in 1985 firm employment increases the termination probability by 0.37 percentage points, and a 10 percent increase in plan assets reduces the termination probability by .52 percentage points.

In Table 12, the three plan status categories are included for 401(k) and DC plans. In column four, which includes industry dummies, and controls for firm employment and plan assets in 1985, an additional 401(k) plan is estimated to increase the termination probability of the accompanying DB plan by 15.5 percentage points relative to a DB plan of a sponsor with no 401(k) or DC plan. This 15.5 percentage point increase in the probability of termination is large relative to the unconditional termination probability of 35 percent. Maintaining the 401(k) plan increases the termination probability by 12.8 percentage points. These two estimates are highly statistically significant. The effect of losing a 401(k) is also estimated to increase the termination probability, but the estimated effect is small and measured imprecisely.

Without any controls in the regression (column one), the termination probability for the base case sponsor with no 401(k) or other DC plan increases from 0.17 to about 0.35 when a new 401(k) is offered. Thus, relative to this base case, the addition of a 401(k) doubles the DB termination probability.

Adding a DC plan reduces the likelihood of the DB plan existing by 24.0 percentage points, and maintaining a DC plan increases the termination probability by 18.6 percentage points. Losing a DC plan is associated with an increased DB termination probability of 12.4 percentage points as in the sponsor-level estimates. These effects are also highly statistically significant.

To summarize, it appears that 401(k) plans and DC plans are substitutes for DB plans over this period. Adding a 401(k) plan essentially doubles the DB termination probability. Losing a 401(k) plan has no effect on the DB plan. Adding (or maintaining) a non-401(k) DC plan also increases DB termination probabilities. But, if a sponsor terminates a DC plan, it is also more likely to terminate the accompanying DB plan. Losses of non-401(k) DC and DB plans move together.

 Back To Top

Changes In DB Participation

Given the legal complications of plan termination, substitution between a defined benefit and 401(k) plan may be a more subtle process. If employers are gradually replacing defined benefit coverage with 401(k) coverage, defined benefit participation may decline over time as employers offer only the 401(k) plan to new employees.(6) Table 13 presents estimates of the effect of changes in 401(k) and DC plan status on participation in the accompanying DB plan.

The estimates in the second column of Table 13 indicate that adding a 401(k) reduces DB plan participation by 105.5 employees. This is 18.25 percent of the average participation in a 1985 DB plan (578 participants), but is half of the participants in the median plan (211 participants). This estimate is statistically significant at the five percent level. The estimated effect of maintaining a 401(k) plan is small (a loss of 35 DB participants) and is statistically insignificant. Losing a 401(k) is estimated to also reduce DB participation (by 111.7 participants) but is also imprecisely measured.

Changes in DC status have a larger effect on DB participation and all three categories are statistically significant at the five percent level. Offering a new DC plan reduces DB participation by 146.0 employees and maintaining a DC over the period reduces DB participation by 288.9 employees. As in the regressions where a loss of a DC plan is associated with a loss of a DB plan, losing DC participants also is associated with losing DB participants (166.9).

When other covariates are added, however, only the coefficients on adding and maintaining a DC plan maintain their economic and statistical significance. Offering a new DC reduces DB participation by 100.4 employees, and maintaining the DC reduces DB participation by 250.4 employees. Overall, the r-squares in these regressions are quite small, indicating that not much variation in DB participation is being explained by changes in 401(k) or DC plan status.

 Back To Top

Changes In The DB Funding Ratio

Table 14 presents estimates of the effect of 401(k) and DC plan status on the DB funding ratio between 1985 and 1991. (See the Data Appendix for a discussion of the funding ratio calculation.) The funding ratio is likely to fall over this period due to OBRA87 funding limits regardless of other sponsor activity. However, changes in 401(k) or DC plan offerings have virtually no economically or statistically significant estimated effect on DB funding ratios. The estimates in column four include industry dummies, and indicate that a 10 percent increase in 1985 firm employment increases the funding ratio by .0021, and a 10 percent increase in plan assets in 1985 reduces the funding ratio by .0015.

 Back To Top

Data Appendix

This Data Appendix addresses two issues: alternative measures of participation from the Form 5500 and construction of DB funding ratios.

 Back To Top

Measures Of Participation And Eligibility

The 1985 Form 5500 includes two questions that are used to measure participation: 7a4 and 17h. Question 7 asks about participants in the plan, and question 17 and its subparts ask about employees of the employer who participate in the plan. The pages from the Form 5500 containing these questions follow.

Question 17 includes subparts that record employee eligibility (17f) and total employees (17b). The ratio of these two is a measure of coverage. In 1985, questions 7 and 17 are completed for most plans.

The 1992 Form 5500 includes an additional question that can be used to measure participation. As in the 1985 form, question 7a4 is included, but in addition, question 7g asks for the number of participants in DC plans with account balances. Participants are considered to have an account balance if they have made a contribution, or the employer made a contribution in their name, in this year or any previous year.

Question 21 on the 1992 Form 5500 is the one most comparable to question 17 on the 1985 questionnaire, but it is expanded considerably to include questions about anti-discrimination requirements. The pages from the Form 5500 containing these questions follow.

Question 17m can be used to measure participation, question 21j can measure eligibility, and with 21h (total employees) a measure of coverage can be constructed. However, two difficulties arise in using question 21 and its subparts in 1992. First, some of the subparts are missing for a large number of plans. Second, as questions 21a-21e make clear, the responses to the remaining subparts of question 21 may not be accurate for the plan in question if the sponsor has temporarily restructured it to meet anti-discrimination requirements.

Use of allowable restructuring techniques to pass anti-discrimination tests is fairly common. Of the 2,694 employers that sponsor at least one 401(k) plan in 1992, 52.23 percent (1,407) report either (1) applying the separate line of business rules of Code section 414(r) when testing the plan for coverage and discrimination tests (question 21a); (2) applying the mandatory disaggregation rules under Income Tax Regulations section 1.410(b)-7(c) (question 21c); (3) aggregating plans to test whether the plan satisfies the coverage and discrimination tests (question 21d); or (4) restructuring the plan into component plans to satisfy the coverage and discrimination tests (question 21e). Of the 7,017 real DB plans that report question 21, over 17 percent of them report using these techniques.

Thus, for these plans, the remaining information on participation (question 21m) and eligibility (question 21j) may not apply solely to the plan in question, but to some aggregation of plans of the employer. These responses will differ from question 7a4 and may be less accurate indicators of plan participation. Unfortunately, question 21 is the only source of questions concerning eligibility and therefore coverage on the Form 5500.

In this paper, question 7 is used as the primary measure of participation since it is asked in the same way in both years. Eligibility and coverage numbers could be used for the plans whose sponsors do not use aggregation techniques, but these characteristics are not the focus of this paper.

 Back To Top

Calculation Of The Funding Ratio

The funding ratio is constructed as the ratio of the accumulated benefit obligation (or termination of plan liability) divided by the market value of plan assets at the beginning of the plan year. The accumulated benefit obligation (ABO) is the present value of pension benefits owed to employees under the plan's benefit formula absent any salary projections and at a nominal rate of interest. These data are reported on Schedule B of Form 5500.

Since the choice of the discount rate used to value plan liabilities is negatively related to a sponsor's profitability, in constructing the funding ratio, I adjust the reported liabilities to a common discount rate used by the Pension Benefit Guaranty Corporation to value immediate annuities in non-multiemployer plans (see Bodie and Papke (1990)).

The formulas used to calculate the funding ratio in the two years are given below. They differ slightly across the years since not all the Schedule B information is included in the 1992 Form 5500 data file. Funding ratios greater than 3 in either year are set equal to missing.

 Back To Top

Calculation Of The 1985 Funding Ratio

i1 = (.94)^(9.57-SchB12c)

i2 = ((100 + SchB12c)/(100 + 9.57))^(15)

pvbr = SchB6d1*i1 (retired liabilities)

pvbnr = (SchB6d2 + SchB6e + (.8*SchB9b) )*i1*i2 (not retired liabilities)

abo = pvbr+pvbnr

fratio=assetboy/abo

 Back To Top

Calculation Of The 1992 Funding Ratio

i1 = (.94)^(6.1875-SchB12c1)

i2 = ((100 + SchB12c1)/(100 + 6.1975))^(15)

pvbr=SchB6di3*i1

pvbnr=(SchB6dii3 + SchB6diii3 + (.8*SchB9b))*i1*i2

abo = pvbr + pvbnr

fratio=assetboy/abo

 Back To Top

Table 1

Single Employer Pension Plans with 100 or More Participants
By Type of Plan, 1980-1992

Year

Number of Plans

Number of Active Participants (thousands)

 

DBs

DCs

401(k)s(1)

DBs

DCs

401(k)s(1)

1980

22,424

12,978

N.A.

21,889

14,798

N.A.

1981

22,877

14,803

N.A.

21,749

16,090

N.A.

1982

23,271

16,247

N.A.

21,664

18,233

N.A.

1983

23,888

18,087

N.A.

22,025

22,520

N.A.

1984

23,033

19,731

17,303

22,484

24,984

7,540

1985

22,619

23,279

29,869

21,591

26,973

10,339

1986

22,426

25,713

37,420

21,427

27,723

11,559

1987

21,432

27,556

45,054

21,504

27,563

13,131

1988

20,870

30,661

68,121

21,348

26,417

15,203(2)

1989

19,531

33,275

83,301

20,595

25,683

17,337(2)

1990

17,521

33,035

97,614

20,228

27,252

19,548

1991

16,750

34,368

111,394

19,726

27,155

19,126

1992

17,019

38,283

139,704

19,752

29,497

22,404

  1. 401(k) plans include all plan sizes and entity types.

  2. Includes some employees who are eligible to participate but have not elected to join.

Source: U.S. Department of Labor, Pension and Welfare Benefits Administration, "Private Pension Plan Bulletin: Abstract of 1992 Form 5500 Annual Reports," Number 5, Winter 1996, Tables E3, E10, and E23.

 Back To Top

Table 2

Summary Statistics for Sample of Primary Defined Benefit Plans, Plan Year 1985

 

Mean & Std Deviation

Median

Obs.

Active participants (Q7a4)

578.05 (4283.95)

211

15,589

Employees participating (Q17h)

772.32 (5,119.05)

216

12,841

Employees eligible (Q17f)

804.17 (5,265.13)

220

12,529

Total employment (Q17b)

11,536.71 (475,230.90)

686

15,476

Actives participation ratio (Q7a4/Q17f)

0.955 (0.152)

1.000

10,920

Employee Participation ratio (Q17h/Q17f)

0.989 (0.069)

1.000

12,468

Coverage ratio (Q17f/Q17b)

0.570 (0.317)

0.650

12,436

Existing 401(k)?

0.170 (0.376)

--

16,597

Terminated or intend to terminate?

0.088 (0.283)

--

16,597

Consolidated or merged?

0.034 (0.180)

--

16,597

Funding ratio

1.63 (0.62)

1.62

9,137

Plan age

18.07 (11.24)

17

16,456

Plan assets ($millions)

12.4 (179.0)

2.33

15,662

Note: Q7a4 refers to question 7, part a4 on the 1985 Form 5500 and other data items are similarly identified. Standard errors are in parentheses.

 Back To Top

Table 3

Summary Statistics for Accompanying 401(k) Plans,  Plan Year 1985

 

Mean & Std Deviation

Median

Obs.

Active participants (Q7a4)

1067.46 (4813.37)

281.5

1,574

Employees participating (Q17h)

1,017.23 (4,930.88)

271

1,424

Employees eligible (Q17f)

1,572.21 (14,223.27)

326

1,387

Total employment (Q17b)

4,392.72 (21,858.39)

650

1,543

Actives participation ratio (Q7a4/Q17f)

0.833 (0.184)

0.882

1,280

Employee Participation ratio (Q17h/Q17f)

0.844 (0.181)

0.897

1,381

Coverage ratio (Q17f/Q17b)

0.637 (0.272)

0.692

1,383

Plan age

2.84 (2.91)

1.0

1,586

Match rate

0.64 (0.78)

0.39

980

Plan assets ($millions)

11.3 (48.9)

1.88

1,508

 Back To Top

Table 4

Summary Statistics for Sample of Primary Defined Benefit Plans, Plan Year 1992

 

Mean & Std Deviation

Median

Obs.

Active participants (Q7a4)

842.80 (4,635.84)

257

7,004

Employees benefiting (Q21m)

475.38 (3,600.00)

0.0

5,573

Employees eligible (Q21j)

1,287.04 (5,457.09

168

6,225

Total employment (Q21h)

2,270.62 (10,740.18)

241

6,260

Actives participation ratio (Q7a4/Q21j)

0.805 (0.308)

0.976

3,244

Employee Participation ratio (Q21m/Q21j)

0.875 (0.258)

1.000

3,898

Coverage ratio (Q21j/Q21h)

0.707 (0.232)

0.766

3,861

Existing 401(k)?

0.389 (0.488)

--

7,017

Terminated or intend to terminate?

0.052 (0.221)

--

7,017

Consolidated or merged?

0.038 (0.191)

--

7,017

Funding ratio

1.01 (0.38)

0.96

5,938

Plan age

24.09 (11.21)

23

2,976

Plan assets ($millions)

32.6 (414.0)

5.25

7,017

Note: Q7a4 refers to question 7, part a4 on the 1992 Form 5500. Standard errors are in parentheses. These calculations do not include phantom DB plans.

 Back To Top

Table 5

Summary Statistics for Accompanying 401(k) Plans, Plan Year 1992

 

Mean & Std Deviation

Median

Obs.

Active participants (Q7a4)

1,592.30 (8,862.04)

353

2,633

Employees benefiting (Q21m)

1,502.17 (5,025.1)

356.5

2,362

Employees eligible (Q21j)

2,584.03 (11,133.69)

395

2,379

Total employment (Q21h)

5,306.11 (33,365.35)

643

2,384

Actives participation ratio (Q7a4/Q21j)

0.819 (0.244)

0.923

1,994

Employee participation ratio (Q21m/Q21j)

0.906 (0.211)

1.000

2,358

Coverage ratio (Q21m/Q21h)

0.703 (0.236)

0.767

2,378

Average firm match rate

0.40 (0.37)

0.33

2,054

Maximum match rate

0.42 (0.38)

0.34

2,145

Plan assets ($millions)

41.7 (251.0)

4.71

2,637

Note: Q7a4 refers to question 7, part a4 of the 1992 Form 5500. Standard errors are in parentheses.

 Back To Top

Table 6

Linear Probability Model - Dependent Variable: Firm Attrition in 1992 - Plan Year 1985

Number of DB plans

-.019 (.006)

-.021 (.007)

-.030 (.007)

Number of 401(k) plans

-.132 (.013)

-.126 (.014)

-.142 (.014)

Number of DC plans

-.087 (.010)

-.087 (.010)

-.099 (.010)

Log (firm employment)

 

.013 (.003)

.134 (.003)

Industry Dummies?

no

no

yes

Constant

.474 (.009)

.377 (.020)

.263 (.024)

Obs.

11,814

10,992

10,992

R2

.0182

.0177

.0341

 Back To Top

Table 7

Linear Probability Models - Dependent Variable: Plan Attrition - Due to Firm Attrition in 1992 - Plan Year 1985

Number of DB plans

-.021 (.003)

-.023 (.004)

-.031 (.004)

-.029 (.005)

-.024 (.007)

# of 401(k) plans

-.109 (.011)

-.102 (.012)

-.114 (.012)

-.079 (.014)

-.091 (.021)

Number of DC plans

-.069 (.008)

-.072 (.008)

-.078 (.008)

-.070 (.009)

-.081 (.014)

Log (firm  employ)

 

.015 (.003)

.015 (.003)

.023 (.004)

.039 (.006)

Industry dummies?

no

no

yes

yes

yes

Participation Ratio

 

 

 

-.176 (.034)

-.100 (.052)

Log (plan assets)

 

 

 

-.036 (.004)

-.039 (.007)

Age of plan

 

 

 

-.001 (.0004)

-.002 (.001)

Plan to terminate?

 

 

 

.182 (.024)

.183 (.038)

Retirees/Actives

 

 

 

 

.062 (.017)

Adm cost/part ($)

 

 

 

 

.0001 (.0001)

Constant

.469 (.007)

.360 (.017)

.252 (.021)

.907 (.062)

.749 (.099)

Obs.

14,862

13,843

13,843

9,565

4,671

R2

.0195

.0184

.0321

.0528

.0522

 Back To Top

Table 8

Summary Statistics for Variables in Linear Probability Attrition Models - Plan Year 1985

 

Plan Level

Firm Level

Means & Std. Dev

Obs.

Means & Std. Dev

Obs.

DB absent in 1992

.383 (.486)

16,583

.413 (.492)

11,948

Number of DB plans

1.700 (1.284)

14,911

1.261 (.743)

11,828

# of 401(k) plans

.170 (.375)

16,583

.133 (.339)

11,948

Number of DC plans

.298 (.622)

16,223

.202 (.484)

11,917

Log (firm employ)

6.944 (1.865)

15,462

6.459 (1.612)

11,118

Particip. ratio

.955 (.152)

10,915

 

 

Log (plan assets)

14.666 (1.694)

15,514

 

 

Age of plan

18.072 (11.242)

16,442

 

 

Plan to terminate?

.088 (.283)

16,583

 

 

Retirees/Actives

.415 (.511)

11,957

 

 

Adm Costs/Part. ($)

67.135 (65.952)

9,145

 

 

 Back To Top

Table 9

OLS Regression - Dependent Variable: Change in the Number of DB Plans - Offered by the Firm Between 1985 and 1992

Change in the # of 401(k) plans

-.210 (.012)

-.205 (.012)

-.184 (.012)

Change in the # of DC plans

-.139 (.013)

-.146 (.013)

-.152 (.013)

Log (firm employment in 1985)

 

-.062 (.005)

-.062 (.005)

Industry dummies?

no

no

yes

Constant

-.285 (.009)

.128 (.036)

.217 (.041)

Obs.

6,854

6,554

6,554

R2

.0462

.0681

.0931

 Back To Top

Table 10

OLS Regression - Dependent Variable: Change in the Number of DB Plans - Offered by the Firm Between 1985 and 1992

Offers at least 1 new 401(k) plan

-.346 (.018)

-.336 (.018)

-.312 (.018)

Maintains its 401(k) plan

-.250 (.026)

-.219 (.026)

-.213 (.027)

Loses 1 401(k) plan

-.040 (.037)

-.010 (.037)

.005 (.037)

Offers at least 1 new DC plan

-.423 (.021)

-.403 (.021)

-.410 (.021)

Maintains its DC plan(s)

-.312 (.027)

-.269 (.027)

-.264 (.027)

Loses at least 1 DC plan

-.193 (.027)

-.148 (.028)

-.134 (.027)

Log (firm employment in 1985)

 

-.050 (.005)

-.050 (.005)

Industry dummies?

no

no

yes

Constant

-.113 (.011)

.206 (.035)

.233 (.040)

Obs.

6,854

6,554

6,554

R2

.1179

.1274

.1441

Note: The omitted category is the firm offers no 401(k) plan or other DC plan.

 Back To Top

Table 11

Linear Probability Model - Dependent Variable: Is the DB Plan Still Offered in 1992?

Change # of 401(k) plans, '85-'92

-.064 (.005)

-.049 (.005)

-.044 (.005)

Change # of DC plans, '85-'92

-.046 (.007)

-.041 (.007)

-.042 (.007)

Log (firm employment in 1985)

 

-.038 (.003)

-.037 (.003)

Log (plan assets in 1985)

 

.054 (.003)

.052 (.003)

Industry dummies?

no

no

yes

Constant

.687 (.005)

.171 (.045)

.253 (.048)

Obs.

9,693

8,819

8,819

R2

.0183

.0586

.0739

The mean of the dependent variable is 0.654 with a standard error of 0.476.

 Back To Top

Table 12

Linear Probability - Dependent Variable: Is the DB Plan Still Offered in 1992?

Offers at least 1 new 401(k) plan

-.183 (.011)

-.166 (.011)

-.155 (.011)

Maintains its 401(k) plan

-.152 (.016)

-.141 (.016)

-.128 (.016)

Loses 1 401(k) plan

-.032 (.021)

-.037 (.021)

-.027 (.021)

Offers at least 1 new DC plan

-.265 (.012)

-.246 (.013)

-.240 (.013)

Maintains its DC plan(s)

-.224 (.016)

-.191 (.016)

-.186 (.016)

Loses at least 1 DC plan

-.138 (.015)

-.128 (.015)

-.124 (.015)

Log (firm employ in 1985)

 

-.025 (.003)

-.026 (.003)

Log (plan assets in 1985)

 

.057 (.003)

.056 (.003)

Industry dummies?

no

no

yes

Constant

.830 (.007)

.166 (.043)

.197 (.047)

Obs.

9,693

8,819

8,819

R2

.0918

.1217

.1276

Note: The omitted category is the firm offers no 401(k) or other DC plan. The mean of the dependent variable is 0.654 with a standard error of 0.476.

 Back To Top

Table 13

OLS Regression - Dependent Variable: Change in DB Participation Between 1985 and 1992

Offers at least 1 new 401(k) plan

-105.537 (50.561)

-81.950 (54.015)

-44.545 (56.142)

Maintains its 401(k) plan

-35.103 (75.237)

24.884 (80.400)

62.588 (81.742)

Loses 1 401(k) plan

-111.719 (99.120)

-38.926 (105.908)

4.491 (106.699)

Offers at least 1 new DC plan

-146.012 (59.380)

-108.826 (63.863)

-100.392 (64.133)

Maintains its DC plan(s)

-288.920 (76.505)

-253.886 (81.461)

-250.409 (81.837)

Loses at least 1 DC plan

-166.851 (70.275)

-84.600 (75.564)

-53.420 (75.943)

Log (firm employ in 1985)

 

-51.652 (14.807)

-41.768 (15.233)

Log (plan assets in 1985)

 

-84.445 (14.874)

-98.133 (15.415)

Industry dummies?

no

no

yes

Constant

61.101 (34.732)

1631.053 (219.160)

1768.62 (237.406)

Obs.

9,262

8,658

8,658

R2

.0031

.0095

.0128

Note: The omitted category is the firm offers no 401(k) or other DC plan. The mean of the dependent variable is -60.731 with a standard deviation of 2,165.519.

 Back To Top

Table 14

OLS Regression Dependent Variable: Change in DB Funding Ratio Between 1985 and 1992

Offers at least 1 new 401(k) plan

.050 (.021)

.042 (.021)

-.017 (.022)

Maintains its 401(k) plan

.047 (.032)

.054 (.033)

.003 (.033)

Loses 1 401(k) plan

-.016 (.042)

-.017 (.043)

-.059 (.042)

Offers at least 1 new DC plan

.104 (.027)

.093 (.028)

.061 (.027)

Maintains its DC plan(s)

.046 (.032)

.042 (.033)

-.00002 (.03216)

Loses at least 1 DC plan

.013 (.030)

.016 (.031)

-.029 (.030)

Log (firm employ in 1985)

 

.032 (.006)

.021 (.006)

Log (plan assets in 1985)

 

-.032 (.006)

-.015 (.007)

Industry dummies?

no

no

yes

Constant

-.699 (.013)

-.426 (.093)

-.623 (.100)

Obs.

3,707

3,631

3,631

R2

.0065

.0176

.0684

Note: The omitted category is the firm offers no 401(k) or other DC plan.

 Back To Top

References

  • Andrews, E.S. "The Growth and Distribution of 401(k) Plans," in Trends in Pensions 1992, J.A. Turner and D.J. Beller, eds., U.S. Department of Labor, Pension and Welfare Benefits Administration, 1992, 149-176.

  • Bodie, Z., and L.E. Papke, Report to the U.S. Department of Labor, Contract J-9-P-0097, 1990.

  • Clark, R.L., and A.A. McDermed, The Choice of Pension Plans in a Changing Regulatory Environment, Washington D.C., AEI Press, 1990.

  • Engen, E., W.G. Gale, and J.K. Scholz, "Do Savings Incentives Work?" Brookings Papers on Economic Activity, Spring 1994, 85-150.

  • Even, W.E., and D.A. Macpherson, "Why Did Male Pension Coverage Decline in the 1980s?," mimeo, Miami University, March 1993.

  • Gustman, A.L., and T.L. Steinmeier, "The Stampede Toward Defined Contribution Pension Plans: Fact or Fiction?" Industrial Relations, Volume 31, No. 2, Spring 1992, 361-369.

  • Kruse, D.L., "Pension Substitution in the 1980s: Why the Shift Toward Defined Contribution Plans?" Industrial Relations, Volume 34, No. 2, April 1995, 218-241.

  • Newsweek, "Retirement Blues: How Safe is Your Pension?" December 6, 1993, pp. 38-40.

  • Papke, L.E., "Does 401(k) Introduction Affect Defined Benefit Plans?" National Tax Association Proceedings of the Eighty-Sixth Annual Conference, 1994, 122-123.

  • Papke, L.E. "Participation in and Contributions to 401(k) Pension Plans: Evidence from Plan Data," Journal of Human Resources, Volume 30, No. 2, Spring 1995, 311-325.

  • Papke, L.E., M.A. Petersen, and J.M. Poterba, "Do 401(k) Plans Replace Other Employer Provided Pensions?" in Advances in the Economics of Aging, D. Wise, ed. NBER, University of Chicago Press, 1996, 219-239.

  • Petersen, M.A., "Pension Reversions and Worker-Stockholder Wealth Transfers," Quarterly Journal of Economics, August 1992, 1033-1056.

  • Poterba, J.M., S.F. Venti, and D.A. Wise, "Do 401(k) Contributions Crowd Out Other Personal Saving?" Journal of Public Economics, Volume 58, 1995, 1-32.

  • U.S. Department of Labor, Pension and Welfare Benefits Administration, Office of Research and Economic Development, "Private Pension Plan Bulletin: Abstract of 1992 Form 5500 Annual Reports," Number 5, Winter 1996.

 Back To Top

Endnotes

  1. See the Employee Benefit Research Institute, 1994, for eligibility and participation rates calculated from the April 1993 Current Population Survey.

  2. See Papke (1995) for a detailed discussion of contribution limits and nondiscrimination requirements.

  3. Between 1980 and 1985, the active participant count in the larger DB plans is relatively stable (21.9 million in 1980 to 21.4 million in 1985), while DC participation increases from 14.8 million to 27.7 million. But from 1985 to 1992, active participation in DB plans falls from 21.6 million to 19.8 million, and participation rises in DC plans from 27 million to 29.5 million. The story is similar if all sizes of single employer plans are considered.

  4. The plan's match rate is calculated as the ratio of employer to employee contributions. See Papke (1995) for a discussion.

  5. If Question 21k ("Do 100 percent of the nonexcludable employees entered on line 21j benefit under the plan?") is answered affirmatively, then question 21m is not completed on the form. The response to question 21j ("Enter the number of nonexcludable employees") is substituted in the data.

  6. Since participation in an available DB plan is usually mandatory, the participation ratio (active participants divided by employees eligible) may not reflect this shift. It is more likely that the coverage ratio (eligibles divided by total employees) will reflect this subtle shift. Unfortunately, as explained in the Data Appendix, the coverage ratio is not available for 45 percent of plans in the 1992 sample.

About EBSA


Laws & Regulations


Compliance Assistance


ERISA Enforcement


Newsroom


Consumer Information


Frequently Asked Questions


Publications/Reports


Forms/Doc Requests


Programs/Initiatives


Related Resources


Contact Us



Phone Numbers