Multiemployer Defined Benefit Pension Plans: Employer Withdrawals and Financial Vulnerability

Abstract

Multiemployer defined benefit pension plans are facing severe funding challenges. The Pension Protection Act of 2006 requires that a multiemployer pension plan with an actuarial funded percentage of less than 80% must take corrective actions to improve its financial health. We use a regression discontinuity design to examine the effect of funding rule requirements on employer withdrawals from multiemployer pension plans. We find that multiemployer pension plans subject to funding rule requirements are about 14 percentage points more likely to experience employer withdrawals in a 1-year period compared to plans not required to take any corrective actions. We also find that plans with ex ante employer withdrawal experiences are more vulnerable to financial shocks such as the 2008 financial crisis. Our study provides important policy implications for regulators concerning best practices to build pension plan resilience to insolvency events.

ACKNOWLEDGMENTS

The authors are grateful to two anonymous referees’ valuable suggestions that improved the article. We also thank Ben Collier, Rhiannon Jerch, Martin Grace, Lisa Posey, and participants at the 2021 American Risk and Insurance Association (ARIA) Conference, the 24th International Congress on Insurance: Mathematics and Economics, and the Brownbag Seminar of Temple University for their helpful comments.

Discussions on this article can be submitted until October 1, 2023. The authors reserve the right to reply to any discussion. Please see the Instructions for Authors found online at http://www.tandfonline.com/uaaj for submission instructions.

Notes

1 DB pension plans sponsored by a single firm are commonly known as single-employer pension plans. They cover over 23 million participants (PBGC 2021). The financial outlook of single-employer plans is drastically different from that of multiemployer plans. For example, we show in Appendix Figure A.1 that the average funding ratio of underfunded single-employer plans has stayed around 80% since 1994, whereas that of underfunded multiemployer pension plans consistently declined from around 80% to 42% as of 2018. In addition, the federal backstop for single-employer DB plans, the PBGC Single-Employer Insurance Program, remains financially healthy and continues to improve (PBGC, 2020).

2 Several reports have already discussed important factors associated with the funding issues of multiemployer pension plans. For example, Munnell , Aubry, and Crawford (2017) studies the characteristics of plans that are projected to run out of money within the next 10 to 15 years. Munnell et al. (2019) assessed the effectiveness of using benefit cuts as a tool to forestall plan insolvency.

3 Prior to PPA, the Multiemployer Pension Plan Amendments Act of 1980 (MPPA) created a “reorganization” index to identify financially troubled multiemployer DB pension plans. These plans are subject to higher minimum funding standards and required to take actions (e.g., by increasing contribution) to improve their financial conditions. However, the index was rarely triggered. Only four plans were reported to be in reorganization status in the 2009 plan year, whereas 934 plans were identified as in the yellow/red zone in the same plan year under the new PPA funding rules (PBGC, 2013).

4 A funding deficiency occurs when a plan’s total credits (i.e., employer contributions and investment gains) are less than the total charges (i.e., benefit costs earned by participants and investment losses) during a year (Topoleski, 2020).

5 In Appendix Figure A.2, we plot benefit reductions (versus funding status in Figure 1) as a function of a plan’s funded percentage. The y-axis shows the fraction of plans that had benefit reductions in a 1-year period within each 0.5 percentage point bin of funded percentage. We do not observe a clear discontinuity at the 80% funded percentage, which implies that plans around the cutoff do not significantly differ from each other when it comes to benefit reductions. The information regarding the degree of any benefit changes and whether and to what extent a plan had any contribution increases is not reported in the data.

6 There are two types of withdrawals in a multiemployer plan: complete or partial. We study complete withdrawals, because Schedule R does not record partial withdrawals (Internal Revenue Service, 2020). More information about complete and partial withdrawals can be found at https://www.pbgc.gov/prac/multiemployer/withdrawal-liability.

7 To ensure that the filtering does not generate bias, in Appendix Table A.1 we compare plan/industry characteristics of 2008 green zone plans before and after filtering. None of the variables are significantly different.

8 We also conduct robustness check by using both a linear functional form and nonparametric local linear regressions to the samples. Our results remain quantitatively similar (see “Robustness Tests” under Subsection 3.3).

9 We run the following regression: $y_i=\alpha I(\text{FP}<0.8)+{\theta }_1({\text{FP}}_i-0.8)+{\theta }_2({\text{FP}}_i-0.8)\times I({\text{FP}}_i<0.8)+{\theta }_3{({\text{FP}}_i-0.8)}^2+{\theta }_4{({\text{FP}}_i-0.8)}^2\times I({\text{FP}}_i<0.8)+{\nu }_i,$ where y_i represents each covariate and FP represents a plan’s actuarial funded percentage. Regressions are clustered by industry. For a direct comparison, the right-hand side of this equation is the same as that of Equation (3)(3) $$\begin{array}{c}f({\text{FP}}_i)={\psi }_1({\text{FP}}_i-0.8)+{\psi }_2({\text{FP}}_i-0.8)\times I({\text{FP}}_i<0.8)\\ +{\psi }_3{({\text{FP}}_i-0.8)}^2+{\psi }_4{({\text{FP}}_i-0.8)}^2\times I({\text{FP}}_i<0.8);\end{array}$$(3) .

10 We do find that the discontinuity estimators on the 2010 dependency ratio and the percentage of union covered employment, respectively, are marginally significant. However, it is possible that the observed significance is random due to multiple testing problems (Lee and Lemieux 2010). To address this, we conduct a seemingly unrelated regression (SUR) by combining each equation with different baseline covariates as dependent variables and perform a single ${\chi }^2$ test for all αs being zeros. For both the 2009 and 2010 plan years, we fail to reject the null hypothesis that all covariates evolve continuously across the cutoff.

11 In Appendix Table A.3, we also explore the relationship between a plan’s non-green zone certification and withdrawal intensity in a 1-year period. We use two measures of withdrawal intensity: the number of withdrawals per plan per year and the share of contributing employers withdrew from a plan per year (i.e., the ratio of the number of withdrawals per fund per year to the number of participating employers). We do not find significant effects of non-green zone certification on withdrawal intensity with/without controls. It is likely that withdrawal intensity is more closely akin to other factors such as participating employers’ individual characteristics, which we do not observe in our data.

12 In Appendix Table A.6, we also explore whether the funding rule effect on employer withdrawal frequency differs by other factors such as a plan size, age, etc. However, none of these factors appear statistically significant.

13 We manually reviewed the proposed schedules under a funding improvement plan or a rehabilitation plan of 2010 non-green zone plans. On average, the maximum proposed contribution increase is 91% for non-green zone plans with a dependency ratio in the top quintile ($\ge$2.3) and 43% for plans with a dependency ratio <2.3. The difference is statistically significant at a 5% confidence level.

14 Current funded percentage is defined as the market value of assets divided by the current liability. The interest rate used to discount the current liability is based on 30-year Treasury securities (instead of a plan’s investment rate of return).