Technical writing
The graveyard of pensions: using PBGC data to track terminated defined-benefit plans
Since 1975, the Pension Benefit Guaranty Corporation has taken over more than 5,000 private-sector defined-benefit pension plans whose sponsors could not or would not honor their obligations. Every one of those trusteeships is a matter of public record. The PBGC publishes the plan name, the employer, the termination date, the number of workers affected, the assets in the plan at the time of termination, and the shortfall the agency had to cover. The database is a comprehensive ledger of broken pension promises — and almost nobody uses it.
This article covers what the PBGC is, why defined-benefit plans terminate, the full data structure of the PBGC termination records, the largest single-employer terminations in the dataset, how the maximum guarantee works and who it fails, where to access the data, three concrete research use cases, the separate multiemployer crisis, and how to cross-reference PBGC data against DOL Form 5500 filings, SEC pension footnotes, and bankruptcy court records.
What the PBGC insures
The Pension Benefit Guaranty Corporation is a federal agency and government corporation established by the Employee Retirement Income Security Act of 1974 (ERISA). It insures private-sector defined-benefit pension plans — plans that promise a specific monthly benefit at retirement based on salary history and years of service. The PBGC does not insure defined-contribution plans such as 401(k) or 403(b) plans, where the employer contributes to individual accounts but makes no promise about the ultimate benefit. The PBGC's insurance covers only the traditional pension: the guaranteed monthly check.
The PBGC operates two distinct insurance programs with different funding structures, coverage rules, and financial conditions. The single-employer program covers plans sponsored by one company or a controlled group of related companies. It is funded by insurance premiums paid by plan sponsors plus the assets of terminated plans it takes over. When the PBGC trustees a single-employer plan, it becomes the plan's administrator, collects the plan assets, and pays benefits to retirees and vested participants up to the statutory maximum guarantee. The single-employer program has generally been financially self-sufficient, though large industrial bankruptcies have strained its reserves repeatedly.
The multiemployer program covers plans maintained by groups of employers under collective bargaining agreements — typically within a single industry and union. These plans pool contributions from multiple employers, so the plan continues even when individual employers exit the industry. The multiemployer program operates differently: the PBGC does not trustee failing multiemployer plans the way it does single-employer plans. Instead, it provides financial assistance to multiemployer plans that become insolvent. Multiemployer coverage limits are far lower — historically around $35 per month per year of service — meaning a worker with 30 years of service could receive at most roughly $12,600 per year from the PBGC if their multiemployer plan failed. The multiemployer program was technically insolvent before Congress acted, a crisis addressed below.
Why plans terminate
Single-employer defined-benefit plans terminate through three distinct legal pathways, each with different triggers, procedural requirements, and implications for workers.
A standard termination occurs when a plan sponsor voluntarily closes a plan that is fully funded — meaning plan assets are sufficient to cover all benefit liabilities. The sponsor purchases annuities from an insurance company to cover all participant benefits, the plan winds down in an orderly fashion, and the PBGC's insurance is never triggered. Standard terminations appear in PBGC records but do not represent insurance claims: no worker loses benefits, and the agency pays nothing. They are nonetheless worth tracking in the data because they mark the voluntary contraction of the defined-benefit system — healthy plans being closed by sponsors who no longer wish to bear the funding and regulatory burden of a pension.
A distress termination is the mechanism through which an employer in financial difficulty — typically in bankruptcy or demonstrably unable to remain in business while continuing plan contributions — terminates a plan that is underfunded. To qualify for a distress termination, the plan sponsor must meet one of four statutory distress criteria: liquidation in bankruptcy, reorganization in bankruptcy where the court determines the plan is unaffordable, inability to continue operations without terminating the plan, or inability to pay plan contributions without unreasonable burden. When a distress termination is approved, the PBGC becomes trustee, takes the plan's assets, assumes the liability for benefit payments, and pursues the sponsor as a creditor in bankruptcy for the unfunded benefit liabilities.
A PBGC-initiated termination occurs when the PBGC itself acts to terminate a plan without the sponsor's consent. The agency has authority under ERISA to initiate termination when a plan's continued operation poses an unreasonable risk of loss to the insurance program — typically when the plan is severely underfunded, contributions have stopped, and the sponsor shows signs of impending financial distress. PBGC-initiated terminations are less common than distress terminations but represent the agency's most aggressive use of its authority to protect the insurance fund before conditions deteriorate further.
Data structure
The PBGC's pension insurance data tables, published at pbgc.gov/prac/data-tables, contain a record for every plan that the PBGC has trusteed or otherwise resolved since the program's inception. The core fields in the termination dataset are:
plan_name # name of the pension plan
company_name # name of the sponsoring employer
ein # employer identification number (IRS EIN)
pn # plan number (3-digit, from Form 5500 filing)
termination_date # effective date of plan termination
termination_type # S = standard, D = distress, P = PBGC-initiated
state # state of plan sponsor headquarters
naics # NAICS industry code of plan sponsor
participants # number of participants at termination (active + retired)
# Financial condition at termination
plan_assets # fair market value of plan assets at termination date
unfunded_vested_benefits # PBGC's claim: present value of vested benefits
# in excess of plan assets at termination
pbgc_claim # amount the PBGC expects to pay over the plan's life
funding_ratio # plan_assets / total_benefit_liabilities at termination
# Outcome
max_guarantee_flag # whether any participants are above the maximum guarantee
settlement_amount # amount recovered from plan sponsor in bankruptcy
recovery_ratio # settlement_amount / unfunded_vested_benefitsTwo fields require careful interpretation. The unfunded_vested_benefitsfigure is the PBGC's claim against the plan sponsor in bankruptcy — the statutory measure of the plan's shortfall. It is not the same as the total benefit promises made to workers: it covers only vested benefits (those workers have earned the right to receive regardless of future employment), discounted to present value using PBGC's actuarial assumptions. Workers who were not yet vested when the plan terminated receive nothing from either the sponsor or the PBGC. The pbgc_claim field represents the agency's estimate of what it will actually pay out over the remaining lifetimes of participants — a number that may be lower than unfunded vested benefits if the plan was far above the maximum guarantee level, since PBGC only pays up to its statutory cap.
Major terminations in the dataset
The industrial bankruptcies of the 2000s produced the largest single-employer pension terminations in the PBGC's history, and their records remain in the dataset as benchmarks for understanding the scale of what the agency absorbs.
Bethlehem Steel (2002) generated one of the largest single pension terminations on record at the time. When Bethlehem filed for Chapter 11 in 2001 and subsequently liquidated, its pension plans — covering steelworkers across plants in Pennsylvania, Maryland, Indiana, and New York — were terminated with a combined PBGC claim of approximately $3.7 billion. The plans covered roughly 95,000 workers and retirees. Many retired Bethlehem steelworkers who had expected full pensions based on final salaries and decades of service found their monthly payments capped at the PBGC maximum guarantee, which at that time provided substantially less than what their collective bargaining agreements had promised.
United Airlines (2005) produced what was then the largest pension termination in American history. After emerging from Chapter 11 bankruptcy, United terminated four pension plans covering roughly 134,000 current and former employees — pilots, mechanics, flight attendants, and ground workers — with a combined PBGC claim of approximately $3.2 billion. The airline pilots' plan was particularly devastating for individual workers: pilot pensions were based on high final salaries and early mandatory retirement ages, making them especially vulnerable to the maximum guarantee cap, which at the time provided far less than promised benefits for high-salaried pilots who retired before age 65.
Delta and Northwest Airlines (2006–2009)added several billion more in PBGC claims during the airline industry's extended financial distress period. The airline industry as a whole — carriers including US Airways, Aloha Airlines, and Frontier Airlines — generated a concentration of PBGC terminations in a single decade that reflects the structural mismatch between pension obligations negotiated during regulated-industry profitability and the competitive economics of post-deregulation aviation.
Delphi Corporation (2009), the auto parts supplier spun off from General Motors, terminated its pension plans during the financial crisis with a PBGC claim of approximately $6.2 billion — one of the largest single terminations on record. Delphi's collapse illustrates the supply-chain dimension of pension risk: the plans covered workers at a company that was itself a product of corporate restructuring, whose pension obligations were inherited from the original General Motors workforce and then left stranded when Delphi could no longer sustain them.
The maximum guarantee
The PBGC's maximum guaranteed benefit is the most consequential number in the termination dataset for understanding worker outcomes. It is the statutory ceiling on what the PBGC will pay any individual participant regardless of what the plan promised. For 2024, the maximum guaranteed benefit for a single-life annuity beginning at age 65 is $83,400 per year ($6,950 per month). The maximum guarantee is indexed to the national average wage and changes annually.
Three factors reduce the effective guarantee below the headline number for many workers. First, the maximum guarantee applies only to benefits that have been in place for at least five years before the termination date; benefit improvements adopted within five years of termination are only partially guaranteed, phasing in at 20 percent per year. Second, the guarantee is reduced for benefits taken before age 65: a benefit beginning at age 55 is guaranteed at only 45 percent of the age-65 maximum. This reduction hits airline pilots hardest — pilots face mandatory retirement at age 65 and typically retire between 60 and 65 with pension calculations based on final salaries of $200,000 or more. Their promised benefits routinely exceeded the PBGC maximum by multiples, meaning the termination of their plan cut their retirement income by half or more. Third, the maximum guarantee is per participant, not per household: a couple where both spouses participated in the same plan each faces the individual cap.
The dataset's max_guarantee_flag field identifies plans where at least some participants were above the maximum. In practice, virtually all large industrial plan terminations have this flag set: any plan covering professional or skilled-trade workers with long tenure and above-median salaries will have participants whose promised benefits exceeded the cap. The concentration of above-cap workers in airline, steel, and auto supply chain terminations is not coincidental — these were industries with strong unions, defined-benefit plans negotiated when the companies were profitable, and high tenure among the workforce at the time plans failed.
How to access the data
The PBGC publishes its termination and financial data through two primary channels, with a third supplementary source for at-risk plans that have not yet terminated.
The pension insurance data tables at pbgc.gov/prac/data-tables are the primary public interface. The tables are organized by year and by data type: terminated plan records, active trusteed plan inventory, premium payment data, and financial results. The terminated plan table is available as a downloadable Excel or CSV file updated annually. It is the closest thing to a comprehensive registry of every plan the PBGC has taken over. The file includes all the core fields described above, though the exact field names and availability of computed ratios vary by data vintage: older terminations have less complete financial detail than recent ones because actuarial data was not uniformly recorded before the 1990s.
The annual pension insurance data book, published each year by the PBGC's Office of Policy and Research, provides aggregate statistics: total trusteed plans by year, claims by industry, average funding ratios at termination, average participant counts, and the PBGC's net financial position across both programs. The data book is essential context for interpreting the plan-level termination file because it provides the distributions against which individual plan records should be compared. A plan with a 30-percent funding ratio looks different if the annual data book shows the median funding ratio at termination that year was 50 percent versus 20 percent.
ERISA Section 4010 filings cover a separate and important population: plans that have not yet terminated but are at elevated risk. Section 4010 requires controlled groups of companies to report to the PBGC when their pension plans are severely underfunded (funding ratio below 80 percent using PBGC's assumptions) or when the controlled group has failed to make required contributions. These filings are a leading indicator of potential future terminations. The PBGC makes 4010 aggregate data available in the data book and certain plan-level information available through FOIA; the full filing detail is confidential but the existence of a filing is disclosed.
Research use case: industry concentration
The most revealing aggregation in the termination dataset is by industry. Which NAICS sectors have generated the most PBGC claims since 1975? The answer is not evenly distributed: manufacturing, airlines, and retail have generated disproportionate shares of both plan counts and dollar claims.
Manufacturing (NAICS sector 31–33) dominates the dataset by plan count, reflecting the historical prevalence of defined-benefit pensions in unionized industrial employment. Steel, auto, rubber, paper, and other heavy manufacturing sectors drove the largest claims through the 2000s and into the 2010s as the American industrial base contracted. Air transportation (NAICS 481) has a much smaller plan count but an outsized share of total dollar claims, because airline pension plans tended to be large, heavily negotiated, and covering high-earning workers whose promised benefits exceeded the PBGC maximum guarantee at high rates. Retail trade (NAICS sector 44–45) appears with elevated plan counts in more recent years as large retail chains have filed for bankruptcy.
import pandas as pd
df = pd.read_csv("pbgc_trusteed_plans.csv")
# Distress and PBGC-initiated terminations only (not standard)
claims = df[df["termination_type"].isin(["D", "P"])].copy()
# Aggregate by 2-digit NAICS
by_sector = (
claims.groupby("naics_2digit")
.agg(
plan_count=("plan_name", "count"),
total_claim=("pbgc_claim", "sum"),
total_participants=("participants", "sum"),
mean_funding_ratio=("funding_ratio", "mean"),
)
.sort_values("total_claim", ascending=False)
)
print(by_sector.head(15).to_string())
total_claims_b = round(claims["pbgc_claim"].sum() / 1e9, 1)
total_pp = int(claims["participants"].sum())
print(f"\nTotal PBGC claims in dataset: {total_claims_b}B")
print(f"Total participants affected: {total_pp}")Research use case: funding ratio at termination
The funding_ratio field — the ratio of plan assets to total benefit liabilities at the termination date — is the cleanest single measure of how underfunded a plan was when it failed. Across the full dataset of distress and PBGC-initiated terminations, the distribution of funding ratios reveals a consistent pattern: the median funding ratio at termination is substantially lower than the mean, because a small number of extremely underfunded plans (funding ratios below 20 percent) pull the mean down.
A funding ratio of 1.0 (100 percent) would mean a plan had exactly enough assets to cover all benefit promises. In practice, plans that reach distress termination are well below that. The PBGC's own data book reports that the average funding ratio of single-employer plans at distress termination has historically ranged between 40 and 70 percent, depending on the year and the industrial composition of terminations in that year. Plans terminating during the peak of the 2008–2009 financial crisis had particularly low funding ratios because asset values had collapsed at the same time that plan sponsors were under maximum financial stress.
import pandas as pd
import numpy as np
df = pd.read_csv("pbgc_trusteed_plans.csv")
claims = df[df["termination_type"].isin(["D", "P"])].dropna(subset=["funding_ratio"])
def pct(x):
return f"{round(x * 100, 1)}%"
print("Funding ratio at termination (distress + PBGC-initiated):")
print(f" Count: {len(claims)}")
print(f" Mean: {pct(claims['funding_ratio'].mean())}")
print(f" Median: {pct(claims['funding_ratio'].median())}")
print(f" P25: {pct(claims['funding_ratio'].quantile(0.25))}")
print(f" P75: {pct(claims['funding_ratio'].quantile(0.75))}")
print(f" Min: {pct(claims['funding_ratio'].min())}")
# Plans with funding ratio below 20% (severely underfunded)
severe = claims[claims["funding_ratio"] < 0.20]
claim_share = pct(severe["pbgc_claim"].sum() / claims["pbgc_claim"].sum())
pp_share = pct(severe["participants"].sum() / claims["participants"].sum())
print(f"\nPlans with funding ratio below 20%: {len(severe)}")
print(f" Share of total claims: {claim_share}")
print(f" Share of participants: {pp_share}")Research use case: geographic concentration
Plan terminations are geographically concentrated in ways that map onto the historical footprint of American industrial employment. States in the industrial Midwest — Ohio, Michigan, Pennsylvania, Indiana, Illinois — account for a disproportionate share of plan counts and participant counts in the termination dataset. This is not merely a function of population size: controlling for the size of the private-sector workforce, these states show higher termination rates because they hosted the manufacturing and heavy industry sectors that drove the largest waves of distress terminations.
The geographic concentration of terminations has implications for local labor markets that extend beyond individual workers' retirement income. When a large industrial employer in a single-industry town terminates its pension plan in bankruptcy, the workers who lose promised benefits are often also losing their jobs at the same company simultaneously. The PBGC termination record is therefore often the pension dimension of a broader labor market shock that can be cross-referenced against Bureau of Labor Statistics mass layoff data and HUD housing market records for the same geography and time period.
import pandas as pd
df = pd.read_csv("pbgc_trusteed_plans.csv")
claims = df[df["termination_type"].isin(["D", "P"])].copy()
# Terminations by state
by_state = (
claims.groupby("state")
.agg(
plan_count=("plan_name", "count"),
total_participants=("participants", "sum"),
total_claim=("pbgc_claim", "sum"),
)
.sort_values("total_participants", ascending=False)
)
print("Top 15 states by participants in terminated plans:")
print(by_state.head(15).to_string())
# Terminations by decade
claims["decade"] = (pd.to_datetime(claims["termination_date"]).dt.year // 10 * 10)
by_decade = (
claims.groupby("decade")
.agg(plan_count=("plan_name", "count"), total_claim=("pbgc_claim", "sum"))
)
print("\nTerminations by decade:")
print(by_decade.to_string())The multiemployer crisis
The multiemployer program represents a different and more complex dimension of the PBGC's role. Unlike single-employer plans, where one company bears the full funding obligation, multiemployer plans pool contributions from multiple employers under collective bargaining agreements. The Teamsters Central States, Southeast and Southwest Areas Pension Fund is the canonical example: at its peak it covered hundreds of thousands of Teamster members across thousands of trucking and logistics employers. When individual employers left the industry — through deregulation, bankruptcy, or contraction — their share of the funding obligation was supposed to be borne by remaining employers, but if enough employers exited, the remaining employers faced contribution burdens they could not sustain.
The PBGC's multiemployer program was technically insolvent years before Congress intervened. The agency's own financial statements, published annually, showed the multiemployer program's net financial position going deeply negative through the 2010s as dozens of large multiemployer plans projected insolvency within 10 to 20 years. The Multiemployer Pension Reform Act of 2014 (MPRA) gave troubled plans authority to cut benefits to remaining retirees — an authority that had never existed in ERISA — as an alternative to insolvency. The Central States fund applied to cut benefits under MPRA; the Treasury Department rejected the application in 2016 after finding the cuts would not be sufficient to prevent insolvency.
The American Rescue Plan Act of 2021 (ARPA) resolved the immediate multiemployer crisis by appropriating approximately $94 billion for Special Financial Assistance (SFA) payments to the most severely underfunded multiemployer plans. Plans receiving SFA are prohibited from cutting benefits and must use the funds for benefit payments rather than investments. The PBGC published a list of plans that have received SFA approvals — this list is a useful complement to the termination dataset because it identifies multiemployer plans that were on the brink of failure and were rescued rather than terminated, and the relief amounts reflect the scale of their funding shortfalls.
Cross-referencing with other data sources
Three external data sources extend the analytical value of the PBGC termination record in ways that materially change the research questions it can answer.
DOL Form 5500 is the annual financial filing that every pension plan must submit to the Department of Labor, IRS, and PBGC. Form 5500 includes the plan's balance sheet (assets and liabilities), contributions received, benefits paid, actuarial assumptions used for funding calculations, and the actuary's certification of the plan's funded status. For any plan in the PBGC termination dataset, Form 5500 filings from the years before termination provide the pre-termination financial trajectory: when did the funding ratio begin to decline, how large were the contribution shortfalls, and was the plan's actuary flagging going-concern risks? The DOL's EFAST2 system provides Form 5500 filings as bulk downloads at efast.dol.gov; the plan EIN and plan number in the PBGC termination record match the EIN and plan number in Form 5500, enabling exact joins.
SEC 10-K and 10-Q pension footnotes provide the plan sponsor's own financial disclosures about its pension obligations in the years before termination. Public companies are required under ASC 715 (formerly FAS 87/158) to disclose their defined-benefit pension obligation, plan assets, funded status, expected contributions, and actuarial assumptions in financial statement footnotes. For publicly traded companies whose plans appear in the PBGC termination dataset, the SEC EDGAR full-text search system enables retrieval of 10-K filings from the years preceding termination. Comparing the company's reported pension funded status to the funding ratio recorded in the PBGC termination dataset frequently reveals discrepancies: companies used more optimistic actuarial assumptions in their financial statements than the PBGC applied at termination, producing a reported funded status that understated the true shortfall by hundreds of millions of dollars in the large industrial bankruptcies.
Bankruptcy court records via PACER are the primary source for understanding the PBGC's recovery in distress terminations. When the PBGC takes over an underfunded plan, it becomes a creditor in the plan sponsor's bankruptcy for the unfunded benefit liability — the same amount recorded as the PBGC claim in the termination dataset. PACER (Public Access to Court Electronic Records) at pacer.gov provides access to federal bankruptcy case dockets, including the PBGC's proof of claim, the plan's recovery under the reorganization plan, and the settlement amounts. The settlement amount divided by the PBGC's claim is the recovery ratio — typically very low for large industrial bankruptcies, where the PBGC's claim is an unsecured creditor behind secured lenders and priority administrative claims. The combination of the PBGC claim amount, the Form 5500 pre-termination financial history, the SEC pension footnote disclosures, and the PACER bankruptcy recovery constitutes a complete picture of how a pension termination unfolded from warning signs to final resolution.
For DOL Wage and Hour Division enforcement data covering the same workforce populations whose pensions appear in PBGC termination records: Wage theft by employer: using DOL Wage and Hour Division enforcement data to find labor violations →
For NLRB union election data covering the collective bargaining relationships behind multiemployer pension plans: Who won, who lost: five years of union elections in NLRB data →
For SEC Form 13F institutional holdings data that reveals who holds equity in the companies whose pension plans appear in the PBGC dataset: Who owns what: indexing SEC Form 13F institutional holdings data →