A Review of the Pension Benefit Guaranty Corporation Pension Insurance Modeling System

Jeffrey R. Brown, Douglas J. Elliott,
Douglas J. Elliott Former Brookings Expert, Partner - Oliver Wyman
Tracy Gordon, and
Tracy Gordon
Tracy Gordon Co-Director and Acting Robert C. Pozen Director - Urban-Brookings Tax Policy Center, Urban Institute
Ross A. Hammond

September 11, 2013

Executive Summary

In December, 2012, the Social Security Administration (SSA) invited proposals from
the Retirement Research Consortium to “independently review and evaluate the
data, assumptions, and methods underlying models of the Pension Benefit Guaranty
Corporation’s (PBGC) pension plan insurance programs and related models of
pension funding and sustainability.” In response to this request, a team of
researchers affiliated with the National Bureau of Economic Research (NBER) and
the Brookings Institution prepared this analysis of PBGC’s Pension Insurance
Modeling System (PIMS).

Our analysis suggests that the PIMS model was, in many ways, “state‐of‐the‐art”
when it was created approximately two decades ago. The use of stochastic
simulation tools is a clear improvement over the deterministic model used
previously. Among other benefits, a stochastic simulation model helps interested
parties understand that there is a distribution of possible outcomes, not just an
average outcome – a fact that is especially important for a program that is largely
insuring against extreme events. It is also clear that the professional staff at PBGC
has a deep understanding of both the capabilities and the limitations of the model. It
is our impression that PBGC staff is committed to the principle that the PIMS model
should be as unbiased as possible and insulated from political considerations.
However, several key components of the model have not been revised to reflect the
availability of new tools, new insights from the academic literature, or even new
data. PIMS has developed into a considerably more important tool for policymakers
than was initially envisioned, but resources for PIMS have not risen
commensurately, and budget and staffing constraints appear to have limited PBGC’s
ability to keep the model up‐to‐date.

Our review also highlights three features of the existing governance system for
overseeing PIMS: (i) some of the model documentation is internally inconsistent and
outdated, (ii) the process for updating data and model parameters appears, at least
to external observers, ad hoc, and (iii) there does not appear to exist any publiclyavailable,
systematic inventory of the robustness checks that have been performed.
Indeed, to the extent that methods or assumptions are tested, this fact is not
documented in any central location, making it difficult to assess which features of
the model are most critical. Other long‐term models that are important to federal
programs – such as the actuarial models underlying the report of the Trustees of the
Social Security and Medicare programs – regularly undergo an external review by a
technical panel of outside experts, a process that has led to continual improvement
of those models over time.

A key finding of our review is that the limited treatment of correlated risk factors
arising from the macroeconomic environment is likely to substantially understate
the degree of fiscal risk to PBGC’s insurance programs. This may be one reason that
actual PBGC results have come out much below PIMS’ median projections. In the
PIMS model, there are very few avenues through which broader macroeconomic
factors can operate directly on the distribution of potential future losses. In reality,
however, macroeconomic factors directly affect many of the key drivers of PBGC’s
finances: for example, during an economic downturn, it is reasonable to expect more
plan sponsors to experience financial distress and more plans to be underfunded.
Consequently, the distribution of possible loss exposure has much “fatter tails” (that
is, the probability of extreme losses is much greater) than is currently captured by
the PIMS model. This matters because PBGC and other insurers have an asymmetric
exposure to fat tails, being hurt more by the negative extremes than they are aided
by the positive extremes.

Although our analysis focuses narrowly on the PIMS model, rather than broader
policy questions about the pension insurance program, it is worth stressing that
these extreme negative events are most likely to occur in states of the world in
which the broader U.S. economy is relatively weak, which means that it would be a
particularly economically painful time for the nation to have to address an
underfunded pension insurance program. Recognizing the true economic costs of
these correlated risks and how they affect the broader fiscal position of the U.S.
government, therefore, has potentially important implications for program design,
the average level of premiums, the question of whether to risk‐adjust premiums,
and other important policy parameters which are well beyond the scope of this
narrow technical review of the PIMS model. Our review provides a number of
specific observations about the model that could be used to guide future revisions to
the model in this respect, particularly with regard to the modeling of the bankruptcy
and financial market processes.