Gary King Delivers Inaugural ISS Distinguished Lecture
As of 2014, Social Security comprised 37 percent of all federal budget expenditures. Meanwhile, one in five senior citizens currently depends on Social Security to keep them out of poverty. Professor King explained why those gloomy forecasts are worse than previously thought. He also stressed the importance of applying the best practices of the social sciences in federal entitlement programs.
King’s research has concluded that while the Social Security Administration’s trust fund forecasts showed relatively little bias before 2000, the same cannot be said for projections released since. Not only have these been overly optimistic—they have also featured statistical errors increasing in magnitude with each passing year.
Inflated optimism in the Social Security forecasts affects other critical areas of the federal budget. Other federal entitlement programs, such as Medicare and Medicaid, depend on those projections to predict future healthcare spending. The Social Security Administration—specifically the Office of the Chief Actuary—uses these projections to score public policy proposals to reform Social Security. Policymakers on both sides of the partisan aisle rely on these scores to determine the future impact of their proposal on federal spending.
Outmoded statistical analyses—not political corruption or a conspiracy of nefarious technocrats—account, King said, for the discrepancy in statistical modeling. The Social Security Administration’s actuarial forecasting models have changed little since the 1940s. While the data science revolution has transformed the ways industry and research institutions make predictions about the future, the Social Security Administration still relies on outmoded statistical models that King described as “ad-hoc,” “jerry-rigged,” and “arbitrary.” In the wake of increased partisan polarization, the Social Security Administration has become more insular, opting to stick with current practices rather than modernize its forecasting models. Moreover, life expectancy has increased since 2000—Americans are living longer and will continue to depend on their Social Security benefits longer than previous generations. The Social Security Administration has yet to adjust their forecasting models accordingly.
King outlined three simple fixes that could improve the accuracy of the Social Security forecasts without provoking battles on Capitol Hill. First, develop models that remove as much human judgment as possible. When it comes to big data, what can be automated should be automated. Second, in instances where human judgment is required, the Social Security Administration should embrace protocols that minimize bias. Double blind experiments in academic and industrial settings are examples of structural procedures that minimize error due to human bias. Finally, he called on the Office of the Chief Actuary to share data and make it available to the public for independent verification. These standard features of modern academic and industrial quantitative research—features which the Social Security Administration has yet to adopt—will, according to King, reduce bias in Trust Fund projections and provide policymakers with an accurate metric of the fund’s solvency.
Learn more about Gary King.
The ISS Distinguished Lecture series brings influential policymakers and eminent scholars in the social sciences to UC Davis. To receive updates on future installments in the series, as well as other ISS news, events and funding opportunities, sign up for the ISS newsletter.