Is the Social Security Administration offering erroneous projections?

By and

Social Security is the U.S. government’s largest program, accounting for almost a quarter of its spending. Many retired workers, disabled workers and their families rely on social security as a key source of their monthly income. In fiscal 2015, about 60 million people will receive benefits totaling $877 billion. The nearly 40 million retirees enrolled in the program get an average of $1,335 per month.

The Social Security program, established in 1935, is funded partly by payroll taxes and income taxes on benefits. Tax revenues are credited to the program’s two trust funds along with interest payments on securities that are held by those funds, according to the Congressional Budget Office (CBO). The funds, in turn, pay for benefits and administrative costs. Lawmakers, government officials and others have expressed increasing concern about the Social Security program because its viability depends on the solvency of the two trust funds. As more members of the baby-boom generation retire, the gap has widened between the amount of tax revenue going into the trust funds and the amount of money being spent from them. The CBO projected in December 2014 that one of the trust funds — the Disability Insurance Trust Fund — will be depleted in fiscal year 2017 and the other — the Old-Age and Survivors Insurance trust fund – will be exhausted in 2032. Meanwhile, the Social Security Board of Trustees has released different projections. In July 2015, the board reported that the Disability Insurance Trust Fund will be depleted in 2016 and that the combined asset reserves of both funds will be depleted in 2034.

Critics, including some scholars, have questioned the accuracy of Social Security Administration (SSA) projections, which legislators depend on when making decisions related to policy and spending. A 2015 study published in Political Analysis, “Explaining Systematic Bias and Nontransparency in U.S. Social Security Administration Forecasts,” examines the SSA’s forecasts to gauge whether they make the two trust funds appear healthier than they are. The three authors — Konstantin Kashin and Gary King of Harvard University and Samir Soneji of Dartmouth College – also analyzed the possible reasons for forecasting errors. To gather information, the authors conducted “a large number” of personal interviews with individuals who are involved with or connected to the federal agency. This study builds on an earlier, related 2015 study published in The Journal of Economic Perspectives.

This study’s findings include:

  • Before 2000, there was no observable bias in SSA forecasts. After 2000, forecasting errors emerged, making Social Security trust funds look healthier than they are.
  • The SSA does not meet best practices in scientific evaluation procedures that are common in other government agencies as well as industry and academia. One failing is that the SSA has not published systematic or comprehensive evaluations of its forecasts.
  • Errors in forecasting likely occurred because of a lack of transparency in SSA’s forecasting methods and a lack of formal statistical models used to create forecasts. The SSA frequently makes qualitative decisions and manual adjustments in its forecasts without providing sufficient information for outside groups to replicate or evaluate the quality of forecasts.
  • The SSA has generated problematic forecasts for financial metrics, including one known as the cost rate, which is equal to the ratio of the cost of SSA programs to the taxable payroll for a given year. From 1978 to 1999, the difference between the forecast five years out and the actual observed value was 0.1 percentage points on average. After 2000, the magnitude of the error increased more than tenfold on average. “Similar results of approximate unbiased forecasts before 2000 and very substantially biased forecasts after 2000 exist for other SSA financial forecasts we studied,” the authors note.
  • Forecasting errors did not occur because of any deliberate effort to fudge numbers. Actually, SSA actuaries “hunkered down” to prevent political pressures from affecting their forecasts. However, in addition to shielding themselves from political pressures, they inadvertently also shielded their forecasts from necessary modifications that would help with accuracy.

The researchers offer suggestions for how the SSA can avoid projection errors. They note that while the SSA Office of the Chief Actuary (OCACT) may aim to be unbiased, it needs to follow well-developed, best practices designed to avoid bias. “The self-conscious public stance of OCACT is as an island of fairness and objectivity amidst a storm of partisans, and so far as we can tell this is precisely what they attempt to do,” the authors state. “Having public servants who try for this level of fairness is certainly ideal but, as indicated above, trying harder to be free from bias is an ineffective way to further reduce bias unless they begin to use the well-tested advice from the scholarly literature.”

Related research: A 2015 annual report of the Board of Trustees of the federal Old-Age, Survivors and Disability Insurance program offers details on the actuarial status and financial operations of trust funds for the Old-Age and Survivors Insurance program and the Disability Insurance program. A 2014 research paper for the Michigan Retirement Research Center looks at the effect of life-expectancy trends on the social-security program. A 2012 study from the Congressional Budget Office and the Social Security Administration examines retirement trends, economic circumstances and the effects of rule changes on beneficiaries.


Keywords: Social Security, forecasting, bias, accuracy, budget, federal government, retirement, retirees, elderly, senior citizens, SSA, Old-Age, Survivors and Disability Insurance

Last updated: December 17, 2015


We welcome feedback. Please contact us here.

Citation: Kashin, Konstantin; King, Gary; Soneji, Samir. "Explaining Systematic Bias and Nontransparency in U.S. Social Security Administration Forecasts," Political Analysis, Summer 2015. doi: 10.1093/pan/mpv011.