Uncertain Forecast for Social Security
Program's health has been overstated for years, studies say
A new study has found that the financial health of Social Security, the program that millions of Americans have relied on for decades as a crucial part of their income, has been dramatically overstated.
The study compared all forecasts made by the Social Security Administrationover the 80-year history of the program with its actual outcome, and found that its forecasts of the health of Social Security trust funds have become increasingly biased since 2000. Current forecasts are likely off by billions of dollars, and the program could be insolvent earlier than expected unless legislators act, the study found.
The study, which appears in the Journal of Economic Perspectives, was co-authored by Gary King, the Albert J. Weatherhead III University Professor at Harvard University; Konstantin Kashin, a Ph.D. student at Harvard’s Institute for Quantitative Social Science; and Samir Soneji, an assistant professor at the Dartmouth Institute for Health Policy & Clinical Practice.
In a second paper, published on the same day in Political Analysis, the Harvard-Dartmouth team points to antiquated, ad hoc methods for creating the forecasts as the cause for the growing bias. They suggest that otherwise laudable efforts to insulate the forecasts from political influence have resulted, somewhat ironically, in insulating the process from data that could improve their accuracy.
“The bias in their forecasts results in a picture that’s rosier than it really is,” King said. “They’re not saying the system is in good health. Pretty much everybody who evaluates Social Security realizes there’s a problem … But the system is in significantly worse shape than their forecasts are indicating.
“This is a major problem,” he continued. “Social Security is the single largest government program. It lifted an entire generation of elderly out of poverty, and today affects the lives of almost every American. The forecasts are essential for ensuring the solvency of the Social Security trust fund, as well for Medicare and Medicaid, which together add up to half of the entire budget of the federal government.”
While forecasting the health of the Social Security trust funds has long been part of the program — each year, the administration creates forecasts that look one, five, 10, 20, and even 75 years into the future — the study conducted by the Harvard-Dartmouth team is the first by anyone inside or outside of the government to evaluate their accuracy.
“It’s typically been difficult to conduct studies that evaluate forecasts, but the Social Security Administration has been around long enough that if they made a 10-year forecast a decade ago, by now we can look to see how they did,” Soneji said. “There’s tremendous scientific value in evaluating real-world forecasts that were made by people who were really trying to figure out what the future was going to be like.”
What they found, King said, was that while forecasts were never perfect, they were largely unbiased for quite some time.
“On average, they were about right until about 2000,” he said. “Sometimes they were too high, sometimes they were too low, but they were able to adjust quickly enough over time and remained fairly accurate.”
Over the last decade and a half, however, those course corrections weren’t made, and the gap between the forecasts and reality has grown steadily. To understand just how wide that gap is, King said, it’s necessary to understand the other key role played by Social Security auditors: evaluating legislation related to the program.
No matter what reforms are put in place, King said, it’s important to understand that the forecasting process will never be foolproof.
“The progress that’s been made in data science formalizing, and thus improving, human decision-making has been spectacular, and these developments need to get to Social Security,” King said. “The rest should be dealt with by social psychologists, who can devise procedures to take the human bias out of the process that must remain qualitative. For example, the late Harvard psychologist Richard Hackman showed that if men and women auditioned for violin spots in an orchestra from behind a curtain, men still won most of the spots. But if you took off their shoes first, so the judges couldn’t hear who had on high heels, the gender bias vanished.”
Soneji explained: “The combination of modern data science, modern social psychology, and modern data sharing can vastly improve the situation.”
Ultimately, however, taking steps to improve the forecasts can’t keep Social Security from becoming insolvent. The debate over how to keep the program afloat must be left to the nation’s elected representatives. But by improving the forecasting process, King said, it is possible to ensure that debate is informed by facts.
“I don’t know how the politics are going to come out,” King said. “There certainly are ways to keep the system from going insolvent: You could slightly lengthen the retirement age, increase taxes on the wealthy, or increase payroll taxes. Our results don’t say which of those to choose, or even whether to choose anything. I think the politicians will do something. There have been grand compromises over Social Security over the years. When the parties sit down to negotiate, all we want is for them to have the real facts. That’s all.”
Click here for the full story as reported in the Harvard Gazette.