In Part 1 of my response to Sen. Bernie Sanders and economist Teresa Ghilarducci, I explained why Sanders’ favorite and most compelling factoid – that nearly half of Americans approaching retirement have zero savings – is outright false, something Sanders should have known for years.
But it’s not about a single factoid. In arguing that the U.S. retirement system doesn’t work, Sanders and Ghilarducci make several other claims that are equally incorrect, albeit for a different reason. The U.S. retirement system – both private savings and Social Security – is both massively important and massively expensive. This isn’t personal to me. I like Teresa very much. And who knows, maybe Bernie and I would hit it off. He’s grumpy, I’m grumpy.
WASHINGTON, DC – OCTOBER 06: Sen. Bernie Sanders answers questions during a press conference at the … [+]
But we very much need to use accurate data to understand if the U.S. retirement system is working well. But only one side of the retirement debate is doing it.
Consider these figures taken from Teresa’s article, two of which quote Sanders while the third is Teresa’s own:
- “Roughly 52 percent of Americans 65 and older ‘are living on less than $30,000 annually, and one in four survive on less than $15,000 per year.’”
- “Nearly 17 million older Americans aged 65 and up are financially insecure, below 200% of poverty — that’s nearly one-third; more than 10 percent are in poverty.”
- “The elder poverty rate has been stubbornly steady at around 10 percent over the past 30 years, and rose from 8.9 to 10.2 percent from 2020 to 2022.”
All of these claims cite data from the U.S. Census Bureau’s Current Population Survey (CPS). In light of these figures, only a fool – for instance, me – would deny the looming retirement crisis.
And yet each of these claims is incorrect, a fact that the Census Bureau itself acknowledges. Accurate data tell an entirely different story.
Let’s start at the beginning. In a 2012 Wall Street Journal op-ed, Sylvester Schieber and I pointed out that the Current Population Survey reported that Americans 65 and older collected $228 billion in income from IRAs, employer-sponsored retirement plans, and annuities. IRS data, however, showed that Americans received $568 billion from these same sources. Now, Americans aren’t telling the IRS they’re receiving more income than they actually do. This implies that the CPS, the source of Sanders’ and Ghilarducci’s factoids, captures only about 40% of the benefits paid by private retirement plans.
The U.S. Census Bureau took these concerns seriously. In 2017, two Census economists, Adam Bee and Joshua Mitchell, published a groundbreaking study that linked CPS survey responses with IRS records. This linkage let the two researchers compare the income the CPS reported for a given household versus what IRS data showed.
What Bee and Mitchell found was startling, probably the most significant piece of retirement research in the past decade: the median retiree’s true income in 2012 was 30 percent higher than measured in the CPS. The true elderly poverty rate wasn’t the 9.1% reported in the CPS, but just 6.9%. Since 2017 the Bee-Mitchell study has been cited almost 150 times by other academics.
And the Census Bureau has not stopped: in 2023 it released the first iteration of its National Experimental Wellbeing Statistics (NEWS
EWS
Social Security Administration economists validated this approach, finding similar results. For instance, while the CPS reported that only 47% of 65+ households received benefits from a private retirement plan, IRS data showed that 69% actually did.
And private retirement plan benefits are the root of the CPS’s income measurement problem. The main reason: the Census Bureau defines “income” as payments received “on a regular basis.” This means that as-needed withdrawals from a retirement account aren’t counted as “income.” And because seniors are increasingly drawing benefits from retirement accounts, the CPS’s underestimation of retirees’ true incomes has grown larger over time.
Bernie Sanders and Teresa Ghilarducci apparently live in a world where none of this research has happened.
So what do more accurate data tell us about the claims Sanders and Ghilarducci make?
The publicly-released Census NEWS data don’t let us precisely check Sanders claim that 52% of 65+ have incomes below $30,000 and one-in-four less than $15,000. But 52% signifies a figure close to the median, and we know that the true median income for seniors is 27% higher than in the CPS data Sanders cites. So Sanders’ claim is almost certainly incorrect.
What about Sanders’ factoid that nearly one-third of older Americans have incomes below 200% of poverty? Well, Bee and Mitchell’s study found that the CPS reported 33.4% of seniors had incomes below 200% of poverty. But IRS data showed that only 23.6% did. So, wrong again.
What about Teresa’s claim that senior poverty “has been stubbornly steady at around 10 percent over the past 30 years.” This is perhaps the biggest problem. According to the Census Bureau’s Bee and Mitchell, 9.7% of seniors had sub-poverty level incomes in 1990. By 2018 the elderly poverty rate had fallen to 6.4%. Over the same period Teresa cites we’ve seen a more than one-third reduction in the risk of retiring poor. That’s an entirely different narrative.
Now, I don’t expect everyone to be aware of these data issues. For instance, for years the Social Security Administration claimed, based on CPS figures, that up to 36% of seniors received nearly all their incomes from Social Security. When Bee and Mitchell checked these statistics using IRS data, they found that just 12% received 90% or more of their incomes from Social Security. That’s a massive difference, but you can’t expect the man in the street to know the research.
But Sanders is a Senate committee chair with a staff capable of producing a 21-page report featuring 94 footnotes. Ghilarducci is a tenured professor specializing in retirement issues.
They should know better than to cite data we know with certainty to be inaccurate. We owe people a better retirement debate than they’re getting.