Does the CBO Report on Income Volatility Undermine Jacob Hacker's "The Great Risk Shift"? No it Doesn't...
Ever since this CBO report on individual earnings volatility came out, there have been questions raised about whether it undermines Jacob Hacker's work in The Great Risk Shift showing increases in economic insecurity over time. For example, this article by David Leonhardt in the New York Times highlights the CBO results, Greg Mankiw notes the results here, and Tyler Cowen says:
Leonhardt agrees with my view that the recent CBO report effectively counters Jacob Hacker on "the great risk shift." Until we see further evidence to the contrary, that thesis belongs in the "simply isn't true" pile.
However, the CBO results do not effectively counter Jacob Hacker. By email, Jacob explains why:
Jacob Hacker: I have received many questions about the Congressional Budget Office’s (CBO) recent report that finds that individual earnings volatility, while extremely high, has not risen since the 1980s. Although this new research significantly expands what we know about individual earnings volatility, it does not challenge my contention that family income volatility has grown, nor is it at odds with my larger argument that the level of economic risk that families face has risen dramatically.
The distinction between individual earnings and family income is often missed, so it is important to emphasize that there’s no reason why trends in individual earnings instability and trends in family income instability need to match up. Unlike earnings volatility, family income volatility hinges on (1) the joint labor supply decisions of workers in the family; (2) family formation, expansion, contraction, and dissolution; (3) the earnings and losses of family-owned businesses and capital holdings; and (4) government taxes and benefits. Each of these could cause individual earnings volatility and family income volatility to follow different paths.
In fact, it’s been clear for some time that family income volatility and individual earnings volatility have followed different paths. For example, Peter Gottschalk and Robert Moffitt, in their pioneering work on male earnings instability, have shown that there has not been a secular rise in male earnings instability since the early 1980s—a finding confirmed by the CBO study (which, because of data limitations, was not able to look at earnings instability among all workers before 1981). Since I essentially adapted Gottschalk and Moffitt’s technique to look at family incomes, it’s not surprising that in my own analyses of individual earnings volatility (not reported in my book, The Great Risk Shift), I too did not find a consistent rise in individual earnings volatility since the early 1980s—even as I did find a consistent rise in family income volatility.
It is enormously valuable that the CBO is looking into this subject and bringing new data to bear on it. We are continuing to see that the United States has extremely high levels of personal financial instability, especially given how favorable and stable our macroeconomic indicators have been. And, in the process, we are gaining a rich new picture of the economy that’s much closer to the lived experience of Americans.
We are also learning much more about the strengths and weaknesses of alternative sources of evidence and alternative techniques—and I have been incorporating what we have learned into a major reanalysis of family income volatility currently underway at Yale, which will incorporate newly available data as well as refine my original estimates. I hope to be able to release all my updated findings soon.
Nearly all research on family income dynamics, including my own, has relied on the Panel Study of Income Dynamics, a longitudinal survey that has followed a nationally representative sample of American families since the late 1960s (with an initial over-sample of low-income respondents). The PSID is unmatched as a source of long-term data on the over-time income dynamics. Still, the PSID is a survey, and thus has the weaknesses all surveys do: people may misreport their income, they may not be reachable, and so on. Fortunately, there is little reason to think these weaknesses have grown over time, so we can be pretty confident that trends in income instability are caused by real changes, rather than changes in the accuracy of reporting.*
By comparison, the CBO’s data source, Social Security wage records, should be more accurate. For this reason, it is good news that the CBO analysis of Social Security wage data finds the same basic trend in individual earnings instability that researchers using the PSID have found. However, Social Security wage records don’t permit examinations of total family income. This is a big shortcoming, because family income is ultimately what most of us care about. We are interested in earnings dynamics, to be sure, but ultimately we want to know how those dynamics affect family’s resources and well-being. How do government taxes and benefits and the joint labor supply decisions of working family members respond to earnings shocks and other income risks? Have they cushioned individual earnings shocks as well as in the past? What the PSID evidence suggests is that they have not—family income instability has risen despite the lack of a clear trend in individual earnings instability.
Finally, while the PSID evidence indicates that family income volatility has gone up, family income volatility is scarcely the only measure of economic insecurity or the “risk shift” that I and others have discussed. Only one chapter in my book is about family income instability. The rest are about pensions, health care, the decline in traditional job security, the increasing debt burdens reflected in families’ financial balance sheets—in short, about the whole range of economic risks that Americans face. Many of these risks, such as health costs, retirement insecurity, bankruptcy, and mortgage foreclosure, either do not show up in the incomes of working-age people or show up only weakly.
As I put it in The Great Risk Shift, “The up-and-down movement of income among working-age families is a powerful indicator of the economic risks faced by Americans today. Yet economic insecurity is also driven by the rising threat to families’ financial well-being posed by budget-busting expenses like catastrophic medical costs, as well as by the massively increased risk that retirement has come to represent, as more and more of the responsibility of planning for the post-work years has shifted onto Americans and their families. When we take in this larger picture, we see an economy not merely changed by a matter of degrees, but fundamentally transformed—from an all-in-the-same boat world of shared risk toward a go-it-alone world of personal responsibility.”
* As Peter Gottschalk and Robert Moffitt explain in a seminal 1994 paper on earnings variance, while estimates of earnings instability based on PSID data are “unquestionably biased upward, the issue is whether the bias has changed over time. We know of no evidence that it has…[N]or is there any reason to suspect that respondents themselves have become more erroneous in their reporting over time; in fact, most survey research indicates that respondents become more accurate the longer they remain in the survey.”
One more note. Jacob tells me, as he notes above, that he is working on new estimates for public release that take account the questions he has received on his work. So if what Jacob has written above does not leave you convinced, I urge you to withhold judgment until you’ve seen the new work.
Update: Two more notes on interpreting the CBO results.
First, the CBO report doesn't include self-employment earnings, which
are probably more volatile. This is worth noting because rates of
self-employment trended upward through the mid-1990s. That is, the CBO
data excludes a source of increasing volatility and therefore likely
underestimates both overall volatility and the change over time.
Second, there is an additional source of volatility excluded from the CBO report. The year-to-year change of labor earnings doesn't capture big periods out of the workforce, because you just disappear from the dataset. Family figures will capture that because you're likely to have some other income (family members, benefits, etc.). This is likely important because of the rise of long-term unemployment, structural displacement, etc.