Is Income Volatility Really Rising? For Whom?

Jacob Hacker‘s Great Risk Shift described rising income risk over recent decades as an important and quite general phenomenon. While there’s been plenty of controversy around that claim, the most careful analysis I have seen roughly supports Hacker’s contention. (The CBO, using different data and a different methodology disagrees.)

What Hacker actually shows is that the average level of income volatility is rising. But we know that an average can hide as much as it reveals. And this is the point brilliantly developed in a provocative new working paper by my Wharton colleague Shane Jensen, and my former colleague Stephen Shore (full paper available here; warning, there are some econometric pyrotechnics involved).

Income volatility is not a single number — some people’s incomes move around over time more than others. And while Hacker and others have documented a rise in the average level of income volatility, Jensen and Shore document changes in the entire distribution of different people’s income volatilities.

A stock market analogy might be useful: some stocks are more volatile, some are less, and it is interesting to see what is happening to the volatility of different types of stocks, and not just some mythical “average stock.”

The Jensen-Shore findings are pretty stark and are sure to stir the policy debate: Despite sharp growth since the 1970’s in the average level of income volatility, median income volatility is basically unchanged. (There are some differences in samples and methods between the Jensen-Shore and Hacker analyses, but I would be surprised if that explains much.)


Indeed, there’s been no change in income volatility for most of the distribution:


Here’s the punch line:

The key driver of rising average levels of income risk is that life among the already risky has become even riskier. Indeed, you really need to look to the riskiest 5 percent of the distribution to find the rise in income risk. And this rise in risk among the already risky is so great as to be responsible for nearly all the rise in average income volatility. And who are these riskiest 5 percent? Jensen and Shore find that they are particularly likely to be self-employed.


The Jensen-Shore analysis yields an interesting scorecard: Hacker was right on average, but wrong for 95 percent of us.

And as for that earlier Hacker v. CBO debate, perhaps Jensen and Shore have a useful reconciliation. The CBO’s analysis was based on administrative income data, which we typically think of as being better than self-reported income. But in this case, one might imagine that the income swings of the very riskiest 5 percent didn’t show up on their W-2’s (think: entrepreneurs or the self-employed), and hence it is possible that the CBO data missed the rising risk among the riskiest.

You might think that the disconnect between Hacker’s thesis and the actual experiences of most people would have undermined its acceptance. But Hacker’s thesis resonated strongly with many policymakers because it somehow “feels” accurate. Here, I suspect that it is too easy to confuse the feeling that the labor market is riskier than we would like with the actual claim that income risk is higher than it was.

Perhaps the debate about the Great Risk Shift isn’t such a big deal after all: the best argument for a social safety net is that there is too much risk, not that risk has grown.

Full details, including technical wizardry, here.


the real stark finding that should stir public debate is real wage stagnation for most of the population (which incidentally effects the volatility data)- the finance industry's giant speculative swings have no bearing on Joe Punchclock's non-volatile wages that are less and less able to buy the staples of a middle class lifestyle


Forget mere "income." The risk shift is about disposable income, after taxes AND health-care deductions, and assets, too. The point is that health care, with fewer people being covered and those who are covered having to pay more for it, is eating into family finances. Elizabeth Warren's "Two-Income Trap" chronicles how many bankruptcy families have huge medical bills that nee to be discharged. And with fewer Defined Benefit pensions, old age, now a third of your life, is now going to see far more risk and volatility as you depend on the stock market's vagaries more. That's why the "risk shift" rings so true.


This paper is actually more or less irrelevant to Hacker's work, since it examines risk to men's labor income, while Hacker looks at risk to family income. It's entirely possible for risk to family income to increase while risk to men's earnings decreases; this can happen, for example, if the covariance between men's earnings and women's earnings increases over time. In fact, if you look at the literature on the subject, that appears to be pretty much what has happened.


Gotta concur with comment #4. I think the median is hiding that data. After the year 2000, there's a huge drop in variance on the 95th percentile, while there's a large rise in the 1st and 5th percentile. The mean variance after 2000 continue to rise, so I think the 1st and 5th percentile are contributing more to the mean variance than the 95th percentile after 2000. So maybe Hacker is wrong on 85 instead of 95 percent of us?


there was an old man who lived in a shoe
that had so much paper income
he didn't know what to do
untill the shoe was on the other foot
and all his notes came due


>You might think that the disconnect between Hacker's
>thesis and the actual experiences of most people
>would have undermined its acceptance.

Huh? What charts are the authors looking at? The first chart seems to me to say that, for the safest jobs, income volatility went from .2 to .25 -- a 25% rise. And for the next safest line, it went from .25 in the '80s to .3 today. In fact, every one of the lines rose at least somewhat. So I'm guessing that's part of why people think the incoming volatility is rising!


Seems to me, that volatility isn't just "risk" in the layperson use of the word (downside), it's also reward (upside). If I make a billion dollars on some crazy innovative business venture, my volatility has sky-rocketed. Interesting that the increased volatility is among the richest and most risk-tolerant.

Pair this with the notion that income disparity is on the rise (rich are getting super-rich) and what do you have: recent times are seeing the riskiest ventures with the hugest payoffs. Where's the need for a safety net?


2 things:

1. When looking at percentiles, you are not measuring the same people over time, particularly near the top of the income distribution. Also, immigrants and those who just enter the work force tend to fill the lower percentiles and often work their way up over time.

2. Those in the upper end of the distribution tend to have a higher percentage of their incomes from investments, so you would expect that higher volatility in the financial markets would affect their incomes more than the relatively stable labor market.


#4 and #6: You should get sense about the magnitudes. It is true that the two safest displayed quantiles' volatilities increased - but only by about 0.005. What we are trying to explain is why the overall volatility increased from 0.06 to 0.13, i.e. by 0.07, an increase by an order of magnitude higher than what you are pointing out.

If the volatility of 1/20 of the population increased by 0.005, by how much is it going to influence the mean value? Virtually zero. The increases in mean volatility must come from the upper quantiles, where the impacts are of orders of magnitude larger.

There are essentially two possible justifications:
1) The fall in the 95th quantile volatility implies that further increases are even more limited to people above 95th quantile.
2) There are covariances at play which mess up the aggregation exercise.

In any case, the impact of the increase in the volatility of the safest income groups are negligible for the means and aggregates.