I’ve had many people asking me whether Jacob Hacker’s results about "the great risk shift" hold up. The CBO weighs in:
Since 1980, there has been little change in earnings variability for both men and women. There is some evidence that, between 1960 and 1980, earnings variability increased for men but was offset by a decrease for women. Those findings are consistent with most existing studies of the topic that use publicly available survey data, which tend to find higher levels of earnings variability for men in the 1980s and 1990s relative to the 1970s, but little change since around 1980.
Here is the paper. I’ll read through it soon, and report back if my deeper impression runs in the other direction. If you know of relevant defenses of Hacker, please do leave them in the comments. I’d like to get to the bottom of this.
My sense is that the weakness of the CBO’s report is its source data. It’s all based on wage data from SSA, which means that it has no demographic information etc. So, for example, if you wanted to look at household income uncertainty you couldn’t see this in the CBO data. And since there are fewer married couples today than in 1970, more people living alone, higher frequency changes in household composition (more divorce, more co-habitation) then you could have no change individual income variability but big changes in family or household income variability.
Another factor (which I don’t believe) is that wages are a smaller share of income than they used to be, or at least there are a greater number of people who receive lots of income from non-wage sources (business income, capital gains etc) and that changes in the variability of these sources of income has increased.
Saez, Song and Kopzuck [sic] also have a new working paper on this that looks at a much longer time frame with the SSA data and comes to similar conclusions about the stability of wages over the last couple decades. In particular they show that the increase in income inequality at the very high end isn’t due to random shocks or higher variance–once you’re that rich you are always that rich. In fact, mobility between the sorta-rich and the really rich has gone down recently.
Since 1980 women’s share of employment has increased significantly.
Women have greater income volatility then men.
Consequently, total or average volatility has increased.
This is true even though volatility for men or women is unchanged.
It stems from the change in composition.
The CBO addresses the question that can be answered with wage data from the SSA, but more interesting (and useful) would be a look at consumption variability, not earnings variability.
FYI, Hacker’s work is based on the Panel Study on Income Dynamics sample, which contains extensive demographic information and the like.
Have you seen this? It is another approach to the issue.
In the past quarter century, the ups and downs of the American economy – that is, its business cycle volatility – have decreased. That’s a good thing: it means less severe recessions, milder swings in the unemployment rate, and possibly fewer business failures. Over the same time period, though, the volatility of employment growth rates and sales growth rates at some 10,000 companies whose securities are traded on various stock markets have risen, on average.
In Volatility and Dispersion in Business Growth Rates: Publicly Traded versus Privately Held Firms (NBER Working Paper No. 12354), co-authors Steven J. Davis, John Haltiwanger, Ron Jarmin, and Javier Miranda seek to explain these apparently contradictory trends. For their study, they use the recently developed Longitudinal Business Database (LBD), which contains annual observations on employment and payroll for some 6 million U.S. businesses. This is a dramatically larger and more comprehensive database than the COMPUSTAT data on publicly traded companies used in previous studies. Publicly traded companies constitute less than 1 percent of all U.S. firms and about one-third of U.S. employment in the non-farm business sector.
valuethinker,
Imagine a group of W’s and B’s, each containing five people. The five W’s score an average of 1200 on the (old, i.e. two categories, not three) SAT’s, and the five B’s score a mean average 800 on the SAT’s. The average for the group is 1000.
Now imagine that ten years later education where W’s live has improved, and the average SAT score for W’s is 1250. The education system where the B’s live also improves, and they score an average of 850. What is the overall average?
If you answer 1050, you gave the wrong answer. The correct answer is that there isn’t enough information, you need to know how many W’s there are and how many B’s there are. If there are still 5 W’s and 5 B’s, then 1050 is correct. But what happens if there are now 5 W’s and 10 B’s? You add 5250 and 8500, making 13750. Then divide 13750 by 15 to get 916 2/3.
So instead of of an overall improvement of 50 points if the ratio of the two groups is the same, you have a decrease in the overall average of 83 1/3 points, quite a swing!
This isn’t to say that your thesis is incorrect of worsening schools. The point simply is that changing the demographics of the sample, of any sample, distorts what is actually happening and can distort what is actually a good thing into what appears to be a bad thing.
Please come to Tibia money, we will give you a great surprise.
Comments on this entry are closed.