USA Today last week published yet another story claiming that public sector workers make more that their private sector counterparts - this one saying that Wisconsin is one of many states where this is the case. Their “analysis” used data from the Bureau of Economic Analysis, and compared total compensation (salary+benefits) between workers in the private sector and state/local government.
No matter how many times they are told that you can’t just make a straight comparison of dissimilar groups of workers, apparently they still don’t get it. Incredibly, this particular article admits as much, and even quotes economist Jeffrey Keefe, who tells them that the gross comparisons don’t account for important sectoral differences in education and other factors. In other words, their numbers don’t tell us much of anything about public versus private sector compensation. Still, there is the headline: "Wisconsin one of 41 states where public workers earn more." How many people saw that headline, and now believe that public workers are “overpaid?"
USA Today, of course, is not alone. These assertions have lately become insidious, coming from governors, commentators, and others. But when a major national newspaper decides to run this story at this politically-charged time, based on their very own “analysis," a separate response seems in order.
I’ve discussed this issue before, but maybe it would be more helpful to show how the data are more properly analyzed in a step-by-step fashion, using 2009 U.S. Census microdata (the American Community Survey, available from the wonderful organization IPUMS.org). Here’s how you make a false earnings gap disappear in five minutes.
I limit the sample to workers who are in the labor force and work at least 35 hours per week (full-time workers). Federal employees, family workers, agricultural workers, and the self-employed are also excluded. This permits a comparison of state/local government employees with workers in the private sector.
It is important to note, however, that these are wage and salary data only – they do not include benefits (which the Census does not collect). I use these data only to demonstrate how and why a straight public-private comparison is invalid. Remember, though, that more educated workers also tend to get better benefits in both sectors - the relationship is less strong, but the discussion below applies for benefits as well. (For careful matched comparisons of public-sector workers to private-sector workers, which include both wages and benefits, see here, here and here.)
Let’s start with the USA Today methodology – a gross comparison of all workers in the public and private sectors. Keep in mind that these workers vary a fair amount in their hours and weeks worked, so I’ll also add a column that very crudely holds constant the number of hours worked per week by restricting the sample to those who work 40 hours (over half the sample).
According to this estimate, public employees earn roughly the same across all workers, and 12.5 percent more among those who work a standard 40-hour week. There’s a lot of underlying variation, but you can’t see it.
A majority of public employees are professionals, compared with only about one in four private sector employees. Many government workers (teachers, bridge engineers, scientists, etc.) are very highly educated. They also tend to be more experienced than private sector workers, on average. In short, their jobs and qualifications are different, and you can’t compare them as if they weren’t. So, let’s see what happens when we actually control for some of the factors that we know influence earnings.
In the table below, the dollar amounts represent the difference between public and private workers’ annual earnings from their jobs, controlling for the other factors listed (in OLS regression models). Negative numbers mean that public employees earn less, all else being equal. See the table stub for details of the models, if you're interested.
Let’s take it one row at the time, very quickly.
First, we’ll try a slightly more refined nationwide version of the USA Today method – a model that compares workers in the public and private sectors while controlling for state (interstate differences in average earnings), as well as hours worked per week, weeks worked per year, and gender (there are more women, as a proportion, in the public sector).
With this rudimentary set of control variables, the public sector “advantage” is about $1,600 per year, which is equivalent to around three percent of the average private sector salary across all states and work schedules, for both men and women (a rough characterization of the "advantage").
Now let’s take a look at the public/private wage gap if we add worker experience to the model (since public employees are, on average, more experienced, and would therefore make more money in either sector). This narrows the gap by about a third – roughly one percentage point. Intersectoral differences in experience, which are ignored in all sweeping comparisons of public/private employees, clearly matter.
But education is by far the most significant factor. If we add that in too, we find that the situation is completely reversed, in a rather dramatic fashion – public employees earn over 20 percent less than comparable private sector workers, a difference of almost $10,000 per year. Using the clumsy methods of USA Today and others, public servants seem to be overpaid. In reality, their earnings fall far short of comparable private sector workers.
Again, this doesn’t include benefits, which, once accounted for, help to narrow this gap substantially (overall, public employees earn around five percent less in total compensation). Nor does it show that some public sector workers – those in the lower-skill, lower-wage jobs – would probably earn less in the private sector. And, finally, this is not a comprehensive model (e.g., it does not include firm size and other factors) or analysis – see the papers linked above for better ones.
But it does show that context matters – a lot. As I’ve pointed put before, it makes little sense to claim that IBM employees are overpaid just by comparing them to Walmart workers.
What it also shows is that the USA Today story, as presented, was irresponsible. In a charged political atmosphere, as Wisconsin protesters were camping out inside the capitol, one of the nation’s largest newspapers chose to run a story under a headline calling them overpaid – when they’re not – based on an analysis that couldn’t have been more simplistic or inappropriate.
That’s bad research. And bad journalism. And we don’t need any more of either.
Comments