We’ve had a fascination with manufacturing productivity statistics for quite some time. Several months ago we began exploring a possible problem with how manufacturing productivity is calculated.
Very briefly, we were concerned that intermediate subassemblies that were built at overseas factories but then turned into completed product at U.S. factories would distort productivity in that it would look like more was being produced with fewer domestic hours. An offshored subassembly doesn’t represent an increase in true output by a U.S. worker, therefore this distortion potentially called into question the oft-heard comment that productivity gains have led to the steep reductions in manufacturing employment.
Our questions got some real press, which led to some comments from real economists. Then we were pointed to some actual studies that showed that the productivity impact of offshored intermediate components was about 3-6% of total productivity gain, so not a huge factor. Since total dollar manufacturing output in the U.S. continues to rise, and the impact of offshored intermediates is small, the reduction of manufacturing jobs can be predominantly attributed to productivity improvements.
Now our friend Don Boudreaux of Cafe Hayek tells us of another way productivity statistics can be misinterpreted. In an article in the Christian Science Monitor titled Statistics Can Mislead As Easily As They Can Enlighten, Boundreaux takes on some popular myths regarding country-to-country comparisons of productivity, such as:
Not long ago, Corinne Maier boasted in The New York Times that "… in many years French workers have a higher productivity rate than their American counterparts." Measures of productivity do regularly reveal French workers to be more productive than American workers.
Stats are stats and facts are facts, right? Well…
France’s labor regulations are much more burdensome than those in the US. By artificially raising the cost of hiring workers in France, these regulations make it unprofitable to hire the lowest-skilled workers. One result is that only higher-skilled workers get jobs in France. But because US labor regulations are less restrictive, a higher proportion of low-skilled workers find jobs in America. With a larger proportion of highly skilled workers, France’s average productivity is bound to be higher. But the French shouldn’t be cheering.
So you can have much higher unemployment primarily of the lower skilled workers creating a higher productivity statistic. Boudreaux clarifies this even further in his article via a couple simple examples. But this is only only one example of why you have to be careful with statistics.
The larger lesson is that proper interpretations of statistics often are surprisingly counterintuitive. After all, our intuition tells us that countries with higher labor productivity do better economically than do countries with lower worker productivity. But our intuition is wrong.
There’s another example that ties in directly with a debate currently going on in the U.S.:
Statistics can also fool us when averages change over time. Suppose that the average real-wage rate in the US falls. Do we conclude that American workers are worse off? That’s one possible explanation. But before jumping to that conclusion, be aware that another, very different, explanation might better fit the facts. If lower-skilled workers enter the labor force in unusually large numbers, the average wage rate will fall without necessarily reducing any worker’s pay. Indeed, the typical worker can even see his real-wage rate rise while the average rate falls!
Obviously this has been happening with the mass influx of immigrants, legal and otherwise. Overall unemployment is at near all-time lows and dangerously close to the point where wage inflation could ignite. Average wage rates have dropped due to the number of new people coming into the country and taking lower skill jobs. But we’re not thrashing statistics in general.
None of this is to suggest that statistics are useless. Quite the contrary, statistics are indispensable to grasp reality better and to distinguish explanations that are correct from explanations that are merely plausible or even downright erroneous. But statistics will assist us in our quest for understanding only if we approach them critically, aware that they can mislead as easily as they can enlighten.
Just be careful, and be aware of the potential for countintuitive explanations.
Paul says
The problem does not lie with the statistics but rather with those who abuse them. Donald Wheeler would say one needs to understand the context from which the data are drawn. Only then would it be possible for meaningful comparisons to be made. What you have done is research the context for the data sets. The original comparison is correct (it is only arithmetic), but it must be understood that the two labor markets are different in a meaningful way. This is why 85% of data analysis effort should go into understanding the context. Conclusions follow after understanding is achieved.