Wednesday, August 12, 2015

Any damned data will do

From Robert Higgs
Among the many problems with using expenditure data to compare standards of living across different countries is the incomparability of their climates. For example, many people in temperate-zone countries, not to mention places farther north, spend thousands of dollars each year just to heat their homes in the winter, which are still not as comfortable as people's homes in tropical regions, where no heating expense at all must be incurred. One can make a long list of such incomparabilities. But econometricians want data, and for many of them any damned data will do, regardless of their substantive suitability for the measurement task at hand.
In an age when we are in the midst of Big Data, there is a curious inattention to the variability in the quality of data, not just in terms of measurement errors for precision and accuracy, but more importantly regarding context. Deracinated data is the bane of good analysis. It is as if a chef were told to "Put the meat in a broth for ten minutes." What kind of meat? what kind of broth? what temperature? You have stripped away too much information for the recipe to be useful.

No comments:

Post a Comment