The furor over the president's recent firing the BLS director over unfavorable revisions to the jobs numbers highlights an issue with all data driven decision making. Firms are increasingly turning to data to inform strategic decision making. However, often managers consider the new requirement of empirical support for a decision to be a nuisance. Instead, they seek to find analyses that support their prior beliefs. A useful check against this "cherry-picking" of the data is to vest the data analysis team with some autonomy from the decision maker.
Reproduced below is a graph from Vincent Geleso (who credits Gary Wagner) that tracks BLS jobs revisions over the last quarter-century. It appears that data revisions are common, are both positive and negative, are not biased up or down, are often large, and most importantly are non-partisan. This suggests that the recent revisions were not out of the ordinary. Non-independence of future BLS data reporting from partisan priorities will likely render their future reports to be more suspect.
No comments:
Post a Comment