By Carl Tannenbaum

Members of my team are required to memorize what have become known as “Tannenbaum’s Laws,” which govern the use of statistical techniques. One entry in the list is: “The best of models is often undone by the worst of data.” Analysts periodically construct elegant algorithms that produce interesting conclusions, only to learn that the underlying information is flawed.

To avoid that outcome, a fulsome understanding of data is essential. I had the opportunity to address this topic recently at an Economic Measurement Seminar put on by the National Association for Business Economists (NABE).  NABE is the premier organization for people who use economics in their work, and has long been dedicated to improving the quality of the data that we all rely on.

Measuring economic activity is not easy.  Important quantities are sometimes not directly observable, and must be inferred through the use of surveys and assumptions. One leading example is employment, which is gauged in many countries by a series of questions that are posed to households.  Respondents are asked if they are working, and if not, why not.

Assembling adequate data for this exercise has become more difficult.  Response rates to surveys used to explore the labor market have plunged since the pandemic. The process relies on telephone calls, and fewer people are answering. This reduces the quality of measures like job openings and the unemployment rate, which play a very important role in the conduct of economic policy. Any models that rely on these variables are subject to a heightened risk of error.

Correcting for this requires an investment in people and processes.  Unfortunately, economic statistical agencies have routinely struggled to sustain the funding needed to fulfill their missions.  Some budgets have failed to keep pace with inflation, while others have been cut outright.  In some cases, this has led to the curtailment of data collection efforts. 

Objective and comprehensive evidence on economic activity is an essential foundation for sound decision-making.  History is littered with examples where this process has been compromised, to ill effect.

Another one of Tannenbaum’s Laws is that increments of money and time should be applied to areas that yield the best returns.  Some additional investment of both in economic data would yield substantial benefits.