One of the challenges of being an economist is that one often works with unreliable data subject to multiple revisions. As pointed out by writer Floyd Norris, errors in the data are most likely to be severe when the economy is turning, such as during the early or late stages of a recession. Figures regarding jobs and unemployment have been particularly scrutinized in recent years. Part of the problem is that labor market statistics come from 2 key surveys, each with its own issues.
Employment numbers emerge from a survey of employers, while the unemployment statistics come from a survey of households. Sometimes, the survey results differ radically though ostensibly they measure the same thing. The employment numbers are revised twice after they are released as late responses are included. These figures include another estimate that comes from what is known as the birth-death model, which reflects the number of jobs created by very new companies and the number of jobs lost when existing companies falter.
Then, more than a year later, the numbers are completely revised to line up with unemployment insurance data. Economists are often faulted for not being able to forecast the future, but it’s hard enough just to know what’s going on right now.
