Back in the late 1970s I did an expose on various government statistics including the unemployment stats, and the consumer price index, and the housing starts stats... they led to many of the major changes instituted by the government.
I coined the phrase "phantom unemployment" in my reports for CBS and The Washington Journalism Review to discuss "discouraged workers" and why they were not included in the count of the unemployed.
But what is really revealing is that the government's unemployment numbers are based on household surveys by the Census Bureau -- and they are not a count of those receiving unemployment benefits, or government reports about layoffs or hires, or anything else. Back in the 1970s the government surveyed 40,000 households each month (I don't know what the current figure is) and the opportunity to skew the numbers with the households surveyed was huge.
I stumbled upon the "phantom unemployment" phenomenon when I was reporting on the Smith Corona (the typewriter company) layoff in Cortland, New York. Smith Corona employed more than 40% of the workers in Cortland but the three week layoff failed to show up in the unemployment stats for Cortland or for New York State. Why?
Well, the Census Bureau conducts its survey in the week containing the 12th of each month. The Smith Corona layoff started the day after that "reporting week" and ended the day before the following month's "reporting week" thereby escaping the statistics.
Why would they plan a layoff that way? The obvious reasons is what I alleged. Would Smith Corona want to be held responsible for a huge increase in the unemployment rate in Cortland County or in New York State? Instead of an unemployment rate of 40+% during the layoff the "official rate" was only about 5%.
Remember: there are lies, damned lies, and statistics.