At 8:30 a.m. on the first Friday of each month, political and financial analysts across the country begin refreshing their browsers maniacally. That's when the Bureau of Labor Statistics releases its monthly jobs numbers — one of the most closely-watched indicators of the state of the U.S. economy. 

Having a hard time counting. (Tim Boyle/Bloomberg)

Yet it turns out that the monthly payroll numbers are often wrong. In fact, over the past year, they've been off by quite a lot. On Thursday, BLS released newly revised data showing that the U.S. economy appears to have added 386,000 more jobs between April 2011 and March 2012 than the monthly payroll survey reports had been showing. That would suggest that, in the last year, those monthly BLS payroll reports have been understating job growth by roughly 32,000 jobs per month, on average.

Those extra jobs can add up: If the U.S. economy had in fact been adding, on average, 194,000 new jobs each month over that period instead of 162,000, then the country would be on pace to get back to full employment by 2021 rather than after 2025, according to the Hamilton Project's calculator. (Though that's still a slow, anemic recovery by either measure.)

Economist Justin Wolfers wonders why these revisions aren't generating nearly as much fanfare. "Amazing how much attention a 30k miss on a monthly jobs number gets," he writes, "and how little follows when 386k more jobs are found in a re-benchmark." Wolfers adds that this revision could help explain why the U.S. unemployment rate has been nudging downward despite the seemingly weak jobs growth.

According to the new BLS numbers, the United States has shed 67,000 more government jobs than thought as of March 2012. Private-sector jobs, by contrast, grew by 453,000 more than originally calculated.

What explains the change? Once a year, the BLS creates a "benchmark" for its jobs numbers by poring over tax records from companies across the country to find out how many people were actually working in March 2012. This helps the agency generate more accurate numbers, but the process takes about 10 months to complete. So, for the monthly payroll surveys, the BLS simply relies on smaller, less accurate surveys of businesses and government agencies. The agency will publish its final benchmark revisions Feb. 1, 2013.

Bill McBride of Calculated Risk provides a table of previous BLS benchmarks revisions going back to 1979. The changes can often be quite dramatic — in both directions. In 2009, for instance, the economy actually lost 902,000 more jobs than the monthly jobs reports had initially suggested.

The fact that these large revisions occur every year suggests that we could all probably stand to put less emphasis on the monthly jobs reports. They're almost never the last word on how the economy's actually performing.

Related: Why August's jobs number is likely wrong.