On Friday, the Bureau of Labor Statistics (BLS) reported that a total of 8.4 million jobs were lost since the start of the recession — revising down almost a year’s worth of survey data by adding 1.4 million lost jobs to a previous estimate of 7 million. Also Friday, BLS released a seemingly contradictory figure, that the unemployment rate fell to 9.7 from 10 percent.

The trouble with understanding the U.S. labor situation is that these two key economic indicators are compiled using multiple sources. “Total jobs lost” comes from surveys of businesses that pay payroll taxes and are required to report their monthly personnel gains and losses. The unemployment rate, by contrast, comes from door-to-door surveys of American households where people self-report their employment situation. On average, about 50,000 households are surveyed a month.

Both are imprecise measures, and as evidenced today, jobs lost numbers are frequently revised as new data comes in.

Many experts dismissed the improving numbers as statistical illusions. “There was an inexplicable decline in unemployment in January,” said economist Heidi Shierholz of the Economic Policy Institute, a non-profit Washington-based think tank. While the lower unemployment number is a welcome sign, the drop is “largely attributed to the higher volatility of the … household survey,” she said.

Marc Lieberman dismissed the improved unemployment rate outright. “In a situation like we have now, where the job market actually worsened, the unemployment rate is going down because people are giving up looking for work.” Lieberman, Clinical Professor of Economics at NYU, and an expert in labor economics, also noted that the number of people who want a job, but have not looked in the last 12 months, is rising, a fact unaccounted for in the unemployment numbers.

The monthly jobs lost rate also provides a rough estimate of the number of jobs created by new businesses, because new businesses are usually left out of the monthly surveys. This data comes in later in the year, at which point the BLS can see how well it estimated job creation in sectors like new business, and revise accordingly. Today’s revision of a year’s worth of data could reflect an over-optimistic estimate for job creation in certain sectors. “There’s a lot of information that comes in later — that was not available when they did those previous months of estimates,” Lieberman said.

The conflicting messages and revisions did show that while the economy may still be losing jobs, there are signs that it is losing them at a slower rate. Only 20,000 jobs were lost in January compared to 150,000 in December.