Jatinder Singh drives his taxi in New York this summer. Singh owned a yellow cab, which he decided to paint black so he could start driving with Uber. He lasted about six months before he sold the car and returned to taxi driving.

The Pew Research Center has a new report today that looks at self-employment in the U.S. and its role in creating jobs. But the data in the report exposes one of the more perplexing mysteries about the modern economy: Despite all the discussion that the U.S. is transforming into a gig economy, the share of people who say they’re self-employed is declining.

This trend, especially among unincorporated, self-employed workers, has been in place for some time. We noted over the summer that it’s hard to find high-quality statistics backing up the idea that the gig economy has grown enormously. In fact, it appears that the overall economy is getting slightly less “giggy” over time.

The Pew report provides a detailed breakdown of the demographics of the self-employed and how many of the self-employed have employees of their own. But the report doesn’t attempt to answer why the supposedly widespread gig economy is so elusive.

One theory is that the data comes out with a lag and the rise of the gig economy will soon become apparent. But Pew’s preliminary estimate of last year’s self-employment finds another slight decrease.

What are possible explanations for this?

One possibility is that while some very high-profile companies are using the gig model, it’s actually not very common when you look across the entire 157 million-person U.S. labor force. (People often use the expression “Uber and its ilk,” as if there are dozens or hundreds of companies just as large as Uber in every industry. It could be that a fairly small handful of these companies actually have any significant scale.)

It could be that gig economy companies are flourishing in industries that have always had a large number of independent contractors. Meanwhile, self-employment in industries like agriculture is in sharp decline. For example, Thumbtack is an online marketplace where people can hire local contractors and other service professionals, and as long as there were yellow pages in the phone book, these industries have had gig workers. Thumbtack’s chief economist, Jon Lieber, recently testified to Congress that “about half of the service professionals on Thumbtack have been in business for themselves for five or more years.” In other words, many people were always “gig economy” workers and now they use new websites to find business.

The Bay Area Council Economic Institute has documented that there’s been a rise in people filing tax returns that report miscellaneous income, the sort you might expect from “gig economy” work. Yet another possibility is that many employees of Uber and its ilk consider themselves employees, even if the companies adamantly insist otherwise.

One final challenge worth mentioning is that the Bureau of Labor Statistics has occasionally conducted a survey of Contingent and Alternative Employment Arrangements, that provides greater detail on the type of jobs in question. Congress, however, has not granted funding for this since 2005.

So for now, the best surveys puzzlingly suggest fewer workers each year, not more, who are self-employed.