It feels like it’s taking forever for the U.S. economy to add back the jobs it lost in the Great Recession. In historical terms, that’s true, and new research suggests it’s partly because of a half-century shift from manufacturing to services as the engine of the economy.

Even with the relatively good news Friday that the economy added 165,000 net jobs in April, the United States is 2.5 million jobs short of where it was when the recession officially began in December 2007. It has been 45 months since the recession ended. Forty-five.

The recession was the deepest one since the Great Depression. But slow job recoveries have become a defining trend of the past three recessions. It took the economy 37 months from the end of the 2001 recession to add back the jobs it lost. For the recession before that, which ended in 1991, it took 23 months. Yet about two decades earlier, it took just nine months to regain the jobs lost in the recession that ended in 1975 and six months for the one that ended in 1970.

New research by economists Martha Olney, at the University of California at Berkeley, and Aaron Pacitti, at Siena College, suggests that the economy is slower to recover jobs today because it has grown far more dependent on people doing things as opposed to making things.

Goods production supplied about three-fifths of economic output in 1950 and about half of its jobs. By 2010, growth in the service sector has accounted for two-thirds of output and seven out of every 10 jobs.

Olney and Pacitti estimate that because of that shift, the march back to pre-recession employment levels will last about a year longer than it would have a halfcentury ago. They base their analysis on a study of 50 years of U.S. recessions, along with 30 years of data on how states — with different economic compositions — rebounded from their own recessions.

To understand why a service-oriented economy would recover more slowly than a production-oriented one, it helps to think about the psychology of recessions.

This is the classic story Olney told her beginning economics students at Berkeley for 30 years: There comes a point, after an economy has been contracting, when factory owners start to anticipate that better times are around the corner. So they ramp up output, which puts people back to work, pumps more money into the economy and creates a virtuous cycle of output, employment and growth.

It’s a different story for service providers. They don’t anticipate new demand; they wait for it to appear and then hire workers to handle it. Think of an ice cream shop owner: “They don’t want to scoop the ice cream and leave it there, hoping that you’ll walk in the door,” Olney said in an interview. “If you don’t walk in the door, it will melt.”

Manufacturers don’t run the same risk, Pacitti said. “Microwaves don’t melt on the store shelf.”

So for a service-based economy coming out of recession, there’s relatively little hiring in advance based on expectations of new business, and there’s no positive feedback loop from that hiring. That delays a rebound in jobs, compared with a manufacturing-heavy economy.

The trend is easy to spot on the state level, particularly in Delaware, the most extreme example of a shift to services. In the late 1960s, manufacturing was nearly three-fifths of the state’s economy. The state shed jobs in the recession that began in 1969, but it needed only three months to replenish them, Olney and Pacitti calculate.

Today, manufacturing has shrunk to less than one-third of the Delaware economy. The state needed more than three years from the start of the 2001 recession to get back to its previous jobs peak. Forty-five months after the end of the last recession, the state has only made up half the jobs it lost.