To hear President Trump tell it, the key economic metric by which he should be judged is employment, and that by that metric he’s a remarkable success.

“Unemployment in New Hampshire has reached the lowest rate in actually 31 years,” the president told an audience in that state last week. “But within a very short period of time, it will be the lowest rate in the history of our country. So think of that, you have the best unemployment, you have the most successful state in the history of your state and in the history of our country — and then you’re going to vote for somebody else?”

The unemployment rate in New Hampshire is 2.5 percent — indeed remarkably low. It hit 2.2 percent in March 1988, 31 years ago. Of course, it was also 2.4 percent as recently as May and has been either 2.4 percent or 2.5 percent for a year. Hitting the “lowest rate in the history of our country” seems unlikely.

Unemployment rates have been fairly stagnant in many places over the past year. In September 2018, the national unemployment rate was 3.7 percent, what it is now.

To get a broader look at this (particularly through the lens of the 2020 election), we pulled data from nine swing states — Arizona, Florida, Georgia, Michigan, Minnesota, New Hampshire, North Carolina, Pennsylvania and Wisconsin — and tracked unemployment and employer payrolls over the past several years. On average, the unemployment rates in those nine states haven’t moved over the course of 2019. In Arizona, the unemployment rate has been flat since mid-2017.

What’s more, the drop in unemployment during the 30 months before Trump took office was faster than the drop since he has been in office in seven of the swing states and in the United States overall.

This isn’t necessarily a bad thing. Economists refer to “full employment,” a point at which everyone seeking a job is able to find one. There’s still some unemployment as people move between jobs or stop looking, but it’s expected that unemployment rates would settle at some nonzero point.

When we look at actual employment, the picture is less rosy. The nine swing states above have added jobs at a slower rate than the country on the whole since March 2018 on average.

(For this analysis, we’re looking at a subset of economic data: the number of non-farm employees, seasonally adjusted. This is the population that’s generally considered in employment numbers.)

The nine swing states vary in the rate at which they’ve added jobs since Trump took office. Four — Arizona, Florida, Georgia and North Carolina — have added jobs at a faster clip than the nation overall since January 2017. The other five have added them more slowly.

(Notice that Arizona, where unemployment has been flat, has added jobs faster than any of the other eight states.)

You may recall that three of the five states — Michigan, Pennsylvania and Wisconsin — are alone responsible for Trump’s 2016 electoral college victory. Swing 78,000 votes in those three states to Hillary Clinton, and she would be president. It’s worth noting, then, that employment in those three states is down since January.

In Pennsylvania, the number of non-farm adjusted employees in July was up about 1,000 since January. In Michigan and Wisconsin, the number was down about 4,000 each. It was down about 2,000 in Minnesota, too, a state that the Trump campaign has indicated it hopes to win in 2020.

There are several caveats here, including that drops in employment (especially seasonally adjusted employment) more than a year before an election don’t mean that the incumbent candidate won’t repeat as the winner in a state. But it’s certainly a different picture than the one Trump generally presents of the economy as a perpetual forward-motion machine.

Incidentally, New Hampshire has added 5,000 jobs this year. In July, though, employment was flat.