You won’t hear many of the points below, however, either because they are too positive, too subtle to capture in a sound bite, or too long-term to worry about now. But if you want the truth about American jobs, you may want to keep reading rather than tuning into a presidential debate or stump speech.

Times Are Good

It’s hardly a bad time for jobs overall—official unemployment is at the lowest rate since 2008, and applications for unemployment insurance were at the lowest level since 1973. You won’t hear that from the Republican candidates, of course—the news is much too sanguine. If there’s a problem, it is with low labor force participation, which has dropped several percentage points over the last decade. As testimony before a Joint Economic Committee suggests, there are a variety of reasons for this, although an aging population is clearly part of the issue.

The Lack of Middle Class Jobs

Growing inequality is another big issue—the hallmark of Bernie Sanders’ campaign, and an issue that is occasionally even mentioned by Republican candidates. An overall lack of jobs is less of a problem for our economy, however, than the shortage of well-paying, middle class jobs.

Many of the new jobs in the US over the last decade or two have been low-end service jobs—feeding, selling to, and taking care of our fellow Americans. These jobs don’t generally produce a lot of revenue or grow much in productivity, so it’s difficult for businesses to pay high wages to workers who perform them. And that situation isn’t going to change much no matter who is president. As a New York Times article points out, our economy is a service economy, and it’s been going in that direction for several decades now.

Free Trade Isn’t the Problem

There is little doubt that some of our well-paying manufacturing jobs have gone to places like Mexico and China. But recent trade deals aren’t the problem, capitalism is. For several centuries now, manufacturing has moved to places where it’s done most cheaply and effectively. And for most industries, that place hasn’t been the U.S. for a while.

Donald Trump’s bluster notwithstanding, it seems highly unlikely that any president could change this much. As a society we don’t generally like to tell businesspeople where they can produce stuff, and we don’t like paying high tariffs for stuff made elsewhere. Corporate moves to low-tax countries do some damage to our tax base, but they don’t tend to involve many jobs.

Automation Is Good

Ironically, our best hope for bringing manufacturing jobs home is automation. Also ironically, the same is true for outsourced service jobs. Since the U.S. is one of the world’s leaders in automation technologies, we’ve got some chance of bringing jobs home if we have the best automation capabilities and the people who are best at working with automation technologies.

In factories, this means technologists who can install, maintain, and optimize robots, CAD/CAM, and flexible manufacturing cells. In services, it means working with tools like robotic process automation and cognitive technologies that can do work previously outsourced to low-wage countries. Whether in factories or offices, all those automation machines need people to configure, install, and tend them. Because most of the actual work is done by machines, there won’t be huge numbers of jobs that will come home this way, but there will be some.

Information Technology is a Drag

Information technology hasn’t eliminated a lot of jobs, but it is keeping some from growing. Take bank tellers, for example. Contrary to what President Obama said in a 2014 interview, automated teller machines haven’t led to many fewer jobs in that field. Mr. Obama may not run into them when he needs cash, but the number of bank tellers in the U.S. has remained pretty constant since 1980, when ATMs really took off. Of course, the U.S. has grown a lot in population since 1980, but the number of tellers hasn’t. The Bureau of Labor Statistics estimates that the number of tellers will decline about 8% over the next decade—not a precipitous decline, but certainly not a growth occupation either.

The Service Sector Slowdown

In the future, many lower-level service jobs are going to stop growing—if not decline—because of technology. There are already some pretty capable technologies that automate truck and taxi driving, hamburger ordering and preparation, lawn mowing, and so forth. They’ll get better and cheaper over time. Human workers will get more expensive. If I were in one of these fields, I’d learn how to sell, install, or fix these machines.

Professionals Should Be Wary

The same is true for many professional jobs. There are already smart machines that can do some of the key tasks of lawyers, doctors, marketers, journalists, and even scientists. Not all of the incumbents of these knowledge work roles will lose their jobs, but some will. If you don’t want to be one of the losers, again, your best bet is to make sure you know how to work alongside these machines and add value to them.

The lessons from all this are pretty clear. If you’re voting for your preferred candidate because you believe they can reverse the economic trends of the last several decades, you might want to reconsider your choice or your reason for it. You may also want to think about whether you want your candidate to bring about more jobs or less inequality. The policies that lead to these objectives are different (though they are both quite difficult to achieve), and candidates tend to emphasize one or the other.

Whether you prefer lower unemployment or higher incomes, helping Americans get more college degrees is still a good idea, although it probably won’t be the boon that it’s been in the past. And if you want to be assured of a job over the next few decades, learn how to work closely with smart machines and to do something they can’t.

Thomas H. Davenport is a Distinguished Professor at Babson College and a Fellow of the MIT Initiative on the Digital. With Julia Kirby, he is the author of Only Humans Need Apply: Winners and Losers in the Age of Smart Machines (Harper Business, 2016).