What exercises produce the best health benefits? The scientific studies don’t provide much help

By Emily Oster

February is the month things go south in our exercise routines. The excitement of New Year’s resolutions fades and the dark and cold keep us in bed, rather than on the treadmill.

This is a shame, because regular exercise is really beneficial. It can control high blood pressure, improve mental health and reduce falls among older adults. A review of 187 randomized controlled trials covering nearly 30,000 people found exercise lowered mortality risk by 13 percent.

Given the fact that people struggle to stick with exercise, the crucial question is: How can we design fitness programs that maximize long-term adherence?

Unfortunately, a lot of research and media coverage is focused instead on asking what the best type of exercise or the optimal amount is. Is walking better than tennis? Is running better than swimming? Is it really important to do 80 percent of your workout at 60-70 percent of your max heart rate (known as Zone 2)? These questions serve an engaged population and promise that with a bit more knowledge, you can maximize your health. The trouble is, they are basically impossible to answer well, and most of the answers we get are misleading and wrong.

The best way to discover the optimal exercise regimen would be with large randomized trials. While we have high-quality studies comparing some exercise intervention (often a combo of aerobic exercise and weight lifting) to doing nothing, it’s far harder to find trials comparing two different types of exercise head-to-head. That’s because you’d need enormous sample sizes to detect the — most likely small — differences in a statistically meaningful way. Big trials are expensive. And not everyone wants to stick to a randomly assigned exercise program.

Because of these challenges, most of the evidence on the so-called best exercise comes from what are known as observational studies. Instead of randomizing people to different programs, these studies ask people about the exercise they already do and compare their health outcomes to those of people who do more or less exercise, or who do different types of exercise. The trouble is that such studies almost always confuse correlation and causation.

To take a specific example: A recent paper found that swimming didn’t lower mortality, but running did. Does that mean swimming isn’t effective? No. When you look more closely at the data in the appendix of the study, it’s clear what is going on — swimmers were much less likely than runners to engage in other positive health behaviors.

For instance, only about 5 percent of frequent runners in the study smoked, but among frequent swimmers it’s 11 percent — just shy of the U.S. average. Swimmers were also more likely to have high blood pressure at base line, weigh more and have a family history of cancer. One explanation is that doctors may recommend swimming as a low-impact activity for people who already have higher health risks. Even if the swimming helps, the other differences among people hide that. (Although the authors adjust for differences in the variables they can see, there are many differences they can’t observe, and those will drive their results.)

It’s also why you should be very skeptical of the conclusion, from the same paper, that walking and running a lot are beneficial, but frequent jogging (essentially slow running or fast walking) is not. A study like this tells us plenty about the types of people who do different exercises, but that’s not actionable. People want to know what would happen if they changed their behavior.

These issues are common in exercise research. A 2019 study argued that the benefits of walking leveled off at 7,500 steps a day, suggesting that this was a threshold number. But when you look at the paper, it’s clear that the people who walk more are also more likely to do other healthy behaviors (like abstaining from smoking and eating well). A 2025 paper used wearable devices to suggest that one minute of vigorous activity is equivalent to an hour of light activity. But it doesn’t sufficiently take into account that the mere ability to engage in vigorous activity is most likely already a sign of better health, or the fact that (like with the earlier study) there are other clear differences across groups. Again, the authors make statistical adjustments to account for the differences they can see in the data, but they can’t control for all the factors associated with different exercise patterns.

I could go on. People love findings like these, but the reality is that when we see such claims we should assume that they don’t tell us much. They can even backfire — making people feel inadequate about exercise programs that are actually fine.

The limits of our knowledge are a tough pill to swallow (especially since it’s not just exercise; the research on nutrition is equally shaky). It can be frustrating to know there are some questions we’ll never answer. I also think it can be freeing, an invitation to let go of optimization and focus on what we do know. In the case of exercise, we can say with confidence that it’s good to regularly do something that raises your heart rate.

People also want to know how much exercise to do; this is also very hard to know. The randomized studies showing benefits of exercise typically aim for 2.5 hours per week. After that, we are mostly relying on lower-quality observational studies. Those show that the benefits of exercise tend to flatten around seven to 10 hours a week.

What we do know is that people are more likely to stick with an exercise program that is tailored to their preferences and lifestyle. When doctors are advising patients about exercise, they should be as open as possible to the variety of ways people might want to get moving. Research should focus on which approaches are best to lead to long-term, sustained exercise habits — ideally, research with randomized trials. The type of flawed observational research discussed above gets in the way of both of these priorities — it takes research time from studying what matters, and it creates misleading and constraining advice. It doesn’t really matter which exercise people do — they just need to do it past February.
________________

Credit: New York Times