Life insurance is losing its appeal in the U.S. In 1965, Americans purchased 27 million policies, individually or through employers. In 2016, a population that was more than 50 percent larger still bought only 27 million policies. The share of Americans with life insurance has fallen to less than 60 percent, from 77 percent in 1989. Why this is happening remains a puzzle.

People buy life insurance for various reasons: to pass wealth along to future generations, to provide liquidity for mortgage payments, or to cover funeral expenses, to name a few. These motivations may become more or less important as the population shifts demographically.

Yet socioeconomic and demographic trends can’t explain the decline in life insurance, a recent analysis from the Federal Reserve Bank of Chicago has found: If various population groups had acted the same way in 2013 as they did in 1989, 78 percent of U.S. households would have had life insurance, not 60 percent.

Other evidence points in the same direction. The observed declines have been steeper for cash value life insurance, which includes a saving component, than they have been for term life, which does not. Another study looking specifically at cash value ownership found that neither changes in demographics nor in the tax law (which can affect the incentives to hold cash value policies) can explain the declines from 1992 to 2010.

The puzzle deepens when one examines life expectancy, which clearly should influence decisions about life insurance. Theoretically, the lower a person’s chance of dying over a given period, the less should be his or her desire for life insurance during that time. And over the past few decades, overall life expectancy has risen.

But this otherwise plausible explanation doesn’t work when you take a closer look and see that life expectancy has been rising rapidly only among higher earners. For lower earners, it has been stagnating or even declining. The top 40 percent of male earners who reached age 50 in 2010 could expect to live seven to eight years longer than those who reached that age in 1980. But there was little to no increase for the bottom 40 percent of male earners across those generations, a National Academies of Sciences panel that I co-chaired found.

If life insurance changes were being driven by life expectancy, we would expect ownership to fall less (or perhaps even rise) among lower earners and to fall more among higher earners. Instead, the opposite has happened.

In 1989, 76 percent of Americans with a high school diploma owned any kind of life insurance. By 2013, that share had declined to 55 percent. For those with a college degree, ownership fell only to 73 percent, from 88 percent. Similarly, among people in the top 20 percent of the income distribution, life insurance ownership fell to 85 percent from 94 percent, while it dropped to 27 percent, from 44 percent, among those in the bottom 20 percent of income.

Perhaps people in low-income households can no longer afford policies, or they don’t consider it as necessary as they once did to protect against financial risk to their families. Another possibility, though, is that policy pricing is having an effect.

Most individual life insurance policies require a medical exam. If the health of lower earners is deteriorating relative to that of higher earners, the price of life insurance for them will rise disproportionately. And if life insurance companies put more weight on the risks to life than the individuals do, they’ll end up with policy pricing that’s unattractive to lower earners.

It is also possible that industry changes have affected life insurance purchases. Over the past two decades, many insurance companies “demutualized” by shifting from being owned by policyholders to being owned by shareholders. Mutual insurance companies appear more inclined to sell life insurance, and so this broader industry trend may be affecting how policies are advertised and sold. Evidence suggests that term life policies became cheaper as they became more widely available on the internet, which may be why term policies have declined less dramatically than cash value policies have.

Finally, although fewer people are buying life insurance, those who do are buying more valuable policies. Apparently, while some families are deciding insurance isn't worth buying, others consider it such a good idea, they're buying more. That only makes the puzzle harder to solve.

This column was provided by Bloomberg News and does not necessarily reflect the opinion of the editorial board or Bloomberg LP and its owners. Peter R. Orszag is a Bloomberg View columnist. He is a vice chairman of investment banking at Lazard. He was President Barack Obama’s director of the Office of Management and Budget from 2009 to 2010 and the director of the Congressional Budget Office from 2007 to 2008.