Hey readers,
Today I'm thinking about the Social Security funding problem in the US. This is, of course, not the main news in the United States, but it remains important, and I have nothing to say on the other news other than the obvious but important points that killing someone is wrong and a politically motivated killing is corrosive to liberal democracy. |
|
|
| Dylan Matthews Senior correspondent and lead writer |
|
|
| Dylan Matthews Senior correspondent and lead writer |
|
|
How to save Social Security without screwing over poor people |
|
|
Raising the retirement age does more to hurt poorer people who live less long. |
|
|
The other day, economist Tyler Cowen made an offhand observation that took me aback a bit: that the French, today, enjoy "the longest financed retirements ever seen in the history of the world." Verifying the "history of the world" part is beyond my historical skill level. That said, the OECD's Pensions at a Glance report from 2023 confirms that French retirees are enjoying a lot of years off the job. French men, per the report, left the labor force at an average age of 60.7. At that point, they have a life expectancy of 84, meaning they can expect 23.3 years in retirement, longer than any of the other countries the OECD examined (mostly rich peer nations plus a few select others). French women can expect 26.1 years in retirement, which is beaten by Luxembourg, Spain, Slovenia, and the world leader, Saudi Arabia, but still very high. (The Saudi case is more about women working fewer and shorter stints than in more liberal polities, as opposed to retirement policy.) French men and women alike can expect over five additional years in retirement compared to Americans. Incidentally, the French government fell this week in part due to opposition parties demanding that the centrist coalition in power go back on its decision to raise the formal retirement age from 62 to 64. Funding 23 to 26 years of retirement per person is expensive, which is exactly why President Emmanuel Macron raised the age in the first place, but when the elderly voter bloc is only growing in size, failing to pay that money out can be politically suicidal.
Retirement, American-style As a non-Frenchman, this fight inevitably makes me think about the coming retirement battle in the US. Our Social Security trust fund is due to be depleted in about eight years. Under current law, when that happens, retirees will see an across-the-board cut of about 23 percent in their benefit levels. Everything I know about how the US government works tells me it will not get to that point. The question, then, is what a deal to prevent those cuts would look like. One obvious way to avoid the French predicament is to do what Macron did: raise the retirement age. There are two components to the aging problem hitting the US and other rich nations' pension systems. One is that, because of the size of the baby boom population, more people are hitting retirement age than ever. The number of retired workers newly receiving Social Security hit 3.4 million in 2022, compared to under 2 million in 2000. Raising the retirement age doesn't solve this issue. But it does partially address the second issue, which is that the average time spent in retirement has risen as nutrition and medicine have improved. A man born in 1900 and turning 65 in 1965 could expect to live 12.9 more years. The Social Security Administration estimates that a man born in 1960 and turning 65 this year can expect 18.4 more years. Even accounting for the trend of people claiming Social Security later in life, that's a good number of additional years that the program has to pay out per male retiree. Between 2000 and 2022, the US gradually raised the retirement age for full Social Security benefits from 65 to 67. But most bipartisan proposals to reform Social Security (that is, proposals with any shot of passage) envision some kind of further age increase. Two years ago, Sens. Angus King (I-ME) and Bill Cassidy (R-LA) floated raising the normal retirement age to 70. The Bipartisan Policy Center brought together some ex-politicians and experts in both parties to put together a plan, which wound up advocating an age of 69. One of the key political virtues of a retirement age increase is that it's a benefit cut that doesn't present itself quite as obviously as a benefit cut. But it does amount to a cut, and potentially a large one. Right now, a 67-year-old woman can expect to live 18.5 more years. Suppose she has to wait until age 70 to claim the same amount of benefits she can now claim at 67. That eats up three of her 18.5 years of expected benefits, an over 16 percent cut. The cut for men, with our shorter lifespans, is even larger in percentage terms. The most important question to ask about it, though, is whether it's an across-the-board benefit cut, or in fact a regressive one. There are strong arguments that it is the latter. Death inequality and Social Security
The eminent Social Security expert and economist Alice Munnell recently highlighted a chart from the program's actuary's office that underlined a pretty concerning gap and trend: |
If you don't speak Social Security jargon, this can be a little hard to parse. Essentially, it's comparing two groups: men born in 1930 considering retirement in 1992 and men born in 1960 considering retirement in 2022. In both groups there is a large gap in life expectancy between the people who earned the least in their careers and those who earned the most. In 1992, the highest-earning men could expect to live 8.4 years longer than the lowest-earning men. In 2022, they could expect 10.3 more years. ("Highest-earning" here means the highest-earning fifth, This is not exactly Elon Musk money: in 2020, being in the top quintile as a man meant an average monthly income of at least $6,391, or $76,692 annually.) Put differently: not only is there a big life expectancy gap between rich and poor people, but also the gap seems to be growing. This puts retirement age discussions in a different light. Suppose we're considering raising not the normal retirement age (now 67) but the early age (now 62), at which point retirees can claim reduced benefits. If we raise the age by three years, then men in the highest income bracket get a cut of 3 divided by 25.6, or about 11 percent. Men in the lowest income bracket get a cut of 3 divided by 15.3, or almost 20 percent. The specific numbers are different if you're considering raising the normal retirement age, or looking at female workers, but the overall takeaway is the same: raising the age of retirement amounts to a bigger cut for poorer workers. Recently, economists Henry Aaron at Brookings and Mark Warshawsky got into a heated dispute about how to make sense of these numbers. Warshawsky argues against using life expectancy numbers like those above on the grounds that they inevitably require one to make projections (we don't know, of course, how long people who retired in 2022 will in fact live, chiefly because most of them haven't died yet), and for restricting analysis to men aged 65-69. Aaron argues that this is too restrictive (everyone, including insurers, relies heavily on life expectancy projections as well) and neglects that women, for instance, have seen lifespan inequality increase. To my non-expert eye, Aaron has the better of this specific dispute. But it's worth emphasizing that the lifespan gap between rich and poor need not be increasing in order for hiking the retirement age to be regressive on net. If, in 30 years, rich men are still living 10 more years in retirement than poor men, an increase in the retirement age will still hit poor men harder than rich men, even if the gap itself hasn't grown. Splitting the difference The traditional Republican approach to Social Security has been to call for its shortfall to be closed entirely with benefit cuts; the traditional Democratic approach has been to rely entirely on tax hikes. Neither of these has any shot in hell of happening, especially if the Senate filibuster remains in place. I highly doubt that there are 50 Republicans in the Senate now willing to vote for major benefit cuts, and there certainly aren't the 60 that would actually be needed. Similarly, I put the odds of Democrats ever electing 60 senators willing to pass a huge payroll tax hike, even just on top earners, at near zero. If there's going to be reform before the trust fund runs out in 2033, it's going to have to be on a bipartisan basis and involve pretty huge concessions by each side. And I suspect some kind of a retirement age increase will be part of the deal. If that happens, the best option out there is one that Wendell Primus, Tara Watson, and Jack Smalligan outline in their recent Brookings reform plan. They would raise the retirement age — but only for the top 40 percent of earners. Most retirees would not see the age rise at all, while the top fifth of earners would see it rise to 70. Those in the 60th to 80th percentiles would see smaller hikes. Along with other progressive benefit cuts and tax hikes, the plan would fix the program's solvency issue. This retirement age change would make the system somewhat more complicated, as people would have to look up what their specific retirement age is based on their income. But it's the only plan I've seen that keeps the most popular kind of benefit cut from being painfully regressive. |
|
|
- I was intrigued by a finding in a new paper by UChicago economists Brian Jabarian and Alex Imas that software suites for detecting whether papers are written by humans or AI actually work very, very well. I had assumed that distinguishing between a human and a sufficiently advanced AI was essentially impossible; it's not like you can put an AI watermark on a block of text. But Jabarian and Imas find that top detectors have very low rates of both false positives (human-written material being marked as AI-written) and false negatives (AI-written material being marked as human-written). The best one they tried, Pangram, had false positive rates of, at worst, 0.75 percent (so human-written text would be wrongly flagged less than 1 percent of the time), and for longer pieces it falls to literally zero. The false negative rates are similarly tiny.
- Some debates never end. Case in point: the orthodoxy vs. revisionism debate in the late 19th century in the German Social Democratic Party. On one side were the orthodox, led by Karl Kautsky, who held to a very literal interpretation of Marx and Engels' writings and predictions that capitalism would trend toward inevitably greater inequality, immiseration of workers, and subsequent growth in support for socialist parties; on the other were revisionists led by Eduard Bernstein, who argued that that spiral toward worker immiseration was not inevitable and that socialists should reorient themselves toward supporting trade unions and advocating pro-worker reforms. The economists Charlotte Bartels, Felix Kersting, and Nikolaus Wolf have a clever new paper boiling down the debate to three testable hypotheses about the German economy, and find that the revisionists come out ahead.
- The most obvious way for the US to deal with people put out of work by AI is through redistribution: tax the gains from AI, give the money to displaced workers somehow. One nice feature of this approach is that it scales up automatically with the scale of the problem: An AI disruption that causes large-scale job loss will, by definition, cause a lot of economic growth too, which means a larger tax base from which to redistribute. While this may be a nice dynamic for the United States, Anton Leicht has a sobering essay explaining why it could screw over middle and low-income countries. Most of the AI-driven economic growth, he argues, will concentrate in rich countries, while most of the job loss will concentrate in poor ones. The "losers" in the US may be compensated, but a Filipino call center worker put out of a job by a voice AI won't be.
|
|
|
Want more Future Perfect in your inbox? Sign up for more newsletters here. Let us know what you think about this newsletter by hitting reply to this email. |
|
|
|
