Monday, December 8, 2008

Does Social Security actually spread risk across generations?

Two of the better arguments for the Social Security program are the ability of a government-run program to spread risk both within and between generations. Spreading risk within a generation means redistributing from higher to lower-earning members of the same birth cohort. Spreading risk across generations means redistributing from relatively high earning birth cohorts to relatively low-earning ones. Both of these risk sharing functions are potentially very valuable, yet neither is easily accomplished through private markets. This creates a potentially strong argument for a pay-as-you-go program like Social Security: it can do things that a simple forced savings plan couldn't do.

I've recently been doing some work on the ability of Social Security to spread risk within generations. (I'll have an AEI paper out on that shortly, or you can click here for a preview of some of the basic thoughts.) Intra-cohort risk – the risk of having low earnings relative to other members of your cohort – is addressed through a progressive benefit schedule. Individuals with low lifetime earnings receive higher benefits relative to their earnings – a higher "replacement rate" – than do higher earners. However, as this chart shows, there's actually a lot of variation in replacement rates even for people with the same earnings, due to quirks in the benefit formula. As a result, Social Security's "insurance" against relatively low earnings doesn't work nearly as well as it could.

But here I'm interested in looking at Social Security's ability to spread risk across generations. Social Security is often called a "compact between generations." As a pay-as-you-go program, in which each working generation funds the retirement benefits of the generation that preceded it, Social Security could in theory smooth transfer resources between richer and poorer generations. A relatively poor generation, for instance, would be subsidized by a relatively rich generation that followed. The risk that you'll belong to a relatively poor generation – one that lives through times in which average earnings and economic growth are low – can't be diversified away through any ordinary means of investment or insurance.

So Social Security can accomplish intergenerational risk sharing in theory. But does it do so in practice? I don't think so. Here's why.

Under Social Security, average retirement benefits for a given cohort are calculated based on average wage growth during that cohort's working years. Here's how it works: when you retire, your past earnings are "wage indexed" to the year you turned 60. For instance, if you earned $10,000 at age 25 and average wages economy-wide grew by 3 percent annually from that year through age 60, then that $10,000 wage would be "indexed" to $28,139. This process is followed for all your past wages, which are then averaged. For the typical worker, Social Security pays a benefit equal to around 40 percent of the average indexed wages. The key here is that your retirement benefit depends both on your own wages and on the growth of average wages in the economy. The faster average wages grew during your working years, the higher your own benefit will be in retirement. The rate of average wage growth is like an interest rate that's paid on your past earnings and Social Security contributions.

But here's the problem: the earnings of future workers have no effect on the benefits owed to contemporaneous retirees. Wage growth for Generation B play no role in the benefit calculation for Generation A. For instance, imagine that average wages grew only 1 percent annually during your working years, but that in the year you claimed benefits wage growth took off and continued at a 5 percent rate throughout your retirement. Unfortunately, you would not share in any of that wage growth.

So it seems that in ordinary circumstances Social Security doesn't do much to spread aggregate earnings risk between generations. It may be financed cross-generationally, but the benefits a given generation is entitled to aren't dependent on the earnings of following generations.

Would it be possible for Social Security to spread risk more effectively between generations? Yes. Consider a reform plan where retirees receive a benefit equal to a given percentage of the average wage at the time. For instance, retirees in 2008 might receive a benefit equal to 25 percent of the average wage earned by workers in 2008. This would produce an annual benefit about equal to the benefits received under current law. However, the benefits received by retirees in 2008 would depend on average earnings in 2008, not on the retirees average earnings during their working years. If earnings rise more rapidly, that extra wage growth will be shared with current retirees. But if earnings grow only slowly, retirees will also share some of that pain.

The advantage here is diversification: most retirees will have some non-Social Security savings, the level of which depends on how much they earned while working and on the interest rates their investments received. Deriving part of their retirement income from a different source, the earnings of future workers, allows them to diversify against the risk of their birth cohort having low earnings growth or receiving poor returns on their investments.

But again, it appears that current law Social Security doesn't really take advantage of this opportunity to smooth risk across generations.


Arne said...

One could argue that the generations most at risk of not having enough were the initial generations. Their risk was spread to future generations.

Andrew G. Biggs said...

Clearly the initial generation was the worst off and also received a significant net transfer from Social Security. But that's a somewhat different issue than the risk sharing discussed here, which is a question of what policies we can put in place to counter variations in relative incomes between future generations. I think it's possible to have policies like that, but Social Security currently doesn't do a very good job of things. There's some small inter-generational risk sharing, but not very much.

MS said...

Why not just index benefits to wages after they are claimed? It's a much simpler way to achieve a constant replacement rate relative to current average wages.

Of course, this type of change may require an initial benefit cut in order to be cost neutral.

Andrew G. Biggs said...

Under the idea I outlined, benefits post-claiming would effectively be indexed to wages. And you're right that to make the numbers add up, initial benefits would have to be lower. But this actually isn't bad policy, since it would encourage people to claim later and would increase benefits for the very old, who are at greater risk of poverty.

I'm not 100% sure that simply indexing to wages post-claiming would be the same as paying a benefit equal to x% of the contemporary average wage, though it might be.