Thursday, November 12, 2009

Social Security Online Benefit Calculator Leads to Faulty Conclusions

Over at AEI's Enterprise Blog, I commented on a Washington Post article by MIT professor Simon Johnson and Yale law student James Kwak arguing that Social Security and 401(k) plans won't provide for a decent income in retirement. While they have a good qualitative case – after all, Social Security is facing insolvency and 401(k) plans face problems regarding participation rates and investment choices – their estimates of Social Security benefits just seemed too low.

As it turned out, the problem was that they relied on one of SSA's online benefit calculators. As I pointed out last year regarding the Social Security Statement, while the calculator claims to show benefits "in today's dollars" it actually doesn't. It shows benefits in "wage indexed" dollars, which can make for a big difference. Johnson and Kwak were quick to acknowledge the error, which after all wasn't their fault, and discussed the issue further on their Baseline Scenario blog. But it's the source of the error that I'm interested in.

Johnson and Kwak estimate benefits for individuals retiring in 2051. Let's say that the nominal benefit – meaning, the dollar amount that's actually paid each month – was $4000. To inflation adjust that back to today, we multiply $4,000 by the ratio of today's CPI to the projected 2051 CPI: based on this table from the 2009 Trustees Report those numbers are 100/315.37, meaning that the inflation adjusted value of $4,000 is $1,268.

Now let's look at the wage-indexed value, which is what the Social Security Statement will give you and what you get when you choose "in today's dollars' from the online calculator. To get that, you multiply $4,000 by the ratio of today's average wage to the average nominal wage in 2051, which from the same table is $42,042/$209,615, which gives you only $802.

Now, one big problem is that the calculator doesn't even tell you you're getting wage-indexed dollars. So when your scheduled benefit is actually worth $1,268 in today's dollars you'll think you're only getting $802. The fact that an MIT professor and a Yale law student couldn't figure this out seems like pretty good evidence that it's confusing. When I raised this issue last year with regard to the Statement, the agency's solution was to take the phrase "in today's dollars" out to the Statement. That eliminates one problem, but leaves the reader to only guess in what form their benefits might be expressed.

Second, even if there were full disclosure that benefit estimates were in wage-indexed dollars, it's not clear to me whether there's any usefulness in the number. I understand how to calculate wage-indexed dollars, but these figures don't mean anything to me (and I'd guess not to other people either). Moreover, anyone who's actually trying to plan their retirement – which, presumably is what the Statement and the online calculators are for – would express their retirement income either in today's dollars (to show their real purchasing power) or in nominal dollars, to make them comparable to benefit estimates from 401(k) plans or DB pension plans. (If anyone can find me a retirement calculator that expresses income in wage-indexed dollars I'll send you a bottle of wine to celebrate.) Throwing wage-indexed dollars into the mix serves only to confuse people or give them mistaken estimates of their retirement income.

As you might have noticed, this is an issue that ticks me off. There's very little substantive case for showing benefits in wage-indexed dollars – try to explain to someone what they mean and you'll see the blank look on their face. That's probably why there's no effort to explain any of this either on the web sites or in the Statement. While it would be a simple task to fix the calculator and Statement so they would show benefits in inflation adjusted dollars, I suspect the reason this isn't done is that people would ask questions why their benefit estimates had changed from last year to this year. What most people would see as an improvement some people in the agency would perceive as an admission that they'd previously been wrong, and some folks don't like to do that. But that mis-serves the public: Social Security is the largest form of retirement income for most Americans and the only way the typical person can know what they're going to get is if they're told. We should do a better job of telling them than we are.


Pops said...

I am somewhat persuaded by the "lifestyle" argument of the SSA. If you are trying to project your income thirty years from now, what do you really want to know? I don't think it is really how many dollars that you will receive measured in either nominal dollars or today's dollars, but whether you will still be able to live a middle class life or drop to a lower economic status in retirement.

Isn't lifestyle the reason benefits are wage indexed? It makes sense that what you want in retirement is a lifestyle, and by presenting you with wage indexed income estimates, the SSA is telling you what you want to know. If a certain lifestyle costs X dollars today, that same lifestyle will cost more than X (real dollars) in the future. If SSA used X in their benefits estimate, they would be overestimating the lifestyle you would be able to afford.

Andrew G. Biggs said...

There are a couple ways to think about this. One is just practical: if people are calculating their other retirement income in nominal or real dollars, you want to provide that. Even if they're planning on relying solely on Social Security in retirement, you still don't want to tell them a figure is "in today's dollars" when it isn't.

More broadly, your point raises a general question of how people measure their relative well-being. What you're thinking about is called the 'relative income hypothesis', in which people judge their well-being relative to other people's income. So if I have an average income today, I want to have an average income in retirement. This idea was put forward by Duesenberry; it's basically the idea of 'keeping up with the Joneses.' It's not implausible, but's been pretty much overshadowed by the life cycle theory or permanent income hypothesis, in which people plan their future consumption to match their current consumption - in essence, smoothing consumption over time. The idea here is that since the marginal utility of income is declining, shifting resources from a year of higher income to one of lower income maximizes welfare over time. This will TEND to move toward smoothed consumption, but there are obviously other things going. Moreover, there are a lot of people (generally lower income) who don't smooth their consumption; they tend to follow a more-or-less Keynesian consumption function, in which they consume a constant fraction of their incomes. At any rate, the broader point is that the RIH approach, while not implausible, doesn't reflect where most economists would be on these issues.

Pops said...

But how does your argument square with wage indexing benefits? I thought the idea was that if you earned a middle class wage and paid middle class taxes, you get a benefit pegged to replacing part of that middle class lifestyle. If workers want to consumption smooth and spend the same amount of money throughout their lifetimes, why shouldn't benefits be tied to inflation adjusted wages?

Just to be clear, I am not in favor of using inflation adjusted wages in determining benefits. I do want lifestyle replacement, not real dollar replacement.

Andrew G. Biggs said...

I'm not sure the wage indexing of benefits really has a lot to do with things here. Wage indexing means that the ratio of benefits to pre-retirement earnings stays constant over time. That would be true in most pension systems where there's a link between contributions and benefits (401ks are effectively wage indexed, assuming similar investment returns).

Individuals doing their own retirement planning don't particularly care how benefits change from cohort to cohort. They want to know how to maintain their own standard of living in retirement. One way of knowing that is to compare the purchasing power of future income to that of current income, but that implies an inflation-adjusted figure.