Todd Ganos at Forbes Magazine is upset that people earning $150,000 a year get Social Security. Hey, I'm upset that people earning $1 million a year get interest on government bonds. In both cases they don't need it, but you know what? They paid for it.

But let's skip the morality play and have some fun with numbers. We want to figure out how much money we could in principle save by taking away Social Security benefits from these rich retirees. Ganos tells readers:

"According to the Congressional Budget Office, there are approximately 25.6 million senior households in the United States.  The average annual pre-tax income of the top 10% of senior households is approximately $197,000.  Does such a household need Social Security retirement income payments?"

Okay, so the average income for this top 10 percent of seniors is $197,000 a year. Let's suppose that the average income for the top 1 percent of seniors is $1 million a year, which is probably pretty much in the ballpark. That leaves the average income for the remaining 9 percent of these wealthy seniors at just under $110,000. And that is still an average. Many will have considerably lower incomes. Yet, Ganos thinks we can painlessly take away Social Security benefits for this group that average (for them) close to $20,000 a year. He must not have heard all the howls of pain last fall about the job creators being devastated by an increase in their tax rate of 4 percentage points on income over $450,000 a year. 

If we want to be serious about this exercise, we would look at the actual distribution of benefits rather than playing games with averages that include the Warren Buffets of the world. We did this a few years back. If you wanted to cover 10 percent of the benefits paid out (not 10 percent of seniors), you had to get to individuals with incomes of less than $50,000, not counting their Social Security benefits. (We took person income to keep things simpler, many seniors do live alone.)

Now we could tell all these people that we are just expropriating their Social Security checks, as Ganos advocates, but there are two problems. First, most of the people affected do not meet our usual definitions of rich or even particularly well off. The other is that we have just created a huge effective marginal tax rate somewhere near $48,000.

If we assume that the average Social Security benefit is $15,000 a year, then anyone who earns one dollar less than the Ganos cutoff will stand to lose $15,000 in income, plus their marginal tax rate, on their next dollar of income. Usually we don't like to create cliffs like this for the obvious reason, people find ways to avoid them.

A more serious discussion would have a phase out. The question then is how rapid we phase out the benefit and where we want it to actually hit zero. If we want to follow the Ganos line and deny benefits to the richest 10 percent of beneficiaries, then we would want the phase out to be complete somewhere at incomes around $50,000 per person. If we want to eliminate $15k in benefits and we use a 30 percent phase out (each dollar of income costs you 30 cents in benefits), then we would have to start cutting benefits at any positive income level. In other words, at $10,000 you would lose $3,000 in benefits, at $20,000 you would lose $6,000 in benefits, etc.

Even in this case, we have effectively added 30 percentage points to the marginal tax rate -- that's a pretty serious disincentive. For those who are still working this would likely push their marginal tax rate over 60 percent. This measure would be a powerful incentive to hide income for example by buying a condo with savings and using that as a vacation home rather than investing money and paying for hotels. (The accounting industry has tons of tricks like this.) 

Of course you could start the phase out a higher income level (like $50,000 per person) and have it at a more reasonable rate (e.g. 10-20 percent), but then you find that you don't save the program much money. In our paper we found that the savings, net of tax, for a 20 percent phase out starting at person incomes of $40,000 would save around 3 percent of benefits. If it was started at a person income of $100k it would save around 0.6 percent of benefits. These numbers give a much more realistic idea of how much can be saved with means-testing. See how much fun math can be?