Friday, January 14, 2005

On the stuff and nonsense about the Social Security Trust Fund guaranteeing payment of currently promised benefits -- or any level of benefits -- from 2018 on.

This is a quick response to an odd idea that's suddenly grown very popular among some defenders of the Social Security status quo, and is circulating among them merrily.

Namely, that the fact that the Social Security Trust Fund holds U.S. government bonds -- which on current law would be redeemed to finance currently promised benefits during the period from 2018 to 2042 or thereabouts -- somehow guarantees the payment of such benefits, because failing to redeem the bonds to fund the benefits would be an unthinkable default on US Treasury obligations.

Claims to this effect are all over the Internet and the liberal press these days. Here's two I just noticed. From Max Sawicky a couple days ago...
[Social Security] can only be insolvent, or "short of cash," if the Trust Fund assets are not redeemed, and to fail to redeem them is to default. [my emphasis]
And in the comments today over at Andrew Samwick's:
"The US government has the unparalleled record of having never missed a debt payment since the founding of the country. If you believe the US government is about to start missing those payments ... you should not be worrying about a fractional reduction in Social Security payments 20 years from now - you should be stocking up on guns and gold."
Look ... if failure to redeem these bonds to provide cash to pay Social Security benefits is an unthinkable "default" by the US Treasury on its obligations, then the Treasury is already defaulting on them every year -- because the bonds in the trust fund are 15-year bonds and none of them have been redeemed in 22 years! Default???

Why, no, not default. Because these bonds -- despite frequent claims along the lines that "they have the same status as U.S. bonds owned by Japanese pension funds and the government of China", as per Krugman -- are in fact very different from Treasury bonds held by the public (including foreigners and foreign governments) in a good number of ways.

One such way is that the trust fund bonds are demand bonds. That is, although they nominally are 15-year bonds for such purposes as setting the interest rate on them, they can be cashed in at the Treasury on demand, before they mature -- and if they aren't cashed in on demand they simply roll over for another term.

You cannot buy US Treasury bonds like this, and neither can Japanese pension funds nor the Chinese government.

Now, let's imagine that come 2018 or so Congress finds itself facing the need to pass one big honking income tax increase to pay off the $5 trillion or so of bonds that will then be in the trust fund (today's current value) over about the next 25 years. That may well be an income tax increase on the order of 20% or more (see below), on top of even larger income tax increases needed simultaneously to finance Medicare.

And let's imagine that the voting taxpaying public for some reason is reluctant to pay all of such a big tax increase -- and that Congress responds to the people's wishes by reducing, to a greater or lesser extent, Social Security benefits, thus reducing the size of the necessary tax increase.

Now fewer (if any) bonds will be presented by Social Security to the Treasury for redemption. Bonds will not be redeemed. Must this failure to redeem the bonds be default, as Max said?

Of course not. The bonds that needn't be cashed to fund current benefits will simply be rolled over -- exactly as they were last year, and will be this year and next year too. What default??

The bonds can be rolled over until the year 2500, until whenever, or until forever, without any bond default ever occurring -- just as they are being rolled over now.

So the idea that the alternative horrible consequences of a bond default somehow secure the payment of the currently promised level of Social Security benefits -- or of any given level of benefits -- until the bonds run out is simply nonsense. Among amateur observers the idea may be described as "naive." But professional economists and political observers who peddle such nonsense deserve, I think, rather harsher criticism.

(All this is just another reflection of how the strange notion that the government can default on an obligation to itself -- any more than anyone else can -- is a logical absurdity.)

OK, so Congress certainly can reduce Social Security benefits after 2018 (or at any time) without incurring the horrors of any kind of bond default, that's settled.

What are the practical, real-world chances that it will want to do so?

Well, let's look at some real-world numbers.

Come 2018 or so, when Social Security is projected to go cash-flow negative relative to the payroll tax, the trust fund is expected to hold about $5 trillion of bonds, current value. To redeem them over the period until 2042, as long as they are projected to last under current law, will take an average of more than $200 billion a year current value just to cover the principal (neglecting interest).

For perspective, total personal income tax collections in 2004 were $811 billion. Add corporate income taxes, and the total came to almost $1 trillion.

So we are talking about a general income tax increase on the order of 20%, just to redeem the trust fund bonds.

Or, alternatively, the Social Security Administration projects the cost of benefits to increase by about 2 points of GDP during the period in which the bonds will be paid off. With total federal income taxes today being a little less than 10% of GDP, this too implies a 20% increase in income taxes to fund the payment of the trust fund bonds.

Or, alternatively, the US government's unfunded liability for promised Social Security benefits over the next 75 years is $12.6 trillion current value [as per the Treasury, pdf. p. 11] This shortfall will have to be financed from general revenue over that period.

(Note that while defenders of the status quo often say the unfunded 75-year liability of Social Security is "only" $3.5 trillion, this is only the Social Security Administration's liability -- it is not the government's liability, and does not include the cost of such things as paying off the trust fund bonds.)

That liability is concentrated (more than 100% so actually) after about 13 years from now, when Social Security is projected to go cash-flow negative relative to the payroll tax. So let's say the $12.6 trillion, current value, will have to be financed from income taxes over those following 62 years.

Again, we get back to an increase in income taxes of about 20% being needed to fund benefits, starting by paying the bonds until they run out. (I'll let you do the math.)

And, unfortunately, this will all coincide with even bigger income tax increases needed to pay the unfunded liabilities of Medicare that are twice as large -- $24.6 trillion current value as of 2004.

So, sometime around 2018 the taxpaying citizenry will have within its line of sight the prospect of something like a 20% increase in income taxes to fund the redemption of Social Security trust fund bonds on top of perhaps a 40% increase in income taxes to pay for Medicare.

Will the prospect of a coming 60% increase in income taxes motivate voting taxpayers and the politicians who answer to them to look for possible reductions in benefits?

We can only say this: In 1983, when promised Social Security benefits last overran the taxes available to pay for them, Congress did significantly reduce benefits -- and the tax increase that was needed to fund promised benefits then was nothing like the increases that will be needed after 2018.