Sunday, August 3, 2008

Responding to Angry Bear: Where does the $17 trillion deficit come from?

Over at Angry Bear, Bruce Webb has a post on the so-called $17 trillion "legacy debt" inherited from early participants in the program. I've argued here, as has Jim Glass in the comments, that the future "infinite horizon" shortfall is basically caused by "over-generosity" to early cohorts of Social Security participants. It's a bit hard to summarize Bruce's argument, and harder for me to respond in the comments section at Angry Bear, so here I'll try to lay out what the legacy debt means and how it affects the system's financing going forward.

Ordinarily, we think about Social Security's finances in cross-section – e.g., how do all taxes paid by all workers in a given year compare to all benefits owed to all retirees or disabled in that year? That's obviously very relevant.

But another way of measuring things is by birth cohort: how did all the taxes paid by individuals born in a given year compare to all the benefits they have or will receive? This approach, best known as "generational accounting," was first proposed by Gokhale, Auerbach and Kotlikoff (see this backgrounder by the Tax Policy Center) and has become a standard way of analyzing how the program treats individuals in different birth cohorts. It's through a generational accounting approach that we can see where the long-term shortfall comes from.

So for each birth cohort taking part in Social Security, we can calculate (or, for future beneficiaries, project) how their total taxes will compare to their total benefits. If we discount all these values to 2008, add them up and then subtract the current value of the Social Security trust fund, we get the total infinite horizon shortfall for the program. Currently, that amount is projected at $13.6 trillion in present value.

Now, how does the claim come about that these future shortfalls are due to over-generosity to past generations? How can this make sense – the system is still in positive cash flow, and it still has a significant trust fund balance?

One way to illustrate is purely mechanically:Based on data from Dean Leimer, an economist at SSA who's probably the leading expert on these issues, the 1885 birth cohort – who retired at 65 in 1950 – received an average rate of return on their taxes of around 25% above inflation. This translates to their receiving roughly 15 times more in benefits than they paid in taxes. As late as the mid-1980s, a typical retiree received returns equaling those paid on stocks, but with none of the risks. (If you want an answer why Social Security was so popular for so long, there it is.)

Now think about future retirees: even if we could pay full scheduled benefits – that is, even if the Low Cost projections came true and we had no solvency problem – future retirees will receive less in benefits than they paid in taxes.

I can sum up the argument in this way:

  1. the infinite horizon shortfall is the sum of the net taxes or benefits paid by each cohort, from 1935 through the infinite future; and
  2. the infinite horizon shortfall is calculated under scheduled benefits; and
  3. we know that early cohorts received more in benefits than they paid in taxes, and current/future retirees will receive less in benefits than they pay in taxes; therefore we can conclude:
  4. that the infinite horizon shortfall is entirely attributable to the fact that early cohorts of retirees received far more in benefits than they paid in taxes.

Now let's give a little more detail and tackle some questions.

Some will argue that since Social Security is still solvent today, and has a large trust fund balance, that it's not possible that we gave large net transfers to early generations. This is confused. A fully funded pension program – on in which each cohort saves while working, earns interest on its savings, and draws them down to pay their retirement benefits – will always have a running trust fund balance. But it would be a lot larger than the one we currently have. Look at it this way: if early cohort hadn't been paid returns so far above the trust fund bond rate, what would have happened to that extra money? It would have gone into the trust fund and earned interest. How much larger would the trust fund balance be today if those early cohorts hadn't gotten such larger returns? Around $17 trillion larger. That extra balance today would be enough to pay full benefits forever (since future cohorts will receive returns below the trust fund bond rate even if we paid full scheduled benefits, they're effectively adding to the trust fund balance in any case).

Bruce Webb also raised the question of the tables in the Trustees Report. These break down Social Security's unfunded obligations between past/present participants and future participants. This isn't the greatest distinction (and I'll see if I can get a better break-down from the actuaries) because it includes in one group everyone from the first retiree in 1940 to the youngest worker participating in the system today, with the second group being people who will start working and paying taxes in the future. The better breakdown would be between past and current beneficiaries (i.e., people who have already died plus those already collecting benefits) versus future beneficiaries. As we saw in the chart, until relatively recently, retirees tended to get returns above the trust fund bond rate and thus were net beneficiaries of the system (benefits > taxes); going forward, most retirees will tend to be net taxpayers (taxes > benefits).

Now, as Bruce warned in his Angry Bear post, some will suspect this formulation of being some sort of right wing trick. Hardly. The first people to stress the importance of the legacy debt for policy questions were actually Democrats, Peter Diamond and Peter Orszag. Here's how they put it:

The benefits paid to almost all current and past cohorts of beneficiaries exceeded what could have been financed with the revenue they contributed, including interest. This history imposes a "legacy debt" on the Social Security system. That is, if earlier cohorts had received only the benefits that could be financed by their contributions plus interest, the trust fund's assets would be much greater today. If those expanded assets existed, they would be earning interest that could contribute to benefits.

I believe this is an important concept for thinking about how to formulate Social Security policy for the future, and I'll be working on it more. But hopefully this at least starts to explain where the idea of the "back transfer" comes from.

I'm sure this will provoke some push back, and I'm also sure I haven't addressed everything, so happy to answer objections either here or at Angry Bear.


34 comments:

Anonymous said...

Biggs

this is long enough and complex enough so as to defy a short, polite, answer.

You are, I believe deliberately, confusing some things because you can be sure the ordinary reader cannot follow the gymnastics.

I will try to raise some questions, if not provide answers, but it make take a while to get back.

I recomment, meanwhile, anyone else who is interested to look up Dean Baker's chapter on generational accounting in his book "The Phony Crisis."

I am not as well educated as Baker, so my comments will take a different approach: essentially the peasants eye view: what do I pay and what do I get, without any trips into along-side land where "what we could have done" is treated the same as "what we did for reasons that seemed important at the time."

coberly

Andrew G. Biggs said...

Coberly,

You're right that it's a bit complicated and maybe too long. Here it is as short and simple as I can make it:
1. Each individual pays taxes and receives benefits over his lifetime; the net value (taxes minus benefits) indicates how he affects the trust fund balance.
2. If we add up the net taxes or benefits of all the people born in a given year, we can see how that birth cohort affects Social Security. Early cohorts received more benefits than taxes; later cohorts more taxes than benefits.
3. If we add up the net taxes or benefits of ALL birth cohorts, from the first people to participate in Social Security until as far in the future as we can project, we get the system's infinite horizon unfunded obligation ($13.6 trillion).

For the basic premise that early cohorts were the ones who generated the long-term shortfall to be wrong, there needs to be a flaw in fact or logic of one or more of the terms above.

Andrew

Patrick R. Sullivan said...

Until these guys can comprehend that SS is merely a joint checking account that everyone has been making deposits to during their working lives, and then having disbursements made to them after they reach a certain age until they die, it's hopeless for them getting it.

In that simple factual situation, all you need is 4th grade arithmetic to see that if some people get more money disbursed to them from the joint account than they deposited into it, it follows as night follows day that someone else has to take less out than they put in.

The entire political dispute is about WHO the latter are going to be. It isn't that complicated.

Bruce Webb said...

I suspect I will have to take the same approach and move my extended response over to AB. I spent some time this morning with the Historical Tables (IV.A2 and IV.A4 specifically) and did the simple exercise of adding up actual benefits paid out.
OAS 1941-1956 $25.6 billion
OASDI 1957-1968 $181 billion
OASDI 1969-1980 $793.1 billion
OASDI 1981-1992 $2.476.5 trillion
OASDI 1993-2000 $2.829.3 trillion
OASDI 2001-2007 $6.330,9 trillion

For a grand total of $12.636.4 trillion. For Glass to state openly as he did here and at his website that
"If you look at this table {IV.B7} you'll see that past/present participants have received $17.4 trillion more in benefits than they paid in taxes." is simply not true, and the problem goes far beyond the tense problem in "have received". Instead he simply skipped over the relevant Trustees' definition of 'current participant' and ignored the 100 year time frame for costs that generate that $17.1 trillion. Of that amount fully $10.4 trillion falls in the period 2083 to 2018. Trying either to blame this on past beneficiaries or scaring current workers into thinking they actually have to come up with this amount above and beyond what they will actually receive in benefits is just nonsense, and if deliberate pernicious nonsense. No worker scheduled to retire by 2041 (interestingly almost exactly 67 years after the 1983 reform) is likely to collect benefits after 2083, meaning no one born before 1974 is going to be making a claim on that $10.4 trillion. Moreover no one born before 2016, which is to say no one currently born, is going to have to make any part of that post 2082 claim good. It is a simple matter of using a long term calender.

So while the following is a perfectly good way of framing the problem:
"But another way of measuring things is by birth cohort: how did all the taxes paid by individuals born in a given year compare to all the benefits they have or will receive?" The answer is not to be found in Table IV.B7 and only in a limited way in IV.B6. Instead the answer is to take those birth cohorts seperately, and if you do you will see that each cohort from the Boomers on back to program inception left or is projected to leave the Trust Fund with a positive TF ratio. Given that it is hard to see where the inequity comes from, all of the unfunded liability and all of the revenue needed to backfill it or contrawise the benefit cuts fall on Gen-X and beyond. To project this future gap backwards approaches dishonesty.

As to the graph I am not particularly impressed. A 21 year old in 1936 would not be eligible for full benefits until 1980 at which point everyone would be phased into the system. Almost all of the drop before that would seem to be attributable to people becoming eligible for retirement even though they had some to many years of non-contribution. The graph properly interpreted just shows what happens when the program is fully phased in, an event that didn't happen until 44 years from program inception.

Too you would need to correct some of those earlier ROI to reflect the cost of paying Title 1 benefits out of the General Fund. The dollar value of taxes needed to pay those benefits should be added to actual contributions from payroll to get a real idea of what the real ratio of tax to benefits was. As Title 1 phased out current workers got a silent tax cut, all things being equal the cost of supplying benefits to then current beneficiaries got that much cheaper. (Of course that silent tax cut may have been totally offset by increases in payroll tax over the time period.)

But even after making those corrections I don't think the following is correct:
"3. we know that early cohorts received more in taxes than they paid in benefits, and current/future retirees will receive less in benefits than they pay in taxes; therefore we can conclude:
4. that the infinite horizon shortfall is entirely attributable to the fact that early cohorts of retirees received far more in benefits than they paid in taxes."

Well no. 4 does not follow from 3. Total benefits paid out from program inception to 1980 were only $999.6 billion. Even now the total of benefits paid out does not equal projected unfunded liability and was funded to boot. The unfunded liability to Infinite Future is entirely due to a gap in projected cost via projected revenue starting in 2041 and not due to past benefit payments at all. I am afraid I am still not able to accept the Biggs/Glass formulation.

Bruce Webb said...

PRS oddly enough both Coberly and I are fully aware that Social Security operates as a savings account. But it is not the simple savings that your 4th grade model would have it.
"In that simple factual situation, all you need is 4th grade arithmetic to see that if some people get more money disbursed to them from the joint account than they deposited into it, it follows as night follows day that someone else has to take less out than they put in."

Hmm. Well no. Your account ignores the adjustment in initial benefit due to real wage increases over the workers lifetime. 4th grade math is a little early to have mastered fractions but lets give it a shot. Suppose you have a six inch economic pie and the elderly are getting 1/4 of it in benefits. Thirty five years later suppose that pie is now 9 inches wide and you and your cohort who are now retired and older cohorts still get a 1/4 of it. If real wages have been positive on balance over your lifetime your initial piece that that 1/4 of pie will be bigger than the pieces of those people who retired before you, they having their benefit only adjusted for inflation over those years. Moreover everybody still in the work force will be enjoying their own bigger slice relative to the overall retiree population. I'd explain further but it might require some ninth grade algebra and tenth grade geometry and so might be a little beyond your current level of math expertise.

Your account tries to take a dynamic system and force it into a static Procrustean bed. You have simply ignored the fact that savings accounts typically draw interest, if that interest adds to the account in ways that more than cover increased withdrawels over time it is perfectly possible for all future account holders to take out more than they put in. The adjustment of initial benefits to real wage functions to largely replace the interest on savings effect (and of course you get actual interest on the Trust Fund Special Treasuries).

"Its not that complicated"
No everything is simple if you just ignore the complexities. And simply skip examining the actual data tables for yourself.

As typical of all your appearances at AB you are here sneering down on us from the bottom of the data well. You don't get it because you have not done the hard work needed.

Bruce Webb said...

"3. we know that early cohorts received more in taxes than they paid in benefits, and current/future retirees will receive less in benefits than they pay in taxes; "

Well no we don't know that (and btw you need to edit this to switch 'taxes' and 'benefits' around'.

An unfunded liability over the Infinite Future need not make any individual cohort actually receive less in benefits than they paid in taxes, if most of that liability remains beyond that cohorts expected period of contributions and then withdrawels it is perfectly possible for everyone to get out more than they put in, the liability continually falling on future cohorts that are not necessarily currently contributing. Which is the case with the $17.1 trillion backward shift, it is a case of a theoretical transfer from the long, long term future to the simple long term future.

Paul Lawin said...

Andrew, I’m getting lost in the various definitions here. Please pardon a long example so that I can be accurate about my perspective.

Imagine the world’s simplest paygo retirement system. People only pay taxes the year they turn 45. They only get benefits the year they turn 75. In every year, taxes equal benefits. We have a stable population and no productivity gains. The program started in year 0. It paid benefits to cohort 1, even though they had never paid taxes. The tax rate doubled in year 30, and has been level since then. Using your “cross-section” view, the pattern of taxes and benefits is then:
Year 0 - 50/50;
Year 30 – 100/100,
Year 60 – 100/100,
All following years – 100/100.

Using the birth-cohort view, we get this pattern of taxes and benefits:
C1 – 0/50
C2 – 50/100
C3 – 100/100
All following cohorts 100/100

So C1 and C2 both got “windfall gains” of 50. Each succeeding cohort expects to exactly break even. However, we know that the program will end some day. Suppose the country is invaded between the time C4 pays its taxes and the time it gets its benefits. In this case, the actual result for C4 is 100/0, so they lose 100. It’s no coincidence that their loss is equal to the C1 and C2 gain.

If the system’s actuaries are doing a forward looking valuation at any time after Year 30, they would say:
The present value of future benefits exactly equals the present value of future taxes. (i.e. the “actuarial deficit” is zero).
The “termination cost” is 100. (That’s the amount of accrued future benefits the current participants would lose if the system were terminated.)

Notice that there is no mathematical requirement that the past windfalls equal the current actuarial deficit. But you seem to be saying that for our SS program, the two are equal, and about $13.4 (or $15.6) trillion.

In my system, if the gov’t foolishly raised future benefits by 10% without changing future taxes, then there would be a non-zero actuarial deficit (the value of in infinite stream of 10’s), and the termination cost would go to 110. By breaking the pure paygo match between taxes and benefits, they would also break the match between past windfalls and the termination cost.

It seems to me that the past SS windfalls will not be repaid until the system terminates. The program could be a "good deal" for almost all future generations in spite of the past windfalls. Any actuarial deficit that the actuaries calculate based on the assumption that the program continues infinitely must come from mis-matches in future tax and benefit formulas, not from past windfalls.

At other times, you seem to be equating the $17.4 trillion termination cost of SS with the past windfalls, even though we know that the system’s current benefit formula does not match the current tax rates. That may be a good approximation, if there is a close depending on how close future taxes/benefits match.

Jim Glass said...

Excellent post. Superior.

One thing I'll add is that "intergenerational equity" was a major concern of FDR himself and the other founders of Social Security, such as FDR's head of the Social Security Administration, Arthur Altmeyer, who was one of SS's chief designers.

One thing FDR famously insisted on was that SS be self-supporting and *not* drop a burden on future generations. He'd be aghast at a $14 trillion unfunded liability being dropped on future workers We know this from both his words and actions.

E.g, when his program was being legislated some working on it tried to slip in "paygo" provisions that would cause SS to require funding in excess of payroll tax in the 1960s. To their surprise FDR actually read all the paperwork, found the provision, yanked it, made them re-write it, and delayed the legislative submission. The final program enacted was largely "funded" and projected to have a very large trust fund accumulation in the 1980s -- and would, have if FDR's original law had stayed in place. (We know what happened then instead.)

FDR specifically and repeatedly insisted on "insurance-basis funding" providing a "fair return" comparable to what people could obtain commercially in an annuity bought from an insurance company -- which very few could in the 1930s -- which basically resolved to the T-bond interest rate, which was stable and sustainable indefinitely into the future.

Look at the actual benefit schedule of the FDR's own Social Security Act of 1935 -- with the payroll tax rapidly rising to 6%, no benefits payable until five years after tax collection started, and benefits then only very slowly rising to their maximum, and always directly dependent on the amount of tax already paid in -- at the same rate across generations. FDR's funded program gave a steady rate-of-return could have lasted forever.

But what happened instead? As soon as the payroll taxes started coming in Congress' Left side said "don't wait, spend those taxes right now!". And it's Right side said "the left is just going to squander all the tax money -- stop that by cutting the tax!" So they made a deal: use all the tax money coming in on faster and bigger benefits right away while stopping the future tax hikes. More spending, less taxes! What politician doesn't like that??

Of course, the arithmetical effect of that was to greatly increase the return on contributions to early participants, while reducing that to later participants and creating a funding shortfall in the future -- creating exactly the kind of burden on the future that FDR loathed.

FDR vetoed the changes for exactly those reasons. It was his only veto in 10 years -- that's how strongly he felt about it. When Congress was getting ready to override the veto he sent his SSA Chief, Altmeyer, to Congress to urge it to uphold the veto. Here's some of what Altmeyer said:
~~~

"It is a mathematical certainty that the longer the present day payroll tax remains in effect, the higher the future payroll tax must be if the system is to remain financed by payroll taxes.

"This will eventually necessitate raising employee's contributions rate later to a point where future beneficiaries will be obliged to pay more for their benefits than they would if they obtained this same insurance from a private insurance company.

"I say it is inequitable to compel them to pay under this system more than they would have to pay to a private insurance company, and I think Congress would be confronted with that embarrassing situation..."...

"If we should let a situation develop whereby it eventually becomes necessary to charge future beneficiaries rates in excess of the actuarial cost of the protection afforded them, we would be guilty of gross inequity and gross financial mismanagement, bound to imperil our social insurance system."
~~~

Congress ignored him, over-rode FDR's veto, kept the tax rate down, raised the benefits, and created the "Ponzi scheme that works" that Samuelson praised.

And by Altmeyer's "mathematical certainty" it also assured that workers born 30 years later would get retirement benefits worth less than they paid for them, taking an outright loss of $14 trillion+ -- which sure as heck is a "rate in excess of the actuarial cost of the protection afforded them", eh?

The result is exactly what Altmeyer predicted, people getting net some $30 trillion less than the prior generations are showing signs of, well, being not happy about it -- with the result "bound to imperil our social insurance system". The man was right, eh?

Political finance is interesting. "Intergenerational equity" was not discovered by Kotlikoff. Back at the founding of SS it was a prime concern of FDR, Altmeyer, et. al. Then it was totally forgotten, wiped from memory, during the easy-money "everybody wins with 'the Ponzi scheme that works'" era. Clearly, it is still wiped from memory among today's left. Then, as the Ponzi game approached its back end, it was rediscovered by Kotlikoff.

So for everybody who self-righteously thinks, "I'm defending FDR's Social Security from those who want to destroy it" -- get a grip: you're not.

That $14 trillion+ loss being dropped on future workers has FDR burning rubber popping wheelies in his chair, wherever his is today.

"I say it is inequitable ... we would be guilty of gross inequity and gross financial mismanagement, bound to imperil our social insurance system...."

Here we are living it -- he was right on all counts.

Andrew G. Biggs said...

Bruce,

Just to respond quickly to two points in your first comment:
If I'm reading right, the dollar figures you're citing from the Trustees Report are nominal dollars; to be comparable to the unfunded obligation they need to be converted to present value dollars.

Second, the reason the Trustees Report measures net benefits to past/present participants extends out 100 years is because of an assumption that the youngest participants are currently 15 years old and no one will live more than an additional 100 years past that. So it's really just a tool for measurement, but isn't comparable to the 75 year and infinite horizon actuarial balance periods.

Third, you say: "the answer is to take those birth cohorts separately, and if you do you will see that each cohort from the Boomers on back to program inception left or is projected to leave the Trust Fund with a positive TF ratio." That's a) actually not true, since early cohorts collected far more than they paid in, and b) not a sufficient measure. The question isn't whether the trust fund has a positive balance, but whether it has a balance at the right level. I'm not sure of an easy way to explain it, but I'm very sure that this line of reasoning is wrong.

Fourth, returning to your final points, again, you have to put dollar figures in present values, since that's what affects the current trust fund balance.
Andrew

Anonymous said...

Biggs

"flaws in logic" are a bit tricky. one needs to set up a logical structure, which you have not done, and which i keep asking for.

but i think i may see your problem: a bit of carelessness with language.

at least if i am reading the graph you gave us correctly, there is never a time when rate of return is negative. what you really mean is that the rate of return falls below the rate of return you think you can get on bonds. That is a VERY different issue.

it would involve us comparing risks, and what risks mean to different people at different stags of their lives. Social Security is an insurance program, not an investment program, and you should know the difference.

coberly

Andrew G. Biggs said...

Bruce,

Responding to your 3:51 comment: Quoting me you say: "3. we know that early cohorts received more in taxes than they paid in benefits, and current/future retirees will receive less in benefits than they pay in taxes;" Then respond that "Well no we don't know that."

Maybe we have disagreement over definitions here. As I'm defining it, which is what matters for long-term system financing, an individual or cohort receives more in benefits than they paid in taxes if the present value of their benefits exceeds the PV of their taxes. Equivalently, we know that the PV benefits > PV taxes if their internal rate of return (as shown in the chart) exceeds the interest rate on the trust fund. When a cohort of retirees is receiving a 25% real rate of return, we can very confidently say that PV benefits > PV taxes.

Anonymous said...

glass and sullivan

you both like to throw out statements like "simple arithmetic shows..."

well, simple arithmetic may show that, but you haven't shown that your simple arithmetic applies to the present situation.

and while Biggs may think that the length and complications are a problem for me... the real problem is how much time do i want to spend trying to teach people one verifiable step at a time reasoning when it is so much fun for them to say "one and one and one is three
all i can tell you is you got to be free!"

Andrew G. Biggs said...

Paul,

You catch a valid point, which I've written about elsewhere but unfortunately didn't include here: the termination cost of the program -- the amount of benefits currently earned, but yet to be paid out -- is around $17 trillion. (See www.ssa.gov/OACT/NOTES/ran1/an2008-1.pdf)I think it's coincidence that this is also around the net benefit paid to past/current participants; the similarity may draw from the fact that it's only in recent cohorts of retirees that we switched from people being net beneficiaries to net taxpayers (i.e., receiving returns below the trust fund interest rate). I'll have to think about this more, plus work to get better data since the available stuff isn't easy to work with. Thanks.

Andrew G. Biggs said...

Coberly,

You're right that on average, returns are never negative. (Individuals might receive negative returns through chance, but entire cohorts won't so long as wage growth + labor force growth is positive).

But what we're looking at here isn't about whether the system is a good deal for individuals, whether rates of return are relevant for insurance programs, etc. The issue is where the system's long-term shortfalls arose. To determine that, you need to know how individual cohorts affected the trust fund, and you measure that by discounting their taxes and benefits at the interest rate earned by the fund. If I wasn't clear enough I'll try to say something like 'net discounted taxes/benefits' or similar.

Bruce Webb said...

Well where to start.

Total value of past benefits paid from progrm inception to full phase in 1980 was $999.6 billion. Total value of all benefits paid to date are around $12 trillion. All of which have in fact been funded. Every single dollar of unfunded liability comes from imbalances between projected costs and projected income after Trust Fund depletion in 2041.

If we simply agreed to accept a 22% cut in benefits in 2041 (to a real benefit better than today) then that unfunded liability resets to zero. In which case blaming this on early cohorts turns kind of silly.

In any event AB readers are awaiting for a direct response.

Bruce Webb said...

Biggs you continually use the simple past 'paid' to refer to what is in reality what the French call 'future compose' 'will have paid'. To repeat total benefits paid to date total $12 trillion. Your attempt to cast $17.1 trillion as the result of excess contributions 'paid' at a minimum violates the rules of grammar. Which given I spent a good deal of time in Grad school studying verbal modalities is kind of painful.

Tense and time matter in language. This constant attempt to confuse past and present effects may fool economists, students of Prof Julian Boyd at UCB (I.e. me) are not amused.

Bruce Webb said...

Pace Biggs past dollar payouts do not need to be adjusted to current dollar numbers simply because they have already been paid. That it would have cost that many more current dollars to pay that past benefit is irrelevant, that debt is now off the books. It makes sense to cast future obligations in both current and constant dollar terms, projecting that backwards doesn't. Done is done. Trying to wedge $13.6 in future obligations onto the back of past obligations fully satisfied and at a much lower level doesn't make logical sense.

Jim Glass said...

If we simply agreed to accept a 22% cut in benefits in 2041 (to a real benefit better than today) then that unfunded liability resets to zero.

Of course! And if you cut benefits by 50% you'd create a surplus! Social Security would then be strong as an ox, fiscally. You could use it for propping up Medicare or fighting another war or something. What would be to blame the earlier cohorts for then?

Oh, but that would be to forget -- most regressively -- that Social Security was designed with a purpose....

Here's Altmeyer, in 1936, promising to American the public it would get a good return on its contributions to SS...

"Moreover, every worker eligible ... will receive a monthly retirement benefit upon reaching the age of 65 larger than he could purchase from any private insurance company with the taxes he will have paid *"

Altmeyer, 1944, warning Congress not to break that promise...

"This will eventually necessitate raising employee's contributions rate later to a point where future beneficiaries will be obliged to pay more for their benefits than they would if they obtained this same insurance from a private insurance company.

"I say it is inequitable to compel them to pay under this system more than they would have to pay to a private insurance company"...

"If we should let a situation develop whereby it eventually becomes necessary to charge future beneficiaries rates in excess of the actuarial cost of the protection afforded them, we would be guilty of gross inequity and gross financial mismanagement, bound to imperil our social insurance system"

I wonder, when you cut benefits only by 22%, how will you square that with keeping Altmeyer's promises? Upon which Social Security was sold to the public?

And how will you avoid being guilty of the exact "gross inequity and gross financial mismanagement, bound to imperil our social insurance system", that he warned against?

Some people here seem to think that these promises and warnings from the creators of Social Security never existed -- or that they had an expiration date circa the cohorts born around 1940.

Jim Glass said...

BTW, Bruce, re:

If we simply agreed to accept a 22% cut in benefits in 2041 (to a real benefit better than today) then that unfunded liability resets to zero. In which case blaming this on early cohorts turns kind of silly.

Friedman pointed out that it is the easist of intellectual exercises to show the past cohorts took out more than they put in, so that the future ones must get back less than they put in and so take a loss.

Say you terminate SS on a given date, perhaps 1/1/09, promising to pay all benefits earned as of that date (using general revenue from income taxes, or national park admissions, or whatever). SS participants will owe nor more tax as of that date but will still be entitled to $X trillion of benefits -- say $15 trillion, whatever -- as of that date. So as a group they have earned benefits of that much more than they paid in. Simple.

Now going forward in real life, after 1/1/09 all SS participants will pay $Z in taxes and receive from SS benefits of $Z - $X, because that $X is gone to pay their predecessors. Ergo they must take a loss of $X. That's pretty simple too. QED.

And it is because the earlier generations received benefits equal to all their taxes plus $X that the later ones must take a loss of $X -- no other reason.

Now you seem fixated exclusively on the idea of getting enough cash-in to balance cash-out for SS benefits -- e.g., "just cut benefits by 22% in 2040". Of course, you could raise payroll raise after 2040 instead. Or do a host of other things. Balancing cash-in, cash-out is a trivial challenge. It is so trivial that it is impossible not to do it. An infinite number of arrangements can do it. Why anyone would fixate on such a trivial thing I don't know.

What's not trivial or easy is arranging the future of SS, including cash-in cash-out, so that participants don't take a loss from it -- so that it does *not* "eventually become necessary to charge future beneficiaries rates in excess of the actuarial cost of the protection afforded them".

You don't seem to care about that at all. Maybe that's what you think is "silly" when you dismiss "blaming this on the early cohorts" as "kind of silly."

You may not think it matters if future particpants pay into SS much more than the benefits they get are worth. But that's a value judgment on your part.

Altmeyer, speaking to Congress for FDR, had a different value judgment on that:

"we would be guilty of gross inequity and gross financial mismanagement, bound to imperil our social insurance system"

Well, maybe we are already to where Altmeyer warned us not to go in that regard, but there is still the major issue of how to spread the loss to the young most equitably, to mitigate its cost -- which could be rather important to the political future of SS, being that when future workers get voting contol of the system they may not be as uncaring about how they take that loss as you are uncaring about how they take that loss.

But to discuss that issue you must realize that balancing cash-in, cash-out is not the only thing that matters to SS ... and that that loss to the young does matter a great deal, just as Altmeyer said it would.

If you are going to just hand-wave that loss to future workers away with "silly", disregard it, then there's no common ground for us to talk about.

shoffy22 said...

Thanks for the great post Andrew, this is the most understandable explanation of the legacy debt that I've read.

Larry Dewitt wrote an article on Social Security Financing 1939-49 that relates well. It points out how congress postponed multiple tax increases during this time and moved the system from it's original design of being part pay-as-you-go and part pre-funded to being almost all pay-as-you-go.

One comment - it seems to me that if one looks at social security from the infinite horizon perspective and argues that the legacy debt is important - then to be conistent one would also argue that the Social Security trust fund is legitimate.

There are certainly specific critiques of whether the current trust fund mechanism is the best investment structure for pre funding Social Security - perhaps most importantly whether investing in the rest of the Federal government will yield enough economic benefits so that the Federal governement is in good enough shape to pay it's interest obligations to Social Security with general revenues - but the overall concept of the Trust Fund as a mechanism to earn interest and pre fund Social Security, and to enable present value calculations and the infinite horizon perspective - to me seems sound.

I'm curious what your thoughts are in regards to this relation between the infinite horizon perspective and the Social Security trust fund.

Thanks again for this great post - what a great way to kick start Monday Morning!

Andrew G. Biggs said...

Bruce said: "Pace Biggs past dollar payouts do not need to be adjusted to current dollar numbers simply because they have already been paid. That it would have cost that many more current dollars to pay that past benefit is irrelevant, that debt is now off the books. It makes sense to cast future obligations in both current and constant dollar terms, projecting that backwards doesn't. Done is done. Trying to wedge $13.6 in future obligations onto the back of past obligations fully satisfied and at a much lower level doesn't make logical sense."

Bruce, this part is just wrong. The actuarial balance is a present value number -- PV of future taxes - PV future costs + PV trust fund balance, over PV future payrolls.

If you want to know how a given cohort's net benefits affect long-term financing, you need to translate it into a present value for the simple reason that a cohort's net gain or loss to the system compound over time at the trust fund's interest rate.

For example, let's say that the first cohort of new retirees in 1940 actually contributed more to the system than they received in benefits. If so, that amount would boost the trust fund balance, earning interest each year.

More broadly, the infinite horizon actuarial deficit can be disaggregated into the present values of each cohort's net taxes/benefits that were or will be paid into the system. That's just how the math works. So numbers that are in nominal or constant dollar terms just aren't comparable.

Andrew G. Biggs said...

Shoffy - thanks for the compliment and a very good question. You said, "if one looks at Social Security from the infinite horizon perspective and argues that the legacy debt is important - then to be consistent one would also argue that the Social Security trust fund is legitimate."

By legitimate, I think what you're saying is that the fund is economically meaningful, meaning that increases in the fund balance correspond to improvements in the overall budget balance, increases in national saving, output, etc.

This is a very important (in fact, crucial) question for reform, but I'm not sure it has to be a central one for measuring the long-term shortfall. (Although I know some smart people who disagree with me on this.)

Let's say there was no trust fund at all. We'd still want to measure long-term financing, but we wouldn't want to count a dollar of obligations 100 years from now the same as a dollar of obligations today. (Time value of money, etc.) So we'd discount that future dollar amount back to the present. From there, we can create a long-term actuarial balance (or something like it) even if we don't have a trust fund or don't think the trust fund is meaningful.

For reform, though, you need to decide whether you think the trust fund actually transfers resources over time through increased national saving, or whether it's really just a book-keeping device that tracks how much we've borrowed.

Thanks again!

Andrew G. Biggs said...

Jim G: Very interesting historical material. The history of Social Security is often a lot different than we imagine it was.
Andrew

Paul Lawin said...

Andrew, I tried to make two points. I probably should have numbered them.

The first point was that there is no guarantee that the termination cost of our SS system is identically equal to the legacy debt. The two would be equal in a pure paygo system like the simple example I constructed. But SS isn’t pure paygo. I can believe that even in SS the two are the same order of magnitude. I think you caught this point.

The second point was that the number you call the “infinite horizon shortfall” (which I’m assuming is what the actuaries call the “open group unfunded obligation”, at least you’re quoting exactly the same $13.6 trillion) has nothing to do with the legacy debt. I don’t think you caught this point. (You don’t use the term “legacy debt”. I used “windfalls”. I mean the “excess” benefits that early generations get in paygo systems.)

The actuaries describe the 75-year OGUO as “The present value of future cost less future tax income over the long-range period, minus the amount of trust fund assets at the beginning of the projection period, amounts to $4.3 trillion for the OASDI program.”
Then they add “extending the calculations beyond 2082 adds $9.3” trillion. This is the $13.6 that I think you are quoting.

Note that the $13.6 trillion is the $15.8 trillion excess of future scheduled benefits over future scheduled taxes, less $2.2 trillion trust fund. One act of Congress could change the OGUO to zero. But that act would have no impact on the legacy debt.

I constructed my simple example so we could see this. Note that in my example, the OGUO is zero, the legacy debt is 100, and the termination cost is 100. I then had the politicians change the future benefits. This increased the OGUO to an infinite stream of 10’s, had no impact on the legacy debt, and increased the termination cost to 110. Again, the OGUO is not connected to the legacy debt in any way. The termination cost is at least related.

Patrick R. Sullivan said...

If we simply agreed to accept a 22% cut in benefits in 2041...

Which is your nomination for WHO gets to pay the legacy cost. Just as I told you.

Andrew G. Biggs said...

Paul,

I'd have to think more about your first point. There's no exact definition of the 'legacy debt', but I would think (off the top of my head) that the termination cost would equal the net legacy debt (net transfers to early cohorts minus net taxes levied on later ones), but I'd have to think more.

I think you're wrong on the second point, or at least I'm not sure if we're disagreeing. You're right regarding how the actuarial balance is calculated, but in this context the legacy debt would enter through a lower current trust fund value. (I.e., if we hadn't given net transfers to early cohorts then the TF balance today would be around $19 trillion, not $2.5 trillion.) You're right that Congress could zero out the future shortfall by cutting benefits, but the need to cut benefits is a function of the legacy debt. Not sure if we disagree here.

Andrew

Paul Lawin said...

Andrew,

I think we’re at least closing in on the issue. I won’t argue that if we hadn’t given the first cohorts “net transfers” the trust fund would be in the $19 trillion range (I assume you’re adding the $17 trillion termination cost to the $2 trillion trust fund here. Recall that I’ve said this isn’t a perfect calculation). If that were the history, then we would have an “advance funded” system, not a paygo system.

But, I don’t see why “the need to cut benefits is a function of the legacy debt”. I provided a simple example of a paygo system with a legacy debt and no need to cut benefits. It provides exactly what a mature paygo system is expected to provide – an IRR equal to the growth rate of the tax base (which I set to zero for simplicity). If all paygo systems with legacy debts have to cut benefits, why does my example work? If only some have them have this connection, what’s special characteristic that makes this happen?

Maybe I shouldn’t complicate this by adding a “what if” to my example, but I’ll risk it anyway. Suppose that after a few years of successful operation, the longevity in my hypothetical society starts to improve. More people live to 75 and collect benefits. If nobody changes taxes or benefits, the actuaries will start reporting “open group unfunded obligations”. They will tell the politicians that either taxes or benefits need to change if they want to continue the paygo format. The program “needs to cut benefits”. But that has nothing to do with the legacy debt. Mathematically, it’s just as plausible to assume that fertility rates go up, the tax base increases, and the actuaries would say the program “needs to increase benefits” to maintain balance. Again, it seems that the legacy debt is irrelevant to the OGUO.

Andrew G. Biggs said...

Paul,
In one sense, the need to cut benefits isn't a function of the legacy debt, it's a function of a benefit formula that promises people a roughly actuarially fair return (PV taxes = PV benefits) when the legacy debt means that under most foreseeable economic conditions it would be impossible for the system to pay such a return on an ongoing basis.

You're right that if longevity improves, that will throw things out of balance. But I'm not saying that ANY foreseeable deviation from balance is a function of the legacy debt, only that the one we have is.

Now, if demographics and economics improved a LOT, such that the rate of aggregate wage growth equaled the interest rate, then Social Security could pay full scheduled benefits despite the legacy debt. But this is a) economically implausible, and b) actually a textbook case where a paygo system would actually pay a higher return than a funded system. But again, it's implausible.

Andrew

Bruce Webb said...

Well the fact remains that all benefits paid up to 1980 totalled somewhat less than a $1 trillion dollars and all benefits to date amount to $12 trillion. Trying to explain a $13.6 trillion or a $17.1 trillion unfunded liability going forward as the result of excess transfers going back fails PRS's 4th grade math test.

I plan a new Angry Bear post on 'Unfunded Liability' as early as tomorrow. Hope to see you there.

shoffy22 said...

Thanks for your response to my comment Andrew. I think your response articulates well the key debate regarding the trust fund - namely whether it is an economically meaningful method for savings and investment that can effectively transfers resources through time.

Regarding PV calculations, I think you're right that we wouldn't need the specific trust fund mechanism that we have now, but I think we would need some sort of mechanism that would enable having an account that earned interest of some sort. From the perspective of Social Security financing, having a Trust Fund with it's assets invested in Treasury bonds has worked well in regards to earning a guaranteed interest rate which can then be used for PV calculations so that we have a good sense of the time value of money that is achievable in practice.

Then it seems to be an open question whether the trust fund invested in treasury bonds is only beneficial as far as keeping track of social security finances but doesn't produce any economic gains since it's just keeping track of other future obligations that general revenues will need to pay back, or whether it does produce additional economic benefits through enabling the rest of the federal government to have lower taxes, higher spending, or less borrowing during social security surplus years.

Perhaps the beauty of this question is that it is impossible to know for sure. For example, in the 1990s when we had large social security surpluses, on one hand it would have been great if these surpluses had been invested in other non-government assets that were earning interest, which would be paid back to Social Security at a later point from non-govt sources and entail no future obligations to the Federal government. On the other hand - if these surpluses had been invested in other assets, then the rest of government would have had to either raise taxes, cut spending, or publicly borrow to make up the difference. It seems to be an open question whether doing one or more of these options would have dampened economic growth and our future ability to pay back these obligations more or less than the new gains that would have accrued from making new investments in non-govt assets.

I may have veered off topic, but I think this relates back to the legacy debt question as well and the possiblity that it's importance could be debated continuosly since it's also really hard to know the ultimate economic effects of the legacy debt.

Similar to the question of the surpluses in the 1990s, the tax rates of the 1940s that DeWitt refers to in his paper bring up a similar question. By freezing the tax rates, congress started a chain of events that led us to continually underfund Social Security and moved it to really become pay-as-you go. From the specific perspective of Social Security financing, it seems clear that this hampered Soc.Sec's ability to pay future obligations. At the same time, keeping social security tax rates low and giving very high benefit levels may have helped contribute substantially to economic growth from the 1940s through the 1960s/1970s while the soc.sec system became mature. While at the end of this stage Social Security had much less direct ability to pay it's obligations than it would have if there had more pre funding, perhaps the economy grew to a large enough degree more than it would have if Soc.Sec had been prefunded that ultimately Social Security's abilitity to pay benefit obligations turned out to be similar to what it would have been had there been more pre funding, due to the larger economic output that enabled more payroll tax revenues?

Or perhaps since there are so many factors to the economy - maybe if there had been more pre funding, we would have still had the same economic growth from the 1940s through the 1960s and also an improved abilitity to pay our benefit obligations.

Wow, it's fun to think about this stuff, but I guess the morning's flying, spose it's time to head in to work. :)

Paul Lawin said...

Andrew, This part of your last comment gave me a glimmer of hope “In one sense, the need to cut benefits .. [is] a function of a benefit formula that promises people a roughly actuarially fair return (PV taxes = PV benefits)” Since the discount you’re using for the PV calculation is the bond rate, you seem to be saying that the current US SS system is promising future retirees an IRR equal to the bond rate.

Only a glimmer, because this directly contradicts your original post: “Now think about future retirees: even if we could pay full scheduled benefits … future retirees will receive less in benefits than they paid in taxes.”

Which statement is correct? The graph says both statements are wrong. It shows future retirees getting something below the bond rate, but clearly above zero.

But your first statement at least gave me an idea of what might be on your mind. Here’s a possibility.
1. Imagine a retirement system that starts out as paygo, with the normal windfalls to the early cohorts.
2. Suppose the tax base growth rate for this economy is “i”, so as the system matures later cohorts can get an IRR of i.
3. But, the politicians don’t like that return. So they decide to raise the benefits to the point that the later cohorts all get “j”, where j>i. (Note that j is purely arbitrary, it doesn’t need to be related to any real demographic or economic factor.)
4. Of course, we no longer have a paygo system, because “paygo” means that benefits equal taxes in all years, and they’ve created something where benefits will always exceed taxes. Maybe we can call this thing “negatively funded”.
5. Today, the actuaries calculate an “open group unfunded liability” by discounting future excesses of benefits over taxes. They decide to discount by j.
6. Next they do a calculation of the “legacy debt”. They look at the benefits the early cohorts got in excess of their taxes accumulated at j. Then they accumulate this excess to today, again using j.
7. If all of the above are true, then OGUO = legacy debt.
(I haven’t taken the time to prove this, or even do extensive examples, but it does seem true for a very simple example.)

So if the US SS actually met all the criteria in 1-7, our OGUO would equal our legacy debt. (For this purpose, “j” would be our T-bond rate.) Is this what you meant?

I feel good that I can at least see a way to get at the equality that you seem to want. However, if this is your approach, then I’ve got two serious problems with your post.

First, there is the contradiction I reference above. This equality only works if the current benefit schedule gives all future generations an IRR equal to the bond rate. But you’re saying future generations won’t even get zero. There is a huge difference. If fact, it seems that if they’re scheduled to get less than zero, the OGUO would disappear entirely.

Second, your first sentence says that the OGUO in our system “is basically caused by” the initial windfalls. But that isn’t the case in my example. The OGUO is clearly caused by the fact that people decided they weren’t going to stick to paygo. The windfalls were there regardless of what happened at step 3. If the politicians had decided to have a properly funded paygo system the OGUO would have been zero, nothing close to the legacy debt. The OGUO isn’t caused by the windfalls, it’s caused by the decision to drop paygo.

The implications for SS are important. The existence of the legacy debt does not doom SS to an unending stream of negative IRRs, nor to an ever-increasing OGUO. The legacy debt is a normal characteristic of paygo systems, and we both know that paygo systems can pay positive IRRs with no unfunded obligations indefinitely.

Andrew G. Biggs said...

Paul,

The average return to scheduled benefits is slightly below the bond rate, which I may not have been sufficiently clear on. The system promises close to an actuarially fair return, but not quite. In the same table that cites the $17 trillion net benefit given to past/present participants, there's a small net tax to future participants even under scheduled benefits. It's not a huge amount, but simply points out that future participants are essentially self-funding.

I think your basic example is correct, though I'd have to work it more carefully to be sure.

Your second problem is a matter of interpretation. Historically, I don't think the problem lately has been Congress raising benefits above the payable rate; the 1983 reforms lowered them pretty significantly. But again, you can interpret them either way.

Finally, you're right that a paygo system need not pay negative returns, although a) I'm not sure I suggested that, and b) the steady-state return isn't related to the legacy debt but to the growth rate of aggregate wages. The projected rate is around 1.43%, I believe, so that's about the IRR people could expect in the long-run.

Paul Lawin said...

Andrew, Thanks for the reply. I find your comments much more in line with my observations than the original post. In particular, the reference to the -$1.5 trillion for the future participants fills in a blank for me.

This comment looks right-on " the steady-state return isn't related to the legacy debt but to the growth rate of aggregate wages. The projected rate is around 1.43%, I believe, so that's about the IRR people could expect in the long-run."

However, regarding this: "you're right that a paygo system need not pay negative returns, although a) I'm not sure I suggested that,". Well, I go back to the OP and find: " even if we could pay full scheduled benefits ... future retirees will receive less in benefits than they paid in taxes." and then this: "current/future retirees will receive less in benefits than they pay in taxes". You apparantly suggested something without realizing that you were. Maybe you could clarify your meaning in the OP by includng some of the language from the comment -- i.e. I'd suggest an edit of the OP.

Regarding my example, the problem with the decisions in step 3 wasn't that they raised the benefit formula, it's that they didn't decrease it enough. Initial paygo benefit formulas have to be very rich to distribute the early year taxes, then they need to be reduced. This hypothetical, like our Congress, just did't want to go down to the sustainable level.

Finally, the unfunded obligation was not "caused by" the initial windfalls. I think you are proposing a syllogism of the form: 1) All humans walk upright, 2) Some humans are heavy drinkers, 3) Therefore, walking upright causes heavy drinking. The facts that 1)all paygo systems generate windfalls, and 2) some have unfunded obligations, do not imply your desired conclusion that 3) the windfalls caused the unfunded obligations. I'm really at a loss for a way to make that plainer.

Andrew G. Biggs said...

Paul,

Re negative returns, receiving less in benefit than you paid in taxes (both in present value) doesn't imply a negative rate of return, only a rate of return below the discount rate. So if the trust fund bond rate is 2.9% above inflation the scheduled return might be 2.6% or something along those lines. It wouldn't be -5% or anything like that, at least on average.

We may have to agree to disagree on what 'caused' the unfunded obligation, since we could say that any step we failed to do that would have reduced it is the cause. Here's a restatement that's probably less ambiguous: future participants will receive returns below the bond rate -- meaning they will pay more in taxes than they receive in benefits -- because past participants received more in benefits than they paid in taxes.