Edward M. Gramlich
The University of Michigan
June 27, 1988
The two studies today deal with what is to me a very interesting question. The 1983 Social Security amendments fixed the Social Security trust fund so that tax and replacement rates stay relatively constant over the next seventy years, in the face of large demographic changes in the share of people working and retired. This leads inevitably to a rise and fall in the cash surplus of the trust fund. Should the rest of the government then validate this rise and fall by keeping the general government surplus apart from Social Security constant, or should the rest of the government in effect spend the surplus by having a general government deficit that offsets the Social Security surplus?
The Social Security trustees have already worked out the details of a seventy year forecast as it regards the trust fund itself. What they do not know is how any incremental saving, if this trust fund surplus is validated, feeds back on income, interest rates, wages, consumption, and ultimately on the trust fund itself. These papers are commissioned to fill the gap. Both start with the moderately pessimistic IIB assumptions and crunch out the results. In my remarks I first provide an armchair guide to what the ICF Report is saying, I give some technical criticisms of their report, and then I give a few personal thoughts on the validation issue itself.
A Reader's Guide to the ICF Report
The report is written in what might be called a laborious mode, as consultant reports probably have to be. Seventeen fiscal scenarios are investigated, though most of what's important can be discerned by studying two. Two fairly complicated models are simulated, but much of the detail of these models is suppressed to make the projections conform to the IIB assumptions. Many possible saving-investment responses are discussed qualitatively, though when we get down to business only the very straightforward cases are investigated quantitatively. Some early tables give results as if GNP growth and interest rates were exogenous, though in fact they are endogenous.
When we strip away details, the ICF report is focused on one central question: what happens if America were to save more for the next forty years, and then less for the following thirty, by validating the upcoming Social Security surpluses and deficits? The question can be answered by comparing scenarios 6 and 7B. In scenario 6, ICF simply runs two models, their long run MDM model and a short-run long-run MGM model. They replace equations or modify these models as needed to replicate the results of the IIB assumptions. Call these paths in the scenario 6 runs the control paths.
These control paths assume that the unified budget deficit of the federal government, now running at about 3.5 percent of GNP, is eliminated by 1995 and then stays at exactly zero for the remaining 65 years. Although the control paths are the low national saving alternatives, it should be understood that getting to a balanced budget by 1995 will not be easy, and already entails a substantial increase in saving, discretionary fiscal cuts of at least two percent of GNP compared to the CBO baseline path. The Brookings control path is that deficits from 1995 forward are 1.5 percent of GNP.
The high saving, or validation, alternative (ICF 7B) begins with the same fiscal posture of a balanced unified deficit in 1995. After that time the trust fund surplus rises, and the unified surplus is allowed to mimic exactly that rise. There is some complex discussion about interest paid from the Treasury to the trust fund in the report: all that means is that the trust fund surplus that equals the overall unified budget surplus is the one that includes intergovernmental interest as a receipt. When the two surpluses are added, interest paid out of the general government and received by the trust fund then cancels out, making the sum equal to the trust fund surplus including intergovernmental interest.
The experimental results come form simulating two models. The MDM model assumes neoclassical equilibration of all markets, so that when national saving is increased in the experimental path, income starts rising right away. The MGM model, by contrast, has both short run IS and LM relations and endogenous prices, so that when national saving increases, output first falls and then rises. In keeping with these differences, the MDM model determines interest rates form the marginal product of capital, while the MGM model takes them from the IS-LM equilibration.
The results of all of this are not very surprising. In the experimental 7B path, the country raises its national saving rate by almost two percent for about thirty years. We already knew that that policy yields trust fund assets of about 28 percent of GNP at the peak in 2020. Now the 28 percent is assumed to go entirely into domestic capital accumulation. In the neoclassical MDM model real GNP rises slowly to a level higher than the control path by 2.4 percent, and then builds down as the trust fund begins dissaving. In the cyclical MGM model real GNP starts off lower for about ten years, rises to a peak higher by 3.9 percent, stays higher as the onset of saving reductions boosts income, and then starts building down more rapidly. In both models GNP in 2050 turns out to be trivially different from control GNP.
Interest rates show a corresponding pattern. In the neoclassical MDM model the rising capital accumulation moves along a production function to give about a thirty basis point fall in interest rates, at the peak capital accumulation point. In the cyclical MGM model the immediate drop is much sharper, but the long run drop is about the same. And when all of this is cranked into the trust fund, the higher saving policy turns out to be moderately costly to the fund under present laws. Payroll taxes go up early, they are invested at lower interest rates, benefits are higher subsequently, and the present value of all future surpluses is slightly lower. On the other hand, national consumption is higher, and the country could certainly afford to lower replacement rates to keep 21st century retirees whole and still raise the present value of future trust fund surpluses.
The early chapter of the report is so complete in its qualitative discussion of what might happen with validation of the trust fund surpluses that one is misled into thinking that the two elaborate models will deal with these complications. In fact, even these elaborate models do not deal with some of the critical magnitudes.
In any examination of higher saving policy, one needs the following central elements:
- an assumed natural rate of growth;
- an assumed production function;
- an assumption of how private saving will respond to any budget shift;
- an assumption of how open the economy is to international capital slows.
The rationale for the first three is well-known. The last is added to modernize the growth models. In modern times, when a country alters its national saving rate, it is not so clear that the investment rate would immediately follow, as was formerly assumed in closed economy models. Saving could show up as increased foreign lending, with no change in interest rates, and a less than complete shift in domestic physical capital.
In these simulations ICF gets the natural rate of growth from the trustee's IIB assumptions and the production function from the models. These may not be correct, but what more can a critic say?
But a critic can say more about the last two. The experiment under investigation is to validate the Social Security saving. Since the treatment of individuals by Social Security in the experiment is unchanged, there is no reason for Social Security to change private saving behavior. But if interest rates fall, or if because the government saves more, Barro private savers save less, there could be at least these other two reasons why private saving may decline in the experimental path. ICF assumed neither reaction, in effect that private savers fully validated the higher Social Security saving by keeping their own saving constant. I have often argued that such an assumption is empirically plausible in my view, so I will not be critical of the ICF choice. But as a systematic matter, I guess I think that they should have done at least one run with some private saving offset.
The previous criticism is grounded in academic pickiness -- it is an important point, others worry about it, and ICF should at least deal with it. The second problem, the open economy, really is important. Both of ICF's simulation models assume a closed economy regarding international capital flows -- any change in asset stocks is completely reflected in higher physical capital, and higher saving lowers interest rates. If the United States should be treated as an open economy, neither assumption is correct. And empirically, neither seems correct. As the US lowered national saving rates in the 1980s, there was a relatively slight interest rate effect and much borrowing to keep investment above saving. If the open economy was alive and well in the 1980s, why not in the next seventy years? I certainly think ICF should have provided at least one open economy variant, as indeed Brookings did.
Let me now turn from the ICF model to the policy issue itself. I want to argue that the country should validate its trust fund surpluses, and to aid in doing that, it should take these trust funds off-budget.
On the validation issue, by this time I suppose that I am a known critic of the low national saving rates that country has had in the 1980s. I have taken shots at these low national saving rates, and the deficit problem, in about ten papers by now.
One can break this issue into several components. I myself believe that the country should have been saving more even before the 1980s, when the ratio of net national saving to net national product was about 7.5 percent. I became obnoxious on the issue when the rate fell to about 3.5 percent, where it is now. And I should become even more obnoxious now that I realize that that rate fell to 3.5 percent at a time when the country should have been saving about 2 percent more than normal because of demographics.
The argument on demographics is straightforward. Our generation has chosen not to have many children, so that we know that our children will have to work harder to support us in our old age, other things equal. Do we blithely lay that burden on them, or do we help them? How can we not?
There are two requirements for helping them. The first is that any trust funds around must be kept in what analysts call close actuarial balance. That means that analysts do projections every now and then, and if the present value of expected receipts does not equal the present value of expected outlays, either tax or benefit rates are altered. In these terms the Social Security trust fund appears to be in close actuarial balance but the Medicare fund not, so this much must be done to satisfy this first requirement.
The second requirement, of course, is that any trust fund surpluses required by close actuarial balance must be validated. It does no good to save for the future with one hand but spend it with the other.
It is conceivable that we could run affairs by putting all trust funds on a pay-as-you-go basis, and just save more or less in the general government budget to help our children weather demographic swings. But we tried that in the 1980s and the results do not look promising -- the country saved much less when it should have been saving slightly more, and we made our children much worse off than they would have been for the demographic change alone. As somebody said, "saving ain't us". If saving ain't us, we then have to rearrange things to protect our children from our consumption proclivities. One way, by far the simplest, is simply to take these trust funds off the budget, and force ourselves to operate on the rest of the budget as if the trust fund surplus did not exist. In so doing, the country has a chance at validating the surpluses. Otherwise, I don't see how we ever will.