INTRODUCTION The discussion of Social Security reform has centered around alternative plans to restore solvency based on the best available current projections. This is the same process that was followed in the past. In 1983 Congress enacted a bipartisan reform intended to ensure 75-year solvency for Social Security. But no sooner had Social Security been saved than the program slipped back into projected insolvency and the reform debate began anew. One reaction to this development, not without controversy, has been the notion that reform should aim at ensuring “sustainable solvency.” That is, Social Security reform should not just put the program on a sustainable footing for 75 years but should aim for projected balance over the indefinite – or even infinite – future. Consistent with this, it has become more common to focus not just on the aggregate 75-year impact of Social Security proposals but also to ensure that the cash flow balance is positive in the 75th year.
The experience since 1983, however, provides the motivation for another reaction that has not had much of an impact on policymakers: an emphasis on what I term “robust solvency.” Based on the projections at the time, the 1983 reforms should have been sufficient to roughly restore solvency through about 2064. The fact that Social Security again faces an imbalance is a reflection not just of the passage of time but also of the fact that reality has turned out somewhat differently than what was forecast just a few decades ago.
This experience raises the prospect that tremendous political effort could again go into “saving” Social Security only to find out ten or twenty years down the road that the problem was significantly under-solved or over-solved. Given the tremendous political difficulties and delays inherent in reforming Social Security it would be better, everything else being equal, to minimize the chance of this occurring. In addition, building contingency into a reform would eliminate one objection to acting sooner: the argument for an option value to waiting.
To date, the vast majority of research and policy work on Social Security has focused on solving the Social Security problem as if the future is known with certainty. Much less work has gone into what statisticians term the “second moments,” that is how the future might turn out differently than what we expect today. To get a sense of just how large these second moments are imagine a forecaster at the inception of Social Security in 1935 needing to predict World War II, the baby boom, the baby bust, and all the other changes in the following decades.
The institutional process of policymaking itself places virtually no emphasis on robust solvency. The Social Security actuaries, for example, score all Social Security plans only on the basis of the “intermediate” projections and do not show what the plan would look like under alternative assumptions about the future. The Congressional Budget Office shows a stochastic range of outcomes but does not provide any metrics to assess whether a plan is more or less “robustly” solvent.
Some recent proposals clearly fail the robust solvency test. “Price indexing” and “progressive price indexing,” for example, both deliver smaller reductions relative to scheduled benefits when productivity growth is lower and Social Security’s long-run imbalance is larger and vice versa. Other proposals, including indexing some combination of benefit levels, payroll taxes or the normal retirement age to longevity would have the right sign: making larger changes if longevity widened Social Security’s imbalances. But even these proposals fall well short of solving the first order financing challenge (standard variants of longevity indexing eliminate less than one-third of the Social Security shortfall) and address only one of the many factors that increases second order uncertainty about future Social Security finances.
Social Security’s tax and benefit structure is already reasonably robust against variations in economic conditions. In the long run, benefits and taxes both rise with average wages and thus variations in the growth rate have little long-term impact on Social Security. Program rules, however, are not robust against variations in demographic conditions, whether through fertility, mortality, or immigration. These demographic conditions are a much larger source of long-run uncertainty in Social Security.
A relatively simple mechanism could be incorporated into any reform and help make Social Security more robust. Specifically, the program could be “dependency indexed” to ensure that it can automatically adjust to overall changes in the demographic situation. Some combination of benefits (specifically the so-called “PIA factors” that determine benefits as a fraction of earnings) and payroll tax rates could vary proportionately with changes in the ratio of the aged to the working-age population. In a simplified system this would be sufficient both to restore long-term solvency and to ensure rough annual cash flow balance in the retirement portion of the program. The basic concept of dependency indexing would be relatively easy to explain and motivate, as evidenced by the fact that both President Clinton and President Bush frequently explain the looming shortfall in exactly these terms, describing the rising number of beneficiaries per worker (the inverse of the dependency ratio).
This paper first motivates the concept of “robust solvency” by examining past forecasting errors and likely future uncertainty. The paper then develops the proposal of dependency indexing and shows how it could applied in practice.