The Model
The basic model we start with is a stochastic model of demographic change developed in our previous work (Lee & Tuljapurkar, 1994). This method of forecasting uses historical trends to model future economic and demographic outcomes. Because of this, we obtain less precise but more accurate estimates of potential insolvency. That is, the range of our estimates is wider but more realistic since it is based on history. For a given starting point, the model generates a large number (about 1000) of alternative future trajectories. If we use the size of the population in the year 2022 as an example of an outcome we are interested in predicting, these trajectories would give us a probability distribution of population values in that year. From this distribution, we are able to obtain statistical descriptors such as a range, probabilities, means, and standard deviations. Population values at the minimum and maximum ends of the range would have the smallest chances of occurring while those toward the middle would be more likely to occur. In the experiments we report on here, our outcome of interest is insolvency of the trust fund expressed as the percent chance (the probability) that insolvency will occur in a given year. The program is launched from initial conditions in 1997 (any date could be chosen) and is calibrated to the current tax and benefit levels reported by the Social Security Administration, using their population projections. The model is run in 1-year steps over the forecast horizon.
Summary of Major Findings
Using the Model to Forecast Insolvency
Starting with the trust fund balance in 1997, we use the model to forecast the probability of insolvency over a 100-year horizon. We use current policy in this baseline model: the tax rate is held constant at 12.4 percent, the NRA is raised to 66 and 67 starting in 2000 and 2017, and there is no investment in equities. Taking into account future mortality, fertility, real wage growth, interest rates etc, we find that the median year of insolvency is 2032 meaning that there is a 50 percent chance that the fund will last beyond 2032. This finding is in accord with the Social Security Administration’s estimate. There is a 3.8 percent chance that the fund will reach insolvency by 2022 and a 90.7 percent chance that it will be exhausted by 2047. In the following experiments we test the effects of the three policy measures separately and then in combination.
Increasing the Social Security Payroll Tax
We evaluated four different hypothetical tax increases that would be levied immediately. The current Social Security payroll tax rate is 12.4%. An immediate increase to 13.4% would extend the median date of insolvency to 2044. An increase to 14.4% would extend that date to 2064, and 15.4% to 2095.
Increasing the Normal Retirement Age (NRA)
Presently, the NRA is scheduled to increase by two months of age per year for six years starting in 2000 (raising to 66 by 2005), and again starting in 2017 (raising to 67 by 2022). Using the same six-year phase-in period, we evaluated the impact of four different hypothetical schedules. The first two scenarios use the current rate of increase (two months per year), but the first scenario reaches the increase sooner and the second achieves an NRA of 70 by the end.
- Scenario One: increase NRA to 66 by 2005, and again to 67 by 2012;
- Scenario Two: to 66 by 2005, to 67 by 2012, to 68 by 2019, to 69 by 2026, and again to 70 by 2033
- Scenario 3: to 66 by 2002, to 67 by 2006, to 68 by 2010, to 69 by 2014, to 70 by 2018.
- Scenario 4: Same as scenario 3 with an additional increase to 71 by 2022.
The next two scenarios result in more dramatic changes. Scenario 3 would extend the median year of