Thursday, August 21, 2008

Treatment of uncertainty in new CBO Social Security projections

There is a great deal of uncertainty in projecting Social Security’s finances out over 75 years or more. We know the basic economic and demographic building blocks that determine Social Security’s finances, but we can’t know for sure the value of each over coming decades. In the Social Security Trustees Report, this uncertainty has traditionally been primarily addressed using “high cost” and “low cost” scenarios to complement the best-guess “intermediate cost” projections. For a number of reasons, I think these high/low cost scenarios aren’t particularly helpful. (The recent Technical Panel agreed.)

CBO doesn’t use high/low cost scenarios. Instead, they illustrate uncertainty using a “stochastic” or “Monte Carlo” simulation which assigns probability distributions to each of the main economic or demographic variables and then illustrates the range of outcomes we could expect. (The Trustees also use a stochastic model, but unfortunately it’s not yet the primary descriptor of uncertainty in the Report.)


But even given a model that can calculate the range of possible outcomes, there’s the important question of how you describe this range. You can’t simply do a data dump, you need to find ways that are easily understandable. The new CBO report has a number of very effective ways of portraying uncertainty in Social Security financing. I’ll run through them here.

CBO's Figure 1 shows the 80 percent probability range for Social Security's annual income and costs. The dark lines indicate the median outcomes for each, while the shaded areas denote the range of outcomes. Only 20 percent of outcomes would fall above or below these shaded ranges.

CBO's Figure 3 is similar, except that it focuses on uncertainty in the value of the trust fund ratio (the ratio of the trust fund's balance in a given year to the system's costs in that year). The dark line indicates the median outcome; it hits zero in the year in which the trust fund is projected to become insolvent. The advantage of Figure 3 over Figure 2 is that it also accounts for uncertainty in interest rates, which do not affect annual income and costs. However, many believe that annual cash flows have more substantive importance than the trust fund balance.

Figure 4 is a new chart that I believe is very helpful. It shows the probability that the trust fund will have been exhausted by a given year. Up through the mid-2020s there is almost a zero chance of the fund being exhausted. By the mid-2040s, the likelihood is around 50 percentg. This probability rises until by the 2070s there is only around a 15 percent chance of the trust fund not having been exhausted. One advantage of this chart is that allows an easy comparison of how much a reform plan might improve Social Security's finances. We could say, for instance, that under current law there is a 50 percent chance of the trust fund being exhausted by 2049, but under the reform plan this drops to 25 percent, etc.

Table 3 is also a new addition to the CBO report. It focuses on the probability of the program running deficits of a given size in a given year. For instance, the chart shows that in the year 2050 there is an 87 percent chance that Social Security will be running a cash deficit, a 53 percent chance that the deficit will exceed 1 percent of GDP, and a 16 percent chance it will exceed 2 percent of GDP.

These figures all derive from the same underlying model, but show different ways of describing different aspects of the model's output. I believe the two new figures constitute a significant improvement to how uncertainty is described and should be considered for inclusion in the Socail Security Trustees Report.