Social Security Finances
Correlations with Unemployment:
A proposal for 'Unemployment Insurance' for Social Security

Printable PDF version

Social Security's basic problem: "It's the economy" -- Unfortunately, many members of the general public, as well as Congress -- and even many economists -- talk about "reducing" Social Security with little apparent understanding of the close relationship between Social Security finances and the health of the general economy.

Social Security benefits are not a tax-financed "entitlement." They are a government-sponsored, primarily self-financed insurance program, like many private corporate pension programs used to be. The program's Trust Fund was built up over decades by contributions (not "taxes") from the people who expect -- as in any insurance program -- to receive the benefits they have paid for. The solutions being currently proposed for the system's alleged financial problems involve various methods for reducing those benefits, but fail to respond to the underlying problem.

The relationship to unemployment [1] -- Every spring, the Trustees of the Social Security Trust Fund publish an annual report [2] (2010 version here), which provides very useful statistical appendices. Over the past decade, these have included a chart [3] of three different long-term projections of the system's finances. Here is the most recent version :

Figure 1

Clearly, projection "I" is the only one with long-term viability -- a fact largely ignored in virtually all public discourse, which tends to focus, like the Trustees' report, on projection "II." These projections -- the so-called "Low-Cost" (I), "Intermediate" (II), and "High-Cost" (III) -- differ in their assumptions about various contributing factors. These factors are shown in Table II.C.1 of the Trustees' 2010 annual report. A key difference is the Unemployment Rate:

  I   Low-cost:   4.5%                   II   Intermediate:   5.5%                     III   High-cost:   6.5%                

The notion that the U.S. should -- or even could -- tolerate an average
5.5% unemployment for the next 75 years is unconscionable.

A surprising number of people still believe the long-ago discredited idea that low unemployment is inherently inflationary. Since 1950, the U.S. has had four periods of 4.5%-or-less unemployment. During those years inflation [4] was low to moderate:

Table 1
Period Unemployment Inflation
1950:Q4-1953:Q4 3.2% 3.1%
1955:Q1-1957:Q3 4.2% 3.2%
1965:Q3-1970:Q1 3.8% 4.0%
1998:Q2-2001:Q2 4.2% 1.9%

The average rate of inflation for all these periods was 3.2%; the average for the entire period from 1950-2009 was 3-1/2%. There is no credible reason why our economy, using appropriate tools, cannot be managed so as to maintain stable full employment without inflation, which would greatly improve the environment for both business and labor -- and for Social Security.

"Unemployment Insurance" for Social Security -- Three significant observations offer guidance on the problem of Social Security financing:

These suggest that the best -- and possibly only -- reliable approach to the Social Security problem is to insulate the Trust Fund from two threats to its solvency

  1. Congressional misappropriation of Trust Fund assets for non-Social Security purposes, including deficit reduction and
  2. the effect of recessions.

The former is beyond the scope of this proposal, which addresses the latter by giving the Trust Fund "unemployment insurance," rather than betraying the trust of the millions of people who have played by the rules and contributed to Social Security for decades. Such an approach would involve

  1. defining a level of unemployment that would keep Social Security solvent -- e.g., the 4.5% level in the Trustees' "Low-cost" projection I in Figure 1 above,
  2. calculating the appropriate "unemployment insurance transfer" -- i.e., the difference between Social Security's actual net income, and what that net income would be under unemployment conditions defined in (a), and
  3. transferring from the Treasury's general fund an amount equal to that difference.

The process could even involve transfers to the general fund when unemployment is brought below the level defined in (a) above, as it was during the periods listed in Table 1, above. The "unemployment insurance" calculation in (b) above is the focus of this discussion of the close relationship between unemployment and Social Security finances.

The Trustees' annual report (Appendix VI, Table A4) provides the financial data [5] that govern the Trust Fund's level of assets, as shown in the Trustees' chart of long-term projections, Figure 1, above.

The data series that is of primary interest here is the Net Trust Fund Increase (NTFI), the difference between the Fund's income and expenses. Because it constitutes the growth of the Trust Fund, it is what must be supplemented in recessions by the proposed "unemployment insurance," and its relationship to unemployment is what determines the appropriate level of such supplementation.

Determining the relationship -- Any relationship between two data series is most easily seen in a scatter diagram. This is a plot of points, with one data series on one axis, the other on the other axis. If the points tend to line up, a relationship is likely; if they form a "cloud," a relationship isn't likely.

In looking for a relationship between the NTFI and unemployment, each point represents a year, from 1984 to 2009 [6]. For each point, the Unemployment Rate is plotted on the horizontal axis and the NTFI on the vertical axis. The NTFI values are scaled down using Potential GDP (PGDP) [7]. The resulting scatter diagram looks like this:

Figure 2
OASDI Net Increase w/Interest

At first glance, this looks like a cloud, not a line. But because the two data series are time series that extend over time, a relationship that exists at one time may be changed to a different one by changes in policies and economic factors, so that a "cloud" may actually consist of several line-ups of points representing different relationships at different times. To spot this kind of situation, the points are connected and each point labeled with its year date. The result is a date-ordered scatter diagram [8]:

Figure 3
OASDI Net Increase w/Interest printable

This chart offers a number of suggestive line-ups [9] for certain periods:

1986-89         1990-92         1995-2000         2000-02         2004-06         2007-09

The points falling within these periods can then be tested for correlation using regression lines:

Figure 4
OASDI Net Increase w/Interest printable

This plot shows the regression lines in periods of recession as dashed red, and those in periods of recovery as solid green. The regression parameters for these lines are shown both on the plot and in Table 2 below. These parameters are:

Calculating the "unemployment insurance" transfer -- As just noted, the slope value for each line indicates the NTFI change for each percent change in the unemployment rate. However, as mentioned above, the chart data uses NTFI as a percent of Potential GDP, which must be converted back to the original billions-of-dollars value to be useful in determining the amount of the transfer. The formula for converting the slope to the dollar value of the NTFI shortfall caused by a one-percent change in unemployment takes this into account:

(1)     Shortfall (transfer amount) per percent unemployment   =   ( - Slope / 100 )
x   Potential GDP

The ultimate aim of this exercise is the determination of how much money should be transferred to the Trust Fund at a given level of unemployment. The above formula provides the amount for each excess percent of unemployment. The actual final amount is obtained by multiplying that by the excess of the actual unemployment rate over the target rate (e.g., the one suggested above: 4.5%):

(2)     "Employment insurance" transfer   =   Shortfall per percent unemployment     (from formula #1)
x   ( actual unemployment   -   4.5 )

Table 2, below, illustrates this process. For each period,

  1. it indicates the degree of certainty of the regression-line relationship (R2 value, column 2) being used for that period;
  2. it uses formula #1 and PGDP to convert the regression line's slope (blue column 3) to the funding needed for each percent of excess unemployment (gray column 5);
  3. it then uses formula #2 to convert this result and the average actual unemployment rate for the period into the final total funding needed per year (yellow column 7):
  Table 2
R-squared Value Color Codes
 99%+  95-98% <95%
1 2 3 4 5 6 7
Regression Line PGDP
(bil$ per
R2 Slope
86-89 0.9932 -0.4195 4959.7 20.8 6.0 30.8
90-92 0.9959 -0.1504 6173.6 9.3 6.6 20.0
95-00 0.9834 -0.4913 8398.3 41.3 4.8 11.2
00-02 0.7838 -0.0329 10030.4 3.3 4.8 1.1
04-06 0.9987 -0.1169 12485.2 14.6 5.1 8.4
07-09 0.9994 -0.1221 14576.5 17.8 6.6 36.8

Yellow column 7 shows the average for the period. In practice, the transfer could be done annually, using year-specific values. If this were to be done for the each of the individual years covered here, the average yearly transfer for each period would remain the same as in yellow column 7 -- but the transfers in the individual years in that period would vary. This is particularly true for the most recent period (2007-09), where excess unemployment amounted to only about 0.3% in 2007, but was 4.8% in 2009. The transfers would have been insignificant in 2007, but quite significant in 2009. [10]

Conclusion -- The decision on what to do about Social Security is not separable from what to do about the economy, nor from the moral and financial responsibilities of the federal government to its citizens.

This proposal -- the provision of "unemployment insurance" to insulate the Trust Fund from fluctuations in unemployment -- clearly maintains the government's moral and financial responsibility by honoring the financial commitment made by the millions of participants in Social Security. The proposal also clearly adds Social Security to the group of economic factors known as the automatic stabilizers, which tend to increase deficits in depressed times and increase surpluses in good times, a manifestation of the fact that appropriate federal government spending is the best -- and sometimes the only -- way to end a depression.

Deficit hawks should take note that the automatic stabilizers create a surprisingly large automatic deficit component during times of high unemployment. Currently, this is creating an automatic debt for taxpayers of about $60 billion for each excess percent of unemployment above 4%, for a total of over $300 billion a year.

We have a choice. This automatic deficit is wasted money, money that could be used to rebuild our on-the-edge infrastructure and create jobs -- that is, to invest in lower future deficits by pulling us out of this depression, and reducing the need for supplementation of Social Security's finances. Yes, we do have a choice.

  1. The relationship between Social Security finances and unemployment is intuitively obvious. In a depressed economy, the Trust Fund's
  2. The Annual Report's full title is Annual Report of the Board of Trustees of the Federal Old-Age and Survivors Insurance and Federal Disability Insurance Trust Funds. (back)
  3. The long-range projection chart has changed shape over the course of the past decade, as the assumptions have changed. For those who wish to look further into this, the charts from the past eleven years are still available on the Social Security site and have been collected here). The page also provides a collection of the Trustees' charts of "stochastic projections" -- their estimates of the probability that reality will match their long-term asset projections. (back)
  4. Inflation -- The measure of inflation used here is the GDP Deflator, which covers the whole economy, rather than the Consumer Price Index, which covers only a "market basket" of products. (back)
  5. The data used in this study covers both Social Security programs -- Old Age And Survivors Insurance (OASI) and Disability Insurance (DI) -- known collectively as OASDI. The summary data series are:

    Net Trust Fund Increase
            Total Trust Fund Income
                    Income excluding Interest (i.e., contributions, taxes)
                    Interest [11]
            Cost (98-99% of which is benefits)

    Of these data series, the first -- the Net Trust Fund Increase (NTFI) -- is the appropriate one for determining the necessary level of "unemployment benefits" for the system, since it represents the year-to-year change in the Trust Fund's assets that requires support in times of high unemployment. It is the focus of the analysis presented here. The NTFI's component income and cost data series, listed below it, are important in an analytical, predictive sense for determining what is affecting the NTFI, but are not relevant to the determination of benefit amounts; their relation to unemployment is treated
    elsewhere. (back)
  6. Date range -- The NTFI values used here are annual data for the period running from 1984 to 2009 -- i.e., from the first year following the major reorganization of Social Security finances by the Greenspan Commission and the Social Security Amendments of 1983, to the most recent data as of the Trustees' 2010 report. (back)
  7. Ratio to PGDP -- The NTFI is subject to two influences -- growth and inflation -- to which the Unemployment Rate is not. Growth and inflation are not only irrelevant to the problem at hand, but make it impossible to compare NTFI values -- either to each other across time, or to the Unemployment Rate. To remove these two factors, the NTFI is divided by Potential GDP, for which we use the Congressional Budget Office's measure, available in the CBO's spreadsheet Supplemental Information on Key Assumptions in CBO's Projections of Potential Output available from the CBO's The Budget and Economic Outlook: An Update, August 2009 page. (back)
  8. Making date-ordered scatter diagrams -- The version of Excel used to create this chart doesn't offer this as a chart type. However, once the Scatter type is chosen, double-clicking one of the data points provides a Format dialog from which a Custom or Automatic connecting line can be chosen. Labeling the data points relies on a third-party program known as XY Chart Labeler, available free from Applications Professionals. (back)
  9. Selecting likely correlation periods -- In examining the date-ordered scatter diagrams for all the data series listed in the data note above, these periods appeared to be the optimal ones for correlation in common across all the data series, some exhibiting a better correlation than others. (back)
  10. The need for further research -- The purpose of this paper is merely to raise awareness of the unemployment/Social Security relationship and suggest the logic of an "unemployment insurance" program. Making use of the former to create the latter requires in-depth analysis of the economic, demographic, and policy factors underlying both the periods of correlation and the transitional periods of apparent non-correlation between them. For instance, why are the NTFI regression-line slopes virtually identical in the 86-89 and 95-00 recovery periods, yet different from the 05-06 recovery period, which is instead similar to the 90-92 and 07-09 recession periods? Those who wish to pursue this line of inquiry may find useful the spreadsheet used in this study. (back)
  11. The role of interest -- The key to the long-term solvency of Social Security, or of any insurance fund, is the compounding of interest on the principal. The reason for the rapid collapse of the High-Cost and Intermediate projections (Figure 1, above) is an acceleration of decline caused by the loss of interest on declining principal. In examining the contributions/benefits behavior of Social Security, it can be useful to look at the effect on the NTFI of excluding interest -- one of the data series shown in the background material.
    Note: In the Trustees' table, the column heading for the interest data used here is "Net Interest." However, despite the term "net," almost all of that is the interest income on the Trust Fund assets. A detailed explanation of the factors that make the published figure a "net" one can be found here.) (back)

Written: December 11, 2010
Last revised: February 22, 2011
To IEA home page