How much trouble is Social Security in if valued with market interest rates?
A few weeks ago, I received a call from Jim Grant of Grant’s Interest Rate Observer. Jim wanted to discuss the status of public (state and municipal) pension plans whose valuations are done using an assumed rate of return which for decades has generally hovered in the 7.0–8.0% range. This differs from the valuation of private sector defined benefit plans which have been valued for over three decades using market interest rates (for accounting purposes, AA-rated corporate bonds) of similar maturities or duration as the forecast benefits. These currently are in the 2.5–3.0% range. Given the rather long horizon that benefits are projected over, small changes in rates have large impact on the value of liabilities. This varies by individual pension plan, but as a rule of thumb, for every 1% drop in rates, the value of the liabilities might go up by roughly 15%. So, with a discount rate that is off by 4–5%, the true current economic value of public plan liabilities could be 60–75% higher than what states and municipalities are reporting!
However, as I said to Jim, “The states and municipalities are only the rhinos and hippos in the room, I’m afraid. The real elephant is Social Security!” Whereas the combined underfunding of municipal plans is a few trillion dollars, I hazarded a guess that given its scale, Social Security’s reported underfunding might be off by tens of trillions. So, I decided to estimate that number.
How bad would Social Security’s underfunding be if we used market interest rates? Figure 1 below compares the present value of social security deficits (PVSSD) under three assumptions to the value of outstanding market-traded US Treasury debt for valuation dates over the past 16 years. The blue bars show the outstanding Treasury debt, and the lines present the PVSSD for three assumptions, Intermediate (orange) and High Cost (gray) as provided by the Social Security Administration’s (SSA) Trustees Report¹, and my estimates (red) using market yields on Treasury Inflation-Protected Securities (TIPs) as of the corresponding valuation dates. Between 2004 and 2020, under the intermediate assumptions, PVSSD has grown from $3.7T to $16.8T, or from 54% to 74% of outstanding public market debt. Under the high cost assumptions, the corresponding numbers triple in value from $10.3T to $30.9T, or are 36% more than outstanding debt this year. Using market yields, the PVSSD has rocketed from $6.2T to $45.6T, more than seven-fold, or from 91% to twice the value of market-traded Treasury securities. It is important to note that the market value numbers are estimated by making adjustments to the Intermediate Cost values. The only assumption changed to the SSA’s baseline assumptions is that for interest rates.
The graph shows that between 2004 and 2010, the market rate-based results were nestled between the Intermediate and High Cost results. After 2010, they diverged and proceeded on a more dramatic upward trajectory. To understand why, we need to examine the path of interest rates and appreciate why the interest rate assumption has such a dramatic impact.
Which interest rates should we use? First, as these are future US government taxes and payments that are being considered, it is logical to use a sovereign rate instead of a corporate discount rate. Second, because Social Security has income that is a function of economy-wide wages, and expenditures that grow with inflation, it makes sense to focus on real interest rates instead of more traditional, nominal rates. Indeed, after 2004, the SSA directly specifies its ultimate interest rate assumptions as real rate. Prior to the introduction of Treasury Inflation-Protected Securities (TIPs) in 1997, economists and actuaries had to use traditional, nominal securities and estimate what portion of the current yield was compensation for (expected) inflation and what was the real rate of interest. With the advent of TIPs, the real rate is specified contractually, and the realized inflation accrues annually. Similar to nominal Treasury bonds, the US Treasury regularly issues TIPs in 5, 10, and 30-year maturities and those securities are actively traded and fairly liquid.
Figure 2 shows the yield on 30-yr TIPs versus the assumed real interest rate used in the SSA Trustees Report for the years 2004–2020. Valuations are as of the end of the federal fiscal year prior, so TIPs yields are as of September 30 of year T-1. The SSA actuaries run 3 sets of assumptions: Intermediate, which is their base case, low cost (not shown) and high cost. The orange line corresponds to the intermediate case, the gray line the high cost case, and the red line is the market interest rate for the 30-year TIPs, the longest maturity, market-determined real rate.
Note that in 2004, the high cost discount rate was actually slightly lower than the market yield on TIPs (2.2% vs 2.29%). Combined with the myriad of other assumptions that go into the valuation, this “high cost” scenario was then probably truly labeled as shown in Figure 1. Also, for the valuations after 2004 through 2010, the high cost discount rate was reasonably close to the market yield if always a bit higher. Combined with other more conservative assumptions than those used for the intermediate baseline, the high cost case clearly produced more conservative estimates of unfunded SS obligations. After 2010, market yields diverge substantially from the SSA discount rate assumptions, and most worrisome, are moving further apart. In 2010 and before, the high cost discount assumption was set 0.8% of a percent below the intermediate assumption. After 2010, this spread was narrowed to ½ percent. This was a judgement call as the Fed was seen to be manipulating the yield curve through open market purchases and keeping long rates “artificially” low. Note however, that long interest rates were low throughout the developed world and not just in the United States! Long real rates continued to fall, dropping precipitously in 2010–2012, and then trending down again from 2014 to today. The acceleration upwards in market valued PVSSD since 2010 as shown in Figure 1, reflects the behavior in market yields. Also bear in mind that the last rate shown was for September 30, 2019, before the Covid-19 pandemic. Since 2014, the assumed discount rates have been reduced, contributing to a substantial rise in PVSSD in both SSA cases, but at a much more attenuated pace than market rates. The spread between the intermediate SSA case and the market yield which was approximately 0.7% in 2004, widened to 1.8% this year.
A word on the time value of money and interest rates…
Readers with a background in fixed income can likely skip this section.
Most homeowners are aware that the amount of house they can afford depends on the monthly payment they make more so than the price of the house. The mechanism that ties the two together is mortgage mathematics. The fundamental inputs are the size of the mortgage, the number of payments, and the interest rate. In the US, most houses are bought with 30-year mortgages with monthly payments. So, for our example, let us assume a 30-year term and say we are looking at a $500,000 initial mortgage, leaving interest rate as the variable. Table 1 shows the monthly payment varying the interest rate from 2% to 5%.
As rates drop, for a fixed amount of mortgage principal, so do the payments. Let us assume our house buyer could just afford a $2500 monthly mortgage payment. If the prevailing rate was 5% — the $500,000 mortgage would be just out of reach — the largest mortgage they could get would be for $465,700. However, if rates were lower, they could afford the house.
Note that the mortgage math can also be run in reverse. Given a known loan term, payment amount and interest rate, we can solve for the corresponding principal balance. Table 2 shows how the principal balance changes as rates are varied given a $2500 monthly payment.
So, with a monthly budget of $2500 for mortgage payments, our buyer can afford more house as rates fall. If rates dropped to 3%, they could afford a mortgage of nearly $600,000 and if rates fell to 2%, they could afford a little over a $675,000 mortgage. However, if rate stay low or continue to trend downward, it doesn’t necessarily mean that they will be able to buy more house, just that the value of all houses will trend upward and the house that once required a $500,000 mortgage now may cost substantially more. This, in a nutshell, is what happened to the housing market in the early 2000’s — interest rates fell, and housing prices appreciated. While good for many, it had pernicious secondary effects as homeowners took out equity that they had in houses they owned and/or levered up to purchase more homes. This, combined with the financial engineering of Wall Street and less than scrupulous mortgage brokers, aided and abetted risky behavior as documented in The Big Short and a host of other books about the financial crisis.
Just know that the same math that works for our house example works for pensions. However, instead of 360 monthly payments that are level in time, we are dealing with something on the order of 75 years for a pension plan — benefits in payment for current retirees, and benefits to be paid in the future for those currently working. In the case of Social Security, benefit payments are increasing in time as Social Security benefits are indexed for inflation. System-wide, the benefit payments are also growing as the population expands and there is wage growth across the economy. These factors make the liabilities of the Social Security system much more sensitive to changes in interest rates than a 30-year mortgage!
There are two fundamental components of the Present Value of Social Security Deficits, the Actuarial Balance and the Present Value of Future Taxable Payrolls.
Actuarial Balance — What’s the tax shortfall to make things balance?
One of the numbers that is featured prominently in the presentation to Congress and policy makers is the Actuarial Balance (AB). This is the tax shortfall (or rarely, surplus) as a percentage of taxable payrolls that would be required to keep the system in balance. The Actuarial Balance is equal to [PV(future contributions) — PV (future benefits)] /PV(taxable payrolls), or AB = PVSSD/PVTP. For our purposes estimating unfunded, we simply re-arrange to get PVSSD = AB x PVTP. To get PVTP(r*) we need to estimate AB(r*) and PVTP(r*), where r* equals the market TIPs yield, from data reported in the report.
In the SSA Trustees Report, the AB is presented repeatedly and analyzed for the intermediate assumption case. In 2004, the Actuarial Balance was -1.89% and in 2020, it was -3.21%. In 2004, someone making $50,000 would have had to pay an additional $900 in payroll taxes — likely split $450 each between deductions and employer contributions. Wages rise in time, so suppose the same individual’s pay went up in line with wages broadly reflecting inflation, productivity growth, etc., they would make roughly $90,000. Today if the tax rate were adjusted to keep up with the Actuarial Balance, they would be paying an additional 3.21%, or nearly $2900. As a former grad school classmate said to me years ago, “A problem deferred, is a problem compounded!” But wait! There’s more!
Buried deep in the Appendices of the report is a table for sensitivity analysis. (Table VI.D6, p.187) The sensitivity varies every year, but for 2020, for every one-percent change in interest rates the actuarial balance shifts by 0.39%. So, if we plug in the September 2019 TIPs yield, the estimated Actuarial Balance would be -3.92%. On our $90,000 example, the $2900 in additional taxes rises to a bit over $3500. Figure 3 shows how the Actuarial Balance has evolved through time and compares that to the estimated AB’s with market interest rates.
If you are as surprised by the picture as I was — I expected the estimated balance based on the TIPs yields to diverge further — there are two reasons why they do not. First, the Actuarial Balance reflects the difference in the present value of future tax receipts (and interest payments on the Social Security Trust Fund) and the present value of future benefits, divided by the present value of future payrolls. Both items in the numerator are going up, but so is the denominator!
Second, and to my mind a bit more insidious, the report only presents a linear estimate of the rate change at the intermediate rate +/- 0.5%. Go back and look at our hypothetical example in Table 2. Going from 5% down to 4%, the value of the mortgage goes up by $57,949 or 12.4%. From 3% down to 2% it increases by $83,398 or 14.1%. This is a property fixed income practitioners call convexity.² What leads me to believe there should be convexity in the Actuarial Balances? From 2012–2020, the alternative rate assumptions were +/- 0.5% from the intermediate case and deltas in the Actuarial Balance per 1% move ranged from 0.38%-0.45%. In 2010 and the prior years, the deviations were +0.7% and -0.8%, and the deltas per 1% rate move were 0.48%-0.66%. Therefore, the estimated Actuarial Balance for market rates should likely be a little bit lower. How much is difficult to say without deeper analysis.
Present Value of Future Taxable Payrolls and the Unfunded Balance
Here is where the numbers grow dramatically. If the Actuarial Balance was the only piece that was sensitive to changes in interest rates, the present value of Social Security benefits would be about $4T higher than the Intermediate case. The present value of taxable wages, however, is quite sensitive to changes in interest rates³, especially as the discount rate gets closer to zero.
Under the intermediate assumptions, PVFP goes from $211T to $555T, an annual growth rate of 6.2%. Under market discount rates they go from $267T to $1167T. Compare those values to 2019 taxable payrolls of $7.6T. At a 2.3% real rate, the PV of 75 years of taxable payrolls are approximately 72.8x the last year’s payroll — the discount rate is just high enough to overcome the growth due to real wage growth, demographic growth, net immigration, and a host of other factors. Using market interest rates, the effective discount rate becomes a net growth rate, and the PV of each future year is higher than the current level of taxable wages.
The SSA Actuarial Report also shows an infinite horizon figure for the unfunded balance which they estimate to be $53T using the intermediate assumptions. It is notable that they do not present the figure using a high cost scenario. Why? By my estimates, the “event horizon” where the interest rate is low enough to generate an infinite cost for the infinite horizon lies just below the assumed 2.3% real rate used for the intermediate case. It will be interesting to see how much the actuaries drop their base case, intermediate interest rate assumption for 2021 and assuming they do so, if they are even able to generate infinite horizon valuation numbers going forward.
As of September 30, 2020, the end of the fiscal year, the 30-year TIPs yield is -0.33%, 0.85% below where it was a year earlier. The PVFP will be quite a bit higher and so will next year’s value of unfunded liabilities.
Implications
To those of you who feel my intent is to bury Social Security, nothing could be further from the truth. Let me give it the praise it is due. Between 1960 and 2018, poverty among the elderly fell from 35% to 9.7%. That is a direct result of changes made to Social Security and the introduction of Medicare in 1964. Social Security provides forced retirement savings for nearly all workers in the US, along with a wealth transfer from high earners to the poor and disabled given its progressive benefits structure. It provides inflation-indexed annuity benefits, something that the private markets have generally failed at. Its administrative costs and fees are microscopic compared to either defined benefit or defined contribution plans. In 2019, 64 million Americans received Social Security benefits — more people than those who voted for our President.
But, and there has to be a “but,” its costs have been mis-measured over the last two decades and dramatically so more recently. As we have seen with public pension plans, persistent bad measurement leads to bad decision-making. These include granting of new benefits when a plan cannot truly afford them, poor policing of participants who game the benefits system, as well as perverse and downright risky investment schemes, among others. For the public plan actuaries who consult to states and municipalities, their livelihoods depend on not learning central tenets of financial economics. Their integrity, and hence their assumptions, are compromised by the incentives of elected politicians who know that raising taxes today to pay for employee benefits due long after they are in office is political suicide. To quote Upton Sinclair, “It is difficult to get a man to understand something, when his salary depends on not understanding it.”
The SSA Actuarial Report runs to 276 pages. It is sent to the Speaker of the House and President of the Senate with a cover letter signed by the Secretaries of Treasury, Labor, and Health and Human Services, three of the Trustees. I doubt strongly that any of those five have read it in total and am skeptical that any of them know what to focus on. Actuaries highlight the Actuarial Balance. Economists and financial analysts will look at the unfunded, but how many of them really dig into the numbers and stress-test the assumptions? As the late Doug Love told me once, “The actuaries have conflated capital budgeting and valuation.” Capital budgeting prizes stability and seeks to avoid year-by-year swings that are large. Valuation allows us to compare things. Markets can and do move a lot when economic circumstances or other important factors change meaningfully. Prioritizing capital budgeting to such a degree that important assumption inputs are divorced from market realities pollutes valuation!
The actuaries also construct a valuation allowing for uncertainty around input assumptions. However, the highest figure given for this stochastic valuation of the unfunded obligation is $37.2T over 75 years. If the figure with contemporaneous market rates is $45.6T, something is seriously flawed with the range of inputs to the simulation. Valuation using market-determined, contractable rates should lie somewhere close to the mean or median of the distribution!
Major modifications to Social Security were last made in 1983 when normal retirement ages were slowly phased higher. President Bush established a bipartisan commission in 2001–2005 and proposed legislation that would have addressed then-known issues. Nothing was done. President Obama set up the Simpson-Bowles Commission in 2010 to study entitlements and other fiscal issues. Nothing was done. Here we are a decade later and the problems are worse than ever — A problem deferred is a problem compounded. The Social Security Trust Fund, a $2.9T “lock box” of intra-government IOU’s, will run out of assets sometime between 2030–2035 according to the actuaries as the gap widens between benefits paid out and payroll tax receipts. Maybe a year or two earlier, but that’s close enough for government work. They will run out before I, and likely many of you, hit Normal Retirement Age.
To summarize:
- The finances of Social Security are in much worse shape than both decisionmakers and the American public realize. By using optimistic, off-market assumptions and burying the most important numbers, the actuaries and Trustees have contributed to complacency. Congress and the executive branch need to make decisions using good numbers. They have yet to see them.
- Washington has kicked the can down the road for too long. It has been 37 years since anything truly substantive was done to address Social Security’s fiscal deficits. Those insufficiencies, once small cracks, have grown to a yawning chasm.
- Tough decisions need to be made in the next administration. Americans would like to see benefits increased! They’d also like to see a second round of stimulus during these difficult times. Is anyone going to tell them the news they don’t want to hear? We have tough decisions to make between cutting benefits and raising taxes now on lifetime promises. If not, our children and grandchildren will be burdened with a bill beyond comprehension.
___________________________________________________________________
Notes:
- https://www.ssa.gov/oact/tr/2020/tr2020.pdf
- In reality, traditional fixed-rate mortgages (and hence mortgage securities) are not convex as the mortgage holder has the right to pre-pay the mortgage at any time to the lender without penalty.
- It should be noted that here a simplifying assumption has been made versus what the SS actuaries do by brute force to get the PV of Taxable Payrolls. With the current payroll and PVTP given the discount rate, one just needs to know some high school algebra, a little arithmetic, and a spreadsheet to find the PVTP under the new discount rate. There would be some difference if the actuaries calculated this using brute force and computers given that their various growth and discount rate assumptions are not constants, but I doubt the order of magnitude would change materially. For the details of the calculations, contact the author or if there is enough demand, I will post a “technical” appendix.
