Monday, January 12, 2015

New Working Paper on Social Security Replacement Rates

Back in July, I co-authored a Wall Street Journal article with Syl Schieber, the former chair of the Social Security Advisory Board, in which we raised questions regarding how Social Security’s actuaries calculate “replacement rates,” which measure retirees’ income as a percentage of their pre-retirement earnings. We argued that SSA’s method significantly overstates individuals’ pre-retirement earnings by indexing them to nationwide wage growth, which is even faster than inflation. Overstating their earnings causes their replacement rates to look lower, which make Social Security seem less generous and encourages the view that Americans face a “retirement crisis.” We argued that comparing Social Security benefits to retirees’ inflation-adjusted pre-retirement earnings gives a better measure of how well Social Security lets retirees maintain their pre-retirement standard of living.

These differences matter. For instance, a recent CBO report measured replacement rates both ways. Relative to wage indexed earnings, the average person born in the 1980s will receives a social security replacement rate of 48%. But relative to inflation-adjusted career earnings, Social Security provided a replacement rate of 64%, one-third higher. If you assume that the typical person requires a replacement rate of around 70% -- a common financial advisors benchmark, as well as calculated in some academic studies – these two figures paint very different pictures regarding the adequacy of Social Security benefits and retirement incomes overall.

Our op-ed coincided with the Social Security Trustees’ decision to delete the actuaries’ replacement rate calculations from their annual report. After that, it was game on: Boston College professor Alicia Munnell, whose National Retirement Risk Index uses a method similar to SSA’s actuaries, pushed back hard on our article. As did SSA’s actuaries themselves, publishing a defense of their methods back in July. More recently, the CBO and OECD issued reports that can be taken to support our point of view. I recently spoke to the Social Security Advisory Board regarding this issue, and the Board’s own Technical Panel on Assumptions and Methods – which is chaired by Munnell – took up the replacement rates question at their opening meeting.

To move the debate along, I have a new AEI working paper on replacement rates co-authored with Syl Schieber and Gaobo Pang, an economist at the pension consulting firm Towers Watson. Among the points we make:

  • Until recently the Social Security Trustees calculated replacement rates relative to career average earnings indexed to wage growth; SSA’s actuaries continue to do so. This measure effectively compares the benefits paid to new retirees to the earnings of today’s workers, not to retirees’ own pre-retirement earnings. These “wage-indexed replacement rates” understate the ratio of retirees’ benefits to their own real pre-retirement earnings.
  • We argue, and recent reports from the CBO and the OECD concur, that calculating replacement rates relative to inflation-adjusted average pre-retirement earnings is a better shorthand representation of the life cycle approach to retirement planning, as well as being more understandable to policymakers and individuals saving for retirement.
  • The SSA actuaries’ method of calculating replacement rates for hypothetical worker examples is calibrated to produce a pre-determined result. Prior to 2002, SSA calculated replacement rates using a different method and using different hypothetical workers. But these previous hypothetical workers had very unrealistic earnings patterns, so SSA updated to more realistic stylized earners. But SSA then calibrated its methods to produce the same replacement rates figures as before: it changed its method of calculating replacement rates and increased the earnings of its stylized workers in order to reduce measured replacement rates to its previous value of around 40%. There is no reason these calculations should be given any special importance.
  • A recent SSA actuarial study using administrative data concluded that wage-indexed replacement rates closely replicate those calculated relative to final earnings, a common practice for financial advisors. But the SSA OACT study excluded all spousal and widow benefits, thereby reducing measured replacement rates. Over one-third of female retired worker beneficiaries receive auxiliary benefits and, on average, auxiliary benefits increase their monthly payments by 78%. An analysis that included all beneficiaries and all benefits received by them would show higher replacement rates.

There is a lot of new information in our paper about what replacement rates mean and how they have been measured and I believe we add a great deal to the current debate.

10 comments:

WilliamLarsen said...

"that calculating replacement rates relative to inflation-adjusted average pre-retirement earnings is a better shorthand representation of the life cycle approach to retirement planning, as well as being more understandable to policymakers and individuals saving for retirement."

Andrew as you should surely understand, inflation adjusted only maintains an income stream based on the year in which the value is set. For example a person who draws SS-OASI in 2000 will have their benefit paid to support those goods and services identified in the basket of goods in 2000. When new products come out after 200, they are not in the basket of goods being indexed for and therefore, are outside the buying ability of the beneficiary. This is why so many beneficiaries say that inflation does not maintain their standard of living - they are trying to maintain their standard of living in relation to that of workers.

That said, if you used inflation indexed wages, then you are forever fixing the buying power of 1975 to those goods in 1975 and those wages in 1976 to 1976. If that is the purpose of SS-OASI then so be it.

However, I would want some type of return above inflation for my tax dollars.

Based on my calculations, those born after 1985 will receive as a cohort basis, 29 cents for each $1 in combined payroll tax and credited US Treasury Rate interest. In simple terms you could burn 71% of theoretical invested payroll tax in US Treasury Notes and still match what SS-OASI will pay.

Arne said...

I understand trying to make "replacement rates" more useful, but changing from one measure that does not match the one used by financial advisors with clients to another measure which also does not match seems like a waste of effort.

From what I know of both engineers' and teachers' salaries, they go up faster than AWI for several years and then drop back to about CPI. This means that an AWI adjusted PIA will be systematically higher than a final salary and a CPI adjusted PIA will be systematically lower.

The use of averaging over 35 years is a compromise to better align benefits with individuals' taxes (since some workers don't follow a single career salray trend). It is not going to conform to retirement planning. Perhaps if the SSA is going to use "replacement rates" they need to just need to use final salaries for hypothetical workers.

Arne said...

"Based on my calculations, those born after 1985 will receive as a cohort basis, 29 cents for each $1 in combined payroll tax and credited US Treasury Rate interest."

This calculation does not conform to analyses I have seen from the SSA. A bit ironic given that the purpose of the original post is to move towards meaningful comparison of diverse analyses.

Andrew G. Biggs said...

Arne, you actually have it the other way around: a CPI-indexed replacement rate will be higher than a wage-indexed one. E.g., see the recent CBO report or my 2008 paper with Glenn Springstead. Your other point is valid, but I think outweighed by the greater accuracy and understandability of a CPI-indexed replacement rate. Probably not one in a thousand people understands how SSA calculates replacement rates, and if they did, they'd know that SSA's approach doesn't make much sense.

WilliamLarsen said...

" Probably not one in a thousand people understands how SSA calculates replacement rates, and if they did, they'd know that SSA's approach doesn't make much sense."

What is the purpose of replacement rate?

I have read the their approach and understand it. The problem is that the method does not agree with reality of funding a Social Security Program.

We are dealing with math here. We have two sides of an equation that in the ideal world equal each other "costs - revenues."

Clearly there are variables that can be measured: cohort life expectancy at ages 0-110. Wage history, US Treasury rates, Inflation to a degree. Female workers, male workers. Beneficiaries - widowed, married, divorced, never married.

Has Social Security ever tried to balance both sides of the equation? Clearly from the previous numerous changes to the OASI program starting in 1950 or if you like 1939 and 1940 when congress delayed the tax increase from 2% to 3% and allowed beneficiaries to start collecting benefits in 1940 v 1941, I have not seen any attempt to balance both sides. The big fix of 1983 only delayed and made the inevitable worse. Now we seem to debating indexing as being something that can save social security. Has no one looked at both sides of the equation?

Andrew G. Biggs said...

I'm not sure how the measurement of replacement rates relates to the program's funding. A replacement rate is designed to measure the degree to which a person's retirement income (from Social Security or other sources) allows them to maintain their pre-retirement standard of living. In general, as life expediencies rise then TARGET replacement rates will fall slightly, but that happens whether you measure replacement rates relative to wage-indexed earnings, real earnings, final earnings, etc.

Arne said...

replacement rate = benefit / salary

A higher PIA is a higher salary leading to a lower replacement rate, so we are actually agreeing.

WilliamLarsen said...

"I'm not sure how the measurement of replacement rates relates to the program's funding. A replacement rate is designed to measure the degree to which a person's retirement income (from Social Security or other sources) allows them to maintain their pre-retirement standard of living."

This may be the entire problem in a nut shell. How do replacement rate affect social security funding.

When you plan for retirement you need to determine what you need in the future. One way is to use a percentage of current income. The replacement rate in the future relative to today's income is a replacement rate. The replacement rate can be anything; sp-500, dow jones, NSADAQ, inflation, wage growth or even zero. It is the rate at what you want today's income to be replaced at in the future. If you do not define this now, then you can not later define it in the future for the years to save have passed you by. If this is the case you need to skip to the last step, How long will it last v How much do I need to save?

I will try formulas that hopefully will identify where replacement rates come into play.

1 Determine the retirement needs per year in current dollars.

2 Determine the following variables:

2.1 Rate of return assets will earn I = __________

2.2 Replacement rate G = __________

2.3 Years till retirement N = __________

2.4 Number of years in retirement X = __________

2.5 Present value of retirement assets PV= __________

3 Calculate future value of retirement needs per year.

FVneeds= (STEP 1) * (1+G)N

4 Calculate future value of assets required to sustain retirement needs in step 3.

Ie = (1+I)/(1+G)-1

Effective Rate of Return

FVassets =(STEP 3)*((1+Ie)X-1)/((1+Ie)X *(1+G)*Ie)

Future value of assets needed.

5 Calculate the amount to save each year to reach goal in step 4. Increase amount saved each year by inflation or own goal. NOTE: If other than inflation a new effective rate must be calculated for this step only.

SAVEyear=((STEP 4)/(1+I)N))*(1+G)*(1+Ie)N *Ie/((1+Ie)N-1)

IF YOU HAVE PRESENT ASSETS

SAVEyear=((STEP 4)/(1+I)N-PV))*(1+G)*(1+Ie)N * Ie/((1+Ie)N-1)

6 Recalculate each year up till retirement to minimize any large fluctuations in rate of return, inflation, savings rate and retirement age.

7 During retirement recalculate yearly disbursements based on the previous years variables determined in step 2. Note: X will decrease each year you are retired.

DISBURSEMENT= (ASSETS*(1+G)*(1+Ie)X * Ie)/((1+Ie)X-1)

Andrew G. Biggs said...

I understand your calculations, but the crucial step (as far as my working paper was concerned) was in figuring out your target. Does a retirement income relative to your current income make sense? Relative to your career-average earnings -- and if so, how indexed? Your expected final earnings, etc? That's the hard part, but different choices at this stage make a big difference later on.

WilliamLarsen said...

I understand your calculations, but the crucial step (as far as my working paper was concerned) was in figuring out your target. .....Relative to your career-average earnings -- and if so, how indexed? Your expected final earnings, etc? That's the hard part, but different choices at this stage make a big difference later on.

It appears your paper is a theoretical paper or as an attorney once told me an intellectual challenge. An intellectual challenge is great, it opens up the mind, makes you think outside the box. However, in this case, social security, it appears that the writing is intended to sway perception in order to make policy changes; "but different choices at this stage make a big difference later on."

At the same time I would hope you also present the cost in real terms to those who are working paying 10.6% of their wages into OASI. As Altmeyer clearly articulated in 1943 and 1944, future generations will pay far more for their benefits than they are worth.

Policy changes, changes to SS-OASI must be presented so that workers now know weather it is worth continued support of Social Security or repealing it.

People should not be like the frog that found a nice cool pot to land but later found as the water increased was actually a cooking pot.

Truth in lending is common place, Truth in Social Security is not.