Tuesday, December 29, 2015

CBO's new replacement rate numbers open Social Security, retirement debate

Over at Forbes, I write about how new figures on Social Security "replacement rates" published by the CBO in response to a recent expert panel report to the Social Security Advisory Board will open questions about the adequacy of Social Security benefits and of Americans' overall retirement savings.

The short story is that CBO's new numbers -- which compare Social Security benefits to the final five years of substantial earnings prior to age 62 -- show Social Security benefits coming far closer to the 70 percent total replacement rate recommended by most financial advisers. To me, this doesn't make a strong case for expanding Social Security or for declaring a broader "retirement crisis."

Click here to read the whole article.


WilliamLarsen said...


A very good reason to scrap social security. No one decided on what the criteria should be. It has always been left up to a politician to decide. Replacement rates as you stated mean different things to different people. Defining the terms is a top priority in any project so that there is no misunderstanding of the objective and design. With out definitions can you have defined criteria?

It is amazing that in 2015, 78 years after SS began, the issue of replacement rates and value of benefits now and in the future have not been nailed down. It is as if we set sail on a bamboo raft and over the years we are trying to build something new out of the same material.

Joe the Economist was correct when he wrote that it is critical we define what social security is before we do anything.

Reading this makes me shake my head in disgust. This late in the game and the rules are still changing. When I was a kid, any time the gang continued to make up new rules we left. We did not stand there and take it. We stopped it.

The value of social security is pretty simple and straight forward. The SSA now has an online benefit calculator (of course if congress changes things, it is totally meaningless). However, it is far better than a bunch of yahoos who are trying to confuse people over terms. Simply take the SSA benefit amount that is estimated and divided it by the workers yearly wage. That is the replacement rate.

In addition, the two bend points have three values 90%, 32% and 15%. Again it is pretty simple knowing the indexed income to determine the benefit (See above from SSA). Divide the calculated benefit by the indexed wage and you get a replacement number.

The only way you get much above 40% is that the wage is relatively low, far less than the average wage.

Arne said...

"The only way you get much above 40% is that the wage is relatively low, far less than the average wage."

I guess it depends on your definition of "much". Anything below $79K per year gets at least somewhat above 40 percent.

To do the math with the bend points you need the Primary Insurance Amount and you need to assume that the worker is full time for the entire 35 years. At that, you would get that anything below $79K per year gets at least somewhat above 40 percent. A minimum wage earner ($20K per year) would get 62 percent. From the table Andrew posted we can see that actual salary histories do not match the PIA all that well (which is corollary to his point).

"No one decided on what the criteria should be."

It is more that everyone wants to be able to define the criteria themselves. If you decide to analyze it as retirement savings instead of as a social safety net with immediate value, you are bound to be dissatisfied.

WilliamLarsen said...


"It is more that everyone wants to be able to define the criteria themselves. If you decide to analyze it as retirement savings instead of as a social safety net with immediate value, you are bound to be dissatisfied."

I do not look at it as retirement savings. I view social security as a cost. With that cost, one expects to get something for it. It may be lower taxes else where, improved standard of living or it offsets saving for retirement in which case it could be viewed as a retirement account.

What bothers me is that each succeeding cohort pays a bit more for their benefits than the previous. This continued downward spiral is leading to a flat to reduced standard of living.

You may like the rules changing all the time. You may also like not planning ahead of time. You may also begin a project with no idea of what it will look like, what materials you will need, the cost of those materials, compatibility of those materials and you may not care that after putting in so much work that in the end it fails. You may be a happy go lucky person who could care less how it affects others.

As an engineer, I attempt reduce unknowns. When I implement a project I have bench marks to measure the success of the project. The Federal Government General Budget has not been balanced in over 58 years, every year after 1957 has been deficit spending.

Reagan took us off the gold standard and from that point on we have paid for imported goods with "Art Work" or "US Treasuries" we exchange for goods and services. When was the last time this country had a surplus in in exports v imports?

It is one thing to have a minor out of balance, but continued imbalances are going to bring it all tumbling down. The government might as well begin putting different characters on the treasury notes making them more collectible. Disney shares now have Marvel and Star Wars themes surrounding the original Disney icons: snow white, cinderella, etc.

If you think the 119 million potential voters under age 46 will forever continue to see their benefits decrease, I think you are wrong. Will they slowly change SS or will they yank it out? If people like you seem to be so careless with a program then you get what you deserve.

JoeTheEconomist said...

I am guessing that the replacement rate doesn't include the projected 40% reduction of benefit levels - CBO 12/15.

Arne said...

"Earnings are “substantial” if they amount to at least half of a worker’s average indexed earnings."

This means that they have included reduced income for anyone who worked more than 6 but less than 12 months. The similar SSA report:
specifically addresses this issue and gets substantially lower numbers - 40 percent instead of 59 percent at the median. (there is obviously something different as well to create such a large discrepancy)

The 70 percent replacement rate for people who can afford a financial planner is also unlikely the be representative of the entire population. (as other numbers in the SSA report demonstrate)

Andrew G. Biggs said...

The SSA OACT paper (if I'm remembering right) should have included more years of lower earnings than the CBO approach. Say, if a person earned 25% of their AIME in a given year, that would be counted in the average by OACT but not by CBO. There's no perfect way to do this, especially if you look at only a few years of earnings, because earnings fluctuate even when someone is working full time, plus many people reduce their work hours as they approach retirement. What you want to know is how Social Security benefits compare to someone's pre-retirement spending/consumption, which is harder to get at.

The SSA OACT paper, in my opinion, has a number of issues. For one, it doesn't include spousal/widow's benefits, which are a big deal for many women. It's distribution figures also aren't easily comparable to CBO's; CBO shows mean replacement rates based on the distribution of earnings; SSA OACT shows the distribution of replacement rates, which isn't the same thing and isn't very meaningful. Syl Schieber, Gaobo Pang and I looked at some of these issues in this paper: https://www.aei.org/publication/measuring-communicating-social-security-earnings-replacement-rates/

Arne said...


Your 2015 critique of Goss 2014 (OACT) showing why you might expect that wage indexing is inappropriate would make more sense if wage indexing showed different results from using the final five years, but they produce (coincidentally?) about he same result. I do not see where your critique indicates a meaningful problem with their 5-year sample data. (Claiming that 2011 is depressed and late 1990s is normal does not pass the smell test.) (Widows' benefits have much more to do with maintaining benefits later in life than in determining replacement rate at retirement.)

So, I am still stuck with why does CBO that purports to use 5 salary year data very so much from OACT data that purports to use 5 year salary data. (The medians of two correlated data sets will coincide if the sample is large enough.) Why should I prefer CBO? In addition to the methodological oddity I pointed to in my earlier comment, I see that CBO says "All workers are assumed to claim benefits at age 65" while OACT using actual claiming at an average of 63.75. It seems clear to me that I should prefer the OACT numbers!

Andrew G. Biggs said...

Note that wage indexing doesn't show the same results as using the final five years of earnings. Wage indexing of career earnings -- INCLUDING zero years -- produces similar results to looking at the final five years of earnings, EXCLUDING zero years. They're not calculated the same way and there's nothing in OACT's memo explaining why you'd include zeros in one measure and exclude them in the other.

You're right that widow's benefits play in more late in life. But spousal benefits don't. To me, there just doesn't seem any good reason to exclude them.

In terms of why the CBO and OACT numbers differ, I obviously can't say for sure. We found that when we restricted our sample to workers who actually "looked like" the stylized medium scaled earner, i.e., people with relative full working careers, replacement rates were (I think) about 10 percentage points higher than what OACT found for their full sample. With CBO restricting to people with at least 20 years of work, versus around 10 years for OACT's sample, that might push things up.

OACT had previously published replacement rates as of the normal/full retirement age, which makes sense to me for several reasons. CBO's gone with claiming at 65, which also seems reasonable as a 'traditional' retirement age. I'd have to think about how big a difference it would make. I'm going to look more into this once I get a couple projects off my plate -- hopefully next week.

Arne said...

You say: "there's nothing in OACT's memo explaining why you'd include zeros in one measure and exclude them in the other"

OACT says, about replacements rates: "For individuals with steady, consistent earnings up to retirement, we generally consider the ratio of their retirement income to their income immediately before retirement."

The paper does not refer to financial planners' 70 percent rule directly, but it is clear that such a rule applies to workers with steady, consistent earnings up to retirement. So how does real data relate for workers who do not have steady earnings or do not have earning up to claiming?

"Another approach, one that more nearly approximates the intent of benefit replacement rates for steady, consistent workers, considers actual earnings levels late in career."

So, they hypothesize that including zeros in the final years salaries is non-representative, then they show that the data do show a difference. Then they show that (empirically) wage indexed data including full salary histories match the preferred (excluding zeros) denominator.

Their conclusion "For the general population, with wide variation in earnings patterns through a career, this approach reflects well the relative standard of living experienced by a worker over a career", does not highlight that the reflection is empirical, opening them to your justified critique, but their search for what I will call a 'usable' denominator to calculate replacement rate seems to be a tighter approach than CBO's.