STUMP » Articles » Geeking Out: Are Women Really Receiving Lower Social Security Benefits? » 8 August 2018, 16:41

Last week, I graphed out retired worker Social Security benefits… and it culminated in this graph:

Look at that — women are paid less!

Well, duh. When it’s based on women’s own work history, and women tend to have lower earned income than men… of course they’ll have lower benefits.

BUT that’s not the only way women can receive Social Security benefits — they can receive based on their spouse’s earnings (and yes, men can do this, too… but there are all sorts of wrinkles. Such as: Stu can’t file for Social Security benefits based on my earnings when he turns 62… because I’ll be only 49 then.)

So let’s see what the possible benefit combinations are.

Spreadsheet is here.

SPOUSAL, DISABILITY, SURVIVORS BENEFITS

Now, one can get Social Security benefits from becoming disabled, though that may be a temporary and not a “permanent” award.

I am going to ignore children for all of this. (But yes, children can get Social Security benefits – I did for a time, when my dad died before I was 18 – and if you’re a minor child of a retired worker.. yup, you can get benefits.) Evidently even parents can get survivors benefits, but I will ignore that because: a. the number is very small (only a few thousand) and b. I don’t even know how that works.

I barely know how the spousal benefit works for retired workers, other than I know Stu can’t get anything from my benefit unless I die or he lasts until I’m 62, and I file for benefits.

So here is the classification of the Social Security beneficiaries in June 2018:

Now, that’s very interesting to me.

About the same number of women as men are receiving retired worker or disabled worker benefits. There is a small difference, but I don’t care enough to check.

Where you see a large difference is aged widows/widowers — that is, widow(er)s over age 60 (why they make that distinction, I don’t know.)

In short, a small number of women have such low earnings records that they are receiving benefits of a retired or dead spouse. This is true of very, very few men.

Of the 32 million women receiving Social Security benefits 67.4% are receiving them as retired workers on their own earnings record; 13.2% are receiving disability benefits based on their own records. Together, that’s 80.6% of women receiving Social Security benefits having nothing to do with their marital status.

For the 19.4% remaining, 11.1% are aged widows. Only 6.9% are receiving benefits based on their retired spouse’s earnings. So that leaves only 1.4% for young widows, widows of disabled workers, or spouse of disabled worker benefits (which are very low benefits).

I will contrast to the same for men. There are 26 million men receiving Social Security benefits — yes, 6 million fewer than women, and almost all that difference comes from men dying earlier.

82% of men receiving Social Security are retired workers on their own record. And 16.7% of men receiving Social Security benefits are receiving disabled worker benefits. That leaves 1.2% for the status of being married or having been married to somebody.

So fine – 19.4% of women receiving benefits are doing so based on having been married, and only 1.2% of men receiving benefits do.

Let’s compare amounts.

ADD UP THE BENEFITS

Here we go with combined benefits, and I’m putting it back to monthly because of something I will be doing in a moment:

There is a problem with this graph — maybe you can see it. Some weird discontinuity in the Women’s line at two points specifically: at the $1,200 – $1,299 mark, and at the $2,100 – 2,199 mark.

Well, that’s because I can’t combine the numbers that way – it causes trouble.

PROBLEM: CENSORED DATA

It’s called censored data, and this, along with truncated data, happen naturally all the time. This data is censored in multiple ways — censoring is when you don’t know the exact value of a data point, but you know it occurs in a range (truncating is when you don’t even know a data point is there, if it falls in a certain range — that can happen with deductibles).

My original retired workers data was censored in two ways. First, I just have number of beneficiaries within a bin (like $1,200 to $1,299) that’s interval censoring. No problem, the other data sets use bins of width 100. The problem is the second bit — the data are right-censored at $4,000. The bin is called “$4,000 or more”. Now, so few have benefits above that point, that censoring is not terribly concerning.

Thing is, for the other beneficiary types, they right-censor the data at much lower points. For “spouse of a retired worker”, it’s censored at $1,200. For aged widows, it’s censored at $2,100. For disabled workers, it’s censored at $2,900. For spouses of disabled workers, it’s censored at $500 (yikes).

FIX ONE: MAXIMUM CENSOR

There are several ways to deal with this, but the “easiest” is to go with is to use the most restrictive set of bins.

Here we go, censoring at $500:

This is not helpful.

FIX TWO: CENSOR AT LOWEST, MOST SIGNIFICANT DIFFERENCE

Okay, when I censored at $500, I squished my data into so small a space, I obliterated a lot of information.

But should I really have done that? The spouses of disabled workers — even for women, that’s only 0.4% of their total grouping.

We should go with the most significant, lowest censoring.

Given about 6.9% of the women are spouses of retired workers, let’s use their censor point: $1,200.

Okay, that’s better, but not by much.

FIX THREE: EXTRAPOLATE

So here is what I’m going to do. I’m going to do the benefit curves for men and women by category, compared against our most numerous category: retired workers.

Women’s curves (tee hee):

Lines for the men:

So there seems to be a general shape – it comes to a peak, and then kind of tails off.

The male shape is a bit odd (tell me about it!) but that may be related to strategies of higher income men when they decide to apply for Social Security benefits.

In any case, I will look at age effects a different time.

Anyway, it looks like exponential decay, doesn’t it? Let’s do a few exponential fits.

Women, retired workers: (I’ll use midpoint)

Eh, not fabulous, but it will do.

Disabled workers:

Better, better.

In both cases, for every step of $100, I hit the the number by exp(-.003*100) = 0.74 — essentially a 25% drop per step.

Let’s check that against the data.

Yeaaaah, doesn’t quite work. But it will do for smoothing out my final data — I make the sums total up to what it was before, and it will be just fine.

RESULT: YES, WOMEN GET LOWER BENEFITS

So yes, the distribution for women’s benefits is skewed lower than that of men. Which didn’t really need that whole song and dance, but I was having fun anyway.

None of this should be all that surprising – spousal benefit formulas do not give the spouse the same exact benefit:

at your full retirement age, your benefit as a spouse cannot exceed one-half of your spouse’s full retirement amount.

This is why so many more women are receiving benefits as retired workers — even if they made less money than their husbands, it wasn’t 50% less. Women do generally have lower earnings (partly because of the professions they work in, partly because of hours worked, etc.), and even then, women tend to have fewer full earnings years — the Social Security benefit is based on highest (indexed) 35 years of earnings. If one worked part-time while the kids were growing up, if one retired a bit early because one had an older spouse… it all leads to reducing the Social Security benefit.

But notice again — there are 23% more women than men receiving Social Security benefits — and that’s basically because women live longer. 32 million women vs. 26 million men.

But talking about ages and benefits… that’s for another time.

Social Security Benefit Spreadsheet