SNAP Rolls: They’re Elevated for a Reason

May 21st, 2013 at 9:09 pm

So I’m driving into work the other day, and since 8-10 hours of this stuff isn’t enough for me, I’m listening to wonk radio where this guy is going on about the SNAP, or Food Stamps, program.  He’s a knowledgeable guy making a lot of sense, until he goes off and says something to the effect of: unemployment’s coming down, so the SNAP rolls should be coming down too.

It’s not an unreasonable thought, and it probably resonated with lots of listeners (or at least with the three other people in the world listening to this sort of thing at 8am in the morning).  The notion that the SNAP rolls are “too high” has also become a bit of a conservative meme.

But it’s wrong in at least two ways.  First, because the labor force participation rate has been dropping, in part due to people dropping of out the labor force due to lack of opportunity, the unemployment rate is a less reliable measure of labor slack right now (it’s artificially low because of the dropouts).  A better indicator of the weakness of the recovery and the continued need for nutritional support for low-income households is the employment rate—the share of the population employed.  And that’s been flat-lining for a while, meaning that job growth has just kept up with population growth.  Under those conditions, you’d expect elevated SNAP rolls.

The figure makes the case.  It shows SNAP recipients as a share of the population compared to the unemployment rate and the employment rate (it’s on the right axis).  As you can see, unemployment drifts down but the employment rate stays flat.  I’d argue that right now, it’s the latter—employment rates—that captures the weakness in labor demand more so than unemployment.*



Second, while SNAP rolls increased sharply in the recession, as you’d expect—along with UI, SNAP was and is highly elastic to increased need—they’ve decelerated of late as the economy has slowly improved, though here again, it’s not improved as fast for those more exposed to food insecurity.  I found a careful academic paper by Hoynes and Bitler (new so no link yet) that tests whether the response of the SNAP program to the increase in unemployment was uniquely large in the great recession, a finding that would support the conservative meme of SNAP over-correcting for the downturn (providing more assistance than historical relations would predict).  It finds this not to be the case—the increase in SNAP rolls over the downturn, at least through 2011 (their last data point) was not significantly out-of-line with the historical record.**

So here’s the thing.  Too many people have trained their eyes to look at a trend like the one of the SNAP rolls in the figure, note the sustained increase, and sharpen their fingers with a tut-tut-tut.   Other less knee-jerkers see the trend and ask whether it’s justified given underlying weakness in the part of the economy faced by the eligible population.  Do that, and this SNAP trend is as it should be.

*Extra wonky section for extra-wonky credit: I regressed monthly changes in the SNAP rolls/per capita on a constant, trends, a few dummies for monthly spikes due to disasters, and changes in both labor market variables: employment and unemployment rates, letting them fight it out to see which had more explanatory power.  The winner, by a knockout, was the employment rate…see output table below (note the change in the unemp rate is insignificant and one-third the size of the emp rate change).


**Bitler/Hoynes: The More Things Change, the More They Stay the Same: The Safety Net, Living Arrangements, and Poverty in the Great Recession, Table 4.

Print Friendly, PDF & Email

11 comments in reply to "SNAP Rolls: They’re Elevated for a Reason"

  1. Kevin Rica says:


    For us old curmudgeons who don’t read the Huffington Post or the Style section of the Post (excluding the comics, of course), what the heck is a meme?

    • Dave says:

      Noun: Meme
      An element of a culture or behavior that may be passed from one individual to another by nongenetic means, esp. imitation.
      An image, video, etc. that is passed electronically from one Internet user to another.

      Bookmark Google and use for quick definitions. Type the word you need with definition: “meme definition”.

      • Kevin Rica says:

        So “meme” by that definition is just a synonym for fad. It’s the conformity of the internet age. It’s what is “trending” on the Huffington Post. I suspected as much. Only it come from the online Style section of the Post as opposed to the print edition. “Side boob” is a meme. “Meme” is a meme.

        • S. D. Jeffries says:

          Not really. A meme is more an item of propaganda which may or may not be true. It’s an item that’s accepted as real or true by a certain segment of the population (usually by someone or some group who stand to profit by its acceptance) who spread it through the remainder of the population until it’s accepted by a majority proportion of the population as true.

          A meme in political terms is usually planted by one side of the political dialogue. The “fact” that the deficit is the biggest economic problem this country faces began as a meme – an idea item that was promoted by one side of a political argument and propagandized extensively until larger and larger portions of the population accepted it as real and true. It spreads more quickly if the meme is based on a premise that is easily accepted because it “sounds” right or is in some respect already attractive to its intended audience. There are also social memes and business memes.

          Whether a meme is actually real and true or not, once it is accepted by the majority, is extremely difficult to dislodge. Usually proof in the form of data or logic won’t do it; dislodging or replacing it takes a substantial event that makes it no longer acceptable.

          • Kevin Rica says:


            Well explained!

            So a meme has the quality of what Colbert calls “truthiness.”


            “Truthiness” though, is a useful word. It expresses something that you can’t describe as well without it.

            “Meme” is a useless, ill-defined, neologism that explains nothing well on its own.

            It’s not the most vacuous thing that we’ve read on the Huffington Post (that might be “side boob”), nevertheless (quand même?) I think we should abandon it unless it has a well-defined and useful meaning.

            In the spirit of Orwell’s “Politics and the English Language,” let’s defend the language.

    • Convert says:

      Meme is the mental/verbal equivalent of a gene. Its a culturally transmitted unit of information. See

      • Kevin Rica says:

        So by that definition, it’s a “rumor” passed along through the internet.

        • S. D. Jeffries says:

          It doesn’t depend on the internet. There have always been memes. Before the internet they were spread by word of mouth or by the press, or by rumor.

          • Corey Yates says:

            In the 19th century, when every newspaper and cartoonist portrayed Irish people as monkeys can be thought of as an old school form of a “meme”

  2. Stephen W says:

    Clarification for any intro stats students reading this: the regression output in this post represents a fairly atypical situation in which you can compare the importance of two variables by directly comparing their coefficients. Changes in employment and changes in unemployment are already in the same units, and it probaby makes sense to compare a 1-unit change in the employment rate to a 1-unit change in the unemployment rate. If you’re new to regressions, remember that you usually cannot make the comparison that Dr. Bernstein makes here without first doing something to convert them into comparable units, like multiplying each coefficient by the standard deviation of its x variable.

    • Jared Bernstein says:

      That’s why I used changes instead of levels. Log changes provide similar results, with emp beating unemp…question for commenter and stat students: which is better here, log changes (which approximate percent changes) or percentage point changes?