“The Narcotic of Government Dependency”

That’s Rick Santorum talking about the American welfare state. But who, really, is hooked—and how does that matter politically?

Yesterday’s New York Times featured a long, meaty article on the distribution of federal benefits. One of the more striking points, drawing on work by political scientist Dean Lacy, is that government benefits constitute a larger share of total income in “red” states than in “blue” states. (The Times article includes a lovely set of interactive maps detailing the geographical distribution of benefits from a variety of specific programs such as Medicare and unemployment compensation; here is the 2009 map of total benefits.)

A friend asks, “Is this true at the individual level? It isn’t, right?”

I imagine that the answer to that question depends a lot on what gets counted as government benefits (tax credits? grants and contracts?) and whether we are talking about gross benefits, benefits minus taxes, or benefits as a share of total income. As Monkey Cagers know (see herehere, and here), political scientist Suzanne Mettler’s book, The Submerged State, includes some excellent analysis of the tenuous relationship between objective and subjective dependency on the federal government.

In a review (forthcoming in Democracy) of political journalist Thomas Edsall’s new book, The Age of Austerity, I raise the question of how Republican policy-makers bent on budget-cutting will come to grips with the actual distribution of government spending:

In a perceptive recent essay in New York magazine, heterodox Republican David Frum sketched a political landscape much like the one portrayed by Edsall. “We have entered an era in which politics increasingly revolves around the ugly question of who will bear how much pain,” Frum wrote. “Conservative constituencies already see themselves as aggrieved victims of American government: They are the people who pay the taxes even as their ‘earned’ benefits are siphoned off to provide welfare for the undeserving.”

However, Frum went on to pinpoint the fundamental contradiction in this conservative worldview. “The reality,” he wrote, is that “the big winners in the American fiscal system are the rich, the old, the rural, and veterans—typically conservative constituencies.” Squeezing the programs conservatives hate won’t bring in much revenue, so balancing the budget would require chopping into programs most conservatives support—including defense, Medicare, Social Security, and middle-class tax breaks.

In Chain Reaction, Edsall recognized that “the anti-tax, anti-government view of the electorate … was directed at programs serving heavily minority and poor populations,” while spending on education, health, Social Security, crime control, and environmental protection “retained unstinting, and in some cases growing, majority support.” That remains true 20 years later; even most conservatives oppose cuts in most major government programs, and they do so even when they are reminded of the perils of deficit spending.

Unfortunately for Republicans—and for Edsall’s analysis of the politics of austerity—“programs serving heavily minority and poor populations” are not where the money is. According to the Census Bureau’s Consolidated Federal Funds Report, less than 8 percent of federal spending in 2010 was for unemployment benefits, food stamps, housing assistance, student aid, and the earned-income tax credit. Almost half was for salaries and wages, grants, and procurement; most of the rest consisted of Social Security and Medicare payments. Large-scale reductions in government spending would require significant cuts in big-ticket programs that mostly benefit the middle class. The political challenge facing budget-cutting Republicans is exacerbated by the fact that beneficiaries of government spending are disproportionately concentrated in red states. Federal expenditures made up almost 30 percent of total personal income in the 29 states that voted for John McCain, a significantly higher dependency level than in the states that voted for Barack Obama.

“The rank and file of the GOP,” Frum concluded, are “caught between their interests and their ideology.” This clash of interests and ideology is left largely unexplored in Edsall’s analysis. While he acknowledges that “substantial numbers of Republican voters have no appetite for cuts in the two programs that virtually every economist and budget analyst says must be chopped down to size: Medicare and Social Security,” he never really comes to grips with the question of how Republican politicians will finesse that fact. It is one thing to carp about the futility and injustice of government programs in the abstract, but something else to deprive voters of their concrete benefits.

20 Responses to “The Narcotic of Government Dependency”

  1. Andrew Gelman February 13, 2012 at 6:13 pm #

    I’m hooked on government spending. I’ve been supported by the National Science Foundation for decades!

  2. Thomas Brambor February 13, 2012 at 7:32 pm #

    “The rank and file of the GOP are caught between their interests and their ideology.” This seems the most interesting upshot of this analysis. Though from surveys of Tea Party supporters few are aware of the inherent contradiction of their calls for small government while at the same time drawing from social security and medicare. Same dissonance holds, I suppose, for the many lower income voters happily supporting tax cuts for the rich. How do we explain this with rational voter theories?

  3. RobC February 13, 2012 at 7:48 pm #

    When considering the amount spent for “programs serving heavily minority and poor populations,” are you including any portion of the Social Security and Medicare disability program, which according to Robert Samuelson represents a large and growing expense and which seems to be going to a lot of people who aren’t disabled in the traditional sense of that word–and most of whom, I’d hypothesize, are not persons who would otherwise be eligible for high-income jobs?

    • RobC February 13, 2012 at 7:49 pm #

      I neglected to include a citation for Samuelson. It’s here.

  4. Bill Bishop February 13, 2012 at 8:49 pm #

    People say (and by people, I mean political journalists and anybody who works at Brookings) all the time that rural areas get a disproportionate share of the federal budget, just as Edsall does, apparently. That happens not to be true.

    The Economic Research Service tracks these things, tracing federal budget expenditures to the county level. It turns out that rural counties receive less per capita than urban counties. Rural people are older, so there is more Social Security spending in rural counties. Otherwise, it’s not even close. Rural communities are not “big winners” in the federal budget free-for-all, as Edsall writes. Anything but. Here are the details: http://www.dailyyonder.com/speak-your-piece-times-brookings-and-rural/2011/06/02/3360

    • Andrew Gelman February 13, 2012 at 9:59 pm #


      I clicked through your link but I’m confused. The graph says, “Total federal spending per capita in rural counties lags spending in urban counties in most years,” with the most recent year being 2009, labeled “Urban: $10,471” and “Rural: $10,186.”

      The graph also provides a link. I followed the link, which lists per-capita Federal funds in 2010 as $10,334 in urbanized counties, $10,010 in less urbanized counties, and $11,437 in rural counties. That is, funding is 10% higher per capita in rural counties!

      I thought that maybe 2010 was special, so I clicked on some more links and got USDA’s data for fiscal 2009: Per-capita spending was listed as $10,061 in urbanized counties, $10,080 in less urbanized counties, and $11,731 in rural counties. That year, funding was over 15% higher per capita in rural counties.

      I don’t know what’s going on, but your graph doesn’t seem to match the numbers you’re citing.

      • Bill Bishop February 14, 2012 at 11:15 am #

        See reply below. You were looking at subsets of the broader metro (urban) and nonmetro (rural) categories. Metro vs. non metro is the standard way of measuring urban and rural. Just as Larry says below.

  5. Larry Bartels February 14, 2012 at 2:48 am #

    Andrew: “Urbanized,” “Less urbanized,” and “Totally rural” are subcategories of “Nonmetro,” which the government (confusingly) also calls “Rural”; the per capita spending ordering in 2010: “Totally rural” > “Metro” > other non-metro counties.

    Bill: I think the question of “disproportional share” depends a lot on what and how you count. As you note, retirement benefits are higher in non-metro areas, but that is part of what David Leonhardt was noting as “unrepresentative” in the piece you criticized in your post. Metro areas get vastly more in defense contracts; are those government benefits? Also, since non-metro areas are generally poorer, similar per capita figures imply higher levels of “dependency” measured as a proportion of total income (as in the map I reproduced, where the dark patches in Arizona and New Mexico, the Ozarks, Appalachia, and northern Maine are obviously not urban areas). Incidentally, the “rural” quote in my piece is from David Frum, not Tom Edsall. (I was somewhat confusingly quoting myself quoting both of them.)

    RobC: My 8% figure included only the programs I listed–unemployment benefits, food stamps, housing assistance, student aid, and the earned-income tax credit. Social Security disability payments accounted for an additional 3.7% of government spending (13.3% of Social Security spending) in 2010. I’m pretty sure that most of those payments went to poor people, but I don’t know how many of them were poor before they became disabled. For that matter, I don’t know how most Republicans feel about that 3.7% of the federal budget (though I seem to recall that the Reagan administration’s attempt to trim the disability rolls touched off a political firestorm and, according to the SSA, a ”huge volume of adverse court decisions”).

    Thomas: Why “rational voter theories,” particularly? (Not that there is anything “irrational,” in a technical sense, about motivated reasoning, or about preferences ungrounded in material self-interest, as long as pristine logical consistency is preserved.)

    • Andrew Gelman February 14, 2012 at 9:14 am #


      Yup. What I’m wondering is, where did Bill Bishop’s graph come from?

  6. Bill Bishop February 14, 2012 at 11:00 am #

    The data comes from the link I placed in my first comment, guys. Right here: http://www.ers.usda.gov/briefing/ruraldevelopment/developments.htm

    Look at Table 1. The standard way of measuring rural vs. urban is to look at metro and nonmetro counties. If you look up there near the top of the very first graph, you’ll see that metro receives more money than nonmetro on a per capita basis. So, that’s where the data on the chart comes from — the first few lines on the very first chart. (In the notes, the ERS writes: “Rural” and “nonmetro” are used interchangeably to refer to people and places outside of MSAs.)

    Further down, you can find other measures of county type. Yes, farming counties might receive a bit more in federal spending because of ag payments. But, overall, rural counties receive less.

    Do urban counties receive more in defense spending? I don’t know. I don’t believe this data includes contracts. Think of the military bases and installations in rural areas. You can pull this data out of the ERS files if you have a mind.

    And, yes, it does depend on how you count. Should we include Social Security spending? After all, would recipients be any younger if they moved to urban areas?

    Leonhardt and others argue that it costs more to the government to have people living in rural (nonmetro) areas. The figures show the exact opposite.

    • Bill Bishop February 14, 2012 at 11:16 am #

      Okay, I’m leaving a reply to my own comment. Crazy. Anyway, Larry is absolutely right. Defense spending is higher in metro counties. It’s in the charts here: http://www.ers.usda.gov/briefing/ruraldevelopment/developments.htm

    • Andrew Gelman February 14, 2012 at 11:26 am #


      Aaah, I see, thanks for the clarification. The urbanized vs. totally rural data that I was finding (from table 2 of your link) are all for nonmetro counties. And these tables do not seem to separate urban from suburban. From these data, we have per-capita spending in 2010 at $10,976 in metro areas (i.e., urban + suburban), as compared to $11,437 in totally rural counties. Nonmetro counties that are urbanized and less urbanized have per-captia spending of $10,334 and $10,010, respectively.

      In summary, we don’t know the answer for urban areas (without suburbs thrown in), and the answer for rural depends a lot on whether we include urbanized and less urbanized areas. Also, we’re only looking at one data source (from USDA), and it’s only at the county level.

      So don’t think you should be so clear that David Leonhardt is wrong when he wrote, “Suburbs and rural areas receive vastly more per-person federal largess than cities.” Not that he’s proved his point with data either; I just don’t think it’s as clear as you say, even given the numbers you link to.

  7. Bill Bishop February 14, 2012 at 11:57 am #

    Well, there is no definition of a “suburban” county. And what the heck is a “nonmetro” county that is “urbanized”? Those counties are already metro. For instance, the Austin metro area contains several counties that “look” rural, but are urban because of commuting patterns.

    You are mixing classifications — metro /non metro and USDA’s continuum codes. Best to pick one. The most common is metro/nonmetro. Urban vs. rural.

    Leonhardt said “rural” areas “receive vastly more per-person federal largess than cities.” We have his assertion (which, he told me, was based on what was spent in “rural states,” whatever they might be). And we have ERS’s data that goes to the hard work of dividing data into rural and urban counties. Seems pretty clear to me, especially given the numbers……

    • Andrew Gelman February 14, 2012 at 12:16 pm #


      You are the one mixing classifications! If you want to entirely use the words “metro/non-metro,” that would be fine. My confusion is that you also use the words “urban” and “rural,” even in the labeling of your graph. Regarding ERS’s data, it seems pretty clear that they have per-capita as highest in totally rural counties. Leonhardt may well have been sloppy with his categorizations, but as we can see from this discussion, it’s not easy. For example, you write that “nonmetro counties that are urbanized” are “already metro.” But ERS doesn’t think so. The definitions just aren’t so clear.

      I think you’d be on safer ground simply stating that Leonhardt is not backing up his claims with numbers. You can put the burden back on him to do so, and then see what he comes up with.

      P.S. You ask, “what the heck is a ‘nonmetro’ county that is ‘urbanized’?” The answer is in your USDA link: “This table uses the 2003 definition of metropolitan areas to distinguish between metro and nonmetro counties. Urbanized = at least 20,000 urban population in 2003; less urbanized = 2,500 to 19,999 urban population; totally rural = less than 2,500 urban.”

      • Andrew Gelman February 14, 2012 at 12:17 pm #

        P.S. to Bill: Thanks for commenting on the blog. It’s great when a blog such as the Monkey Cage can be a venue for people to clarify points and move a discussion forward. Traditionally we would do such things via email exchanges but I think it’s better to have it in the open where others can see and participate.

      • Bill Bishop February 14, 2012 at 12:32 pm #

        No, you are still mixing classifications — and yes, these classifications are a bane to everyone who has to deal with the rural/urban area. The chart shows metro and nonmetro counties. This is the most common way to determine urban and rural. The other categories are subsets, and that’s what you’re putting into the mix. So, yes, if you take a subset of “rural” that has very high crop payments and an older population, it will show higher federal payments than a subset of urban.

        But we aren’t comparing subsets. We are comparing rural and urban — nonmetro and metro.

        The definitions are an absolute hairball. But, believe me, Leonhardt didn’t get into one tenth of this detail. We had quite a long back and forth on email. He was looking at STATES — ones that looked rural (you know, Montana and Wyoming, but never Vermont). We asked him for his data. He had zip. He was looking at spending at ONLY the state level.

  8. Joel February 14, 2012 at 1:49 pm #

    who benefits most from government spending on roads? from the existence of the us treasury? etc.

    even if you can decide the argument focusing on direct (or even indirect) payouts, it’s important to remember that is not the entire argument.

  9. Andrew F. E. February 14, 2012 at 11:52 pm #

    I think the assumption that the GOP will be increasingly squeezed between their ideology and their interests is a bit too exaggerated. Research has shown that representatives have increasingly been unresponsive to the interests of their average constituents without paying any political price.

    US politics is about organization and activism and the majority of those red-state government dependents will most likely remain as misinformed about what is exactly going on in Washington as they are now and eventually mishit the target.

    I believe if you hold the average red-state Republican and force him to choose between ending some subsidy for the oil industry or cutting social security, they would opt for the latter for the simple reason that organized interest has a far more influential organizational base to harm the candidate than do scattered disorganized average voters.

    • Larry Bartels February 15, 2012 at 2:11 am #

      Frum’s claim is that rank and file Republican voters, not elected officials, are “caught between their interests and their ideology.” My own research makes me sympathetic to your point that misinformation, partisan rhetoric, and even false consciousness may blunt the apparent contradiction–but not entirely, and not indefinitely.

      Similarly, my work provides empirical support for the notion that elected officials’ behavior depends more on their own ideological convictions than on the views of “average voters.” But the views of voters (especially affluent voters) do seem to matter, especially for Republican officials.

      I’m not sure what research you have in mind showing “that representatives have increasingly been unresponsive to the interests of their average constituents without paying any political price.” I interpret work by John Sides and others on the 2010 election as implying that marginal Democratic MCs paid a very real political price for supporting health care reform. And I believe that Martin Gilens’s forthcoming book on the responsiveness of the policy-making process as a whole suggests that, if anything, responsiveness to public opinion has increased over time.

      It will be interesting to see whether Republican MCs run for reelection this year on the Ryan budget plan. But even that will not be a real test of what would happen if they actually implemented it. As Kasper Gutman says in _The Maltese Falcon_, “this is genuine coin of the realm. With a dollar of this, you can buy ten dollars of talk.”

  10. Philip February 16, 2012 at 4:22 pm #

    It is amazing that you all scrupulously avoid mentioning Medicaid, thus entirely ignoring the single biggest and most fiscally problematic of our low-income federal programs.