Skip to content

Is it really true that candidates who are perceived as ideologically extreme do even worse if “they actually pose as more radical than they really are”?

Most of Kruggy’s column today is about macroeconomics, a topic I’m pretty much ignorant of.

But I noticed one political science claim:

It’s easy to make the political case that Democrats should nominate a centrist, rather than someone from the party’s left wing. Candidates who are perceived as ideologically extreme usually pay an electoral penalty; this is especially true if, like Bernie Sanders, they actually pose as more radical than they really are.

The research I’ve seen shows a small average electoral benefit for moderation (see here, for example), so I’m with the Krugmeister on that one. But where did he come up with the claim that extremists pay more of an electoral penalty if they actually pose as more radical than they really are? It’s hard for me to imagine there’s enough data to estimate that; also I’m not quite clear what the claim is, also I’m not sure how you’d go back and measure candidates’ posed and real ideologies. So all that makes me skeptical.

As always, I’d be happy to be corrected if there’s something I’m missing here.


  1. Ben Hanowell says:

    it would be so nice if prominent journalists would cite their god damn sources.

  2. Chris Wilson says:

    The funny thing about ‘extremists galvanize the opposing party’s base’ is that in 2016 the rather “centrist”/”moderate” HRC galvanized Republican opposition like crazy. My guess is that negative partisanship is becoming more and more common, and the difference between moderates and “extremists” (all relative, Sanders e.g. is actually just more or less a New Deal social democrat, but I digress) in outcomes more of a wash.

    • jim says:

      “in 2016 the rather “centrist”/”moderate” HRC galvanized Republican opposition”

      HRC was centrist in her run against Obama, but in 2016 she was much further left, following Obama’s turn to the left in his second term.

  3. Joe says:

    This seems like a straightforward implication of the finding that extremists pay a penalty?

    Assuming that extremists pay a penalty as they move away from the median (which you appear to agree with), then it follows that if you’re actually x units away from the median but you tell everyone you’re x+1 units away from the median, then you’ll pay a bigger penalty. I don’t understand how the opposite could possibly be true. You don’t need data to estimate this once you have the first regularity unless you want to twist yourself into real contortions. If having characteristic z is harmful to a candidate, then I think we’re safe to presume that pretending to have z is harmful barring some really convoluted common knowledge game.

    There’s another question here: how does Krugman know that Bernie Sanders is pretending to be more radical than he is? I don’t know the answer to that.

    • It’s a commonly expressed belief among at least some of my Democrat friends that Bernie calls himself a socialist but is really just a mainstream left-leaning Democrat. Like, Bernie doesn’t advocate for the elimination of private property, or really even for some things that Krugman advocates like the government printing limitless amounts of money to cover limitless spending, etc.

    • Greg says:

      I understood it differently : given two candidates “posing” at x units from the median, if one of them is “really” at only x-1, he will pay more. There would be a price associated with the difference between the “pose” and the “reality”.

    • JimV says:

      That (what Joe says, I’ll label it view J) is exactly what I thought. I suppose you could make other interpretations fit what Dr. K wrote, but J certainly fits and makes sense. Why not read with charity and not make a federal case out of it?

      It is very difficult to write something without leaving any room for misinterpretation. I try sometimes, and it turns a five-minute comment into a two-hour one.

      • Andrew says:


        You might be right. I have no idea what Krugman meant with that phrase. I was reading with as much charity as I could—as a blogger who writes something like 400 posts a year, I’m well aware how hard it is to write clearly—but I really couldn’t figure out what he was saying. Also, whatever it was that Krugman was saying, I didn’t see the evidence for his claims. It’s fine—not every newspaper column has to be perfect—but when there’s political science in the newspaper, I notice it. To notice a claim and wonder where is the evidence for it, that’s not “making a federal case.”

  4. Terry says:

    It’s easy to make the political case that Democrats should nominate a centrist, rather than someone from the party’s left wing. Candidates who are perceived as ideologically extreme usually pay an electoral penalty; this is especially true if, like Bernie Sanders, they actually pose as more radical than they really are.

    Isn’t this just about tautological and therefore empty?

    “extreme” = not very popular, i.e., not many people agree
    “more extreme” = even less popular than extreme
    “pose as more extreme” = appear less popular than there positions “really are”… if a pose is taken (at least somewhat) seriously by voters, shouldn’t that matter?

  5. I’d guess fairly confidently that this is just bad phrasing that editors who didn’t think very hard about what Krugman was saying missed. Krugman just means to say that extremists pay a price, and that Bernie will probably also pay that price — even though Krugman himself doesn’t think Bernie is all *that* extreme — because he believes Bernie poses as more extreme than he is.

  6. One of the ambiguous phrases here is “especially true.” It doesn’t necessarily mean that those who are posing as more radical than they are pay a measurably greater penalty than those who are just as radical as they claim to be. It could mean that, but it could also mean that the one posing as more radical gets penalized more than others in proportion to his or her *actual* degree of radicalism. (This came out a bit convoluted, but the point is: “especially true” can be taken in more than one way.)

  7. Zhou Fang says:

    I think you can probably get to that conclusion from the paper he links in that statement. The theory being claimed here is that ideologically extreme candidates do badly because they serve to galvanise their opposition harder than they make their base turn out.

    In the case of a candidate that is perceived overall as extreme but is merely ‘posing’, this can be worse because opposition supporters who know the candidate poorly will still perceive the candidate as very extreme, and hence turn out to vote. Meanwhile people closer to the candidates will be more likely to perceive the candidate as boringly moderate, and hence not turn out. In other words, the “pose as more radical than they really are” is about a candidate who appears more extreme the less you know about them.

    Mind you, I doubt there’s enough data to show this.

  8. David Marcus says:

    Krugman has never liked Bernie (e.g., Krugman was a strong Hillary supporter, even though it was clear (to me, at least) that Hillary had problems as a candidate.

    • jim says:

      Personally I like HRC, I think she’s a sharp chick. Last time out way to far left. But that’s the problem right? As a candidate her biggest problem – how come no one ever says this? It was talked about during both of her presidential runs – was her well-earned reputation for changing her position according to the political wind. She is intelligent, but she didn’t articulate a vision in either run. She repeatedly tried to lane-dodge her positions into the polling fast lane.

      • David Marcus says:

        Hillary is a corporate Democrat, as opposed to a Progressive Democrat. Also, she was part of the Obama administration, so she was basically offering more of the same. Hard to get excited about that. Trump offered something different. Of course, he lied, but many voters didn’t realize that.

        • Zhou Fang says:

          Except that more people voted for Hillary than did for Trump, and the people who did vote for Hillary were more likely to say they voted *for* Hillary, while the people who voted for Trump were more likely to say they were only supporting Trump to oppose Hillary.

  9. paul alper says:

    People are very touchy these days and antisemitism is the third rail of American politics. Andrew might be immune to referring to a Jewish Nobel Prize winner as “Kruggie” or “the Krugmeister,” but would others be so fortunate? And only in the last couple of weeks did I find out what is now the new meaning of the word “thug.”

  10. gdanning says:

    This is a question I have had for a long time: The confidence intervals in the graphs on page 8 of the linked paper are huge. Given that, how sure are we that the true relationship is not the reverse?

    Eg on the left half of graph, the point (-0.2, 0.54) is within the confidence interval, and the point (-0.01, 0.47) also is within the confidence interval, and a line connecting the two has a negative slope, not a positive slope. Does the data suggest 95% confidence that that the true relationship is positive? Less than that? There is no way to tell?

    PS: I am reposting the here, for convenience:

    • Zhou Fang says:

      This is a correct read and why the text of the paper does not discuss the relationship between the *margin of victory* and the turnout (which as the confidence interval shows, cannot be reliably estimated), but rather the difference across the 0 discontinuity (i.e. between a narrow moderate win and a narrow extremist win). The former is really there in the model to try and eliminate a potential confounding variable – removal would open accusations of p-hacking. For the latter effect, there seems to be some fairly decent evidence, and the effect size is large enough to be worth some attention anyway.

      • gdanning says:

        Thank you. That certainly makes sense, yet I rarely see it explicitly acknowledged that the relationship cannot be reliably estimated (and I have had a stats prof or two who did not seem to understand my question when I raised this issue).

  11. Terry says:

    The Hall and Thompson paper cited by Krugman does not inspire confidence.

    A number of z stats hovering just above 2, and even the strongest z stats are only about 3.

    A discontinuity analysis.

    Big standard errors in the discontinuity analysis (mentioned by gdanning).

    And some third and fifth degree polynomials.

    As usual, the discontinuity analysis looks great when you focus on the blue regression lines, but gdanning and Lakeland have shown this certainty to be … flimsy.

Leave a Reply