In a recent op-ed for the New York Times, political scientists Christopher Achen and Larry Bartels claim that the unexpected success of Bernie Sanders in the US Democratic primaries is not driven by policy positions to the left of Hillary Clinton, but rather identity and symbolic factors. However, their argument is only partially consistent with the data they cite, and inconsistent with data they do not consider. Presenting a deeper look into the 2016 ANES Pilot Study, I show that, relative to Clinton supporters, Sanders supporters are notably more supportive of abolishing the death penalty, more favorable toward campaign finance restrictions, more opposed to free-trade agreements, and less convinced that black people need to “try harder.” I also find that Clinton and Sanders supporters are characterized by deep ideological differences over the policies most important to them. While Achen and Bartels are probably correct to argue the Sanders movement is not necessarily a massive shift of policy preferences in the American electorate, neither is it a merely identitarian or symbolic blip; it is a rebellion, by an American left marginalized for decades, against an increasingly hopeless Democratic Party.
Achen and Bartels argue that “it is very hard to point to differences between Mrs. Clinton and Mr. Sanders’s proposed policies that could plausibly account for such substantial cleavages,” referring to the cleavages between supporters of Sanders and those of Clinton. Although Achen and Bartels are world-class political scientists, both of whom I greatly admire, we are living through a period when long-standing truisms in political science may at times mislead more than clarify. While the text of their article is careful and responsible, its overall effect suggests the divergence between Sanders and Clinton supporters is not about fundamentally different political visions. They suggest that Sanders supporters are not even necessarily to the left of Clinton supporters in their policy preferences. Surprised and intrigued, I went further into some of the data they cite and I found a very different picture.
Now, if we look at survey data on policy preferences, it is true that we should not expect Sanders and Clinton supporters to look starkly opposed. Overall, we would expect them to look like one big bunch of lefties. One can only support a minimum wage hike so much, for instance, and both camps of supporters could be predicted to largely support it. In this sense, Achen and Bartels are correct to suggest that Sanders and Clinton supporters are not going to be distinguished by drastically different policy preferences. But if that was their only argument, their piece would not have been published with a title as provocative as “Do Sanders Supporters Favor His Policies?” Achen and Bartels actually argue that, compared to Clinton supporters, Sanders supporters are less likely to support left-wing policies, naming a few in particular:
However, they were less likely than Mrs. Clinton’s supporters to favor concrete policies that Mr. Sanders has offered as remedies for these ills, including a higher minimum wage, increasing government spending on health care and an expansion of government services financed by higher taxes.
As far as I could determine, the data show this claim to be correct only in a limited sense, and contradicted by a number of additional measures I brought into consideration. Figure 1 plots the distribution of policy preferences for Sanders and Clinton supporters over 11 policy questions gauged in the 2016 ANES Pilot Study. When appropriate, I reordered the scales of each measure to read more intuitively, as reflected in the header for each facet of the plot. The variables are measured on an ordinal, categorical scale (e.g., “favor a great deal,” “favor moderately,” and so on). It is not technically ideal to treat these as numerical responses (for instance, support for a policy on a scale from 1 to 5), but this is a convenient and intuitively understandable way to compare the distributions for two different groups. In the analysis that follows, when I refer to Sanders and Clinton supporters as being statistically different or indistinguishable in policy preferences, I am referring to whether one group has a different mean preference in this sense; by convention, if there is more than about a 5% chance we would observe a particular difference by chance alone even if there is zero difference in the population at large, then we do not consider that difference to be very meaningful.
Unless I selected different variables than Achen and Bartels, their interpretation of the data is not necessarily incorrect but it is certainly underwhelming. The first three facets represent the policy preferences mentioned by Achen and Bartels as evidence that Sanders supporters do not support his policies. First, Clinton and Sanders supporters are overwhelmingly likely to support raising the minimum wage (the right-most bar in that facet); Clinton supporters are only very slightly more likely to be in that category. Regarding government spending on healthcare, Sanders supporters are marginally more likely to be in the most supportive category relative to Clinton supporters, but they also have marginally more folks in the least supportive category. For both of these variables, however, the more important point is that these distributions are not very different in either direction. Assigning them numerical scales, on the question of minimum wage the mean for Sanders and Clinton supporters is both exactly 3.7 on a scale from 1 to 4; on health spending the means are 5.4 and 5.6, respectively, but there is about a 30% chance we would observe such a difference by chance alone (i.e., it is not statistically significant by conventional standards). I have zero doubt that Achen and Bartels arrived at their claims legitimately by some slightly different approach, the point is to see how fragile and minor appears this difference. The data perhaps most favorable to the argument of Achen and Bartels is for preferences on government spending for more services. The difference in means is statistically significant in the direction suggested by Achen and Bartels, but again the difference between these distributions are not especially stark to my view. To be clear, Achen and Bartels are more experienced political scientists than I am, and this style of political analysis is always debatable, but as far as I could find, I personally would not have much confidence inferring from these three measures that Clinton supporters are more likely to support those policies to any statistically or substantively significant degree.
If we look at a larger sample of policy questions, however, Sanders supporters are clearly farther to the left than Clinton supporters. Achen and Bartels only discuss three of the policy questions in that pilot study. To be fair, it was only an online op-ed so I would not fault them, but I broadened the analysis to a healthier number of 11 diverse issues. This was not an exhaustive analysis of all policy questions in the data, but I did my best to avoid a fishing expedition for counter-evidence. Sanders and Clinton supporters are statistically indistinguishable on six of the eleven (global warming, migration, gender pay gap, minimum wage, health spending, and vaccines). Clinton supporters are statistically more to the left only on greater spending for services, and to a very slight degree, as already discussed. Sanders supporters are statistically distinct from Clinton supporters, and in line with Sanders’ differing policy positions, on four of the eleven policies, and by more notable margins than with respect to spending on services. Sanders supporters are more opposed to the death penalty, more in favor of campaign finance restrictions, less likely to say black people need to “try harder” (an implicit policy position), and they are more opposed to free-trade agreements. In short, Clinton and Sanders supporters all mostly support left-leaning policies but, on average, Sanders supporters are distinguishably farther to the left of Clinton supporters on four of the five policy questions for which there is any statistically or visually notable difference between the two camps.
But we can go even further in assessing whether Sanders supporters favor different policies than Clinton supporters and, importantly, the degree to which Sanders supporters are more likely to favor the policies proposed by Sanders.
We can hypothesize that Sanders and Clinton supporters are probably most different in their policy priorities, that is, which policies they think it is most important for their elected leaders to pursue. Such differences are just as well core, ideological, policy differences but here we are conceptualizing and measuring them differently. This is where I think the conventional wisdom in political science–that “voting behavior is primarily a product of inherited partisan loyalties, social identities and symbolic attachments,” most seriously understates the degree to which the Sanders phenomenon is indeed a meaningful signal about a latent but hitherto repressed ideological component in the American electorate.
To consider this, I look at the series of questions from the 2016 ANES Pilot Study asking respondents to list the four top policy issues most important in their choice of candidate. I calculate the mean ranking assigned to each issue, across all the respondents within each camp. For instance, if all Clinton supporters said “Economic Growth” is their top issue, it would earn a value of 4 in the Clinton camp. If the gist of Achen and Bartels is correct, Sanders and Clinton supporters should not have especially different issue priorities; or, if they do have different priorities, they should not be predictable by differences in emphasis from the two different candidate’s policy programs but rather they should be non-policy rationalizations of their different identitarian and symbolic attachments.
As revealed in Figure 2, Sanders and Clinton supporters have markedly different issue priorities. To be honest, I was surprised how different they are. One of the obvious, large differences is regarding income inequality, the top priority for the Sanders bloc but only the thirteenth for the Clinton bloc. Nobody will need a citation showing that Sanders has made income inequality one of the top goals of his policy agenda, more so than Clinton. Morality issues and women’s rights are other very differently ranked items in the two camps’ hierarchies of policy priorities. But one interesting and perhaps somewhat surprising, massive difference in issue priorities is regarding the environment and climate change. If Achen and Bartels are correct, then we would expect little difference in the policy positions of Sanders and Clinton on this issue; according to their theory, this difference is likely just Sanders voters rationalizing the identitarian and symbolic factors making them lean toward Sanders. But as Mother Jones reports, Sanders’ policy program on climate change is more aggressive than Clinton’s on multiple, specific, sub-domains of this general policy area. Sanders supporters desire political action on income inequality and climate change much more intensely than Clinton supporters, and they appear to be choosing the candidate who seems most likely to deliver the most policy change on those issues. Do Sanders supporters favor his policies? As far as I can tell, they do.
I am also skeptical that the Sanders phenomenon represents a shift to the left in the Democratic Party. It is probably not a shift but rather the re-emerging tip of a longstanding historical iceberg. Critics might say the Sanders movement is a shot in the dark, but observers should at least acknowledge what is most likely behind it: an American left previously forced into hibernation by a Democratic Party that for decades now has gleefully abandoned most poor, working, and minority folks in favor of global capitalism, white supremacy, and neoliberal feminism. Hillary Clinton is a model of precisely this headlong capitulation to everything left parties were once upon a time supposed to oppose, and if Bernie Sanders has attracted an unexpectedly large movement behind him then we should infer it is simply because he is one of the only remaining holdouts from that by-gone era. The Sanders movement probably also contains identitarian and symbolic aspects, of course, as Achen and Bartels argue. But even if Sanders supporters are enjoying some symbolic opposition to the Democratic Party, the data do not discourage us from believing they are nonetheless doing so because of real and deep dissatisfaction with the policy agenda of that party and a corresponding preference for the anti-systemic, egalitarian political vision articulated by Bernie Sanders.
Share this post:
Murphy, Justin. 2016. "Differences Between Supporters of Clinton and Sanders," http://jmrphy.net/blog/2016/06/03/differences-between-supporters-of-clinton-and-sanders/ (April 24, 2017).