Are liberals more likely to drink lattes than conservatives? Despite the billions of dollars spent on political polling in the United States, this association has yet to be examined. In an article entitled, “Why Do Liberals Drink Lattes?,” sociologists DellaPosta, Shi, and Macy (Reference DellaPosta, Shi and Macy2015) were unable to address whether this empirical assertion is true, thus rendering the question of why somewhat premature.

Nonetheless, this trope is now a political and journalistic staple. On social media, the phrase “latte liberal” is used, on average, once per minute (socialmention.com 2017). The phrase has enjoyed ongoing popularity in the news, beginning roughly when it was popularized by journalist David Brooks in The Weekly Standard. Borrowing Alan Ehrenhalt’s term, “latte towns,” Brooks (Reference Brooks1997) defined latte liberals as upscale, politically left-of-center people who have abandoned political activism for creature comforts:

If anything, the phrase has gained steam (or perhaps foam) since that time.^{Footnote 1} Beyond journalistic usage, the line between market and political research is now blurred, thereby making it increasingly of interest whether political views are predicted by consumer purchasing behavior (Fischer and Mattson Reference Fischer and Mattson2009; Hersch Reference Hersch2015). President Obama once unabashedly raised his cup in what became known as the “latte salute.” This was viewed as an alienating symbolic gesture because, as one journalist put it, Obama “broke the cardinal rule of not playing to his negative stereotypes” (Rogers Reference Rogers2014).

In this study, we begin by examining the hypothesis that liberals are more likely to drink lattes than are conservatives. To be clear, no one to date has suggested that this oft-asserted relationship is causal. It is unlikely that drinking lattes causes a more liberal ideology or that a more liberal ideology compels one to drink lattes. Instead, we assume this relationship is spurious; that is, some third factor related to both latte drinking and liberal ideology induces a relationship between liberalness and coffee-drinking preferences.

Assuming there is empirical support for a positive relationship between latte drinking and liberal ideology, the challenge is understanding why. Toward that end, we examine four possible explanations. We use the results of these analyses to address the underlying meaning of “latte liberal,” a phrase intended as a derogatory term describing those who claim to care more about the poor than their conservative brethren, yet indulge in consumption of high-priced beverages when they could probably make do with a regular cup of joe.

Our first hypothesis is that this relationship occurs because latte consumption is a function of the sheer availability of coffee shops. Although chain coffee shops are everywhere in contemporary America, they are more prevalent in urban areas, where liberals are more likely to live. Because lattes require espresso, and people in urban areas are more likely to have nearby coffee shops, lattes may be simply less available in rural (and more conservative) areas. Figure 1, panel A, diagrams this possibility. The positive relationship between latte consumption and liberalness occurs because both are a function of living in areas with more coffee shops.

Figure 1 Four Possible Explanations for a Spurious Relationship between Liberal Ideology and Latte Drinking

A second possibility, illustrated in figure 1, panel B, is that the cost of purchasing one’s coffee beverage at a coffee shop means that both latte consumption and liberal ideology are functions of income. According to a 2015 survey, consumers will spend $3.28, on average, for a cup of regular coffee at a coffee shop; for barista-prepared beverages, the cost can run much higher (Brown Reference Brown2015). As a result, those with higher incomes find it easier to afford lattes than those with limited incomes. This interpretation fits most squarely within Brooks’ (1997) original usage—namely, that latte liberals have upscale incomes and live with high levels of creature comforts, but still think of wealth as somewhat embarrassing (Kelly Reference Kelly2014).

Third, we assess the possibility that this spurious association occurs for reasons related to gender. It is well established in American politics that women are more liberal than men (Kittilson Reference Kittilson2016). What is less well known is that women also are more likely to drink lattes. Based on coffee-preference data collected by Zagats, lattes are the favorite drink among women at 22%, with regular coffee trailing closely behind at 19% (Brown Reference Brown2015). The preferred drink among men is regular coffee at 30%, with espresso next best at 14% (Hernandez Reference Hernandez2015). Thus, latte-drinking liberals may be a spurious function of women’s preferences for lattes.

Why are men averse to lattes? According to qualitative interviews conducted at Starbucks (Tourjee and Ettachfini Reference Tourjee and Ettachfini2017), men view lattes as “girly” drinks, ordered in hushed tones when ordered at all. This is clearest in the case of the politically-charged pumpkin spice latte, which is apparently “inexorably coded as feminine” (Tourjee and Ettachfini Reference Tourjee and Ettachfini2017). Indeed, some deem hating on pumpkin spice lattes to be a form of sexism that invalidates women’s preferences relative to men’s (Timpf Reference Timpf2015). Thus, those who describe lattes as frothy and insubstantial are implicitly indicting women as the same.

A fourth possibility is that conservatives may be less likely to drink lattes as a function of their disdain for globalization. The difference between liberals and conservatives on this issue has come to the fore with Donald Trump’s presidency, but the pattern dates back to at least the early twenty-first century. In response to a Chicago Council on Global Affairs national survey asking whether “globalization, especially the increasing connections of our economy with others around the world, is mostly good or mostly bad for the United States,” liberals have long been more likely to say globalization is good. As illustrated in figure 2, the size of the gap between liberals and conservatives was only 7 percentage points in 2004; however, by 2017, that gap became an enormous 30 percentage points.

Figure 2 Support for Globalization by Ideology

Note: Percentages are from the Chicago Council on Global Affairs’ representative national surveys.

On the one hand, lattes consumed in the United States are domestic goods. It would be difficult, if not impossible, under FAA rules, to import a cup of hot latte. On the other hand, the name of a product may be as important as, if not more important than, its actual country of production. For example, in 2003, when the US conflict with France over whether to invade Iraq escalated, there were calls from people including Bill O’Reilly to boycott French products. Even the US House of Representatives cafeteria temporarily renamed its French fries and French toast, “freedom fries” and “freedom toast” (Loughlin Reference Loughlin2003).

According to recent evidence in the Review of Economics and Statistics, consumers are not particularly adept at identifying foreign products and often misidentify products’ national origins due to cues embedded in product names (Pandya and Venkatesan Reference Pandya and Venkatesan2016). For example, Tresemmé shampoo, Raison D’Être beer, and other French-sounding products took a hit from consumers in 2003 even though they were American-made products. Based on an analysis of supermarket products, the authors concluded that in 2003, sales decreased for products with French-sounding names—an effect that was particularly strong in areas where American citizens were heavily concentrated (Pandya and Venkatesan Reference Pandya and Venkatesan2016).

A burgeoning literature on “consumer ethnocentrism” suggests that high levels of nationalistic sentiment and opposition to trade promote consumer avoidance of foreign-sounding products (Shimp and Sharma Reference Shimp and Sharma1987). “Latte” is an Italian word and, most importantly, it is clearly not American in terms of its linguistic roots. Although the United States has no immediate conflict with Italy as it did with France in 2003, nationalistic sentiments are on the rise. People claim to prefer American products and to shun “foreign” goods, even though many products with foreign-sounding names are produced in the United States. Nonetheless, the mere appearance of foreignness may be sufficient for consumption to be affected by nationalistic sentiments. Nationalism is higher among conservatives than liberals; therefore, America-centric purchasing behavior by conservatives may be responsible for an association between latte drinking and liberalness.

Thus, in addition to evaluating the empirical basis of this basic claim, we test four potential hypotheses to explain this spurious relationship. The relationship between liberalness and latte drinking occurs because:

RESULTS

First, we examine whether and how hot beverage preferences are related to ideology. Figure 3 reports the percentage of liberals, moderates, and conservatives who prefer various forms of coffee. There were no differences between liberals and conservatives in their propensity to like coffee in general. Overwhelmingly, most Americans prefer regular brewed coffee, even when offered an array of more exotic choices. This is true of liberals, conservatives, and moderates. However, this is not to suggest that there are no important ideological distinctions. As shown in figure 3, liberals are significantly less likely than conservatives and moderates to prefer regular brewed coffee. Most importantly, the pattern so widely proposed turns out to be true: although only 16% of liberals prefer lattes, this percentage is significantly greater than the 9% of conservatives who prefer lattes. Moderates fall in between with approximately 11% preferring lattes. Among other types of specialty coffees, there are no significant ideological differences among drinkers.

Figure 3 Coffee Preferences by Ideology

Notes: Data are from a representative national probability sample interviewed online by GfK, Ltd., in 2014 (N=1,053, coffee drinkers only). Comparisons among these three groups demonstrated significant differences by ideology groups for latte drinking (F=4.07, p<0.05) and for regular coffee drinking (F=7.04, p<0.01). Post hoc comparisons indicated that conservatives are significantly less likely to prefer lattes relative to both liberals and moderates; liberals are significantly less likely than conservatives or moderates to prefer regular brewed coffee. Other hot beverages were not preferred more by one ideology than another among a representative sample of American coffee drinkers.

To simultaneously evaluate multiple predictors of latte drinking, we used probit regression, given our dichotomous outcome measure. As confirmed in the first column of table 1, liberalness on a 7-point scale significantly increases the probability of latte drinking. But what accounts for this pattern? One common approach to testing causal relationships is to dump a collection of demographic variables into the equation to determine whether the relationship survives these control variables. However, this approach is problematic for two reasons. First, when simultaneously incorporating many variables into an equation—even if the central relationship of interest reduces in strength and/or fades to insignificance—it is difficult to know which particular variable explained the spurious relationship. Second, incorporating more parameters into a model reduces the number of degrees of freedom, automatically making it more difficult to achieve statistical significance, even if the variables included are not the reason for the spurious relationship.

Table 1 Predictors of Latte Drinking, Controlling for Potential Explanations

Notes: Entries are probit coefficients with standard errors in parentheses. Results are based on probit regressions with the dependent variable coded as Prefer Lattes (1) or Not (0). Liberalness is coded on a 7-point scale. Analysis based on the subsample of self-identified coffee drinkers.

To avoid these problems, we began using the most parsimonious approach, individually testing each potential explanation illustrated in figure 1 to determine whether any one succeeds in reducing or eliminating the spurious relationship. If no single explanation accounts for the relationship, then more complex models can be considered.

First, in table 1, model A, we consider differences in the availability of lattes by controlling for the number of Starbucks in each respondent’s ZIP code area. Availability definitely matters; people residing in areas with more coffee shops are more likely to drink lattes. However, after including the availability predictor, the size of the coefficient corresponding to liberalness does not change; liberal ideology still predicts greater preferences for lattes to the same extent.

Second, in model B, we instead include income in the model as a predictor. Higher household incomes do, indeed, predict greater latte consumption. But again, the coefficient corresponding to liberalness does not change. Thus, there is little evidence that either income or the availability of coffee shops is the reason that liberals are more likely to drink lattes.

Third, we test the hypothesis that gender may account for the positive correlation between liberalness and latte drinking. As shown in model C, women are more likely to prefer lattes than men. Nonetheless, gender has no effect on the predictive value of liberalness. The coefficient remains constant.

Fourth, in model D, we test the hypothesis that this relationship results from the fact that liberals are less nationalistic than conservatives. Conservatives are more likely to shun foreign-sounding products; thus, nationalism may account for the positive relationship between latte drinking and liberalness. As shown in model D, nationalism does indeed negatively predict latte drinking. Moreover, unlike the previous analyses, once nationalism is taken into account, the coefficient for liberalness diminishes from 0.08 (p<0.05) to 0.01, and it is no longer statistically significant. Although nationalism may not be the only possible explanation for the spurious relationship between liberalness and latte drinking, it is clearly the most powerful and parsimonious among those we examined.

The final column of table 1 shows the fully saturated model including all variables brought to bear on explaining this relationship. Interestingly, all of the additional variables—availability, income, gender, and nationalism—continue to significantly predict latte drinking. Nonetheless, it is purely the inclusion of nationalism that explains why this spurious relationship exists.

Our goal has been to explain why latte drinking is associated with liberal ideology. We did not seek a rigorous causal explanation for this relationship because it is most likely to be spurious. Instead, we sought an underlying reason for this somewhat illogical association. Our analyses suggest that nationalistic attitudes toward the superiority of one’s country—and, by implication, its products—most parsimoniously explain this relationship.

Affinities between lifestyle and ideology contribute to greater stereotyping and social fragmentation. If conservatives forego lattes because they sound Italian, then how can Americans ever hope to come together over beer (domestic or imported) and pizza?