Once More, With Feeling: Millennials Haven’t Given Up on Democracy

In the months after the 2016 US presidential election this figure was widely circulated:

The figure and others from a paper Roberto Foa and Yascha Mounk in the Journal of Democracy, was widely reported to show that Millennials globally were “rapidly losing interest” or “ha[d] lost faith” in democracy. This interpretation of the argument Foa and Mounk in fact made was far too strong, as they themselves acknowledged in an online exchange shortly after their paper went viral. However, the idea that there is a crisis in commitment to democracy among young people has nevertheless become part of the “conventional wisdom.” The inspiration for this post, in fact, was hearing a prominent British politician invoke the trope in a speech today.

It will be helpful to concentrate for purposes of clarity on “Figure 1” from Foa and Mounk’s 2016 paper “The Democratic Disconnect.” In particular, it is a visualization of their headline claim about declining numbers of young people believing it is “essential” to live in a democracy. Similar points apply to other claims Foa and Mounk make and can be found in several response papers from political scientists.

The data for Figure 1 comes from the 2005 and 2014 editions of the World Values Survey. Specifically, it is based on the following question:

“Q: How important is it for you to live in a country that is governed democratically? On this scale where 1 means it is “not at all important” and 10 means “absolutely important” what position would you choose?”

Note that the word “essential” appears nowhere in the formulation of this question. What Foa and Mounk do is aggregate responses from 2005 and 2014 and show the percentage of respondents giving a 10/10 response on a 0/10 scale by birth cohort.

Unfortunately, Figure 1 doesn’t tell us anything about the numbers of those who didn’t give a 10/10, or indeed the mean responses by birth cohort. While we might intuitively expect that the means for earlier cohorts would be higher than later ones, we don’t know how large the difference is. And if you look at the WVS Wave 6 data for the United States, for example, you see that the mean for respondents < 30 is a 7.67, 30–49 is 8.24, and >50 is 8.89. Perhaps we should be concerned by this difference in means by age group, but we shouldn’t interpret this as evidence that young people are giving up on democracy. This is all assuming, of course, that the scales of different birth cohorts are similarly calibrated.

The next thing to note is that I just made a slight equivocation by switching from birth cohorts to age groups. Birth cohorts age, age groups don’t. However, given that Figure 1 is based on data from surveys completed only six years apart, Figure 1 also runs together cohort differences with age differences. To see this, compare this figure from one of the response papers to Foa and Mounk:

Note how the percentage responding “very good” rises by age group in each year of the survey. This reflects in part what political scientists call a “life-cycle effect” in political attitudes: people’s political behavior and attitudes change as they age. In particular, attitudes tend to become more favorable and people tend to participate more as they get older. What Figure 4 does show is a downward shift in the age-approval gradient (at least measured by percentage answering “very good”), but the shift in fact seems more pronounced for Boomers than other generations. Compare the difference between the 45–54 age range in the 1996 data with 65+ in 2011 (Boomers), for example, with the difference between the 15–24 age range in 1999 and the 25–34 in 2011 (end of GenX/beginning of Millennials). I emphasize *seems* because the WVS is not panel data and therefore not tracking the same people over time.

Figure 1 also obscures variation in the age or birth cohort-approval gradient across countries. The relationship between birth cohort and approval of democracy is nonexistent in up to half of post-industrial democracies, depending on the question in the World Values Survey you consider. There does appear to be a real birth cohort-approval gradient in English-speaking democracies, however.

Finally, Figure 1 does not tell us much about overall levels of support for democracy. As Alexander and Welzel write in their response to Foa and Mounk:

… levels of democratic support are astoundingly high even for the least supportive cohorts and only vary within a narrow corridor. Indeed, from the third round of the WVS in 1995–98 (i.e. the first round in which the respective items were fielded) to the sixth and most recent round in 2011–14, average support for democracy fell only from 81 to 77 percent. Age-related differences in support only vary between 79 percent in the oldest birth cohort (i.e. people born before 1930) and 74 percent in the youngest one (born after 1980). Even though these age-related differences are statistically significant, they account for a negligible 0.9 percent of the variance in support for democracy. In a nutshell, the decline in support for democracy across birth cohorts and over time is by no means as dramatic as Foa and Mounk propagate.

The bottom line: there is no good evidence that Millennials as a generation have lost faith in democracy and opened themselves up to populists and military juntas. People with public platforms should stop repeating this unsupported claim.