In a previous post, I presented some simple data on “subjective class identification," which is the practice of asking people to place themselves within a class structure. The data show that, despite constant political rhetoric appealing the U.S. “middle class," more people actually consider themselves to be working class than middle class, and that this hasn’t changed much over the past thirty years.

I also noted that there is even a fairly significant “working class presence” – about 25 percent – among the highly educated (those with a bachelor's or higher). This struck me as interesting, given the fact that having a college degree is sometimes called “the ticket to the middle class," and also given that the income advantage for college graduates – the “college wage premium” – is substantial (and it's actually increased over the long term). I found myself wondering whether the relationship between having a college degree and “gaining entrance” to the middle class (at least by one’s own judgment of his or her class position) had changed over time. In other words, when it comes to subjective class identification, is college less of a middle class “ticket” than it used to be?

I couldn’t resist taking a quick look.

In the graph below, I show the percentage of people with bachelor’s or graduate degrees who consider themselves to be “at least” middle class (i.e., middle or upper class). I’m using the same data source as in the previous post (the General Social Survey), and limiting the sample to respondents 21 or older who are still in the labor force. In order to get better estimates, the data are pooled into five-year categories (note that the y-axis begins at 50 percent).

Between 1973 and 2010, there is a modest decline in the percentage of degree holders who identify as middle/upper class, going from about 80 percent in the 1973-75 period to roughly 72 percent in 2006-10. Much of the decrease occurs during two periods: Between 1986-90 and 1991-95, and an even steeper drop between 2001-05 and 2006-10. So, at least judging by this raw tabulation, there is some small extent to which a college degree is not the path to the (self-identified) middle class that it was during the 1970s.

Still, it’s important to note that this raw relationship between time and class identification is almost certainly being confounded by other factors. For example, women are more likely now to attend college (and be in the labor force) than they were in the 1970s; thus, if women’s class perceptions differ from men’s, part of the trend above might be driven by gender. Similarly, to the extent that earnings affect class identification, the slope in the graph might in part be due to changes (or perceived changes) in college graduate earnings.

So, let’s do a quick estimate that takes some of these underlying dynamics into account: Gender, race, income, marital status, age, and, of course, year and education. I would have liked to have added homeownership (which is often associated with middle class status), but the GSS variable is not available in all years.

The graph below presents the trend in predicted probability of identifying as a member of the middle/upper class among college graduates (or those with advanced degrees), while controlling for all the other variables. You can read the notes in the graph for more details on the model.

When looking at the graph, remember that the probabilities apply to only one specific point in the distribution of all the variables (a hypothetical respondent whose characteristics are the mean of each variable except education and year). The utility of the graph is to identify the change in probabilities over time, while “purging” other variables that may be clouding up the relationship.

That said, since the early 1970s, there has been a discernible decline in the odds of college graduates self-identifying as “at least” middle class, even controlling for key factors such as gender and income. The trend was fairly flat up until the 1986-90 period, and a fair amount of the decrease (about seven percentage points) occurred during the 2000s.

It goes without saying that class identification is complex and my model is simple (and uses cross-sectional data). In addition, the returns to a college degree are substantial. Still, these results suggest that, while college-educated Americans are very likely to identify as middle class or higher, the relationship may have weakened slightly over time. In other words, when it comes to people’s perception of their own class position, college is a bit less of a "ticket to the middle class" than it used to be.

- Matt Di Carlo