Education is the silver bullet

i'd control for demographics. lots of the current youth cohort are the children of immigrants for whom it might be a "reach" to go to graduate school, if their own parents didn't complete secondary school, or make it past elementary.

in fact, i'll look now and brb.

N too small for hispanic. can't tell definitively.

"What's interesting, and possibly troubling, is the decline in number of people completing grad school"

1) I'm having trouble convincing myself that that trend is real. The moving average has a downward slope, but it's a very gentle one in a lot of noise. Has anyone done any statistical tests on the data?

2) I'm not so sure that it would be a bad thing if it were real. It depends a lot on what those graduate degrees are. I often see articles and blog posts about how we're producing more PhDs than there are jobs for them. Keep in mind that "graduate degrees" doesn't mean science graduate degrees. We have far more English PhDs out there than there are jobs for them.

the trend in the GSS may not be real, but there's been a fair amount of reporting about the leveling off and possible decline of higher ed completion recently in the USA. this is disturbing because 1) education = skills for modern economy 2) there is a major wage premium for education, so people aren't responding to incentives.

but as i said, you really need to correct for the children of immigrants and their parental educational backgrounds. children of high school/education immigrants do go to university, but if you have parents who never made it past 8th grade, getting some post-secondary vocational training would be a major increment. but that would drag down the aggregate.

How were the solid lines fit to the data? They appear to be incorrect for birthdates after about 1970. The lines start coming down too soon and too gradually, whereas the data points themselves come down later and more steeply.

"What's interesting, and possibly troubling, is the decline in number of people completing grad school. High school graduation rates seem to have stabilized or even be declining, college graduation rates seem to be up until you get to populations to college-aged people, and there's no doubt that the fraction doing graduate programs has declined over the last couple of generations (since the cohort graduating college in the '70s)."

Parts of your post sound like you are saying that fewer people are doing grad school than used to. That can't be right. Field after field complains of too many grad students, too few jobs, etc. There are way more grad schools than there used to be probably in most fields, and definitely if you count huge fields like MBAs and nursing masters. And grad school is *way* more open to e.g. women, minorities, etc., than it would have been in the 1960s or 1970s.

So something seems wonky...the grad school pattern might be explained if those surveyed who were born in the 1940s & 1950s were predominantly e.g. white & professional, whereas those born in the 1970s & later were a more representative sample of the US population. Perhaps land-line vs. cell-phone bias could give such a bias?

The absolute number of people getting graduate degrees may well be growing, but the fraction of the population with postgraduate degrees is declining. I find that worrisome.

Rosie: It's a loess regression, locally weighted. It won't track the data perfectly, but finds a line which minimizes root mean square error within a certain window around each point. Hence the line's "too soon and too gradual" decline.

I'll second the "which graduate degrees?" line. Last I looked, the overwhelming majority of master's degrees awarded are in primary education for the simple reason that it's a requirement for tenure in most states. A PhD in education is also required for many school administrative positions.

Now subtract law and medicine, which also don't have much to do with the "innovations" issue.

The number of research PhDs in the residue is simply not going to show up on the chart.

Given that the error bars on the percent going to grad school are about half the actual percent, I'm pretty skeptical of drawing trendlines from these data and concluding that such a gradual slope is real. Of course, there's nothing in the figure to give us any idea what the error bars are (95% confidence interval, maybe?), so I can't really draw a conclusion. I'd be very interested if someone could run and analysis to see whether the trend is statistically significant.

Either way, I'm not convinced that a decline in the percentage of people receiving grad degrees is a bad thing. We may need more people with certain types of grad degrees, but that's not the same as needing more people with grad degrees in general.

I would argue that there are many fields with too many graduate students, where many of the students have chosen grad school with a flawed understanding of the demand for the degree or with a youthful belief that they will be the exception. The data don't show us whether the decline (if it's real) is somewhere where it's harmful or beneficial. It's a more complicated picture than fewer grad degrees = bad, and there's just not enough information here to judge.

I just posted a new graph showing the breakdown by race, so have fun with that. The error bars are 95% binomial confidence intervals. They're small around each point because I'm going by year, and that means small sample sizes at the tails. The trend line is a loess regression, which doesn't give many options for further testing. The important thing with that regression, though, is not the error bars around each point, but the sheer volume of data. I feel comfortable that the trends are real.

I agree, too, that what fields people are getting degrees in matters, especially since I had been talking about science literacy (that's why I didn't cite these data in the essay). But in general, the decreasing number of people seeking higher education is worrisome regardless of which field they choose to pursue. There are lots of factors driving that, of course, including the inadequate number of academic positions for newly minted PhDs. The incentive structure is all screwed up, and this trend is probably more a symptom of other structural problems than anything causative.

The loess fit may not be very statistically useful, but that doesn't mean that some sort of statistical test can't be performed. for example, one could pool the data from those born in the 40s, separately pool the data from those born in the 70s*, and compare the two. The 95% confidence interval for each would be reduced by about a factor of 3. I'm still not certain that that would be significant, but it might be.

That would still leave plenty of things that I'm not sure are accounted for. For example, lifespan is correlated with education level. Someone born in the 40s is more likely to still be alive if they were highly educated.

Regardless, I'm still not convinced that a decline in the number of people receiving graduate degrees would be a bad thing. Going through 24 years of schooling to end up in debt and doing a job that requires much less education is not generally a good thing. Education is not just for job training, of course, but that doesn't mean that more is always better.