What's an education worth - Data note

This post is meant to accompany What's an education worth - Part II. I originally intended it to be a note at the end of that post, but the note turned out to be almost as long as the post, so I hived it off into its own write-up.

The data on college completion are from the Census Bureau, via this page.

In addition to the data shown in the main post on college completion among people 25 and older, it's interesting to look at specifically the 25-to-29-year-olds, which gives as close a picture as this dataset allows as to what portion of young people are in college.

In 1947, male and female attendance is pretty much equal. It looks like the war kept men out of college who would otherwise have gone. But then the GI Bill kicks in, and men's rate of college jumps rapidly.

It goes up again in the 1960's, presumably at least in part a reflection of men going to college to avoid the draft.

But women's attendance rises sharply at the same time, and while men fall off in the 1970's, women keep going at higher rates, passing men in the 1990's and not looking back.

For expenditure on higher education, I combined a couple of sources tables available through the National Income and Product Accounts of the Bureau of Economic Analysis, with Congressional Budget Office data obtained from the Federal Reserve Economic Data (FRED) site maintained by the St. Louis Federal Reserve Bank, as well as their Archival Economic Data (ALFRED).

BEA's Table 3.16 gives government expenditure, broken down into various different purposes, and shown as either all government together, or federal vs. state and local. Within the category of “Education,” you can distinguish between “Elementary and secondary” and “Higher.” The data run from 1959 through 2016.

Table 2.4.5 from BEA shows personal consumption expenditures broken down by different things people spend money on, including on “Education services” and, within that, “Higher education.”

I’m not sure whether that captures all the money spent on higher education in the country, but it seems to me it should give a reasonably accurate overall picture. Education is basically paid for by the students (or their families) and the government, and personal consumption expenditures reflect the first, while government expenditure shows the second.

The first question I asked was how important higher-ed spending was within government budgets. This chart shows each of the three groupings (“All levels,” “Federal,” “State/Local”), and each line is higher-ed expenditure by that level of government as a percentage of the total expenditure of that same level of government.

You can see that spending by all levels of government together rose from 1.2% of their expenditure in 1959, to 3.0% in 1977, and has fluctuated between 2.5% and 3.0% ever since.

Most higher-ed spending by government comes at the state level rather than federal (think state universities), so it plays a larger role in State/Local budgets, rising from 3.9% in 1959 to 6.6% in 1976, and falling thereafter.

Note the precipitous decline from 2001 to 2003—I don’t have a good explanation for that, and would be curious if anyone either knows why it happened or has a possible explanation.

Federal higher-ed spending rose from less than 0.4% of overall federal spending in the early 1960’s to 0.9% in 1981, holding relatively stable from there until 2008.

The early Obama years saw a marked increase in federal higher-ed spending, with a return back to 1% by 2016.

Another way to look at spending is as a share of GDP. BEA's Table 1.1.5 gives nominal GDP (that is, not adjusted for inflation), which is the correct benchmark, since the government and household expenditure figures from the other two tables are also in nominal terms.

A striking feature of this chart is the jump in spending by government overall from 2007 to 2010, driven primarily by a jump in federal spending. But remember that from late 2007 to mid-2009, real GDP (i.e., inflation adjusted) actually shrank, and it didn't regain its previous peak until early 2011. When you combine increased spending as part of the stimulus package of 2009 with a shrinking GDP, that gives you a result of a large increase in spending as a share of GDP.

In other words, a piece of that jump in spending is an arithmetical artifact of the recession. The easiest way to control for that is to use potential GDP instead of actual.

Potential GDP isn’t something we can observe or measure. It is, rather, a calculation of what the GDP in some sense “should” be, what it “would” be if unemployment were at its “normal” level. And so it should in principle be immune to business-cycle fluctuations.

In reality, when a recession hits, the economists who calculate the potential GDP go back and revise it and say, “Well, potential didn’t fall like actual did, but it did start to rise more slowly during the recession.”

But every estimate of potential GDP from the Congressional Budget Office includes a forecast of potential GDP 10 years into the future. The ALFRED site mentioned above provides older versions of the data currently available through FRED, so you can go see what the CBO potential GDP estimate was in 2006, which includes the forecast to 2016. Once we adjust that from inflation-adjusted to nominal terms, we can compare it with the expenditure items.

The 2006 estimate was made before the recession of 2007-09, so it doesn’t have the slowdown in growth that you see if you use the current estimate. Using those older, higher figures for potential GDP, the increase in spending is a little less dramatic, though it certainly doesn’t go away.

The next chart separates out government and private spending, using the 2006 potential GDP as the point of comparison. You can see the run-up in government spending from 1959 to 1975, the growth in private spending, particularly from 1979 to 2010, and the increasing share of spending done directly by households, as their total spending increases while government's stays flat.

While the data on spending and actual GDP are from the Bureau of Economic Analysis provide the data on expenditures and actual GDP, the potential GDP data are through the data sites of the St. Louis branch of the Federal Reserve.

FRED provides potential GDP in both nominal form (series label NGDPPOT) and real (series name GDPPOT).

ALFRED provides older estimates of real potential GDP. The one I used was from August, 2006, and has the series name GDPPOT_20060817.

There are a few steps involved in making the 2006-vintage real potential GDP commensurable with the 2018-vintage nominal potential GDP, but explaining that would take another page or two. If anyone is curious, let me know in the comments section and I will oblige.