On Comparing Education Spending Across Time

I’ve noticed a lot of back-and-forth recently on social media about education spending – specifically, on how spending has changed over the years in the United States.

The usual context is someone complaining about how spending in K-12 schooling has soared over the past few decades, but outcomes haven’t improved. I and others have repeatedly pointed out just how dumb such claims are, so no need to rehash it. Let’s instead set aside outcomes for the moment and focus instead on inputs: how much more has the U.S. spent on schools, across the years?

When I see people sling around the numbers, I find they tend to break down into measures that range from valid and useful to completely worthless (and probably deliberately deceptive). Let’s arrange these from worst to better, with the goal of producing the most reasonable estimation of how much K-12 school spending has changed.

-       Total spending per year. This is simply the total amount spent on schooling in any one year. Anyone who tries to use this measure is either hopelessly inept or a con artist. The most obvious flaw is that the number of students changes in any year; total spending makes no attempt to account for this. Any time you see this measure being used, ignore it.

-       Per pupil spending per year. This is barely an improvement on above, because there is no adjustment for changes in costs over time. The cost of a textbook or a gallon of gas or an hour of a teacher’s work is different in 1970 than it is in 2020. Again, ignore anyone who cites this figure.

-       Per pupil spending per year in “real” dollars. This is probably the figure you’ll see referred to most often by folks making the claim that we spend so much more than in the past but still suck. Spending is given in a per pupil figure, and the figure is adjusted over time for the changes in the price for goods, usually consumer goods. This means the figures are in “real” or “constant” dollars. Because this measure does account for changes in student populations and in the prices of goods across time, it seems to be a valid measure…

Until you start digging in.

First, and less important: while “real” dollars are calculated by estimating changes in costs across time, they rarely account for changes across space. A plate of pasta at a nice restaurant in New York City, for example, tends to cost a different amount than that same meal in Omaha. This problem is that populations can migrate, with proportionally more or fewer people living in less or more expensive areas of the country year-to-year. Constant dollars, in the aggregate, almost never account for these changes.

Second, and more important: changes in the prices of consumer goods, such as those reflected in the Consumer Price Index (CPI), do not necessarily reflect changes in the costs of schooling. K-12 education is a labor-intensive endeavor, and labor costs do not shift perfectly in sync with energy or food or other consumer costs. In other words: a big-screen TV may cost less this year than last, but that isn’t an accurate reflection of the change in cost of a well-qualified teacher, which may well cost more.

“Real” dollar spending per pupil is, therefore, highly problematic as a measure of K-12 spending. In general, education policy stakeholders should avoid using it.

-       Share of total or new spending. This measure has the advantage of having a built-in adjustment for inflation and population growth over time. We can measure the changes either for all spending, or just increases in spending from a designated starting point. The first question we need to address, however, is: Total spending on what? Total governmental spending? Total spending on all goods and services?

Simply asking the question reveals the problem: we’re still not accounting for differences in the relative costs of things across whatever total spending we’re measuring. If we limit the measure to governmental spending, we have to assume the relative costs of different services of the government never change. This is a very big assumption: information technology advances, for example, have made some parts of the government more efficient than others. So education spending may rise relative to, say, administrative costs for Social Security, simply because computers can replace clerical workers but not teachers.

-       Share of the economy/GDP. In my work with Bruce Baker and others on school finance, we refer to this as effort. It’s a useful way to compare different jurisdictions, although it has its limits. States with wealthier economies don’t have to put forth as much effort as states with less wealthy economies to generate the same amount of revenue for schools. So a state may look to be making less effort than another, but the amounts raised for schools are equivalent.

This measure also has the same problem as measuring shares of spending across time: differences in costs across different sectors of the economy can’t be accounted for. There’s also the issue of how the overall economy can shrink and grow, but spending on education could remain the same. This would mean that effort would also rise and fall without any change in how much is actually spent on schools.

-       Wage adjusted per pupil spending. Somewhat complex, this measure is one of the better ways to deal with the problem of differences in education costs across time and space. The premise is this: because education is labor intensive, we should try to determine how labor costs vary over time. However, we don’t want to simply look at educator wages: if we do, we won’t see how changes in the relative compensation of educators might vary in ways that also change the quality of people entering the profession. In other words: spending on teachers may go down compared to other workers, but so might the quality of people who choose to become teachers.

The solution is to look at the changes in wages of other workers who are similarly educated (and have other similarities, such as age). If it costs less to employ a college-educated worker in one place and/or year than another, we can fairly assume it will cost just as much less to employ an educator, without having to expect their quality will be different.

There are, of course, many assumptions and limitations built into this type of measure. Educators wages may fall or rise relative to other wages due to things like job satisfaction, which means relative wages might change but teacher quality does not. And while about four-fifths of K-12 spending is on staff salaries and benefits, that still leaves one-fifth of expenditures that will not necessarily track with labor costs. I would argue, however, that this is still better than trying to adjust costs through the CPI or some other consumer price adjustment.

All of this highlights an important point: the cost of an education is not the same as the spending on education. Spending is simply the funding shelled out for schooling. Cost, however, is how much must be shelled out to meet a certain standard. We can easily cut spending for schools, but we can’t then expect schools to meet the standards they were meeting before (unless we think they were inefficient to begin with – we’ll save that discussion for later…).

Over the past several decades, we’ve expected schools to do more – much more. The Individuals with Disabilities Education Act, passed in 1975, requires a “free appropriate public education” for students with disabilities. State and federal laws passed in the last several decades have required schools to set more stringent curricular standards, accompanied by tests that have grown more rigorous over the years. School shootings have raised the bar for school safety. Parents have demanded more programs and a wider curriculum. Now the pandemic puts new demands on schools for health and safety.

Is it any wonder school spending has increased? And the spending would not necessarily lead to commensurate gains in things like test scores; the outcome measures we use aren’t going to pick up things like expanded arts programming or more inclusive environments for children with special learning needs.

When it comes to changes in school spending, we have to take all of these things into account. Simple spending measures with inflation adjustments are not going to cut it. If people are interested in a serious conversation about public school finance – and they should be – they’re going to have to do better than throwing out flawed measures of school spending with no discussion of their inherent limitations.

ADDING: The economist Richard Rothstein has a nice explanation of the problems with using CPI in school spending measures here: https://www.epi.org/publication/books_wheremoneygone/ (p.9) Included is a discussion of “Baumol’s disease,” the phenomenon of uneven productivity gains across the economy.

-->