A front-page story in Chron. Higher Educ. Jan 2011 claims that newly elected Republican governors are more likely to cut spending than Democrats or even incumbent Republicans. Since education (which includes both K-12 and higher ed) is one of the largest components of state budgets, this implies that the higher education expenditures will fall in states where governors plan to make budget cuts.

As evidence, the article contains the graphic shown below. The chart shows the total projected budget shortfall in 2012 for six states compared to the higher education appropriation in 2010.

CollegeBudgets

Look at all the pretty circles. Image from Chron. Higher Educ.

There are so many problems with this graphic that it is hard to know where to begin. But I will try to critique it.

1. The most serious problem is that this chart displays a meaningless comparison. There is no a priori reason to believe that higher-education appropriations in FY2010 and total projected budget shortfall two years later in FY2012 should be correlated. Nor should one expect the two numbers to be about the same size. My guess is the fact that the two numbers appear to be the same magnitude for these six states for these two years is just a coincidence.

2. The chart should include more years. If expenditures and budget gap are correlated, then the relationship should be stable over time. Prove it by showing time series data of total budget and higher-education appropriations during previous years. It would be even better if data from previous recessions was available. My guess is that the numbers vary greatly and the relationship disappears.

3. The chart should include more states. The article and the chart predicts newly elected Republican governors will cut higher education budgets more than other classes of governors. But the chart only shows data for six states with newly elected Republican governors. There is no comparative data for incumbent Republicans or any Democrats. Again, if the relationship is due to party affiliation and incumbency of governor, then showing data for all states will emphasize it. My guess is such data would show there is little or no relationship between governor status and budgets.

4. The chart should be adjusted to aid comparisons. The size of state budgets and higher education expenditures vary because of many factors. At a minimum, the data should be adjusted for population. A few other possible adjustments that come to mind are median household income, population of 18-25 year olds, proportion of population with college degrees (presumably, grads will be more likely to favor higher ed spending), enrollment at in-state public colleges (presumably parents of current students will be more likely to favor higher ed spending), etc.

5. The chart should not use areas to represent linear values. Look at the two circles under Ohio. The red circle representing the budget gap of $3,000 million is more than twice as large as the orange circle showing the higher education appropriation of $1,900 million even though $3,000 million is less than 60% bigger than $1,900 million. That’s wrong. The area of a circle is proportional to the square of the diameter. One of the basic rules of drawing a graph (for example, see Edward Tufte’s The Visual Display of Quantitative Information) is to make sure the size and shape of data markers actually assist in understanding the values being measured.

****

My proposed improvement to the original chart is shown below. The data come from the same sources used in the Chron. Higher Educ. chart, state budget gaps from the Center on Budget and Policy Priorities, state higher education expenditures from the  Illinois State Univ. Grapevine Project, and political alignment from the Nat. Conf. State Legis. I also added U.S. Census annual estimates of population and U.S. Census counts so that I could calculate per capita state budget gaps and higher education expenditures. The data are posted on SkyDrive in case you want to create your own charts.

My scatterplot shows the cumulative budget gap per capita at the end of FY2009 for each state on the horizontal axis and the increase or decrease in higher education expenditures for state monies (excludes federal stimulus monies) from FY2009 to FY2010. My hypothesis is that deficits at the end of one fiscal year may impact changes in expenditures in the following year.

Each data point is colored to show the political affiliation of the governor (using blue for Democrat and red for Republican). The marker shape indicates if the party changed hands (triangle) or if it did not (square or diamond) in the 2010 election. The green line shows the least squares line through the data (not weighted for state population).

EducByGap

Look at all the tiny dots. Chart by George Taniwaki

I didn’t spend much time analyzing this chart but a few things caught my eye. First, the slope of the regression line is negative meaning that states with the largest deficits tended to cut education expenditures the least. Specifically, California, with the largest deficit at the end of 2009 at $1,004 per capita actually increased higher education expenditures in FY2010 by $7 per capita.

Second, the states with the largest budget gaps at the end of FY2009 were more likely to have Republican governors. They were also more likely to elect a Democratic governor in 2010. The three states that fit this pattern are California, Rhode Island, and Connecticut.

Finally, the states that made the biggest cuts in higher education expenditures in FY2010 were more likely to have a Democratic governor. They were also more likely to elect a Republican governor in 2010. The three states that fit this pattern are Wyoming, Iowa, and New Mexico.

It could be that these correlations are spurious. But my conclusions are no more unreliable than those claimed based on the weak analysis in the original chart.

Much thanks to Susan Wolcott who pointed out the original chart to me.

Advertisements