Analysis: Correlation between education spending and graduation rates is … .006 percent
Yet another piece of evidence that there’s little to no correlation between spending and student achievement — let alone causation — and the source of this evidence comes from an unlikely place … the Las Vegas Sun.
A little background is in order.
Yesterday, the Sun had a story by Paul Takahashi with this headline, “Analysis finds correlation between high education spending, high graduation rates.”
In his opening paragraphs, Takahashi wrote:
The Sun analyzed federal education data on school expenditures and high school graduation rates and found the majority of states see a correlation between high education spending and high student performance.
But as the story progressed, the evidence Takahashi cited to support the Sun “analysis” actually hinted (more like screamed) that there wasn’t much of a correlation between spending and graduation rates.
Of the bottom half of states that spend the least on education, 14 states — or 56 percent — are also among the bottom half of states with the lowest graduation rates.
If there the relationship between the two items was random, you’d expect that percentage to be 50 percent. And then there was this.
The following nine states spent the most on education but have among the lowest graduation rates nationally: Alaska, Delaware, Hawaii, Louisiana, Michigan, Minnesota, New York, Ohio, Rhode Island and West Virginia.
That’s a lot of exceptions for a small sample size.
So instead of cherry-picking data, and I called Takahashi and got the source of his per-pupil-spending data and graduation-rate data. Both are federal government reports for the school/fiscal year 2010-11. Here are the results:
Assuming a 95 percent confidence level, this chart yields a R2 value of 0.00006. Also called the coefficient of determination, R2 is one of the best means for evaluating the strength of a relationship.
In these cases, the correlation between per-pupil spending and graduation rates is, at best, only .006 percent. This virtually non-existent relationship is graphically demonstrated by the fact that the data points are not closely grouped along the regression trend line, but are literally all over the chart.
“.006 percent” is not a typo. For some perspective on how microscope that correlation is, consider that .006 percent of $1,000 is six pennies. And again, correlation is worlds away from causation.
Just another reason — to add on top of 50 years of history — that spending more won’t increase student achievement or graduation rates.
* The chart only includes 47 states because the federal report did not include graduation-rate data for Idaho, Kentucky and Oklahoma.