SAT scores and Income

[older material below...]

Just found these data for NC: The fraction of students taking free or reduced lunches (a measure of poverty) and the composite performance score for 2,535 schools in NC. These data are pruned from larger sets to make sure both items exist for each school.

A particularly large argument has erupted in Wake County with the election of a new school board determined to eliminate busing for diversity. Some have made comparisons with Charlotte-Mecklenburg schools, so here’s a plot pulling out that data:

[click on the plot for a larger image] Clearly, the C-M school district has more economically challenged school districts than Wake County. Why? Below I show data indicating that Wake County has the second-highest per capita income in the state. Ahhh, how about that: I just found out that Mecklenburg County (it holds Charlotte) has the highest per capita income. So, the question might be, has Wake County’s busing reduced the high poverty schools and increased performance? Wake’s average performance is 77.9 whereas Mecklenburg’s is 74.4, giving Wake about a 5% higher score despite having a lower income. Is busing for diversity increasing the scores of students whose neighborhood school would have a greater concentration of lower income students and lower performances? I don’t know. Would Mecklenburg do better by transferring the students from schools with, say, more than 60% free lunches to higher income schools? I don’t know. Downey (2007) provides results on segregation and minority income for major U.S. cities. Both Raleigh and Charlotte have B/W income ratios of 0.591 (blacks have 59.1% of the white population’s income), but the segregation in Charlotte is greater: 55.2% of Charlotte’s black population would have to move to reach a “uniformly mixed population”, whereas only 46.2% of Raleigh’s black population would have to do so (apparently that’s a standard measure for segregation).

So, both metropolitan areas have similar economic conditions for blacks, Charlotte is more segregated, and Wake has bused for diversity. Presumably, both factors could be responsible for Wake’s historically better performance. I guess we’ll have to wait and see how the new Wake school board’s experiment plays out. It would be nice to see the data from the studies they have that form the basis of their changes.

Here’s the spreadsheet with NC schools with information on free lunch and performance. I’ve pruned some data where one or the other wasn’t available: combined

School performance decreases with this particular measure of low income. Here are data from three counties around Durham:

All three counties show similar trends with low income, though Wake County seems a little higher (I didn’t do the stats yet, but the points sure look higher). That is until the new Wake County School Board gets done doing what they’re doing with their schools.

Just to be clear, higher family income promotes better scores on standardized tests. Here I show data from North Carolina counties, plotting SAT scores versus family income, measured over the 100 counties:

satincomeOverall, a pretty dramatic dependence. These data don’t really belong in “environmental inequity”, in the sense that family income doesn’t fall under what I’m thinking of as the natural environment. However, as I’ve posted under the Durham category, family income also influences tree canopy levels. Wake County prides itself on higher education scores, yet that pride should be tempered by acknowledging it’s level of wealth.

It’s not just a matter of spending money on students. This plot below shows SAT scores varying as functions of federal, state, and local expenditures per pupil. All of the monies add up to essentially equivalent total EPPs, with state and federal sources making up for lower local sources, but it’s really the above plot — family income — that seems to determine student achievement.


The data was/is available on the web, but I’m happy to provide it to you.

Here are the SAT scores across school systems in NC, 2004-6. 2006Table6

(I’m looking in my files for my county income source.)

I’d like to collect more examples of income and academic performance: If you have one or more, please send me a comment with a link to the data or publication. I’ll take a look, plot it up, and post it here.

9 Responses to “SAT scores and Income”

  1. Lance Olive says:

    Thanks for taking the time to plot this data.

    What your chart above shows is that there is appears to be a very loose non-linear relation between per capita income and SAT scores. I say that because below $26k there is no correlation at all. It does not demonstrate that income is a direct contributor to SAT scores. It could be indirect, and not in the way you might think, or not really related at all.

    For example, suppose having a comfortable family income allows more families to choose to home school (or even better, private school), and what if there is data show that wealthier counties have a higher percentage of home/private school students, and what if there is data that shows that these students have a higher average SAT score than those in public school?

    In that hypothetical example (I don’t have data to support or refute, it’s just stated as an alternate explanation to the theory that higher income supports better scores) it can be argued that family income, especially government expenditures, has less to do with achieving the high scores than is being presumed.

    I applaud your inclusion of the EPP charts. This is no way someone can look at them and justify a federal or state level, taxpayer funded, program in an attempt to raise scores. On the contrary, the argument should be that spending less on education at those levels of government would lead to higher scores.

    But there aren’t many government officials in education who would listen to that because it defies their core belief that the way to fix our education system is to spend more money.

  2. Will says:

    The correlation is quite strong, and correlation doesn’t identify causation. I have no information on home school/private school influence, nor have a horse in that race, as it were. Again, the correlation is strong and exists, no doubt. Federal and state expenditures mostly make up for insufficient local funds, taking absolute $ out of the picture. Your conclusion about the EPP charts is off base. One could look at the federal EPP and see that the more the feds spend, the lower the scores, but then you need to think about mechanism. The feds spend less because local funds increase because property values are high where people have high income where students earn high SAT scores. That source of funds goes up only because local funding sources go down. Again, blatant correlation doesn’t identify causation.

    The point is that roughly constant dollars are spent over all the counties, but higher family income promotes higher SAT scores. Yes, one could use the SAT score variation at the low income range to uncover approaches that lead to student success.

  3. Lance Olive says:

    Perhaps one way to look at the chart is higher income has some relation to lack of failure (indicated by the lack of data points in the lower right quadrant), but that successful county scores are attainable without it (indicated by the left third of the chart).

    And, yes, I agree 100% that correlation doesn’t identify causation. That was actually the point that I was making. It’s great to expose data, especially as a picture (chart), but it’s another thing entirely when people draw conclusions from a bit of data, unless many possible explanations are considered and either shown to be consistent or faulty.

  4. Kevin Davis says:

    Hi Will,

    Interesting data. Thanks for sharing.

    Since high Federal/state funding tends to happen in disadvantaged impoverished districts to make up for a lack of local funding, it would be interesting to see a plot of total expenditure per pupil against SAT score. What does that show? Also, have you done any multivariable analysis to examine the relative impact of EPP vs. family income?

    I don’t quibble with your basic assumption, by the way, which reiterates the real problems inequality plays in educational outcomes.

  5. Will says:

    I made that plot for earlier data (2000 math SAT scores and 1999 income data) and there’s just no clear correlation. Perhaps I’ll plot that for this data later (other stuff happening now), but don’t expect any dependence. Family income drives so much, like life expectancy (a several year variation across NC counties correlated with income), I’m just not surprised by academic performance showing the same thing. And, no, I haven’t done the stats, but I expect SAT scores won’t show any variation with total EPP; in that sense, it’s all income.

  6. Jim says:

    What was R, and is a linear relationship the best fit?

  7. Will says:

    Thanks. r^2=0.21 (p<0.0001). I didn’t check any regressions other than linear.

  8. Howie says:

    Hey Will,

    Thanks! This is really interesting. I’m wondering how you account for the outliers–like the dots/counties on the upper left-hand corner of the map. What counties are those? I wonder what happens there to produce such high scores even with the fairly low family income in the area?

    Also, where’d you get the data?

  9. Will says:


    I don’t understand the source of the variation. I’ve edited the post to include the SAT data I found on the web (presumably available for all states), and I’m perplexed about where I have the income data on my computer.


Leave a Reply