Top performing cities on NAEP test have the least poverty — but some poor cities do surprisingly well

This article also appeared here.

This article also appeared here.

Do cities with less poverty test better? Yes, but the correlation is not as tight as you might guess, according to 2013 test scores released Dec. 18, 2013. I put together a spreadsheet looking at the percentage of students who qualify for free and reduced priced lunch in each of the 21 urban school districts that volunteered to be part of a National Center for Education Statistics assessment (known as NAEP TUDA). I then compared these poverty rankings to each city’s standing in fourth grade math. My original data came from here and here.

Charlotte, NC with the smallest percentage of poverty on the list (only 56%) has the top math score — as you might expect. But it’s interesting that Jefferson County, KY, which has the 4th smallest percentage of poverty (65% low income students) ranked 11th in math. You would have expected it to post a higher math score. Similarly, Atlanta and Washington DC post lower math scores than their poverty rankings would suggest. Conversely, Boston has higher poverty than most of the other cities. Yet its fourth graders posted the 5th highest score in math.

Here is my table….

City 4th Grade Math 4th Grade Math Ranking % Poverty Least Poverty Ranking(Most poverty = 21)
Charlotte 247 1 56 1
Hillsborough County, FL 243 3 58 2
Austin 245 2 62 3
Jefferson County KY 234 11 65 4
San Diego 241 4 66 5
Atlanta 233 12 73 6
Miami-Dade 237 6 74 7
Washington DC 229 14 76 8
Albuquerque 235 9 77 9
New York City 236 8 79 10
Houston 236 7 83 11
Milwaukee 221 18 83 12
Chicago 231 13 84 13
Los Angeles 228 15 84 14
Boston 237 5 85 15
Baltimore 223 17 87 16
Detroit 204 21 88 17
Fresno 220 19 91 18
Dallas 234 10 94 19
Philadelphia 223 16 94 20
Cleveland 216 20 100 21

Related stories:

Low-income inner-city achievement gap starts to close, test scores of urban school districts improve faster than nation over past 10 years, Washington D.C. stands out

POSTED BY Jill Barshay ON December 18, 2013

Comments & Trackbacks (8) | Post a Comment

Paul Bruno

Interesting, but FRPL eligibility isn’t really a measure of poverty. You can make well over the poverty threshold and still be eligible for reduced-price lunch.

[…] ranking, and DC and Atlanta and Jefferson County (KY) did worse.  Read the full post here: Some poor cities do surprisingly well.  What do you think?  Is it fair to compare cities much smaller like Boston to cities much bigger […]

bruce baker

Pictures are helpful here:

full post:

Current level scores are related both to prior averages and to poverty – seemingly more tightly associated with Census Poverty estimates than with school district low income enrollments (Free Lunch).

But, it’s really important to understand that these poverty rates across cities aren’t really comparable because they rely on the same income thresholds for poverty. So I wouldn’t try to make too much of those who fall above, or below the trendline.


So I am not sure why this is not strong correlation. Your table has 5 columns starting at the left with column one holding labels. If we compare columns 2 and 4 using correlation (the statistical procedure), which you can perform in spreadsheets, nothing special needed, the correlation is
r= -0.7456

If we square this to get the share of variation “explained” we get

r square = 0.5556 or roughly 56%.

You are right, it is not 100% (no one would ever think it would be) and an adequate regression model would consider a lot more, so the share associated with poverty may change in a more complete model, but this is very strong association.

Jill Barshay

@Dan Thank you for pointing out the rather high r-squared (especially in an education model). And you remind me that it’s very little effort to be doing a basic regression like this in excel. I will try to do that more in the future. But still interesting to me the cities, such as Boston, whose NAEP scores defy their poverty ranking. And I was surprised that Washington DC, despite its outsized gains in the past couple years, is still below where its poverty ranking would place it.

Jill Barshay

@Bruce Thank you for sharing your NAEP score vs. poverty graph. Curious to hear your thoughts on why Census poverty rate is tracking NAEP scores better than lunch eligibility? Why would Detroit have the highest poverty rate according to the Census, but less eligibility for lunch program than Cleveland? I’ve heard from other researchers that Latino families often don’t admit to being eligible for lunch even when they qualify. Perhaps that’s a factor in Detroit?

Brian Preston

Jill, thanks for the raw data. Using that, I ran a correlation between the math scores and the percent poverty in Excel using the CORREL function, which yielded a correlation of -0.74564, which rounds nicely to -.75. In the social sciences, of which education is one, this is considered a very strong correlation. The negative number, for the non-statisticians, means that as poverty increases, scores decline. I’ve written on poverty and educational attainment, and your blog gave me inspiration to do it again on my website. Keep up the good work!

Your email is never published nor shared.