This map supports equity in DeKalb County, Georgia. The aim is to find the structural causes of vulnerability so they can be removed. "Structural" means conditions built into how housing, jobs, health care and money work, not traits of the people who live there. The no-diploma layer narrows this to one structural driver, lack of a high school diploma. Across four of the five Dunwoody cluster elementary schools there are 673 unused seats (capacity minus enrollment, from an undated DeKalb planning table). This app is meant to help understand opportunities to fill them in a way that targets vulnerability reduction in other census tracts.
The CDC Social Vulnerability Index (SVI) shows which places may need the most help in a disaster or crisis. It uses 16 measures, such as poverty, people without health insurance, and households with no car.
The map works with census tracts. A census tract is a small neighborhood area set by the Census Bureau, usually with 2,500 to 8,000 people. DeKalb has 203 tracts. 201 of them have SVI data, and this map uses all 201.
Each measure is turned into a percentile. A percentile tells you how a tract compares to every other tract in the United States. A percentile of 0.80 means the tract ranks higher than 80% of US tracts on that measure. The 16 percentiles are added up and ranked again. That gives one overall score from 0 to 1. Higher means more vulnerable.
Which tracts are included? You first tried leaving out the least vulnerable tracts, those in the lowest quarter nationally (a score below 0.25). Then you chose to use all tracts. So every result below covers all 201 tracts.
Which measures make up a quarter of the score? In most tracts, 3 measures add up to 25% of the score (128 tracts). In many, it takes only 2 (65 tracts). In 6 tracts, one measure alone is enough. A few need 4 (2 tracts).
The top 5 contributors. These are the measures that most often rank in a tract's top 5: Minority status (145 tracts), Multi-unit housing (101), Uninsured (98), Housing cost burden (97) and Single-parent households (78). Housing cost burden means a household spends 30% or more of its income on housing.
Structural causes vs population traits. We split the 16 measures into two groups. Structural measures are things policy can change, such as poverty, unemployment, housing costs, missing high school diplomas, no health insurance, limited English (language access), housing type, crowding and no car. Population traits describe who lives there, such as age, disability, single-parent households and group living quarters. This split is a judgment call. At your request, Minority status is counted as a structural driver too. Structural measures make up about 70% of the score in the average tract.
Top structural driver in each tract. The biggest structural driver is Minority status in 55 tracts, Multi-unit housing in 49, Uninsured in 25, Unemployment in 24, Limited English in 18 and Housing cost burden in 15. Counting each tract's top 3 structural drivers, the most common are Minority status (127 tracts), Uninsured (91), Multi-unit housing (88), Housing cost burden (68) and Unemployment (66).
The most vulnerable tracts. 58 tracts have a score of 0.75 or higher. That is the top quarter nationally. In these tracts the top driver is Minority status in 15, Housing cost burden in 11, Uninsured in 9 and Unemployment in 8. Health insurance, housing costs and jobs show up again and again. Structural measures make up about 75% of their score.
North vs south. We split the county at its middle latitude. The south is more vulnerable on average (score 0.58 vs 0.49 in the north). The causes also differ. In the north, the top driver is most often Multi-unit housing (40 tracts), then Uninsured (18) and Limited English (18). In the south, it is most often Minority status (48 tracts), then Unemployment (16).
Why did I first call minority status a "marker" and not a cause? I first treated it as a sign of other causes rather than a cause itself. You pointed out that if it matters, it matters. That is fair. It is now listed simply as "Minority status," the same as the other structural drivers.
What did the regression test show? A regression is a math test that checks whether one measure still adds information once others are known. We left minority status out of the score for this test, so it was not being compared with itself. Once we know a tract's poverty, insurance and housing cost levels, minority share adds only a tiny bit of new information (it raises the share of the score explained by half a percentage point). With unemployment, education and car access added too, the extra is even smaller and no longer clear. That is because these measures overlap a lot. It does not prove that race does not matter.
Does that mean race cannot be part of the fix? No. Unfair treatment in lending, hiring and investment may be a root cause of the poverty, insurance and housing gaps. This data cannot test that, because it only counts conditions, not what caused them.
How big are the gaps? Rates are averaged by population.
| Measure | 75%+ minority tracts | Under 50% minority tracts | Most minority quarter of tracts | Least minority quarter of tracts |
|---|---|---|---|---|
| Below 150% of the poverty line | 27.9% | 10.6% | 27.0% | 9.0% |
| Uninsured | 17.0% | 5.0% | 16.3% | 4.5% |
| Housing cost burden | 37.9% | 19.7% | 37.7% | 18.6% |
So the poverty rate is about 2.6 times as high in mostly minority tracts, and the uninsured rate is about 3.4 times as high. Mostly minority tracts (75% or more) hold 56.5% of DeKalb's residents. But they hold 73.1% of the people below 150% of the poverty line and 72.6% of the uninsured.
Every tract's popup has an Explain this tract button. Tap it to get a plain-language explanation of that tract's own numbers. Here is what one popup means, line by line, for Census Tract 213.06 (FIPS code 13089021306).
The name and ID. A census tract is a small neighborhood area drawn by the Census Bureau. Its ID number (FIPS code) is 13089021306. The first 2 digits (13) mean Georgia. The next 3 (089) mean DeKalb County. The last 6 (021306) mean tract 213.06.
Overall vulnerability score (0-1): 0.582 (Moderate-high). This tract is more vulnerable than about 58% of all census tracts in the United States. Scores are split into four groups, called quartiles. A score between 0.50 and 0.75 puts it in the Moderate-high group.
Top 5 contributors. The score adds up 16 measures. If they all counted the same, each would be 6.25% of the score. Here the biggest one is Uninsured at 11.9%, about 1.9 times the average share. The top 5 are Uninsured (11.9%, people with no health insurance), Housing cost burden (11.0%, households that spend 30% or more of their income on housing), Limited English (10.9%, people who speak English less than well), Multi-unit housing (10.5%, homes in buildings with 10 or more units) and Crowding (9.2%, homes with more people than rooms).
Top 3 structural drivers (33.8% combined). Structural drivers are conditions that policy can change. Here they are Uninsured, Housing cost burden and Limited English. Together they make up 33.8% of the score. These are the policy levers: the places where action could lower vulnerability the most.
Structural share of the score: 86.0%. Structural measures make up 86.0% of this tract's score. That is higher than the DeKalb average of 70.5%. So more of the vulnerability here comes from things policy can change.
Measures that make up 25% of the score: Uninsured; Housing cost burden; Limited English. The first two reach 22.9% (11.9% plus 11.0%), so a third is needed. With Limited English, the total is 33.8%. So these three measures together make up a quarter of the score.
The rates. The last lines show the percent of residents who are a minority, below 150% of the poverty line, uninsured, and cost-burdened. The number after ± is the survey's margin of error. For example, ±5 means the true value is likely within 5 points either way.
Ask Grok link. Each popup also has an "Ask Grok about this tract" link. It opens Grok in a new tab with a question that includes that tract's numbers. The question is filled in for you, and those numbers go to Grok only if you press send.
This layer highlights tracts where lack of a high school diploma is a structural source of vulnerability. A tract is outlined in magenta when all three of these hold: no high school diploma is one of its top 3 structural drivers, its share of the score from that measure is in the top quarter of DeKalb tracts, and its overall score is above the county median (0.58). The share is used as a stand-in for the diploma rate, which is not in this data.
The blue outlines mark the tracts that contain the Dunwoody cluster schools: Austin, Vanderlyn, Dunwoody, Chesnut and Kingsley Elementary, Peachtree Middle, and Dunwoody High. These are tracts 212.10, 212.24 and 212.29. Two of them, 212.10 and 212.29, are among the least vulnerable tracts in the county, with scores of 0.04 and 0.01. The third, 212.24, scores 0.61, just above the median. None of the magenta tracts is in the cluster.