Texas Data Visualizer

Analysis runs on county, tract, block group, ZIP, city, and metro data. Pick one of those geographies in the Layers tab.

Every tool here uses all areas in Texas, not just what's on screen, for the period chosen in the Layers tab.

I want to…

None selected· Shift-click areas to select

Build an indexCombine measures into one score

Combine using

↑ means a higher value raises the score; ↓ lowers it. Sliders set relative weight.

How to read this

An index turns several measures into one 0–100 score so you can rank places. Each ingredient is put on a common scale, flipped if lower is "worse," weighted, and averaged.

Percentile ranks: each area's position among all Texas areas. Easy to explain; ignores how far apart values are.
Z-scores: distance from the Texas average in standard deviations. Keeps magnitude, but extreme areas pull harder.
Min–max: rescales each measure to 0–1. Sensitive to the single highest and lowest area.
First principal component: lets the data pick weights from how the measures move together. Objective, but harder to explain.

Stability reports how many top-10% areas stay in the top 10% when every weight is randomly shifted up to ±50%. Above 80% means the result doesn't hinge on your exact weights; below 60% means it does.

Create a variableWrite a formula from existing measures

How to read this

Write math with measure ids. A few useful patterns:

pct_poverty - pct_poverty@2014 change in poverty rate (percentage points) since 2010–14
pop / pop@2014 * 100 - 100 percent population growth
dens(jobs) jobs per square mile
rank(med_income) income percentile among Texas areas
pct_renter - lag(pct_renter) how different an area is from its neighbors

Change is measured on the same 2020 boundaries for every period, so it compares like with like. Older periods are estimated onto today's boundaries by area.

RelationshipsScatterplot, correlation, regression

Log X Log Y

How to read this

r (Pearson) measures how closely the dots follow a straight line: 0 is none, ±0.3 weak, ±0.5 moderate, ±0.7 strong. The sign gives direction. ρ (Spearman) does the same using ranks, so it's robust to outliers and curves. If ρ is much bigger than r, try a log axis.

R² is the share of the variation in one measure that a straight line through the other accounts for.

These are relationships between places, not people. A strong tract-level link doesn't mean the same people have both traits (the ecological fallacy), and results can change at a different geography.

Regression outcome Predictors Cmd/Ctrl-click for several Robust standard errors (HC1) Weight by population

How to read this

Regression estimates how the outcome changes with each predictor while holding the other predictors constant.

Coef.: change in the outcome for a one-unit increase in the predictor (for shares, one percentage point).
p: how surprising the estimate would be if the true effect were zero. Under 0.05 is the usual bar, but with thousands of areas almost everything clears it, so focus on size.
Std. β: effect of a one-standard-deviation change, so you can compare predictors measured in different units.
VIF: above 5, predictors overlap so much that their separate effects are unreliable.
Robust SE guards against uneven scatter. Weight by population stops tiny areas from counting as much as big ones.

After running, check Moran's I of residuals. If errors cluster on the map, the model is missing something about place, and the p-values are too optimistic.

Spatial patternsMoran's I, clusters, hot spots, spatial lag

How to read this

Spatial tools ask whether where a value is matters. Each area is compared with its nearest neighbors (pick 4 to 12; more neighbors means smoother, broader patterns).

Global Moran's I: one number for all of Texas. Near 0, values are scattered randomly. Positive means similar values sit together (0.3 is clear clustering, 0.6+ very strong). Negative, rare in Census data, means a checkerboard.
Clusters (LISA): maps which areas drive the pattern. High-high and low-low are clusters; "high among low" and "low among high" are outliers that stand apart from their neighbors.
Hot spots (Gi*): asks whether an area and its neighbors together run high or low. Better at finding broad regions; LISA is better at spotting outliers.
Spatial lag: each area's neighborhood average. Useful for smoothing noisy small-area estimates.

Significance is not adjusted for testing thousands of areas, so some flagged areas will be chance. Trust large, consistent clusters over single areas.

Summarize around a point 10 km

How to read this

Draws a circle and recombines every area whose center falls inside. Shares are rebuilt from counts (for example, all renters who are cost-burdened divided by all renters), not averaged, so big and small areas count correctly. The gray figure is Texas as a whole for comparison.

RankingsTop and bottom places for the current measure

Min. population

Leave out college-student areas (median age under 24)

How to read this

Lists the highest and lowest areas for whatever the map is showing, in the period chosen in the Layers tab (or the change between periods in Change mode). Click a name to fly there.

The population floor keeps tiny areas, whose survey estimates are the noisiest, from crowding the extremes. The summary line shows the typical value, the middle half of areas, and how far apart the top and bottom tenths are.

Charts and dataChart builder, exports, side-by-side

Export data

About this map