The exact formula
For each selected measure, take the minimum and maximum point estimates among the 50 states. For “higher scores more,” compute 100 × (value − minimum) ÷ (maximum − minimum). For “lower scores more,” subtract that result from 100. A constant measure receives 50 points for every state.
Divide each slider value by the total of all slider values to get its effective weight. Multiply each normalized score by its effective weight, then add the contributions. This produces a 0–100 score. Zero weight excludes a measure. With all weights at zero, no ranking is calculated. Ties share a rank using the unrounded score; the table rounds scores to two decimals.
What a composite leaves out
These six measures are a limited selection, not a complete account of a state. Min-max scores depend on the observed extremes: an outlier changes the spacing of everyone else. Excluding DC avoids comparing a city-like jurisdiction against the state extremes.
Related measures can count the same underlying pattern more than once. We flag selected pairs whose absolute Pearson correlation across state point estimates exceeds 0.8. That warning is descriptive and does not prove causation. You can reduce or zero one weight.
Source uncertainty is not propagated into a composite confidence interval. A two-decimal score is arithmetic precision, not evidence that two states differ meaningfully. Changing weights can change the order; try a single-measure emphasis and equal weights to inspect sensitivity. We do not publish our own overall “best states” ranking.