When researchers say a community gives more even after accounting for income, they mean a partial correlation: the co-movement that remains between two measures once a third, income, is held steady. It matters because raw giving rates rise with wealth, so a place that gives a lot may simply be a place with more money, and the partial correlation is the tool that tells the two apart. In the metro analysis behind these reports, the giving association with congregational density stays at about 0.46 with income held constant, the firmest of the three goods measured.
This explainer defines a measure, the partial correlation, in plain terms, and explains why it is the honest way to make a giving claim. It introduces no new finding of its own; the figures it uses are illustrative of the method and are attributed to the survey programs they come from.
What does a partial correlation actually hold constant?
A partial correlation measures how two things move together after the influence of a third is removed, and the third thing here is income. Raw giving rates rise with wealth: in the Census and AmeriCorps Civic Engagement and Volunteering Supplement, higher-income households give at higher rates, so any place with more money will look more generous before any adjustment. Holding income constant asks a narrower question: among places at the same income level, does giving still track with the thing being studied? The figure that comes back, between minus 1 and 1, is an association across cases, not a measured effect on any person.
Because giving climbs with income on its own, a claim that some other factor, congregational density, neighborliness, education, is associated with giving has to clear income first. Otherwise the claim might just be restating which places are wealthier. The partial correlation is what clears it.
Why does the distinction matter for a giving claim?
Without the adjustment, a giving claim risks measuring wealth twice. A raw correlation between, say, congregational density and giving could be high simply because denser-congregation metros also happen to be wealthier, and the giving would be tracking the income, not the congregations. Reporting the partial correlation of about 0.46, with income held constant, says something stronger and narrower: among metros at comparable income, giving still moves with congregational density. The number is smaller than the raw association, and that shrinkage is the point, it shows how much of the raw pattern income alone explained.
A partial correlation still does not prove a cause. It rules out one alternative explanation, income, but others remain (who chooses to live where, what else those places share). The honest reading is associational: the link survives an income adjustment, which is more than a raw rate can say, and less than an experiment could.
Methodology and limitations
This is an explainer about a method, not a report of an original finding. The illustrative giving-by-income figures motivate the adjustment and are drawn from the pattern in the Census and AmeriCorps Civic Engagement and Volunteering Supplement, 2023; they are shown to make the method legible, not as standalone measurements.
Method. A partial correlation between two measures, with a third held constant, is computed across cases (here, metros). It removes the linear association each of the two has with the control variable and reports the correlation of what remains, on a scale from minus 1 to 1. It is descriptive: it summarizes co-movement across cases, not an effect on any individual.
Limitations. Holding income constant addresses only income. A surviving partial correlation is consistent with several explanations and proves none; it does not establish causation, and it does not rule out controls that were not included. The underlying giving rates are survey self-reports with the recall and social-desirability caveats of any self-report, except where administrative records (such as IRS charitable-deduction totals) supplement them. The method is a tool for honesty about a giving claim, not a substitute for an experiment.
Conclusion
So what does it mean to say a place gives more even after income? It means a partial correlation: the part of the giving pattern that does not simply restate which places are wealthier. The adjustment shrinks the raw association on purpose, and what survives, about 0.46 in the metro work behind these reports, is the claim worth reporting.
The distinction is small and it is the whole game. A giving rate by itself can flatter a wealthy place; a partial correlation asks the harder question and reports the humbler answer. It is the discipline that lets a generosity claim be both interesting and honest, and it leaves the cause, as every correlation must, an open question.
Sources
- Indiana University Lilly Family School of Philanthropy, ongoing. Giving USA and the Philanthropy Panel Study. givingusa.org; philanthropy.iupui.edu. Established research institute and survey program.
- U.S. Census Bureau and AmeriCorps, 2023. Current Population Survey, Civic Engagement and Volunteering Supplement. census.gov; data.americorps.gov. Official statistic.