Only 16% of leaders let parents shape what gets taught. Where they do, reported parent engagement is twelve points higher — the largest positive gap we measured.
Ministry leaders talk constantly about parent partnership. We wanted to know how many of them hand parents any actual influence — not a newsletter, not an invitation, but a say in what their kids get taught.
Sixteen percent. And in that sixteen percent, parents are described as engaged far more often than anywhere else in the data.
| Outcome | Parents have input vs They don't | Raw | Adjusted | Signif. | Direction |
|---|---|---|---|---|---|
| Parents engaged with the ministry | 57.7%vs 45.9% | +11.7% | +11.8% | z = +4.1 | higher |
| Parents described as not engaged | 8.8%vs 15.9% | −7.1% | −6.9% | z = -3.4 | lower |
Adjusted pools the difference inside strata of church size and volunteer status, because both drive many answers at once. A claim only appears on this site if the adjusted figure keeps its direction and clears two standard errors (|z| ≥ 2). Group sizes: 385 and 2,014.
The direction of cause is genuinely unclear here and we won't pretend otherwise: engaged parents are easier to invite in, and being invited in probably produces engagement. One thing this is not about is event volume — these leaders host parent touchpoints no more often than anyone else. Influence and activity are separate things, and only one of them tracks with engagement.
Based on 2,399 survey responses. This page compares parents have input (385 respondents, 16.0% of the sample) against they don't (2,014).
Church size, tenure and paid-versus-volunteer status jointly drive a general tendency toward more structured ministry, which touches most practice questions in this survey. Raw comparisons therefore overstate almost everything. Each claim here is re-tested inside strata of church size (q2) x volunteer status (q1) and only published if it survives — see what didn't survive.
Percentages are shares of each group who selected that answer. Every figure on this page is generated from the dataset at build time, and a verification script re-derives each claim independently. Check our work in the explorer →