Cohorts

What does a slice of the country think, and is the slice big enough to say?

Why this screen exists

Every cross-tab a broadcaster asks for is a cohort. The failure mode is that cutting a sample four ways produces a cell of forty people, and a number gets read on air as though it means something. This builder computes the effective sample after weighting, not the raw count, and suppresses the estimate below a floor rather than printing it small.

Two dimensions are locked. Caste is not fielded, so there is no data to cut on. And any combination that would isolate fewer than a hundred effective respondents is refused, because at that point a cohort stops being a segment and starts being a description of identifiable people.

Build a cohort

Region

Age band

Settlement

Education

Gender

Community

Neither is asked, stored or inferable. The chips are shown rather than hidden so a buyer sees the constraint instead of assuming an oversight.

All adults, nationalAbove floor

Nominal n

2,14,600

Effective n

1,38,846

Margin of error

±0.3

Design effect

1.55

Voting intention, weighted

Incumbent alliance 41.2%Opposition 36.8%Other and undecided 22.0%

Interval is the sampling component only. Total survey error is wider. Refusals are reported as refusals and no preference is imputed for them.

First-mentioned concern

1. Employment and daily wages · 26.4%
2. Price of essentials · 19.8%
3. Law and order · 14.1%

Available as

The third is disabled at build time on every cohort, at every size.