Every election season, a handful of technical words escape into everyday conversation and lose their meaning on the way. Here is what six of them mean when the work behind them is done properly.
Booth
The smallest unit at which votes are counted - typically eight hundred to a thousand voters. Everything serious in campaign research eventually resolves to the booth, because that is where an election is physically won or lost. If advice cannot be expressed at booth level, it is a slogan, not a plan.
Sample
The set of people actually interviewed, standing in for everyone who was not. A good sample is a scale model of the constituency - by region, community, age, and past behaviour. Its size matters far less than its shape: a thousand well-chosen respondents beat ten thousand convenient ones every time.
Weighting
The correction applied when the sample's shape drifts from the constituency's. If young voters answered less often than they vote, their responses count proportionally more. Weighting is honest arithmetic, not manipulation - refusing to weight is the real distortion.
Swing
The shift in vote share between two elections. Swing is a description, not an explanation: a two-point swing can hide four points gained in one community and two lost in another. The useful question is never "how much swing" but "whose, and why".
Margin of error
The honesty attached to every survey number. A projection of 44% with a margin of three points means the truth plausibly lives anywhere between 41 and 47. Any research report that omits it is asking to be trusted rather than checked.
Turnout
The share of registered voters who actually vote - and the most underestimated variable in Indian elections. Preference decides which way a voter leans; turnout decides whether that lean becomes a vote. Campaigns that only measure preference are reading half the instrument panel.
"Precision in language is precision in thought. A team that uses these words loosely will read its data loosely too."
Field manual, Political Laboratory
