House Effect

A house effect is the systematic lean a pollster shows relative to the polling average — the consistent direction (D-leaning or R-leaning) of its results, after accounting for which races and weeks the pollster surveyed. A pollster with a +1.5 D house effect tends to show Democrats doing 1.5 points better than the polling average suggests they are.

House effect is not the same as bias. It can come from honest methodological choices — likely-voter screen design, weighting decisions, mode (live phone vs. online vs. IVR) — that systematically advantage one party’s voters by oversampling them or underweighting them less. Two pollsters with opposite house effects can both be methodologically defensible.

Aggregators like Silver Bulletin and Thin Gold’s polling average correct for measured house effects when computing trend lines. A poll showing R+5 from a pollster with a known +2 R house effect counts as roughly an R+3 reading, all else equal. The corrections are imperfect; an aggregated average is more reliable than any single poll, even one with no measurable house effect.