Political economics predicts that the rich oppose redistribution and vote for conservative parties. Although this seemingly fits the data well in most countries, I show that the relationship breaks down when we control for unobservable characteristics. Using Norwegian survey data, I study to what extent voting is caused by income. Although a positive association between income and conservative voting persists when controlling for unobservables, the magnitude of the effect is reduced by a factor of five. To correct for measurement error, I instrument income with average income by profession. The magnitude of the coefficients becomes higher, but the main conclusion remains.