I examine the effect of marriage penalties in the US income tax system on marital status. I construct a simulated instrument that exploits variation in the tax code over time and between US states to deal with potential endogeneity between the marriage penalty a couple faces and their marital status. I find that a $1000 increase in the marriage penalty faced reduces the probability of marriage by 1.7 percentage points, an effect four times larger than previously estimated. Those in the lowest education groups respond by as much as 2.7 percentage points, with the average response declining as education increases.