Voters in U.S. elections receive markedly different representation depending on which candidate they elect, and because of incumbent advantages, the effects of this choice persist for many years. What are the long-term consequences of these two phenomena? Combining electoral and legislative roll-call data in a dynamic regression discontinuity design, this study assesses the long-term consequences of election results for representation. Across the U.S. House, the U.S. Senate and state legislatures, the effects of ‘coin-flip’ elections persist for at least a decade in all settings, and for as long as three decades in some. Further results suggest that elected officials do not adapt their roll-call voting to their districts’ preferences over time, and that voters do not systematically respond by replacing incumbents.