We study the effects of social media political advertising by randomizing subsets of 36,906 Facebook users and 25,925 Instagram users to have political ads removed from their news feeds for six weeks before the 2020 US presidential election. We show that most presidential ads were targeted toward parties’ own supporters and that fundraising ads were most common. On both Facebook and Instagram, we found no detectable effects of removing political ads on political knowledge, polarization, perceived legitimacy of the election, political participation (including campaign contributions), candidate favorability, and turnout. This was true overall and for both Democrats and Republicans separately.