Cite as: 596 U. S. ____ (2022) 17 Opinion of the Court ance of corruption. Essentially all the Government’s evidence, however, concerns the sort of “corruption,” loosely conceived, that we have repeatedly explained is not legitimately regulated under the First Amendment. The academic article—cited for various propositions by both sides—concludes that “indebted politicians” are “more likely to switch their votes” if they receive contributions from the banking or insurance industries. Ovtchinnikov, Debt 31. But the authors explicitly note that they cannot distinguish between voting pattern changes traceable to legitimate donor influence or access, and voting pattern changes as part of an illicit quid pro quo. See A. Ovtchinnikov & P. Valta, Self-Funding of Political Campaigns, Management Science, Articles in Advance 18 (April 7, 2022) (Ovtchinnikov, Self-Funding). As noted, our precedents demand adherence to that distinction. See, e.g., McCutcheon, 572 U. S., at 209. The authors also state that their analysis is merely a “first step” in understanding whether politicians’ self-funding decisions impact voting behavior, because they cannot “pin down a causal link” yet. Ovtchinnikov, Self-Funding 21. The online poll the Government asks us to consider similarly misses the mark. The poll, conducted at the Government’s behest for this litigation, reports that most respondents thought it “very likely” or “likely” that a person who “donate[s] money to a candidate’s campaign after the election expect[s] a political favor in return.” App. 351–352. But it failed to ask whether those same respondents thought it likely that donors who contribute to a campaign before the election also are likely to expect political favors in return. Nor did the poll mention that the individual base limits still apply to such contributions. And it failed to define the term “political favor,” leaving unclear the critical issue whether the respondents associated such contributions with the direct exchange of money for official acts,