Now, we should be cautious about reading too much into these surveys. After all, we are talking about small overall movements that lie inside the margin of error for each poll. That consideration also applies when trying to analyze who shifted, given that margins of error are larger for subgroups within a survey’s overall sample. More broadly, surveys conducted this far out from the general election historically have had little predictive value. Additionally, third-party candidates often poll better farther away from Election Day, when the stakes of the election are lower — and millions of dollars in general election advertising have yet to be spent. And finally, not every third-party voter would vote for a Democrat or Republican if their preferred candidate didn’t run, so we can’t assume that, say, a Green Party voter would back a Democrat or a Libertarian would vote Republican.