Donald Trump has seen a recent surge in support. Was this driven by advertising?

Last month, I suggested that Donald Trump’s campaign might have an ace in the hole. It was losing but hadn’t spent much money yet, and it might be holding off on advertisements until late in the election, when such spending might actually be useful.

The Trump campaign followed through with that tactic recently, spending massively across several competitive states. It may have even caught the Clinton campaign off guard, dominating spending in states like Virginia and Colorado. Those traditional swing states had seemed out of Trump’s reach but have become much closer recently. Was it the spending that did it?

To test this, I used the spending-per-electoral-vote figures reported by Ken Goldstein, John McCormick, and Andre Tartar for 14 states. For each state, I calculated Trump’s percentage of the money spent by both Trump and Hillary Clinton for the week of October 25. I then compared that with Trump’s vote surge in the same states between October 25 and November 6, allowing a week for polls to capture changes in voter sentiment. These polling figures come from the aggregate scores reported by FiveThirtyEight. A scatter plot appears below.

The big takeaway from this graph is that the trend line is nearly flat. Trump saw an average increase in his polling position of 2.6 percentage points across all these states. However, he did about as well in the states where he dominated spending (Colorado and Virginia, notably) as those where the sole spending was by Clinton (Arizona and Texas, for example). Trump’s biggest boost was in New Hampshire, where Clinton spent slightly more.

What this suggests is that whatever caused Trump’s support to increase during this time period was likely tied to national events rather than local spending. This could have been a result of the FBI’s renewed interest in Clinton’s emails, an inevitable decline in support for her after her debate bump wore off, or Gary Johnson’s support bleeding to Trump as part of a typical third-party candidate decline as the election approaches.

That said, this does not mean that local ad spending can’t make a difference in a race. Indeed, considerable evidence suggests that an advertising advantage can move the polls toward a candidate, if only temporarily.

In that light, it seems the Trump team was playing it smart while the Clinton team was perhaps becoming overconfident. Notably, the Trump team was spending heavily in states where it was trailing in the polls but that it would need should the national race become more competitive. Clinton, rather than shoring up those traditional swing states, was pressing an ephemeral advantage in Arizona and Texas, where Democrats almost never win. That said, she made up for many of those swing state advertising deficits the following week with new spending.

There’s little evidence here or elsewhere that these tactical decisions are changing the outcome of the election. But it suggests a surprising degree of competence from a campaign that hasn’t demonstrated much of it this year, and a surprising lapse from an otherwise very professional campaign.

(Thanks to John Zaller and Lynn Vavreck for suggestions and feedback on this piece.)