One way studies in science are 'faked' is by p-hacking. It turns out if you wiggle variables around enough, you can often find statistically significant data that isn't really relevant, but sounds like it could viably be. 538 has an excellent (and highly relevant) p-hacking demo based on real world data. [1] You want to "prove" that this party or that being in power has a statistically significant effect on the economy. It almost certainly doesn't, but if you tweak your sample enough you can "prove" that it does, in a way that sounds completely justifiable.

So this study limited itself exclusively to people intending to buy an EV and within the next year. The fun thing about p-hacking is that it sounds reasonable. For instance on that page limit the data to Democrat presidents, employment, exclude recessions, and you can prove that having a Democrat president is good for the economy. And that data selection even sounds reasonable. But it's not.

Musk not being fond of the political establishment is not exactly breaking news. Yet Tesla continues to dominate EV sales worldwide and domestically to the point that sales figures are broken down into "Tesla" and then "everybody else." And as others have mentioned Toyota dominating in this sample is quite indicative of the fact that they ended up surveying something besides EV favorability.

[1] - https://projects.fivethirtyeight.com/p-hacking/