The research I would like to see is things like carstereo.com vs car.stereo
Yes there is no .stereo tld but that is what I want to see - a researcher who owns both, identical site on both, posts links around web randomly choosing one or the other and see if there is a trend.
My guess is there won't be.
Without an experiment designed to be repeatable to test the hypothesis, you can cherry pick data to match whatever conclusion you want.
And the experiment should have same keywords in both domain names (car and stereo in above example)