I'm not going to comment on the quality of the article. It's certainly not exhaustive research and is only a small insight into some sales numbers. But it does make a kind of intuitive sense.

Personally, if I play a game and see how its core game mechanics work then I am much less likely to purchase it.

A lot of my game purchases are impulsive — I see something cool in the trailer and want to try it. Once I've tried it I lose interest. Demos give me the opportunity to lose interest without purchasing the game.

When I know I really want to play a game all the way through I will always purchase it without trying the demo (if available). So in my limited gaming experience demos have never been a step to a sale.

I generally avoid buying games for the same reason. Few really advertise their core mechanics, and even the ones that do, I have no way to trust them on how well executed they are. For cheap games, I am willing to take the risk because even if the game is 'bad' I will still generally come out with a decent $/hr ratio. But the only times I get 'expansive' games are if I have seen a lot of people play it to the point where I have a good sense of how it works. Most games do not have a large enough network effect to get to that point.

What would be a good control for this then?

Dev budget? Marketing budget? Purchases by people who don't ever play demos? Purchases in areas the demo wasn't available? Analysis of whether studios always do demos, or only for certain games? Weighting by external review scores?

Any number of things you could take into account.