Read on Twitter
I've been thinking a lot today about how much advertising effects behavior, but it has ended up in different threads. Organizing it here:
It started with this picture from a recent #polisci paper on the effects of political ads (shown as part of a survey) on attitudes. https://twitter.com/eleafeit/status/1301508168485396480
As Garrett points out, there is a lot of literature in marketing that shows small effects sizes of advertising. https://twitter.com/garjoh_canuck/status/1301720304859983874
As Avi points out, there are subtle selection effects in the review process for papers on advertising. Often the data comes from companies who have disincentives to publish null effects. https://twitter.com/avicgoldfarb/status/1301849811457437697
Or it could just be sampling variation, as Brad points out: https://twitter.com/btshapir/status/1301862463676260353
What I'm really interested in is the wide range of effects we see when we look at well-controlled studies. Sure, I'd expect effects to vary across media or across categories, but even within the same retailer, you sometimes see a wide range of positive and negative effects.
Ads are a funny thing and we really don't understand them very well yet.
But if you want more of an "ad effects are small" perspective, Garrett posted this really nice piece https://twitter.com/garjoh_canuck/status/1301890901376208898
And getting back to the original graph I posted, Arman read the paper closely and points out that the definition of "ad treatment" is a little murky. https://twitter.com/StableMarkets/status/1301524016650485760
And, finally, Ron points out that small effect != negative ROI (especially when you are talking about elections). https://twitter.com/marketsensei/status/1301512471199592451
~*~ Fin ~*~
Oh, and here is one more plot of advertising effects. I told you I couldn't get enough of them. 
https://twitter.com/garjoh_canuck/status/1194291531932418048
https://twitter.com/garjoh_canuck/status/1194291531932418048
