lecture 5 (1)

Question 1

Estimating the treatment effect

Answer 1

we can estimate the average treatment effect

assumption: parallel trends

If treatment group did not receive treatment, average would also have increased by beta1

This is crucial and untestable

Question 2

do advertisers care where their advertisement is answers

Answer 2

ad context matters to advertisers

they are willing to pay more for it

willing to pay for right placement

Intutively
Ad for makup conveys a different message when placed in a high end fashion magazine versus on the business page of a newspaper

Question 3

heterogeneity; chains and ratings: findings

Answer 3

Smaller effects for chains
Consistent with ads informing consumers and signalling quality

Larger effects for higher rated

Question 4

Heterogeneity; traffic and reviews findings

Answer 4

larger effects if higher pre-experiment traffic

Larger effects if larger number of existing reviews

–> consistent with a reminder effect of advertising

Question 5

Heterogeneity: age & price tier findings

Answer 5

Larger effects for younger restaurants
Larger effects “medium” price tier

Question 6

Who benefits more from digital advertising

Answer 6

Independent firms and higher quality firms and larger firms (in terms of traffic)

Question 7

Ghost ads presents an experimental framework to study causal effects of online display ads

Answer 7

Creates a log entry where an ad wins an auction but shows the second ad

I.e. a competitor in the ad auction

Which is better control group than serviec PSA ads

Online display ads are more effective when placed in context that better match the brand/product

the idea that context matters is relatively new