Internet Advertising with Information Congestion

I estimate a structural model of the demand for the display advertising on Korean Internet portals that accounts for the limited attention that website users devote to ad messages. The theoretical model assumes heterogeneity in the attention span distribution of users. I simulate the probability that a user pays attention to a message posted on the website. This, in turn allows me to identify the extent to which the information congestion affects the advertising demand. The estimated model enables me to predict the probability that a message is seen by a user of one particular website in a given month. These probabilities range from .36 to .67 by websites. The probability is decreasing in the intensity of the users’ activity on the website and is increasing in the age and the education level of website’s users. Male users are shown to have lower probability than females do. The results are compared with those given by a model where the information congestion would be ignored. The congestion being ignored, the platform’s market power turns out to be underestimated. I separately estimate the user demand with mixed-logit model and find that 92.17% of users have negative coefficients of the ad level on a platform. Finally, the estimated user-advertising elasticity is shown to be lower than the estimated attention-advertising elasticity.