According to the paper, there's a decrease in demand by the primary home device for activities trackable by comscore. This is the data used (not technically an assmption). I do not see the paper claiming more than this.

The conclusions do not apply to aggregate demand from all device types. Only certain types of activities are tracked.

My personal opinion is that the inferiority of online time is due to the increase in use of and substitution to additional devices and apps that were not tracked in the dataset - and a higher income better affords such alternative devices, apps, and their data plans.