Effects of TV advertising on keyword search

This paper investigates the possibility that television advertising influences online search using the AOL search dataset. It uses a novel keyword mining technique to classify keywords as brand related, category related (generic), or unrelated, distinguishing between category search and consumers' tendency to search a branded keyword. A three-level conditional choice model is estimated to determine whether hourly changes in brands' television advertising expenditures are related to deviations from baseline trends in search behaviors. The results indicate a statistically significant relationship between TV advertising and consumers' tendency to search branded keywords (e.g. “Fidelity”) rather than generic category-related keywords (e.g. “stocks”) in the dataset. The effect is largest for relatively young brands during standard business hours with an elasticity, .07, comparable to extant measurements of advertising's impact on sales. However, television advertising is not found to influence category search incidence and has limited effects on click-through rates.