Generalizable and Robust TV Advertising Effects
Bradley T. Shapiro, Günter J. Hitsch, and Anna E. Tuchman, 2020, 20-100
When deciding how much to advertise, having a prior belief about the extent of likely advertising effectiveness is of utmost importance to managers. Yet published case studies may not be generalizable because they only consider positive and significant results.
To answer the question, “How much does advertising generally work?” Bradley Shapiro, Günter Hitsch, and Anna Tuchman study the effect of advertising on sales. Their intended result is a distribution that can serve as a benchmark to decision makers.
They pre-select 288 brands that are available in the Nielsen Scanner data (RMS) and in Nielsen advertising data (Ad Intel). They include all brands in the top 500 in RMS revenue where TV advertising and sales can be matched.
Importantly, because these brands are selected before estimating any effects and all are included, they overcome the problem of publication bias (non-representativeness in published case studies) by generating a full distribution that is not selected on outcome.
In order to estimate the causal impact of advertising on sales, they adjust for confounding factors using “baseline” and “border” identification strategies.
For the baseline strategy, they adjust for easy-to-target potential confounding factors -- such as seasonality, geography and long-run trends -- using fixed effects. They argue that residual variation is driven by factors outside of manufacturer control, such as television station slot constraints, bulk discounts to advertising agencies, and randomness in ad slot availability (e.g., due to a sporting event going long).
For the border strategy, they focus only on the borders of television markets, and use border-time fixed effects to make comparisons between otherwise similar consumers who happen to live just on opposite sides of TV market borders.
These two strategies use different sources of variation, so the comparison of them is important for robustness.
They allow advertising to affect demand for 52 weeks following the ad being aired and specify the advertising response function to be constant elasticity with respect to advertising stock. They include own and competitor prices as well as own and competitor advertising stock (In this way, they adjust for so-called “prisoner’s dilemma” concerns that if a firm stops advertising, a competing firm will steal its share. Thus, their estimates may be interpreted as the effect of own advertising, net of competitor advertising.).
They do sensitivity analysis around all modeling decisions, to help ensure robustness and generalizability.
They find a median, long-run advertising elasticity of about 0.014 and a mean long-run advertising elasticity of about 0.025. This effect of advertising on sales is considerably smaller than has been found in the published literature.
About two-thirds of their estimates are not statistically distinguishable from zero, a number which is not reflected in the published literature.
Their results are consistent with publication bias. Only considering positive and significant results censors almost 75% of their total estimates and pushes the median estimate to almost what is seen in the published literature.
Put into Practice
- Managers should be careful to generate their prior on advertising effectiveness from a distribution that is not selected on outcome. In particular, firms should be sensitive that negative results are often suppressed, leading to a true average effect that is smaller than what is published.
- While most brands advertise “too much”, for many brands some advertising is still better than nothing. For about half the brands in the sample of 288 brands analyzed, the observed level of advertising is more profitable than not advertising at all.
- Corporate: FREE
- Academic: FREE
- Subscribers: FREE
- Public: $18.00
3 WAYS to GET CONNECTED
Employees of MSI Member Companies enjoy the benefits of complete online access to content, member conferences and networking with the MSI community.
Qualified academics benefit from a relationship with MSI through access to msi.org, conferences and research opportunities.
The public is invited to enjoy partial access to msi.org content, a free e-newsletter, selected reports and more.