Should you disclose ad spending?

Happy with the best paper award at this weekend's American Marketing Association conference, here is a one-page executive summary of the key insights.

Since 1994, the US Securities and Exchange Commissions (SEC) no longer requires public firms to disclose their ad spending, and two thirds have chosen not to:

For instance, Barnes & Noble did not disclose ad spending between 1997 and 2006, while Apple did not disclose in 2016? Why? Likely because their advertising was ineffective, and increased versus decreased their firm's systematic risk.

How did we reach this conclusion? We use a machine learning technique called “adversarial validation” to tackle the systematic differences between voluntary disclosing and non-disclosing firms. Next, we build a machine learning ensemble of 200 models consisting of eXtreme Gradient Boosting (XGBoost) machines, random forest, generalized linear model (GLM), gradient boosting machine (GBM), and deep neural nets to predict values of undisclosed advertising.

The results reveal that firms who disclose (vs not disclose) advertising, enjoy:

1) higher ad effectiveness on customer-based brand equity metrics;

2) higher firm value impact of advertising;

3) lower systematic risk thanks to advertising.

For market research, we show how to combine the strengths, and overcome the weaknesses, of the 2 dominant ad spending information source: Compustat and Kantar media.

For regulators, we show that non-disclosure biases the empirical generalizations about advertising effectiveness. Because non-disclosing firms have systematically lower ad effectivenes, research based on Compustat data:

1) overestimates advertising elasticity on sales;

2) overestimates advertising effect on firm value;

3) overestimates ad effects on reducing firm systematic risk.

Interested in the full paper? Download it here on SSRN: