Linear Regression models can be causal only if we control for all factors

To prove causality in Linear Regression models, one has to build more regression models

Linear Regression models can be causal only if we control for all factors.

In Marketing Mix Modeling, we inevitably face the question - 'Is there any causal relationship between Marketing spends and Sales?'

To answer causality, one has to build more regression models.

This becomes evident in following two cases:

1) Addressing Endogeneity by 2SLS model.

2) Using Frisch-Waugh-Lovell (FWL) theorem when we want to prove causality between only 2 variables of interest while also controlling for other variables.

Overall an interesting pattern that I have observed is that, to prove causality in Linear Regression models (including MMMs), one has to build more regression models.

I will expand on FWL theorem and how we use it in MMM in future posts. Stay tuned.

Resources:

• Check out our causal inference work in MMM here https://www.arymalabs.com/proving-efficacy-of-mmm-through-difference-in-difference-did/

• Event Studies : https://arymalabs.substack.com/p/event-studies-and-its-utility-in?utm_source=post-email-title&publication_id=1880035&post_id=145174120&utm_campaign=email-post-title&isFreemail=true&r=1cg11&triedRedirect=true&utm_medium=email

  ---

Thanks for reading.

For consulting and help with MMM implementation, Click here

Stay tuned for more articles on MMM.