Mauboussin on Making Better Comparisons

Michael Mauboussin Guide to Making Better Comparisons (Link)

Comparing via analogies (steps and common mistakes)

1. Select the source for analogy
   • Common pitfall at this stage: Undersampling due to availability bias
2. Map the source to target to make inferences usually looking for similarities
   • A common pitfall at this stage: mistaking correlation for causality
     • Consider the first attempts at flight were people putting feathers on their arms, not studying lift
     • Confusing a star performer with talent in a volatile field
3. Adjust for differences between source and target
   • Tversky showed we place more emphasis on similarity than differences
   • The framing problem: We are influenced by which differences and similarities we are prodded to focus on (ie framing)
4. Learn by the success or failure of the analogy
   • Pitfalls in heuristics and intuition
     • Intuition works great in chess since it is a form of pattern recognition, but works poorly in investing where outcomes are non-linear
     • Recency bias influences sample
     • Choice-supportive or confirmation bias taps into our need to be consistent; we create stories after the fact to validate our decisions
     • Hyperbolic discount rates prevail when we study inter-temporal decision making
       • Stress is good response for crisis because it focuses on immediate needs; chronic stress causes decisions to be short-sighted
   • Using poor reference points in comparison
     • 2 companies in different industries can be more similar than their peers within their own industry; this shows how we can conflate attributes with what is actually driving value
     • Not recognizing that widely varied values can be justifiable. For example, 2 companies with the same earnings growth can trade at justifiably different valuations if underlying returns on invested capital are very different
     • Anchoring bias
     • When I went to Capital Camp, Mauboussin discussed T Theory. The top row of a category have more in common with each other than the average in the category. This articulates how I think about investors! Warren Buffet, Annie Duke, and Sam Hinkie have more in common with each other than other people in the same category.

So how to get better?

Instead of relying on analogies drawn from memory we can use “similarity-based” forecasting.

• Inputs
  1. A wide sample for the reference class rather than 1 or 2 examples from memory
     • Additional refinement by weighting the results of the most similar samples more heavily
       • Ways to quantify the similarity
         • “nearest neighbor” algorithm (requires identifying relevant axes for the dimensions)
         • “connectionist” technique for weighting features by similarities and differences
  2. A statistical “base rate” drawn from the outcomes of varied reference classes