What lack of diversity has to say about estimates

What would be your guess for the weight of the following ox? Likely you’re not a livestock specialist, your guess may be way off, and you think that relying on a specialist would be the most accurate way to estimate its weight. I’d go a bit further and say that averaging out the specialists’ guesses with regular people’s ones would result in an estimate worse than the one from a specialist, right?!? Well, that may not be the case.

Photo by Macau Photo Agency on Unsplash

In his book “The Wisdom of Crowds”, James Surowiecki describes an experiment done in 1906 when averaging out all the 787 guesses resulted in a better weight estimate than the guesses of so-called specialists. The crowd average was only 1 pound off! Surowiecki states that “under the right circumstances, groups are remarkably intelligent, and are often smarter than the smartest people in them.”

The average of a large number of forecasts reliably outperforms the average individual forecast. However, that depends on how independent the participant’s opinions are, a statistics concept known as “signal independence”. Independence in this context means uncorrelated.

The more correlated people are, the less independent they are and then their opinion will be biased towards a specific value that may be different from the actual result. Diversity can play an important role to drive that independence.

Correlation can be originated from different sources. If people discuss a subject in advance, their opinions may not be independent during a meeting with other team members. If they always talk to the same people, even about different topics, their opinions may be correlated. If team members have the same background, that’s another source of correlation.

Diversity in a team brings a variety of ways to think, multiple cultural experiences, conflicting ideas to stimulate discussions, and a learning environment that is not possible if people are all alike. That reduces correlation. Diversity is also a key factor when running surveys that can support strategic business decisions. Including people from different races, gender, experience level, religion, sexual orientation, nationalities, and many other dimensions, when combined with a proper selection of a large sample, reduces biases and improves accuracy.

While diversity by itself does not guarantee independent opinions, neglecting it will most likely affect your organization’s capacity to come up with estimates that are closer to reality.


  • The Wisdom of Crowds, James Surowiecki, 2005
  • People Analytics, Coursera, Wharton University of Pennsylvania

You can also find other articles in my Medium or LinkedIn.

Leave a Reply