Your model of the world is a continuously updated set of generalizations based on the things you see in real life or read about from the experiences of others. It couldn't be any other way, of course; you have eyes, you have ears, but you don't have some sort of sensory apparatus that lets you directly observe all of reality. So the single most important tool in your mental toolbox is to ask: to what extent are my observations driven by underlying reality, and to what extent are they driven by the subsets of reality I encounter every day?

That's a broad question, and it applies at every conceivable level:

• Your real-life observations about the range of human behavior will be artificially close to the middle of the bell curve; ridiculously high-functioning people are busy being high-functioning, so they're filtered out of many normal neighborhoods, companies, online hangouts, etc. Extremely dysfunctional people end up in prison or dead, so you don't encounter them much, either.

• Meanwhile, your observations of the media will focus too much on the tails of the distribution rather than the part that you interact with or are a part of. There aren't professional sports leagues that specifically aim for 60th percentile rather than 99.99th percentile athletes; there aren't business magazines covering moderately successful companies or slightly above-average investors (unless there's some other outlier trait that makes these interesting).

• The set of public companies you can analyze as an investor consists of whichever companies have a good enough business (or story) to go public, but don't have such favorable economics that they stay private so existing owners can benefit from the cash flows themselves...

• ... unless there's some other selection effect at work, like an industrywide tradition of high equity compensation, coupled with economies of scale and network effects so favorable that a buyout is infeasible.

• Your dating pool is limited to the people you're attracted to, who find you attractive. The more efficient the dating market, the more range restriction will create artificial negative correlations between traits that positively correlate in the real world. (On average, physically attractive, outgoing, high-income, and mentally stable all positively correlate to some degree, but almost everyone's experience is that, if they're in a position to choose, they're trading off some of these traits against others. A corollary to this is that the more efficient the market, the more rewarding it is to have idiosyncratic preferences.)

• It's very hard to tease out the difference between generations having different norms and mere aging. In just a few short decades, Millennials went from doing stereotypically twentysomething activities (weird fashion choices, complaining about how hard school or work are) to stereotypical fortysomething activities (conventional fashion choices, complaining about how hard it is to afford a house in a good school district and about how entitled the Zoomers at work are).

• The feedback you get on a business idea comes from friends, who generally want to be nice to you and aren't voting with their wallet. When they all say "Awesome idea, you should quit your job and build that," there's a chance that they're future anchor customers or investors, but a better chance that they're being encouraging.

• The news stories you read are selected based on your propensity to read them, and to keep reading other things from the same publication. Whether or not you're aware of them, you have biases that shape the kinds of stories that you see, so you should generally assume that even if your worldview is right, you'll be slightly overconfident in it.

• The fields you pursue because they're a way for mavericks to achieve extraordinary success will have more strivers and LARPers than you expected, because other people heard about that field, too, and joined for more boring reasons. (Then again, if you're genuinely interested and your competition is just out to make a buck or have a job that impresses other people, you'll probably outperform them over the long run. You will almost certainly underperform them when that field's at the peak of its popular interest, though. Amazon only went up about half as much as the Nasdaq in 1999, and that is in some ways psychologically harder to deal with than losing 90% of your money alongside your peers.)

We're always sorted on various traits, and the higher-impact the trait, the harder it is to ever have a random sample of what people with that trait are like. Your stereotype of what extremely smart people are like is strongly determined by which cohort of extremely smart people you hang out with—maybe they all know Ancient Greek, or they're all surgeons, or every single one of them loves arguing about compiler design, or they're universally obsessed with market microstructure. For people who have direct real-world experience talking to smart people who have a variety of interests, there's a good chance that a shockingly high percentage of them went to Stuyvesant or Yale or whatever with whoever is doing the observing.

In some ways, this is a liberating framework. It means that a sufficiently strict filter will overpower any preexisting biases. And if you've passed such a filter, there's a good chance that you genuinely meet the criteria and care a lot about what's being filtered for. Conversely, when a filter is biased, it can distort the perceived value of that bias in the opposite direction within that specific context. For example, if there's a hiring manager at your company who likes to hire from his alma mater, then having attended that school becomes a (slightly) negative signal within the company even if it's a positive signal in the aggregate.

One irony of all of this is that these filters are self-reinforcing: if you select into a particular peer group that's really excited about political activism, recreational math, vegan cooking, cryptography, or whatever, you get constant evidence from your daily life that these are universally interesting topics. And if you've selected into these groups, you'd probably rather live in a world where that perception is true! So being aware of selection effects has tradeoffs: you'll know more about the rest of the world if you periodically check yourself in that way, but a lot of what you learn makes the world a slightly less friendly place.

Read More in The Diff

This post was prompted by the realization that The Diff is constantly talking about selection effects in particular cases, but hasn’t done a breakdown of why they’re so important and ubiquitous. Some examples:

• Media selection effects mean we pay more attention to industries at their cyclical peak, and during the initial decline.

• This idea is closely related to looking for boring explanations before leaping to something more interesting and narrative-driven.

• They also help explain why companies don’t get as many complaints as they should when they buy a lavish headquarters.

• Selection effects also mean that CEOs tend to be ambiguous during upswings and cautious after downturns ($), because the CEOs whose personalities aren’t aligned with the cycle are more likely to get fired.

And note that, for a limited time, new Diff paid subscribers who choose an annual plan can get a free copy of Boom.

Join the discussion!

1  And also, the biggest returns come from the ideas people actively discourage, though the typical outcome for those is that your friends were right all along.