If this is your first time reading, please check out the overview for Poor Ash’s Almanack, a free, vertically-integrated resource including a latticework of mental models, reviews/notes/analysis on books, guided learning journeys, and more.

Executive Summary Of Base Rates + Inside/Outside View Mental Model:

If you only have three minutes, this introductory section will get you up to speed on the base rates + inside / outside view mental model.

The concept in one sentence:

the statistical frequency of something over a large sample size is known as the “base rate,” and it serves as a useful starting point on which to anchor our evaluations.

A formal definition: Michael Mauboussin, in “ The Success Equation” , defines the base rate with the following example: “If 85% of the taxicabs in a city are green, then 85% is the base rate.  Absent any other information, you can assume that whenever you see a taxicab[,] there’s an 85% chance that it will be green.”
Key takeaways/applications: when evaluating situations in our lives or work, we have a tendency to focus only on the factors we see – the “inside view” – which often excludes important information thanks to selective perception; we can improve our probabilistic outlooks if we incorporate the “outside view” by basing our analysis off the “base rate.”

Three brief examples of base rates:

Taking Munger literally.  Base rates can help us apply a margin of safety to our recreational activities: one of the best pieces of advice in Laurence Gonzales’ “Deep Survival” is a recommendation to read published accident reports and statistics for your specific hobby (climbing, surfing, skiing, etc) to understand how and when dangerous situations can crop up.  Gonzales notes the example of some climbers who failed to take into account the high likelihood of lightning on mountain peaks circa 3 PM on summer afternoons, often despite previously clear/beautiful weather, to deadly consequences.  It’s a literal interpretation of Munger’s advice:

“All I want to know is where I’m going to die so I’ll never go there.”

Where do businesses die? Statistical blindness is why so many strategies fail; they ignore the “base rate” of what, on average, actually sinks a company, a topic three-time Turnaround of the Year award winner Jeff Sands explores in Corporate Turnaround Artistry when identifying 23 common “Business Killers” that he sees leading to business distress.

Don’t forget the missing planes.  A knowledge of base rates helps develop understanding of counterfactuals and prior probabilities, allowing us to be a filter, not a sponge and separate good advice from bad – “signal” from “noise” – when we encounter new information.  Phil Rosenzweig points this out compellingly in “The Halo Effect” , explaining how many popular business case studies fail to take survivorship bias into account.

If this sounds interesting/applicable in your life, keep reading for deeper understanding and unexpected applications.
However, if this doesn’t sound like something you need to learn right now, no worries!  There’s plenty of other content on Poor Ash’s Almanack that might suit your needs. Instead, consider checking out our discussion of the n-order impacts,compounding, or sunk costs mental models, or our reviews of great books like “Zero To One” by Peter Thiel , “ Ordinary Men” by Christopher Browning , or “ Internal Time” by Till Roenneberg .

A Deeper Look At The Base Rates + Inside/Outside View Mental Model

“Unrealistic optimism is at its most extreme in the context of marriage.  In recent studies, for example, people have been shown to have an accurate sense of the likelihood that other people will get divorced (about 50 percent).  
But recall the fact that they have an absurdly optimistic sense of the likelihood that they themselves will get divorced.  It’s worth repeating the key finding: nearly 100 percent of people believe that they are certain or almost certain not to get divorced!
It is in these circumstances, and in part for that reason, that people are immensely reluctant to make prenuptial agreements[,] believing that divorce is unlikely, and fearing that such agreements will spoil the mood…”

That quote, from Cass Sunstein and Richard Thaler on page 226 of “Nudge,” powerfully illustrates how the “inside view” – i.e., the story we construct based on the evidence we see – can diverge wildly from the “outside view,” i.e. the statistical “base rate” that applies to the situation we find ourselves in.

The obliviousness of newlyweds is a clear case of overoptimism: of all couples that estimate there’s only a 10% chance they will get divorced, there’s actually a 50% chance they will end up divorced.  (Depressingly, we don’t learn: the statistics for repeat marriages are not much better, and maybe even worse.)

Base Rates + Inside/Outside View x Salience / Vividness x Margin of Safety x Disaggregation

Base rates can be helpful because they’re cold, hard data that counter  salience / vividness bias.  While that’s a mental model worth internalizing thoroughly, the short version is that we perceive danger with emotion, not reason.

We often tend to overestimate “scary” risks (like murder, or Ebola) while remaining blase about the real killers.  Here’s data sourced from the CDC and other credible entities on the base rates of causes of death:

U.S. Deaths from Heart Disease (2016)	~ 630,000
U.S. Deaths from Suicide (2016)	~ 43,000
U.S. Deaths from Car Crashes (2016)	~ 37,000
U.S. Deaths from Flu (average year)	~27,500
U.S. Deaths from Homicide (2016)	~ 17,000
U.S. Deaths from Airline Crashes (CUMULATIVE, Feb 2009 – June 2018)	46
U.S. Deaths from Ebola (2014)	2

Sources: CDC.gov, CDC.gov, NHTSA.gov

Of course, as we’ll get into, it’s important to use the “right” base rate – for example, your likelihood of dying of the flu is meaningfully higher if you’re elderly or have a compromised immune system. 

Nonetheless, focusing on “base rates” can help us make more appropriate decisions about what risks are worth worrying about and mitigating, or which opportunities are worth pursuing and which are long shots without enough payoff to have an expected value higher than the  opportunity cost.

As usual, this is often discussed in a probabilistic business/prediction/statistical analysis context, as in Mauboussin’s “The Success Equation” or Tetlock’s “ Superforecasting” , which are excellent books that thoroughly explore those angles.  (Ellenberg’s “How Not To Be Wrong” focuses on it less, but it’s still included.)  So, as usual, I’ll present a more novel take here. Let’s get a little more technical: what is the appropriate base rate to use?

Since it’s scary and salient and I’m going to bore you to death if I talk about coronary artery disease and the dietary steps that can reduce your risk, let’s go back to our lightning example.  How likely is it that you will be struck by lightning?  You could, for example, start with the number of people who are struck by lightning in the U.S. every year – NOAA data seems to estimate this at circa-500 (with about 10% of those strikes fatal, and the rest generally resulting in long-term neurological damage, which is also bad).  
So, on any given day, 500 ÷ 365 = ~1.5 people are struck by lightning.  There are 300 million people in America, so your chance of being struck on any given day is 1-in-50-million… i.e., too small to worry about.
But that’s a profoundly unhelpful base rate: the number of people who are struck by lightning in a properly grounded building or a car with the windows closed probably approximates zero.  NOAA data indicates – unsurprisingly – that everyone killed by lightning was outside; it’s also not a stretch to imagine that there was also a storm (they weren’t struck from clear blue sky).  So, our denominator is wrong – it shouldn’t be “all Americans,” but rather “all Americans who happen to be outside during a thunderstorm.”

Philip Tetlock discusses this topic in “ Superforecasting” , explaining that the eponymous “superforecasters” integrate outside-view, base-rate thinking into their forecasting process, with an emphasis on using the right base rate via a process of disaggregation.

Given how few people are fatally struck by lightning, the sample size is too small to really conclude a lot of useful further lessons from looking at this sort of data.   Inversion, on the other hand, can be helpful: let’s flip the conditional probability around.  Rather than focusing on where people were, given that they were struck by lightning, what if we focused on where lightning strikes – or, even more specifically, where lightning strikes and at what times of day?

That’s exactly the sort of analysis Gonzales touches on in various places in “ Deep Survival” – it turns out that afternoon lightning is quite common high up on mountains in the summer.  Of course, very few people out of all total Americans are up on the tops of fourteeners during summer afternoon storms, so not very many people die – but if you’re planning to be there at that location, season, and time of day, your “base rate” of being struck is orders of magnitude higher than it would be generally.

It is difficult, of course, to come up with a precise estimation, but it doesn’t matter because this is a precision vs. accuracy type situation: any activity that gives me a relatively meaningful chance of dying or sustaining permanent, irreversible injury, with a low opportunity cost delta to my next-best alternative (going hiking at another time in the year when storms are rare, or in another location where, again, storms are rare), would seem to be a pretty easy call.

If you conduct this sort of probabilistic analysis, it does become clear that a high-altitude backpacking trip this summer in the Colorado Rockies (which I was originally planning) is probably not a good idea.

So, a key takeaway is that while precise and quantified base rates can be helpful, they’re not absolutely necessary.  I like/love this quote from a colonial governor in John Lewis Gaddis’s 300-page masterpiece “On Grand Strategy”:

“I imagined, like most young beginners, that… I should be able to make a mighty change in the face of affairs, but a little experience of the people… has absolutely cur[e]d me of this mistake.”

See also, of course, Buffett/Munger’s commentary on their poor track record of changing people’s minds.  You don’t need to know exactly how unlikely it is that you’ll be able to change someone’s mind to know that it’s not the best approach to rely upon.
Base Rates + Inside/Outside View x Humans vs. Econs = Planner-Doer / Hyperbolic Discounting

Like Richard Thaler, I steal good ideas when I find them. In Misbehaving, Thaler presents a hilarious story about cashews at a dinner party that violates classical “rational-actor” economic theory and demonstrates that we’re humans, not econs.  (I won’t spoil the story here – you gotta read the book for that.)

Thaler extends this theme into the idea of hyperbolic discounting, presenting a “planner-doer”model that applies the idea of local vs. global optimization to our current and future decisions.  This is obviously important in the context of life decisions (as Thaler and Sunstein note above), but it’s important in more trivial contexts too.

Have you ever shopped at Costco while you were hungry… and come home bearing fifteen five-pound bags and 48-ounce boxes of various goodies, which then proceeded to sit in your pantry and freezer for a year before you were able to get over the sunk costs, admit you were never going eat them, and throw them away?  Well, that’s an example of the hot-cold empathy gap – our tendency to consistently overestimate (when we’re hungry) how much food we will want when we’re noth ungry, a sort of recency bias schema bottleneck, if you will.

As with many things, admitting we have a problem is half the solution; knowing that we have a tendency to overpurchase at Costco – a “base rate,” or “outside view,” – we can therefore adjust our “inside view” (what we want now).  Yeah, that giant bag of veggie straws may look tempting*, but so did that value-pack of croutons that’s still going stale…

*Said no one ever.

Jokes aside: in “The Happiness Advantage” – a serious psychology book by Shawn Achor that drives home the power of agency – Achor implicitly discusses base rates as an underlying component of behavioral modification.  If we notice that we’re consistently not achieving our goals, that may be a sign that either our goals need to be changed, or we need to reduce the activation energy between us and our goals.

Application/Impact: from contexts ranging from trivial (learning guitar) to vitally important (starting a business, choosing a spouse), knowing the appropriate “base rate” – whether that is the statistical likelihood that someone in our position will succeed/fail, or our own likelihood based on historical experience of actually doing something we plan to do – can help us make better decisions based on more realistic information than the limited amount we can immediately see.  If you want to explore this in more detail, check out the Bayesian reasoning / “be a filter, not a sponge” mental model.

Base Rates: The AI Angle

Tools like Perplexity make base rates accessible; simply type in a natural language query, and Perplexity will return sources with statistics, allowing us to instantaneously access the “outside view” on nearly any topic.

Interestingly, large language models can be thought of as massive base rate engines. They operate on a massive “outside view” of training data, calculating the statistical probability of the next word in a sequence.

This “massive base rate,” however, is a powerful starting point, not an ending point. If you ask AI for investment ideas, it will likely return the sum of all consensus – a reflection of the “average” view found in its training data, rather than a truly differentiated or novel insight.

This limitation is fundamental. LLMs don’t think in the human sense; they don’t possess judgment, consciousness, or the ability to generate truly novel insights. They are masters of statistical mimicry, assembling plausible-sounding text based on patterns they’ve “seen” before.

Moreover, LLMs are inherently biased toward their training data. They don’t know “truth”; they only know correlation. Just as a human might believe something is true simply because they’ve heard it repeated many times, an LLM’s “base rate” reflects all of our collective biases, inaccuracies, and consensus fallacies as captured in the text it was trained on. Therefore, we must never accept their outputs uncritically.

This is where the Dunning-Kruger effect creates a dangerous trap for the user. We are easily swayed by the sophisticated polish of an AI’s answer. The confident tone, perfect formatting, and pithy quotes create an aura of authority. But this is a classic example of the product vs packaging mental model. Polished formatting can easily hide a complete lack of underlying substance or, worse, “hallucinate” facts that are dangerously incorrect.

We should create our own A/B tests to spot this. I once ran a research report on a company I knew intimately using a supposedly reliable AI tool; the output was riddled with fundamental errors so basic that it was almost unbelievable (the AI tool completely misunderstood what part of the industry the company played in – as an analogy, it might have believed that a company was selling ICE engine components instead of electric motors).

This experience highlights the critical danger: if the AI can be so confidently wrong about a subject where I know the ground truth, what happens when I use it on a new company where I don’t? I’d have no filter to catch the plausible-sounding (but entirely false) information. That’s a useful base rate to keep in mind when using AI as a productivity enhancer.

Application/Impact: Treat an AI as an engine for accessing the starting base rate or the broad consensus. It’s an accelerator for gathering the “outside view.” But it is not a substitute for the hard work of building your own “inside view” through critical thinking, independent verification, and genuine judgment.

Base Rates x (Marginal) Utility

I’ll keep this last one short, since we’ve already covered a lot.  Here’s a pithy saying:

“How do you get good judgment?”

“Experience.”

“How do you get experience?”

“Bad judgment!”

This is why reading is so powerful: it allows us to, as John Lewis Gaddis mentions in “The Landscape of History,”  “experience vicariously what we can’t experience directly: a wider view.”

As Rob Henderson discusses in Troubled, reading became his lifeline out of chaos. Growing up in foster care, he had little stability, few role models, and almost no consistent guidance. Books became his mental escape hatch; a way to construct order and meaning in a world that offered neither. He describes making “makeshift role models” from fragments of stories, using fiction and ideas as a substitute for the absent adults in his life. That habit of reading didn’t just teach him facts; it gave him a coherent narrative to inhabit, a scaffolding to build a self.

Johann Hari explains in Stolen Focus that reading also develops something deeper: empathy. When we immerse ourselves in another person’s thoughts, we “imagine what it is like to be inside another person’s head,” a process that literally trains our minds to understand others. As Hari writes, fiction provides a “kind of empathy gym,” expanding our ability to perceive complex interior worlds. Reading, in this sense, doesn’t just sharpen judgment: it enlarges the soul.

Thaler’s Misbehaving includes a lot of analysis on the idea of (marginal) utility: an important model you should read about more in depth, but (summarily) the idea that the enjoyment we get out of things follows a declining curve; it’s nonlinear.  As David Einhorn might say,

Two Jelly Donuts are an indulgent breakfast.

Twelve Jelly Donuts is fraternity pledge hazing.

It is, of course, sometimes difficult to know today what it is we will want tomorrow, let alone in five or ten years.  That’s where “base rates” can come in – we all unquestionably have our own idiosyncratic preferences, tendencies, and desires, but we’re all also human.  Therefore, it can be helpful to contextualize our “inside view” of what we want out of life with the “outside view” of those who’ve gone before us.

Pillemer’s “30 Lessons for Living” is a unique read. Some of the takeaways are “predictable” – but maybe that’s the point. Common sense isn’t so common!

Research, such as that presented in Shawn Achor’s “The Happiness Advantage” or Karl Pillemer’s “30 Lessons for Living” , tends to find that humans generally derive less satisfaction than they expect from reaching financial or material milestones, and far more than they realize or anticipate from a sense of meaning/purpose/ agency as well as social connection, particularly with close friends and family members. This deep-seated need for connection is a powerful “base rate” of human nature, one shared by our best friend.  Clive Wynne, in Dog is Love, argues that love is the defining evolutionary trait that binds humans and canines

The “experienced citizens” surveyed by Pillemer rarely expressed that more money would have made them happier: instead, it’s better relationships with their kids, having the courage to do things for themselves that seemed risky, etc.

Application/Impact: As Tavris/Aronson note in “ Mistakes were Made (but not by me)” , it can be easy to ignore the “outside view” and pretend that cognitive biases, self-justification, or other traits intrinsic to human nature are merely “other people problems.”

While that may be true sometimes, using the average human experience as your “base rate”  will likely lead to more informed and accurate predictions for maximizing your own utility rather than only using the “inside view.”