Risk vs. Uncertainty: Knight's Distinction and the Kinds You Face

"Risk" and "uncertainty" get used as if they mean the same thing. In everyday conversation that is harmless. In planning, it causes real trouble, because the tools that work well for one can mislead you badly on the other.

The distinction goes back more than a century, to the economist Frank Knight. It has since been refined by statisticians, auditors and government analysts, and the refined version is surprisingly practical. Once you can tell which kind of unknown you are dealing with, you can stop arguing about whether a number is "right" and start asking what kind of response the situation actually needs.

This guide explains Knight's original idea, a modern way to sort uncertainty into types, and how to match your response to each type.

What Knight actually said

In his 1921 book Risk, Uncertainty, and Profit, Knight drew a line between uncertainty that can be measured and uncertainty that cannot. To keep the two apart, he wrote, "we may use the term 'risk' to designate the former and the term 'uncertainty' for the latter."

He grounded this in three kinds of probability situations:

1. A priori probability, like the throw of a fair die, where the cases are essentially identical and the odds can be worked out in advance.
2. Statistical probability, like fire losses across many buildings, where the odds come from grouping similar cases and counting what happened.
3. Estimates, where, in Knight's words, "there is no valid basis of any kind for classifying instances."

Knight argued that the first two can be handled by grouping enough cases together, which is what insurance does. The third cannot. And he was blunt about where business sits: business decisions, he wrote, "deal with situations which are far too unique, generally speaking, for any sort of statistical tabulation to have any value for guidance."

That does not mean business judgment is guesswork. It means judgment, not a formula, carries the load whenever there is no meaningful set of comparable past cases.

Why the distinction still matters

Modern risk standards use the word "risk" more broadly than Knight did. NIST's online glossary defines risk as the "effect of uncertainty on objectives," a definition it traces to ISO's risk management vocabulary and to the U.S. Office of Management and Budget's Circular A-11. Under that definition, both of Knight's categories are "risk," and that is fine for organizing a risk program.

The practical danger is treating every item in a risk program as if it were Knight's measurable kind. A probability-times-impact score looks equally precise whether it rests on years of claims data or on one person's hunch. If you do not track which is which, the scores for genuinely unknowable items borrow credibility they have not earned.

A modern map: three kinds of uncertainty

The Uncertainty Toolkit for Analysts in Government, which the UK's AQuA Book describes as produced by a cross-government group, offers a clear classification that many organizations can borrow. It describes three broad types.

• Aleatory uncertainty: the inherent randomness that is always present due to underlying probabilistic variability. It can be quantified, usually with a probability distribution, but it cannot be completely removed. The toolkit's example is a coin toss: you know the possible outcomes and their odds, but you cannot know the next result before it happens.
• Known unknowns (epistemic uncertainty): gaps in knowledge about the system you are trying to understand. Assumptions are used to plug these gaps. This kind is reducible: new data, research or wider data collection can narrow or remove it. In the coin example, you do not yet know whether the coin is fair, but every toss tells you more.
• Unknown unknowns: things you do not know you do not know, usually arising from situations you have not experienced before. The toolkit says these cannot be quantified, although their likely effect can be assessed qualitatively. Its example: someone swaps your coin for a weighted one without your knowledge, and your earlier analysis no longer applies.

The toolkit also notes "deep uncertainty," where you cannot anticipate the problem you will later need to understand. In that case, it says, a judgment is required on when a decision must be made, with analysis done on the best information available by then.

Roughly speaking, Knight's "risk" lines up with the aleatory type, while his "uncertainty" spans the other two.

Match the response to the kind

Each type of uncertainty calls for a different move. Here is a practical way to think about it.

For aleatory uncertainty, plan for the spread rather than the average. Use buffers sized to the observed variation: cash reserves, spare capacity, safety stock, insurance. Arguing about the next outcome is pointless; preparing for the range is not.

For known unknowns, buy information. Run a pilot, survey customers, test an assumption, talk to someone who has done it before. The toolkit points to sensitivity analysis, which tests how much your results change as you vary assumptions. If a conclusion flips when one assumption moves a little, that assumption is where your next research dollar should go.

For unknown unknowns and deep uncertainty, build robustness instead of precision. Favor options you can reverse, keep some slack, watch for early signals, and decide in stages rather than all at once.

For example, imagine a hypothetical 20-person landscaping company deciding whether to open a second location. Weather-driven swings in weekly demand are aleatory: the owner has years of records and can size a cash buffer. Whether the new town's homeowners will pay current prices is a known unknown: a season of subcontracted work there would answer it cheaply. A sudden change in local regulations or an unexpected new competitor is closer to the third type: the owner cannot estimate it, but can lease rather than buy the new yard, which keeps the exit cheap.

The uncertainties you cannot control

The UK National Audit Office's 2023 guide Managing uncertainty makes a useful split. Some uncertainties can be reduced with time and research; they can be analyzed, managed and mitigated. Others sit in the wider environment, where a team may have little control and which cannot be managed or eliminated, yet can still affect whether the effort delivers value for money.

The NAO found that decision-makers "do not always take time to understand and consider these wider uncertainties or can underestimate their impact." The result, it warns, can be plans that give "a false sense of certainty" or are not resilient to change.

Its guide suggests questions that translate well outside government, including:

• Have we taken a structured approach to identifying uncertainties, involving people with a broad range of perspectives?
• Have we identified the range of scenarios we might need to operate under, and how sensitive our outcomes are to each uncertainty?
• Have we identified our interdependencies, such as reliance on other projects, and the uncertainty they bring?

A quick exercise for your next plan

Take your current plan or budget and list the five assumptions it depends on most. For each, write down which kind of uncertainty it is: inherent variability, a gap you could close, or something you genuinely cannot foresee. Then check that your response fits the type. Many teams discover they are spending effort predicting things that are random, while ignoring knowledge gaps that a few phone calls could close.

Key takeaways

• Knight separated measurable uncertainty ("risk") from unmeasurable uncertainty ("uncertainty") in 1921, and argued most business decisions fall in the second group.
• A useful modern map has three types: inherent variability, known unknowns and unknown unknowns.
• Buffer against variability, buy information to shrink knowledge gaps, and build robustness against what you cannot foresee.
• Label which kind of uncertainty sits behind each estimate so precise-looking numbers do not hide guesses.
• Ask explicitly about the wider uncertainties you cannot control; they are the easiest to underestimate.

Sources

• Frank H. Knight, Risk, Uncertainty, and Profit (1921), Library of Economics and Liberty edition
• Uncertainty Toolkit for Analysts in Government, Chapter 3: Defining and identifying uncertainty
• National Audit Office, Managing uncertainty: Questions for decision-makers to ask in an uncertain environment (August 2023)
• UK Government Analysis Function, The AQuA Book (July 30, 2025)
• National Institute of Standards and Technology, CSRC Glossary: risk