The numbers don’t add up
Economists love equations the way conjurors love top hats. Both create the impression that something rigorous and reliable is happening inside, when often what's inside is guesswork dressed up in tidy notation.
The physical sciences earn their mathematics honestly. An electron does not develop opinions. It does not read the newspaper, change its mind, or panic when its neighbours start selling. Physics can model matter precisely because matter is uniform, indifferent, and incapable of surprise. Human beings are none of these things. They carry histories, form loyalties, absorb rumours and act on hunches. A model that reduces them to interchangeable maximizers of utility has already discarded the very thing that makes them worth studying.
None of this means economics should abandon mathematics and take up interpretive dance. Numbers still earn their keep when dealing with large aggregates, where individual eccentricities wash each other out and broad tendencies emerge. Push the price of something down and more people will buy it. Impose rent controls and housing will grow scarcer.
These are not laws of nature. They are durable regularities, reliable enough to build policy on, though never immune to being upended by a change in circumstance or a shift in what people believe about tomorrow. A downward-sloping demand curve describes a pattern. It does not command obedience from the people producing it.
Behavioural economics is where the swagger really outruns the substance. Kahneman and Tversky found genuine, replicable quirks in how people judge risk and value. Anchoring is real. Loss aversion is real. But somewhere between the laboratory and the policy paper, a modest finding about undergraduates guessing at wheel spins became a universal law of human choice, confidently applied to pensions, organ donation and diet.
The replication crisis has been merciless to a good many of these findings, and merciless too to the nudge units that built careers on them. A tendency observed in sixty psychology students under artificial conditions is not a lever you can pull on sixty million adults living complicated lives.
The trouble is not that economists count things. Counting is fine. The trouble is that an equation looks precise even when the reality behind it is soft, contingent and shot through with human unpredictability. A regression coefficient carried to three decimal places implies a confidence that the underlying data cannot support. It is precision borrowed from physics and spent on psychology, where it does not belong. It has not earned the white coat of science that it wears.
There is a proper, useful role for mathematical models in economics, and it is a modest one. They can sketch tendencies. They can test whether an intuition survives contact with data. They can rule out claims that are flatly inconsistent with the numbers. What they cannot do is predict, with the confidence their decimal points suggest, how a particular population of free and unpredictable people will behave next year.
The economist who says a policy will probably raise employment, other things being equal, is telling the truth as far as it goes. The one who says employment will rise by exactly 2.3% has mistaken a guess for a measurement, and dressed the guess up in a suit it did not earn.
Markets do not work because they are mathematically optimal but because they let millions of separate judgments, made by people with knowledge no central model could ever gather, correct each other in real time. That is not a formula. It is closer to a conversation, conducted in prices instead of words, and no economist has yet written an equation that listens as well as a market does.
Madsen Pirie