09 · Thomas Sowell

Thomas Sowell

Thomas Sowell (b. 1930) grew up in Harlem, left school at 17, and after the Marines studied at Harvard, Columbia and Chicago, where Milton Friedman taught him. He was a Marxist in his twenties; studying how policies actually worked changed his mind. For decades at the Hoover Institution he wrote about how knowledge is used in society (Knowledge and Decisions, 1980), the assumptions that divide political camps (A Conflict of Visions, 1987; The Vision of the Anointed, 1995), and economics without equations (Basic Economics, 2000; Applied Economics). Late in his career he returned to his lifelong subject, why groups differ (Discrimination and Disparities, 2018). His method is the same throughout: ignore intentions, follow the incentives, and keep asking what happens next. Below, each of those questions is a model you can run.

There are no solutions, only trade-offs

sowell-trade-offssowell-visible-vs-statistical-victims
A drug regulator decides how much evidence to require before approving a new medicine. Too little and some unsafe drugs reach patients; too much and patients die waiting for good drugs that were held back. No setting removes both. The best available choice is a point on a curve, and who gets blamed for which deaths decides where an agency will actually put it.
visible victims count

The frontier of possible outcomes (deaths per year, illustrative)

frontier with $0 budgetfrontier at this budget
020040060080005001kdeaths from unsafe drugs approved (seen)deaths waiting for good drugs (unseen)the 'solution': zero of both (unreachable)fewest total deathsyou

Every point on the cyan curve is a choice. Moving along it trades one kind of death for the other; money moves the curve inward, but never to zero, and that money is not available for anything else.

unsafe approvals
273
visible, traceable
deaths from delay
205
statistical, unseen
total deaths
478
best possible here: 468
budget spent
$400M
foregone elsewhere
Total deaths are lowest at 57% strictness. Even there, 468 people a year die from one error or the other. That is the best available trade-off, not a failure to find the solution.

Try: drag strictness to 100% (no unsafe drugs, but the unseen deaths are at their maximum), then back to 0%. Raise the budget and watch both fall while the bill rises. Then switch to "10× in the headlines" and see where the agency's own incentives pull the choice.

A price does two jobs at once: it tells people how scarce something is, and it decides who gets it. Hold the rent below the market level and both jobs pass to something else. Drag the ceiling and watch the shortage open up. Then let the years run, as rent control's effects mostly appear slowly.
time horizon

Rental market (drag the ceiling line)

$0$500$1,000$1,500$2,000$2,500$3,0000k20k40k60k80k100k120kapartments (thousands)rent per monthwhat the marginal renter would pay: $1,700shortage 25.0kdemandsupply (short run)market rent $1,600ceiling $1,200 ⇕
units offered
55.0k
vs 60k at market rent
units wanted
80.0k
shortage
25.0k
households searching
key money
$500
side payment per month
lost trades
$15M/yr
deadweight loss
saved by lucky tenants
$264M/yr
if no side payments

Who gets an apartment? (each square = 2,000 households)

■ housed at $1,200 · □ want a unit at that rent but there is none. With price no longer deciding who gets the scarce units, something else does: waiting lists, connections, discrimination, bribes, or sitting tenants never moving.

Try: drag the ceiling to $1,000. In the short run the shortage is mostly people newly wanting units at the lower rent. Now switch to the long run: owners convert and stop building, the units offered collapse, and the shortage and key money grow.

Housing stock · year 30

free-market cityrent-controlled citywant a controlled unit
0501001502000102030yearapartments (index)

Real rent and building quality

free rent (vs cost)controlled rent (vs cost)upkeep / quality
0%50%100%0102030yearcost of supplying a unit
controlled stock
64
free city: 139
real rent
50%
free city: 103%
quality
10%
waiting list
150
vacancies / yr
1.0
free city: 20.8
newcomer's wait
50+ yr

Try: with inflation at 4% and a 2% allowed rise, press Play. The ceiling starts below cost and real rent erodes every year. Then upkeep is cut, new rental building stops, units leak away, and sitting tenants stop moving, so a newcomer's wait stretches to years. Set the allowed rise equal to inflation and the decay slows a lot, but any ceiling below the market rent still stops new rental building.

Rent control is one of the few questions on which economists of the left and right largely agree. Assar Lindbeck, a Swedish social democrat, wrote that it "appears to be the most efficient technique presently known to destroy a city — except for bombing." Defenders argue that controls protect sitting tenants from displacement, which they do, at the cost of the tenants who come later.
A minimum wage does not make anyone more productive; it makes it illegal to hire someone whose work is worth less than the minimum. Here 160 job seekers differ in what they can add per hour, and the young and inexperienced cluster at the bottom. Raise the floor and see who is no longer hired, and what that does to them over the following years.

160 job seekers, placed by the value they add per hour

$0/hr$10/hr$20/hr$30/hr$40/hrmust add ≥ $11.11/hr to be worth $10.0016–1920–2425+

Filled: hired. Hollow: priced out, because the law forbids a wage they would be worth. Green ring: still hired and paid more than before (happens only when employers had wage-setting power).

employed
144/160
got a raise
0
unemployed 16–19
33%
unemployed 20–24
5%
unemployed 25+
1%
hourly payroll
$2,885
no floor: $3,016

The first rung: today's 16–19-year-olds over the next 10 years

no minimum wageminimum $10.00
01020300246810years from nowavg. value added ($/hr)
10-yr earnings per teen
$367k
no floor: $436k
teens never hired in 10 yrs
7%
never got on the ladder
Skills grow on the job (+$1.80/hr a year for teens here) and only by 30¢ without one. A lost first job is also lost experience.

Try: drag the minimum from $10 down to none and back. At $10 adults barely notice, but a third of teens are hollow. On the ladder below, some never catch up, because the experience they would have gained never happens. Then give employers 20% wage-setting power: a floor around $8 now raises many workers' pay at little cost in jobs, until it climbs past what the least productive are worth.

This is the most contested section on this page. Card and Krueger (1994) found no job loss from a New Jersey increase, and many economists think employers have some wage-setting power, which the second slider models. Sowell's reply is that the effects show up over years and among the least experienced: fewer hours, less training, and teenagers who are never hired.
The first effect of a policy is usually the one it was designed for, and the one everyone sees. People then adjust, and their adjustments show up later, spread over different people, and are usually blamed on someone else. Step through the stages and compare the policy's net effect with the credit it gets.

Goal: Make housing affordable

1stage one+2 for the intendedvisibility 100%2stage 23stage 34stage 4
  1. stage 1Tenants in controlled apartments pay less rent.
    falls on: sitting tenants
stages revealed
1/4
net for the intended
+2
sum of all stages so far
credited to the policy
+2.0
visible effects not blamed on others
gap
0.0
why it stays popular
Real case: New York after 1943; San Francisco's 1994 expansion (Diamond, McQuade & Qian 2019 found covered landlords cut rental supply 15%).

Try: pick a policy and keep pressing "And then what?". The net effect on the people it meant to help turns negative, while what gets credited to the policy barely moves, because later stages are less visible and get blamed on someone else.

A conflict of visions: constrained vs. unconstrained

sowell-constrained-visionsowell-unconstrained-visionsowell-process-vs-results
Sowell argued that people who disagree about crime, war, poverty and justice often disagree about something underneath: what human beings are like and how much anyone can know. He called these 'visions', a sense of how the world works held before any argument. Switch lenses and see one fact read two ways, then try placing thinkers and claims on the spectrum.
read the world through the

Assumptions, mostly unstated

Human nature
Fixed and limited: self-interested, short of knowledge, the same in every age. Work with it, don't hope to change it.
Knowledge
Scattered across millions of people, mostly unarticulated, embodied in traditions, prices and habits.
Social processes
Systemic: markets, law, families and customs coordinate without anyone designing the outcome.
Key question
What are the trade-offs, and which institutions make the best of them?
Justice and equality
Process: the same rules for everyone, whatever outcomes result.

Same evidence, different reading

constrained reading of “crime rises”

Restraints have weakened: punishment became less certain and families and norms frayed. Ask what incentives changed.

Neither reading is forced by the fact itself. Each is what you would expect given the assumptions on the left, which is why the two sides so often argue past each other.

Sorting game: place each thinker or claim on the spectrum

C constrained · c leans constrained · · mixed · u leans unconstrained · U unconstrained

  • Adam Smith
  • William Godwin
  • 'Prosperity, not poverty, is what needs explaining'
  • The Federalist (Hamilton, Madison)
  • 'Fairness means equal results'
  • Edmund Burke
  • Karl Marx
  • 'Judge a system by its rules, not its results'
  • Condorcet
  • F. A. Hayek
  • John Rawls
  • 'Experts should design the solution'

Try: pick "Poverty exists" and flip the lens. Then place each card on the spectrum and press Check. Sowell's point is not that one vision is always right, but that most political conflicts come from these premises, not from the evidence.

Knowledge and decisions: how fast are errors corrected?

sowell-knowledge-and-decisionssowell-feedback-mechanisms
Every institution makes mistakes. Sowell's question is how each one finds out, and what forces it to change. Here, what people want drifts and sometimes jumps. A market, an elected government and a central plan each try to supply it, and each learns from a different kind of feedback: losses every period, an election every few years, a scheduled plan revision.

What people want vs. what each system supplies · period 90

what people wantMarketElectoral politicsCentral plan
050100020406080periodquantity

Average miss per period

Market
1.4
Electoral politics
4.6
Central plan
7.6
  • ■ Market: profit and loss, every period
  • ■ Electoral politics: elections, if voters notice
  • ■ Central plan: scheduled plan revisions

Feedback log

  1. t80 plan revised to 69.8 using data 6 periods old
  2. t80 election: gap of 2.8 too small to notice; no change
  3. t64 election: voters notice a gap of 8.8; policy changes
  4. t60 plan revised to 67.7 using data 6 periods old
  5. t48 election: voters notice a gap of 21.4; policy changes
  6. t40 demand jumps up by 25

Try: press Play, then Demand shock. The market is noisy but back on target within a few periods, because every period of error costs producers money. Politics waits for an election and moves only if the gap is big enough for voters to notice; the plan corrects on schedule, using stale numbers. Lower the attention threshold or shorten the cycles and the gaps narrow, but never to the market's speed.

The model compares feedback speed only. It doesn't say markets should decide everything: some goods, like defence or courts, have no market price to learn from. Sowell's point is narrower and harder to escape: wherever a decision is made, ask who bears the cost of being wrong, and how soon.

Disparities are not discrimination

sowell-disparities-not-discrimination
Two groups are paid by exactly the same rule: pay rises with age up to about 50, is higher in big cities, and higher in some fields of study. Group B is younger, lives elsewhere and studied different subjects. No one is biased, and yet B earns less. Then switch on a real bias and see what changes in the statistics.

Income by age (every 4th person shown)

● group A · median age 41● group B · median age 31
050k100k2030405060ageannual incomeA average $54kB average $46k
raw gap
$8k
B earns 84% of A
gap among like people
-$430
same age, place, field

Where the raw gap comes from

age
$5k
location
$2k
field of study
$2k
unexplained
$0
luck only
Not one employer here treats anyone differently, yet B earns less. Every bit of the gap comes from differences in age, place and field.

Try: with discrimination at zero, set B 12 years younger and watch the raw gap appear with no bias at all. Then set the three factors to zero and add a 10% bias: now the raw and adjusted gaps match. Finally combine them: the raw gap alone can't tell you which world you are in.

Discrimination has a cost, and markets make the discriminator pay it

sowell-cost-of-discrimination
Following Gary Becker, Sowell stresses that discrimination is not free. A firm that turns away better workers because of their group gets less output for the same wages, and in a competitive market rivals who don't discriminate outgrow it. The cost disappears when wages are set above market, so applicants queue, or when a regulator passes costs on to customers. Those are the settings where, historically, discrimination lasted longest.

16 firms, sized by market share · round 0

—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
—
■ hires group A first, B only if A runs out■ hires the best applicants
discriminators' market sharegroup B share of jobs
0%50%100%01020304050roundB share of applicants
discriminators
50%
of market share
cost of bias
0.0%
output lost per job
firms exited
0
replaced by new entrants
B share of jobs
—
B is 40% of applicants
Passing over a better B applicant costs output. In a competitive market that cost comes out of profit, and the market shrinks firms that pay it. New entrants keep discriminating at the old rate, and keep losing.

Try: press Play in the competitive market and watch the red firms shrink and exit. Then switch to "wage set above market" and to "regulated" and play again: the same preferences survive because they no longer cost the firm anything.

  • takeawayAsk "at what cost?" and "compared to what?". Every policy is a point on a frontier of trade-offs, and who bears the unseen costs often decides which point gets chosen.
  • takeawayControlling prices doesn't make anything less scarce. Ceilings create shortages, deterioration and black markets; floors price out the least experienced. Most of the damage arrives in stages two and three, where nobody connects it to the policy.
  • takeawayJudge institutions by their feedback: who decides, what they know, and how quickly being wrong costs them. Before treating outcome gaps as proof of bias, check age, place and skills, and notice that competition itself punishes discrimination.

Key concepts · 20

Thomas Sowell
  1. Trade-offs, not solutions

    Every choice gives up something; policy picks a point among imperfect outcomes rather than eliminating a problem.

    Framing an issue as a problem with a solution hides what the solution costs. Asking 'at what cost?' turns slogans into choices that can be compared.

    A Conflict of Visions (1987); Basic Economics (2000)

    FDA approval delays vs. unsafe drugsSafety regulation vs. price and availability
    ↑ see it in the visualization
  2. Visible vs. statistical victims

    Some costs of a decision fall on identifiable people, others on people nobody can name.

    Decision-makers are blamed for the first and not the second, so their incentives tilt toward avoiding visible errors even when that causes more harm overall.

    Knowledge and Decisions (1980); The Vision of the Anointed (1995)

    Thalidomide (1960s) and the drug-lag that followedBeta-blockers approved in the US years after Europe
    ↑ see it in the visualization
  3. Price ceiling

    A legal maximum price set below the market-clearing level.

    At the lower price more is wanted and less is offered, so a shortage appears. Price stops rationing the scarce good, and something else (queues, connections, bribes) takes its place.

    Basic Economics (2000)

    US gasoline lines, 1973–79Venezuela's price controls, 2010sSoviet food queues
    ↑ see it in the visualization
  4. Shortage vs. scarcity

    Scarcity is permanent (there is never enough for everyone to have all they want); a shortage is people unable to buy at the current price.

    Price controls don't increase how much exists. They turn scarcity, which prices ration, into a shortage, which queues ration.

    Basic Economics (2000)

    ↑ see it in the visualization
  5. Rent control

    A ceiling on rents, usually with limited annual increases.

    Short-run benefits go to sitting tenants. Over years, less building, conversions, deferred maintenance and frozen mobility shrink and degrade the rental stock that newcomers can get.

    Basic Economics (2000); Economic Facts and Fallacies (2008)

    New York City since 1943San Francisco's 1994 expansionStockholm's decade-long waiting lists
    ↑ see it in the visualization
  6. Hidden price rises: quality deterioration

    When the money price is held down, sellers cut quality, service and maintenance instead.

    The real price of a unit includes its condition. A controlled rent can rise in effect even when the number on the lease doesn't.

    Basic Economics (2000)

    Abandoned buildings in the 1970s South BronxSoviet-era housing upkeep
    ↑ see it in the visualization
  7. Black markets and key money

    Side payments and illegal resales that reappear when a legal price is below what buyers would pay.

    The gap between the ceiling and the marginal buyer's willingness to pay doesn't vanish; it goes to whoever controls access, legally or not.

    Basic Economics (2000)

    Key money for rent-controlled flatsTicket scalpingVenezuelan bachaqueros
    ↑ see it in the visualization
  8. Price floor

    A legal minimum price set above the market-clearing level.

    More is offered than is bought: surpluses of farm goods, or of labour in the form of unemployment.

    Basic Economics (2000)

    US and EU farm price supports and butter mountains
    ↑ see it in the visualization
  9. Minimum wage

    A price floor on labour.

    A worker whose output is worth less than the legal minimum becomes unemployable, and those are mostly the young and inexperienced. Whether modest increases cost many jobs is contested (Card and Krueger 1994 found no job loss in New Jersey fast food).

    Basic Economics (2000); Applied Economics (2004)

    Puerto Rico under the federal minimumBlack teenage unemployment in the US after 1950s minimum-wage increases
    ↑ see it in the visualization
  10. Entry-level jobs as training

    Much of the value of a first job is the experience and work habits it gives, not the pay.

    Pricing someone out of a low-wage job also denies them the experience that would have raised their later earnings.

    Basic Economics (2000); Economic Facts and Fallacies (2008)

    ↑ see it in the visualization
  11. Stage-one thinking

    Judging a policy by its immediate, intended effect without asking what happens next.

    People adjust, and the later stages often reverse the first. Sowell's test is to keep asking 'and then what?'

    Applied Economics (2004)

    Price-gouging bans after hurricanesUS steel tariffs, 2002
    ↑ see it in the visualization
  12. Constrained vision

    A view of human beings as fixed in their moral and intellectual limits, with knowledge dispersed among many people.

    It favours systemic processes (markets, law, tradition) that work with human nature, and sees policy as a matter of trade-offs.

    A Conflict of Visions (1987)

    Adam SmithEdmund BurkeThe FederalistF. A. Hayek
    ↑ see it in the visualization
  13. Unconstrained vision

    A view of human nature as improvable, with the knowledge to solve social problems available to the wise and well-intentioned.

    It favours deliberate, articulated solutions, and treats bad outcomes as evidence of a problem someone should fix.

    A Conflict of Visions (1987)

    William GodwinCondorcetRousseau (partly)
    ↑ see it in the visualization
  14. Process vs. results

    Whether fairness and justice are judged by the rules people followed or by the outcomes that came out.

    The two visions differ most clearly here, which is why they disagree on equality, justice and poverty even when they agree on the facts.

    A Conflict of Visions (1987); The Quest for Cosmic Justice (1999)

    ↑ see it in the visualization
  15. The vision of the anointed

    Sowell's term for the outlook of an intellectual elite that sees itself as morally and intellectually superior and its policies as self-evidently right.

    Because the vision is held for self-flattering reasons, its policies are insulated from feedback: failures are blamed on others or on not going far enough.

    The Vision of the Anointed (1995)

    ↑ see it in the visualization
  16. Knowledge and decisions

    The key question about any institution is who decides, with what knowledge, under what incentives.

    Knowledge is costly and scattered. Institutions that leave decisions with people who have the knowledge and bear the consequences tend to do better than those that don't.

    Knowledge and Decisions (1980)

    ↑ see it in the visualization
  17. Feedback mechanisms

    The ways an institution learns that it is wrong and is forced to change: losses in markets, elections in politics, and very little in many bureaucracies.

    What matters is not avoiding errors, which no one can do, but how fast and how cheaply they are corrected.

    Knowledge and Decisions (1980)

    Firm bankruptcy vs. programmes that outlive their purposeSoviet five-year plans
    ↑ see it in the visualization
  18. Disparities are not proof of discrimination

    Groups differ in age, location, education, culture and choices, so unequal outcomes appear even when everyone is treated the same.

    Statistical gaps are often read as proof of bias. Sowell argues that even-outcome benchmarks rarely exist anywhere in the world, so the first question is what produced the gap. Critics answer that the 'explanatory' factors can themselves be shaped by past discrimination.

    Discrimination and Disparities (2018); Economic Facts and Fallacies (2008)

    Median ages of US ethnic groups ranging from the 20s to the 40sUneven ethnic representation in professions worldwide
    ↑ see it in the visualization
  19. The cost of discrimination

    In a competitive market, an employer who passes over better workers for group reasons pays for it in lost output and profit.

    Competition penalises discrimination; price controls, surplus applicants, regulated monopolies and government employment remove the penalty, which is where discrimination has historically persisted.

    Discrimination and Disparities (2018); after Gary Becker, The Economics of Discrimination (1957)

    Southern streetcar companies resisting Jim Crow seating lawsApartheid laws restricting employers who wanted to hire Black workers
    ↑ see it in the visualization
  20. The economic role of middlemen

    Traders who stand between producers and consumers cut the costs of finding, storing, and distributing goods.

    Middlemen are often resented as people who 'add nothing', and middleman minorities have been persecuted for it, but removing them usually raises costs, because someone still has to do the work.

    Basic Economics (2000); Black Rednecks and White Liberals (2005)

    Chinese merchants in Southeast AsiaIndians in East AfricaSoviet state distribution