Careful data beats ideology on our hardest policy fights
Good Economics for Hard Times is for voters, students, and policy readers who want what the evidence actually says on trade, immigration, welfare, and climate—not slogans from either side.
IJR · Nov 4, 2026 · 13 min read
In 60 seconds
- Economists lost public trust by selling trade, immigration, and tax cuts as settled win-wins when the research was full of caveats, regional losers, and genuine uncertainty.
- Workers rarely move toward higher wages; a small travel subsidy and sudden refugee inflows show migration is sticky and local labor markets adjust slowly and incompletely.
- Trade can raise aggregate welfare while specific communities take durable hits to jobs, wages, marriage, and health that last a decade or more.
- Nobody has a stable recipe for growth, so large tax cuts sold on investment and productivity payoffs rest on weaker evidence than politicians claim.
- Robot job losses so far are real but modest and regional; tax rules that cheapen machines relative to workers accelerate substitution beyond what pure productivity requires.
- Resistance to redistribution runs on moral stories about deservingness more than self-interest, and climate and safety-net designs fail when they ignore who specifically pays.
The big idea
Banerjee and Duflo argue that free-market orthodoxy and populist backlash both rest on bad evidence about immigration, trade, growth, automation, inequality, and climate. Good economics, in their view, is humble, local, and evidence-first: it designs policy around how people actually attach to place, work, and fairness instead of textbook models or ideology.
The book in brief
After 2016, confident economic predictions about trade, migration, and taxes looked foolish or dishonest to many voters. Banerjee and Duflo open on that wound. They say the discipline’s careful micro work was not the failure; the public voice was. Textbook absolutes were sold as settled science on questions the literature treated as contested, full of regional variation and real losers. That overconfidence, they argue, helped populists dismiss expertise altogether rather than demand better expertise.
From there the book walks flashpoint by flashpoint. Labor mobility comes first. Standard models assume wage gaps get arbitraged away by people moving; the authors show people often stay put even through steep local decline, which changes what immigration and trade shocks actually do on the ground. Trade policy is next. Aggregate gains can be real while losses stay concentrated and durable in the places most exposed to import competition, so “net positive on average” is not the same as a painless adjustment.
Growth and automation get the same skeptical pass. Cross-country growth stories prove unstable, which weakens policies justified mainly by promised GDP payoffs. Measured effects of industrial robots are real but narrower than popular panic suggests, yet tax rules can still tilt firms toward machines over workers. In each case the simple political story—markets fix everything, or technology ends work—fails contact with the data.
The second half turns to why repair is hard. Publics resist redistribution less from pocketbook math than from moral narratives about who deserves help. Cash transfers perform better than paternalists feared, yet a full universal basic income may miss the social meaning of work. Climate is framed as the starkest test: standard cost-benefit tools understate irreversible tail risk, and efficient taxes collapse politically when costs fall on people with no alternatives.
The book lands on a posture, not a master theory. Economists should act less like prophets of universal laws and more like plumbers—diagnose a local problem, test a fix, check results, adjust—because institutions and behavior vary too much for one-size ideology from either the free-market right or the populist left.
Context
Banerjee and Duflo are empirical economists whose careers center on building and testing anti-poverty interventions in the developing world through randomized trials. The book arrived after Trump’s election and Brexit, when public trust in economic expertise had cracked and both market orthodoxy and populist backlash were claiming certainty the authors say neither had earned.
The key ideas
1
Trust broke on overconfidence
Economics lost public trust less because its core findings were wrong than because economists spoke in textbook absolutes on contested questions like trade and immigration.
Public voice
Absolutes borrowed from textbook models; trade, immigration, and tax cuts sold as settled win-wins
Research literature
Caveats, regional variation, contested evidence, and real losers on the same questions
Banerjee and Duflo name the post-2016 wound directly. Economists who had confidently predicted disaster or prosperity from big policies looked foolish or dishonest to ordinary voters after Trump and Brexit.
They separate two things often blurred in public. Decades of careful microeconomic research still beat ideology when decisions affect real lives. The failure was the public voice, which borrowed certainty from simplified models instead of admitting caveats, regional variation, and genuine losers.
Free trade, immigration, and tax cuts had each been sold as settled science with guaranteed win-win outcomes. The research literature on each was messier. That gap, the authors argue, created space for populists to dismiss expertise entirely rather than demand better expertise.
Their answer is not to abandon the discipline. It is to practice what they call good economics: slow, evidence-first, comfortable with uncertainty and local variation, and resistant to sweeping claims from free-market boosters and protectionist populists alike.
In practice: After 2016, confident forecasts tied to Trump’s election and Brexit made economists look either wrong or dishonest to voters who then tuned out expertise.
2
Workers rarely relocate
People stay put even through steep, sustained decline because the psychological and social costs of leaving dwarf the financial upside models count.
Classical trade and labor theory assumes wage gaps get closed as workers move toward higher pay. Banerjee and Duflo show that rarely happens in practice.
Leaving means losing social networks, caregiving arrangements, local status, and identity tied to place. Standard models ignore those costs, so they treat staying as irrational when it is often a reasoned response to what migration actually takes.
The stickiness has a direct political bite. If natives are this reluctant to move even when income would rise, the feared floods of migrants swamping local workers are smaller in practice than simple models predict. Labor markets adjust far more slowly and incompletely than theory assumes.
Policy that plans only for people who leave will fail the many who stay. Friction has to be designed around, not assumed away.
In practice: A 2014 Bangladesh experiment offered rural households roughly $8.50 in travel subsidies for seasonal city migration during the lean season; migration and incomes rose sharply, yet uptake needed the subsidy and many returned home. Card’s study of the 1980 Mariel boatlift found roughly 125,000 Cuban refugees arriving in Miami barely moved wages or employment for existing low-skilled workers there.
3
Trade’s losers stay damaged
Trade liberalization can produce real aggregate gains while losses stay concentrated, durable, and unpaid for decades in the hardest-hit places.
- 1990–2007China shock: surge of Chinese manufacturing imports into the US
- Years after exposureHard-hit counties: lower manufacturing employment and wages, lower marriage rates, higher opioid deaths
- A decade laterNo meaningful local recovery for many of the most exposed places
Once immobility is on the table, trade policy looks different. Diffuse winners and visible, lasting local losers can coexist.
Banerjee and Duflo lean on Autor, Dorn, and Hanson’s work on the China shock—the surge of Chinese manufacturing imports into the US between 1990 and 2007. Counties most exposed saw sustained drops in manufacturing employment and wages, plus knock-ons including lower marriage rates and higher opioid-related deaths, with no meaningful recovery even a decade later.
Textbook theory expected workers to shift sectors or regions. Many of the affected were older, less mobile, and short on transferable skills, so entire local economies stagnated instead. “Trade is good on average, let the market sort it out” does not survive that record.
Protectionism is not their answer either. Tariffs impose their own costs and rarely reverse shifts in comparative advantage. The live debate, they insist, is compensation, place-based investment, and honest transition support—not whether trade happens at all.
In practice: US counties most exposed to Chinese import competition from 1990 to 2007 took lasting hits to manufacturing jobs and wages, with lower marriage rates and higher opioid deaths and little recovery a decade on.
4
Growth has no sure formula
Economists lack a reliable theory of what makes economies grow, so policies justified mainly by promised growth effects rest on far weaker evidence than their backers claim.
Across the political spectrum, leaders talk as if they know how to engineer GDP growth. Banerjee and Duflo attack that confidence.
Decades of cross-country growth regressions look notoriously unstable. Variables that predict growth in one dataset lose significance in another. The same country’s spurts and slowdowns often defy the usual suspects—investment rates, institutions, openness.
They apply the skepticism to arguments prominent around the 2017 US Tax Cuts and Jobs Act: that cutting top marginal rates spurs investment and growth. Historically the US grew rapidly in the 1950s and 1960s with a top marginal rate above 90 percent. Work by Piketty, Saez, and Zucman finds little empirical link between top-rate cuts since the 1980s and higher investment or productivity growth.
Taxes are not irrelevant. The point is narrower. Growth-worship crowds out interventions with stronger evidence behind them—health, education, direct transfers—that improve well-being even when GDP effects are uncertain or small. If growth cannot be engineered with confidence, sacrificing equity for a promised growth payoff looks much weaker.
In practice: The US grew rapidly in the 1950s and 1960s with a top marginal tax rate above 90 percent, undercutting the claim that high top rates choke growth.
5
Automation fears outrun measurement
Mass job destruction by robots is overstated relative to measured effects so far, but tax and subsidy rules needlessly accelerate capital substituting for labor.
Popular narrative
Robots will empty the labor market in a sweeping wave
Measured pattern
Modest, regionally concentrated manufacturing losses; tax rules that cheapen machines vs workers make substitution worse
Popular commentary often treats empty factories as destiny. Banerjee and Duflo separate the technology from the policy around it.
Drawing on Acemoglu and Restrepo’s work across US commuting zones, they find real but modest job losses from industrial robots, concentrated in specific manufacturing regions—not sweeping mass unemployment. Earlier waves, from tractors to computerization, displaced occupations and eventually generated new ones, though often too slowly to rescue the workers actually hit.
The sharper claim is distortion. US tax law, through accelerated depreciation and favorable capital-gains treatment, makes machines artificially cheap relative to hiring people—even when a human worker would be more efficient without the subsidy. Some automation then reflects the tax code’s tilt, not pure productivity.
China’s rapid automation in textiles and electronics shows a parallel pattern of displaced workers struggling to find comparable new work. The urgent problem, on this view, is tax neutrality between labor and capital plus support for the displaced—not banning the machines or pretending losses are imaginary.
In practice: Acemoglu and Restrepo find industrial robots caused real but modest, regionally concentrated job losses across US commuting zones rather than economy-wide emptying of the labor market.
6
Deservingness blocks the safety net
Opposition to redistribution is driven less by rational self-interest than by deep beliefs about who deserves help, shaped by effort narratives and identity.
American pattern
Welfare attitudes shaped heavily by beliefs in individual effort and mobility; often racially coded deservingness
European pattern
More weight on structural explanations for poverty
If markets do not self-correct the way theory promises, why do publics still resist the policies that might repair the damage? Banerjee and Duflo turn to attitudes, not only pocketbooks.
Survey and experimental work separates views of the poor as unlucky versus lazy. American welfare attitudes lean hard on individual effort and mobility; European attitudes lean more toward structural explanations—a divide that echoes comparative work by Alesina and Glaeser. In the US those narratives are often racially coded.
Simply showing people inequality or poverty statistics rarely moves them. Framing does. Portray aid recipients as trying hard versus passive, and support shifts more than need or effectiveness numbers do.
Technically sound programs can still die or get gutted after passage when they ignore fairness intuitions voters actually hold. Design has to work with those psychological realities, not against them.
In practice: Experiments show that portraying aid recipients as trying hard versus passive moves support for programs more than statistics about need or effectiveness.
7
Cash helps; work still means more
Unconditional cash transfers work better than skeptics feared, but a full universal basic income may undermine the social meaning of work that a check cannot replace.
Evidence on direct cash has grown fast. GiveDirectly’s large unconditional programs in Kenya found recipients did not significantly cut work effort or spike spending on alcohol and tobacco, undercutting common paternalistic objections.
That record has fed interest in universal basic income, including a Finnish pilot and a since-cancelled Ontario trial. Banerjee and Duflo push back on treating UBI as a full fix for rich-country labor disruption.
Work carries status, structure, purpose, and community beyond the paycheck. They tie the decline in US prime-age male labor force participation to rising despair, family breakdown, and opioid deaths—damage lost income alone does not explain and a check may not fully repair. A UBI generous enough to replace lost wages would also be enormously expensive if paid to everyone rather than targeted.
Their preferred direction leans toward targeted transfers plus public job guarantees or subsidized employment that keep work’s social function while still flooring income, with less stigmatizing safety-net design overall. They leave the debate more open than other chapters, on purpose.
In practice: GiveDirectly’s unconditional cash programs in Kenya did not show significant drops in work effort or rises in alcohol and tobacco spending among recipients.
8
Climate math misses tail risk
Standard cost-benefit climate models understate the case for urgent action by treating catastrophic, irreversible risk as an ordinary discounted-value tradeoff.
Nordhaus-style path
Higher discount rates; “optimal” warming targets around 3 to 4°C
Stern-style urgency
Much lower discount rate; aggressive near-term action because future welfare weighting is a moral choice and tails are irreversible
Banerjee and Duflo treat climate as the starkest test of whether economics can face a genuine emergency.
They contrast Nordhaus-style integrated assessment models—which use relatively high discount rates and land on “optimal” warming around 3 to 4 degrees Celsius—with the Stern Review’s 2006 case for a much lower discount rate and far more aggressive near-term action. Heavy discounting of future generations’ welfare is a moral choice, they stress, not a neutral technical setting. They side with the more urgent camp because tail scenarios are catastrophic and irreversible in ways ordinary expected-value tools handle poorly.
Efficient designs still fail politically when distribution is an afterthought. France’s 2018 gilets jaunes protests met a fuel tax that was tidy on paper but hit rural and lower-income drivers who lacked transit alternatives and saw little benefit from the revenue.
The repair they favor is carbon pricing paired explicitly with revenue recycled to affected households, so distributional politics sit inside the design. Aggregate efficiency is necessary and not sufficient.
In practice: France’s 2018 gilets jaunes protests erupted against a fuel tax increase that was efficient on paper but fell hard on rural and lower-income drivers with no transit alternatives.
9
Be plumbers, not prophets
Economists should diagnose specific local problems, test incremental fixes, and adjust—because institutions and behavior vary too much for universal laws.
- 1DiagnoseName the specific local problem, not a universal law
- 2Try a fixField a concrete intervention at testable scale
- 3Check resultsMeasure whether it worked for the people affected
- 4AdjustChange the design; do not defend the theory
Every flashpoint in the book—migration, trade, growth, automation, redistribution, climate—resists sweeping answers and demands local context, distribution, and psychology.
Duflo has described the ideal posture as plumbing: find the leak, try a fix, see whether it worked, adjust. Derive fewer grand solutions from first principles and expect fewer of them to travel everywhere unchanged.
The authors point to their own developing-world trials—conditional cash designs, deworming, microfinance variants tested against each other empirically rather than by ideological taste—as the template rich-country policy should use more often. Humility here is not small ambition. It is how cumulative progress happens instead of conviction-driven swings between pro-market and anti-market fashion.
Good economics, on this closing claim, fights two enemies at once: a complacent mainstream that oversold simple market fixes, and a reactionary populism that abandoned expertise. Neither helps people caught in trade shocks, automation, or climate disruption. Patient tinkering around how people actually behave does.
In practice: The authors’ own randomized trials compared conditional cash transfer designs, deworming programs, and microfinance variants empirically instead of choosing by ideology.
Key terms
- Good economics Slow, evidence-first policy work that accepts uncertainty and local variation and resists sweeping ideology from either market boosters or populists
- China shock The surge of Chinese manufacturing imports into the US between 1990 and 2007 and the lasting job, wage, and social damage in the most exposed counties
- Mobile worker assumption The textbook idea that workers will move to close wage gaps; the authors show social and identity costs often block that move
- Deservingness Moral beliefs about who has earned help—often tied to effort stories and identity—that shape support for redistribution more than self-interest math
- Unconditional cash transfer Money given to people with no strings on how they spend it; evidence the authors cite finds less waste and work collapse than critics expect
- Universal basic income A cash floor paid to everyone; the authors doubt it as a full answer where work’s social meaning and fiscal cost matter
- Discount rate How heavily climate models weigh future generations’ welfare; a higher rate pushes “optimal” warming higher and delays aggressive action
- Economists as plumbers A posture of diagnosing local problems, testing fixes, checking results, and adjusting rather than proclaiming universal laws
What to do
- Ask what specific study backs any confident claim in the news about tax cuts, trade, or robots—and whether that study’s setting resembles yours.
- Build support for people who stay in declining places rather than plans that work only if workers relocate.
- When someone opposes redistribution, address their beliefs about deservingness directly instead of opening with inequality statistics.
- Treat every policy proposal, including your favorites, as a hypothesis to test at small scale before defending it as settled.
- Pair any efficient climate price with visible revenue returned to the households who would otherwise bear the cost.
Questions to think with
- When did a confident economic prediction you trusted fail once local losers and adjustment frictions entered the picture?
- If workers will not move for higher wages, what does fair help after a trade or robot shock actually require in a specific town?
- Can a cash floor without work preserve dignity, or does joblessness damage something a check cannot repair?
- Whose costs get ignored when a policy is sold only as aggregate net benefit?
- Where should economists admit they cannot run the decisive experiment—and what should guide policy then?
The other side
The book’s RCT comfort works for cash and migration tweaks but cannot randomize growth paths or climates, so later chapters fall back on the cross-country correlations, cases, and simulations the authors elsewhere call unreliable; it underweights entrenched power and institutions, says little about building coalitions against organized opposition, and can feel like sharp field chapters awkwardly joined to more provisional Western literature reviews without a unified theory of policy success.
Who it's for
Voters trying to separate evidence from talking points on trade, immigration, or welfare; policy professionals and students who want the empirical state of play rather than slogans; readers shaken by 2016-era populism who want a clear account of what mainstream economics got wrong and still gets right.