All podcasts / People I (Mostly) Admire / Summary

The Economist Who (Gasp!) Asks People What They Think

2025-08-30 - source - Read full transcript

Steve Levitt (host)Stefanie StantchevaMorgan Levey

Key insights

Economists have historically dismissed survey research because they trust revealed behavior over stated beliefs, but this discards the only direct way to learn what people understand, believe, and want.

Stantcheva argues the profession's norm of only believing 'what people do, not what they say' has real limits: behavior alone can't reveal people's constraints, misperceptions, or reasoning, which are necessary inputs for designing better policy. Pursuing survey methodology was treated as close to career suicide when she started, but the field's attitude has shifted as the value of well-designed, large-scale surveys became visible.

survey-methodology-in-economics

Americans wildly overestimate the share of immigrants in their country, and the gap is far larger than can be explained by living in a low-immigrant neighborhood.

In Stantcheva's six-country survey (US, UK, Sweden, Germany, Italy, France), the actual US documented-immigrant share was about 10 percent at the time of the study, but the average perceived share was 36 percent; even accounting for people conflating first- and second-generation immigrants, the gap barely closes. The misperception is correlated with local immigrant density (more accurate near immigrants) but the overall bias is still enormous and driven heavily by media salience.

misperceptions-and-policy

People systematically overestimate immigrants' unemployment rate and welfare reliance relative to native-borns, even though in the US the true unemployment gap is close to zero.

A naive question about immigrant unemployment gets inflated answers partly because respondents don't share economists' technical definition of 'unemployed.' Benchmarking against a parallel question about non-immigrant unemployment isolates the real bias: people still believe immigrants are disproportionately more unemployed and more reliant on welfare than they actually are, feeding a 'free rider' narrative not supported by the US data.

misperceptions-and-policy

Merely being asked questions about immigration - with no new information provided - causally reduces stated support for redistribution generally, not just toward immigrants.

Using randomized question order, Stantcheva found that respondents asked immigration-perception questions before policy questions became systematically less supportive of progressive taxes and welfare programs than respondents asked in the reverse order, even though no new facts were given. This suggests that priming people to think about an out-group ('who am I redistributing to?') generalizes into a broader backlash against redistribution to everyone, consistent with theories that generosity travels less well across ethnic/national lines than within them.

misperceptions-and-policy

Correcting immigration misperceptions with pure facts does almost nothing to change views, but an anecdote about a hardworking immigrant meaningfully increases support for both redistribution and immigration.

In a randomized information experiment, groups given accurate immigrant-share or origin-country statistics showed essentially no shift in policy views (people believed the facts but didn't update their preferences), while a narrative describing a single hardworking immigrant's day produced real increases in support. Stantcheva attributes this to immigration being an unusually narrative-driven topic, where the underlying concern (perceived free-riding) is better addressed by a counter-story than by statistics.

facts-vs-narratives

For the estate tax, the opposite pattern holds: correcting the facts is highly persuasive, unlike with immigration.

Americans surveyed believed roughly a third of people pay the estate tax when the true figure is under one in a thousand (the threshold at the time was about $11 million per person). Simply telling respondents the true incidence sharply increased support for the tax, including for raising it and lowering the exemption threshold; adding a narrative element (a photo of a mansion) added only a small effect on top of the fact itself, showing facts alone can be powerful when the topic isn't already narrative-saturated.

facts-vs-narratives

Policy support is driven much more by perceived fairness, distributional impact, and trust in government than by narrow self-interest or straightforward economic argument.

Across Stantcheva's studies, people's stated views on tax policy or a specific tax are shaped less by whether the policy directly costs or benefits them and more by whether they see it as equitable and whether they trust the institution implementing it, which Levitt says undercuts the common economist assumption that clear economic arguments alone should move public opinion.

misperceptions-and-policy

Support for climate policy depends heavily on perceived equity of the specific policy design, not on convincing people climate change is real.

In a 20-country survey, most respondents were already convinced climate change is a serious problem, so the binding constraint on support was whether a given policy (like a carbon tax) was seen as regressive. A carbon tax with revenues simply absorbed into the general government budget polled poorly everywhere; the same tax with revenues either returned progressively to lower-income and vulnerable households, or earmarked for green infrastructure, polled substantially better across all 20 countries.

misperceptions-and-policy

'Zero-sum mindset' - the belief that one person's or group's gain necessarily comes at another's expense - is a cross-cutting psychological factor that predicts a distinctive bundle of policy views that don't fall neatly on the left-right spectrum.

A high zero-sum score predicts both support for taxing the rich (because the rich are seen as gaining at the poor's expense) and support for restricting immigration (because immigrants are seen as gaining at natives' expense) - two positions that cross conventional partisan lines. This makes zero-sum thinking a distinct axis of political psychology, not reducible to standard left/right ideology.

zero-sum-mindset

Zero-sum thinking correlates strongly with a person's (or their family's) lived economic history rather than being a simple cognitive bias, and it runs in opposite generational directions in rich versus developing countries.

In the US and other rich countries, younger generations hold more zero-sum views than older generations, tracing to lower growth and mobility in the economic environment they grew up in; the reverse pattern holds in developing countries now experiencing higher growth than prior generations did. At the individual level, people whose families experienced upward mobility, or whose ancestors were not subject to enslavement or oppression, are less zero-sum, suggesting the mindset is calibrated (sometimes over-calibrated) from real historical experience rather than pure misperception.

zero-sum-mindset

Levitt's 'get paid for your data' vision goes beyond current market-research monetization models like Verb.AI toward outcome-based revenue-sharing between consumers, platforms, and advertisers.

Verb.AI pays users a flat $50/month for phone-activity access, mainly to feed traditional market research; Levitt argues the bigger opportunity is realigning incentives so a platform (e.g., Google or an AI assistant) and the consumer become 'teammates' who jointly negotiate with manufacturers for a cut of the actual sale (e.g., a $400-600 cash-back offer per brand), splitting the resulting surplus rather than the platform capturing all ad-placement revenue alone.

data-ownership-monetization

Media referenced

Why Do We Dislike Inflation? (working paper) - paper - Stantcheva's study on why people hate inflation more than economists think they should; finds it's driven by perceived wage stagnation, employer discretion, and inequity, not just the headline CPI number
Immigration and Redistribution study (six-country survey, ~20-25k respondents) - paper - Stantcheva's large cross-country survey on perceptions of immigrants and how those perceptions shape support for redistribution; the episode's central case study
Estate tax perceptions study - paper - Survey finding Americans believe about a third of people pay the estate tax when the true figure is under one in a thousand, and that correcting the fact sharply increases support for the tax
Climate change policy study (20 countries) - paper - Recent survey work finding people's climate policy support hinges heavily on perceived equity/progressivity of the policy, including carbon tax revenue use, more than on convincing them climate change is real
Zero-sum thinking study - paper - Recent research showing a 'zero-sum mindset' (belief that one group's gain requires another's loss) predicts a distinctive, cross-partisan bundle of policy views and correlates with a person's or their ancestors' history of upward mobility versus scarcity/oppression

Companies

Verb.AI - Tech startup discussed in the listener-question segment; pays users $50/month to access and monitor phone activity for market research, cited as a 'baby step' toward Levitt's vision of consumers being paid for their data
Social Economics Lab - Stantcheva's research lab (socialeconomicslab.org), referenced as where listeners can find her papers and survey methodology

Techniques and frameworks

Large-scale in-depth economic surveys - Stantcheva's core methodology: 20-25 minute surveys with thousands to tens of thousands of respondents, combining factual-perception questions with policy-view questions, designed over a year or more to avoid leading or confusing phrasing
Question-order randomization (priming) - Randomly varying whether respondents see factual/perception questions on a topic (e.g., immigration) before or after policy questions, to isolate the causal effect of having just thought about that topic on stated policy preferences
Information-treatment randomized experiments - Randomly giving different survey groups different information (raw facts, an anecdote/narrative, or nothing) before asking policy questions, to test whether facts versus stories move people's views
Benchmarking perception questions - Asking a comparison question (e.g., perceived unemployment rate among non-immigrants) alongside the target question (perceived unemployment among immigrants) to strip out a general definitional misunderstanding and isolate the actual perception gap

Summary

Steve Levitt talks with Harvard economist Stefanie Stantcheva, this year's John Bates Clark Medal winner, about the unconventional research path that made her one of the field's most influential young economists: asking people directly what they think, via large, carefully designed surveys, at a time when doing so was widely viewed within economics as close to career suicide. Economists have long trusted revealed behavior over stated preference, on the theory that people's actions are more honest than their words. Stantcheva pushed back on that norm, arguing that behavior alone can't reveal the constraints, misunderstandings, and reasoning that shape how people respond to policy - and built a research program around 20-25 minute, carefully piloted surveys run on tens of thousands of respondents to get at exactly that.

The bulk of the conversation walks through what these surveys reveal, using her six-country immigration study as the central case. Respondents in the US, UK, Sweden, Germany, Italy, and France dramatically overestimate the share of immigrants in their country (36 percent perceived versus roughly 10 percent actual in the US) and believe immigrants are far more unemployed and welfare-reliant than they actually are. Using randomized question ordering, Stantcheva shows this isn't just passive misperception: merely being asked to think about immigration, with no new facts provided, causally reduces stated support for redistribution broadly, not just toward immigrants, consistent with theories that generosity travels less well across group lines than within them. A companion randomized-information experiment then reveals a sharp asymmetry in what actually changes minds: giving people accurate immigration statistics does almost nothing to their policy views, while a simple anecdote about one hardworking immigrant meaningfully increases support for both immigration and redistribution.

That asymmetry doesn't hold everywhere, which is part of what makes the methodology powerful. On the estate tax, the pattern flips: Americans believe roughly a third of people pay it when the true figure is under one in a thousand, and simply correcting that fact sharply increases support for the tax, with a narrative element (a photo of a mansion) adding only a small effect on top. Similarly, a 20-country climate survey finds people are already largely convinced climate change is real, so the constraint on policy support is perceived fairness of design - carbon tax revenue that's returned progressively or earmarked for green investment polls far better than revenue that disappears into general government budgets. Across studies, Stantcheva and Levitt land on a shared conclusion: policy support is driven far more by fairness, distributional concern, and institutional trust than by narrow self-interest or clean economic logic, which is a genuine blind spot for a discipline that often assumes people respond rationally to costs and benefits.

The most novel finding discussed is "zero-sum thinking" - the belief that one person's or group's gain necessarily means another's loss - which Stantcheva's recent work shows is a distinct psychological axis that predicts an unusual, cross-partisan bundle of views: support for taxing the rich and support for restricting immigration tend to travel together in zero-sum thinkers, even though those positions don't sit on the same side of the traditional left-right spectrum. The mindset isn't a simple bias; it tracks people's real economic history. In the US, younger generations are more zero-sum than older ones, tied to lower growth and mobility in the era they grew up in, while the pattern reverses in developing countries now experiencing faster growth than previous generations did. It's also tied to family history of enslavement or oppression, suggesting zero-sum thinking is a rational-enough response to lived scarcity that can then get over-applied to situations that aren't actually zero-sum.

The episode closes with two lighter threads: Stantcheva's personal story growing up in communist Bulgaria and East Germany before her family settled in France, which she says gave her an early, pre-academic fascination with economics through everyday hyperinflation and currency-watching; and a listener-question segment with producer Morgan Levey about Verb.AI, a startup paying users $50/month for phone-activity access. Levitt frames Verb.AI as only a "baby step" toward his longer-standing idea that consumers should be paid for their data, sketching a model where a search engine or AI assistant becomes the consumer's negotiating "teammate" against manufacturers, splitting outcome-based payouts (e.g., cash-back on a car purchase) rather than just selling ad placement.

Notable Quotes

"For those who are non-economist, essentially we, economists, do not believe what people say. We only believe what they do or what they show you through their behavior." - Stantcheva

"It's naive at best and idiotic at worst to think that economic arguments are going to win the day when it comes to policy outcomes." - Levitt

"This is like the epitome of Trump's view on every issue I can think of - immigration, tariffs, everything." - Levitt, on zero-sum thinking

"It is not a bias or misperception. So people have lived through things that have been zero-sum and so they have adopted that mentality." - Stantcheva, on zero-sum thinking

"You and Google are essentially teammates who are trying to play off the auto manufacturers in a way that maximizes the joint surplus to you and Google." - Levitt, on getting paid for personal data