Raj Chetty
What this covers
Raj Chetty discusses his research philosophy and methodology in this conversation, arguing that understanding economic opportunity requires moving beyond data access to ask better questions and execute relentlessly. He covers work spanning childhood neighborhoods and teacher quality to corporate finance and consumer behavior, showing how natural experiments and administrative datasets can answer causal questions that were previously thought intractable. The conversation centers on how specific factors—where children grow up, who teaches them, whether communities are integrated—demonstrably shape adult earnings and life outcomes, and how to measure these effects despite the lack of randomized trials.
The discussion branches across several territories. On geography and mobility, Chetty presents evidence that neighborhood effects operate through exposure—each year matters—and that relative mobility has stalled while absolute mobility has collapsed, a distinction that reframes the American dream debate. On education, he argues that kindergarten teacher quality improves adult earnings through non-cognitive skills rather than test scores, and that early test performance explains only a quarter of adult variation, undermining broad tracking. On integration, he suggests socioeconomic mixing is likely causal, not merely descriptive of place quality. He also ventures into behavioral economics, explaining how salience shapes consumer responses to taxes and discounts, and into corporate finance, where he discusses dividends as commitment devices and the principal-agent problem in manager compensation. Throughout, he emphasizes that surprising patterns—Salt Lake City's upward mobility, girls' marriage rates in the Bay Area, stagnant teacher pay despite clear value-added differences—often reveal where intuition fails and where policy levers actually exist.
Chetty argues that the determinants of economic opportunity—neighborhoods, teachers, integration—are causal and measurable through creative use of big data, and that his research edge lies not in unique data access but in conceptual ingenuity plus relentless execution across every stage of research.
- Childhood environment causally shapes adult outcomes, demonstrated via randomized adoption studies and exposure-time effects
- Effective kindergarten teachers improve adult earnings by building non-cognitive skills rather than test scores
- Socioeconomic integration (low segregation) is a strong and likely causal predictor of upward mobility
Absolute mobility has collapsed while relative mobility stagnated despite offsetting investments, revealing a fundamental American economic decline.
- Relative intergenerational mobility (your odds of moving up or down in percentile ranks) has not changed significantly over the past ~40 years; this stasis reflects two offsetting forces—rising investment by affluent families in their children versus decades of War on Poverty programs (Head Start, college access) aimed at low-income kids—so the lack of improvement is itself surprising.
“relative intergenerational mobility... has not changed significantly over the past 40 years or something like that.”
- In contrast to flat relative mobility, absolute mobility—a child's chance of earning more than their parents, arguably the true 'American dream'—has fallen dramatically over the past 50 years.
“absolute mobility — your chance of doing better than your parents, which arguably is what people really think of as the American dream — that has fallen dramatically over the past 50 years.”
- While resource booms (e.g., North Dakota) can temporarily raise upward mobility, cross-place mobility differences are mostly not driven by economic growth: holding growth fixed, some places have much higher mobility (Atlanta booms yet has low mobility), and mobility rates are highly persistent over decades (Iowa stays high despite brain drain), indicating something durable rather than random booms.
“Atlanta is a city that's booming in terms of jobs and economic growth overall. But Atlanta's one of the places with the lowest levels of upward mobility for kids growing up in low-income families there.”
- Girls who grow up in affluent Bay Area families have low household income not because of low individual earnings—their individual earnings are very high—but because they are much less likely to be married, with each additional year of childhood in the Bay Area lowering marriage likelihood, so marriage-pattern differences, not earnings, explain the household-income gap.
“girls growing up here in the Bay Area do extremely well. However, when you look at household income, they don't do so well, and that's because they're much, much less likely to be married”
- Areas with concentrated poverty (e.g., Baltimore) produce especially poor outcomes for boys relative to girls—boys growing up there lose roughly 30 percent of earnings versus an average place—likely because boys are more drawn into crime and gangs, whereas girls' outcomes are only slightly negative.
“growing up in a place like Baltimore turns out to be extremely detrimental for boys. We estimate that you lose something like 30 percent of your earnings relative to if you've grown up in an average place in America.”
Neighborhoods shape adult earnings through childhood exposure duration; integrated places with cross-class contact produce better outcomes than segregated ones.
- Chetty's group advanced the 30-year neighborhood-effects literature by showing the mechanism is exposure: every year of childhood spent in a better neighborhood matters, an intuitive hypothesis they documented clearly enough to change how the field studies neighborhoods.
“every year of exposure during childhood seems really critical, something that struck us as being an intuitive hypothesis, but then figuring out how to document that clearly in the data”
- Places that are more integrated across socioeconomic groups (lower segregation) tend to produce better outcomes for children; small towns like those in Iowa force shared schools and activities and thus more cross-class contact, whereas big cities allow self-segregation via gated communities and private schools—and Chetty believes this integration is likely causal, not just a restatement of place quality.
“places that are more integrated across socioeconomic groups, that have lower segregation, tend to have better outcomes for kids.”
- Technology like driverless cars could deepen residential segregation by making it easier to live further away in secluded places and reduce cross-class interaction, so urban planning and architecture must be deliberately designed to make mixed-income communities attractive rather than passively facilitating more segregation.
“if you have access to driverless cars, it makes it all the more easy to go live further away in a secluded place, further reduce interaction”
- Salt Lake City, Utah consistently ranks at the top of US data for upward mobility and health outcomes despite not being an obvious economic powerhouse, plausibly due to community cohesion and the Mormon Church—an outcome that was hard to predict ex ante.
“Salt Lake City, Utah, is maybe not a place you'd think of as having the most vibrant economy and so forth, but it's a terrific place for upward mobility. It always shows up in the top of our data”
- Technology has transformed private life through the private sector but has had relatively little impact on social issues to date, creating a large 'arbitrage opportunity' to apply technology to problems like upward mobility and urban integration.
“technology has completely transformed the way we live our lives through the private sector. I think it's, to date, had less of an impact on social issues, and there's a big arbitrage opportunity there.”
Teacher quality, especially early kindergarten, builds durable non-cognitive skills that predict earnings; performance measures outperform experience for evaluating teacher impact.
- Current US teacher pay rises continuously with experience and ignores performance; since value-added and other performance measures predict teacher impact far better than experience beyond the first few years, tying pay, bonuses, or retention efforts to performance measures would be more sensible.
“tying pay or bonuses or efforts to retain teachers to performance measures — whether they're test-score-based, value-added measures, or principal ratings or something else — is quite sensible. That is not something that's currently done in most US school systems.”
- High value-added and more-experienced kindergarten teachers improve students' adult earnings even though the test-score gains fade by later grades; the best hypothesis is that teachers impart durable non-cognitive skills—discipline, social skills, getting along with others—that are rewarded in the labor market rather than academic content.
“Our best hypothesis is that, what's going on is that the teacher is not teaching you skills that end up showing up on later tests... Rather, it's that they are teaching you things like discipline and social skills and other things that are valued in the labor market directly.”
- Kindergarten teaching quality plausibly declined as high-skilled women gained better labor market opportunities; earlier cohorts, when teaching attracted more high-skilled women, had higher teacher quality, consistent with the finding that teachers with 10+ years of experience (from earlier cohorts) appear better.
“the quality of teachers in earlier cohorts, where women were more likely to go into teaching — high-skilled women — was much higher. So I think that's plausible.”
- Using New York City data with 2.5 million observations, teacher effectiveness improves most in the first two years of experience and then follows a roughly flat profile; the original above/below-10-years split in the kindergarten STAR study was merely a statistical convenience for a small sample, not evidence that experience matters only after a decade.
“it's the first couple of years of experience that matter the most, so novice teachers don't do as well as slightly more experienced teachers, but then you see essentially a flat profile.”
Early test scores predict outcomes only modestly; broad tracking is unjustified but gifted programs for disadvantaged top performers yield large spillovers.
- Early childhood test scores have real but limited predictive power for adult outcomes—explaining perhaps a quarter of the variation seen in adulthood—so they are strong predictors but do not determine a child's fate, which is why broad early tracking is unjustified.
“You can explain maybe a quarter of the variation that you'd see in adulthood with early childhood test scores and things like that.”
- Early tracking is detrimental across most of the distribution because test scores imperfectly predict trajectories and children's outcomes are not fixed early; but at the very top tail, gifted-and-talented programs—especially for disadvantaged kids—can be highly effective and justified given the large positive spillovers of developing future innovators.
“while I would not advocate tracking at the broadest level across 95 percent of the distribution, I think there can be a role at the very top to give attention to gifted and talented kids, especially, I would say, from disadvantaged backgrounds”
- Non-cognitive skills such as discipline and focus—of the kind drilled by strict early schooling—appear highly predictive of later life outcomes in Chetty's research.
“those kinds of non-cognitive skills — something we see in our research — discipline, focus, things like that — seem to be highly predictive of later outcomes”
Environment shapes life trajectories even conditional on genetics, as adoption studies and quasi-random peer composition experiments demonstrate causally.
- Studies of adopted children, especially Bruce Sacerdote's work on Korean adoptees effectively randomly assigned to families, provide clear causal evidence that conditional on genetics, environment matters greatly: children placed with more educated, affluent families in better neighborhoods do substantially better in adulthood; twin studies are less convincing because the variation driving one twin's outcome versus another is not experimental.
“he shows very convincing evidence that, if you happen to be an adopted kid who ends up in a family that's more educated and more affluent and so forth, you do substantially better in adulthood than if you end up randomly assigned to a less educated, less affluent family”
- Social network data like Facebook offers the most precise lens yet for measuring social capital and integration; a path to causal inference is to exploit that children befriend same-grade peers but not adjacent grades, so the random grade-to-grade variation in peer composition acts as quasi-random exposure to test whether one's social connections affect life trajectory.
“kids tend to be friends with children in the same grade, but not friends with kids who are one grade ahead or one grade behind... that's essentially going to be a random variation from the perspective of a given child.”
Behavioral framing—salience, commitment devices, tax visibility—shapes financial and consumer decisions more than rational magnitude considerations alone.
- Although dividends are tax-inefficient relative to share repurchases, they persist because they function as a commitment device that constrains managers from empire-building with excess cash; cutting a dividend sends a strongly negative signal, so managers are forced to return money to shareholders.
“you need to constrain managers — the amount of capital managers have on hand — because otherwise there's a tendency to empire-build and invest in projects that aren't as efficient from the shareholder's perspective”
- Managers who own a larger fraction of shares—and whose interests are therefore more aligned with shareholders—tend to pay out more dividends, providing a rational explanation for cross-firm variation in dividend behavior.
“managers who own a larger fraction of shares and thereby have interests that are more aligned with shareholders', they tend to pay more dividends — exactly as you might think.”
- Retail sales work primarily through salience rather than the actual dollar discount: the 'on sale' tag itself attracts attention and triggers purchases, more than the magnitude of the price reduction does.
“It's not so much the dollar price reduction. It's the fact that the thing says it's on sale, I think, that attracts people's attention and makes them want to buy.”
- Consumers respond less to a tax shown as a separate sales tax added at the register than to the same cost embedded in the displayed price, because people pay little attention to taxes day-to-day; companies favor the add-on form because including the tax in the price would force them to cut prices and bear part of the tax burden.
“Companies recognize that, when the tax is added at the end, it's in their favor because consumers don't cut back on demand as much as if the tax was included in the price, and that would force companies to basically cut prices and bear some of the costs of the tax.”
- Checklists are highly useful for staying organized and tackling complex problems; the way to make progress on complex research is to divide it into smaller manageable problems and work down a list solving each.
“The way to make progress on complex research problems, for example, is to divide them into smaller, manageable problems, and then go down a list and solve that checklist.”
- Martin Feldstein's practice of granting Chetty independence at age 17-18—encouraging him to pursue his own ideas rather than assigned RA tasks—built confidence and was formative, suggesting early autonomy from a respected mentor has outsized developmental value.
“Hearing that when you're 17 or 18 years old from somebody like Marty Feldstein builds a lot of confidence and was very inspiring.”
- Chetty thanks Cowen for having him on the show.
“thanks, Tyler, for having me. Pleasure to be here.”
- Raising the status of teaching is highly important for K–12 quality; in the US, top-college graduates avoid teaching due to both pay cuts and low prestige, and Teach for America succeeds at attracting top talent precisely because it is hard to get into—talented students are drawn to difficult-to-obtain positions—though they tend not to stay.
“the reason it's so effective in attracting top talent is precisely because it's very difficult to get into Teach for America. I think kids at top institutions like to do things that are hard to do, basically.”
Economic growth alone does not drive cross-place mobility differences; durable cultural or institutional factors like community cohesion and institutional strength matter.
- A region's tradition of mathematical prodigies, exemplified by Ramanujan in South India, may perpetuate itself because children grow up hearing about and looking up to such figures, and this role-model mechanism strongly influences what children choose to pursue.
“you look up to that kind of thing, and maybe that influences subsequent generations of children as well. I think that kind of mechanism is actually extremely important in terms of what children choose to do.”
- In developing countries historically, families lacking funds to educate all children would pick one child to educate, and that single decision had multi-generational effects on opportunity—Chetty's parents were each the chosen child in their villages, shaping outcomes for the next generation.
“what was common in developing countries at that time was to pick one child to educate because most families didn't have enough money to educate all of their kids.”
- The Baumol cost-disease story does not yet bind US K–12 education because large productivity gains are available without major cost increases—e.g., retaining the most effective teachers while retraining or dismissing the least effective—and high-performing systems like Scandinavia and certain affordable colleges show good outcomes need not require dramatically higher spending.
“if we were to try to keep the most effective teachers in the classroom and either retrain or dismiss the teachers who are less effective, we could substantially increase productivity without significantly increasing cost.”
Research impact multiplies when findings are written with extreme clarity and accessibility for non-expert audiences rather than technical narrowness.
- Chetty's competitive advantage is not unique data access or technique—both replicable—but combining classic important questions with newly available big data, then making a conceptual advance that lets the data test something others missed, executing well across numerous stages (data, concept, team management), and writing with extreme clarity.
“What hopefully our contribution and scale is, is showing how you can take those large datasets and not get lost in them, and bring out the key lessons that are relevant for thinking about these classic questions.”
- Writing economics papers with clarity and accessibility—so findings can be explained to non-experts—broadens the set of readers and increases impact, contrasting with technically oriented papers that narrow the audience; Chetty attributes his appreciation for this clarity to his science background.
“There are ways to write papers in economics that are more accessible to the public and thereby have greater impact, and there are ways to write papers that are more technically oriented and narrow the set of readers.”
- Because US data linking parents' and children's income only exists for kids born in the 1980s, the 'fading American dream' paper's key conceptual advance was devising a statistical method to estimate the fraction of children earning more than their parents going back to 1940 despite the inability to directly link parents and kids.
“the key conceptual idea of that paper was to figure out a way, even though you could not link parents and kids, a statistical method of determining the fraction of kids doing better than their parents despite that data limitation.”