Daniel Carpenter
What this covers
Daniel Carpenter, the leading expert on the FDA, ranges across medicine, history, and political economy in this conversation. The spine of the discussion is a counterintuitive claim about what approval regulation actually does: rather than being primarily about safety, it creates indispensable incentives for rigorous evidence—randomized controlled trials above all—and sustains the credibility that makes treatments work in practice. This rests on a prior claim that drugs are credence goods, not like cars that drivers can evaluate through use; because patients cannot know whether a drug caused their cure, collective belief and public trust are wired into the product's actual value. Cut regulatory corners, and you erode that trust rationally and measurably, not through psychological panic.
The conversation moves outward into the prerogatives and failures of regulatory agencies themselves. Carpenter contests the overuse of regulatory capture as an explanation—it is hard to prove, he argues, and patterns favoring large firms often arise from uncertainty avoidance rather than corruption. He defends approval regulation against a Good Housekeeping seal alternative by pointing out that the former preserves incentives for ongoing study of long-term drugs, while the latter would not. But he also acknowledges real costs: labeling decisions on products like SSRIs can drive prescribing patterns in ways that harm users, and the FDA was plausibly too slow on COVID-19 testing. The discussion branches into 19th-century petitioning history, the power of the postal system as a platform for democratic culture, and the distinction between directed and plebiscitary petitions as forms of voice—territory that reveals how trust, credibility, and institutional design shape outcomes across domains.
Carpenter argues that FDA-style 'approval regulation' is not primarily about safety but creates indispensable incentives for credible evidence (RCTs) and that regulatory credibility, corporate reputation, and public trust are deeply endogenous to whether medical products actually work in practice—so cutting procedural corners can backfire by destroying the trust that makes treatments and vaccines effective.
- Drugs are credence goods, not experience goods, so collective beliefs and credible evidence are central to a product's value
- Procedural irregularity at the FDA can produce rational hesitancy (e.g. insurers refusing to cover Exondys) and drops in vaccine confidence
- Capture is overused as a term and hard to prove; agency behavior favoring large firms can arise from uncertainty avoidance with zero capture
Regulatory capture is overused rhetoric that obscures the role of uncertainty avoidance and institutional credibility in agency decisions.
- An asset-pricing study by Brian Libgober found that banks which met with the Fed before a proposed rule enjoyed a statistically and financially significant abnormal positive stock return in the first hour after the proposed rule was announced—totaling roughly $40–70 billion in positive abnormal returns from those meetings alone.
“He totaled up, I think, $40–70 billion of positive abnormal returns just from those meetings with the Fed”
- Patterns where larger, older firms enjoy quicker or more favorable regulatory decisions than smaller, newer firms can arise even when the regulator has zero implicit or explicit preference over firm type—because of uncertainty avoidance rather than capture or even risk avoidance.
“that can be true even when the underlying regulator... has zero implicit or explicit preference over the kind of firm it’s regulating.”
- 'Capture' is too seductive a term—thrown around whenever a pundit dislikes a regulatory decision—but it is orthogonal to whether an agency does a good or efficient job, and it is genuinely hard to prove because you need a counterfactual of an uncaptured agency ceteris paribus, while existing empirical studies suffer from omitted-variable bias.
“the problem with capture is it’s too sexy a term.”
- The Nuclear Regulatory Commission is a plausible case of capture, partly through professional training: in France, regulators come through formidable technical schools that incline them to view nuclear power favorably—suggesting capture can arise from how regulators are trained rather than from designed corruption.
“one agency that has been captured or plausibly captured would be the Nuclear Regulatory Commission.”
- Academics benefit from an activity so absorbing they almost cannot think about work while doing it; fly fishing serves this purpose for Carpenter, who rejects the framing of it as a Zen practice and values it instead for its absorption in beautiful places.
“sooner or later, you need something where, when you do it, you can’t think about work.”
Deliberative institutional design including legislative petition days strengthens democratic voice and agenda-setting beyond plebiscites rewarding viral momentum.
- Making state-level referenda too easy is problematic—especially allowing constitutional amendments by a bare 50-plus-epsilon majority of whoever shows up in a small-turnout election—because plebiscites weaken the separation of powers, deliberation, investigation, and compromise by forcing a binary choice between a proposition and the status quo with no third option; directed petitions are institutionally more valuable than plebiscitary ones.
“it weakens the separation of powers. It weakens deliberation.”
- Modern petitions—especially electronic and plebiscitary ones—reward select groups that know how to make something go viral quickly (as with the Bank of America card-fee petition), but the US has lost the legislative petition institution that once structured robust public debate, which Carpenter regards as a problem even when individual viral petitions are effective.
“What you’re going to see with petitioning nowadays... is that select groups and interests that know how to make something go viral.”
- In the crucial 1820s-1850s period across the US, Canada, and parts of Mexico, the petition let people without existing organizations, parties, lobbies, or interest groups make their voices heard, set the agenda, and organize—so that anti-slavery societies and new parties like the Canadian reform parties and American Whigs drew much of their energy and structure from early petitioning campaigns.
“people who didn’t already have organizations... were able to successfully harness the petition to make their voices better heard, to set the agenda”
- Petitions were key to 19th-century women's rights movements internationally because women signed not just individually but collectively, gaining what Carpenter calls 'the accountability of voice'—the sense that their voices mattered through both their arguments and the sheer weight of numbers—which motivated bolder and bolder claims for the vote.
“what I call the accountability of voice, the idea that their voices could matter, not only in the arguments they put forth, but also in the weight of the numbers”
- Legislatures at all levels should consider reintroducing petition days—as governed the US until the 1950s and as New Zealand and Canada still do—because, though inefficient, they would make the democratic republic more deliberative, increase social trust by giving people a sense of being heard, and open up legislative agendas whose closure (which makes parties efficient) often keeps the best ideas off the policy table.
“I don’t think it would be such a bad idea if legislatures at all levels of government had to reintroduce petition days such as governed the United States until the 1950s”
- Early 19th-century French Canada was central to petitioning history: a single petition of roughly 90,000 names—among the largest per capita in the Atlantic world—helped fight off colonial fusion (explaining why Canada remains bilingual), deposed Governor Dalhousie via mass protest rather than elite channels, and inspired British chartists and working-class reformers who called the French Canadians 'our brothers in oppression.'
“a single petition with over 90,000 names attached — excuse me, 87,000 names... that was signed in what is today Quebec”
FDA approval regulation creates powerful R&D incentives and ongoing post-market study that weaker information-only regimes cannot sustain.
- A key advantage of approval regulation over a Good Housekeeping seal model is that it preserves incentives for continual study of products well after approval—important for drugs like Prozac or statins taken for decades—whereas an information-only regime would not generate ongoing study.
“what I want, also, is a world where the unknown, deeply unknown product studied or being potentially approved gets studied massively, thoroughly, and where the studies continue well after approval.”
- FDA 'approval regulation'—the FDA acting as a veto player granting access to profitable markets—creates very powerful incentives for R&D and for running randomized controlled trials; a Good Housekeeping seal-of-approval model would weaken or eliminate those incentives because firms could simply sell offline and advertise their way into the market, as the patent medicine industry did.
“the FDA as a veto player granting access to profitable markets creates very, very powerful incentives for R&D.”
- Drugs are credence goods, not experience or inspection goods like cars: even if cured, a patient cannot really know whether the drug caused the cure, so collective beliefs and Bayesian priors about whether treatments work are central and deeply endogenous to the value of the product itself—making the car/R&D analogy inapplicable.
“cars are experience goods, maybe even inspection goods, and drugs are credence goods.”
Civil service merit protections and reduced political appointees insulate agencies from authoritarian capture better than ethics laws targeting revolving doors.
- Modernizing the Pendleton Act today should mean reducing the number of political appointees at the top of agencies—keeping presidential appointment of agency heads as an Article II power, but restricting the many secondary appointees (communications, undersecretaries) whose added value is doubtful.
“I would actually reduce the number of political appointees at the top levels of the civil service.”
- Newer agencies like the Consumer Financial Protection Bureau are less likely to be captured when set up with the right staffers and merit-service protections, reflecting Marver Bernstein's 'life cycle of agencies' view that capture evolves over time in agencies that were not originally captured, rather than being designed in.
“An agency that is newer and less captured would probably be the Consumer Financial Protection Bureau.”
- Among ethics laws, revolving-door constraints may be overrated—there's evidence firms seeking talent are most likely to hire the toughest early-career regulators—while restrictions on small favors are underrated, because the psychological literature shows small gifts are sometimes the most powerful influence.
“The psychological literature has demonstrated that small gifts are sometimes the most powerful ones.”
- An independent civil service that cannot be commandeered by a party or a leader with authoritarian tendencies is valuable, and the prior presidential administration demonstrated that value—while still needing to avoid the worst public-employee union abuses seen in some police departments.
“the last presidential administration showed us the value of having an independent civil service, one that cannot be used by a party or a leader who verges on authoritarian energies”
Pre-notice-and-comment agenda-setting and early industry meetings overwhelmingly favor large firms, requiring broader stakeholder embedding rather than exclusion.
- Most regulatory influence occurs before notice-and-comment begins: studies of the Volcker rule found hundreds of pre-proposal meetings, roughly 90 percent of which were with large bank holding companies (JP Morgan, Goldman, Citi) and their law firms and associations, not smaller regional banks—an early agenda-setting stage the Administrative Procedure Act does not regulate at all.
“Pre-proposed rulemaking, earliest agenda-setting meetings are dominated by industrial interests.”
- Rather than removing industry meetings, rule writers should be induced to more regularly meet with consumer, labor, and other advisory committees, because embedding interests that wouldn't otherwise be represented opens up the conversation—even if it's uncertain whether outcomes become objectively better or worse.
“I do think we need to embed consumer and nonindustry perspectives more.”
- Building/zoning restrictions are better understood as housing policy than as part of a unified theory of regulation; coastal and certain Southeastern places—driven partly by racial, ethnic, and wealth discrimination—will crack down more heavily on new residential building, even if rural depopulating areas may have more freedom to build.
“I see what you’re talking about as a kind of regulation, although to me, I see it much more as a kind of housing policy.”
Vaccine confidence and product credibility drive adoption as much as supply; weak institutional credibility produces rational hesitancy distinct from psychological resistance.
- When the FDA cuts procedural corners or approves on weak evidence—as in the controversial eteplirsen/Exondys approval for Duchenne muscular dystrophy over advisory-committee and division-director objections—it can produce rational hesitancy among informed purchasers, exemplified by insurers refusing to cover the drug because they don't know whether it works; this differs from psychologically driven vaccine hesitancy.
“when you have procedural irregularity or corners cut or bad evidence underlying approval decisions is, you get, in some cases, what might be rational hesitancy on the part of treatments.”
- The correlation between vaccine-hesitancy factors in public opinion and first-dose vaccination of the eligible US population is 0.6—an extremely high correlation—indicating that vaccine confidence, not just supply, is a very strong predictor of actual vaccinations, contrary to a story that getting good vaccines out quickly alone drives uptake.
“the correlation between vaccine hesitancy factors as expressed in public opinion and first-dose vaccination of eligible population in the United States is 0.6.”
- Rushing vaccines out in summer or fall 2020—when expressed willingness to be vaccinated had dropped 10-15 percentage points, concentrated among the young who are most needed for herd immunity—would have produced a social catastrophe of tens of millions refusing the vaccine, costing thousands of lives, because vaccines with positive externalities are only as valuable as the share of the population that actually takes them.
“you would have had, I think, a social catastrophe — tens of millions of people not taking the vaccine.”
- Vaccine hesitancy toward AstraZeneca is not merely a regulatory construct created by politicians badmouthing it; AstraZeneca had genuine credibility issues all along—managerial and manufacturing problems, study-design and data-reporting issues flagged by its own Data Safety Monitoring Board—demonstrating that corporate reputation matters for whether a product is accepted and used.
“AstraZeneca has had credibility issues all along.”
- Pharmaceutical trials in poorer countries should be allowed, but wealthy-country regulators are right to decline to rely on them as pivotal when a country has weak scientific and credibility-based institutions—as evidenced by repeated FDA crackdowns on quality problems among Indian generic-drug manufacturers; institutional credibility of scientific institutions matters as much as broader institutional quality.
“If we think that a country has weak scientific and credibility-based institutions, then I’m fine with allowing the trials there, but I’m also fine with regulators... saying, ‘Yeah, we’re not going to accept those trials as pivotal for approval.’”
Universal public services like the postal system require cross-subsidizing mechanisms to prevent geographic and cultural concentration in wealthy areas.
- Fully privatizing the postal system without a universal service standard would accentuate and accelerate the bicoastal concentration of culture, wealth, talent, and economic innovation, cutting off tens of millions of people—especially rural users like Express Scripts customers who rely on the mail rather than FedEx—from a part of their culture they still use.
“you would accentuate and accelerate the bicoastal concentration of culture, wealth, talent, and economic innovation in the United States”
- The 19th-century postal system's near-zero cost for letters cross-subsidized newspapers, which was crucial because newspapers created a far more informed society and served as a platform for advertising and broad cultural dialogue—an open-access model that anchored American political culture, economy, and education.
“The letters subsidized the newspapers, which was crucial because the newspaper created a much more informed society.”
- Progressives saw the Postal Service as the exemplar of an open-access universal public service—something the national government could actually do well, providing a platform for democratic deliberation, markets, and a more universalized cultural dialogue—at a time when most other functions (even the army) were decentralized to states and militias.
“Progressives saw at one moment in the Postal Service, well, here’s actually something that the national government can do”
Surgical innovation decentralized through tort and professional societies rather than FDA approval suggests regulatory design varies across domains without settled health-benefit superiority.
- Surgical innovation is regulated very strongly but in a decentralized way—primarily through the tort system, juries, and medical professional societies rather than FDA approval—and although new surgical devices still require FDA approval, it is unclear whether surgical innovation has delivered more aggregate health benefits than pharmaceutical innovation under that biased, state-level regime.
“in the case of surgery, it’s the tort system that regulates that.”
- At the federal level, employee unions are politically weak because, outside a few western states and parts of Virginia, they are not strong enough to affect the electoral calculations of a congressional majority—unlike state and local public-employee unions.
“At the federal level, I actually think the federal employee unions are pretty weak.”
- FOIA laws are underrated, especially strongly enforced, as one of the most important laws against creeping authoritarianism—though they changed an equilibrium by removing reporters' need to cultivate agency insiders, a relationship that had made reporters better informed about agency workings.
“FOIA is — and strongly enforced because it was just weakened by the Supreme Court — is one of the most important laws that we have.”
- Native Americans petitioned at very high per-capita rates in the 19th century—on a per capita basis perhaps second only to the French Canadians—yet political scientists understudy them despite there being millions of indigenous people with great variation in tribal governments, especially among Alaska First Nations.
“on a per capita basis — maybe the French Canadians aside — few people are petitioning as much as they are.”
- The narrative that regulation is a one-way ratchet that only increases is easily falsified in the US: governments now regulate prices far less, antitrust/M&A regulation has declined, environmental deregulation has occurred, and even parts of the FDA's medicines approval have been deregulated.
“I think that narrative is pretty easily falsified in the United States”
Labeling and off-label regulation decisions carry unquantified health tradeoffs that cannot be reduced to simple cost-benefit calculations.
- The current FDA system has substantial flexibility around off-label prescribing—the FDA approves a molecule knowing off-label use will follow—but whether something is on the label remains a significant decision (as the bevacizumab breast-cancer case showed) that industry fought hard over because money was at stake; in oncology, a tightly closed research community governed by ASCO further shapes practice.
“That’s not to say, however, that off-label prescribing is not regulated. Whether something is on the label or not is still a significant decision.”
- The FDA was plausibly too slow on COVID-19 testing—and requiring a prescription for tests made no sense under any worldview—but this does not verify Sam Peltzman's critique, because the testing division is not the drug-approval division Peltzman studied, and Peltzman's original argument was about new chemical entity introductions and the research process, much of which came from Congress (e.g., the 1962 Kefauver-Harris amendments).
“I don’t think it verifies Sam Peltzman, and it’s not clear to me”
- Black-box drug labeling definitely changes behavior, sometimes for the worse—adding suicidality warnings to SSRIs led psychiatrists and primary-care physicians to prescribe them less to teenagers and adolescents, which one argument links to increased suicides—so labeling decisions require a perspective weighing the health costs, not just a simple cost-benefit analysis.
“black-box labeling definitely changes behavior, sometimes for the worse.”