What this covers
Nick Bloom discusses why scientific and economic progress, despite growing research investment, is slowing measurably while the true gains available today lie in diffusing known management practices worldwide. The conversation spans his research on why firms fail to adopt obvious improvements, how organizational inertia blocks the payoff from transformative technologies, and where legal systems become the bottleneck preventing poor countries from scaling modern management. Bloom draws on his empirical work measuring Management Practices Measurement across firms and countries, his studies of Adjustment Costs, and his work with collaborators like John Van Reenen and Raffaella Sadun on why American multinationals outperform their foreign peers.
The conversation ranges across several distinct claims. Research productivity has fallen roughly 5% annually—maintaining progress on Moore's Law alone now requires 18 times more scientists than in the 1970s. General-Purpose Technology like electricity and information technology do not automatically boost productivity; organizational restructuring must follow, which happens slowly, as with electricity's delayed reshaping of factory design. Lean Production systems pioneered by Toyota were superior to Fordist mass production but their advantage was non-obvious, illustrating why management consultants add real value. The conversation turns contentious on whether weak legal systems prevent large professionally-managed firms in poor countries—Bloom argues Trust as Legal Enforcement matters far more than cultural factors, contradicting predictions that China could not grow large firms. He also disputes whether unmeasured quality-of-life improvements explain the productivity slowdown, argues that full remote work threatens long-run innovation despite short-run gains, and explains why the shift from government-funded basic research to private development spending threatens long-run growth.
Bloom argues that scientific and productivity progress per researcher is steadily declining because ideas get harder to find, while the large measurable gains available today lie in diffusing known management practices across firms and countries—a diffusion bottlenecked by weak legal systems and slow organizational adjustment.
- Research productivity falls ~5% a year; holding output constant now requires vastly more scientists (e.g., 18x for Moore's Law).
- American firms outperform largely through superior management, not tax arbitrage, and these practices transfer abroad.
- Management improvements don't scale in developing economies mainly because weak courts undermine the trust needed for large professionally-managed firms.
Research productivity per scientist is falling despite massive resource increases into R&D.
- US productivity growth peaked around 3-4% a year in the 1950s and has been falling since; the effective research productivity per researcher is declining at roughly 5% a year, offset only by pouring ever-more resources into R&D.
“technological progress, in terms of productivity growth, peaks in the 1950s at something like 3 to 4 percent a year, and then it’s been falling ever since.”
- Sustaining Moore's Law (doubling transistor density every two years) has required pouring in vastly more researchers—about 18 times more scientists since the 1970s just to hold the rate of progress constant—demonstrating falling research productivity per scientist.
“We estimate since the ’70s, there are 18 times more scientists just to hold that constant.”
- Because the stock of existing knowledge keeps growing, researchers must train longer before reaching the frontier (economics PhDs went from four years to six plus pre-docs), which contributes to aggregate research slowdown—a 'death of the Renaissance man' effect.
“We’re taking three or four years longerjustto get to the research frontier. There’s so much more knowledge before us, it just takes longer to train up.”
- Progress in any individual scientific or technological field eventually slows as it matures, and since the 1950s in the US new fields have not been emerging fast enough to offset the decline of maturing ones, producing an aggregate slowdown in scientific progress.
“It seems that field after field eventually starts to decline. And there aren’t enoughnewfields that are growing to offset the bulk of the current fields that are declining.”
- The share of US R&D funded by government has flipped from about two-thirds in the 1960s to roughly one-third today; since government R&D concentrates on basic research (the 'R') with larger spillovers while private R&D favors development (the 'D'), this shift reduces the long-run-growth-driving research that benefits everyone.
“in the US, when you go back to the ’60s, roughly two-thirds of it is funded by the government and one-third by private firms. Now it’s the reverse.”
In-person interaction is essential for generating novel ideas and collaboration across fields.
- Bloom credits broad rather than deep reading—scanning every NBER paper title and abstract weekly, reading the Economist, and listening to podcasts on daily walks—as a key source of new research ideas, alongside in-person seminars and serendipitous coauthor conversations.
“I really like to read broadly rather than deeply — sounds an odd thing to say.”
- Working from home raises short-run productivity for those who can do it (given good internet and a dedicated room), but full-time remote work for many months threatens long-run productivity because innovation depends on in-person 'water-cooler' sparks; a hybrid of roughly three office days and two home days captures the best of both worlds.
“in theshortrun, that increases productivity... what’s the impact on longer-run productivity, which is the concept... about creation and innovation.”
- The initial spark or idea for collaboration is much more effectively generated in person—often over lunch or coffee—than remotely, after which the work can be continued by email; this is why Bloom predicts a drop in patenting in 2021-2022 if full remote work persists.
“the initial spark or idea is much more effectively generated in person. Often, it’s over lunch or over coffee.”
Superior management practices drive productivity but require functional legal systems to scale.
- The measured productivity advantage of American multinationals is driven primarily by superior management practices (information-driven improvement processes and meritocratic HR) rather than transfer pricing or offshore tax manipulation; these intangible management capabilities travel abroad and generate large profits, explaining 'dark matter' in trade and investment balances.
“American multinationals around the world are well managed, and they make a lot of profits where they’re located... American firms are primarily driven, actually, by better innovation and better management.”
- Management improvements fail to scale in developing economies primarily because weak legal systems—e.g., Indian courts so undersupplied with judges that cases stall—make it impossible to enforce contracts or stop money leaking illegally out of firms, so even private equity buyers like Blackstone cannot turn companies around despite improving practices.
“the big constraint is processing cases through the courts. The courts are dramatically undersupplied in terms of judges”
- Well-run companies tend to share identifiable inputs: large size, high exposure to competition, openness to international trade, educated employees, and an absence of inherited family control by primogeniture—and China's manufacturing sector now ticks most of these boxes, making its manufacturing management quite good.
“a lot of what drives good management practices is being large, being around for a while, being open to competition, and having educated employees. And China has those inputs.”
Knowledge spillovers are geographically localized within firms and universities, not global.
- Knowledge spillovers are localized—people are far more likely to coauthor with colleagues in their own university or firm—so the transmission of ideas across countries (e.g., China to the US) is less effective than within a country, complicating the assumption that rising global R&D fully offsets falling US public R&D.
“there’s plenty of evidence on knowledge spillovers being localized... you’re more likely to coauthor with your colleagues in your own university or the same firm.”
- Lean production—stopping the line to fix defects immediately, pioneered by Toyota out of postwar capital scarcity—was a slow-burn system that by the 1980s delivered lower costs and higher quality than Henry Ford's mass production, yet its superiority was not obvious to most firms, which is precisely where management consultants add genuine value by transferring such non-obvious practices.
“it isn’t obvious to many firms that lean was a far better way to run your factory... consultants come in and help you adopt them.”
Weak rule of law forces family-based firms that cannot scale; strong legal systems enable growth.
- Interpersonal trust that enables large firms derives mainly from rule of law—reliable contractual enforcement—rather than from cultural or religious traits; World Values Survey trust is highly correlated with legal-system effectiveness, which is why China can grow large firms despite Fukuyama's prediction, while weak-law countries force reliance on less-competent family members.
“It’s not based in some cultural, religious thing. It’s based on the fact that the legal system works.”
- In countries with weak legal systems where outsiders can't be trusted, founders staff firms with family members for trust reasons, but as firms grow they exhaust competent relatives and must appoint less-capable ones, structurally limiting how large firms can grow without professional managers.
“I start to stuff it full of sons, daughters, brothers... these people aren’t naturally the best managers to run the place.”
National productivity depends on policy stability, education, regulation, and brain-drain retention.
- Italy's near-total lack of per-capita income growth for two decades is driven by poor government feeding into bad policy—burdensome regulation, political instability, weak education, and a large negative brain drain of able workers to the US and UK—rather than any single firm-level deficiency.
“a lot of it comes down to poor government that then feeds through into all these policies that make it hard for firms to innovate.”
- Pre-COVID economic performance under Trump roughly netted out to neither higher nor lower growth than Obama's second term: the corporate tax cuts aided growth, but heightened policy uncertainty from Trump changing his mind and not listening to advisers was a drag, and US physical investment was weak even before COVID.
“It seems to have netted out to about zero. It was no higher or no lower than in Obama’s second term.”
General-purpose technologies require slow organizational and social reorganization to boost productivity.
- Information technology has not boosted measured productivity much because exploiting a general-purpose technology requires reorganizing firms and society, and organizational practices and social norms change slowly—just as electricity took decades to reshape factory architecture from tall steam-driven buildings to flat, electric-motor layouts.
“you need to change society in order to exploit this... firms and society don’t change that rapidly.”
- Technology has improved quality of life in ways not captured in productivity figures—e.g., cameras, ID cards, and online tracing have sharply reduced football hooliganism and racism at UK grounds, raising attendance—but Bloom judges that such unmeasured quality-of-life gains do not primarily explain the measured productivity slowdown.
“technology’s improved attendance at sports games... That’s something that doesn’t show up in productivity figures.”
High adjustment costs amplify the harm of uncertainty because firms cannot easily reverse decisions.
- When adjustment costs are high—when it is expensive to hire and fire or invest and disinvest—uncertainty becomes especially costly because firms cannot easily change their minds, which is the link between Bloom's early work on adjustment costs and his later work on uncertainty; the same inertia means much of pandemic-era remote work will stick.
“if you have high adjustment costs — as in, it’s expensive to hire someone and fire, invest and disinvest — uncertainty is really costly because you can’t change your mind.”
- People and firms underinvest in cheap high-return improvements not because they're irrational but because of 'unknown unknowns'—they don't know what they don't know—so even obvious-in-hindsight gains (LED bulbs paying off in 2-3 years, a microphone gain knob, factory quality-defect indices) are missed until someone points them out, explaining persistent inefficiency in the market for talent.
“there’s unknown unknowns. The problem is you don’t know what you don’t know.”
Perception of past certainty is unreliable; long-run quantitative indices better measure true uncertainty.
- People systematically perceive the past as more certain than the present because they observe what actually happened and forget the alternative scenarios that could have occurred; long-run quantitative indices are therefore more reliable than anecdotal recollection for judging which eras had high policy uncertainty.
“It’s actually phenomenal to think the past was more certain than the present because you see the past having happened. You forget all the alternative scenarios that could have been.”
Anecdotal case studies in management books suffer selection bias; large-scale data is required.
- Popular management books are weak as evidence because they rest on individual anecdotes and case studies subject to severe selection bias—firms that do badly refuse to be studied or threaten lawsuits and legal sign-off requirements—so a book can be found supporting almost any theory; large-scale survey data is needed to discipline which management hypotheses actually hold.
“it’s very hard for them to get proper information on firms that do badly because they refuse. They threaten lawsuits.”
Market and economy linkage is weak because equity exposure is tilted toward tech, not broad employment.
- The stock market and the state of the US economy are only weakly linked, because the market is roughly 30% high tech (only ~7% of US jobs) and rises when interest rates drop as the economy slows, making equities a relatively better investment.
“when interest rates drop because the economy slows, it makes the stock market go up because it’s suddenly a relatively better investment.”
Unconventional high-variance career paths can pay off individually despite lower average returns.
- Bloom characterizes his own unusual career path (leaving a PhD for McKinsey and the UK Treasury before returning to academia) as on average a worse path that happened to work out due to luck—framed as first- and second-order stochastic dominance, where the conventional path is safer but his higher-variance route paid off individually.
“it’s like first and second orders of the stochastic dominance. On average, the path I took was probably a less good path to take.”
Top-five PhD programs' measured value-added is unclear; smaller programs offer better mentoring access.
- Top-five economics PhD programs recruit by far the best students, so their measured value-added is unclear and not uniformly positive; smaller programs (like UCL in the late 1990s) can be better for some students because thin cohorts give far more direct access to faculty than a 25-per-year cohort allows.
“the top five schools recruit, by far, the best students... The question is, what’s the value added? It’s never been obvious to me what that is.”