Liberal Techno-Optimism: The Case for Markets, Progress, and a Safety Net

I call my worldview liberal techno-optimism. “Liberal” carries the meaning The Economist gives it: individual liberty, open markets, free trade, the rule of law, and a government competent enough to do the jobs only government can do. “Techno-optimism” means expecting science and new technology, AI included, to keep raising living standards, and wanting more of the things that make that possible: homes, power plants, transmission lines, laboratories, and data centers.

The term needs three boundaries.

• Unlike libertarianism, it supports a strong safety net and regulation aimed at specific, demonstrated harms.
• Unlike Marc Andreessen’s 2023 Techno-Optimist Manifesto, which names “trust and safety” and “risk management” among its enemies, it treats safety work as part of building things well.
• It carries no stock tip. Followed into a portfolio, it points to owning the whole market rather than the companies most associated with AI.

A personal finance site has a stake in this because a household’s largest costs (housing, health insurance, electricity) and its largest long-term asset (a share of the world’s businesses) both sit downstream of the choices this worldview is about.

Ends, Means, and Guardrails

The worldview has three layers, and the order matters.

• Ends: the freedom to choose your own life, material welfare, and prosperity that reaches most people.
• Means: competitive markets, trade, science, entrepreneurship, and the entry of new firms.
• Guardrails: courts and property rights, regulation aimed at specific harms, public goods such as basic research, and social insurance.

Markets and innovation sit in the middle layer. They earn their place by serving the ends. In “The Use of Knowledge in Society” (1945), Friedrich Hayek argued that prices pass along what millions of people know about scarcity and demand, and no planner can collect that knowledge. The same standard applies to government. An intervention earns its place by beating a feasible alternative once its costs and side effects are counted, and a worthy goal alone does not meet that bar.

My working rule: default to open competition and decentralized choice. Intervene when the problem is specific, when there is evidence that a feasible alternative does better, and when the remedy respects individual rights.

Where this sits among related views

None of the parts are new. The Economist was founded in 1843 “to support the cause of free trade,” and its 2018 manifesto for renewing liberalism called for “recasting taxation, welfare, education and immigration” and for freeing the economy from monopolies and “the planning restrictions that shut people out of the most prosperous cities.” The closest recent relatives are Ezra Klein and Derek Thompson’s Abundance (2025), which argues for clearing the procedural barriers that block housing, energy, and infrastructure, and the Niskanen Center’s free-market welfare state, where Samuel Hammond argued in 2018 that “a more universal welfare state, far from being at odds with innovation and economic freedom,” can support economic dynamism.

What this post adds is carrying the abundance and free-market welfare state combination into AI and data centers, and following it into a household portfolio.

Growth Comes From Ideas, and Ideas Spill Over

A design, a formula, or a piece of software can be used by any number of people at once without being used up. Paul Romer’s 1990 model of endogenous growth put that property at the center: an economy that only adds machines runs into diminishing returns, while one that keeps producing ideas can keep growing.

Because ideas leak, the firm that pays for research captures only part of the payoff. Using a panel of U.S. public companies, Bloom, Schankerman, and Van Reenen (2013) estimated that the gross social returns to R&D “are at least twice as high as the private returns.” Their specifications put the social return around 55 to 58 percent against a private return near 21 percent. That gap is the economic case for public funding of basic research. It does not show that any particular subsidy pays for itself.

The 2025 economics Nobel went to three researchers of this process. Joel Mokyr was recognized “for having identified the prerequisites for sustained growth through technological progress,” and Philippe Aghion and Peter Howitt “for the theory of sustained growth through creative destruction.” Creative destruction means new firms displace old ones. Policies that shield incumbents from entrants protect today’s profits at the expense of tomorrow’s firms, which is why this worldview treats market entry as a core commitment.

Institutions decide whether any of this happens. In his 1993 Nobel lecture, Douglass North wrote that “it is adaptive rather than allocative efficiency which is the key to long run growth.” Courts, property rights, and predictable rules lower the cost of investing over long horizons. Adaptive efficiency is the ability to change those rules when technology and circumstances change, so they do not freeze around the industries of the last century.

Political institutions show up in the data as well. Acemoglu, Naidu, Restrepo, and Robinson (2019) estimated that “democratizations increase GDP per capita by about 20% in the long run,” with more investment and schooling among the channels. That is an average across many transitions, and it promises nothing about any single one.

Why a Safety Net Belongs in a Pro-Market Program

Some risks private markets insure badly. When buyers know more about their own risk than insurers do, the people most likely to claim buy the most coverage, prices rise, and lower-risk buyers drop out. Pooling everyone solves that. Health insurance, disability insurance, and lifetime income are the standard examples, and most social insurance systems are built around them.

A good safety net can also make the market economy work better. When health coverage comes only through an employer, people stay in jobs they would otherwise leave. Brigitte Madrian (1994) estimated that this “job-lock” cut voluntary turnover among workers with employer insurance from 16 percent to 12 percent a year, though Douglas Holtz-Eakin (1994), using different data, found no evidence of job-lock. Fairlie, Kapur, and Gates (2011) found that business ownership among male workers rose from 24.6 percent just under age 65 to 28.0 percent just over it, when Medicare begins, with no comparable jump at other ages between 55 and 75 (figures from their working paper version). A 2014 working paper by Gareth Olds found that families who became eligible for children’s public health insurance (SCHIP) were more likely to be self-employed and to own incorporated businesses; it has not been published in a peer-reviewed journal.

For early retirees this is concrete. Before Medicare at 65, health coverage usually comes from an ACA marketplace plan, COBRA, or a spouse’s employer, and marketplace subsidies are one of the largest swing factors in an early retirement budget. With the enhanced premium tax credits gone in 2026, the subsidy cliff at 400 percent of the poverty line is back, and a dollar of extra income can cost thousands in premiums. That cliff is an example of design that punishes work and saving.

The Affordable Care Act itself moved behavior less than this logic might suggest. Levy, Buchmueller, and Nikpay (2018) found no change in retirement or part-time work among workers aged 50 to 64 after 2014. Duggan, Goda, and Li (2021) found that coverage among 60- to 64-year-olds rose 4.5 percentage points while their labor force participation fell 0.6 points, and Heim and Yang (2017) found that the ACA did not increase self-employment overall in states without similar insurance rules before 2014, though self-employment rose in the second year among people eligible for subsidies. The case for decoupling insurance from employment is sound; the measured effects of this particular reform are modest.

Denmark’s labor market is the best-known model of the alternative design. Andersen and Svarer (2007) describe “flexicurity” as easy hiring and firing combined with generous unemployment benefits and heavy spending on retraining and job placement. The system protects workers’ incomes while letting firms adjust, and the authors caution that it does not guarantee low unemployment by itself.

The costs are debated. Peter Lindert’s Growing Public (2004) found no clear GDP cost from the high tax-funded social spending of rich democracies, an argument he previewed in a 2003 NBER paper titled “Why the Welfare State Looks Like a Free Lunch.” Bergh and Henrekson (2011) surveyed the later evidence: in rich countries, “an increase in government size by 10 percentage points is associated with a 0.5% to 1% lower annual growth rate,” a correlation they caution against reading as cause and effect. Benefit phase-outs work as extra marginal tax rates on low earners, and programs that are hard to apply for fail the people they are meant to reach. The standard to apply: judge a program by coverage, poverty reduction, work incentives, administrative cost, and long-run financing, rather than by how much it spends.

Three Tests: Housing, Data Centers, and AI at Work

Housing

Housing is the clearest case and the largest category of household spending in the BLS Consumer Expenditure Survey. When high-wage cities restrict building, prices and rents rise, fewer workers can move to the jobs, and the gains flow to people who already own.

The most cited estimate of the cost needs a warning label. Hsieh and Moretti (2019) modeled 220 metropolitan areas and concluded that housing constraints lowered aggregate U.S. growth by 36 percent between 1964 and 2009. In a 2026 comment in the same journal, Brian Greaney reported errors in their code, showed that the model’s results depend on an arbitrary choice of population unit, and found that a corrected version yields an effect “two orders of magnitude smaller than what they report.” Hsieh disputed an earlier draft in a 2023 referee report. The direction of the argument holds up: blocking homes near high-wage jobs has a cost. Its size is unsettled.

The policy that follows is to legalize more homes, with fewer minimum-lot-size and parking mandates and faster permits, while keeping building codes that prevent real harm. For the household-level decision, see Rent vs. Buy.

Data centers and your electric bill

The International Energy Agency’s Energy and AI report put data-center electricity use at about 415 terawatt-hours in 2024, roughly 1.5 percent of world consumption, and projected around 945 TWh by 2030 in its base case. Its April 2026 update reported a 17 percent jump in 2025 and still expects use to double by 2030. These are scenarios, and power plants and transmission lines take longer to build than data centers do.

Computing capacity is productive capital, and this worldview supports building it. The condition is that each project pays its full costs, and the live dispute is who pays for the grid. In December 2025 the Federal Energy Regulatory Commission found the PJM grid operator’s tariff unjust and unreasonable because it lacked clear rates, terms, and conditions for large loads, such as data centers, located at power plants. Among its concerns was the risk that some grid-service costs could be “passed through to residential customers.” It ordered new transmission services for those loads and sent the rates to further proceedings. A June 2026 rehearing order upheld the findings and noted that states still control how retail costs are allocated.

Voters are asking the same question. In an Economist/YouGov poll from August 2026, 61 percent of Americans said they would oppose a data center in their community and 62 percent expected one to raise their electricity bill. A project that pays for its own interconnection, contracts for the generation it uses, and leaves other ratepayers no worse off is easy to defend. A project that shifts its costs onto households is a subsidy, and a liberal should oppose it like any other.

AI at work

The best-identified evidence comes from single workplaces. In the Quarterly Journal of Economics, Brynjolfsson, Li, and Raymond (2025) studied 5,172 customer-support agents and found that an AI assistant raised issues resolved per hour by 15 percent on average, with the largest gains for less experienced workers. The most experienced agents saw small gains in speed and small declines in quality.

Economy-wide estimates are far smaller. Daron Acemoglu’s “The Simple Macroeconomics of AI” estimated that AI’s effect on total factor productivity over ten years “should be no more than 0.66% in total,” about 0.064 percent a year, with GDP up roughly 0.9 to 1.2 percent over the decade. The two results can both hold: one measures a task, the other an economy in which most tasks are not yet affected.

The International Monetary Fund estimated in 2024 that “almost 40 percent of global employment is exposed to AI,” and about 60 percent of jobs in advanced economies. Exposure measures overlap between a job’s tasks and what AI can do. In advanced economies, the IMF says roughly half of exposed jobs may benefit. The policy this worldview favors is fast deployment combined with help for the people who have to change jobs: portable benefits, retraining judged by placement results, and competition so that cost savings reach customers.

The Strongest Objections

The gains from technology can concentrate. In Power and Progress (2023), Daron Acemoglu and Simon Johnson argue that new technology raises wages broadly only when workers and institutions have the power to demand a share, and they want policy to steer research toward tools that complement workers. A safety net and competition spread gains after the fact; they do not change which technologies get built. Their argument is strongest against automation that replaces workers while adding little productivity, which Acemoglu and Pascual Restrepo call “so-so automation.”

Liberalism has a mixed record. Reviewing Alexander Zevin’s history of The Economist, Pankaj Mishra noted that founder James Wilson responded to the Irish famine with “a homeopathic remedy: more free trade.” A tradition with that history has to keep checking its prescriptions against outcomes.

Competition and innovation have an unsettled relationship. Aghion and coauthors (2005) found an inverted-U in U.K. data: innovation rises with competition up to a point, then falls. Hashmi (2013) found a mildly negative relationship in U.S. data. “More competition brings more innovation” is a reasonable default to test case by case.

Incumbents capture rules. George Stigler argued in 1971 that regulated industries often shape their own regulation. Licensing regimes, permitting rules, and safety standards can all become moats for the firms that can afford compliance, AI rules included.

Efficiency and justice can diverge. A policy can raise output while violating someone’s rights, and efficiency cannot settle whether that trade is acceptable. That is why liberty sits in the ends layer and efficiency in the means layer.

Half of Americans are more worried than excited. In a Pew survey published in September 2025, 50 percent of U.S. adults said they were more concerned than excited about AI, and 10 percent more excited than concerned. A case for faster deployment has to win over that half with results people can see.

What This Means for a DIY Investor

Three beliefs are easy to blur together:

1. Economic: competitive markets and new technology raise prosperity over time.
2. Financial: owning a diversified share of productive businesses earns a return above cash over long periods.
3. Portfolio: a particular stock, sector, or country will beat the market from today’s price.

This worldview supports the first two. Neither implies the third.

Faster growth has not meant higher returns

Jay Ritter (2012) found that across developed countries from 1900 to 2011, the correlation between inflation-adjusted per capita GDP growth and stock returns was −0.39; for emerging markets from 1988 to 2011 it was −0.41. A correlation leaves the cause open, but this one contradicts the simple story that the fastest-growing economy makes the best investment. Growth often comes from new firms rather than existing shareholders’ companies, existing owners are diluted by new share issuance, much of the gain goes to workers and customers, and prices already reflect expected growth. The Stock Market Is Not the Economy walks each link from GDP to per-share returns.

New technology has had mixed results for shareholders

Britain’s railway mania of the 1840s built a national network, and Andrew Odlyzko notes that “while shareholders lost heavily, hardly any of the lines went bankrupt.” American railroads went the other way over a longer span: in the Dimson, Marsh, and Staunton data, $1 invested in U.S. railroad stocks in 1900 grew to $62,716 by the end of 2014, against $39,134 for the overall U.S. market, though railroads trailed badly until the early 1970s. Warren Buffett’s 2007 letter described the worst business as one that “grows rapidly, requires significant capital to engender the growth, and then earns little or no money. Think airlines.” The record is uneven, and it cannot tell you which side today’s leaders will land on. Are We in an AI Bubble? covers how to think about that in real time.

The wealth comes from a few stocks you cannot pick in advance

Hendrik Bessembinder’s March 2026 working paper, “One Hundred Years in the U.S. Stock Markets,” covers 29,754 stocks from 1926 to 2025. Long-term holders of nearly 60 percent of them ended with less wealth than Treasury bills would have produced, the median stock’s lifetime return was −6.9 percent, and 46 firms accounted for half of the $91 trillion in net wealth the market created. Most Stocks Lose to T-Bills explains why that skew favors owning the whole market. A broad index holds future winners once they list and qualify, though it misses gains that accrue to private companies and to customers.

Optimism about AI is a view about technology; AI stocks also depend on price

Expecting AI stocks to beat the market requires believing that AI will do better than what current prices already assume. A company can transform an industry and still return less than the market if the price you paid already counted on it. Stock Valuation for DIY Investors includes a “what has to be true?” explorer for this, and Growth Stocks Do Not Mean Higher Expected Returns shows the arithmetic.

The portfolio that follows

• Own the world at low cost. A total-world fund such as VT, or a total U.S. and total international pair such as VTI and VXUS, holds the innovators in every sector and country. See The Case for Global Equity Diversification.
• Make active bets earn their place. S&P Dow Jones Indices reported that 79 percent of active U.S. large-cap funds trailed the S&P 500 in 2025, and about 90 percent over 15 years. William Sharpe’s 1991 arithmetic explains why: before costs, active investors as a group hold the market.
• Watch hidden tech concentration. A broad index already holds large technology companies. Adding QQQ or a semiconductor fund on top raises concentration; see The Tech Bro Portfolio and Why I Avoid QQQ.
• Count your job as an asset. A tech worker paid in employer stock who also overweights tech has the same bet in three places. See Human Capital Risk for Tech Workers.
• Match safe assets to dated spending. Stocks fund growth; known bills on known dates belong in Treasury bills, bonds, or a TIPS ladder. See Asset-Liability Matching.

Liberal techno-optimism does not choose between 90/10 and 60/40. That depends on your spending, horizon, and capacity for loss. The liberty part of the worldview does show up in one place: the purpose of a portfolio is to fund the life you choose, and financial independence is the amount of choice you can afford.

Frequently Asked Questions

Is liberal techno-optimism left-wing or right-wing?

Neither maps cleanly. In American usage “liberal” means center-left; in The Economist’s usage it means pro-market and pro-liberty. This view takes support for markets, trade, and building from the classical liberal side, and support for social insurance from the center-left. Its closest relatives sit near the center from opposite starting points: Abundance was written by two writers on the left, and the Niskanen Center was founded in 2014 by former libertarians.

How is this different from the Abundance agenda?

It shares Abundance’s housing and energy agenda and adds two things: an explicit grounding in classical liberal institutions such as the rule of law and free trade, and a focus on AI and data centers, which Abundance says little about.

If I am optimistic about AI, should I overweight tech stocks?

Only if you also believe AI companies will beat the expectations already built into their prices. A total-market index already holds them at their market weights, so optimism about AI is already expressed in a broad portfolio.

If faster growth does not mean higher returns, why invest in stocks at all?

Because owners of a diversified set of businesses have historically earned a premium over cash in most markets with long records, as Dimson, Marsh, and Staunton’s century of global data documents. The premium comes from bearing risk and buying at reasonable prices, and it does not require predicting which economy grows fastest.

Does supporting data centers mean supporting subsidies for them?

No. The position is to make building faster and more predictable while each project pays for the grid upgrades and generation it requires. Tax breaks and discounted power rates need their own case, and costs shifted onto residential ratepayers fail the test.

Key Takeaways

• Liberal techno-optimism pairs classical liberal institutions and markets with support for building (homes, power, labs, data centers) and a strong, paid-for safety net.
• Ideas drive long-run growth, and their social return exceeds the private return, which justifies public research funding and open entry for new firms.
• Social insurance can make markets work better by covering risks private insurers handle badly, and it can free people to change jobs or start businesses, though the measured effects of the ACA were modest. Program design should avoid steep cliffs.
• Builders should pay their full costs. Data centers that pay for their own grid upgrades are easy to defend; ones that shift costs to households are subsidies.
• Optimism about progress is a reason to own the whole market. Faster-growing economies and exciting industries have not reliably delivered higher returns, and the wealth comes from a few stocks no one can reliably pick in advance.

Related Guides

• The Stock Market Is Not the Economy traces why GDP growth and shareholder returns come apart.
• The Stock Market Rewards Optimists, But Not the Naïve Ones covers Elroy Dimson’s conditions on long-run optimism.
• Most Stocks Lose to T-Bills. The Market Still Wins. explains Bessembinder’s skewness findings.
• Are We in an AI Bubble? applies William Bernstein’s four signs of a mania.
• Will AI Lower Interest Rates? covers what an AI productivity boom could do to rates.
• The ACA Subsidy Cliff Is Back calculates the cliff for early retirees.

Sources

Each entry is labeled by type. Working papers have not completed peer review.

• Romer, P. (1990). “Endogenous Technological Change.” Journal of Political Economy 98(5): S71–S102.Link. (Peer-reviewed)
• Hayek, F. A. (1945). “The Use of Knowledge in Society.” American Economic Review 35(4): 519–530. (Peer-reviewed)
• Bloom, N., Schankerman, M., and Van Reenen, J. (2013). “Identifying Technology Spillovers and Product Market Rivalry.” Econometrica 81(4): 1347–1393.Link. (Peer-reviewed)
• Aghion, P., Bloom, N., Blundell, R., Griffith, R., and Howitt, P. (2005). “Competition and Innovation: An Inverted-U Relationship.” Quarterly Journal of Economics 120(2): 701–728. (Peer-reviewed)
• Hashmi, A. R. (2013). “Competition and Innovation: The Inverted-U Relationship Revisited.” Review of Economics and Statistics 95(5): 1653–1668. (Peer-reviewed)
• North, D. C. (1993). “Economic Performance Through Time.” Nobel Prize lecture; published in American Economic Review 84(3), 1994.Link. (Lecture)
• The Royal Swedish Academy of Sciences (2025). Prize in Economic Sciences press release.Link. (Report)
• Acemoglu, D., Naidu, S., Restrepo, P., and Robinson, J. (2019). “Democracy Does Cause Growth.” Journal of Political Economy 127(1): 47–100.Link. (Peer-reviewed)
• The Economist (2018). “A manifesto for renewing liberalism,” September 13.Link. (Essay)
• Klein, E., and Thompson, D. (2025). Abundance. Avid Reader Press. (Book)
• Hammond, S. (2018). “The Free-Market Welfare State.” Niskanen Center.Link. (Essay)
• Andreessen, M. (2023). “The Techno-Optimist Manifesto.” a16z.Link. (Essay)
• Buterin, V. (2023). “My techno-optimism.”Link. (Essay)
• Madrian, B. (1994). “Employment-Based Health Insurance and Job Mobility: Is There Evidence of Job-Lock?” Quarterly Journal of Economics 109(1): 27–54.Link. (Peer-reviewed)
• Holtz-Eakin, D. (1994). “Health Insurance Provision and Labor Market Efficiency in the United States and Germany.” In R. Blank (ed.), Social Protection versus Economic Flexibility. University of Chicago Press. (Book chapter)
• Fairlie, R., Kapur, K., and Gates, S. (2011). “Is Employer-Based Health Insurance a Barrier to Entrepreneurship?” Journal of Health Economics 30(1): 146–162.Link. (Peer-reviewed)
• Olds, G. (2014). “Entrepreneurship and Public Health Insurance.” Harvard Business School working paper. Summary in HBS Working Knowledge, “Food Stamp Entrepreneurs”. (Working paper)
• Levy, H., Buchmueller, T., and Nikpay, S. (2018). “Health Reform and Retirement.” Journals of Gerontology: Series B 73(4): 713–722.Link. (Peer-reviewed)
• Duggan, M., Goda, G. S., and Li, G. (2021). “The Effects of the Affordable Care Act on the Near Elderly: Evidence for Health Insurance Coverage and Labor Market Outcomes.” Tax Policy and the Economy 35: 179–223. Link. (NBER conference volume)
• Heim, B., and Yang, L. (2017). “The Impact of the Affordable Care Act on Self-Employment.” Health Economics 26(12): e256–e273.Link. (Peer-reviewed)
• Andersen, T., and Svarer, M. (2007). “Flexicurity: Labour Market Performance in Denmark.” CESifo Economic Studies 53(3): 389–429.Link. (Peer-reviewed)
• Lindert, P. (2004). Growing Public. Cambridge University Press. Preceded by NBER Working Paper 9869 (2003).Link. (Book)
• Bergh, A., and Henrekson, M. (2011). “Government Size and Growth: A Survey and Interpretation of the Evidence.” Journal of Economic Surveys 25(5): 872–897.Link. (Peer-reviewed)
• Hsieh, C.-T., and Moretti, E. (2019). “Housing Constraints and Spatial Misallocation.” AEJ: Macroeconomics 11(2): 1–39.Link. (Peer-reviewed)
• Greaney, B. (2026). “Housing Constraints and Spatial Misallocation: Comment.” AEJ: Macroeconomics 18(2): 409–428.Link. (Peer-reviewed)
• International Energy Agency (2025). Energy and AI.Link. (Report)
• International Energy Agency (2026). Key Questions on Energy and AI, press release April 16.Link. (Report)
• Federal Energy Regulatory Commission (2025). Order on PJM co-location, Docket EL25-49, 193 FERC ¶ 61,217, December 18. (Report)
• Federal Energy Regulatory Commission (2026). Order on rehearing, Docket EL25-49-002, 195 FERC ¶ 61,209, June 18.Link. (Report)
• YouGov (2026). Many Americans think data centers are bad for the U.S.; more oppose them locally. August.Link. (Report)
• Brynjolfsson, E., Li, D., and Raymond, L. (2025). “Generative AI at Work.” Quarterly Journal of Economics 140(2).Link. (Peer-reviewed)
• Acemoglu, D. (2025). “The Simple Macroeconomics of AI.” Economic Policy 40(121): 13–58.Link. (Peer-reviewed)
• Georgieva, K. (2024). “AI Will Transform the Global Economy. Let’s Make Sure It Benefits Humanity.” IMF Blog, January 14.Link. (Report)
• Acemoglu, D., and Johnson, S. (2023). Power and Progress. PublicAffairs. (Book)
• Acemoglu, D., and Restrepo, P. (2019). “Automation and New Tasks: How Technology Displaces and Reinstates Labor.” Journal of Economic Perspectives 33(2): 3–30. Link. (Peer-reviewed)
• Mishra, P. (2019). “Liberalism According to The Economist.” The New Yorker, November 11.Link. (Essay)
• Stigler, G. (1971). “The Theory of Economic Regulation.” Bell Journal of Economics and Management Science 2(1): 3–21. (Peer-reviewed)
• Pew Research Center (2025). How Americans View AI and Its Impact on People and Society, September 17.Link. (Report)
• Ritter, J. (2012). “Is Economic Growth Good for Investors?” Journal of Applied Corporate Finance 24(3): 8–18.Link. (Peer-reviewed)
• Bessembinder, H. (2026). “One Hundred Years in the U.S. Stock Markets.” SSRN 6438198, March. (Working paper)
• Dimson, E., Marsh, P., and Staunton, M. (2015). Credit Suisse Global Investment Returns Yearbook 2015.Link. (Report)
• Odlyzko, A. (2010). “Collective hallucinations and inefficient markets: The British Railway Mania of the 1840s.”Link. (Working paper)
• Buffett, W. (2008). Berkshire Hathaway 2007 shareholder letter.Link. (Letter)
• S&P Dow Jones Indices (2026). SPIVA U.S. Year-End 2025.Link. (Report)
• Sharpe, W. (1991). “The Arithmetic of Active Management.” Financial Analysts Journal 47(1): 7–9. (Peer-reviewed)