In the 19th and early 20th centuries hospitals were places where the indigent went and were regarded as degrading, while the non-poor were treated at home by doctors for a fee—so everyone of that era would have preferred a paid doctor at home over the charity hospital, a preference completely inverted today where university (nonprofit) hospitals are considered the best.

Donors prefer giving money to producers (charities) rather than directly to indigent recipients because they want to constrain how it is used—funding a specific need like health care or clean water—so giving consumers vouchers or cash, which economists typically recommend, would generate far fewer donations.

Goods like news, podcasts, and websites where the marginal cost of one more audience member is near zero can be financed better by giving the product away free and soliciting donations than by charging subscriptions, because donors are attracted by the large reach a free product achieves—inverting the standard Samuelsonian public-goods story.

There are not two but at least three (and arguably four) organizational models for addressing social problems: the pure for-profit model motivated by money, the nonprofit model funded by donations/volunteers giving the product away free, and government coercion through taxation—plus a hybrid model where for-profit and nonprofit approaches coexist in the same industry.

Monetary incentives are uniquely powerful in certain settings: you would rather rely on a paid cab driver who fears losing her job to make a 4:30am airport pickup than on a sibling acting out of love, because there is more at stake and the obligation is enforceable—illustrating that not just any incentive but specifically monetary incentives can be most reliable.

Gifts in kind are generally preferred to cash because gift-giving is a relationship transaction, not a financial one; cash or gift certificates are culturally inappropriate in many settings (a spouse's birthday, a dinner guest) precisely because they signal the absence of personal thought and care.

It would be perverse but possibly true that converting the New York Times from a for-profit into a fee-charging nonprofit would make consumers more willing to pay subscription fees, because the nonprofit framing changes the psychological/cultural meaning of the transaction.

Cultural expectations that internet content (and historically Red Cross doughnuts) 'should be free' cause people to forgo transactions whose value to them exceeds the price—an economically irrational behavior driven by a sense of entitlement and the psychological difficulty of paying for something previously obtained for nothing.

Giving a public service away free (public schools, and by extension other government-provided goods) causes people not to treat it as valuable in the way they would if they had to earn money to pay for it, which Roberts argues is part of the problem with the public school system.

A 'pile up free food in a warehouse' welfare scheme (like the Carrboro 'really really free market') is appealing for its minimal administrative cost and self-targeting, but fails the Hayekian knowledge problem: without prices to signal what people actually want, the system would supply goods (e.g., Cheerios) that recipients don't want, whereas markets steer information via prices.

You want the Idaho potato farmer motivated by profit rather than love, because the love-motivated supplier would actually be less reliable and more expensive—profit gets the person up early in unpleasant weather, whereas love would yield worse and more expensive potatoes.

The best gift is one the recipient didn't know they would like but the giver, through investment in the relationship, correctly anticipates—or one the recipient wanted but would never buy for themselves because it seemed self-indulgent; receiving it as a gift removes the guilt while the recipient would genuinely refuse the equivalent cash plus the option to buy it.

Much nonprofit activity (opera, museums, universities) functions as tax-exempt subsidy of services consumed largely by the rich, allowing wealthy patrons to claim social credit while window-dressing programs (e.g., inner-city outreach) obscure the regressive reality of the tax benefit.

The expansion of government provision in the early 1930s at the federal level crowded out private charity to the indigent, sick, and hungry, so today almost no private charity goes to the general poor—surviving private charity (soup kitchens, food banks) mainly serves the homeless who fall outside the bureaucratic systems like food stamps.

Paying donors for blood (as opposed to relying on altruistic donation) historically risked attracting desperate, less healthy donors—drug addicts or people with diseases like hepatitis or AIDS who might lie—but improvements in testing technology have substantially mitigated this problem, allowing paid and volunteer systems to coexist (as with paid plasma donation).

Wikipedia, open-source software, and similar projects demonstrate that many people will contribute time and resources without direct monetary compensation—motivated by love, mission, or reputational benefits—and such voluntary, non-market organization can work well.

Promoting one's own paid website, book, or service on a podcast is not shameful but ordinary commerce—everyone is selling something (reputation, brand, products)—and hostility to it (as seen in reactions to Art De Vany charging for his site) reflects a growing sense of entitlement bred by the abundance of free content.

Sophisticated statistical techniques like two-stage least squares (instrumental variables) are designed to correct for reverse causation and confounding, but they are often applied to data that is simply not good enough for the task, so we can never control for all factors and may falsely attribute correlations to an unobserved variable.

Roberts proposes that to address bias, empirical work might be restricted to actual experiments, to cases where statistical techniques give less scope for researcher creativity, or conducted as 'empirical tournaments' where both ideological sides analyze data simultaneously and in the open.

Researchers' practice of trying many specifications and discarding those that fail, then reporting only the surviving result, invalidates the standard reliability measures (confidence intervals, 95% significance), because what appears in the paper is the last of hundreds of attempts rather than a first effort—a point Roberts draws from Ed Leamer.