Signals, screening and ordeals

Employers infer unseen ability from costly choices such as education. A signal "can succeed only if the signaling cost differs sufficiently among the senders" (as summarized in the Nobel committee's 2001 scientific background). It works even if the education adds nothing to productivity.

A buyer signal (a call, a paid pilot, a data upload) means something only if it costs a poor-fit prospect more than a good-fit one.

The screening side of signaling: the uninformed party designs the test. What a person gains from being identified can exceed what the economy gains, so markets can over-invest in screening.

Qualification hurdles cost money. Check whether they add value or just relabel who gets served.

For "experience" goods, which can only be judged by using them, heavy advertising is itself information. Makers of better products earn repeat purchases, so each ad dollar returns more and they advertise more.

Your own advertising sends a signal: what you offer tells the high-value buyer who you're for.

Firms keep quality promises when price is "sufficiently above salvageable production costs," so cheating would forfeit future rents. "Advertising investments thereby become a positive indicator of likely performance."

A price premium plus visible, sunk brand investment is what makes a quality promise believable.

An introductory price and "directly 'uninformative' advertising or other dissipative marketing expenditures" can signal the quality of a new product; "repeat purchases play a crucial role."

Buyers read your pricing and visible spend as quality signals, not just as acquisition levers.

Zahavi argued that the peacock's tail is an honest signal precisely because only a fit bird can afford it, and Grafen's models were widely read as proving it. Penn and Számadó (2020) argue that the "handicap principle" is wrong as a general rule.

A great story with a live dispute. For business decisions, Spence's version (a signal must cost the wrong customer more) is safer ground than "costly means honest."

A seller that can't tell buyers apart offers a price–quality menu and lets customers sort themselves. Compared with competition, this "almost always reduces the quality sold to any customer," because worse lower tiers keep high-value buyers from trading down, and it "frequently prices customers with the least intensities of demand out of the market."

Good/better/best tiers are a screening device, and the entry tier's limits protect premium revenue.

Programs can reach the people who need them by attaching requirements ("ordeals") that cost the well-off more than the needy, so recipients sort themselves.

The theory behind "some friction sorts." A phone call or a booked appraisal is an ordeal that serious buyers will put up with.

"adding a small application cost to a transfer program can substantially improve targeting through self-selection," mostly because better-off households didn't bother to apply. "Marginally increasing application costs … does not further improve targeting."

One small ask filters well. Piling on more friction adds cost without improving the mix.

Closing Social Security field offices caused "a persistent 16 percent decline in the number of disability recipients," with the largest drops among moderately severe, low-education applicants. Targeting got worse.

Friction can screen out the people you want. Check who drops out, by tier, every time you add a step.

Adopting credit scoring "appears to have increased profits by roughly a thousand dollars per loan," split about equally between screening out high-risk applicants and offering more to low-risk ones.

Scoring pays twice, by saying no to the worst and by investing more in the best. That's "sort, don't reject."