Ten things economists have measured that sound wrong

1

Cash for the poor does not buy more alcohol or tobacco

The worry is old and sensible: give poor households money with no strings and some of it will go on drink and cigarettes. In 2014 two World Bank economists, David Evans and Anna Popova, reviewed the evidence. They found 19 studies that measured what happened to spending on alcohol, tobacco and similar goods when households received cash, in ten countries across Latin America, Africa and Asia, and 11 more in which recipients were asked how they had used the money.

Almost without exception, the studies found no significant effect or a significant fall in spending on those goods. The only two studies that found a significant rise were non-experimental, and in both the rise was small and significant only at the 10% level.

The finding is an average across programmes, mostly cash transfers aimed at poor households. It does not say that no household anywhere spends a transfer on drink, and “temptation goods” here means, principally, alcohol and tobacco.

Evans and Popova, “Cash Transfers and Temptation Goods: A Review of Global Evidence”, World Bank Policy Research Working Paper 6886 (2014); published in Economic Development and Cultural Change 65(2), 2017. Source 1

2

The “third of traffic is hunting for parking” statistic is a misreading

It is repeated in policy papers and news stories: about 30% of city traffic is drivers circling for a space. The number traces to Donald Shoup, an urban planner at UCLA. He reviewed sixteen studies, from 1927 to 2001, of cars cruising for kerbside parking in congested downtowns. Finding a space took between 3.5 and 14 minutes, and between 8% and 74% of the cars were cruising. The plain average of the sixteen was about 30%.

In 2020 Rachel Weinberger, Adam Millard-Ball and Robert Hampshire traced the popular “30% of urban traffic” to that straight average, taken without regard to the size of the town, the date, or how each study counted traffic. They call it a misinterpretation and a large overestimate. Their own measure uses GPS traces of real trips in San Francisco and Ann Arbor. About one car trip in twenty included some cruising, and a cruising trip added two to three minutes. Extra driving for parking came to under 1% of all car travel.

The two numbers do not measure the same thing: the old one counts cars in the traffic stream on the streets studied, the new one counts extra distance across a whole city, and the authors call it an upper bound because a detour can have other causes. Cruising is also concentrated. Around Ann Arbor’s south university area, more than 30% of trips ending there included some. A different counting method, by Hampshire and Shoup, put it at 15% of cars on twelve central streets in Stuttgart. What fails is the general claim that a third of traffic anywhere is looking for a space.

Shoup, “Cruising for parking”, Transport Policy 13(6), 2006; Shoup, “Progress in immobility”, UC Transportation Center working paper (2011); Weinberger, Millard-Ball and Hampshire, “Parking search caused congestion: Where’s all the fuss?”, Transportation Research Part C 120, 2020; Hampshire and Shoup, Journal of Transport Economics and Policy 52(3), 2018. Source 2

3

Winning the auction is bad news

Picture a jar of coins that is worth the same to every bidder, but nobody knows the amount. Each bidder makes an honest guess, some too high and some too low. The winner is whoever guessed highest, which means the winner is the one most likely to have overestimated.

In laboratory auctions run by John Kagel and Dan Levin, bidders in groups of six or seven bid so aggressively that winners lost money, while groups of three or four made profits. In 1989 Douglas Dyer, Kagel and Levin gave the same game to executives from the construction industry and to students. Both groups fell into the trap. A follow-up found that construction firms avoid it in real bidding through industry practices, so the lab result does not say that firms routinely lose.

Kagel and Levin, American Economic Review 76(5), 1986; Dyer, Kagel and Levin, Economic Journal 99(394), 1989; Dyer and Kagel, Management Science 42(10), 1996. Source 3

4

Cheaper rice, less rice eaten, but not by the poorest

Economics textbooks have long carried a curiosity: a Giffen good, one that people buy less of when it gets cheaper. For more than a century there was no real evidence for it. The standard example, potatoes in the Irish famine, has been discredited.

Robert Jensen and Nolan Miller found it in the field. In Hunan, China, they gave randomly chosen households from the local urban poverty lists, typically below the World Bank’s dollar-a-day line, vouchers that cut the price of rice, and tracked what they ate for five months. The result depended on how poor a household was. Families that still took at least a fifth of their calories from foods other than rice ate less rice when it was cheaper, and more when the vouchers stopped: a 1% rise in price went with a 0.45% rise in rice eaten, significant at the 1% level. A cheaper kilo freed money, and they spent some of it on meat and other food in place of rice. Families getting more than 80% of their calories from rice behaved normally. They ate almost nothing else, had nothing to swap into, and bought less rice when it cost more.

Across all the households taken together, the effect had the Giffen sign but was weak: significant only at the 10% level once controls were added, and not significant without them. The evidence for wheat in the second province, Gansu, was weaker still, because most households there did not fit the conditions. It is one experiment in two provinces over five months; nothing here says everyday goods behave this way.

Jensen and Miller, “Giffen Behavior and Subsistence Consumption”, American Economic Review 98(4), 2008 (figures from the working-paper version linked below). Source 4

5

Pharmacists rarely buy the brand name

Brand-name and store-brand painkillers contain the same active ingredient. Four economists matched household shopping records to the occupation of the main shopper. For headache remedies, pharmacists chose the national brand over the store brand only 9% of the time. The average consumer chose it 26% of the time.

Chefs showed the same pattern in pantry staples such as salt and sugar: 12 percentage points less of their purchases went to national brands than for similar non-chefs. Across 50 health categories the authors conclude that misinformation and similar mistakes explain a substantial share of the price premium on brands. For most of the 241 food and drink categories the share is much smaller.

Bronnenberg, Dubé, Gentzkow and Shapiro, “Do Pharmacists Buy Bayer? Informed Shoppers and the Brand Premium”, Quarterly Journal of Economics 130(4), 2015. Source 5

6

Footballers play close to the textbook

A penalty kick is a one-shot game with a known solution. If the kicker favours one corner, the goalkeeper learns to cover it, so the best play is to mix unpredictably, and to mix so that scoring odds come out equal whichever way you shoot.

Ignacio Palacios-Huerta studied penalty kicks in professional football. Kickers’ scoring chances were statistically indistinguishable across the sides they chose, and their choices were serially independent: what a player did last time did not predict the next kick. A few individual players did fail the tests, about as many as chance alone would produce (three of 42 on equal scoring odds at the 5% level, where 2.1 were expected).

It does not hold everywhere. Kovash and Levitt (2009) tested baseball pitchers and American football play-callers and found they did not play the equilibrium: pitchers threw too many fastballs and football teams passed less than they should.

Palacios-Huerta, “Professionals Play Minimax”, Review of Economic Studies 70(2), 2003; Kovash and Levitt, “Professionals Do Not Play Minimax”, NBER Working Paper 15347 (2009). Source 6

7

Money should flow to India. It does not.

Rich countries have a lot of machinery per worker; poor countries have little. If a machine adds more output where machines are scarce, investors should pour money into poor countries. In 1990 Robert Lucas did the sum. Take the gap in output per person between the United States and India, a factor of about 15 in the Summers and Heston figures Lucas used (their table is for 1985), and assume both countries use the same technology and that capital earns 40% of income. Capital in India would then earn about 58 times as much as in America. At that return, almost no investor would put a dollar anywhere else.

Nothing like that flow happened. Lucas weighed four ways out of the puzzle. Differences in the skills of workers shrink the 58 to about 5. Add the benefits that spill over to other workers from the education around them and the figure falls to about 1. The other two are imperfections in capital markets: political risk, and monopoly control of capital flows, as under empire. The debate over why capital stays home continues, and later work points to institutions, among other factors.

The 58 is a thought experiment built on identical technology. Lucas used it to show that the assumption fails, not as a measurement of actual returns.

Lucas, “Why Doesn’t Capital Flow from Rich to Poor Countries?”, American Economic Review 80(2), 1990; the 58 is restated in Alfaro, Kalemli-Özcan and Volosovych, NBER Working Paper 11901 (2005), published in Review of Economics and Statistics 90(2), 2008. Source 7

9

People choose to lock their own money away

In Butuan City, in the Philippines, a bank offered a randomly chosen group of its clients a savings account with a catch: the money could not be taken out until a goal they set was met. Of 710 clients offered the account, 202 (28.4%) opened one. After twelve months the average bank savings of those offered the account were about 411 pesos higher than in the comparison group, roughly four-fifths of the 503 pesos the group had saved at the start.

Standard theory says a person should never choose a rule that takes away options. The authors’ reading is that some people know they will be tempted to spend, and want a way to stop themselves. Women whose answers showed more impatience over an immediate choice than over the same choice a few months ahead were more likely to open the account (significant at the 10% level). Impatience alone did not predict it.

One bank in one city, and 71.6% of those offered the account declined it. The 411-peso gain is significant only at the 10% level, and against a group that received only a marketing visit the difference was not significant.

Ashraf, Karlan and Yin, “Tying Odysseus to the Mast: Evidence from a Commitment Savings Product in the Philippines”, Quarterly Journal of Economics 121(2), 2006. Source 9

10

A dollar sells for more than a dollar

In 1971 the game theorist Martin Shubik described an auction for a dollar bill. Anyone can bid, and the top bidder gets the dollar. The twist: the second-highest bidder also pays what they bid and gets nothing. Each extra bid looks sensible: at first it offers a chance of winning the dollar for much less than a dollar, and once bids pass a dollar it offers a smaller loss than quitting would. The result, as Shubik described it from playing the game at parties, is that the two bidders’ payments added up to three to five dollars for the one dollar.

Shubik’s three to five dollars is the total paid by both bidders, from informal play with no sample reported. The bot is a simple rule, an illustration, and real bidders behave differently.

Shubik, “The Dollar Auction Game: A Paradox in Noncooperative Behavior and Escalation”, Journal of Conflict Resolution 15(1), 1971. Source 10

How these ten were chosen

The aim was findings that are well established in the research literature, that most curious readers will not already know, and that carry no party-political charge. The famous ones (the prisoner’s dilemma, comparative advantage, the broken-window fallacy) are left out. Three candidates were dropped because follow-up studies disagree with the originals: long-run colonial history, the cost of gift-giving, and airline price dispersion.

The figures in items 1 to 7 and 9 and 10 were compared with the papers they come from in October 2026, in full text, except that Shoup (2011) and Dyer and Kagel (1996) were read as abstracts, and Shoup (2006) was not opened: the range and average attributed to Shoup come from the 2011 abstract and from the 2020 paper’s account of his review. Item 4 was read in the working-paper version. The 1964 papers by Balassa and Samuelson were not opened; the one-sentence account of them follows the standard textbook account. Two items carry limits that matter most: item 2, where the popular number is the thing being corrected, and item 4, which rests on one experiment. The simulation, the dollar-auction bot and the price-level statistics in item 8 are calculations, not results from the cited papers.