When states in the US banned cousin marriage, it led to more people immigrating to the cities and changing into higher paying professions. More evidence for how kinship networks hold back development. 1/ It used to be that 7% of marriages were between first cousins in the late 1700s -- this was down to 1.5% in the early 1900s.

A company gave its employees a large menu of options. What did they do? They spent 24% more than they had, with a majority of employees picking a plan for which a strictly better version existed. There is no explanation besides people making mistakes. 1/ They are working with a large, anonymous US firm that gave its employees the option to "build your own plan". Everything about their features was identical, except for varying deductibles, copays (a flat fee), coinsurance (a percentage fee), and an out of pocket maximum.

Workers in India are remarkably unwilling to do tasks which do not fit their caste identity. In this field experiments, workers turned down offers of 10 times their daily wage for doing just 10 minutes of work. 1/ The experiment makes hiring offers to 630 men to work assembling paper bags, and one other task, making multiple offers to as to elicit people's willingness to work on those jobs.

Suppose that the government of a developing country only cares about the welfare of workers. Under plausible assumptions, the best thing you could do for them is tax the workers to give to capitalists. Yes, really! This is why unionization in poor countries is short-sighted. 1/ We should be clear about what assumptions make this result. If you don't like it, challenge the assumptions, and show how it changes the results. (This is general advice). The main friction is that firms have a borrowing constraint, and can only be levered so much.

Caste is old -- really old. It was emphatically not the creation of the English colonists, but something that emerged thousands of years ago. The high castes, who oppressed and exploited the low castes, are not originally from India -- they came down from the steppes. When we look at the genetic origins of the caste, in the North they overwhelmingly from the Caspian Steppes. The Dalits, the untouchable mass of serfs, are what remain of the indigenous people of Northern India.

This paper is so unbelievably good. Absolutely blown away. Why is it the case in Kenya that the cost of food is so high, but the prices received by farmers so low? It's the intermediaries -- they all know each other, and they're colluding. 1/ The normal trouble for studies of "conduct", or the manner in which firms are competing with each other, is that you need exceptional data in order to tell things apart. What could be collusion might actually just be high marginal costs, or a particular demand curve.

A surprise to me -- in coal mines of the early 1900s in Pennsylvania, gun powder for blasting was not bought by the mining firm, but was purchased and brought by the workers. Why? It's likely an attempt to resolve a principal-agent problem. The coal miner of the day was rather independent. They were paid by piece rate, by weight, in their own little room. Now consider, what happens if the coal mine gives them powder for free? Well, they're gonna be incentivized to use all of it.

This incredible paper uncovers massive collusion between Indian textile firms to hold down wages. This collusion is enabled by "pro-worker" laws making it hard to hire and fire. Instead, a higher minimum wage would actually increase output and wages. 1/ Identifying collusion is one of the perennial problems of industrial organization, and we have not come to a satisfactory general test. The problem is that you cannot easily identify the way in which firms compete without knowing their costs.

Auctions are surprisingly infrequent, given that under many perfectly reasonable conditions they are the optimal mechanism. This paper provides a partial explanation -- if participation is costly, then sequentially shopping around is actually better for everyone. 1/ Let's get some context first. If you have some number of risk-neutral bidders with independent valuations who draw their valuations from known distributions and face no cost of bidding, the optimal mechanism is an auction with a reserve price at the midpoint of the distribution.

It is consistently the case the productivity is correlated with competition. But why? Some think that it's due to bad firms being chased out, because wouldn't every firm want to maximize profit? And yet, that's not what we see. Competition simply makes companies better. 1/ The correlation between competition and productivity is one of the most thoroughly documented findings in all of economics. We'd really like to know why this is the case, though. There are two main hypotheses:

Women bus and train drivers in Boston earn less than men, even in a unionized work place where conditions are all identical. It is difficult to think of this as being due to anything other than different preferences. Equalizing wages would make us worse off. 1/ The MBTA is a service with which I have gained great familiarity. They operate three subway lines (Red, Orange, and Blue), a partially underground set of tram lines (Green), bus rapid transit (Silver), and a thick network of 173 buses.

This team of researchers built an AI algorithm to recommend jobs to 2 million users . It substantially increased employment of the people who got to use it, at little cost to the people unable to use it. The biggest benefits went to poorer and less educated workers. 1/ This is exciting, because it's not necessarily clear whether better recommendation algorithms increase employment on net. This is a core problem for any sort of active labor market policy -- how do you know you haven't just pushed out another worker?

I see this line of thought brought up a lot when trying to reason about monetary policy. Unfortunately, it's completely wrong, but understanding *why* it is wrong will likely be useful. The short version is this: the price level is an outcome, not a causal primitive.

https://twitter.com/online_entropy/status/2039161343429808624

The idea is that if you expect prices to fall, then you won't spend now, which in turn will cause prices to fall. Suppose money is fixed. Why is the price level falling? If buyer and seller are symmetric, there's no reason to lower your prices in a second period.

Do you care about market power in the ready-to-eat cereal market? Probably not. I get that. But that's okay, that's not the point of this paper. The point is to understand market power in markets with differentiated products. You cannot evaluate mergers without it! 1/ The ready-to-eat cereal market is largely the breakfast cereal market with which you are familiar from childhood -- Cheerios and Raisin Bran and Frosted Flakes. It's dominated by a few large firms which produce many different products. Do they possess market power? How much?

In Brazil, investments into research doubled the productivity of agriculture in that country, a 17 fold return on investment. The case for public R&D is incredibly strong -- let's look at on great example. 1/ In Brazil, yields per acre had been flat for decades. In spite of its vast land, it was a net importer of food, and even needed donations from abroad. To solve this, the Brazilian government created Embrapa, an agricultural research group, in 1973.

Fertility has fallen in the United States. At the same time, housing has gotten substantially more expensive. Are these related? How so? Benjamin Couillard attributes fully half of the fertility decline of the 2000s and 2010s to housing prices, a loss of 13 million children! 1/ The paper is a substantial improvement over the reduced form approach to estimating the effect of housing prices on fertility. Using local variation is not ideal, because housing supply changes somewhere affect prices everywhere. We're not really sure what we're measuring.

Being able to sue your doctor for negligence increases costs by 5% while having no effect whatsoever on patients outcomes. All liability laws cause is ass-covering. 1/ "Defensive medicine" is ordering extra tests and procedures so that nobody can accuse you, after the fact, of not doing enough. This may look like a doctor ordering a cancer test needlessly, or perhaps an unnecessary biopsy.

It turns out that public transit is extremely important for reducing traffic. Because the relationship of traffic and congestion is extremely non-linear, a shutdown of transit can cause massive increases in congestion -- justifying substantial subsidies for trains and buses. 1/ tThis is a methodologically simple paper, but very clean. In October 2003, the LA public transit workers went on strike for 35 days. We can consider the change in congestion, after sanity testing it, to be the effect of the MTA.

What killed the local newspaper? Craigslist. The actual reporting function of the newspaper was never what made it profitable -- it was simply a way to get people to look at the classified advertisements. With that gone, the work with positive externalities died. 1/ This is a classic staggered roll-out paper -- if you are willing to buy the assumption that Craigslist expanded to different markets in a way that is unrelated to future trends, conditional on basic demographic controls, then you buy the paper.

Why we so often behave as though the world is zero-sum when it isn't? This paper shows that even the slightest bit of asymmetric information is sufficient to cause us to go away from the optimal outcome, and lead to people to vote for things they don't want in expectation! 1/ Consider three voters, Alice, Bob and Carol. They are deciding whether to enact policy P, and policy P*. P is a payoff of zero to all people. Policy P* gives two voters a pay off of 2, and one voter a payoff of -3. Which voter loses depends on the state of the world.

The marginal propensity to consume (or MPC) is one of the most important parameters to know if we want to estimate the effect of fiscal stimulus. Yet, we may not be able to know the real effect -- even from perfectly identified studies! Stimulus is less effective than we think.1/ Previous micro work on the MPCs take two basic paths. With Parker et al (2013), the identifying variation is coming from random variation in which people actually received their checks.