Working Papers

When does violence inside a village damage its ability to act together afterward? I ask this of South Korean villages twenty years after the Korean War. The war began in June 1950, when the North Korean army invaded South Korea and took most of it in six weeks. United Nations forces then landed at Incheon, behind the front, and the North Korean army retreated; some of its units in the southwest could not return north. In the villages those units still held, the communist side killed 59,964 civilians by the state's count of 1952, and local accusations helped decide who they were.

The evidence is a set of records I digitized. The 1972 survey grades how well every rural village in the country worked together. The state gave each village free cement and let its households decide together what to build. The survey records no coordinates. I located all 34,665 of its villages from its own sketch maps and from map series of 1918 and 1970. I read the 1952 register of the dead from 1,784 scanned pages by three independent machine transcriptions, reconciled against a hand-checked sample. I extended a colonial census of village lineages of 1930 to the whole country.

Villages beside the roads the army marched down were graded lower for cooperation than villages beside similar roads it could have taken and did not. However, two such villages in the same township do not differ. The difference is therefore at the township level, not the village level. The killings were heavily concentrated: 72 percent of the register's victims lived in one province, South Jeolla, in the southwest. There, villages split between two large lineages before the war were graded lower for cooperation where killings per household were heavier. Neither condition alone lowers the grade. Reshuffling the lineage census among the villages of a township rarely reproduces this pattern, but reshuffling the killings among them often does. The data tie the division to the village and the killings only to its township. In the rest of the country, where the killings were not concentrated, the pattern is weaker.

In October 1948 the South Korean army declared the interior of Jeju a prohibited area, and over the following winter it burned the villages there and moved their people to the coast. The government’s investigation counts 14,442 recognized victims by the village where they lived. We ask what the campaign left behind. We use houses detected on colonial and 1970 map sheets, the 1972 survey of every rural village, census population and credit-bureau records. Between 1918 and 1970 cells of 1 km more than 5 km inland lost houses relative to coastal cells, and on three islands that the campaign did not reach the same contrast is much smaller. Matching cells on elevation places the loss between 50 and 300 m, where the mid-mountain villages stood. Whether a cleared cell disappeared depended on how large it was in 1918: almost all single-house cells were gone by 1970 and no cell with fifty or more houses was. That is the pattern a settlement model with two stable states implies. Jeju also grew more slowly than the comparison islands, so the campaign did not only move people within the island. At the village level, villages that lost more people have fewer residents and less spending today, and we find no interaction with the pre-war division of a village between two lineages. The cells and villages are the population of Jeju, so we report design-based p-values from reassigning the treatment rather than standard errors.

In the early 1970s, about two thirds of rural Korean schoolchildren had intestinal helminths. Schools examined pupils each spring and each fall, and infected pupils received drugs after the examination. We use the examination results of 6,098 rural primary schools in 1971–1977 and link them to the 34,665 villages in the 1972 Saemaul village survey. We split the change in infection into two parts: the drop between the spring and the fall examination, and the rebound between the fall examination and the next spring. On average, infection falls by 12.6 percentage points between spring and fall and rises by 9.5 points by the next spring. The drop is larger for Ascaris, which the drugs in use treat, than for Trichuris, which they do not treat. The drop grows from 10.2 points in 1972 to 16.6 points in 1976, and the rebound does not grow. The decline in infection therefore comes from a larger drop. Schools that gave more tablets per infected pupil have a larger Ascaris drop and no larger Trichuris drop. Village conditions matter mainly through reinfection. For the same infection level after treatment, infection returns faster around villages with more thatched roofs, without electricity, and far from the township office. These conditions do not predict the size of the drop. Saemaul support in 1974–77 and clan concentration in 1930 predict neither the drop nor the rebound. In a first test of long-run outcomes, schools with higher baseline infection do not produce relatively more notable people, as listed in Korean Wikipedia, in the cohorts that were in school during 1971–77. A second test with credit-bureau data on a 1 km grid also finds no credible effect.

We study fertility in a dynamic spatial equilibrium model with endogenous location and residential choices, alongside investments in children's education. We estimate the model using near-universal administrative microdata from South Korea, drawing on census records, credit bureau data linked to mobile phone mobility records, and government administrative sources. Our estimates indicate that high commuting and housing costs, combined with intensive education investments in large cities, where the amenity of living is large, substantially reduce fertility. We conduct a decomposition exercise to explain the sharp decline in South Korean fertility over the past decade, finding that rising education and childcare costs play a significant role. Finally, we conduct a series of counterfactual exercises considering place-based policies to reduce commuting costs, increase housing supply, or lower education costs. We find that policies focusing exclusively on large cities, such as Seoul, may have the opposite effect on fertility than intended due to spatial sorting. However, a combination of nationwide policies targeting the fertility margin directly can boost fertility significantly.

As young adults in advanced economies have become increasingly pessimistic about their economic prospects in recent years, gambling-like financial activities have proliferated. We link these trends through a Friedman-Savage (1948) motive: when life goals such as homeownership or marriage seem out of reach, individuals turn to financial gambles. We test the theory using an original survey combined with millions of linked records from Korea’s leading credit bureau, telecom provider, and credit card issuer. Consistent with the theory, individuals who place greater importance on hard-to-reach financial goals are more likely to exhibit risk-seeking preferences and to engage in high-risk investment. Conversely, when goals become attainable, risk-taking declines: exploiting Korea’s housing lottery, which quasi-randomly allocates subsidies for home purchase, we show that attaining homeownership reduces interest in cryptocurrency. We conclude with a stylized calibration that shows how goal-driven risk preferences can amplify the welfare costs of overoptimism.