Publications

Land reforms are among the largest asset redistributions that governments undertake, but their long-run consequences are rarely traced at the level of the individual parcel. We link the mapped parcels of one Korean village through the 1950 land reform, using a colonial tax ledger, cadastral maps, the modern cadastral register, and geocoded business-license records that together span a century. The analysis is descriptive. We observe the 1945 ledger owner class of each classified plot but not its statutory reform path, so we report associations with owner class and not effects of redistribution. Our main finding concerns spatial structure before the reform. The mapped holdings of the village's large owners were already scattered before 1950. There were few large blocks of land for the reform to break up. Complementary comparisons show small adjusted differences in later official assessed values, while estimates for commercial activity and land-use conversion remain uncertain. Recorded ownership events provide partial evidence on the intervening decades, including preservation entries clustered in 1965, but they cannot establish when all parcels acquired clear title. The parcel-level reconstruction tests an assumption about pre-reform spatial structure that ownership totals alone cannot test.

We analyze the sellers, buyers, and ownership-transfer patterns of ex-ante disposals, or pre-reform private land sales that occurred during the South Korean land reform of 1950. Our analysis is based on the Jise Myeonggijang, or land-tax registers, compiled in Anryong-myeon, old Suwon-gun. The redistribution of farmland during the South Korean land reform proceeded through three channels: the disposal of vested land (formerly Japanese-owned land), the purchase and distribution of Korean landlord-owned land under the Land Reform Act, and pre-reform private land sales between landlords and peasants. The ex-ante disposals accounted for over half of the redistributed tenanted land, amounting to approximately 710 thousand jeongbos, and have thus been treated as a major driver of the reform. Despite this magnitude, empirical analyses of the specific realities of these transactions seem to remain relatively scarce. To examine this issue, we trace parcel-level changes in land ownership in the Anryong area from 1943 to 1949. We find that land transactions in Anryong were concentrated around 1948, when the disposal of vested land was executed and the Land Reform Act was taking shape. Notably, a substantial share of pre-reform private land sales occurred within local same-surname kinship groups of the Park and Yun families. Furthermore, we find that ownership transfers through ex-ante disposals contributed more to the maintenance or expansion of landholdings by existing landowners than to land acquisition and upward mobility by the landless poor. These findings suggest that while the ex-ante disposals partially accelerated the dismantling of landlordism and the redistribution of tenanted land, they may also have constrained the land-to-tillers principle and the efficient allocation of resources. Furthermore, this study suggests that when analyzing the long-term impacts of the land reform on rural societies, future research should consider that even if the total apparent scale of redistribution was similar, the relative proportion of ex-ante disposals versus legal redistribution may have carried significantly different implications for the local economy.

We construct a new grid-level historical dataset for South Korea by applying YOLO-based object detection to 1:50,000-scale topographic maps from 1918 and 1970, yielding 1 km×1 km counts of residential houses, rice paddies, and other annotated land-use symbols on the Korean national standard grid. The house variable captures the standardized small-square symbol used for individual rural dwellings and should be interpreted as a measure of dispersed rural settlement rather than total housing stock. Internal consistency checks, a cell-level temporal-persistence test between 1918 and 1970, and an external cross-check against the 1930 colonial census at the township level support the dataset’s use as a baseline spatial measure of rural settlement and agricultural land use. The dataset is made publicly available, providing a new resource for long-run regional analysis in Korean economic history.

We examine the short- and long-run effects of historical Protestant churches on human capital and female empowerment in South Korea by combining historical data on religious facilities and contemporary population census and social survey data. In the short run, we find a positive and significant effect of Protestant churches on women’s literacy and employment, but not on men’s. After almost a century, we find no strong relationship between historical Protestant churches and women’s educational attainment or economic and political empowerment.

Ethnic and linguistic heterogeneity are widely studied as determinants of social capital, conflict, and institutional quality. In many cultures, another important dimension of heterogeneity is family clan membership. I study the relationship between family clan diversity in South Korean villages and the voluntary production of public goods and contributions of private resources for village projects. Under the 1970–1971 New Village Beautification Project, the government distributed resources to each village for the production of village public goods. Subsequently, the government systematically evaluated how well these resources were applied. I combine these data with information on village family clan structures collected by the Japanese Colonial Government, as well as records of land donations for village projects between 1970 and 1980. I find an inverted-U-shaped effect of group heterogeneity on the improvement of public goods.

We investigate the impact of an increase in immigrant inflows on natives’ residential choices in a large metropolitan city. We utilize an administrative dataset containing information about residential address changes and self-reported reasons for relocation. Exploiting the expansion of a special visa program as an exogenous shock, we find that immigration inflows are both a push and a pull factor for natives. While neighborhoods in Seoul lost more than 6 natives for every 10 additional immigrants between 2006 and 2015, certain native workers were drawn to areas with immigrant inflows for job-related reasons.

This study explores how authoritarian distributive policies may not only generate political support for the autocrat but may also help sustain powerful and lasting authoritarian legacies. We use microlevel data from South Korea’s New Village Movement, a 1970s rural development program implemented under dictator Park Chung-hee and widely touted as contributing to the country’s rapid economic development. Our analysis shows that townships in receipt of larger cash transfers cast more votes for Park’s incumbent party in the subsequent election. More importantly, we show that the effects of the subsidies still appeared almost four decades later in 2012, when the dictator’s daughter was democratically elected as the president of South Korea. We show that these effects were not driven by villagers’ long-term income gains or enhanced social capital due to the program but by the unwavering support of the beneficiary villagers for the dictator, whose legacy remained strong long after democratization.

Machine learning approaches using satellite imagery are providing accessible ways to infer socioeconomic measures without visiting a region. However, many algorithms require integration of ground-truth data, while regional data are scarce or even absent in many countries. Here we present our human-machine collaborative model which predicts grid-level economic development using publicly available satellite imagery and lightweight subjective ranking annotation without any ground data. We applied the model to North Korea and produced fine-grained predictions of economic development for the nation where data is not readily available. Our model suggests substantial development in the country’s capital and areas with state-led development projects in recent years. We showed the broad applicability of our model by examining five of the least developed countries in Asia, covering 400,000 grids. Our method can both yield highly granular economic information on hard-to-visit and low-resource regions and can potentially guide sustainable development programs.

The increasing frequency and severity of water-related disasters such as floods, tornadoes, hurricanes, and tsunamis in low- and middle-income countries exemplify the uneven effects of global climate change. The vulnerability of high-risk societies to natural disasters has continued to increase. To develop an effective and efficient adaptation strategy, local damage assessments must be timely, exhaustive, and accurate. We propose a novel deep-learning-based solution that uses pairs of pre- and post-disaster satellite images to identify water-related disaster-affected regions. The model extracts features of pre- and post-disaster images and uses the feature difference with them to predict damage in the pair. We demonstrate that the model can successfully identify local destruction using less granular and less complex ground-truth data than those used by previous segmentation models. When tested with various water-related disasters, our detection model reported an accuracy of 85.9% in spotting areas with damaged buildings. It also achieved a reliable performance of 80.3% in out-of-domain settings. Our deep learning-based damage assessment model can help direct resources to areas most vulnerable to climate disasters, reducing their impacts while promoting adaptive capacities for climate-resilient development in the most vulnerable regions.

We conducted a randomized controlled field study to explore the effectiveness of sensory marketing on beverage consumption patterns in a real bar setting. Specifically, we examined (a) the effect of visual elements (i.e., consumption-inducing text messages on coasters), (b) the effect of social density, and (c) the joint effect of visual elements and social density. We manipulated coaster type (visual consumption-inducing messages either present or absent), measured social density, and collected sales data. The results show that visual elements have a significant effect on beverage consumption, but social density does not. The joint effect between the two factors is significant such that the effect of visual elements is higher when social density is low. This study contributes to the sensory marketing literature by revealing the interaction between visual and spatial cues in a field setting.

Do the effects of a mega-event on a local economy persist even after the event has ended? Using a novel administrative big data set of buildings in the host regions, we analyze the long-term effects of the 2018 PyeongChang Olympics. For data collection, we converted building-level GIS data into a township-level data set. We used the synthetic control method which enables causal analysis even with a small number of treated units. Our analysis reveals that, despite a short-term increase in the number of local buildings due to the Olympics, effects did not persist 1 year after the event.

While measuring socioeconomic indicators is critical for local governments to make informed policy decisions, such measurements are often unavailable at fine-grained levels like municipality. This study employs deep learning-based predictions from satellite images to close the gap. We propose a method that assigns a socioeconomic score to each satellite image by capturing the distributional behavior observed in larger areas based on the ground truth. We train an ordinal regression scoring model and adjust the scores to follow the common power law within and across regions. Evaluation based on official statistics in South Korea shows that our method outperforms previous models in predicting population and employment size at both the municipality and grid levels. Our method also demonstrates robust performance in districts with uneven development, suggesting its potential use in developing countries where reliable, fine-grained data is scarce.

High-resolution daytime satellite imagery has become a promising source to study economic activities. These images display detailed terrain over large areas and allow zooming into smaller neighborhoods. Existing methods, however, have utilized images only in a single-level geographical unit. This research presents a deep learning model to predict economic indicators via aggregating traits observed from multiple levels of geographical units. The model first measures hyperlocal economy over small communities via ordinal regression. The next step extracts district-level features by summarizing interconnection among hyperlocal economies. In the final step, the model estimates economic indicators of districts via aggregating the hyperlocal and district information. Our new multi-level learning model substantially outperforms strong baselines in predicting key indicators such as population, purchasing power, and energy consumption. The model is also robust against data shortage; the trained features from one country can generalize to other countries when evaluated with data gathered from Malaysia, the Philippines, Thailand, and Vietnam. We discuss the multi-level model’s implications for measuring inequality, which is the essential first step in policy and social science research on inequality and poverty.

Reliable and timely measurements of economic activities are fundamental for understanding economic development and designing government policies. However, many developing countries still lack reliable data. In this paper, we introduce a novel approach for measuring economic development from high-resolution satellite images in the absence of ground truth statistics. Our method consists of three steps. First, we run a clustering algorithm on satellite images that distinguishes artifacts from nature (siCluster). Second, we generate a partial order graph of the identified clusters based on the level of economic development, either by human guidance or by low-resolution statistics (siPog). Third, we use a CNN-based sorter that assigns differentiable scores to each satellite grid based on the relative ranks of clusters (siScore). The novelty of our method is that we break down a computationally hard problem into sub-tasks, which involves a human-in-the-loop solution. With the combination of unsupervised learning and the partial orders of dozens of urban vs. rural clusters, our method can estimate the economic development scores of over 10,000 satellite grids consistently with other baseline development proxies (Spearman correlation of 0.851). This efficient method is interpretable and robust; we demonstrate how to apply our method to both developed (e.g., South Korea) and developing economies (e.g., Vietnam and Malawi).