What do we know about poverty in North Korea?

The North Korean economy is notoriously opaque, lacking reliable official data on income or production since 1990. Existing GDP per capita estimates are based on extrapolations and lack precision. This study leverages the growing use of remote sensing data, particularly night-time light emissions, as a proxy for economic activity in data-scarce environments. The authors utilize monthly night-time light data from the Suomi National Polar-orbiting Partnership Visible Infrared Imaging Radiometer Suite (NPP/VIIRS) to estimate GDP at the subnational level in North Korea. This is combined with distributional assumptions based on comparable economies to estimate poverty rates. The study aims to provide novel empirical evidence on the level and dynamics of economic development and poverty in North Korea, offering a methodology applicable to other data-limited contexts. The authors hypothesize that existing poverty estimates may underestimate the true extent of extreme poverty in North Korea and aim to test this hypothesis using novel data and methodologies.

The introduction cites several studies that use remote sensing data, particularly night-time light data from the DMSP-OLS satellite (now superseded by VIIRS), to analyze socioeconomic trends globally and in data-scarce regions. The authors acknowledge previous attempts to estimate poverty in North Korea (Crespo Cuaresma et al., 2018), suggesting their method may refine these earlier estimates. The study references the Williamson hypothesis (Williamson, 1965), predicting an inverse relationship between development and regional inequality, which is used as a comparative framework for the analysis of regional income dynamics within North Korea. The literature review highlights the lack of robust data and the need for innovative approaches to understanding the North Korean economy.

The study uses monthly VIIRS DNB night-time light data from 2012-2018, focusing on built-up areas identified using the World Settlement Footprint 2015. Median luminosity values are used to minimize outlier effects. Corrections are applied to 2017 and 2018 data to account for a radiance intensity shift. The relationship between night-time light and GDP, established in Li et al. (2013) using Chinese prefecture data, is applied to convert luminosity data into GDP estimates for North Korean subnational regions. Population data from the 2008 census are extrapolated to 2018 using national-level growth rates. To estimate poverty rates, the authors employ a method based on Beta-Lorenz curves and match North Korean socio-demographic and economic data (sectoral employment, educational attainment, and GDP per capita from luminosity data) with those of similar economies. The Euclidean distance between North Korea and other economies is calculated based on these characteristics, identifying the closest economies (k=7). The median poverty rate of these closest economies is used as an estimate for North Korea. The authors test several variations of this approach, including different values of k and weighting schemes, using a cross-validation exercise involving 40 country/year observations with available poverty data. Root-mean-squared error (RMSE) is used to evaluate the different models, selecting the best performing one (k=7) for the final poverty estimates for North Korea.

The luminosity-based estimate of GDP per capita in North Korea for 2015 was approximately $790 (2011 PPP-adjusted). Income per capita showed an overall downward trend from 2012 to 2016, followed by recovery in 2016-2018. Regional income dynamics exhibited divergence from 2012-2016 and convergence from 2016-2018. The study estimates a national poverty rate of around 60% in 2018, significantly higher than previous estimates. This estimate is based on the median poverty rates of the seven most similar economies to North Korea, which include Romania (1999), Albania (2002), Madagascar (2002), Georgia (1996), Bangladesh (2016), Vietnam (2010), and Armenia (1996). The number of people living in extreme poverty is estimated between 14.9 million and 17 million. Regional poverty rates vary substantially. The study’s poverty rate estimates show a negative correlation with GDP growth rate estimates from the Bank of Korea for 2013-2016. The analysis also revealed high poverty levels in southern regions, potentially linked to conflict with South Korea. The relatively low poverty rates in Yanggang-do are cautioned due to assumptions about zero population growth differentials.

The findings challenge previous estimates of poverty in North Korea, suggesting a substantially higher incidence of extreme poverty than previously thought. The use of night-time light data, combined with the method of matching income distributions from similar economies, provides a novel and potentially more accurate assessment of the situation. The observed divergence and subsequent convergence in regional income and poverty rates highlight dynamic shifts in the North Korean economy, though the specific causes require further investigation. The negative correlation between the study's poverty rate estimates and the Bank of Korea's GDP growth rate provides added support for the study's findings. The methodology developed offers a valuable approach for analyzing economic indicators in data-scarce environments. Future research could explore the underlying drivers of regional disparities and the impacts of potential policy changes on poverty reduction.

This study presents the first subnational estimates of poverty in North Korea using novel methodologies. The findings reveal a significantly higher poverty rate than previously reported, highlighting the extent of extreme poverty in the country. The analysis of regional disparities reveals dynamic shifts in the economic landscape. The study’s methodology, employing night-time light data and comparative income distributions, offers a valuable tool for analyzing economic conditions in data-scarce settings. Future research should focus on refining these estimates, understanding the drivers of regional disparity, and investigating the impact of economic policies on poverty reduction in North Korea.

The study acknowledges limitations stemming from data scarcity. The reliability of the GDP-luminosity relationship derived from Chinese data when applied to North Korea is an important consideration. The assumption of constant population growth across regions and the lack of information on internal migration could bias the assessment of income convergence. The reliance on comparable economies to estimate income distributions introduces uncertainty, though the cross-validation exercise demonstrated the robustness of the chosen method. Further validation with additional data would strengthen the conclusions.