Health and Labour Market Outcomes of North Korean Refugees in South Korea

ABSTRACT

This study utilises the National Health Insurance Service of Korea (NHIS) data which include the results of annual and bi-annual health screening of North Korean refugees together with indicators of their employment and earnings and investigates the relationship between the refugees’ health and labour market outcomes. We find that the refugees with poor health significantly underperform healthy refugees in the labour market, and the impact of poor health at the time of entry lasts longer than 10 years. We also find that poor health has greater detrimental effects on refugees than on non-refugees.

KEYWORDS:

• North Korean refugees
• health and labor market outcomes
• National Health Insurance Service of Korea data
• health screening
• employment and earnings

JEL CODES:

• J15, I13
• I15

Acknowledgements

We would like to thank Chung Choe, Klaus Hornetz, Jongnam Hwang, Jeffrey Nilsen and participants at the 2021 Korea and the World Economy Conference and the 2023 Korean Association of Health Economics and Policy Conference for providing useful comments to earlier versions of this paper.

Disclosure Statement

No potential conflict of interest was reported by the author(s).

Notes

1 Literature suggests that other factors contributing to North Korean refugees’ difficulty in the labour market include lack of specific human capital (Jeong 2018), unfamiliarity with market capitalism (Kim and Kim 2019) and lack of fluid intelligence (Kim and Lee 2018).

2 One disadvantage of the NHIS data is that refugees’ socio-economic circumstances prior to their entry to Korea (e.g. education, work experiences, and family wealth in North Korea) are not available.

3 While these data have the advantage of containing pre-arrival socio-economic information as well as health status, they are not suitable for the study of integration since they do not contain labor market outcome in the years following refugees’ relocation.

4 The programme covers employed persons of all ages, self-employed or unemployed persons of all ages who are the heads of households, and all other people over the age of 40. In 2016, more than 17 million people out of 43 million adults were eligible for the programme. See National Health Insurance Service and Health Insurance Review and Assessment Service (2022).

5 Lee et al. (2022) report detailed statistics regarding refugees’ participation in health screening.

6 North Korean refugees also have low estimated glomerular filtration rates (Song and Choi 2018), which is known to be associated with high blood pressure, one of the four chronic diseases used to construct our health indicators.

7 In Section 4, we have a larger number of observations since we fill some observations with health data from adjacent years. Such filling-in scheme has not been adopted for the calculation of Figure 2.

8 We determine the year of entry from the first year that an individual appears in the NHIS database. As the first year in our database is 2002, we are unable to determine the exact year of entry if an individual entered South Korea prior to 2002.

9 Matching to nonrefugees is necessary for the analysis presented in Section 5. We have 15,562 refugees who have taken part in national health screening between 2009 and 2018. 1,233 of them are not matched to nonrefugees, resulting in the sample of 14,329 refugees. The data of matched nonrefugees are further described in Section 5.

10 Refugees participated in 36,370 health screenings between 2009 and 2018. The number of person-year observations in our data set is much higher because we have filled many person-year observations with health screening data from adjacent years. See three paragraphs down in the main text for further explanation.

11 Since its inception in 1977, the national health insurance system of Korea has rapidly expanded its enrolment and, after a short period of 12 years, in 1989, reached the population coverage. Thus, since 1989, everyone in Korea is covered by the national health insurance system.

12 This category also includes unemployed persons who have substantial assets. Since refugees are unlikely to have substantial assets, this complication is unlikely to affect our analysis.

13 Unlike in the US, being self-employed does not mean a high-income status in South Korea. The self-employed category includes irregular workers; moreover, many small businesses are only marginally profitable, making their owners’ earnings less than average employed persons’ earnings.

14 Those belonging to group (ii) dependents of employed persons, group (vi) dependents of self-employed, and group (v) low-income and other people receiving governmental aid make insignificant earnings, if any, and we treat their earnings to be zero in constructing our earnings variables.

15 One of our right-hand-side variables ($\mathrm{Femal}{e}_i$) does not vary over time. Three other right-hand variables ($\mathrm{Heigh}{t}_{i,t}$, $\mathrm{YearsSinceEntr}{y}_{i,t}$, $\mathrm{Ag}{e}_{i,t})$ do not exhibit independent variation over time. Inclusion of such time-invariant variables makes the fixed effect estimation impossible (since time-invariant variables cannot be distinguished from individual fixed effects). Wooldridge (2010, 326) and Greene (2018, 415) suggest random effect estimation to control for individual effects under such this circumstance; random effect estimation, however, is not favoured by applied researchers because of the restrictive assumptions required. We present random effect estimation of Equation (2) in the Appendix.

16 The estimated coefficients on $\mathrm{Ag}{e}_{i,t}$ and $\mathrm{Ag}{e}_{i,t}^2$. suggest that the effect of $\mathrm{Ag}{e}_{i,t}$ on $\mathrm{HaveEarning}{s}_{i,t}$ is the greatest around the age . of 33–37.

17 Being self-employed does not mean a high-income status. See endnote 13.

18 We allow a single nonrefugee observation to be matched with multiple refugee observations. Thus, the mapping from refugees to nonrefugees is “many-to-one” rather than “one-to-one.”

19 This matching procedure results in a group of nonrefugees having the same characteristics as refugees and makes refugee-nonrefugee comparison meaningless. In the appendix, we adopt a different procedure which allows us to compare refugees and nonrefugees in a meaningful way.

20 We report the summary statistics of Y variables ($\mathrm{HaveEarning}{s}_{i,t}$ and $\mathrm{EarningsPercentil}{e}_{i,t}$. of nonrefugees in the Appendix. Given the way we constructed nonrefugees data, the distribution of X variables (gender, age, height, weight, BMI, and health status) of nonrefugees is identical to that of refugees shown in Table 1.

21 Random effect estimates are reported in the Appendix. See endnote 15 for why panel estimation is not our preferred estimation method.