The impact of city epidemics on rural labor market: The Korean Middle East Respiratory Syndrome case.

This study compared the changes in employment in urban areas in Korea, where a large number of people were quarantined by the Middle East Respiratory Syndrome epidemic, to those in rural areas, where only a small number of people were quarantined using the difference-in-difference approach. The results indicate that the urban labor market experienced a direct effect in terms of a reduction in employment of the group vulnerable to the epidemic while the rural labor market experienced an indirect effect on its economy through a reduction in employment resulting from a decline in consumption and leisure activities. If one looks into the employment in the accommodation and leisure industry, which sustained the most severe blow, dropped to its lowest level right after the Middle East Respiratory Syndrome outbreak. The rural leisure and accommodation industries are highly likely to be dependent on consumption and spending from urban areas. The results suggest that the rural labor market was influenced by the spillover/external effects caused by behavioral changes among people in urban areas due to fear of infection. Thus, this empirical analysis can be used to customize policy to support regions that can be negatively impacted by spillovers due to epidemic in order to respond against economic stresses.

Introduction

In June 2015, people in Korea avoided seeing people in public or going to meetings, and schools were temporarily closed. When people had to go outside, they wore masks, and people who sneezed without wearing masks were finger-pointed as lacking common sense. This was due to the Middle East Respiratory Syndrome (MERS) epidemic that was first confirmed on May 20, 2015. In fact, MERS is a contagious disease that has low morbidity and mortality rates, but the MERS epidemic in Korea had a significant impact on the labor market and on the wider economy due to the large number of quarantined people (about one in every 3000 Koreans) and excessive media coverage (Bank of Korea, 2015, Mediaus, 2015).

School notices on ‘preventive measures against the MERS’ instructed students to “Report to their teachers in case they came in contact with people with MERS symptoms, to avoid using computers and smartphones too much, and to not be disturbed by misleading information on MERS”. The former chairman of the Korean Medical Association said, “Korean society was faced with two daunting challenges. One is the MERS itself, and the other is the overblown fear of MERS” (Noh, 2015). According to Noh (2015), fear spread quickly and was amplified to the extent that it turned Pyeongtaek, one of the cities adjacent to the Korean capital city of Seoul, into a ‘ghost city’. He stressed, “If one compares MERS to tuberculosis, the latter is a more prevalent airborne disease and is more lethal than MERS, given that it claimed 2466 lives in 2013 alone. However, few people refrain from going outside and meeting people or wear masks all day out of fear of tuberculosis, and temporary school closure will not be issued either. Such an excessively chaotic situation as now is not desirable.” On June 11, 2015, after all, the Bank of Korea took a drastic measure to lower its base interest rate by a 0.25% point. The main reason for the reduction interest was not a delayed economic recovery but the economic losses caused by MERS as well as concerns on the worsening economic situation. Previous studies have said that in addition to a high mortality and confirmation rate, the fear of infection spread through rumors and mass media could have a significant influence on the economy (Eichelberger, 2007, Fenichel, 2013). Therefore, despite its low morbidity and mortality, the MERS epidemic might have a great impact on the labor market and on the entire economy of South Korea.

There is a close correlation between employment and disaster (specifically, an epidemic disease) (Lee and Warner, 2005, Shama and Krishna, 2007, World Bank, 2014, Lee and Cho, 2016). Infectious diseases can deal a blow to employment of vulnerable groups (Lee and Cho, 2016), and the standard of living, disposable income, savings, education, and health and overall psyche of a society are determined by current levels of employment and unemployment (Shama and Krishna, 2007). Therefore, it is very important to study employment in relation to disaster (infectious diseases). Moreover, given that urban areas are closely connected with rural areas, they inevitably influence each other (Srivastava and Shaw, 2012), and as such, an unexpected externality can occur (O’Sullivan, 2007). Here, the term ‘externality’ refers to people in a community that can benefit from or suffer loss due to a decision made by another community, i.e., an external decision. Even if an infectious disease takes place in an urban area, decisions made by people in the urban area can have an unintended influence on people in rural areas.

When an infectious disease like MERS breaks out, its possible impact on a country’s labor market and on the wider economy can appear in the form of direct or indirect effects (Lee and McKibbin, 2004a, Lee and McKibbin, 2004b, World Bank, 2014). The direct effects caused by an infectious disease include a decrease in the labor supply of vulnerable groups due to their relative high morbidity and mortality rate or fear of contagion. On the other hand, the indirect effects are those caused by people’s behavioral changes. These behavioral changes include avoiding consumption and leisure activities for fear of contagion, which in turn results in a decline in the consumption of goods and services, ultimately leading to a drop in domestic employment. Therefore, the groups of people vulnerable to the MERS epidemic (with a high confirmation and mortality rate) in an urban area where a large number of people are quarantined may have their employment activities directly affected. In contrast, a rural area may suffer from a decrease in the consumption of goods and services due to urban dwellers’ behavioral changes, such as a decrease in leisure activities and consumption, which is then expected to negatively impact rural employment. In particular, accommodation and leisure industry can play an important role in stimulation the rural economy (Kadiyali and Kosová, 2013). Unlike local employment, which is defined as that in workplaces that operate to fulfill local consumption, those employed in the rural accommodation and leisure industry are highly dependent on non-local(urban area) consumption and spending (O’Sullivan, 2007). Moreover, there is general belief that leisure industry (tourism) generates spillovers/externalities to other industries, thus boosting local economies (Kadiyali and Kosová, 2013). Eventually, due to such spillovers, there is a possibility that a negative externality occurs in the urban-rural labor market.

This study compared the damage aspects of a labor market which is directly hit by an epidemic with those which is not directly hit. To this end, this study analyzed the changes in employment in urban areas in Korea, where a large number of people were quarantined, to those in rural areas, where only a small number of people were quarantined. In particular, since the urban and rural areas are closely linked, they inevitably influence each other. However, the scenarios that unfold in urban and in rural areas may be distinct. An epidemic can occur at any time in an unexpected way, and so this study investigates the effects that disasters, such as epidemics, can have on employment, which directly influence peoples’ livelihoods in both urban and rural areas. In summary, the results of this study provide basic data to improve the response against economic stresses caused by urban-rural disasters.

Related literature

Eichelberger (2007) stated that when SARS occurred in 2003, the American public, particularly those near New York City’s Chinatown, had become infected with an epidemic of fear, not of disease. The news media speculated the possibility of a domestic epidemic as a rumor that a restaurant employee had been inflected with SARS spread through e-mail. 14% of survey respondents said they avoided Asian businesses, and New York City’s Chinatown experienced a heightened anxiety and fear of stigmatization. Eventually, the rumor and its news coverage caused a tremendous drop in business and tourism in Chinatown. Even without a single confirmed case of SARS in New York City’s China town, the community was quickly identified as a contagion and risk site (Eichelberger, 2007). As shown in this study, it is not only the high mortality and confirmation rate of an infectious disease that can have a significant impact on the economy, but also the fear of ‘contagion’ spread through rumors and the mass media. Fenichel (2013) analyzed the effect that contagious disease can have on social distancing and social welfare. The author maintained that social distancing and quarantine policies tend to show an ‘over-done’ phenomenon, which leads to a decrease in welfare. This can cause economically undesirable outcomes, and can also potentially exacerbate the health outcomes.

Studies on the post-disaster employment are very important. Shama and Krishna (2007) indicated that since the urban and rural poor are dependent on their daily wages, they are the groups that are vulnerable to disasters due to the strong correlation between disaster and employment. They added that the standard of living, disposable income, savings, education, health and overall psyche of a society are determined by current levels of employment and unemployment. The World Bank (2014) reported that an infectious disease can exert an impact on the economy through two channels. First, a direct effect of sickness and mortality can reduce the labor supply temporarily or even permanently. Second, people’s fear of contagion can result in behavioral changes where fear of contagion through contact with other people reduces the labor force participation and leads to the closure of workplaces. The SARS epidemic outbreaks between 2002 and 2004 and H1N1 flu epidemic outbreak in 2009 resulted in behavioral effects that were responsible for 80–90% of the actual total impact of the contagious diseases on the economy (Lee and McKibbin, 2004a, Lee and McKibbin, 2004b, World Bank, 2014). In Liberia, for example, many hotel and restaurant employees in the service industry were laid off, and the number of available jobs were reduced to half (World Bank, 2014). Lee and Warner (2005) analyzed the effects of SARS on Hong Kong’s economy. After investigating employment and unemployment conditions in the labor market, the study concluded that the SARS epidemic caused a drastic drop in room occupancy rates in hotels, which in turn led to an increase in unemployment or underemployment and also to a surge of no-pay leaves.

Moreover, Kadiyali and Kosová (2013) analyzed the effect of inter-industry spillovers generated by tourism/outside visitors of rural industry on employment across other (non-accommodation and food) industries. Hotel industry provided by generating jobs in various rural industry spillovers/externalities to other industries, thus boosting rural economies. Also, unlike local employment, which is defined as that in workplaces that operate to fulfill local consumption, those employed in the rural accommodation and leisure industry are highly dependent on non-local(urban area) consumption and spending (O’Sullivan, 2007). Therefore, accommodation and leisure industry play an important role in stimulation the rural economy (Kadiyali and Kosová, 2013).

Few studies have investigated the effect that infectious diseases have on both urban and rural labor markets. Lee et al. (2003) analyzed the impact that the IMF crisis had on the urban and rural labor markets. The results of the analysis indicated that the IMF crisis had a negative impact on employment, especially for elderly workers in the rural labor market. Lim (2000) analyzed the effect of the East Asian crisis on employment in the Philippines by dividing it into rural and urban areas. The results of their analysis indicated a reduction in unemployment in the urban areas that mostly benefited young people aged 15–34 for females and aged 20–34 for the males. There was also some reduction in the rural unemployment rates for females, especially those aged 20–34. However, there was no improvement in employment rates for males in rural areas. Thus, rural and urban areas that are closely linked to each other are inevitably subject to influence by a certain shock, but the impact they suffer may be different for each case. Srivastava and Shaw (2012) said that even though difference scenarios unfold in urban and rural areas, they are interrelated to the extent to which one has influence over the other, predominantly after a disaster. In addition, they showed a strong interrelation in a negative aspect but a weak association in a positive aspect (Lee et al., 2003).

The MERS epidemic

The MERS epidemic first occurred in 2012 in Saudi Arabia and then spread to the Arabian Peninsula. Then in 2015, the Republic of Korea experienced a large MERS outbreak, producing the largest number of MERS patients after Saudi Arabia (Zumla et al., 2015). MERS is an epidemic that spreads through MERS-CoV in homes, hospitals and even through close contact between people. The impact was higher due to the psychological unfamiliarity with the disease, unspecific causes and non-existence of a vaccine (Assiri et al., 2013a, World Health Organization, 2013). MERS can range from asymptomatic to severe pneumonia with acute respiratory distress syndrome (ARDS) (Assiri et al., 2013b). The number of people infected around the world stood at 1599 persons, with a mortality rate of 35.9%, or 574 persons (WHO, 2015).

On May 20, 2015, the first confirmed MERS patient in South Korea was a Korean male who had visited the Middle East. The total number of confirmed patients was recorded at 186 persons, and among them, 38 people lost their lives. In addition, 1164 schools were temporarily closed, and the biggest reason for which the MERS epidemic had a huge impact was the quarantine of 16,752 people. When considering the entire population in Korea, one in every three thousand people was quarantined. In addition, the lack of information and the perceived loss of control amplified fear and anxiety among people (Fig. 1 ).

Fig. 1

Status of MERS isolated in Korea.

Fig. 2 shows the number of confirmed cases and the mortality rates by age. Although the infected patient population was evenly distributed throughout different age groups, the number of confirmed patients and the mortality rate increased among those aged 50 years or above. Therefore, people of all age groups can be assumed to inevitably be fearful of the MERS epidemic. Among them, those aged more than 50 years were found to have a relatively higher risk rate than the other age groups, and it can be assumed that their fear level might be relatively higher than that of other age groups.

Fig. 2

Deaths and infections from MERS in Korea by age groups.

Fig. 3 shows the distribution of quarantined people in urban and rural areas. The quarantined people in Korea were heavily concentrated in a certain urban area, and the number of quarantined people increased dramatically in areas adjacent to the place of the first outbreak. In urban areas with a highly dense population, the number of quarantined people was extremely high. As of June 9, 2015, the distribution map shows that 90% of the quarantined people were concentrated in urban areas. Thus, the MERS epidemic dealt a severe blow to urban areas in South Korea. Urban areas close to the South Korean capital city of Seoul had a high concentration of quarantined people. After June 9, the number of quarantined people surged, and those who came into contact with confirmed patients or previously quarantined people were also put into isolation. In other words, the quarantine was additionally expanded to urban areas that the distribution map indicated already had a presence of quarantined people (The Chosunilbo, 2015). The main reason for the lower number of people quarantined by the MERS epidemic in rural areas was that the disease was mainly transmitted in hospitals in cities (urban regions), so the number of quarantined people was heavily concentrated in these urban areas. All in all, urban areas suffered the direct impact caused by the disease itself.

Fig. 3

Distribution of quarantined people by urban and rural (as of June 9, 2015).

Note: It is the national distribution of people in quarantine by urban and rural areas, and the number of the quarantined people was recorded at 2538 persons as of June 9 (Rural areas: 9.93% N = 212, Urban areas: 90.07% N = 1923).

Methods

Conceptual framework and empirical strategy

The first case of MERS in Korea occurred in the urban areas adjacent to the capital city of Seoul, and the disease quickly spread to the neighboring regions due to the highly dense population. 90% of the quarantine cases occurred in urban areas. According to Local Area Labour Force Survey data officially provided by Statistics Korea, 98% of employees have residences located in the same region as their workplaces. In other words, there was a slim chance for the risk of contagion between regions to increase due to work activities of urban dwellers commuting to workplaces in rural areas (1.90%) or of rural dwellers commuting to workplaces in urban areas (0.60%). Despite that, since the urban and rural areas are closely linked to each other, they inevitably influence each other and are subject to the influence of a certain shock (Strivastava and Shaw, 2012). In the case of the onset of an epidemic disease, it is natural for people show behavioral changes, including a decrease in travel and leisure activities. The leisure (travel) and accommodation industries in rural areas, which took the brunt of the MERS epidemic, are highly likely to be maintained through non-local (urban) consumption and spending. Therefore, if an epidemic breaks out in a city, urban dwellers’ decisions can have an unintended effect on people in rural areas.

After all, it can be assumed for an employment in labor markets of rural areas to suffer a negative impact because leisure activities and consumption will decrease due to behavioral changes caused by urban dwellers’ fear of contagion. In contrast, in urban areas subjected to the direct impact of a contagious disease, it is the group vulnerable to MERS that can suffer a direct blow to their employment condition. Middle-aged and elderly people can suffer from a more severe impact (Lee and Cho, 2016) because they have a high mortality risk when infected with the MERS disease through human contact. Therefore, the middle-aged and elderly have a high probability of a reduction in their labor supply either through voluntary or involuntary means. The age group vulnerable to the MERS (those aged more than 50) suffered a direct shock, which is defined in this study as a direct effect (S ). The other age group (between 15 and 49 years), a relatively more economically active group, are subject to an indirect shock due to behavioral changes, including a decrease in consumption and leisure activities, and this is defined as the Indirect effect (S ). In this study, the hypothesis about the effect of an infectious disease outbreak in an urban area on rural and urban labor markets was defined as Eq. (1).

The following empirical analysis was conducted to verify this hypothesis. First, the study analyzed the negative impact of the MERS epidemic on the labor market through employment status variables. This study also divided it into urban and rural areas and analyzed the respective influence of the MERS epidemic on urban areas that were subjected to a direct impact and on rural areas that were under an indirect impact.where y is the outcome for person i at time t, post denotes a dummy variable that equals one if the observation comes from any month after the MERS epidemic (e.g., June 2015 and July 2015) and zero otherwise, and X denotes a vector of individual characteristics including age, gender, educational attainment, marital status (Table A1). trend controls time-varying influences (starting from June 2010). A falsification check was conducted, and the results were directly compared to those in previous trends to determine the effect after the MERS epidemic. As a falsification check, we incorrectly specify the timing of the MERS epidemic. To do this, we estimate models using pseudo intervention dates with a range from 2010 to 2014. For instance, the epidemic event that actually happened in June 2015 was defined as one that occurred in June 2011, and then the pre-period June to July 2010 and the post-period June to July 2011 were set up for the analysis.

Table A1

Lists of covariates.

Age	Respondent’s age
Gender	Reference: Male
	Female
Education attainment	Reference: Element school
	Middle school
	High school
	College
	University
	Graduate school
Marital status	Reference: Single
	Married with spouse
	Widowed
	Divorce

Second, our central method is a difference-in-differences approach where we compare the change in the employment status of a target group to that of a comparison group.where y is the employment status for person i at time t, post denotes a dummy variable that equals one if the observation comes from June 2015 and July 2015 (after MERS), treated equals one if i is in the treatment group, and X denotes a vector of individual characteristics including gender, educational attainment, marital status (and age). The parameter β 3 provides the estimated effect of MERS on the labor market treated relative to the control group.

As seen in Eq. (3), an analysis of the regression-adjusted difference-in-differences(DD) effects of the MERS epidemic on employment of rural residents relative to urban residents was performed to determine the differences in the impact of the contagious disease on the rural and urban labor markets. This analysis was also performed using a DD approach by defining the vulnerable age group of 50 years or above as the treatment group and those aged less than 50 years as the control group.

To use the DD approach,1 the treatment is assumed to be exogenous. The hypothesis is constituted considering that an epidemic can occur in an unexpected manner. Also, it is very important to assume a common trend between the treatment and control groups before an epidemic occurs. To hold the common trend assumption, this study conducted an analysis according to the method used by Ratcliffe and Scholder (2015).

A robustness check was conducted by controlling the differential trends between the treatment and control groups for the common trend assumption, as seen in Eq. (4). We rerun the DD analysis with the additional inclusion of linear time trends for the treated and control groups. trend is a time trend (starting from June 2010) to account for any differential trends in the labor market outcome prior to the MERS epidemic. We examine the effects of MERS on the labor market by using pre-treatment years of data (i.e., from June 2010) to include monthly (2010m06, 2010m07, 2011m06, …) time trends.

Data

The data used in this study is obtained from the Economically Active Population Survey (EAPS), the official monthly labor force survey of Korea.2 The EAP is the most widely used micro-level labor survey that provides basic information on unemployment on Korea. It is similar to official labor force surveys in other countries, and it represents the Korean labor market with a sample of approx. 33,000 households in Korea. EAPS provides basic data required to analyze macro-economic conditions and set up policies to nurture human resources by researching the characteristics of economic activity (employment, unemployment, labor force, demographic characteristics, etc.). Using sample weights, raw data are inflated to reflect the relevant population as well as to avoid sampling differences across the different waves of EAPS.

To ensure that the labor market outcome follows a similar trend and to control the seasonal variation in treated and control groups prior to MERS, we restrict our main analysis to a time window consisting of two months after the MERS and the same period of the previous year (i.e., June to July 2014 and June to July 2015). To take the differential trends into consideration, data starting from June 2010 (e.g., monthly; June 2010, July 2010, June 2011, July 2011…) were additionally used. The reason to determine 2010 as the starting year was to select an analysis period at a time when the aftermath of the 2008 financial crisis had disappeared. The age of the subjects ranged between 15 and 70 years old. The low limit in the study was 15, which is the minimum age for work according to labor law, and the high limit was 70, the actual age of retirement in Korea stated in the OECD (2015).

In Korea, rural areas are usually distinguished from urban areas by using administrative units, such as ‘Si’ (city) or ‘Gun’ (county) (Korea City Administration Institute, 2006). The term ‘Si’ (an urban area) is a local administrative unit that refers to a municipal autonomous district with more than 50,000 people whereas ‘Gun’ (a rural area) is a local administrative unit that is bigger than ‘Eup (town)’ or ‘Myeon (township)’ but is smaller than ‘Do’ (province). This analysis was conducted by categorizing rural and urban areas according to these definitions. The dependent variable in the analysis is the current employment status (employed or unemployed). Other studies on labor markets are usually conducted by dividing subjects into categories of employed and unemployed (Lee and Zhee, 2001).

Results

Visual evidence

Table 1 shows the economic activity according to period and region, as was used for the analysis. Right after the onset of MERS in June and July 2015, the unemployed status increased relative to the same period of the previous year and of 2010–2014 across the nation in rural and urban areas. Also, the rural areas showed a greater increase in the unemployed status than in urban areas.

Table 1

Summary Statistics.

	2010 June, July–2014 June, July		2014 June, July		2015 June, July
Overall	Mean	S.D.	Mean	S.D.	Mean	S.D.
Employed	0.637	0.481	0.645	0.479	0.648	0.478
Unemployed	0.022	0.148	0.024	0.153	0.026	0.159
Out-of-labor force	0.341	0.474	0.331	0.471	0.326	0.469
Inflated sample size	379,770,000,000		77,106,979,549		77,799,529,310
Raw sample size	579,626		109,759		107,629

If we look into the actual unemployment rate ([the number of unemployed persons/the economically inactive population] * 100) during the analysis period, the urban areas showed a higher absolute unemployment rate than rural areas. However, right after the onset of the MERS epidemic, the rural areas showed a higher unemployment variation over the same period of the previous year than in the urban areas. However, the unemployment rate in June and July 2014 was higher relative to the same period of the previous year. The reason for this was presumed to be the Sewol Ferry disaster in April 2014. This was one of the most serious human-made disasters, with 291 deaths and 13 missing individuals as of June 8, 2014. The disaster caused nationwide sadness and sluggish demand and employment in the accommodation, restaurant, arts, sports and recreation sectors (Ministry of Strategy and Finance, 2014). Also, the study was set with the year 2010 as the starting point because the ramifications from the 2008 financial crisis had almost disappeared by that time, which we can assume had ripple effects that remained up until 2010. If we compare the overall trends, both the rural and urban areas showed a remarkable increase in unemployment after the occurrence of the MERS epidemic. In particular, the rural areas revealed a dramatic surge in unemployment, and the urban areas showed the highest unemployment rate in June 2015 when the fear of contagion was at its peak. The unemployment rate in July 2015 decreased relative to that of June of the same year since the fear of contagion had subsided. However, the rural areas displayed a remarkable increase in unemployment in June 2015 and continued to show an increase until July. Therefore, the rural areas can be assumed to have sustained a different type of impact when compared to urban areas that suffered a direct shock (Fig. 4 ).

Fig. 4

Unemployment rate for urban areas and rural areas (%).

Note: Unemployment rate(%) = (The number of unemployed persons/The economically active population)*100

Fig. 5 showed the results of a comparison of unemployment rates between the vulnerable group of those aged more than 50 years (with a high morbidity and mortality group regarding MERS) and the non-vulnerable group of those aged less than 50 years (a relatively more economically active group), both in urban and rural areas. After the onset of the MERS epidemic, the urban areas showed a dramatic increase in unemployment among those aged more than 50 years. If one looks into unemployment trends in urban areas since 2010, those aged 50 years old and above displayed the highest unemployment rate after the onset of MERS. In contrast, if we look into unemployment trends in rural areas since 2010, both those aged more than 50 years and those aged less than 50 years showed the highest unemployment rates after the outbreak of MERS. If one then looks into the year-on-year unemployment variations for the same time period, those aged less than 50 years showed a more remarkable increase in unemployment after the MERS epidemic. Therefore, it can be assumed that the vulnerable group in urban areas might experience a dramatic increase in unemployment since they suffered the direct impact of the contagious disease. In contrast, the economically active population in rural areas underwent a remarkable surge in unemployment due to the slowdown in domestic demand caused by a decrease in leisure activities and consumption by urban dwellers.

Fig. 5

Unemployment rates by age group.

Note: Unemployment rate(%) = (The number of unemployed persons/The economically active population)*100

The study then examined the industries that suffered losses due to the onset of the MERS epidemic. Fig. 6 shows the variations in year-on-year production output by industry classification, as included in the Monthly Industrial Production Statistics issued in June 2015 by Statistics Korea. The figure showed that the ‘Accommodation & Food’, ‘Entertainment & Leisure’ and ‘Information & Communication’ industries sustained the most severe losses. Previous studies had reported that the accommodation & food, and entertainment & leisure industries were hit the hardest by the epidemic (Lee and Warner, 2005, World Bank, 2014). In the case of a disease that can spread through human-to-human contact, it is natural for people to cancel their tourism packages and travel plans. Therefore, rural areas experienced severe indirect losses due to domestic demand slowdown and dwindling consumption caused by behavioral changes among people, even when they otherwise suffered a mild direct impact of the contagious disease when compared to urban areas.

Fig. 6

Index of industry production after MERS break.

Note: The index of all industry production in June 2015 from the same period of the previous year.

Fig. 7 shows the employment ratio in rural and urban areas corresponding to those aged more than 50 years and those aged less than 50 years in various industries affected after the onset of the MERS epidemic and industries that were relatively unaffected. In urban areas, the affected industries showed similar age ratios when compared to those of the relatively unaffected industries. For both the affected and unaffected industries in urban areas, those aged less than 50 years old had a higher employment ratio than those aged more than 50 years. In contrast, those aged less than 50 years in rural areas had a higher employment rate in the affected industries than in the unaffected industries. Therefore, those aged less than 50 years were actively engaged in economic activities in the accommodation and leisure industries.

Fig. 7

Age-specific employment ratios of the affected industries for urban and rural areas (%).

Note: Shock refers to the proportion of employed persons in Accommodation & Food, Entertainment & Leisure, Information & Communication, Wholesale & Retail, Transportation Industry. Non refers to the proportion of employed persons in manufacturing, construction, electric & air, finance & insurance, real estate, human health & social work, education, water supply & waste, scientific & technical, administrative & support, other service, agriculture, mining, public administration industry.

A. Accommodation & Food, Entertainment & Leisure, Information & Communication, Wholesale & Retail, Transportation

B. Manufacturing, construction, electric & air, finance & insurance, real estate, human health & social work, education, water supply & waste, scientific & technical, administrative & support, other service, agriculture, mining, public administration

Fig. 8 shows a comparison of the employment trends for affected industries that suffered severe losses in production output due to the MERS epidemic with those of unaffected industries. The urban areas did not show any significant differences in employment trends between accommodation and leisure and the other industries. The year-on year variations showed a similar trend when compared to the previous trend. In contrast, the rural areas showed that those aged less than 50 years showed a significant drop in employment over the same period of the previous year after the onset of the MERS epidemic, and the absolute number of unemployed persons reached the lowest since 2010. The employment did not show an instant change as unemployment did. Despite that, rural areas can be assumed to contain those aged less than 50 years who were employed in the affected industries, including accommodation and leisure, and suffered a negative impact.

Fig. 8

Employment changes in the affected and unaffected industries.

Note: (The number of employed persons/the number of working age population)*100

Regression results and robustness checks3

Table 2 shows the results of the analysis performed by dividing the unemployment status before and after the MERS epidemic into the entire nation, the urban areas and the rural areas. The unemployment status increased after the occurrence of the MERS epidemic. The probability of unemployment was increased by 0.24%4 across the entire nation, with 0.17% increase in the urban areas, and 0.56% increase in rural areas. The results of the falsification analysis indicated that the probability of unemployment had decreased until 2013 while the probability of employment showed an increasing trend. In 2014, the urban areas displayed a high possibility of unemployment, and as mentioned above, the reason for the increase in unemployment was due to the Sewol ferry disaster on April 16, 2014. When the actual trends were controlled, the probability of unemployment was increased by 0.45% across the nation, by 0.43% in urban areas, and by 0.53% in rural areas, after the onset of MERS. Therefore, the MERS epidemic is assumed to have had a negative impact on the Korean labor market, not only in urban areas which suffered a direct impact from the contagious disease but also in rural areas.5

Table 2

Impact of MERS epidemic on unemployment: trend from pre- to post- MERS.

	Overall		Urban		Rural
	coefficient	S.E.	coefficient	S.E.	coefficient	S.E.
A. Pre: June-July 2014, Post: June-July 2015
MERS	0.034	0.00002	0.023	0.00002	0.123	0.00004
B. Falsification analysis
2011	−0.036	0.00002	−0.039	0.00002	−0.015	0.00005
2012	−0.016	0.00002	−0.015	0.00002	−0.027	0.00005
2013	−0.001	0.00002	−0.009	0.00002	0.060	0.00005
2014	0.048	0.00002	0.057	0.00002	−0.023	0.00004
C. Time trend control
MERS	0.066	0.00002	0.059	0.00002	0.119	0.00004
Time(×100)	−0.003	0.000003	−0.003	0.000004	0.002	0.000010

Note: (a) S.E. indicates the standard error. (b) The year 2011 denoted that pseudo-intervention is coded as occurring in June 2011, with the data running from June 2010 to July 2011, and similarly for the years 2012–2014. (c) Time estimated a linear time trend. The period of analysis spans June 2010 to July 2015, with a total 12 months (Pre: June and July 2010–June and July 2014, Post: June and July 2015). (d) The estimated coefficient in the bold number is statistically significant at 1% level. (e) Pseudo R-square values and the number of observations: Overall A. R2 = 0.061, N = 104,026,234,710. B. (2011) R2 = 0.059, N = 98,196,859,254. (2012) R2 = 0.051, N = 99,349,727,582. (2013) R2 = 0.059, N = 100,551,662,046. (2014) R2 = 0.067, N = 102,154,850,374. C. R2 = 0.059, N = 302,774,756,010. Urban A. R2 = 0.059, N = 86,297,150,528, B. (2011) R2 = 0.058, N = 80,987,873,477. (2012) R2 = 0.049, N = 82,037,196,299. (2013) R2 = 0.057, N = 83,001,295,929. (2014) R2 = 0.065, N = 84,510,253,880. C. R2 = 0.057, N = 250,286,319,934. Rural A. R2 = 0.074, N = 17,729,084,182. B. (2011) R2 = 0.074, N = 17,208,985,777. (2012) R2 = 0.081, N = 17,312,531,283. (2013) R2 = 0.087, N = 17,550,366,117. (2014) R2 = 0.072, N = 17,644,596,494. C. R2 = 0.075, N = 52,488,436,076.

Table 3 shows the results of an analysis of whether the MERS epidemic actually dealt a more severe blow to the labor market in urban areas, where a large number of people were quarantined for fear of contagion, than to the labor market of rural areas. Contrary to the expectations, the probability of unemployment for the rural labor market was increased by 0.70% when compared to that of the urban labor market after the onset of MERS. The robustness check was performed by controlling differential trends to hold the assumption that there might be a common trend in outcomes between the urban and rural areas, and the rural areas exhibited a higher probability of unemployment than the urban areas by 0.37%. As seen in Table 3, the rural labor market suffered greater losses than the urban labor market although the latter experienced a more direct impact of the MERS epidemic than the former.

Table 3

DID effects of MERS epidemic on unemployment of rural areas relative to their urban areas.

	Coefficient	S.E.
A. Pre: June-July 2014, Post: June-July 2015
Rural × post	0.095	0.00005
post	0.020	0.00002
B. Falsification analysis
2011	0.023	0.00005
2012	−0.010	0.00005
2013	0.070	0.00005
2014	−0.080	0.00006
C. Differential time trend control
Rural × post	0.054	0.00005
post	0.059	0.00002
time(×100)	−0.006	0.000003
Rural × time(×100)	0.007	0.00001

Note: (a) S.E. indicates the standard error. (b) The year 2011 denoted that pseudo-intervention is coded as occurring in June 2011, with the data running from June 2010 to July 2011, and similarly for the years 2012–2014. (c) Time estimated a linear time trend, and time × treated is the differential time trend. The period of analysis spans June 2010 to July 2015, with a total 12 months (Pre: June and July 2010 ∼ June and July 2014, Post: June and July 2015). (d) The estimated coefficient in the bold number is statistically significant at 1% level. (e) Pseudo R-square values and the number of observations: Overall A. R2 = 0.064, N = 104,026,234,710. B. (2011) R2 = 0.063, N = 98,196,859,254. (2012) R2 = 0.055, N = 99,349,727,582. (2013) R2 = 0.062, N = 100,551,662,046. (2014) R2 = 0.069, N = 102,154,850,374. C. R2 = 0.062, N = 302,774,756,010.

Table 4 shows the results of an analysis examining whether the negative impact suffered in the rural and urban labor markets appeared with different characteristics. The analysis indicated that in urban areas that were under the direct influence of the MERS epidemic, the probability of unemployment for those aged more than 50 years who were vulnerable to the MERS epidemic increased by 0.60% compared to those aged less than 50 years, ceteris paribus. If we examine the results by controlling for differential time trends, the probability of unemployment for those aged more than 50 years old was increased by 0.72% compared to those aged less than 50 years after the MERS outbreak. These results can be assumed that in urban areas, which accounted for an overwhelming majority of the quarantined people, the vulnerable age group suffered direct impact in the labor market. If we look into the labor market in rural areas that experienced a minor direct shock from contagious disease, those aged less than 50 years showed a higher increase in unemployment than those aged more than 50 years. If one examines the past trends in the rural labor market, those aged more than 50 years showed a gradual increase in unemployment. However, after the onset of MERS, the probability of unemployment among those aged less than 50 years was higher by 0.18% than that among those aged more than 50 years. Table 3 shows that considering that the rural labor market suffered greater losses than the urban labor market, all age groups in the rural areas might suffer an impact from the MERS epidemic, and that among them, those aged less than 50 years sustained a more severe blow. The results of further analysis after extracting only non-farming households from the entire subject population with differential time trends controlled show that those aged less than 50 years had a higher possibility of unemployment by 0.3% than those aged more than 50 years. As seen in Fig. 6, if we examine the age groups of employees in the affected industries, including accommodation, food and leisure, those aged less than 50 years in rural areas showed a higher employment ratio than those aged more than 50 years. Therefore, it is assumed that the economically active age group in the rural labor market suffered more due to the indirect effects caused by the decrease in people’s consumption and leisure activities out of fear of contagion.

Table 4

DID effects of MERS epidemic on unemployment of age groups for rural and urban areas.

	Urban		Rural
	coefficient	S.E.	coefficient	S.E.
A. Pre: June-July 2014, Post: June-July 2015
Over50s × post	0.077	0.00004	−0.034	0.00009
Post	0.0003	0.00002	0.128	0.00005
B. Falsification analysis
2011	0.030	0.00004	0.008	0.00011
2012	0.019	0.00004	−0.006	0.00011
2013	−0.126	0.00004	0.024	0.00011
2014	0.015	0.00004	0.066	0.00010
C. Differential time trends control
Over50s × post	0.094	0.00004	−0.045	0.00010
Post	0.036	0.00002	0.128	0.00005
time(×100)	−0.0002	0.000004	−0.006	0.00001
Over50s ×  time(×100)	−0.022	0.000009	0.023	0.00002
Non-Farm
D. Pre: June-July 2014, Post: June-July 2015
Over50s×post	0.073	0.00004	−0.033	0.00010
Post	0.001	0.00002	0.097	0.00005
E. Differential time trends control
Over50s × post	0.090	0.00004	−0.054	0.00010
Post	0.035	0.00002	0.103	0.00005
time(×100)	−0.0004	0.000004	−0.005	0.00001
Over50s × time(×100)	−0.022	0.000009	0.026	0.00002

Note: (a) S.E. indicates the standard error. (b) The year 2011 denoted that pseudo-intervention is coded as occurring in June 2011, with the data running from June 2010 to July 2011, and similarly for the years 2012–-2014. (c) Time estimated a linear time trend, and time × treated is the differential time trend. The period of analysis spans June 2010 to July 2015, with a total 12 months (Pre: June and July 2010–June and July 2014, Post: June and July 2015). (d) The estimated coefficient in the bold number is statistically significant at 1% level. (e) Pseudo R-square values and the number of observations: Urban A. R2 = 0.057, N = 86,297,150,528, B. (2011) R2 = 0.056, N = 80,987,873,477. (2012) R2 = 0.047, N = 82,037,196,299. (2013) R2 = 0.055, N = 83,001,295,929. (2014) R2 = 0.063, N = 84,510,253,880. C. R2 = 0.055, N = 250,286,319,934. D. R2 = 0.056, N = 84,495,405,138. E. R2 = 0.055, N = 244,980,442,574. Rural A. R2 = 0.068, N = 17,729,084,182, B. (2011) R2 = 0.063, N = 17,208,985,777. (2012) R2 = 0.074, N = 17,312,531,283. (2013) R2 = 0.084, N = 17,550,366,117. (2014) R2 = 0.073, N = 17,644,596,494. C. R2 = 0.069, N = 52,488,436,076. D. R2 = 0.056, N = 13,267,203,705. E. R2 = 0.054, N = 37,850,141,770.

We further analyzed samples excluding workers of age 60 and over. In Korea, 60 is the legally mandated retirement age, and thus the hiked unemployment rate might not be caused by MERS vulnerability, but possibly the substantial number of compulsory retirement persons. For this reason, workers, who aged over 60, were excluded from the analysis and the result is shown in Table 5 . The results of the analysis show that there is no significant difference with Table 4. In addition, when we consider the different time trends from June, July 2010 to Jun, July 2014, the unemployment effect of rural area becomes relatively mild. It is presumed that these results are drawn by higher proportion of persons, who aged 60 and over, in rural than urban area, as the rural area is relatively more aged. Nevertheless, the direction of our results is consistent.

Table 5

Robustness check: Excluding workers over the compulsory retirement age.

	Urban		Rural
	coefficient	S.E.	coefficient	S.E.
A. Pre: June-July 2014, Post: June-July 2015
Over50s × post	0.070	0.00004	−0.036	0.00011
Post	0.0007	0.00002	0.129	0.00005
B. Differential time trends control
Over50s × post	0.076	0.00004	−0.016	0.00012
Post	0.036	0.00002	0.128	0.00005
time(×100)	−0.0001	0.000004	−0.006	0.00001
Over50s × time(×100)	−0.005	0.000027	0.013	0.00003
Non-Farm
C. Pre: June-July 2014, Post: June-July 2015
Over50s × post	0.068	0.00004	−0.026	0.00012
Post	0.0013	0.00002	0.097	0.00005
D. Differential time trends control
Over50s × post	0.071	0.00004	−0.016	0.00013
Post	0.035	0.00002	0.104	0.00006
time(×100)	−0.0004	0.000004	−0.005	0.000013
Over50s × time( × 100)	−0.0036	0.000010	0.0014	0.000030

Note: (a) S.E. indicates the standard error. (b) Time estimated a linear time trend, and time × treated is the differential time trend. The period of analysis spans June 2010 to July 2015, with a total 12 months (Pre: June and July 2010 ∼ June and July 2014, Post: June and July 2015). (c) The estimated coefficient in the bold number is statistically significant at 1% level. (d) Pseudo R-square values and the number of observations: Urban A. R2 = 0.060, N = 78,248,800,556. B. R2 = 0.059, N = 229,198,124,643. C. R2 = 0.060, N = 76,961,386,133. D. R2 = 0.059, N = 225,411,043,333. Rural A. R2 = 0.067, N = 14,682,640, B. R2 = 0.063, N = 43,409,400,304. C. R2 = 0.060, N = 11,905,355,972. D. R2 = 0.057, N = 34,144,964,873.

Moreover, we further analyzed the temporary quit for voluntary reasons and the temporarily laid-off from June, July 2014 to June, July 2015.6 The “temporary quit for voluntary reasons” is defined as “personal affairs (temporary illness, accident, and annual leave)”, and the “temporarily laid-off” is defined as “slump, sluggish or shut down workplaces”. In accordance with the definitions, the analysis result indicates that for the entire labor market, the temporarily laid-off are hiked after the MERS incident. In rural areas, after the epidemic, the probability of the temporarily laid-off increased in both the group aged between 15 and 49 and the group over 50s. In urban area, after MERS outbreak, the group aged between 15 and 49 shows a higher probability of temporarily laid-off while the possibility for temporary quit by voluntary reasons was higher for the group of over 50s. We could argue that in the urban labor market, the population over 50 s tends to quit their jobs temporarily more to avoid the MERS infection (Table 6 ).

Table 6

Temporarily laid-off and temporary quit by voluntary reasons.

	Temporarily laid-off (ref: temporary quit by voluntarily)
	Overall		Urban		Rural
MERS	0.117 (0.00010)		0.108 (0.00011)		0.127 (0.00023)
	15–49	Over50s	15–49	Over50s	15–49	Over50s
MERS	0.312 (0.00014)	0.033 (0.00014)	0.352 (0.00015)	−0.016 (0.00016)	0.170 (0.00034)	0.179 (0.00033)

Note: (a) Pre: June and July 2014, Post period: June and July 2015. (b) The estimated coefficient in the bold number is statistically significant at 1% level. (c) Standard errors are reported in parentheses. (d) The reference group is the temporary quit by voluntary reasons. The temporal quit for voluntary reasons is defined as personal affairs (temporary illness, accident, annual leave), and the temporarily laid-off is defined as slump, sluggish or shut down workplaces. (e) Pseudo R-square values and the number of observations, Overall: R2 = 0.121, N = 991,928,442, (15–49) R2 = 0.128, N = 593,210,237, (Over50s) R2 = 0.034, N = 381,525,408. Urban: R2 = 0.130, N = 830,075,642, (15–49) R2 = 0.141, N = 510,319,616, (Over50s) R2 = 0.043, N = 306,584,274. Rural: R2 = 0.101, N = 161,852,800, (15–49) R2 = 0.091, N = 79,441,185, (Over50s) R2 = 0.094, N = 73,618,319.

According to the analysis results, despite its low morbidity and mortality rates, the MERS epidemic had a negative effect on the labor market. The fear of contagion can be said to have become amplified when many people were put in quarantine, which exacerbated the employment condition in the labor market. As Fenichel (2013) pointed out, social distancing can result in the deterioration of the welfare of people who do not belong to the target group when a quarantine policy is ‘overdone’. As Eichelberger (2007) argued in his study, the contagion of fear that becomes widespread through rumors and the mass media could have a significant impact on the economy. Particularly for a country with a small territory, like the Republic of Korea, the mere fear of contagion can cause a great jolt to the labor market.

In urban areas that suffered a more direct impact of contagious disease, the direct effect from the disease dominated its indirect effect (S  >  S ) because the group vulnerable to MERS endured a direct impact to their employment conditions. In contrast, although rural areas experienced a less direct impact of the infectious disease, the rural labor market actually suffered more severe losses than the urban market did. These results confirm the hypothesis that the indirect effect caused by economic losses due to behavioral changes among urban people for fear of contagion might dominate the direct effect of the disease (S  <  S ). This argument can thus be corroborated by the fact that the economically active age group, that is, those aged less than 50 years, suffered a relatively more severe blow in the rural labor market. These results are similar to the argument posed by the World Bank (2014) that a contagious disease can have an impact on the economy through two channels: a direct effect of the disease and an indirect effect of people’s behavioral changes.

Discussion

When a contagious disease breaks out, it has: (a) a direct effect of reducing employment when individuals belonging to a vulnerable group with a relatively higher morbidity and mortality rate avoid employment activities out of fear of contagion, and (b) an indirect effect of reducing employment due to a slowdown in domestic consumption caused by behavioral changes such as a decrease in consumption and leisure activities. Even if the outbreak of the infectious disease is heavily concentrated in a certain area (e.g., urban areas) the decision-making of people in the affected region can have an unintended impact on other areas (e.g., rural area). Such externalities can have an influence on the labor market, especially for the leisure and accommodation industries in rural areas that are highly likely to be maintained by non-local (urban area) consumption and spending. In other words, urban and rural areas are closely linked, and they inevitably influence each other, but the scenarios that unfold in urban and rural areas may appear in different ways. To this end, this study analyzed and compared the employment changes in the urban areas that were directly hit by the epidemic with the ones in rural areas that remained relatively unaffected by the disease itself.

The results of the analysis indicate that although the MERS has a low morbidity and mortality rate, a high number of people in quarantine fed the fear of contagion, which in turn had a negative shock on the labor market in Korea. Although the urban areas accounted for about 90% of the quarantined people, the rural areas (which took up only 10% of the quarantined) suffered a more negative effect. In urban areas, the vulnerable age group of more than 50 years old showed a higher possibility of unemployment than the non-vulnerable age group of less than 50 years old. In contrast, rural areas showed the opposite results. In the case of the accommodation, food and leisure industries, which were severely damaged by the MERS epidemic, the rural areas showed that those aged less than 50 years old, an economically active group, experienced a more severe blow than those aged more than 50. In fact, the share of the number of employed persons (of those aged less than 50 years who engaged in the rural accommodation and leisure industries) in the number of working age population reached the lowest point since 2010, and the year-on-year variations over the same period also showed the steepest drop.

Therefore, the direct effect can be considered to have been dominant in the urban areas since the vulnerable age group in the urban areas experienced a worsening in their employment conditions (S  >  S ). The rural areas actually sustained a more severe blow in the labor market than the urban areas, and it can be assumed that the indirect effect became dominant because economic losses caused by people’s behavioral changes had a more negative influence on rural employment (S  <  S ). This argument can be further corroborated by the fact that those aged less than 50 years who were economically active in the accommodation, food and leisure industries in rural areas showed the lowest level of employment after the onset of the MERS epidemic.

A shock occurring in a certain area can spill over into other regions through externalities when decisions are made by the people in the affected region that have an influence on other regions. In other words, the rural and urban areas are closely linked to each other, and they inevitably influence each other, but the way one influences the other can appear through different mechanisms.

Due to limitations in the available data, this study could not narrow down the sample by selecting the cities that suffered severe losses from contagious disease among urban areas and comparing them with their neighboring rural areas. As such, it can be inferred that the results of the analysis could have been underestimated. If the results were supported with further data, these effects are expected to appear in a more remarkable and clear way.

Conclusions

Disasters, such as contagious diseases, can fall on us at any time in an unexpected way. This empirical analysis provides information that can be used to customize policies to support regions that can be negatively impacted by the spillover of epidemics in order to respond against economic stresses. Thus, we need to formulate methods through which both urban and rural areas can cooperate with each other to remedy stresses caused by disasters, and further efforts should be made to strengthen the positive aspects of the interrelation between rural and urban areas and to reduce the unexpected negative aspects of such interconnection.

Funding

This research did not receive any specific grant from funding agencies in the public, commercial, or not-for-profit sectors.

Table A2

Impact of MERS epidemic on unemployment: Considering tourists from China.

	Sightseeing Pass or Short-term Visitors		Sightseeing Pass or Short-term Visitors /Total incoming Chinese
	coefficient	S.E.		coefficient	S.E.
Pre: June-July 2014, Post: June-July 2015
MERS	0.056	0.00002	MERS	0.099	0.00005
Tourists from China(×100)	−0.003	0.000002	Tourists/Incoming Chinese	−0.612	0.0004
Time trend control
MERS	0.038	0.00004	MERS	0.132	0.00012
Tourists from China(×100)	−0.002	0.000001	Tourists/Incoming Chinese	−0.549	0.0004
Time(×100)	0.017	0.00002	Time(×100)	−0.039	0.00007

Note: (a) S.E. indicates the standard error. (b) Time estimated a linear time trend, the period of analysis spans June 2013 to July 2015, with a total 6 months (Pre: June and July 2013–June and July 2014, Post: June and July 2015). (b) The estimated coefficient in the bold number is statistically significant at 1% level. (c) The data (Number of incoming foreigners) is provided by the Statistics Korea from 2013, Korean Statistical Information Service, http://kosis.kr/index/index.jsp.