Prospective Cohort Study of Sociodemographic Factors, Including Occupation and Subsequent Unemployment, Under COVID-19 in Japan

Background: We examined the relationship between sociodemographic factors, including occupation and unemployment, among workers during COVID-19 in Japan. Methods: We conducted a prospective cohort study using a self-administrated questionnaire. We surveyed the socioeconomic status, personal characteristics, and occupation of recruited workers at baseline (December 22–25, 2020); subsequent unemployment was examined at follow-up (February 18-19, 2021). We determined the odds ratio (OR) of unemployment for sociodemographic status and occupation. The multivariate model was adjusted for sex and age. Results: Among the 19,941 participants, 725 (3.6%) had experienced unemployment. Multivariate analysis showed that the OR and 95% condence interval (CI) of unemployment associated with sex were 1.35 (1.14–1.60) for women compared with men. With increasing age, the OR for unemployment was lower (OR, 0.98; 95% CI, 0.97–0.99; P <0.001), adjusted for sex. The OR and 95% CI for the association with marital status were as follows: 1.33 (1.03–1.71) for being married (spouse not working); 2.09 (1.65–2.64) for bereaved or divorced; and 1.29 (1.07–1.56) for unmarried compared with married (spouse working). The respective gures for the association with annual household income were as follows: 4.05 (3.00–5.46) for <2 million yen; 2.12 (1.62–2.78) for 2–4 million yen; and 1.46 (1.11–1.93) for 4–6 million yen, compared with >10 million yen. The gures for the association with education were 1.73 (1.12–2.66) for junior high or high school and 1.83 (1.19–2.83) for vocational school, junior college, or technical school. The association with occupation was 2.01 (1.63–2.48) for temporary or contract employees, 1.35 (1.02–1.78) for self-employed, and 3.02 (1.68–5.42) for agriculture, forestry, or shing, compared with general employees; it was 0.56 (0.40–0.79) for public employees, faculty members, or non-prot organization employees. The association with job type was 1.25 (1.04–1.51) for jobs mainly involving interpersonal communication and 1.85 (1.55–2.21) for mainly manual or physical labor, compared with mainly desk work. Conclusions: COVID-19 appears to have created diculties for vulnerable This suggests the for and support


Background
Unemployment is associated with a substantial risk of poor physical and mental health. It has consistently been shown to be signi cantly associated with increases in chronic heart disease, acute myocardial infarction, poor mental health, mental disorders, substance-related disorders, and suicide [1][2][3]. Unemployment is an important factor with regard to public health: it poses a health risk for individuals; it also leads to family poverty and, ultimately, constitutes a burden on social security.
COVID-19 has spread around the world and continues to exert a profound impact on global economies. Economic activities have changed drastically as a result: people have refrained from travel and outside eating, drinking, entertainment; they are encouraged to stay at home. The resulting unemployment has become a major concern: according to the Organisation for Economic Co-operation and Development (OECD), countries' unemployment rates rose signi cantly around the same time as the outbreak of COVID-19 [4]. In this regard, Japan's unemployment rate appears to have remained relatively lowaround 3%. However, O cial data on the unemployment rate may be underestimated, as it is de ned as "people looking for work. Actually, data for Japan clearly show a marked decline in the number of people in employment and a fall in household income [5,6].
With disasters and crises in the past, the risk of unemployment was found to be particularly high among socially vulnerable groups [7]. The OECD has identi ed young people, women, middle-aged and older individuals, and migrants as vulnerable groups in the labor market [8]. Those workers are also reported to be more likely to leave the labor force owing to unemployment, disability, or economic inactivity [9]. By contrast, civil servants, teachers, and employees of non-pro t organizations in Japan have been found to have lower turnover rates and greater job security [10].
It is assumed that individual socioeconomic status is still signi cantly associated with unemployment during COVID-19. A cross-sectional study conducted in the United States examined adverse outcomes associated with COVID-19 and the country's stay-home policies. It found that African Americans, Hispanics, women, and low-income households were more likely to experience unemployment, food insecurity, mental health problems, poor access to health care, and rent or mortgage delinquency [11]. However, the relationship between unemployment and socioeconomic status among Japanese during COVID-19 is unclear.
The present investigation was a cohort study of the relationship between socioeconomic status and unemployment during COVID-19 among workers in Japan.

Methods
This prospective cohort study about COVID-19 among Japanese workers was conducted under the Collaborative Online Research on the Novel-coronavirus and Work (CORoNaWork) Project. Details of the study protocol are described elsewhere [12]. Brie y, we administered a baseline questionnaire on December 22-25, 2020 and a follow-up questionnaire on February 18-19, 2021, when Japan was in its third pandemic wave.
For the baseline survey, we recruited 33,087 workers throughout Japan from 605,381 randomly selected panelists registered with an Internet survey company.
The inclusion criteria for participants were being currently employed and aged 20-65 years. This study excluded health-care professionals and caregivers. We applied cluster sampling with strati cation by sex, job type, and region. We excluded 6051 invalid responses owing to the following: response time <6 minutes; body weight <30 kg; height <140 cm; inconsistent answers to similar questions; and incorrect answers to questions intended to identify fraudulent responses.
We distributed the follow-up questionnaire to the 27,036 people with valid responses to the baseline questionnaire. In total, 19,941 participants completed both questionnaires (follow-up rate, 73.8%).
This study was approved by the Ethics Committee of the University of Occupational and Environmental Health, Japan (reference nos. R2-079 and R3-006). Informed consent was obtained from all participants.

Baseline characteristics
We retrieved the following data from the baseline survey for inclusion as explanatory variables: age; sex; marital status; socioeconomic status (based on annual household income and education); occupation; and job type. We categorized age into the following ve groups: 20-29; 30-39; 40-49; 50-59; and ≥60 years. Marital status was classi ed into four groups: married (working spouse); married (spouse not working); divorced or widowed; and never married.
Annual household income was classi ed into the following six groups: under 2 million; 2-4 million; 4-6 million; 6-8 million; 8-10 million; and >10 million yen. We categorized education into ve groups: up to junior high school; up to high school; up to junior college or technical school; up to university; and graduate school. We categorized occupation into 10 groups: general employee; manager; executive manager; public employee, faculty member, or non-pro t organization employee; temporary or contract employee; self-employed; small o ce/home o ce; agriculture, forestry, or shing; professional occupation (e.g., lawyer, tax accountant, medical-related work); and other. Job type was classi ed into three categories: mainly desk work; work mainly involving interpersonal communication; and mainly manual or physical labor.

Measurement for unemployment
We ascertained unemployment as follows. First, the baseline survey included only people who were employed at the time of response. In the follow-up survey, in answer to the question "Have you changed your place of work since December 2020?" respondents were asked to select one of the following six options: "no change"; "I was transferred to another company"; "I resigned and got a new job right away"; "I stopped working and was unemployed for a while but am now working"; "I stopped working and started a business (e.g., managing a company, running a sole proprietorship, or engaging in self-employment)"; and "I stopped working and am not currently working (including job seeking)." We de ned unemployment as participants choosing one of the following: "I resigned and got a new job right away"; "I stopped working and was unemployed for a while but am now working"; "I stopped working and started a business (e.g., managing a company, running a sole proprietorship, or engaging in self-employment)"; or "I stopped working and am not currently working (including job seeking)."

Statistical analysis
We determined the odds ratios (ORs) of unemployment for sociodemographic status and occupation using a multilevel logistic model for the prefecture of residence. The multivariate model was adjusted for sex and age. We undertook a trend test by conducting the analysis using age, annual household income, and education as continuous variables. We also used the incidence rate of COVID-19 by prefecture as a prefecture-level variable. The multilevel analysis was performed within this incidence rate of COVID-19. We considered P values under 0.05 statistically signi cant. All analyses were conducted using Stata (Stata Statistical Software release SE16.1; StataCorp LLC, College Station, TX, USA).

Results
The basic characteristics of the respondents appear in Table 1. There were 11,170 men in the sample, accounting for 56% of the total. The mean age was 48.0 years. Table 1 Basic characteristics of the subjects Table 2 shows the associations of sociodemographic status (including occupation) with unemployment: 725 (3.6%) workers had experienced unemployment. Table 2 The association between sociodemographic factors including occupation and unemployment.
Multivariate analysis showed that the OR of unemployment associated with sex was 1.35 (95% con dence interval [CI], 1.14-1.60) for women compared with men. With increasing age, the OR for unemployment was lower: OR, 0.98; 95% CI, 0.97-0.99; P <0.001, adjusted for sex. The respective OR and 95% CI gures for the association with marital status were as follows:

Discussion
Through a cohort study, this investigation examined the association between sociodemographic factors and subsequent unemployment during COVID-19. Under the pandemic of COVID-190, we found that unemployment was associated with sociodemographic factors such as age, sex, marriage, income, education and occupation.
We observed that the risk of unemployment was highest among young people. International Labour Organization (ILO) has reported that worldwide, young people were subjected to the greatest loss of labor opportunities through COVID-19: in 2020, young workers suffered 8.7% job losses compared with 3.7% for adults [13]. The present study conducted in Japan similarly found that young people were more than twice as likely to be unemployed as middle-aged and and contract workers were employed in lifestyle-related industries, travel and entertainment services, which were heavily affected by the pandemic [16,17]. Temporary and contract workers are thought to be used to adjust employment and our results clearly con rm this.
We observed a high unemployment rate in the agriculture, forestry, and sheries sector. The reasons for the high unemployment rate in agriculture are unclear: agriculture is an employment sector that is less susceptible than others to economic uctuations. One possibility is that the economic downturn in the food service and food industry caused by COVID-19 had an impact on agricultural production [16]. Another explanation for the nding in agriculture, forestry, and sheries could that it was re ecting a trend for casual work in Japan: temporary employment during the farming season is common, and over 80% of temporary workers have short-term contracts of under 1 month [18,19]. A further factor could be that the number of farmers in Japan has been declining annually since before COVID-19, and the high unemployment we found in agriculture may not have been due solely to the pandemic [20].
We found that women were more likely to be unemployed than men: the OECD has stated that women are also more vulnerable in society [8]. Japan's Labour Force Survey during COVID-19 observed a gender difference in the decline in the number of people in employment: women were more likely to be unemployed [5]. The 2021 survey reported that 68% of people in informal employment were women; many of them worked in the accommodation, catering, and lifestylerelated service industries [5]. In the present study, we found that 14.7% of women (compared with 7.4% of men) were in informal employment.
The unemployment rate was also signi cantly higher for divorced or bereaved people than with dual-earner households. In economic terms, marriage is held to be a rational behavior that seeks economic gain [21,22]. It has long been pointed out that low-income earners and those with unstable employment are less likely to get married [23,24]. The unmarried participants with their precarious employment situation may became unemployed owing to the pandemic. Also, if the divorced or bereaved had a child, it is possible that they were forced to leave the workplace due to school or after-school care leave, because school closures by the pandemic [25].
With regard to income and education, we observed that the lower the income and lower the education, the greater was the likelihood of unemployment. It is widely known that such socially vulnerable groups are at higher risk of unemployment [7,9,26]; we found a similar trend during COVID-19. The impact of unemployment on the lives of those with lower incomes is accordingly greater, and they constitute the group in highest need of social support.
Overall, COVID-19 appears to have increased di culties for a previously vulnerable group. Previous studies have shown that people of lower socio-economic status are more likely to face di culties in the event of a pandemic. However, much of the research has focused on the higher risk of contracting infectious diseases and, as a result, being more likely to face problems such as healthcare costs and unemployment [7,[27][28][29][30]. Our ndings suggest that in the event of a major epidemic, resulting in unemployment among vulnerable segments of the labour market, regardless of whether workers themselves are infected. Thus, there is a need for employment and nancial support for socially vulnerable groups in the event of a major epidemic.
There are several limitations of this study. First, it was unclear why the participants had experienced unemployment: we did not know whether it was due to the effects of COVID-19, company bankruptcy or nancial di culties, or the participants' voluntary decision to change jobs. Second, this study was conducted as an Internet survey; thus, the generalizability of our results is unclear. It is possible that individuals who were genuinely penurious did not have Internet access and could not participate in the survey. If such people had taken part in the survey, the bias would have been stronger. We attempted to reduce subject bias as much as possible by sampling by region and occupation based on infection rates. Third, of the 27,036 individuals who participated in the baseline survey, 7095 did not respond to the follow-up survey (non-participation rate, 26%). That may have led to further bias.

Conclusion
We con rmed the relationship between sociodemographic factors and subsequent unemployment in the case of COVID-19. There is a need for widespread, sustained support for socially vulnerable groups in the form of both short-term and long-term vocational training and health care.