3.1. Impact of COVID-19 on Viet Nam’s labour market

Viet Nam experienced the COVID-19 pandemic in four main waves. Following the registration of a small number of cases in early 2020, the Government decided to close all schools in February and the international borders in March. Domestic travel was also restricted between provinces. Efforts to contain COVID-19 in 2020 were mostly successful. The Government announced a stricter quarantine policy during the 2021 Lunar New Year, but a large outbreak in April was unavoidable, with 1.2 million infections recorded. The country aggressively pursued a zero-COVID strategy, using contact tracing, mass testing, quarantining and lockdowns. Support measures and policy packages were mostly implemented in 2021.²

The economy was mildly affected in the first quarter of 2020 (World Bank 2020) owing to early border restrictions and containment measures. The labour market rebounded very quickly, almost reaching its pre-pandemic level of employment in late 2020, and the country was praised for its management of the dual crisis (ILO 2020c and 2020d). However, the harder-hitting 2021 waves caused more damage, affecting over 9 million workers (in a 50-million-strong labour force) and eliminating the previous rebound momentum. Among those adversely affected, 540,000 people lost their jobs, and 2.8 million were furloughed or had to suspend production or business operations. In addition, 3.1 million workers reported reducing their work hours or moving to alternate shifts, and 6.5 million reported a loss of earnings (GSO 2021).

In Viet Nam, informal work is determined by workers’ sector and whether their employers have registered their business and are paying taxes. In addition, regulations require formal workers to have a labour contract and social security (Nguyen et al. 2021). These two conditions result in various types of informal employment, ranging from own-account workers to the self-employed and even wage employees working on a non-contractual basis. They are distributed over a wide range of occupations, such as small-scale farmers, vendors, small-scale traders and domestic workers.

A comparison of formal and informal employment shows that the dynamic responses to the different pandemic waves varied drastically. In 2021, over 33.5 million workers were employed in the informal sector, equivalent to 68.5 per cent of total employment (GSO 2022). The household sector is often separated from the informal sector in Viet Nam’s statistics, even though household work is considered informal employment according to the definition established by the ILO and is not covered by labour protection.³ A study by the ILO (2022) provides an overview of the changes in employment that affected formal and informal workers in Viet Nam. Compared to 2019, informal jobs were immediately and adversely hit by the initial reports of the COVID-19 virus in early 2020, while the number of formal jobs was still increasing at the yearly rate of 2.7 per cent. Surprisingly, during the long period of social distancing, while formal employment was undergoing considerable fluctuations, informal jobs maintained steady levels in four consecutive quarters until the second quarter of 2021 (figure 1). This suggests that the informal sector was resilient in adapting to new circumstances and played a buffer role in coping with the crisis. However, later in 2021, when subsequent waves of the pandemic reached Viet Nam and led to lockdowns and more restrictive measures, informal jobs declined substantially and more so than formal jobs (6.7 per cent versus 4.1 per cent in the third quarter of 2021). In addition, there was a big difference in informal employment between the rural and urban areas. From the third quarter of 2020 to the second quarter of 2021, there was some job growth in informal work in urban areas (3 to 5 per cent), while 2 to 3 per cent of informal rural jobs were lost (figure 2), suggesting more resilience and flexibility in urban jobs.

Figure 1

Change in employment compared with 2019 (percentages)

Source: ILO (2022).

Figure 2

Change in the number of jobs by area compared with 2019 (percentages)

Source: ILO (2022).

Nonetheless, a growing body of evidence indicates that Viet Nam faced considerable disruptions due to the pandemic, particularly in terms of labour market outcomes and adverse effects on poverty (Nguyen et al. 2021). With more data available now, this article extends that literature by delving into the gendered impacts of COVID-19 on rural employment, especially among informal workers.

3.2. Women and informal employment in rural Viet Nam

Women’s share of the labour force in Viet Nam had remained stable since 2000, at around 45 to 47 per cent. Although the decline in male and female LFP in 2020 was similar, the recovery in employment was stronger among women than among men (5.4 per cent versus 0.7 per cent) (GSO 2021). The urban labour force also experienced a more robust recovery in 2021, with 5.5 per cent growth, compared with 1.4 per cent in rural areas. Initially, however, urban workers were more severely affected than rural workers (15.6 per cent compared with 10.4 per cent), reflecting the fact that agriculture was the sector least affected by the pandemic (only 7.5 per cent of workers). The services sector was the hardest hit (20.4 per cent of workers affected), followed by the manufacturing and construction sectors (16.5 per cent of workers affected) (GSO 2021).

As regards informal employment, figure 3 shows that male and female workers experienced different degrees of upheaval in different waves of the pandemic (ILO 2022). In 2019, some 14.4 million women were employed in rural areas of Viet Nam. Women’s informal employment experienced a sharp decline in the first two quarters of 2020, down by 4.7 and 6.1 per cent, respectively, whereas changes in male informal employment were more moderate. There was a strong recovery starting in the second half of 2020 before another plunge in the third quarter of 2021. This time, female and male workers in the informal sector were similarly affected, losing 6.5 to 6.8 per cent of employment.

Figure 3

Change in the number of jobs by sex compared with 2019 (percentages)

Source: ILO (2022).

By economic sector, women in rural areas were distributed across agriculture, forestry and fisheries (40 per cent), manufacturing (26 to 27 per cent) and wholesale, retail trade and repair services (14 to 15 per cent), followed by smaller sectors and communications, education and other services. It would therefore be interesting to delve deeper into the effects of the pandemic on rural women across these three main sectors.

4.1. Empirical approach

In order to measure the effects of the pandemic on rural employment and identify their differentials between formal and informal workers, who account for 62 per cent of the sample) and between sexes, we employed ordered logit models (McCullagh 1980). The (response) variable is categorized on an ordinal scale of the severity of the employment effect. This is more efficient than multinomial logit models, which ignore the ordinality of the response variable and treat it as nominal, which may fail to use some information of the outcome available, while estimating many more parameters than necessary and increasing the risk of obtaining non-significant results (Williams 2006; Williams and Quiroz 2020). In our case, we consider the severity of the pandemic effects on employment based on subjective self-reported information from survey respondents.

We used two models with different sample sizes. The baseline model covered a larger sample of individuals who had been employed before the pandemic and were either employed, unemployed or inactive at the time of the survey in 2021. The four ordinal outcomes for the dependent variable were “no effects”, “work hours reduced/furlough”, “job lost but employed now” and lastly “job lost and still unemployed”.⁴ Explanatory variables covered only individual characteristics (X′) and not work characteristics. The proportional odds model is identical to the linear logistic model in McCullagh (1980), expressed in equation (1):

$log\left[\frac{{\gamma }_j(X)}{1–{\gamma }_j(X)}\right]={\theta }_j– {\beta }^{\prime }X$       (1)

where 1 ≤ j < 4.

The cumulative probability up to and including j for a covariate vector X is expressed in equation (2):

      (2)

where θ_j is the cutoff point for the jth category.

Second, we restricted the extended model to those who were working at the time of the survey and thus only the first three categories of the response variable (k = 3) were considered. Therefore, more covariates on work characteristics (Z) were added to this model. We also explored the interaction effects between individual and work characteristics to implement intersectionality analyses (e.g. gender with informality and gender with economic sectors). Some interaction terms were represented by X′Z in the model. The functional form was similar to the above, with the condition 1 ≤ j < 3, and the covariate vector X was replaced by a combination of X + Z + X′Z.

4.2. Data

We used data from Viet Nam’s nationally representative LFS for 2021, when the pandemic started to affect the labour market. In order to restrict the scope of the study to rural areas, we limited the sample to residents who had been living in their locality over the previous five years. This was to rule out the external effects of internal migration caused by the pandemic, when urban workers who lost their jobs migrated back to their home provinces to take up agricultural work.

To measure employment effects, we defined four levels of the dependent variable based on retrospective information from respondents about how COVID-19 affected their work, coded as 1 for those with no effects, 2 for workers who experienced work hour reduction, furlough or business suspension, 3 for workers who lost their job but were employed by the time of the survey and 4 for those who lost their job and were still unemployed due to the pandemic. This was confirmed by an extra question in the survey, asking respondents if their work was still affected by the pandemic. We excluded workers for whom the pandemic had a positive impact. We thus limited our initial sample of respondents aged 15 or older in rural Viet Nam to 96,413 employed, unemployed and inactive workers.

Regarding the explanatory variables, the two key dichotomous variables were “female”, coded 1 for women, and “informal”, coded 1 for those in informal employment. The latter included own-account workers, freelancers, unpaid family workers and household or wage employees without compulsory social insurance or whose employers were not registered (GSO and ILO 2016). A summary of individual characteristics is presented in table SA1 in supplementary online Appendix 1. We used technical skills rather than education levels, incorporating vocational and professional training for those who did not have a university degree.⁵ The provincial dummies represent, to some extent, the various COVID-19 situations in 2021 in different parts of Viet Nam. Although we could not access information on the survey months to enrich our analysis, we created quarterly dummies to control for seasonality in employment, based on the four quarterly rounds of the 2021 LFS.

Following the ILO’s request to enhance statistical capacity to measure total working time, both paid and unpaid, Viet Nam added a light time-use module to its existing LFS (ILO 2021a; GSO 2022). The 2020 and 2021 datasets were the first LFS containing additional information on unpaid work at home. We are particularly interested in the amount of time rural workers spent caring for their family in three major categories: (1) domestic work, including cooking, cleaning, laundry and grocery shopping; (2) adult care and support (for those over 18 years old); and (3) childcare (for those under 18 years old). Each is measured in hours per week with a summary in table SA1 in supplementary online Appendix 1.

5.1. Employment effects of the pandemic

This section provides a descriptive analysis of the employment effects of the pandemic on rural workers based on the retrospective responses in the survey. It is followed by econometric modelling in section 5.2. Negative effects varied from job loss to furlough, business suspension, reduction of work hours and loss of earnings. Outcomes also included some positive effects (i.e. income increases) and no changes. Respondents were allowed to choose more than one answer. Table SA2 in supplementary online Appendix 1 indicates the share of workers whose work and income were affected. Urban workers were harder hit, but they also had a greater chance of being able to telework (work from home) than those in rural areas, where agricultural work requires physical presence and there are fewer opportunities for online occupations and platform tasks.

In rural areas, we find some gender gaps for various effects. While women seem to have experienced a lower risk of job loss in the informal and household sectors, they were more vulnerable to furlough in the formal sector and to business suspension in both the formal and informal sectors (table SA3). However, the overall shares of workers affected by the COVID-19 pandemic in rural areas (42 per cent) are relatively similar for men and women (table SA2). We further analysed gendered effects in relation to sectoral informality to tease out the challenges faced by informal workers in rural areas. Table SA3 reveals that the household sector (almost all farming businesses) was least affected in terms of job loss or work hour reduction, since farms continued to operate. However, it was more susceptible to income reduction as production and sales of agricultural produce declined, owing to lower market demand and social distancing measures. Both men and women in the informal sector were vulnerable to all the negative labour market effects of the pandemic. For instance, the share of respondents reporting job loss (3.2 for men and 2.6 per cent for women) was higher than the national unemployment rate of 2.5 per cent. In contrast, those same rates were lower in the formal sector (1.77 and 1.88 per cent, respectively).

We also find that more informal female workers were affected by temporary discharge or business suspension (27.4 per cent), or by a loss of earnings (43.9 per cent) than any other group. In addition, the shares of women switching to online or platform work were low, though slightly higher than for men.

In the following econometric models, we investigate factors contributing to the propensity of employment change. The first column in table 1 presents the baseline ordered logit model in which only the signs rather than the magnitude of coefficients are meaningful. The positive and highly significant coefficient of the “female” dummy indicates that women were more likely to experience stronger (more serious) employment effects during the pandemic than men. A non-linear relationship between age and employment effects is also found in this model, with younger workers tending to suffer from more serious employment effects than older workers up to a certain age, after which the tendency is reversed. When single workers are defined as the reference group, we find that married workers were less likely to face severe employment outcomes, while divorced, separated or widowed individuals were more vulnerable to the worst employment changes. Holding other factors constant, technical skills play a significant role in mitigating the negative employment effects of the pandemic on rural workers. Taking workers without skills as the reference, highly skilled individuals were less likely to be affected by the crisis, especially those with university degrees, followed by those with technical skills from post-secondary non-tertiary education.

The marginal effects of each explanatory variable are shown in the last four columns of table 1 for the four categories of the response variable, respectively. Without controlling for work characteristics, on average, women in rural areas were 0.06 percentage points more likely than men to lose their jobs and remain unemployed during the pandemic. Meanwhile, rural workers with a university degree were 2.8 percentage points more likely to be unaffected in their employment than workers without skills, whereas the likelihood of those graduates being furloughed, dismissed or remaining unemployed decreased by 1.1, 0.4 and 1.3 percentage points, respectively. The findings for rural workers with post-secondary degrees are similar though smaller in magnitude. To further understand those fundamental differences, the extended models consider other work characteristics, economic sectors and the degree of formality of work.

There are also substantial seasonal differences in those employment changes, with more detrimental effects in the second and third quarters compared to the first base quarter, while there were some mitigations in the last quarter of 2021. This reflects the seasonal pattern of high unemployment halfway through the year in Viet Nam, reinforced by widespread COVID-19 infections from April to August 2021.

In the extended model with three levels of the outcome variable, apart from provincial effects, quarterly seasonality and work skills, we also controlled for workers’ economic sector, their employment status and occupations according to ten groups (1-digit codes of the 2008 International Standard Classification of Occupations – ISCO-08). Results differ from the previous baseline model as the intersectionality between sex, informality and economic sector is allowed.

The first column of table 2 shows a significant and negative coefficient for the female dummy, while the estimate of informal employment effects is highly and positively significant. This is also consistent across models in other columns of the table. It suggests that informal workers were more vulnerable to the risk of being furloughed or losing jobs in the overall sample compared to their peers with formal jobs. Similarly, rural men also bear higher risk than women, regardless of being in formal or informal work. The interaction between the two dummies is shown in the second column of table 2 with a highly significant coefficient of –0.253. This confirms the finding that, although informal workers were more likely to be furloughed or unemployed during the pandemic, the incidence was less prevalent among women than it was among men. This is consistent with the descriptive results we previously observed. Table SA.4 in supplementary online Appendix 1 shows the marginal effects of this extended model. For instance, on average, being female with informal employment additively increases the probability of the individual’s work being unaffected by 2.48 percentage points [(0.0271 – 0.0023) × 100]. Meanwhile, women in the informal sector are 0.78 percentage points less likely to lose their jobs than their male counterparts with similar types of work [(–0.0099 + 0.0077) ×100].

However, the vulnerabilities of rural women are more nuanced in terms of economic sector and family situation. The last two models in table 2 show that being a female household head or being involved in non-farm activities in the industry or services sectors tends to worsen employment outcomes. Those multiplicative effects are all highly significant and positive, as are their additive effects. This is an important finding in the case of rural Viet Nam, indicating the multiple vulnerabilities faced by women struggling with their non-farm work while taking care of their family as the household head.

From the sectoral perspective, across all models in table 2, we find that the coefficients for the industry and services sectors are consistently and significantly positive. This suggests that non-farm workers were more likely to lose their jobs or be furloughed than farmers, in line with the macro view that agriculture was less affected. In supplementary online Appendix 1, table SA4 also shows that farmers were less likely to be affected by the pandemic, while workers in the industry and services sectors were more likely to be furloughed (13 to 14 percentage points) or dismissed (3 to 4 percentage points).

Moreover, we also find a gender differential when comparing these sectoral effects. Positive coefficients for the interaction terms “female × industry” and “female × services” and a negative coefficient for the female dummy in the extended model (2) imply, at first glance, that women were less likely to be hit hard by the pandemic. However, this was probably true only in the farming sector. In non-farm sectors, bigger gender gaps favoured men. Rural women were considerably vulnerable to more serious employment effects in non-farm work. Again, this result confirms some of the gendered gaps in rural employment that we found earlier (see table SA2 in supplementary online Appendix 1). While women seem to have experienced a lower risk of job loss both temporarily and persistently during the pandemic, they were more vulnerable to furloughs, business suspension and work hour reduction than men. The marginal effects presented in table SA4 in supplementary online Appendix 1 elaborate on this finding.

Given the findings highlighted above, we explored the drivers of these results by looking at how men and women allocated their time between labour market work and unpaid work at home during the 2021 pandemic and economic crisis.

5.2. Time for care and the intricacy of intertwined crises

As mentioned in section 4.2, we were particularly interested in the amount of time that workers spent on caring for their family members, in total and in three major categories: (1) domestic work including cooking, cleaning, laundry and grocery shopping; (2) adult care and support for those over 18 years old; and (3) childcare for those under 18 years old.

Overall, urban female workers spent more time caring for family members than their peers in rural areas, at almost 14 hours compared with 12.5 hours. Both domestic work and childcare took up more time in cities, but rural people tended to spend slightly more time caring for adult family members. This might be because more elderly persons co-reside with their adult children under the same roof in rural areas than in cities. Figure SA1 in supplementary online Appendix 1 further shows that a gender disparity in the three types of unpaid work favours men and burdens women similarly across rural and urban areas. The largest gender gaps were found in rural domestic work (almost six hours) and rural childcare (about three hours), implying a higher level of gender disparity in unpaid care work in rural Viet Nam.

Ideally, empirical studies measuring the impacts of unpaid work on employment and earnings should pay attention to endogeneity problems where employment can have reverse effects on caregiving. For instance, when workers lost their jobs or were obliged to work fewer hours in their paid jobs, they would stay at home and do more chores or care work. The solution for endogeneity concerns is to find instrumental variables that influence employment only through unpaid work variables. Various studies have used information on household demographics or community care facilities as valid instruments (Nivakoski and Mascherini, 2021; Sinha et al. 2024). Our dataset, unfortunately, does not contain or is not adequate to provide such useful information. For instance, the household, commune or district identification variables and survey months are not published by the General Statistics Office. As a result, we only attempt to present a correlation and descriptive analysis rather than draw causal effects of unpaid work on paid employment or vice versa. We aim to illustrate the bidirectional influence of the double hardship that rural workers were facing.

Figure 4 does not indicate much variation across economic sectors in total care time but only a consistent gender gap, as previously described. However, we do find a clearer heterogeneity across workers in agriculture, industry and services with regard to adult care. Moreover, from the perspective of (in)formality, figure 5 shows that the gender gap in total care work is wider among formal workers than among informal workers. We find that domestic work absorbed the same amount of time for male and female workers regardless of their formality status. In terms of childcare, although we find large gender gaps in both the informal and formal sectors, they are wider in the latter (over 2.5 hours). We observe a very wide gap in total care work among women in formal employment and those in informal employment (over six hours).

Figure 4

Time spent on domestic work, adult care and childcare by sex and economic sector in rural areas, 2021 (hours per week)

Source: Our own calculations based on the 2021 Viet Nam LFS.

Figure 5

Time spent on domestic work, adult care and childcare by sex and formality status in rural areas, 2021 (hours per week)

Source: Our own calculations based on the 2021 Viet Nam LFS.

To further explore the association between the employment effects of the pandemic and the time spent on unpaid care work, we examined the distribution of the total hours of unpaid care work in four categories of workers who experienced different employment effects. The boxplot in figure 6 shows a comparative analysis between rural men and women by marital status for the four employment effects. We observe for each effect that women spend more hours per week than men in the median, the first and third quartiles, and most clearly in the fourth. The largest gender gap is found among married workers, where the median and mean hours that women spent on unpaid care work are double those for men. For instance, married female workers who experienced furlough or work hour reductions spent a mean of 19.6 hours per week (median of 16 hours) on unpaid care work compared with 10.1 hours (median of 8 hours) for married men experiencing the same employment effects.

Figure 6

Distribution of total unpaid care work per week among rural men and women by employment effect and marital status, 2021 (hours)

Note: Excludes outside values.

Source: Our own calculations based on the 2021 Viet Nam LFS.

Moreover, those gaps increase according to the adversity of the pandemic employment effects. In the case of married workers who lost their jobs but were later employed, women spent a mean of 20 hours on unpaid work (median of 18 hours), while men spent a mean of 10 hours on unpaid work (median of 7 hours). The biggest gap is then found among the married workers who were still unemployed at the time of the survey, with women working a mean of 21.3 hours (median of 18 hours) and men working a mean of 10.9 hours (median of 7 hours). It is also intriguing to find that, among married workers whose employment was unaffected, there is nevertheless an apparent average gender gap, with women spending a mean of 16.9 hours on unpaid work compared to and 7.1 hours for men.

Family responsibilities in all three types of care work played an important role among married workers, especially when compared with the distributions for single workers. While gender gaps are visible in all measures, similar distributions of care work are noticeable among single males, regardless of their COVID-19 employment effects. Median averages are about 5 hours (mean of 5 to 6 hours), which is equivalent to less than 1 hour per day. In contrast, it is highest among married women who were persistently unemployed. This confirms the interlinkages between adverse effects in paid employment and unpaid work during the pandemic, where gender gaps were amplified as the work employment effects became more severe, especially among married workers.

In figure 7, we distinguish the connections between employment and care work for formal and informal workers. Although the cluster for women shows a higher distribution of caregiving hours in the first and third quartiles and in the medians than the cluster for men, the difference in formality status is not gender-neutral. We find no significant disparity between formal and informal male workers who experienced different employment effects, with means of around 7 hours and medians of 8 to 9.3 hours per week spent on unpaid care work. Meanwhile, there is a larger gap among female workers, especially those who lost their job as a result of the pandemic but had found work by the time of the survey. On average, women spent 17 hours on care work per week if they were in the informal sector, but only 14 hours if they had a formal job. The medians show a similar gap of 3 hours (19.6 versus 16.5). This result implies an additional burden of unpaid work for women in the informal sector, besides the existing gender gaps.

Figure 7

Distribution of total unpaid care work per week by employment effect, sex and formality status, 2021 (hours)

Note: Excludes outside values.

Source: Our own calculations based on the 2021 Viet Nam LFS.