3.1. Definition of green, brown and neutral jobs

There is no universally accepted definition of green jobs, although the literature consistently highlights the importance of skills, particularly in the context of labour market transitions (for more details, see Duman and Ananian 2024). One common approach defines the “greenness” of occupations based on their tasks, drawing on the O*NET database developed by the US Department of Labor. Tasks are classified according to their contribution to environmental sustainability and emissions reduction. Dierdorff et al. (2009) categorize green jobs into three groups: “green new and emerging” jobs, “green enhanced skills” jobs and “green increased demand” jobs. Building on this, Vona et al. (2018) developed a continuous measure of occupational greenness based on the share of green tasks within each occupation at the eight-digit Standard Occupational Classification (SOC) level.⁶ Occupations with no green tasks score zero. These scores are then averaged at the six-digit level to align with available employment data. Brown occupations, by contrast, are defined by their over-representation in pollution-intensive industries. An occupation is classified as brown if it is at least seven times more likely than average to appear in such industries (Vona et al. 2018). Occupations that are neither green nor brown are considered neutral.

This methodology has been adopted by several researchers, who have mapped the occupations to ISCO-08 (Scholl, Turban and Gal 2023; Bluedorn et al. 2022; Elliott et al. 2021; Valero et al. 2021). The green task intensity and brown scores of occupations at the eight-digit SOC level are translated into ISCO-08 with different weighting strategies and at various levels of aggregation. For some occupations, owing to the crosswalk with employment weights, both green and brown intensity scores can be positive under ISCO-08, while some occupations contain neither green tasks nor brown scores. Based on their occupation, each worker is assigned a green and brown intensity score. This allows various labour market and economy-wide outcomes to be analysed, such as the share of green jobs over time, innovation, the role of green and brown intensity in transitions between different types of employment status, and the relationship between “greenness” of employment and productivity.

The definition of green and brown occupations in this article follows the measure obtained through the SOC-to-ISCO crosswalk outlined in Scholl, Turban and Gal (2023). As mentioned above, several other studies implement a similar approach. However, Scholl Turban and Gal provide the most detailed green and brown scores for occupations at the four-digit ISCO-08 level. Categorizing each occupation as either green, brown or neutral serves our aim of examining the trajectories of workers to and from green occupations and considering whether social protection can have an impact on the transitions to green occupations. We therefore classify each of the 433 occupations at the ISCO-08 four-digit level based on the “greenness” and “brownness” scores in Scholl, Turban and Gal (2023). Occupations with a positive green task intensity and a brown score of zero are labelled green, while those with a positive brown score and a green intensity of zero are categorized as brown. Most remaining occupations involve tasks without significant green or brown characteristics, and these are classified as neutral. In cases where occupations exhibit non-zero intensities for green tasks and brown scores, classification is based on relative magnitude: if one clearly exceeds the other, the occupation is assigned to the corresponding category (green or brown). However, when the green intensity and brown scores are of comparable magnitude, rendering a clear distinction impractical, the occupation is also assigned to the neutral category. This classification results in 83 green, 50 brown and 300 neutral occupations at the four-digit level.

This methodology has the advantage of standardization across countries, as the ISCO classification is widely used. Nonetheless, identifying green occupations by task intensity has a few disadvantages. First, tasks evolve over time, whereas O*NET measures are typically static and do not reflect changes within occupations. Second, the SOC-to-ISCO crosswalk requires aggregating many occupations because SOC provides more detailed occupational groupings. Third, classifications derived from O*NET assume that technology is the same as in the United States. To partially address these issues, the World Bank has recently developed a dictionary of green terms to identify green tasks and occupations through text analysis (Granata and Posadas 2024). These terms are classified into two categories: “green terms”, which are clearly and directly linked to green tasks, and “green potential terms”, referring to tasks that could be considered green depending on the technology or context in which they are performed. This classification leads to 36 green occupations and 108 occupations with green potential. Future research could apply a similar methodology in Viet Nam and compare the findings with those presented in this article. As the methodology proposed by the World Bank is still recent, few studies have undertaken such comparisons for other countries. Nevertheless, Ham, Vázquez and Yánez-Pagán (2025) estimate the proportion of green jobs in India to be 14 per cent using O*NET-based methodologies and 9 per cent using the World Bank approach.

3.2. Panel construction and econometric technique

Our empirical analysis draws on data from the Viet Nam labour force survey (LFS), made available through the ILO Harmonized Microdata collection.⁷ In Viet Nam, the LFS adopted a rotating structure in 2019. Individuals are tracked over two consecutive quarters, exit the rotation for the next two quarters and rejoin for two quarters of the following year. This provides access to the same respondents over six non-consecutive periods, but not all the variables required to identify individuals are consistently available in the LFS data. Accordingly, we are able to construct a panel over two waves, from the first quarter of 2021 to the third quarter of 2022. Table 1 presents the number of respondents and attrition rates for each wave.⁸ Both waves include roughly the same number of individuals, with similar attrition rates. It is worth noting that the observation period spans 15 months, or five quarters. Furthermore, some seasonality may affect our findings, as individuals’ initial labour market state is observed only in specific quarters (Q1 and Q3).

One challenge in estimating transition probabilities from rotating panels is non-random attribution bias. We use two weighting strategies to account for the possibility of such bias. First, we weight the transitions and regressions using sampling weights provided by the National Statistics Office of Viet Nam, which represent the frequency of each observation in the national population. Second, we estimate the attrition probability for each sample group and reweight each observation using inverse probability of treatment. A similar approach has been adopted by other researchers (see Samaniego and Viegelahn 2021; Fallick and Fleischman 2001). We begin by computing the probability of matching the same individual across two periods for each wave, based on age, sex, education, marital status and location. These probabilities are then used to reweight observations through an inverse probability of treatment weighting procedure.⁹ All transition probabilities in the next section are presented using attrition-adjusted weights.

To estimate how social protection coverage influences the likelihood of transitioning to green jobs over the specified period, we use the following logistic regression model:

$\text{logit}\left(\text{P}\left(G_i=1\right)\right)=\alpha +{\beta }_{1}S{P}_i+X_i^{\prime }\beta +{\delta }_t+{\epsilon }_i$ 1

where G_i is a binary variable that equals 1 if individual i transitions from a brown or neutral occupation (depending on the analysed transition) to a green one between t₀ and t₁ (= t₀ + 15 months) and 0 otherwise. α is a constant, SP_i is a binary variable that equals 1 if i is covered by social protection at t₀ and 0 otherwise, and X_i represents a set of control variables observed at t₀: age, sex, marital status, location (rural/urban), aggregate sector of activity,¹⁰ informality of the individual’s unit of production, working time arrangement, as well as the logarithm of monthly earnings. We also include wave fixed effects (δ_t). Similar models are implemented to analyse the effect of social protection on the likelihood of transitioning from a green occupation at t₀ to a brown or neutral occupation at t₁.

The introduction of socio-demographic and location characteristics in the specification allows us to take into account the confounding factors linked to specific groups. Such controls are also used in similar studies on labour market transitions (see, for instance, Bluedorn et al. 2022; Causa et al. 2024). In addition, the inclusion of variables reflecting workers’ quality of employment enables us to disentangle the effect of social protection from other possible favourable working conditions, such as higher wages or employment in the formal sector.

Lastly, as noted earlier, social protection can have different effects on workers with low and high levels of education. We therefore split our sample accordingly and estimate the relationship between social insurance and transitions in and out of green jobs separately for workers with less than tertiary education and those with tertiary education.

4.1. Green jobs and transitions

Before turning to the analysis of transitions to green jobs, table 2 presents the number of green jobs and their share among Vietnamese workers tracked over the study period. Among the 432,759 employed individuals surveyed, 62,676 – nearly 15 per cent – work in occupations classified as green. Additionally, 5.6 per cent of jobs are identified as polluting, while the remaining 79.5 per cent fall into occupations that are neither green nor polluting. While there are no comparable statistics for Viet Nam in the literature, demand for green skills is reportedly growing and the share of job postings requiring a worker to have at least one green skill rose to 22 per cent in 2023. Moreover, green jobs in Viet Nam are increasing across all sectors, extending beyond traditional green sectors, such as renewable energy and environmental protection (Vietnam Investment Review 2023). In contrast, a recent World Bank study using a green task intensity approach estimated that green jobs account for just 3.6 per cent of total employment in Viet Nam (Doan et al. 2023). However, this share rises significantly – to 41 per cent – when occupations with the potential to become green are included. This variation underscores how estimates of green employment in Viet Nam largely depend on the methodology used.

In table 3, we present the key characteristics of workers in green, brown and neutral jobs. The age distribution indicates that prime-age workers account for the largest share in green jobs (83.31 per cent), with a similar share for brown jobs (80.48 per cent). While prime-age workers also dominate neutral jobs, this category has a noticeably higher proportion of older workers, at 16.28 per cent, compared with 10.88 per cent in green and 7.31 per cent in brown jobs. Young people represent only 5.81 per cent of workers in green jobs. Gender composition varies significantly across job types. Green jobs are the most male-dominated, with men making up 77.17 per cent of the workforce. Male workers also form the majority in brown jobs (65.03 per cent), while neutral jobs stand out for having a female majority (53.87 per cent). Educational attainment is generally low across all three groups, with most workers having less than tertiary education. Brown jobs have the highest share of workers without tertiary education (93.02 per cent), followed by neutral jobs (85.87 per cent). The highest levels of education are found in green jobs, with 25.97 per cent of workers having attained tertiary education or higher. Marital status shows less variation: in all job categories, most workers are married. In terms of location, green jobs are more urban based, with 51.11 per cent of workers living in urban areas. Brown and neutral jobs, by contrast, are predominantly rural, with 60.36 and 61.05 per cent of workers residing in rural locations, respectively.

There is no single indicator for job quality, but we use informality, part-time work and earnings as proxies. We adopt the ILO’s definition of the informal sector, based on enterprise characteristics, which does not necessarily reflect workers’ individual status regarding informality (ILO 2023). Under this definition, the informal sector consists of unincorporated production units, emphasizing the nature of the workplace. As a result, there is no systematic alignment between a worker’s access to social protection and the legal status of the enterprise. It is therefore possible for workers in informal enterprises to be covered by protection,¹¹ while workers in formally registered enterprises may lack such coverage. All job types exhibit high levels of informality. Employment in the informal sector is most prevalent in brown jobs (76.6 per cent), followed closely by neutral (75.05 per cent) and green (74 per cent) jobs. Green jobs have the highest share of full-time workers (85.26 per cent), closely followed by brown jobs (83.70 per cent). Neutral jobs have a significantly larger proportion of part-time workers (31.69 per cent), possibly indicating more flexible or less secure employment patterns. Social protection coverage is similarly limited, particularly in neutral jobs, where 73.61 per cent of workers lack coverage. Green jobs fare only slightly better, with 69.51 per cent uncovered. Brown jobs have the highest share of workers with social protection at 48.50 per cent. Earnings also vary across categories. Green jobs offer the highest average income per month, with mean earnings of 8,242,366 Vietnamese dong (approximately US$317). Brown jobs follow at 7,195,409 Vietnamese dong (almost US$277), while neutral jobs have the lowest mean income at 6,490,824 Vietnamese dong (a little less than US$250).

In summary, green jobs are concentrated in urban areas, dominated by men and typically held by prime-age adults working full-time with, on average, higher levels of education. They offer the highest average earnings but remain largely in the informal sector and lack widespread social protection coverage. Brown jobs are slightly more rural and inclusive of younger workers, with marginally better access to social protection. Neutral jobs, on the other hand, feature a greater share of older, part-time and female workers, with lower levels of educational attainment, limited social protection entitlement and the lowest earnings among the three categories.

As can be seen from figure 1, green job holders are relatively more likely to transition on the labour market. However, transitions from green to brown jobs are quite limited, at only 2.4 per cent. By contrast, 28.2 per cent of green job holders move to neutral jobs, while another 4.6 per cent either exit the labour force (4.3 per cent) or become unemployed (0.3 per cent) over the course of five non-consecutive quarters. Nearly 65 per cent of the workers in green occupations remain in the same state. The high rate of transitions to non-green jobs could reflect the large share of neutral occupations in the country (79.5 per cent, see table 2). Brown jobs, however, have higher churn. While 50.5 per cent of brown job holders are still in a brown occupation after 15 months, a notable 37.4 per cent transition to neutral employment, suggesting a shift away from environmentally harmful occupations. Neutral jobs are the most stable category overall, with 86.6 per cent of workers remaining in such jobs over the period.

Figure 1

Distribution of individuals across labour market states at the end of each wave, according to their state at the beginning of each wave (percentages)

Note: Shares are estimated using sampling weights.

Source: Our own calculations based on 2021–22 Viet Nam LFS data.

The transition rate to unemployment is relatively low for workers, regardless of whether they hold a green, brown or neutral job. This reflects Viet Nam’s low unemployment rate, estimated at 2.8 per cent on average during the period from 2021 to 2022.¹² By contrast, unemployed individuals exhibit high mobility: only 9.2 per cent remain unemployed, while most transition into employment or exit the labour force. Individuals outside the labour force tend to remain there, with a retention rate of 73.3 per cent.

Analysis by age group shows that green job retention among young people (aged 15–24) drops to 55.5 per cent, reflecting less stability in early careers. The transition rate from brown to green jobs is also lower for young workers (4.3 per cent) than for prime-age and older workers (7.9 and 9 per cent, respectively). Youth unemployment persistence is relatively high at 16.7 per cent, indicating structural barriers to labour market entry. Moreover, 15 months after withdrawing from the labour force, 82.3 per cent of young people are still outside it, suggesting difficulties in entering the workforce. Older individuals (aged 55 and above) show high stability in neutral jobs (86.2 per cent) but also a high rate of transition from employment to inactivity. Among older unemployed individuals, only 2.5 per cent are still unemployed in the next period, while 22.6 per cent exit the labour market altogether, likely reflecting retirement or health-related barriers.

When comparing men and women, women are less likely to stay in green jobs (53.3 per cent compared with 68.0 per cent of men). This may have various explanations, including the attrition of women in male-dominated occupations, driven, for instance, by the structural and workplace factors that they endure in these jobs (Torre and Jacobs 2021). Women are also more likely to exit the labour force from any employment category, a pattern that reflects gender-specific constraints and persistent challenges in balancing labour market participation. Men in brown jobs have a higher likelihood of remaining in the same sector (53.1 per cent) compared with women (45.6 per cent). However, the transition rates from brown and neutral jobs to green jobs are also higher for men (10.1 and 8.3 per cent, respectively) than for women (2.8 and 3.8 per cent).

The data also reveal important differences according to educational attainment. Individuals with less than tertiary education follow trends similar to the overall rates, particularly in terms of stability in neutral employment. By contrast, those with tertiary education show distinctive patterns. They are more likely to transition from brown to green jobs (17.9 per cent), suggesting that education may facilitate environmentally sustainable career shifts. They are also slightly more likely to remain in green occupations (67.1 per cent of workers with tertiary education still hold a green job after 15 months, compared with 64 per cent of those with lower educational attainment). However, unemployed individuals with high levels of education face greater persistence in unemployment (18.8 per cent) than their less educated counterparts (7.1 per cent), possibly reflecting a mismatch between qualifications and available opportunities or more selective job-seeking behaviour. Tertiary-educated individuals also exhibit a slightly higher labour force entry rate, with lower retention in inactivity compared with those with lower levels of educational attainment (29.2 per cent compared with 26.4 per cent), suggesting greater labour force dynamism among the educated.

4.2. Green transitions and social protection in Viet Nam

The Viet Nam LFS includes a question regarding whether the respondent is covered by social insurance. Given that only individuals who hold a job in each period answer the question, it is not possible to examine the relationship between social protection and movements from non-employment into green or brown jobs. While there are data on individuals who are in green, brown or neutral employment with social protection and who move into unemployment or outside the labour force, the number of observations is too small to draw any inferences. The above descriptions also indicate that there are substantial differences between the green, brown and neutral job categories, especially in terms of work arrangements. Green jobs, in particular, are associated with higher earnings than brown and neutral jobs. Furthermore, brown jobs – such as those in fossil fuel extraction and other high-emission industries – are increasingly associated with declining demand due to tightening environmental regulations. In contrast, green jobs in areas such as renewable energy, sustainable agriculture and waste management are expanding, driven by supportive national policies and international climate investments (Vandeplas et al. 2022).¹³ Overall, in some key working conditions, such a as job security and remuneration, green employment might appear, on average, to be relatively advantageous. In this context, social protection, and particularly the unemployment benefits scheme, could facilitate reallocations to such occupations.

As highlighted previously, social protection can have differential effects depending on the level of educational attainment. Education may play a critical moderating role in shaping how social protection affects employment mobility. Workers with high levels of education, possessing broader skill sets, higher levels of adaptability and greater access to information, are more likely to leverage social protection mechanisms. Even in developed economies with high social protection coverage, a potential link has been found between the underutilization of unemployment insurance among eligible individuals and a lack of awareness of the procedures for using this scheme (Hentzgen et al. 2022). Hence, we disaggregate the relationship between social benefits and movements in and out of green jobs by educational attainment. We conduct this analysis by estimating logistic regressions following the specification given in equation (1).

Table 4 shows the average marginal effects (AMEs) of social protection for workers with and without tertiary education on transitions between employment types in Viet Nam.¹⁴ For individuals with less than tertiary education, the estimated relationship between social protection and transitions from brown to green employment is small and statistically insignificant. By contrast, among tertiary-educated individuals, it is larger and statistically significant. This finding suggests that social protection plays a more substantial role in facilitating transitions into green employment for those with higher levels of education. A similar pattern emerges in the transition from neutral to green employment. Among individuals with lower educational attainment, social protection has a very small, statistically significant adverse effect. For highly educated individuals, the association is small, positive and statistically significant, further supporting the interpretation that the benefits of social protection in promoting access to green employment are amplified by higher education. One possible factor contributing to the difference in outcomes between workers with and without tertiary education may be the heterogeneous nature of the green employment opportunities available to these two groups, which may vary along dimensions such as job quality and career progression. These findings are robust across different specifications.

A particularly salient divergence is observed in transitions from green to brown employment. For tertiary-educated individuals, social protection is negatively and significantly associated with the likelihood of making this transition. In contrast, no such relationship is evident among those with lower levels of education. This suggests that social protection is more effective in helping highly educated workers maintain green employment than it is for other workers. This pattern is repeated with respect to transitions from green to neutral employment.

On the whole, the evidence suggests that social protection mechanisms function differently across educational strata in Viet Nam. Among highly educated individuals, social protection is positively associated with increased mobility into green employment and a corresponding decline in transitions from green jobs to brown or neutral jobs. This pattern suggests that social protection may play a facilitating role in enabling job reallocations linked to the greening of the economy. In contrast, for those with lower levels of education, social protection does not seem to have a meaningful effect on transitions between brown and green occupations. For this population group, transitions from neutral to green jobs appear to be the only ones significantly impacted by social protection, reducing the likelihood of such mobility. However, the effect is small in magnitude (1 percentage point).

Table 5 examines the link between staying in the same type of employment and social protection across educational levels.¹⁵ For people with tertiary education, social insurance appears to have, on average, a positive impact on the probability of maintaining green employment. This relationship becomes statistically insignificant for workers with low levels of education. Thus, for highly educated individuals in Viet Nam, social protection appears to facilitate both transitions into green jobs and staying in those jobs. For workers with less than tertiary education in brown jobs, social protection is positively and significantly associated with the probability of staying in the same category. By contrast, for individuals with tertiary education, the marginal effect is statistically insignificant, suggesting no meaningful relationship between social protection and remaining in a brown job. Among individuals in neutral jobs, social protection shows a positive and statistically significant effect on keeping the same type of employment for both levels of educational attainment. In short, where social protection has a significant impact on employment stability, the effect is consistently positive, implying that access to social protection tends to reduce employment transitions more broadly and might reflect a form of “job lock”.

Taken together, these findings suggest that education plays a pivotal role in both transitions to and stability in green jobs for Vietnamese workers. Individuals with higher levels of education are more likely to enter green jobs and remain in them over time, highlighting the significance of educational attainment in mitigating risks associated with the green transition. Moreover, our results indicate that social protection tends to be more effective in supporting individuals with tertiary education. These individuals may have greater awareness of available benefits, better access to formal employment sectors where such protections are provided, or stronger administrative capacity to navigate the system, thereby enhancing the impact of social insurance on maintaining employment within the green economy. Nevertheless, other factors may also account for the limited impact of social insurance on job reallocations among those with less than tertiary education. For instance, employment with social insurance provisions beyond unemployment benefits may be considered highly valuable by this group, reducing the incentive to search for another job. Specifically, since social insurance in Viet Nam protects against health risks, less-educated workers may be particularly keen to maintain such coverage, especially as they may be more exposed to health issues than other workers. Similar “job-lock” mechanisms have been described by Madrian (1994), among others. These findings underscore the importance of integrating social protection policies with complementary schemes – in particular, education and skills development initiatives – to ensure more equitable access to green employment opportunities across educational groups.