Part 2. Skills for transformation and resilience

Building on Part 1’s broad understanding of lifelong learning, Part 2 focuses on the world of work, identifying which skills matter most for better jobs and adaptability in changing labour markets. Using innovative big data analysis and novel methods, it shows that core and technical skills across cognitive and socio-emotional domains work best in combination, driving higher wages, career progression and adaptability. Skills are a critical lever for resilience, inclusion and opportunity amid major transformations – demographic, environmental and digital. However, to meet these challenges, they must be developed, recognized and valued in ways that reflect labour markets and societal needs.

3.1. Introduction

The first two chapters set out the concept of lifelong learning, presenting a working definition and a conceptual framework for tracking its development across three main dimensions: education, the world of work and broader societal learning. As illustrated in Chapter 2, the gaps in lifelong learning and skills development are particularly significant in the world of work. In parallel, major transformative phenomena – such as technological advancements, the green transition, demographic shifts and changing trade patterns – are reshaping labour markets globally. Beginning with this chapter, the report therefore adopts a more specific focus on the world of work.

These transformations emphasize the pivotal role of skills in determining employment opportunities. Possessing or lacking certain skills affects the resilience of some workers while rendering others vulnerable to deteriorating working conditions or job loss. Consequently, efforts to foster skilling, upskilling and reskilling are important. To do so effectively, it is crucial to gain a deeper understanding of the skills and skill sets needed to secure better employment opportunities. This chapter addresses this topic by exploring the following questions:

• What are the observed patterns of firms’ demand for skills and the supply of skills by workers?

• How are skills interconnected and related to each other?

• What role do these skills and skills bundles play in securing higher-quality jobs?

These insights are crucial for addressing the policy challenges explored in the rest of this report, including: (i) how to make workers more resilient in the face of structural transformations (Chapter 4); (ii) how lifelong learning systems can best allow workers to build the demanded skills sets; and (iii) which role policies play in this regard (Chapter 5).

This  chapter builds on and complements the existing literature. Particularly in the last 20 years, academic research has contributed to a much deeper understanding of how skills are related to employment outcomes, albeit confined to selected high-income countries in Europe and North America. Early literature distinguished between occupations with high and low skills intensity, often proxying this by educational requirements or workers’ experience (for example, Tinbergen 1974). Autor, Levy and Murnane (2003) introduced a new framework, where jobs are constituted of combinations of routine manual, routine cognitive, non-routine manual and non-routine cognitive tasks to be performed by workers through the application of related skills. The framework has advanced the understanding of routine-biased technological change and job polarization (for example, Autor, Katz and Kearney 2006; Goos, Manning and Salomons 2014; Spitz-Oener 2006) and recovery patterns after recessions (Acemoglu and Autor 2011; Jaimovich and Siu 2020).

The recent rise of natural language processing (NLP) methods, which enable computer-based text analysis, makes it possible to use big data to analyse skills. Studies increasingly use NLP methods to analyse vacancy postings and extract skills information by processing job descriptions and identifying keywords or patterns from the data (Ash and Hansen 2023). Atalay et al. (2020), for example, show that the task and skill requirements of jobs have changed drastically within occupations in the United States. Hershbein and Kahn (2018) show that during recessions, firms tend to hire more skilled workers, accelerating workforce changes already initiated by routine-biased technological change. Deming and Kahn (2018) show, also using US data, that demand for cognitive and social skills is heterogeneous even within occupations or industries and that combinations of skills explain differences in wages, illustrating the importance for workers to develop the skills demanded for higher-quality jobs.

Policy-oriented institutions have also expanded their research, accompanying their skills development agendas with publications and data collection. For example, the EU provides data on skills development and shortages (Cedefop et al. 2021), the OECD conducts the Survey of Adult Skills PIAAC (OECD 2019), and the World Economic Forum has published reports on skills demand based on interviews of company executives (WEF 2023) and skills-based hiring strategies (WEF 2024).

Despite these advancements, considerable research gaps remain. Often, analyses focus on one specific sector, occupation or set of skills, either due to data limitations or to answer a specific research question (for example, Debortoli, Müller and vom Brocke 2014; Deming and Noray 2020; Marconi, Vergolini and Borgonovi 2023). Analyses on the role of skills for employment outcomes other than wages are regularly missing. Most notably, studies tend to concentrate on single high-income countries, while cross-country comparisons and knowledge for other countries are scarce. This is largely due to a lack of reliable data and of universally implementable skills frameworks.

To address these limitations, some studies rely on one-time surveys in selected middle-income countries (for example, Lewandowski et al. 2022; Lewandowski, Park and Schotte 2023, using data from OECD’s PIAAC survey and the World Bank’s STEP Skills Measurement Program38), while others use labour force surveys and assess skills by applying the occupation-level information on skills from the US O*NET Program39 (Almeida, Corseuil and Poole 2017; Bhorat et al. 2020; Reijnders and de Vries 2018). The former approach means that trends over time cannot be detected, and the latter relies on the problematic assumption that skills compositions of occupations in the analysed country are identical to those in the United States. Access to granular, longitudinal skills information from low- and middle-income countries is therefore key to understanding skills dynamics worldwide, especially since existing work shows significant deviations in these countries from the skills-related patterns in high-income countries (Atencio-De-Leon, Lee and Macaluso 2025; Hjort and Poulsen 2019; Lewandowski et al. 2022; Lewandowski, Park and Schotte 2023).

A few studies already use online data from emerging economies – for example, Archibong et al. (2022) for Nigeria, Matsuda, Ahmed and Nomura (2019) for Pakistan and Nomura et al. (2017) for India. However, these studies cover only specific segments of the labour market or focus on specific skills in the respective country.40 A challenge is the absence of an internationally applicable taxonomy for skills, as existing frameworks are tailored to the United States and similar contexts (and thus less suited for application in low- and middle-income economies), primarily by neglecting manual skills.

Given these research gaps and challenges, this chapter complements the existing knowledge, supported by original big data analysis for six selected middle- and two high-income countries in regions so far neglected by the literature. It also acknowledges that future research should further advance the understanding of skills dynamics in low-income countries.

This chapter provides a summary of the key findings. Detailed methodological explanations and in-depth analyses are available in the report-related articles and briefs included on the report’s website.41 Section 3.2 presents the methodology and data used. Section 3.3 explores skill demand and supply and their evolution over time. Section 3.4 analyses whether skills are interconnected as skills bundles. Building on these findings, section 3.5 explores the role of skills in promoting job quality.

3.2. Data and approach

This chapter uses online vacancy and applicants’ data from job portals and aggregators. Such data provide granular information on workers’ previous job tasks and the tasks sought in vacancies, enabling the identification of country-specific trends in skills supply and demand. Moreover, online vacancy and applicant data are typically collected in real-time or at high frequencies. This avoids the time gaps associated with traditional data collection, including surveys, which are often conducted at longer intervals (Escudero, Liepmann and Podjanin 2024).

This chapter presents the first attempt to systematically explore these data in the context of emerging economies (see box 3.1 for details on the data). Vacancy data have been obtained from online job portals and aggregators in six middle-income countries: Brazil, Egypt, Jordan, Morocco, the Russian Federation and South Africa. Additional vacancy data have been obtained from two high-income countries, the United Arab Emirates and Uruguay. The two high-income countries are worth examining, as their labour markets are clearly distinct from the cases the existing literature has focused on.42 In general, countries were selected to cover examples from diverse regions and with heterogeneous labour market features that had not been studied from this perspective before, thereby looking outside of the United States and the EU. For Brazil and Uruguay, applicant profile information was available in addition to vacancy data.

While these datasets offer a rich foundation for exploring skills dynamics, they entail several limitations to their representativeness (see Fabo and Kureková 2022, a study conducted for this report to provide further methodological considerations). The analysis focuses on the online labour market, excluding: (i) jobs for which recruitment occurs exclusively through “off-line” methods, like word of mouth or local notices, and (ii) applicants without online profiles. The prevalence of such data is influenced by the size of the informal economy, cultural norms, disparities in internet access and digital literacy. Online data also tend to be more common in high-income countries and urban areas.

Additionally, the online data used originate from privately owned job portals, which typically attract more postings for professional occupations, compared to public employment service portals (Kureková et al. 2016). Nevertheless, analyses for Uruguay show that the data sufficiently represent jobs requiring medium and even low formal qualification levels to allow for meaningful analyses. Furthermore, the methodology assigns a substantial number of manual skills relevant to these occupations (Escudero, Liepmann and Podjanin 2024).47

Another challenge is that applicants tend to use significantly fewer words than those used in job vacancies to describe their periods of employment (so-called “jobs spells”, which contain starting and ending dates for each employment period and an open-text description). Due to the brevity of these texts, some skills used in applicants’ current or previous employment experiences may not be captured, suggesting that applicants might need to better tailor their profiles in the application process (Luu and Escudero, forthcoming).

Nevertheless, the data present a valuable opportunity for analysis, especially in countries where more comprehensive skills data are scarce or unavailable and similar analyses could otherwise not be conducted. To capture and systematically classify skills in online data, and building on the ILO’s Global Framework on Core Skills for Life and Work in the 21st Century (ILO 2021),48 a new skills taxonomy geared towards broad applicability across countries was developed. This taxonomy aggregates skills into three overarching categories: cognitive, socio-emotional and manual, and 15 subcategories (table 3.1). Each skill subcategory is defined using keywords that are mutually exclusive and tailored to online data. The taxonomy includes skills related to tasks that workers perform on the job and skills referring to individuals’ personal attributes. The keywords are grounded in literature from economics (for example, Deming and Kahn 2018; Deming and Noray 2020; Hershbein and Kahn 2018) and psychology (for example, Almlund et al. 2011), along with complementary analyses. For more details on the taxonomy’s conceptual underpinnings, see Escudero, Liepmann and Podjanin (2024).

Skills are hereby understood as the “ability to carry out a manual or mental activity, acquired through learning and practice” (Strietska-Ilina et al. 2011, 176) and “the ability to carry out the tasks and duties of a given job” (ILO, n.d.). To ensure comprehensiveness and measurability, the focus is on technical and non-technical skills that are transferable and thus applicable across occupations (see box 3.2). As jobs evolve due to technological advances and other economic shifts, workers with strong transferable skills are better equipped to navigate occupational transitions, as demonstrated below. Given the chapter’s aim to provide high-level analyses for countries in different regions, the analysed skills are defined at a relatively broad level. Nevertheless, the underlying methodology enables more granular analyses, such as those related to specific skills demand in the growing digital, green and care sectors. Digital skills are already captured by the taxonomy in general computer skills, specific software skills, and machine learning and artificial intelligence (AI). Green skills are added and analysed in Chapter 4. Going forward, similar extensions could be possible, such as care skills.49 Moreover, the methodology allows for granular analyses at the keyword level, as illustrated in a few instances in this chapter and Chapter 4.

The taxonomy is both comprehensive and compact, which results in distinct advantages over other skills categorizations. First, its broad theoretical foundation makes it more comprehensive than existing taxonomies. For example, the 35 aggregated skills categories from O*NET are tailored to the United States and do not capture manual skills in as much depth (Fleisher and Tsacoumis 2012). Second, its compactness makes the taxonomy well-suited for research purposes. Thus, it differs from highly detailed taxonomies, uncategorized lists or complex ontologies. For example, the European Skills, Competences, Qualifications and Occupations ontology contains nearly 14,000 skills (European Commission, n.d.), which are too complex to be directly utilized in research.50

The taxonomy was applied to the online vacancy and applicant data using NLP techniques, which allows systematic access and categorization of unstructured information. Open-text descriptions of vacancies were pre-processed to align with the structured format of the skills taxonomy and its associated keywords. A skill was identified if at least one relevant keyword – or one of its synonyms – from the taxonomy appeared in the vacancy. For applicant data, the uploaded employment histories were similarly prepared and categorized. For more detailed methodological explanations, see Adamczyk et al. (2025).

3.3. Assessing employers’ demand and workers’ supply of skills

The demand for specific skills varies across countries and over time, shaped by factors such as work organization and firm structure, occupational structure, economic conditions, technological advancements and shifts in global trade patterns. Similarly, the supply of skills is influenced by multiple determinants, including the effectiveness of lifelong learning systems, sociocultural norms and demographic factors, like a population’s age distribution and labour force participation rates. This section explores the skills sought by firms and those offered by workers, examining potential differences across countries and over time and providing the basis for subsequent sections.

3.3.1. Skills requirements in employers’ vacancies

Based  on the analysis of the vacancy data, employers place significant emphasis on socio-emotional skills across all countries considered. The relative frequency of skill occurrences in vacancies reveals that socio-emotional skills are highly demanded in all countries, even accounting for 59 per cent of all skills in Morocco, 52 per cent in the Russian Federation and 51 per cent in the United Arab Emirates (see figure 3.1).51 Among the socio-emotional subcategories studied, social and customer service skills are most frequently required. Character skills also rank prominently, appearing in third or fourth place in all countries. This echoes a broader global trend noted by Sostero and Tolan (2022), who highlight the importance of communication and customer service skills in Europe, and the increasing demand for social skills in the United States (Deming and Kahn 2018). It also reflects the prominence of the service sector in the countries studied (for example, Naudé, Szirmai and Haraguchi 2016), where socio-emotional skills are critical.

• Figure 3.1. Skills compositions of vacancies (percentage)

Note: The figure displays the share of a specific skill relative to the total occurrences of all skills across time. Labels below 3 per cent were omitted.

Source: Analysis based on vacancy data from BuscoJobs for Brazil and Uruguay (2010–23), Adzuna for the Russian Federation and South Africa (2016–21), ESCWA for Egypt, Jordan and the United Arab Emirates (2020–24), and Veille+ for Morocco (2022–25).

Other commonly required skills include core cognitive skills, such as planning and strategic thinking (see box 3.3 on how these skills are identified), and general computer skills. The emphasis on general computer proficiency underscores the importance of basic digital literacy, which has become a foundational skill for thriving in increasingly digital societies and economies (ILO 2023a). Overall, the importance of the broader category of cognitive skills ranges from 31 per cent in Morocco to 48 per cent in South Africa. Manual skills are also in demand, accounting for 16 per cent in the Russian Federation, where their share is the highest. In contrast, Egypt has the smallest share of online manual skills demand at 6 per cent (figure 3.1).

Box 3.3. Interpreting skills

To provide a clearer understanding of how the chapter characterizes and interprets each skill, table 3.2 lists the top three words most associated with each skill in South Africa. The table shows that some skills are predominantly defined by single keywords; this is especially true for machine learning and AI skills (“automation”), writing skills (“report” and “write”), and finger dexterity skills (“equipment”). In contrast, other skills – such as core and sophisticated cognitive, character and people management skills – are more broadly identified by multiple keywords.

The results for most skills are similar in Brazil, the Russian Federation and Uruguay. However, skills in South Africa tend to be associated with relatively more keywords, possibly due to the larger English language vocabulary (Erichsen 2019) and different communication styles across countries (Kureková et al. 2016). South African employers predominately use a low-context communication style, where information is explicit and detailed, whereas in Brazil (Piroşcă 2016) and the Russian Federation (Burmann and Semrau 2022), a high-context style, where much is implied, is more common.

• Table 3.2. Top three keywords associated with each skill as a share of the total number of matches for that skill, South Africa (percentage)

Cognitive skills			Socio-emotional skills
Core cognitive skills			Character skills
Plan	19.0		Achieve	11.2
Design	12.7		Attention detail	10.1
Strategy	8.4		Initiative	8.6
Sophisticated cognitive skills			Social skills
Science	22.5		Team	35.8
Research	22.2		Communication	24.2
Evaluate	11.5		Presentation	5.5
Computer (general) skills			People management skills
Computer	33.2		Leadership	20.8
Software	20.2		Coach	11.5
Microsoft	13.5		People management	5.6
Software and technical support			Customer service skills
SQL	26.8		Client	36.8
HTML	16.5		Customer	24.5
JavaScript	16.1		Sales	19.9
Machine learning and AI skills			Manual skills
Automation	63.6		Finger dexterity skills
Apache	12.4		Equipment	46.3
Machine learn	7.9		Pack	15.8
Financial skills			Single	6.6
Accounting	25.8		Hand–foot–eye coordination skills
Finance	20.1		Drive	44.1
Cost	18.2		Transport	19.4
Writing skills			Clean	8.4
Report	51.2		Physical skills
Write	31.7		Walk	39.8
Recommendation	5.6		March	18.8
Project and process management skills			Physically fit	10.6
Project management	30.3
Operation management	6.3
Management process	5.8

▪Q1 (0.0%–10.9%) ▪Q2 (11.0%–18.8%) ▪Q3 (18.9%–25.9%) ▪Q4 (26.0%–64.0%)

Source: Analysis based on Adzuna vacancy data (South Africa, 2016–21).

Another notable difference relates to character skills. The four selected countries use different keywords to characterize these skills, which may reflect varying work cultures (Minkov and Kaasa 2022; Taras, Steel and Kirkman 2011). Employers in the Russian Federation tend to emphasize diligent, self-independent characteristics, while employers in Uruguay mostly require dynamic and proactive traits. In Brazil, there is a focus on organization, while in South Africa, character skills are characterized by a diverse set of attributes, including achievement, competence, initiation, attentiveness, innovation and organization (figure 3.2).

• Figure 3.2. Word clouds for the subcategory character skills

Note: The size of a word reflects how often it appears in the data: the bigger the word, the more frequently it is found. For the Russian Federation, terms were translated to English to express them in Latin script.

Source: Analysis based on vacancy data from BuscoJobs for Brazil and Uruguay (2010–23) and Adzuna for the Russian Federation and South Africa (2016–21). 

The composition of skills demand has remained relatively stable over time across most countries (see figure 3.3). Brazil, Morocco, the Russian Federation and the United Arab Emirates show modest shifts, characterized by a decline in the share of social skills and an increase in manual skills. In contrast, Egypt, Jordan and Uruguay have experienced an increase in the demand for social skills, often accompanied by a slight decline in cognitive or manual skills. These modest changes may seem counterintuitive considering the prevalent discourse on rapid technological changes and their impact on workplace skills (for example, Atalay et al. 2020; Autor et al. 2024; Deming and Noray 2020).

• Figure 3.3. Evolution of the skills compositions of vacancies (percentage)

Note: The figure displays the share of a specific skill relative to the total occurrences of all skills in given years. The two years shown were selected based on data availability and to maximize the timespan. Labels below 3 per cent were omitted.

Source: Analysis based on vacancy data from BuscoJobs for Brazil and Uruguay (2010 and 2023), Adzuna for the Russian Federation and South Africa (2016 and 2021), ESCWA for Egypt, Jordan and the United Arab Emirates (2020 and 2024), and Veille+ for Morocco (2022 and 2025).

There are several possible reasons for this stability. In particular, the time span covered is relatively short: 13 years at most. Major changes related to the introduction of computers had likely been completed, while the full impact of recent developments related to generative AI can be expected to unfold in the coming years. This suggests that skill requirements do not change dramatically in the short or medium term. Additionally, technological change may progress more slowly in middle-income economies due to a sectoral composition less focused on technology-intensive sectors, unreliable access to electricity and the internet, and greater challenges in securing funding for new technologies (Gmyrek, Berg and Bescond 2023).

However, the apparent stability hides more nuanced, gradual transformations, including within skill subcategories. For example, in South Africa, the demand for social skills remained stable between 2016 and 2021, but software-specific skills underwent significant shifts (figure 3.4). Skills like employing Python (a programming language for data analysis and automation) and TypeScript (a version of JavaScript coding that yields fewer mistakes) have gained prominence. In contrast, skills broadly used in web development (for example, HTML, JavaScript and PHP) have lost relative relevance. Similar trends are observed in Brazil and the Russian Federation, as well as the United Kingdom, as noted by Costa et al. (2024).

Therefore, while core competencies and broader skill categories remain consistently vital over the medium term, specific skills – particularly in fast-evolving sectors like ICT – are rapidly changing. As another example, while the overall demand for machine learning and AI skills remains relatively low, its growth is noteworthy. The share of vacancies requiring such skills has seen significant increases in Brazil (from 0.5 to 1 per cent), the Russian Federation (from 0.3 to 0.7 per cent) and South Africa (from 0.5 to 0.8 per cent). This upward trend is expected to continue or accelerate with the advent of generative AI technologies.

• Figure 3.4. Word clouds for software-specific and social skills in South Africa, 2016 and 2021

Note: The size of each word represents the frequency in the data: the larger the word, the more often it appears in online job vacancies.

Source: Analysis based on Adzuna vacancy data for South Africa (2016–21).

Occupations and skill demand

Occupations are an important dimension of the demand for skills. For Brazil, table 3.3 shows that cognitive skills are more demanded at higher levels of the occupational hierarchy, while manual skills are predominantly sought in occupations assigned lower skill levels in the ISCO classification.52 Specifically, core and sophisticated cognitive skills and financial skills are most frequently required for managerial and professional positions, alongside the full set of socio-emotional skills, reflecting a complex skills portfolio. Customer service skills are most required for service and sales workers positions, while plant and machine operators and assemblers are most often required to possess hand–foot–eye coordination skills. Importantly, socio-emotional skills are central across all occupations and are the most frequently demanded skills in almost all of them. This trend is also observed in the other countries considered.

Further insights emerge when examining skills demand within occupations, which are exemplified here with Brazil. The skills mix demanded within some broad occupations – such as managers (ISCO 1), craft and related trade workers (ISCO 7) and elementary occupations (ISCO 9) – tends to remain relatively consistent. However, there is significant variability within other occupations, reflecting a broader range of tasks. For example, among clerical support, vacancies for customer service clerks (ISCO 42) primarily require customer service skills (82 per cent of vacancies) and social skills (73 per cent), with relatively fewer general computer skills (22 per cent). In contrast, vacancies for general and keyboard clerks (ISCO 41) emphasize general computer skills (45 per cent), alongside customer service skills (54 per cent) and social skills (71 per cent).

Moreover, vacancy descriptions for certain occupations often omit essential skills for the role, particularly for elementary occupations that typically require physical and manual skills. For example, in Brazil, the proportion of vacancies explicitly requesting these skills is notably low, ranging from 4 per cent for physical skills to 35 per cent for hand–foot–eye coordination skills (table 3.3). This could be due to differences in vacancy styles across sectors and occupations. Some vacancies may be considered largely self-explanatory regarding their skill requirements, emphasizing the importance of implicit knowledge among jobseekers, which can be specific to the country, sector or occupation (Fabo and Kureková 2022).53 As another example, only 15 per cent of Brazilian vacancies for personal care workers (ISCO 53) mention social skills, likely because the related job titles – such as child care worker or teacher’s assistant – are often considered self-explanatory. The omission of such occupation-specific skills makes accurate assessments of these skills challenging. Additionally, job postings for lower-skilled occupations tend to be shorter, possibly because employers invest less effort into recruiting for such jobs (Pellizzari 2011).

3.3.2. Skills supplied by workers

This section shifts the focus to the supply side, specifically the skills supplied by workers in Brazil and Uruguay, for which data are available in the sources used in this report. As such, the frequency of skills listed in applicants’ job histories is aggregated at the applicant level, capturing all skills acquired throughout their employment history. This approach provides a broad view of the skills available in the labour market during the period of analysis.54 It also provides insights into how applicants report skills on their profiles.

The skill distribution among applicants mirrors that of vacancies. Socio-emotional skills are the most prevalent, closely followed by cognitive skills, while manual skills are less common. Within specific skill subcategories, core cognitive, social and customer service skills account for around 50 per cent of all skills identified among applicants. In contrast, machine learning and AI, project and process management, software-specific skills, and physical skills are observed less frequently (figure 3.5).

• Figure 3.5. Skill compositions of applicants in Brazil and Uruguay (percentage)

Note: The figure shows the share of a specific skill relative to the total number of skill occurrences. All skills mentioned across employment spells are aggregated at the individual-applicant level.

Source: Analysis based on BuscoJobs applicants’ data for Brazil and Uruguay (2010–23).

While Brazil and Uruguay exhibit a similar overall emphasis on cognitive and socio-emotional skills, there are notable differences in the subcategories of these broader skills groups. Uruguayan applicants showcase a higher prevalence of both core and sophisticated cognitive skills, as well as general computer skills. They also show a greater focus on customer service skills, whereas social skills tend to play a more important role in Brazil. The prominence of cognitive skills among Uruguayan applicants aligns with the strong demand reflected in vacancy data and a relatively high share of applicants with job spells in professional occupations. In contrast, the prevalence of social skills in Brazil appears relevant independently of the occupational composition or demand-side requirements.

Workers’  skill sets can be as diverse as workers are. In Brazil and Uruguay, men and women exhibit fairly similar skills profiles, though men generally use more words to describe slightly broader skill sets (see the top panel in figure 3.6). The most notable differences between women and men appear in customer service skills, which are more frequently reported by women (especially in Brazil), and manual skills, which are more often cited by men. In Brazil, men also report more people management skills, while in Uruguay, men highlight digital skills more prominently, particularly among professionals and technicians and associate professionals (ISCO 2 and 3). For manual skills, Brazilian men report more instances of finger dexterity skills, while in Uruguay, men more commonly report hand–foot–eye coordination and physical skills. These findings complement the existing literature on differences in skills between women and men (for example Black and Spitz-Oener 2010; Encinas-Martín and Cherian 2023) and may reflect occupational gender segregation (Cortes and Pan 2018; Hegewisch and Liepmann 2013).

• Figure 3.6. Heterogeneity in applicants’ skill composition in Brazil and Uruguay, by sex, birth cohort and education

Note: The outermost ring of corresponds to 100 per cent of applicants mentioning a skill. Therefore, a marker further away from the centre means the given skill is more frequently reported in applicants’ employment histories. “Higher education” refers to individuals with vocational, technical or university education. “Lower education” includes individuals who do not report having completed any of these education types.

Source: Analysis based on BuscoJobs vacancy data (2010–23).

Age also plays a role in shaping workers’ skill sets. The middle panel in figure 3.6 compares the skills compositions of two cohorts – those born during the 1961–70 and 1981–90 periods – and reveals that older workers in Brazil have slightly broader skill sets, likely reflecting the accumulation of skills over their job histories, as the measure does not account for any potential skill decline over time. In contrast, differences are less pronounced in Uruguay. The most notable distinction is in people management skills, which are more prevalent among older workers, while general computer skills are more common among the younger generation. Other skills show minimal differences by age, such as sophisticated cognitive and character skills, suggesting that these skills remain relatively stable over the life cycle (Breit et al. 2024; Gander et al. 2020).

Lastly, notable differences exist in the distribution of skills by education level. The bottom panel of figure 3.6 shows that workers with higher education (with vocational, technical or university education) in Brazil and Uruguay report a broader range of skills, particularly the different types of cognitive and socio-emotional skills, compared to those with lower levels of education (individuals not reporting any of these types of education). It is possible that some of these differences come about because individuals with lower formal education levels do not describe their activities in as much detail: on average, the descriptions of their employment histories are shorter. Yet, for customer service skills in both countries and manual skills in Uruguay, individuals with lower levels of education are slightly overrepresented. When comparing university graduates with those who have vocational or technical training, the differences in skill sets are relatively small. This suggests that different types of education prepare individuals for distinct careers paths without necessarily resulting in large gaps in overall abilities. However, certain skills, such as software, financial or writing skills, are more frequently reported by individuals with university degrees. It is important to note that a significant share of the population aged 15 and above – estimated at 74.5 per cent in Uruguay – do not complete either of these educational paths (Escudero, Liepmann and Podjanin 2024).

Occupations and skill supply

Importantly, in contrast to the demand for skills, the skills composition reported by jobseekers varies significantly across occupations, as it is closely tied to the type of work performed. As shown in table 3.4, there are notable differences in the concentration of skills between occupations classified by ISCO-08 as requiring higher or lower skill levels. Managers (ISCO 1) and professionals (ISCO 2) report a broader range of skills, while plant and machine operators (ISCO 8) and workers in elementary occupations (ISCO 9) report a more limited set of skills. Clerical support workers and services and sales workers (ISCO 4 and 5) also have relatively concentrated skill sets, though they exhibit a comparatively higher prevalence of complementary cognitive skills.

In general, the skill sets highlighted by applicants across different job spells align with expected core skills for their respective occupations. For example, individuals who have worked as plant and machine operators (ISCO 8) or in elementary occupations (ISCO 9) tend to emphasize manual skills, with little mention of other competencies. Similarly, clerical support workers (ISCO 4) and services and sales workers (ISCO 5) predominantly emphasize customer service skills – the core component of their jobs – while reporting few additional skills. The analysis indicates that skills compositions reported by jobseekers in their profiles are more narrowly focused on the core skills of their occupations (for example job-specific skills) compared to the skills compositions of occupations found in job vacancies, which often list additional requirements, particularly character and social skills (see table 3.3).

Against this background, Luu and Escudero (forthcoming) examine skills mismatches in Uruguay. First, they capture “skills tightness” as the ratio of vacancies requiring a given skill to the number of applicants showcasing that skill on a monthly basis. The second measure predicts the skills workers are expected to have, based on the occupations they report and the skills that vacancies require for those occupations in the data. The results suggest that, while a significant number of workers in the online labour market of the country possess skills that are relevant to employers’ needs, they often lack – or fail to effectively communicate – the complete set of skills employers are seeking. This indicates that barriers beyond skills gaps – such as the accuracy of skills reporting – may also be hindering effective labour market matching. Addressing these issues could involve supporting individuals in creating more complete and accurate applications that not only reflect their full range of skills but are also tailored to the specific requirements of the jobs they are applying for – ultimately improving the labour market matching process.

3.4. Skills bundles and skills interconnectedness

While the previous analysis assessed skills individually, skills do not exist in isolation. Instead, jobs typically require a diverse set of skills that work together to perform tasks effectively. By assessing these combinations using network analysis and complexity methods, the remainder of this chapter helps gain a better understanding of the interplay between different skill types and how these interconnections lead to better labour market outcomes. This answers the ultimate question of which skills need to be fostered for workers to access better jobs.

The relevance of skills bundles is reflected in the online job vacancy data analysed in this report. In Brazil, the Russian Federation, South Africa and Uruguay, for example, the majority of vacancies require three or more skills, and most applicants list two or more skills in their applications (figure 3.7). In line with these findings, the interconnectedness of skills is becoming an active area of investigation. Studies emphasize that the demand for, and remuneration of, a given skill often depends on the type of skills it is combined with (Deming and Kahn 2018; Rodrigues, Fernández-Macías and Sostero 2021; Stephany and Teutloff 2024).

• Figure 3.7. Share of vacancies and applicants in selected countries, per number of identified skills (percentage)

Source: Analysis based on BuscoJobs data for Brazil and Uruguay (2010–23) and Adzuna data for the Russian Federation and South Africa (2016–21).

3.4.1. Identifying skill interconnections

Figure 3.8 shows that there are strong connections between the skills listed in vacancies, as represented by the density of the lines linking them. Notable connections are visible among the four types of socio-emotional skills, as well as between these skills and cognitive skills (both core and sophisticated) and computer skills, across all countries considered.

The connection between social skills and core cognitive skills is most pronounced in Egypt, Jordan, the United Arab Emirates and Uruguay. There are some additional country-specific patterns. Brazil, for example, displays the most pronounced links between the various socio-emotional skills. For South Africa, the relevance of skills bundles is most evenly spread across the various skills. Even skills that are less frequently mentioned in vacancies are typically required in combination with core cognitive, character and social skills, respectively. For instance, of all South African vacancies that require machine learning and AI skills, 48 per cent also require core cognitive skills, 52 per cent character skills and 76 per cent social skills.

• Figure 3.8. Skill networks within vacancies, by country

Note: Each skill network shows the relationships between skills in each country. Thicker lines indicate higher frequency of both skills being jointly required in job adverts.

Source: Analysis based on vacancy data from BuscoJobs Brazil and Uruguay (2010–23), Adzuna for the Russian Federation and South Africa (2016–21), and ESCWA for Egypt, Jordan and the United Arab Emirates (2020–24).

The frequency of some of these connections between skills is evident across different occupations, as shown for Brazil, the Russian Federation, South Africa and Uruguay in figure 3.9. For example, vacancies for both managerial and elementary occupations typically require core cognitive skills and general computer skills, combined with social and character skills, respectively. This underscores the fundamental importance of these skills bundles throughout the labour market. However, differences across occupations arise when looking at other combinations of skills. Vacancies for elementary occupations tend to emphasize a blend of manual and socio-emotional skills, while managerial ones naturally show fewer connections between manual and other types of skills. Instead, managerial occupations exhibit comparatively diversified skills bundles, including links between financial and writing skills, alongside core cognitive and all socio-emotional skills. In general, higher-level occupations within the occupational hierarchy require a more diverse and complex set of skills. They can therefore be considered more sophisticated from the skill perspective, as illustrated by the two measures of skills complexity and occupational sophistication described in box 3.4.

• Figure 3.9. Skill networks for vacancies as managers and elementary occupations, by country

Note: Each skill network illustrates the relationships between skills within each country, separately for managers and elementary occupations. Thicker lines indicate higher frequency of both skills being jointly required in job adverts.

Source: Analysis based on vacancy data from BuscoJobs for Brazil and Uruguay (2010–23) and Adzuna for the Russian Federation and South Africa (2016–21).

For applicants in Brazil and Uruguay, the patterns of skill interconnections broadly mirror those observed in vacancies. The intensity of these interconnections is generally lower, given the previous finding that applicants tend to report their skills less comprehensively in comparison to the skill descriptions in vacancies. When looking at different groups of workers by sex and education, the overall patterns remain consistent between these groups (see figure 3.10, which focuses on Uruguay). However, there are some distinctions: male applicants are slightly more likely to report skills bundles involving manual skills, whereas female jobseekers emphasize stronger combinations of customer service with core cognitive, financial and social skills. This pattern aligns with occupational gender segregation, where women are overrepresented in clerical support roles that require a mix of financial, customer service and complementary skills. In contrast, men are more frequently found in elementary occupations where manual and complementary skills are required jointly (see section 3.3).

When it comes to education, applicants with vocational or university education do not display strong differences in their skill networks compared to those without such qualifications. However, applicants without vocational or university education tend to highlight stronger connections between customer service skills and other skills, including general computer skills. The linkages among manual skills and between manual and customer service skills are also notable. In contrast, individuals with higher educational attainment tend to show slightly stronger links between financial and both sophisticated cognitive and people management skills, as well as between various cognitive and socio-emotional skills.

• Figure 3.10. Skill networks among applicants in Uruguay, by sex and education level

Note: Each skill network illustrates the relationships between skills mentioned in jobseekers’ employment histories, categorized by sex and education level. Thicker lines indicate higher frequency of both skills being jointly mentioned by jobseekers. Skills are aggregated across all job spells of a worker. “Higher education” refers to individuals with vocational or technical education, or university degrees. “Lower education” includes individuals who do not report having completed any of these education types.

Source: Analysis based on BuscoJobs applicants’ data for Uruguay (2010–23).

3.4.2. Characterizing skill interconnections

Having established the prominence of various skills bundles, this section further explores the nature of these interconnections in vacancy data. Using the bipartite network projection method introduced by Zhou et al. (2007), it examines the co-occurrence of skills in vacancies, the strength of the connections and their directionality. This analysis provides a more nuanced understanding of the underlying network structure, which is essential for determining how skills should be taught and learned, and for mapping the progression from prerequisite skills to more complex ones.

Figure 3.11 reveals two important patterns: first, several pairs of skills exhibit asymmetrical relationships, where the strength of the connection from one skill to another is significantly greater than in the opposite direction.56 This means that other skills are frequently observed alongside this skill in job vacancies, but not necessarily the other way around. This suggests that proficiency in one skill is often a prerequisite for acquiring or applying another, but not vice versa. Second, when these asymmetrical relationships are considered alongside the position and size of each node in the network, they reveal key dimensions of skill complexity. As noted earlier (box 3.4), skills that depend on many others – that is, those built on a broad base of foundational competencies – are assigned higher complexity scores.

In figure 3.11, such complex skills are represented by smaller nodes positioned at the top of the network. These more complex skills typically receive many incoming links from prerequisite skills but generate few outgoing links, except to other equally complex skills.57 Together, these patterns indicate that the most complex skills are those that require the support of other skills for mastery but are not themselves prerequisites for mastering other skills. Conversely, foundational skills may be less complex but are critical for the development of other more complex competencies.

• Figure 3.11. Skills networks showing asymmetrical relationships based on vacancy data complexity scores, by country

Note: Skills are arranged by complexity scores (see box 3.4) across three layers, from left to right and bottom to top. Smaller nodes represent more complex skills. Thicker lines indicate stronger connections between skills. Arrows denote directional importance: an arrow from skill A to skill B indicates a strong link in that direction. An asymmetrical relationship (a link going only in one direction) can be interpreted as a skill requirement: an arrow from skill A to skill B means that skill B requires skill A, with nearly all job vacancies that require skill B also listing skill A. Links going in both directions denote a symmetrical relationship that can be interpreted as a skills bundle. For each skill, only the three strongest connections are shown.

Source: Analysis based on vacancy data from BuscoJobs for Brazil and Uruguay (2010–23) and Adzuna for the Russian Federation and South Africa (2016–21).

The  complexity of skills and the direction of their relationships are country-specific, as can be observed from the changing complexity and position of finger dexterity skills across countries. However, several common patterns emerge across countries. Machine learning and AI skills, as well as managerial skills (particularly project and process management) are consistently identified as the most complex. Machine learning and AI skills rely heavily on general computer and software-specific skills, as well as cognitive skills, and they also contribute to the development of other complex skills. Project and process management skills are among the most challenging to develop because they require a diverse mix of cognitive and socio-emotional skills. Core cognitive skills, general computer skills, customer service, and social and character skills are located at the base of the networks, showing their foundational role in supporting the development of higher-complexity skills.

Importantly, asymmetrical relationships between skills may vary in nature. Rodrigues, Fernández-Macías and Sostero (2021) distinguish between two types of asymmetrical relationships, both illustrated in figure 3.11. First, some skills evolve into more complex ones. For instance, software-specific skills require a foundation in basic computer skills but can later progress to more complex AI and machine learning skills. Second, some complex skills are composites that consist of multiple lower-level skills. Managerial skills exemplify this type of relationship, as they require a broad array of cognitive and socio-emotional skills.

In addition to asymmetrical interconnections, some pairs of skills exhibit reciprocal relationships, where skills have strong bidirectional links, indicating a high level of mutual reinforcement and complementarity in their development and use (the analysis is available upon request). Socio-emotional skills are a key example. They not only contribute to the development of other skills but are also enhanced by them. For instance, social skills (such as empathy, flexibility and assertiveness) can improve customer service skills, while customer service skills (like selling and advertising) can, in turn, strengthen social skills, such as communication and negotiation.

Both reciprocal and asymmetrical skill networks tend to remain stable over time, reflecting the general trend of stable consistent compositions over the years (see section 3.3.1). Understanding these relationships is crucial for distinguishing between skills that are fundamental to the development of other skills and those that work as integral components of a skills bundle, where skills are jointly needed to perform a job. This distinction provides valuable insights for designing more effective and comprehensive strategies to support skills development.

The analysis shows that socio-emotional skills, along with core cognitive and basic computer skills (to a slightly lesser extent), play a key role in acquiring and enhancing other, more complex skills. The importance of these prerequisite skills is supported by existing literature (Cedefop 2018; Cedefop et al. 2021; Weinberger 2014). For instance, Cedefop (2018) finds that adults employed in jobs requiring basic ICT skills also need strong basic skills (for example, literacy, numeracy – also referred to as foundational literacies) as well as planning and organization skills. Establishing a solid foundation in these “basic or foundational skills […] that are necessary for acquiring many important higher-level skills” (Rodrigues, Fernández-Macías and Sostero 2021, 10) enables workers to adapt more readily to evolving job requirements – though the pace of change varies across countries. It also facilitates upward occupational mobility and transitions into jobs with better work quality, as will be demonstrated in the next section.

Which skills are crucial for advancing to higher-level occupations?

Building on the complexity and occupational mobility literature, which explores task similarity and skill networks to understand career pathways and human capital portability (Gathmann and Schönberg 2010; Del Rio-Chanona et al. 2021), figure 3.12 illustrates potential occupational progression based on the skills composition of occupations. For each occupation, the outgoing arrows indicate the three most accessible career paths, determined by the skills most relevant to target occupations, as indicated by to vacancy requirements.58 Occupations with higher ISCO classifications and lower sophistication scores – representing occupations with less complex skill content (box 3.4) – are concentrated in the bottom left-hand corner and are largely dominated by manual skills. In contrast, occupations with lower ISCO classifications and higher sophistication scores, located at the top right-hand corner, tend to require broader and more diverse sets of skills. This convergence between the data-driven approach used in this chapter and the ISCO classification – which is developed by labour statisticians based on their deep knowledge of national labour markets – lends credibility to both methodologies, as it reproduces conceptually grounded occupational distinctions through independent, data-based means, providing a novel contribution to understanding occupational complexity.

Figure 3.12 additionally reveals that possible transitions between occupations that share similar skill sets do not occur necessarily within narrowly defined occupation families. This pattern suggests that skills are a more reliable predictor of occupational mobility than the ISCO family classification. For example, occupations that primarily require manual skills generally lead to other physically intensive roles, although some transitions to occupations dominated by socio-emotional skills or a mix of socio-emotional and manual skills are possible. As such, machine operators and assemblers (ISCO 82) may transition to other physically intensive occupations – such as metal, machinery, and related trade workers (ISCO 72) – or to agricultural, forestry and fishery labourers (ISCO 92), which require both social and physical skills. These roles can then lead to more complex occupations, such as market-oriented skilled agricultural workers (ISCO 61 and 62). Similarly, life science and health professionals (ISCO 22) – categorized as physically intensive roles – can transition to customer service clerks (ISCO 42), which require a mix of socio-emotional and manual skills.

• Figure 3.12. Projected network of occupations (ISCO two-digit) based on the interconnectedness of skills (bundles), South Africa

Note: Nodes are two-digit occupations arranged along the x-axis by ISCO classification numbers (from highest to lowest) and along the y-axis by sophistication (that is, the sum of the skills required, weighted by complexity – see box 3.4). Occupations with lower ISCO classification numbers and higher sophistication scores are thus placed in the top right corner. Larger nodes represent more complex occupations. Only skills appearing more frequently than the average across occupations are retained. An occupation is considered to require a skill category if at least half of the skills in that category are above the average across other occupations. Occupations are shown as “below average” if there are no skills categories with more than half of the skills above the average across occupations. Arrows indicate potential occupational paths based on the skills composition of occupations, with outgoing arrows pointing to the three most accessible occupations.

Source: ILO calculation based on Adzuna vacancy data for South Africa (2016–21).

When considering the skills necessary for advancing to higher-level occupations, socio-emotional skills appear to serve as a bridge for occupations dominated by manual skills, enabling them to transition to more complex roles. Adding cognitive skills further expands the pathway to the most sophisticated occupations. Workers in occupations demanding a balanced combination of cognitive, socio-emotional and manual skills occupy the upper sophistication quadrant, enabling them to transition into roles that require a flexible skill mix (orange nodes). For example, general and keyboard clerks (ISCO 41) can leverage their skill endowments to move into other mixed-skill roles within information and communications technology occupations (ISCO 35 and 25).

However, in South Africa, vacancy data suggest the possibility of more occupational transitions between closely related roles than in the other countries, reflecting the characteristics of the South African skill network. Brazil, the Russian Federation and Uruguay exhibit less defined clusters of occupations with similar skills, resulting in potential transitions among a wider range of occupations. This means that in these countries, skills are more diffusely spread across different occupations, making it easier for workers to move between jobs that may not appear closely related but share overlapping skill requirements. In the Russian Federation, for example, there is more interaction between physically intensive occupations, those requiring a combination of social and manual skills, and those that require a mix of skills from all categories. In this setting, machine operators and assemblers (ISCO 82) could potentially transition to careers as science and engineering professionals (ISCO 21), which require a blend of cognitive and manual skills. However, the feasibility of such transitions depends – among other factors – on whether individuals require additional training or education and, if so, how long this lasts. On the other hand, in all countries considered, domestic workers (included within ISCO 51 and 91) show some potential for mobility, though transitions tend to occur among occupations with similarly low skill intensity or a strong reliance on manual tasks.

3.5. Skills for better-quality employment opportunities

Having seen how prominent skills bundles are – and how strong some of their interconnections – it is important to understand which skills and skill combinations are linked to better jobs. Identifying the skills and skills bundles associated with more favourable worker outcomes is essential for understanding how skills development can lead to higher-quality employment. This section thus investigates how developing certain skills and their bundles can support individuals in accessing better-quality employment opportunities. The ILO defines job quality as “the sum of work and employment conditions that influence an employee’s well-being” (2023b, section 1.3) This concept can be measured using objective criteria – such as the respect and fulfilment of labour rights, including fair wages, social protection and safe working conditions – or subjective criteria that reflect workers’ own assessments of job quality, including job satisfaction.

Online vacancy data allow capturing several aspects of job quality. From these data, it is possible to derive indicators such as advertised wages and non-wage job characteristics, some of which reflect core labour rights, while others represent attributes commonly associated with better job quality. This section explores the connection between these indicators and specific skills, helping to identify which skills should be developed through lifelong learning initiatives to prioritize to improve individuals’ employment prospects.

3.5.1. How skills influence wage levels

While vacancy data do not allow for a direct measurement of adequate earnings that guarantee decent living standards for workers and their families (ILO 2023b), the data enable the assessment of the wage brackets – that is, minimum and maximum offered wages – associated with advertised positions. These wage brackets provide a valuable proxy for evaluating the earning potential of jobs, which is a crucial aspect of job quality, and offer insights into how different skills are associated with higher wages.59

Higher-paying jobs require more skills

Comparing the skills required for the highest-paying jobs (top 20 per cent) with those needed for the lowest-paying jobs (bottom 20 per cent) across the four reviewed countries shows that jobs requiring more skills offer higher wages (figure 3.13). While there is variation across countries, vacancies offering wages in the top 20 per cent tend to require a greater number of core and sophisticated cognitive skills, specialized and general computer skills, financial skills, social skills and people management skills. However, the specific skills most commonly associated with higher-paying vacancies vary across countries. For instance, in South Africa, the largest gap in skill requirements between higher- and lower-paying jobs is in financial and core cognitive skills, whereas in the Russian Federation, certain skilled manual roles are paid higher wages. Jobs requiring specific manual skills – such as finger dexterity for working with equipment, packing and sorting, or hand–foot–eye coordination for repairing, restoring, transporting and driving – can command higher wages. Jobs in building (ISCO 71), metal and machinery trades (ISCO 72), as well as machine operators (ISCO 81), assemblers (ISCO 82) and drivers (ISCO 83) often offer wages in the top income range. These roles tend to pay well because they are critical to industrial sectors (such as oil, gas and mining), where specialized expertise and the physical demands of the work drive up wages for qualified workers.

Some skills are more commonly required for lower-paying jobs. In most countries, this is the case for customer service skills, except for the United Arab Emirates. For other skills, this pattern varies more strongly by country. For example, character skills are more prominent in lower-wage postings in South Africa, while general computer skills are more common in lower-paying jobs in the Russian Federation. However, it is possible that the prevalence of certain skills in lower-paying jobs may not necessarily determine wages directly. Instead, differences in remuneration across occupations, as well as the importance of skills bundles, may influence these wage patterns, as will be examined next.

• Figure 3.13. Skills required for the 20 per cent highest and lowest posted wages, by country

Note: The outermost ring corresponds to 100 per cent of vacancies mentioning a skill. Therefore, a marker further away from the centre means the given skill is more frequently mentioned in vacancies. Wages (in 2023 values) refer to the minimum wages advertised in the vacancies.

Source: Based on Adamczyk et al. (forthcoming) and El Hage Sleiman, Liepmann and Mokdad (2025). The analysis uses vacancy data from BuscoJobs for Uruguay (2010–23), Adzuna for the Russian Federation and South Africa (2016–21), and ESCWA for the United Arab Emirates (2020–24).

Skills matter for wages even within occupations, with cognitive skills suggesting the highest returns

The analysis has so far assessed country-level results across occupations, where it is possible that the relationships between skills and wages is influenced by occupational wage differences. Figure 3.14 includes occupational control variables, such that the estimated effects show how skills relate to wages within a given occupation. While providing informative insights, this analysis does not detect causal relationships between skills and wages, given that additional factors (such as firm or sectoral effects) are not accounted for.

• Figure 3.14. Relationship between skills and wages controlling for occupation-specific factors, by country

Note: The figure shows the results from vacancy-level regressions of logged wages (defined as the lowest monthly fixed wage advertised, adjusted to 2023 values) on 15 skill dummy variables (equal to 1 if a skill is mentioned in a job advert, 0 otherwise). Each graph comes from a separate regression and displays estimated coefficients with 95 per cent confidence intervals. Coefficients indicate the wage difference, in percentage, associated with a given skill, compared to otherwise observationally identical vacancies that do not reporting that skill. All regressions control for two-digit occupations, the number of words in a job advert, year and calendar month fixed effects, and part-time status (except for the United Arab Emirates, where this variable was unavailable). Large sample sizes mean coefficients are estimated with high precision for the Russian Federation and South Africa.

Source: Adamczyk et al. (forthcoming) and El Hage Sleiman, Liepmann and Mokdad (2025). The analysis is based on vacancy data from BuscoJobs for Uruguay (2010–23), Adzuna for the Russian Federation and South Africa (2016–21) and ESCWA vacancy data for the United Arab Emirates (2020–24).

Clearly, skills matter for wages even when one abstracts from occupational factors. Across countries, many of the cognitive skills are associated with higher wages. In South Africa, for example, wages are 20 per cent higher when financial skills are required in a posting. In the United Arab Emirates, posted wages are 18 per cent higher in vacancies requiring machine learning and AI skills. Meanwhile, posted wages are 7 per cent higher and 14 per cent higher, respectively, for vacancies in the Russian Federation and Uruguay requiring project and process management.60 Research from high-income countries reaches similar conclusions. Digital skills, for example, carry a significant wage premium in the United States (Alekseeva et al. 2021) and the United Kingdom (Sostero and Tolan 2022). Yet, the explicit requirement of basic cognitive skills can, in fact, be associated with lower wages. This is true for general computer skills, which – in comparison to otherwise observationally identical vacancies – yield a 10, 6 and 3 per cent decrease in wages in South Africa, the United Arab Emirates and the Russian Federation, respectively.

For socio-emotional skills, the findings are more varied. These skills are generally associated with a wage premium in the Russian Federation. In South Africa, by contrast, character and customer service skills are negatively related to wages. In the United Arab Emirates, character and social skills are those associated with a wage penalty. The same applies to social skills in Uruguay.

Why do certain socio-emotional skills not yield wage premiums, at least in some countries? As discussed in Chapter 4, skills wage premiums are shaped not only by market demand and supply, but also by socially constructed value judgements. For example, social skills are often associated with tasks and skills typically performed by women, such as in care work. Such skills – though essential – tend to be undervalued and underpaid (ILO 2023c; Osterman et al. 2022). One contributing factor is their connection to the unpaid labour that women disproportionately provide for their families, which can lead to the perception that these skills are less demanding. Additionally, the public good nature of care work results in market inefficiencies and underinvestment, further depressing the value of these skills (see section 4.2). Yet, as subsequent results in this section will show, socio-emotional skills still play a significant role in raising wages, as they complement cognitive skills (see also Deming and Kahn 2018; Weinberger 2014).

Manual skills are associated with lower wages in South Africa and the United Arab Emirates. The same is true for the Russian Federation, with the exception of finger dexterity skills. Despite variation across countries, manual skills generally do not yield noticeable wage premiums. This is relevant from a policy perspective, given that workers with low qualification levels often perform manual tasks (Roys and Taber 2022). Again, wage effects for manual skills can be positive when these skills are combined with complementary skills, as discussed below. A similar pattern was already shown in figure 3.12, where the presence of other skills enabled manual-skill occupations to move up the complexity ladder.

To complement this analysis, Adamczyk et al. (forthcoming) go one step further in abstracting from occupational factors. In theory, differences between jobs – such as varying levels of productivity or differences in firm structures – may influence observed relationships between skills and wages in ways unrelated to the true skill-related wage premium. To better isolate the effect of skills on wages, it is therefore essential to account for these job-specific factors. Deming and Kahn (2018), for example, assess returns to skills within narrowly defined occupations, while Marinescu and Wolthoff (2020) show that controlling for job titles is the most precise approach to understanding application dynamics. For South Africa, it is possible to include detailed controls for almost 4,000 job titles. While these controls slightly reduce the magnitude of skill effects on wages, the overall patterns remain consistent, indicating that skills continue to influence wages even within the same job.

To access high-paying jobs, complementary skills are required, particularly those belonging to the socio-emotional category

It could seem easy to conclude from the previous analysis that acquiring cognitive skills alone is sufficient for securing a high-paying job. However, as shown in section 3.4.1, skills do not exist in isolation. Many cognitive skills are strongly reinforced by others, mostly socio-emotional skills. The following therefore explores whether such skills bundles are also associated with higher wages.

Table 3.6 displays the five top-paying skills in four countries, together with the three skills that are most frequently demanded to complement the top-paying skills. In a majority of vacancies, these complementary skills are social, customer service and/or character skills. In South Africa, for example, financial skills – which yield a wage premium of 20 per cent – regularly require additional social and customer service skills (in 66 and 64 per cent of the vacancies, respectively). In Uruguay, general computer skills are associated with a 10.5-per-cent increase in wages. Of these vacancies, 54 per cent also request social skills, while 52 per cent search for complementary customer service skills. Among the skills with the highest remuneration, there are also a few examples of socio-emotional skills, which, in turn, require cognitive skills. In the United Arab Emirates, people management skills are related to a 20-per-cent increase in wages and are jointly required with core cognitive skills in 79 per cent of vacancies.

These  findings highlight the fundamental role of socio-emotional skills as enablers of higher-paying, usually cognitive, skills. Even if some socio-emotional skills may not be valued in isolation – such as character, customer service and/or social skills, which can even carry a wage penalty as in South Africa, the United Arab Emirates and Uruguay – they are nevertheless required in conjunction with high-paying skills. As discussed in section 3.4, socio-emotional skills play the dual role, both enhancing and relying on other skills. Literature for the United States confirms this finding: Cortes, Jaimovich and Siu (2023) show that the relevance of socio-emotional skills has increased in higher-paying occupations, while Deming and Kahn (2018) and Weinberger (2014) demonstrate that the combination of socio-emotional and cognitive skills leads to significant wage premiums.61 Moreover, jobs intensive in both math and social skills have seen employment growth, whereas employment in jobs intensive in only math skills have declined (Deming 2017). Goulart, Rodríguez-Menés and Caroz Armayones (2022) shed further light on these relationships. In their qualitative study, hiring managers in Spain explain that applicants are often initially screened based on technical skills, requiring certifications or good results in a programming test. Subsequently, the most suitable candidate is chosen out of this pool based on the social and character skills demonstrated during interviews.

While manual skills do not tend to feature among the top-paying skills analysed in table 3.6, skill complementarities can also affect their remuneration. In their study of Indian firms, De Marzo, Mathew and Sbardella (2023) find that manual skills are better remunerated when combined with digital skills (box 3.5). Similarly, Senegalese workers with low formal qualification levels, who performed mostly manual labour for the construction of tracks of an express train, saw their earnings increase after they had participated in a training focused on conscientiousness (Allemand et al. 2023).

These findings have important implications for policy. While transitioning to entirely new roles can be complex, even for individuals with the required skills, our analysis demonstrates that such drastic changes may not always be necessary to achieve higher wages. Often, the key to improving wage potential lies in acquiring complementary skills, particularly those that are socio-emotional in nature. By recognizing the value of these complementary skills, policymakers can design targeted training programmes and initiatives that enhance workers’ socio-emotional competencies. This approach not only increases individual earning potential but also fosters a more skilled, adaptable and integrated workforce.

3.5.2. The relationship between skills and non-wage attributes

Beyond wages, various employment conditions and job characteristics shape job quality. Online vacancy data enable the identification of non-wage job attributes62 that reflect aspects of job quality, including certain labour rights and other desirable features of employment. However, they do not allow comparison between formal and informal jobs. Non-wage job attributes include various forms of variable earnings, fringe benefits, working conditions and broader workplace characteristics. Adamczyk, Delaporte and Escudero (2025) rely on empirical literature (mainly Maestas et al. 2023; Sockin 2021), along with their own adaptations, to develop a taxonomy of non-wage job attributes extracted from online vacancies (table 3.7). This taxonomy, underpinned by a keyword-based dictionary, was applied to vacancy data from Brazil, the Russian Federation, South Africa and Uruguay, using NLP techniques akin to the ones used for the creation of skill variables.

As presented in figure 3.15, the prevalence of advertised non-wage attributes in job vacancies varies across the four countries analysed, highlighting differences in employer practices and job market expectations. The share of vacancies mentioning at least one of these attributes is highest in the Russian Federation (75 per cent), followed by Uruguay (51 per cent), Brazil (49 per cent) and South Africa (35 per cent).

• Figure 3.15. Share of advertised non-wage attributes in vacancies

Source: Analysis based on BuscoJobs vacancy data for Brazil and Uruguay (2010–23), and Adzuna vacancy data for the Russian Federation and South Africa (2016–21).

Despite these differences, certain commonalities emerge. Opportunities for career development consistently ranks among the most frequently mentioned attributes, particularly in Uruguay (59 per cent) and the Russian Federation (57 per cent), reflecting an emphasis on training and career growth opportunities. Working in teams is also prominent, especially in Uruguay (45 per cent) and Brazil (14 per cent), where collaborative work environments are emphasized. Additionally, work environment and impact on society is widely advertised in Uruguay (36 per cent) and to a lesser extent in Brazil (9 per cent), suggesting a focus on workplace culture and corporate social responsibility. On the other hand, financial incentives, such as bonuses and commissions, are prominently featured in Brazil (25 per cent) and the Russian Federation (25 per cent), as well as food, subsidies and discounts accounting for 31 per cent in Brazil and 22 per cent in the Russian Federation. Interestingly, location and commuting is an important advertised factor in the Russian Federation (25 per cent) but is less important in Brazil and Uruguay, and almost absent in South Africa (0.4 per cent).

In contrast, some attributes appear less frequently in online job postings. Labour rights and social protection benefits, such as health insurance, paid time off and retirement contributions, are seldom mentioned, particularly in Uruguay, where legal requirements may make explicit advertisement unnecessary. Similarly, office space and amenities and work equipment and allowances are rarely highlighted across all four countries. Only in Brazil is health insurance mentioned more prominently (in about 10 per cent of postings), while in the Russian Federation, paid time off appears relatively more frequently (20 per cent).

Given these variations, 11 non-wage attributes were selected for the analysis of the relationship between skills and job quality, as summarized in table 3.7. Seven of these attributes are positively associated with job satisfaction and employee well-being and are therefore called “amenities”, with some also aligning with labour rights. For contrast, two disamenities were included, indicating low job quality. These classifications are based on a literature review, including sentiment analysis and willingness-to-pay studies (Goldin 2015; Maestas et al. 2023; Mas and Pallais 2017; Sockin and Sockin 2025; Wiswall and Zafar 2018, among others). Additionally, two job characteristics were included without prior classification as amenities or disamenities, as their effects can be either positive or negative depending on the group and context. Their categorization will therefore be determined empirically.

Relationship between skills and amenities

The relationship between required skills and advertised workplace attributes varies by country.63 Yet, certain patterns consistently emerge.

In terms of workplace characteristics, jobs that emphasize opportunities for career development, working in teams and work environment and impact on society are more likely to require socio-emotional skills – particularly social and customer service skills – and cognitive competencies. However, these jobs show little connection – or even an inverse relationship – with manual skills. In Brazil, these associations are especially strong, suggesting that vacancies highlighting professional growth and collaboration typically seek candidates with a combination of cognitive and socio-emotional skills (table 3.8).

On the other hand, as expected, physical effort and pace of work tends to be associated with manual skills while being less relevant for cognitive or socio-emotional competencies. Meanwhile, location and commuting shows little connection with skills.

Within the realm of working conditions, work schedule flexibility tends to align with roles requiring technical cognitive skills – such as computer, software, and machine learning and AI skills – which suggests that professionals in ICT and related fields may enjoy greater autonomy over their schedules. However, this pattern is not equally strong across all countries, indicating that flexibility in technical jobs may depend on local labour market conditions. In contrast, vacancies requiring manual skills consistently lack schedule flexibility, where structured shifts may dominate.

Workplace safety is more frequently mentioned in jobs requiring cognitive skills and people management abilities in Brazil and South Africa – consistent with the idea that labour rights are more commonly reflected in higher-skilled vacancies – though correlations are small. Job security, meanwhile, has little connection to skill requirements, except in Uruguay, where it is negatively correlated with all skills.

Finally, within variable earnings and fringe benefits, jobs offering bonuses and commissions are more likely to emphasize customer service skills in Brazil and Uruguay. In the Russian Federation and South Africa, this extends to all socio-emotional skills and more complex cognitive skills – such as sophisticated cognitive and financial skills – indicating that commission-based pay structures are particularly common in jobs that require a specific combination of cognitive and strong socio-emotional abilities. This pattern may also reflect the types of occupations in which such pay structures are more prevalent – such as sales and finance roles – which tend to demand these skills. Similarly, food, subsidies and discounts tend to be more common in jobs that do not require advanced cognitive skills, reinforcing their association with less specialized roles. Meanwhile, hourly work and overtime appear across roles of varying skill levels, with no strong link to specific competencies.

More complex jobs come with greater and higher-value amenities

The  analysis of the relationship between non-wage attributes and job complexity64 shows that vacancies requiring more complex skills consistently offer higher-value amenities, particularly opportunities for career development, working in teams and work environment and impact on society (figure 3.16). This is in line with above correlations showing that these amenities are more prevalent in roles demanding advanced cognitive and socio-emotional skills (table 3.8).

• Figure 3.16. Distribution of advertised non-wage attributes in the 20 per cent most and least complex vacancies, by country

Note: The outermost ring corresponds to 100 per cent of vacancies mentioning an attribute. Therefore, a marker further away from the centre means the attribute is more frequently mentioned. Vacancy complexity is calculated as the sum of the square root–transformed complexity scores of all skills required in the vacancy, by country (see box 3.4).

Source: Analysis based on BuscoJobs vacancy data for Brazil and Uruguay (2010–23), and Adzuna vacancy data for the Russian Federation and South Africa (2016–21).

Beyond these core trends, certain amenities are linked to more complex jobs in specific countries. Work schedule flexibility is more commonly found in Brazil and the Russian Federation, suggesting that roles with more complex skills in these markets allow for greater autonomy. In South Africa, hourly work or overtime and workplace safety are more frequently associated with jobs requiring more complex skills. Bonuses and commissions are also more common in higher-complexity jobs, though this trend is strongest in the Russian Federation and, to a lesser extent, in Brazil and South Africa.

Conversely, less complex jobs are more likely to feature food, subsidies and discounts, as well as the disamenity of physical effort and pace of work. In the Russian Federation, location and commuting also play a more prominent role in lower-complexity jobs.

These patterns hold when comparing non-wage attributes offered in vacancies across higher- and lower-skilled occupations (for example, managers and elementary occupations), reinforcing the link between job complexity and the benefits provided.

However, more amenities do not necessarily mean higher wages

Interestingly, while more complex skills are typically linked to higher-paying jobs (section 3.5.1) and more complex jobs tend to offer greater and higher-value amenities, the association between salary levels and amenities depends on the strategic role that these attributes play at different job levels. Jobs in the top 20 per cent of wages do not consistently provide more or better job attributes compared to those in the bottom 20 per cent (figure 3.17). In some cases, amenities commonly associated with complex roles are also prevalent in lower-paying jobs. For instance, opportunities for career development is more frequently advertised in the lowest-paid jobs in Brazil and the Russian Federation, suggesting that lower-wage roles may emphasize training and skill development as a means of career progression. Conversely, location and commuting benefits are more commonly found in the highest-paid vacancies in the Russian Federation, likely reflecting the need to attract highly skilled professionals by compensating for geographic constraints.

• Figure 3.17. Distribution of advertised non-wage attributes for the 20 per cent highest and lowest salaries, by country

Note: The outermost ring corresponds to 100 per cent of vacancies mentioning an attribute. Therefore, a marker further away from the centre means that the attribute is more important. Wages (in 2023 values) are measured as the lowest advertised wage in the vacancies.

Source: Analysis based on BuscoJobs vacancy data for Brazil and Uruguay (2010–23), and Adzuna vacancy data for the Russian Federation and South Africa (2016–21).

This discrepancy arises because amenities serve different strategic purposes depending on the job level. In lower-wage positions, employers may emphasize training opportunities to attract workers by offering career advancement prospects rather than immediate financial rewards. In contrast, high-paying jobs may provide commuting allowances to secure top talent, particularly when roles require relocation or frequent travel.

These descriptive patterns provide new evidence on how skills, wages and non-wage job attributes interact across different contexts and job types.65 They align with the extensive literature on compensating differentials, which suggests that jobs with less desirable characteristics (such as low wages, long hours or relocation requirements) may offer non-monetary benefits to attract and retain workers (Lamadon, Mogstad and Setzler 2022; Sorkin 2018; Taber and Vejlin 2020). Such compensating differentials are evident in lower-paying jobs of Brazil and Uruguay, which tend to offer a greater number of valuable amenities. In contrast, for higher-paying jobs in the Russian Federation, higher wages are often accompanied by a mix of attractive amenities and attributes more commonly associated with less complex or lower quality jobs – such as location and commuting support and food subsidies. This suggests that beyond classic compensating differentials, other labour market factors – such as industry norms, job competition and regulatory requirements – also shape how non-wage job attributes are distributed across wage levels.

3.6. Conclusion

This chapter has shown how developing specific skills – and the right combinations of them – can support individuals in accessing better-quality employment opportunities, with a focus on countries in regions often overlooked in the skills literature. It studied skills dynamics in Brazil, Egypt, Jordan, Morocco, South Africa, the Russian Federation, the United Arab Emirates and Uruguay, while drawing on existing research from countries such as the United States for additional context. Given the scarcity of granular and longitudinal data on skills in the selected countries, the analysis leveraged big data from online vacancies (and applicant data, where available). This was possible thanks to the application of ILO methodologies, which were specifically developed for this report to categorize and elicit skills and amenities variables from unstructured text data. Moreover, the chapter benefited from a collaboration with UN ESCWA and Veille+ in Morocco to enlarge the number of countries studied.

While the data come with certain limitations – for example, the online labour market tends to overrepresent professional occupations and differs in sectoral composition from the broader labour market – they nonetheless provide a critical source of insight, particularly in contexts where traditional skills data are scarce or absent.

In summary, the chapter identified key findings in four areas.

First, socio-emotional skills are the most strongly demanded, followed by cognitive skills. While manual skills are less frequently observed in vacancies, they remain relevant, with differences by occupation. Overall, the shares of broad skill categories demanded are relatively stable over time. However, amid this stability, specific skills – especially advanced software, and machine learning and AI – are evolving rapidly, even though they still represent a small share of total demand. On the supply side, the skill distribution among applicants closely mirrors that of vacancies. Interestingly, men and women report broadly similar skill profiles – though women more frequently cite customer service skills, while manual skills are more often reported by men – reflecting their occupational distribution.

Second, skills do not exist in isolation but rather as part of interconnected skills bundles. Enterprises require a blend of different skills, where various socio-emotional skills are strongly interconnected among themselves and also with cognitive skills, and, to a lesser extent, manual skills. Network analysis allowed to further characterize the interconnectedness of skills: socio-emotional skills, along with core cognitive and general computer skills act as foundational competencies that support the development of higher-complexity skills. Prominent examples of higher-complexity skills are machine learning and AI as well as managerial skills, in particular project and process management.

Third, higher wages are generally associated with a greater number of higher-complexity skills, with cognitive skills yielding the highest wage returns. Socio-emotional skills show more varied results and sometimes even negative wage effects – a conclusion that will be revisited in section 4.2. However, socio-emotional skills still play a significant role in boosting wages, as they are often required alongside cognitive skills. These findings underscore the fundamental role of socio-emotional skills as enablers of more complex, higher-paying skills – a pattern also observed in the United States (Deming and Kahn 2018; Weinberger 2014) and India (De Marzo, Mathew and Sbardella 2023). By contrast, manual skills rarely yield wage premiums in any of the countries considered. While manual skills may not be valued on their own, complementarities with other skills can enhance their remuneration.

Finally, skills are a pathway to jobs offering better amenities. Vacancies demanding more complex skills – often a mix of advanced cognitive and socio-emotional skills – consistently offer higher-value amenities, such as opportunities for career development, working in teams and impact on society. However, some of these amenities commonly associated with complex roles are also prevalent in some lower-paying jobs, suggesting that jobs with undesirable characteristics may compensate for low wages by offering non-monetary benefits to attract workers.