ESCO vs O*Net

ICATUS Level Choice

As discussed here, the ICATUS taxonomy comprises three levels that represent different levels of disaggregation, namely the major division, division and group levels. For Tabiya’s matching procedure, the group level activities pertaining to the unseen economy are matched to ESCO Occupations. There are a total of 60 ICATUS group level activities that comprise the Framework’s ICATUS input. The group level is chosen, rather than the major division or division level primarily due to the fact that the group level attains the highest level of precision in the ICATUS taxonomy. Thus, we are able to glean the most detail when matching from this level, relative to the major division and division levels.

The challenge, though, has been in the matching of group level activities to ESCO occupations that occasionally define activities and occupations at differing levels of granularity. In particular, we find that there are some instances where an ESCO Occupation comprises more than one ICATUS group level. For instance, the ICATUS group level activities, “Cleaning up after food preparation/meals/snacks'' and “Preparing meals/snacks” are each matched to the ESCO occupation, “kitchen assistant” (defined as “Kitchen assistants assist in the preparation of food and cleaning of the kitchen area.”). Clearly, the occupation “kitchen assistant” is made up of these two ICATUS group level activities. Theoretically, one way to circumvent this would be to match ICATUS activities at the division level, rather than at the group level. However, this method is ultimately deemed too information costly since at the division level, the chance of omitting relatively rare but relevant ESCO occupations (and in turn, skills) from the ICATUS to ESCO match is very high. This does harm to the end user because it limits the eventual list of skills available to represent the work that they do. After discussion between the research team, and the Harambee team, we mutually agreed that the best way to do no harm in this case is to match ICATUS activities to ESCO occupations at the group level.

ESCO occupations as a substep to skills (NLP, Word2Vec etc.)

The reason that ICATUS activities are matched to ESCO occupations, and then to skills as opposed to a match from ICATUS activities immediately to skills requires a summary of the conceptual roadmap taken to ultimately come to this decision. Initially, since the goal is to assign skills to activities, the most intuitive course of action seemed to be to directly match ICATUS activities to ESCO skills. In order to do this, Natural Language Processing (NLP) techniques were adopted. In particular, to compare ICATUS activities to ESCO skills, a technique called Word2Vec was adopted. Word2Vec is an algorithm used in NLP that uses word embeddings to assign numerical values to a given word in relation to the surrounding context of this word which is informed by the relationship that this given word has with other words in a dictionary used to train the algorithm (Mikolov et al., 2013). After several attempts to match ESCO skills directly to ICATUS activities, NLP techniques failed due the incompatibility in nature between ICATUS/ESCO as inputs and the required input for the algorithm to function optimally. Since ICATUS comprised short phrases pertaining to activities rather than single words as the ESCO taxonomy does for skills, Word2Vec could not optimally reconcile these taxonomies to produce a meaningful skills match. From this arose a precision-reproducibility tradeoff. On the one hand, NLP methods allow for near perfect reproducibility but score low on the precision scale in this context. A decision was taken to prioritize precision throughout the conceptualizing of the Tabiya Framework. Thus, the prevailing methodology that uses ESCO occupations as a crosswalk to skills proved to strike the precision-reproducibility scale optimally for our purposes. That is, the manual assignment of ESCO occupations to ICATUS activities to derive skills associated with unseen economy proved to score high on the precision scale and high enough on reproducibility scale, since this work was independently checked. Hence, the manual matching procedure prevailed for the Tabiya Framework’s creation.

Le capital humain - l'ensemble des compétences, connaissances, et expériences d'un individu - est un déterminant important de la productivité et du niveau de revenu potentiel. Le niveau de capital humain dont dispose un individu est souvent directement corrélé à sa capacité à dégager un revenu. Il est aussi intimement lié à la liberté de l'individu à prendre des décisions et à choisir son mode de vie. Tabiya et ses partenaires visent à rendre le capital humain visible et utilisable.

Nos motivations sont bien résumées par les propos de l'un de nos partenaires, Harambee Youth Employment Accelerator :

Les jeunes Sud-Africains manquent souvent de ressources, de réseau, d'éducation et d'expérience professionnelle pour être des candidats crédibles aux emplois formels. Mais, dans les 12 dernières années, notre travail nous a montré que les jeunes ont le potentiel nécessaire pour occuper ces emplois si la chance la leur est donnée! Et si les jeunes étaient les mieux placés pour identifier et mettre en avant les compétences qu'ils ont acquises en dehors de l'économie formelle ? Et si il pouvaient signaller cette expérience acquises à travers le volontariat ?

Harambee Youth Employment Accelerator, Afrique du Sud

Objectifs

Les organisations qui aident les jeunes en recherche d'emploi à accéder à des opportunités économiques reposent généralement sur des taxonomies décrivant les emplois existant dans une économie donnée, ainsi que les compétences et qualifications qui leurs sont associées.

Or, les taxonomies traditionnellement usités pour mesurer l'activité économique et décrire les marchés du travail ignorent une large part du capital humain et de la production, comme par exemple les tâches ménagères (cuisiner, ranger, s'occuper des enfants), le volontariat, et le travail informel. Ces activités représentent une part importante des modes de vie et de subsistence et contribuent largement à l'acquisition de capital humain, mais elles sont souvent sous-évaluées et ignorées car elles ne donnent pas directement lieu à des transactions monétaires.

Nous visons à rendre visible toutes les connaissances, compétences et expériences pouvant permettre de produire de la valeur économique. Cela implique deux types d'efforts:

1. Rendre visible toutes les activités économiques et les compétences, connaissances et expériences retirées de ces activités, en particulier dans l'"économie invisible", c'est-à-dire l'économie informelle, le travail ménager non rémunéré etc...).

2. Rendre utilisable ces compétences, connaissances et expériences accumulées grâce à ceux activités en les organisant et en les intégrant dans les taxonomies traditionnellement utilisées pour décrire l'activité économique.

Afin d'atteindre le premier objectif, nous suivons une méthodologie décrite ici, basée sur les méthodes des sciences sociales et sur la collection de données empiriques. Nous travaillons au second objectif en créant une taxonomie inclusive des modes de subsistance pour que nos partenaires puissent l'amender et l'utiliser pour construire des plateformes d'intermédiation. Notre taxonomie de référence peut être parcourue, adaptée, et mise à jour au travers de notre Plateforme de Taxonomie inclusive.

Résultat: Une Taxonomie de référence inclusive comme base pour l'intermédiation et la recherche

Une taxonomie de référence offre une langue commune pour comprendre, catégoriser, et relier différentes informations à propos du marché du travail. C'est une système de classification structuré aidant l'interprétation et l'analyse de vaste et diverses data-sets portant sur les marchés du travail. Une telle classification peut être une manière de cartographier l'ensemble du marché du travail. Elle décrit l'univers des emplois existant dans une économie et l'anatomie de chaque emploi - en termes de compétences requises et de qualifications requises pour occuper cet emploi.

Cette taxonomie est une fondation pour que les gouvernement, ONG, et les autres acteurs puissent aiguillier efficacement les actifs en recherches d'emploi vers certaines carrières ou certains services d'aide. Le cadre ci-dessous présente un avant-goût des usages possibles d'une telle taxonomie.

Challenge: Les classifications traditionnelles ignorent de larges pans de l'économie

Les taxonomies traditionnellement utilisées pour analyser l'offre et la demande d'emploi négligent nombre d'activités qui permettent d'acquérir des compétences, connaissances, et expériences. Elles n'offrent donc pas un paysage complet et fidèle de la diversité des moyens de subsistence.

Commonly used labor market taxonomies typically overlook many economic activities that generate important skills, knowledge, and experiences; they fail to accurately portrait the often complex and diverse livelihoods of people.

One of the most widely used frameworks to think about economic activity in a structured way is the System of National Accounts (SNA). It serves as an international statistical standard for the measurement of national economic activities, employed by many countries around the globe. The SNA traces its roots back to the works of 20th-century economists, who sought to systematize the understanding of economic activities through a structured set of separate accounts. The SNA typically includes quantifiable, market-based economic activities.

However, several significant areas of human capital investment and productivity fall outside this boundary, such as household work for own production (e.g., cooking, cleaning, and childcare), volunteer work, and informal work. These activities form important parts of individual livelihoods and contribute significantly to human capital for millions of people around the world, but are often undervalued or ignored because they do not involve direct monetary transactions. We call these activities the "unseen economy".

Intuitively, "seen" activities are paid, and thus typically considered as work, whether they are formal or informal, if not illegal. "Unseen" activities include all productive activities, such as cooking for a family member, that are unpaid and thus typically not considered as work. One other way to qualify unseen activities is that they are unpaid activities, that one could pay someone else to do. Therefore, it does not include leisure, as one may not pay someone to enjoy leisure activities for them.

Recognizing and accounting for the human capital in the unseen parts of the economy is vital for accurately representing people's contributions to the economy, and providing a basis for policies that protect and support all forms of work, thereby enhancing individual agency.

Our Approach: Expanding the Map

With Tabiya's Inclusive Livelihoods Taxonomy, we aim provide a more inclusive map of the labor market – one that includes activities from the "unseen economy." A more inclusive map of the labor market will allow more inclusive matching, the identification of more diverse career and skill development pathways, and richer data analysis. A more detailed description of our methodology can be found here.

We are currently developing the Open Taxonomy Platform, which lets partners flexibly adapt our reference taxonomy in a transparent way. The platform also provides access to all taxonomies through an API, which allows users to query the platform for information on specific occupations or tasks.

The platform is currently under active open-source development. You can follow along and contribute on Github. If you would like to learn more, please get in touch.

Our Challenge

As explained here, Tabiya's work builds upon widely used taxonomies of skills and competences to ensure that our inclusive taxonomy is compatible with existing intermediation tools used by governments and organisations. However, there exist multiple taxonomies serving as basis for national matching platforms and public policies, among which ISCO (International Standard Classification of Occupations), ESCO, O*NET (Occupation Information Network) or OFO (Organising Framework for Occupations in South Africa).

The challenge when picking a taxonomy is the following. On the one hand, the taxonomy needs to be broad enough to account for the occupations and human capital of job-seekers on the whole African continent and elsewhere. On the other hand, it needs to be adaptable to local contexts, which implies using a language understandable by local users, and encompassing context-specific skills and occupations. Finally, it needs to be open-source and regularly updated, to account for the rapid evolution of labour markets, as well as the integration of new skills and occupations, such as the ones linked to AI or the green economy. Based on these considerations and discussions with our localised partners, ESCO appeared as the most appropriate taxonomy.

The Advantages of ESCO

Accessibility to job-seekers:

1. When it comes to user experience, one of our main concerns is to make sure that demographics who may be traditionally marginalised from labour markets et intermediation solutions can use our inclusive taxonomy. In some contexts, the mastery of official languages, such as English in South Africa, is limited. Compared to O*NET, ESCO relies on skill tags that are easier to understand.

2. ESCO contains a list of ‘similar titles’ for each occupation – making it easier to search for an occupation from a broad starting point. For instance, "data engineer" is saved as an alternative title for "data scientist", which ensures that one may use the platform by typing out "data engineer".

Highlighting job seekers' skills

1. Contrary to O*NET, ESCO includes "attitudes and values", which are soft-skills missing in O*NET. Therefore, it covers a broader range of skills that are sought after by employers.

2. ESCO is a European Commission project administered by the Directorate General Employment, Social Affairs and Inclusion. This ensures that the taxonomy is frequently updated and will be used by major actors in the long run. On the contrary, the last update of ISCO was made in 2008, and the last update of OFO happened in 2013. Frequent updates of ESCO allow it to include skills and occupations associated with major evolutions in labour markets worldwide, such as the greening of economies and, in the next few years, the predictable growing reliance and AI.

Compatibility with Compass

1. Technically, ESCO is designed to be used by third-party organisations. ESCO is specifically built for others to use and embed the taxonomy in their own technologies.

2. Other developing countries in Latin America have also adopted ESCO as their skills framework rather than ONET, and they are contributing to the open-source tools around ESCO which can be leveraged (for instance, ML-classification models which take in free-text job-titles and assign the ESCO occupations).

3. ESCO provides a Local API in addition to their web-based API. This means that Tabiya can host this API locally and adjust it as need be without depending on third-party performance when the volumes of requests are scaled up. This also gives Tabiya the ability to edit and adapt the framework freely.

Tabiya's work on the inclusive livelihoods taxonomy can be divided in two streams of work. First, making sure that the "seen" part of the pre-existing taxonomies such as ISCO and ESCO fit local contexts. Second, broadening existing taxonomies so that they include the unseen part of the economy, namely the activities that are typically not considered as productive and the skills that are associated with them. The first stream of work is thoroughly described for .

Motivation of the Methodology

The challenge is the following: Tabiya aims at creating an inclusive taxonomy of livelihoods, while ensuring that this taxonomy is compatible with existing ones such as ISCO or ESCO. Indeed, this allows to ensure that the inclusive taxonomy can be used by institutions that rely on widely-used taxonomies. Along with our partners, we chose to base our work on the European Skills, Competencies, Qualifications and Occupations (ESCO) taxonomy, for reasons highlighted here. Consequently, Tabiya's work entails evaluating the inclusiveness of the existing ESCO taxonomy, to then improve this inclusiveness.

This work's intellectual underpinning builds upon the "Counting Women's Work" literature, that aims at assigning a monetary value to the tasks done by women, especially in their households. Instead of focusing on monetary values, Tabiya aims to highlight the human capital gained from these tasks. Moreover, Tabiya's work covers all job-seekers, and not just women, although depending on local contexts they may de facto represent the majority of job-seekers.

Making Visible the "Unseen Economy"

For the “unseen economy,” i.e. activities outside of the specific SNA production boundary, we rely on an existing classification of all time-uses and tasks one may gain human capital from. Therefore, our works builds on the International Classification of Activities for Time Use Statistics (ICATUS) taxonomy, which lists all time uses someone may have throughout the day. ICATUS represents the internationally applicable classifications of activities that people engage in during their 24-hour days. Countries and regions, including South Africa, Ethiopia, and Ghana, have used either this ICATUS framework in their national data collection efforts.

ICATUS is made up of three levels. The first digit level of disaggregation (highest level of aggregation) is called the major division, the second digit level is called the division level, and the third digit level, which is the most granular level, is called the group level. We first start by comparing ICATUS with the System of National Account, to define the boundaries of the "seen" and the "unseen" economies.

Based on the System of National Account, we thus consider the following ICATUS major division to encompass the unseen economy:

Within these divisions, we exclude the following activities:

Matching unseen ICATUS activities to the ESCO framework

Once ICATUS activities have been associated with the "unseen" part of the economy, the challenge is to make sure the unseen part of our inclusive taxonomy follow the same structure as the existing ESCO taxonomy. Namely, we need to assign skills to the ICATUS activities forming the unseen part of the economy. Therefore, Tabiya conceptualises a framework that links the most granular level of unseen economy ICATUS activities (3-digit level) to a set of non-exhaustive candidate ESCO skills and knowledge tags per ICATUS Activity.

ICATUS 3 digit level activities comprise more specific activities that lie within ICATUS Divisions Three, Four, and Five. For example, “Preparing meals/snacks'' is a Group Level Activity that falls within the Division 3 of ICATUS, namely Unpaid domestic services for household and family members. In addition, at the Group Level, ICATUS includes a definition of each activity, and a non-exhaustive list of the tasks that each activity includes and does not include. Finally, ICATUS presents at least one example of each Group Level Activity.

The ESCO taxonomy is made up of a set of occupations at level five or lower that are derived from the four ISCO-08 occupations hierarchy. For the purpose of the Tabiya Framework, we use only ESCO occupations at level five or lower for our analysis, since this provides the desirable granularity and metadata to adequately link ESCO to ICATUS. The ESCO taxonomy provides a brief description of each occupation it comprises, the occupation’s alternative labels as well as access to regulatory information pertaining to the occupation. What makes the ESCO taxonomy particularly desirable are the skills and competences it assigns to each occupation at level five or lower. These skills and competences include a predetermined list of essential and optional skills and competences, and a list of essential and optional knowledge required to perform a given occupation.

ICATUS Group Level activities are used as one of the inputs for the matching procedure entailed in the Tabiya framework. ESCO occupations and skills, competences and knowledge (skills, competences and knowledge will be referred to as skills henceforth for simplicity) are used as the other input for the matching procedure. In particular, each ICATUS Group Level activity is manually matched to at most four ESCO occupations by a team of researchers. ICATUS, even at the Group Level, is more broad than ESCO occupations and so, to ensure comprehensiveness in the Tabiya matching procedure, up to four ESCO occupations were potentially matched to each ICATUS Group Level activity. Since the ESCO taxonomy assigns skills to occupations, these ESCO-assigned skills then form the initial list of candidate assigned skills per ICATUS Group Level activity.

The assignment of ESCO occupations to ICATUS activities is based on the relative similarity definitions between ICATUS Group Level activities, and ESCO occupations. Comparing definitions of ICATUS activities to ESCO occupations is relatively easy to do amongst a team of researchers, since activities and occupations are similar in nature, and in turn, in the manner that each of these are defined. Once at most four ESCO occupations are matched to each ICATUS Group Level activity by a team of at least two researchers, these matches are then checked by another independent researcher. Throughout this process, there were no instances where the independent researcher disputed an occupation without mutual agreement from the pair of researchers who initially made the match. This system is adopted to ensure that person-specific biases do not prevail during the matching process. This notwithstanding, while we endeavor to minimize biases and human error through this process, we cannot completely rule out the possibility that other (group level) biases and errors may have occurred throughout this process. Over time, as Tabiya accumulates more data and learns from this, the framework will use these learnings to improve and ultimately, not succumb to any currently prevailing biases and errors.

Once the ICATUS to ESCO occupations match is done and verified, the next stage of the Tabiya framework formulation can commence. By design, the ESCO taxonomy pre-assigns skills to each ESCO occupation at disaggregation level five and below. Thus, we exploit this structure for the Tabiya framework. In particular, once ICATUS activities are matched to ESCO Occupations, we adopt the pre-assigned ESCO skills as the first set of candidate skills assigned to each unseen economy ICATUS activity. Once duplicate skills are removed, each ICATUS activity now has a comprehensive list of skills assigned to it. The procedure is shown below:

Reducing the Candidate List of Skills

The taxonomy obtained following this method is hardly usable. Indeed, each ICATUS activity ended up being associated with numerous skills. In order to provide a manageable list of skills to job seekers on the Harambee platform 0.1, the team decided to reduce this tentative list of skills. Reducing this also allowed to addressed the transferability and signalling issues inherent to the unseen economy. To do so, an approach based on a panel of professionals involved in the South-African labour market and in intermediation instances (among which Harambee) was chosen. The panel approach is meant to gain context-specific insights from a wide range of stakeholders. This would make it possible to compensate holders' biases pertaining to the transferability and credibility of skills. In order to test this approach, we first implemented a panel at the scale of Harambee.

For the panel to happen efficiently (in a time limit of 5 hours for the pilot panel), the Tabiya team proceeded with a first round of skill selection in order to delete skills that were deemed obviously not consistent with the definition and description of ICATUS activities. To do so, the team followed 4 easily applicable rules.

Isolating Knowledges:

The ESCO classification distinguishes between "skills" - that all describe an action - and "knowledges". For instance, the occupation "cook" is associated with the skill "use cooking techniques" and the knowledge "cooking technique". As a first step, we chose to isolate knowledges and to mainly focus on skills. However, we chose to let the panel decide to keep certain knowledges when they brought important new informations. For instance, "common children's diseases" may be deemed an essentio

Deleting irrelevant skills:

In ICATUS, each activity is associated with a definition, and explicit list of tasks included in the activty, a list of tasks that are not included, and one or more examples. ESCO is built similarly. However, IACTUS activities are typically not only broader, but conceptually different from ESCO occupations. For instance, "Budgeting, planning, organizing duties and activities in the household" is not a formal sector activity that is included in ESCO. Therefore, to assign ESCO skills to this ICATUS activity, it was matched to "Office clerk", "Accountant", "Bed and breakfast operator", three ESCO occupations that encompass all tasks involved in "Budgeting, planning, organizing duties and activities in the household". The issue is that they encompass more. For instance, because of "bed and breakfast operator", the skill "serve beverages" endded up being associated to "Budgeting, planning, organizing duties and activities in the household", even though serving atsks are not included in the description of the ICATUS activity. When we came across such cases, we decided to exclude the skills that had been unduly associated to the ICATUS activities.

Deleting skills that are deemed too formal: some of the skills associated with ESCO occupations directly refer to situations or tasks only imaginable in the seen economy, whether it is formal or informal. For instance, "maintain customer service" cannot be appropriately associated with "serving meals and snacks", as it describes serving meals and snacks to one's own children/family members.

Deleting redundant skills: ESCO occupations are typically associated with numerous skills, and each ICATUS activity was associated with mutltiple ESCO occupations by Tabiya's team. The lists of skills from each ESCO activity was added to a list of skills for each ICATUS actity, with deletion of skills that appear multiple time (where we only kept one appearance). Therefore, the temptative lists of skills associated to each ICATUS activty contain skills tht may be deemed redundant. For instance, "Outdoor cleaning" was associated with both "prune plants" and "prune hedges and trees". We considered that asociating both skills to "outdoor cleaning" did not bring new information.

This makes up the generalisable unseen economy framework methodology. From this candidate list of skills associated with each unseen economy ICATUS activity, differing contexts can begin to localize this framework.