Contact Us  |  Search  
The National Academies of Sciences, Engineering and Medicine
Partnerships for Enhanced Engagement in Research
Development, Security, and Cooperation
Policy and Global Affairs
Home About Us For Applicants For Grant Recipients Funded Projects Email Updates
Cycle 5 (2015 Deadline)

Data Science for Improved Education and Employability in Morocco

PI: Ghita Mezzour (, International University of Rabat
U.S. Partner: Kathleen Carley, Carnegie Mellon University
Project Dates: February 2017 - January 2020 

Project Overview:

The mismatch between the job supply and demand creates major social, political, and economic problems in Morocco. Every year, many graduates are unable to find jobs, and the resulting youth unemployment causes major social and political tensions. Paradoxically, at the same time, employers are unable to find candidates with the required skills, and this skills shortage results in missed economic opportunities for the country. Despite the importance of the studying skill mismatch in Morocco, the topic attracts very limited attention in the literature. Moreover, there is a lack of large data sets that researchers can use to systematically study the issue and identify effective interventions to alleviate it.

5-648 Mezzour
The PI and her team at the USAID career center in Tangier. Photo courtesy of Dr. Mezzour
The goal of this project is to measure the skill mismatch in Morocco and identify measures to align university training with the job market. More specifically, these researchers will collect and analyze multiple large data sets about higher education and the job market in Morocco.

They will build profiles of university graduates and job openings in Morocco and identify areas of misalignment between the two. They will also interview human resources staff from multiple organizations to learn about their concerns in more detail. Finally, they will collect traditional and social media discussions about higher education and jobs in Morocco in order to learn about the general population’s concerns about the topic.

This project should lead to advances in both education and computer science, and the analysis to be conducted should yield deep and novel insight about areas of mismatch between higher education and the job market in Morocco.

Summary of Recent Events

In the quarter ending March 2018, Dr. Mezzour and her team finalized their work on analyzing the needs of the automobile sector in Morocco. The analysis is based on around 8000 job ads collected from 22 Moroccan recruitment websites during the period January 2017 to January 2018. They found that most automobile jobs are located in Casablanca Settat region, followed by Rabat Sale Kenitra Region, then Tanger Tetouan Al Houcima region. They also found that 40% of jobs require 2 years of higher education and that 23% of jobs require 5 years of higher education and that 14% of jobs require no higher education. Finally, their results indicate that 53% of jobs require 1-2 years of experience and that 20% of jobs require less than 1 year of experience. Based on this work, they submitted a paper to the IEEE Conference on Computer Science and Education.

With the rise of automation and artificial intelligence, soft skills are increasingly important as more and more jobs are replaced by robots and intelligent machines. Extracting soft skills needs a different way of processing since recruiters use a different term designing the same skill (such as teamwork and collaboration), which makes them hard to measure. They are currently developing algorithms to extract and standardize soft skills listed in job ads. They combine and use multiple algorithms and data sets including word2vec, the Newman clustering algorithms, stemming algorithm, DBpedia, job ads from Moroccan and French job recruitment websites in order to build a hierarchical taxonomy of soft skills. The combination of these techniques gives a taxonomy that appear more precise that approaches suggested in prior work and that rely on only one or a subset of these techniques. Finally, they used the techniques they developed to analyze soft skills required by the offshore sector in Morocco. Their preliminary results indicate Business Process Operation (BPO) jobs put more emphasis on soft skills that Information Technology Operation (ITO) jobs. Moreover, the top soft skills required by BPO jobs are “dynamic”, “commercial”, and “communication”. On the other hand, the top soft skills required by ITO
jobs are “communication”, “rigorous” and “respectful”.

The team has continued working on analyzing collaborations and relationships of employment stakeholders based on survey data collected at the opening days of USAID career centers in Tangier, Casablanca and Marrakech. They find that most stakeholders know each other and express a desire to collaborate with each other. However, these stakeholders do not tend to collaborate on employment related matters. For example, their results indicate little exchange of information about the job market needs between the private sector and universities in the 3 cities. Youths do not tend to exchange information with other stakeholders in Tangier and Casablanca. On the other hand, youth in Marrakech seem to be well connected to other stakeholders. They are currently finalizing a journal paper about that analysis.

In the next 3-6 months, the PI and her team will continue working on extracting soft skills from job ads. Extracting soft skills is more challenging than extracting hard skills because the same soft skill can be referred to using different terms. They are currently investigating the use of Dbpedia (a project that aims at extracting structured content from Wikipedia) and Word2Vec (a powerful text mining algorithm that uses deep learning). They are refining their analysis of the needs of automobile sector in Morocco.

They will also be conducting a deeper analysis of the survey data from 2016 and 2017 about the relationships between employment stakeholders in Morocco. They plan to submit the results of that analysis to a journal. In addition to surveys, they are investigating the use of other data sources such as reports, social media and stakeholder websites to gain insight into stakeholders’ mental models. Analyzing stakeholders’ mental models could help explain why some stakeholders do not collaborate. For example, it could be that two stakeholders do not collaborate because their views of the problems are very different.

Back to PEER Cycle 5 Grant Recipients