Can LinkedIn data help us measure the glass ceiling?

March 21, 2023

Gender disparities persist globally, particularly in positions of power and higher pay, where women face cultural and professional barriers to advancement, which are often referred to as “the glass ceiling”. For example, the latest WEF Global Gender Gap report estimates that it would take 132 years to achieve overall gender equality and around 150 years to achieve equal political representation of women and men across the globe. 

Inspired by trailblazing 2021 research paper from Oxford University, UNDP Serbia Accelerator Lab decided to use LinkedIn Data to test how thick of a “glass ceiling” women are facing in Serbia, Bosnia and Herzegovina, North Macedonia, Montenegro, Albania,  Kosovo[1], especially when it comes to the IT sector. We also looked at LinkedIn Data from Georgia and Lithuania, upper middle-income countries with similar population size as the mentioned 6 countries and territories, also known for high representation of Women in science, technology, engineering and mathematics (STEM). 

We found that in the Western Balkans women are underrepresented at the very top of professional hierarchies and company ownership. While the IT sector in the region is developing rapidly, offering many opportunities for economic and social development, it is not uncommon for women to remain “trapped” in lower status positions due to systemic, unobservable biases. 

In this blog we focus on our methodology and the results of our analyses from Serbia.




Our methodology 

Since the official data describing individuals on the labour market is not publicly available, we performed our analysis by using the Linkedin Ads-targeting data. This tool enables companies to advertise new positions by targeting individuals who comply with specific criteria, for example regarding to education or years of experience in a specific industry. We used the Linkedin Ad-targeting data from November 2022 for all our target countries and territories.  

In Serbia, our analysis covered 890 thousand LinkedIn profiles (450 thousand women and 440 thousand men), which is roughly 25% - 30% of economically active population. 

Following the previously mentioned Oxford University research, we first used the LinkedIn Ad-targeting data to measure the Gender Gap Index (GGI), which shows the representation of women, in the overall economy and particularly in the IT sector. 

GGI is calculated as a ratio of the number of women over the number of men with the same specific characteristic, such as age group (e.g. 25-35), level of education (e.g. MSc) or the level of position they hold (e.g. Manager) in the same field of economy (e.g. IT). When calculated this way, GGI above 1 indicates greater representation of women, while lower values indicate more men being in a specific category. 

In order to understand how much various attributes, such as age, level of education, or seniority affect the GGIwe built a predictive machine learning model, based on Random Forest algorithm. When asked to quantify how much each individual attribute contributed to representation of women the machine learning model showed us that in the Western Balkans most differences in GGI values are predicted by the field (IT or not), followed by the seniority, and by the country/territory (economy).

What we learned about representation of women in the Serbian labor market

Across the Serbian LinkedIn, in all industries there is a major fall in the number of women in higher age groups indicating decreased participation in the labor market after the age of 35 (or lack of activity on LinkedIn), as GGI falls below 1 (figure 3).



In terms of representation at higher levels the results are even more striking. Women tend to be underrepresented in the C-suite positions and as owners, despite the fact that they are more educated than men in Serbia (figure 4). 


The results are similar in the IT sector, despite Serbia being more equitable in terms of women’s IT employment and ICT education compared to the EU. One nugget of good news is that women’s participation in the younger age groups is higher, but interestingly, representation of women above 55 is negligible. The analysis also shows the small number of women holding C-suite positions and owning IT companies in Serbia (graph 1).


How can this data be used? 

LinkedIn is a useful tool for conducting professional gender gap analysis due to its wealth of professional profiles and demographic information. However, it is important to note that not all LinkedIn users provide complete or accurate information on their profiles, so the results of an analysis based on LinkedIn data should be taken with a grain of caution. Furthermore, there are biases in LinkedIn's user base, such as underrepresentation of certain groups, that may impact the accuracy of analysis. 

This analysis aimed to provide estimates of observable gender gaps and it does not indicate nor suggest the reasons for the gap. Nevertheless, it allows us to pinpoint where the gaps are, which gaps need to be further analyzed and, if they do show discrimination, remedied. 

The main benefit of using LinkedIn data is that it can complement more in depth, but costly and time-consuming data gathering approaches, and allows “nowcasing”  gender gaps across regions, sectors and labor market characteristics (age, education etc.). 

As noted in the Oxford University research, LinkedIn data shows strong positive correlations with the International Labor Organization’s ground truth professional gender gaps, while the results for the Serbian IT sector align with previous local research

Given the history of women facing the professional glass ceiling, it is good to be able to have a tool to track how various policies and economic shocks affect the gender gap. For example, the recent lay-offs in the IT sector seem to have worsened the representation of women. Using easily available labor force data, such as the one from LinkedIn, can help us nowcast how women’s representation has been affected and devise appropriate policies. 

What’s next?

Previously, social network data provided valuable insights and improved decision making in a wide range of areas such as healthcareeconomic development and disaster relief. In Serbia, UNDP Accelerator Lab used it to understand the country’s changing demographics

We plan to replicate our analysis for other countries and territories for which we gathered the data. This data can help us benchmark the representation of women between similar economies, and see which interventions (for example inclusive hiring, promotion of STEM for women) can contribute to more gender equality in the labour market. 

More importantly, we plan to use the insights we gained throughout our women in STEM initiative in Serbia, which aims to contribute to building knowledge-based economy that offers equal opportunities for carrier development for both men and women. While it is heartening to see more women in ICT in Serbia, especially in junior roles, it is necessary to track their progress and ensure they are not affected by the leaky pipeline phenomenon which we observed in previous research

Finally, we think that the ability to have up to date data on such a critical issue like the gender gaps at your fingertips can eventually help shatter the glass ceiling and help women unleash their potential in the economy. 

More details about the results of our research can be found here:

[1] References to Kosovo shall be understood to be in the context of Security Council resolution 1244 (1999).