How we are leveraging LinkedIn data to explore labour dynamics — and what we find when we zoom in
by Shumin Liu
The World Bank Group recently published data extracted from the employment records from 2015 to 2018 in LinkedIn. I was curious to find out whether this new dataset could shed any light into the labour dynamics in the region, and how this might compare to traditional statistics.
I decided to take Thailand as an example. Not surprisingly, the #futureofwork is an area that has already attracted the attention of hundreds of researchers. So what was my process? Through official data, I could see that researchers were already able to identify mismatches between the educational level attained and requirements of a job in the Thai labour market. Yet it was important to layer in new context-specific data too: like the kind here where researchers from Chulalongkorn University collected data from Google Trends and Thai job portals to (1)understand dynamics in job vacancies, (2) gaps in educational or vocational degree between the labour supply and employment demand, plus (3) the mismatch of wage and productivity.
I then returned to the World Bank datasets- using the following questions to guide my analysis:
- What are the employment trends across industries in general?
- What does labour connectivity between Thailand and other countries look like?
- How does labour connectivity impact various industries?
- Which skills are most correlated with labour connectivity?
As the World Bank makes clear, and as any researcher knows, there are limitations to ‘the data’. In this case, the datasets might not depict a comprehensive picture of the labour market for many reasons, yet especially due to assumptions around internet access behaviour. Especially as we consider that whoever is updating their online employment information doesn’t necessarily represent the general population. However, to reduce bias and increase representativeness, the datasets were validated based on the distribution of age, sex, and industry with ILO data. (Something I am keeping in mind when interpreting the results).
- Which employment trends did I discover across the multiple industries?
The visualization below shows the trends in employment growth in industries across six sectors in 2018, where the industries of outsourcing/offshoring, venture capital & private equity, and online media were having the highest employment growths; while the industries of computer games, writing & editing, and photography were impacted least by employment growth.
If you are also curious about the shift in employment growth among different industries and sectors over time, you may find more insights from the visualization below. The predominant orange areas indicate that these industries were consistently attracting labour. However, employment growth was lagging behind in some industries, reflected by the negative rates in green. The bigger the size of a bubble, the greater effect of the positive/negative employment growth it represents.
So how might we interpret the employment rate here? The negative employment rate in some industries may imply that people are less inclined to work there or the demand for manpower is reducing. However, there could be another possibility that employers are struggling to attract skilled workers since the existing pool of labour cannot meet the rising demand in the industries as a new requirement for skills emerges. If that is the case, is it a hint for us to drill down to those industries and identify the skills in need?
One option to dig further could be to corroborate this research with the job vacancy data to spot the industries with large employment gap between demand and supply, and starting from there, to further investigate the skills in high demand through getting qualitative data from employers in the industries and employees themselves.
2. What does labour connectivity between Thailand and other countries look like?
It is estimated that there are 4 to 5 million international labour migrants working in Thailand. However, nearly 1.5 million migrant workers in Thailand are undocumented. And in 2016, 114,437 Thai nationals departed for work abroad. Can we gather any insights into these movements using the LinkedIn data?
When a LinkedIn user updates the working location from one country to another, it is considered as a labour migration. The labour connectivity maps below illustrate the flows of migrant workers between Thailand and other countries in 2015 and 2018. The circle filled with orange colour denotes the inflow from country X to Thailand; while the circle filled with green colour means outflow from Thailand to country X. A larger circle implies a stronger effect of inflows/outflows.
In 2018, Thailand attracted most of the labour from India, the US, and the UK; meanwhile, a majority of labour were transferred from Thailand to Germany, Myanmar, and Singapore. From 2015 to 2018, the biggest changes in net flow effect on Thailand had been observed in the US (still inflow effect, but reducing), France (changing from inflow to outflow), and Myanmar (still outflow effect, but reducing).
Interestingly, the labour outflow from Thailand to Myanmar was more significant than the labour inflow from Myanmar to Thailand. However, among migrant workers in Thailand, Myanmar is one of the biggest ‘sending countries’, accounting for 69% of the total number of low-skilled migrant workers holding work permits in 2017. The seemingly conflicting data is signifying that the composition of low-skilled migrant workers might not be proportionally represented on LinkedIn.
If we are interested in exploring the future of skills and the weak signals, LinkedIn data might be able to provide some good indicators. But if we think from the perspective of an inclusive future for society, we should have the general public’s footprint in the landscape.
3. How does labour connectivity impact various industries?
Labour migration also tells us something about change underway in terms of industries at large— in this sense, net industry migration is recorded and represented by the net flows between Thailand and other countries in a certain industry at the time of migration. The visualization below describes the effect of the net industry migration in Thailand in 2018, where the labels around the periphery denote the name of industries which are grouped by sections in different colours; the filled bubbles in the nodes stand for the effect of inflow, while the hollow bubbles imply the effect of outflow.
In 2018, Thailand attracted most labour to the industries of international affairs, renewables & environment, and sports, while the effect of labour outflow was more outstanding in the industries of defense & space, maritime, professional training & coaching.
Related to what has been happening since 2018, the Smart Visa programme, introduced by the Royal Thai Government, is designed to attract highly skilled professionals and investors to work and invest in 13 target industries. To what extent the Smart Visa programme is effective in enabling knowledge exchange through solicitation of highly skilled labour and creating an impact on the growth of those industries? For example, the industry of environmental management and renewable energy is one of the target industries, and it is also one of the strongest industries appealing to migrant workers as indicated in the analysis. Is it a coincidence or a corroboration?
Questions for future analysis that might help inform, include for example: how might government policy include real-time data from online platforms such as LinkedIn or job boards to identify emerging trends and gaps in the market for better targeting of priority industries? Can we use this data to spot new emerging professions which are not captured by programmes like the Smart Visa example?
4. Which skills are most correlated with labour connectivity?
When migrant workers are carrying the existing skills or obtaining new skills from a new job opportunity — it induces the effect of skill migration in terms of net flows between sending countries and recipient countries. The visualizations below illustrate the effects of skill inflows (in orange) and outflows (in green) in Thailand’s 2018 labour market by the following five groups (defined by the World Bank Group — LinkedIn data): specialized industry skills, business skills, tech skills, disruptive tech skills, and soft skills.
In terms of the skill breakdown, the group of specialized industry skills is the biggest in both graphs in terms of variety, composing of a diverse spectrum of technical skills. Within this group, we can see the skills inflow effect is most predominantly related to physiology, sociology, and organic chemistry. Under the group of business skills, the top three skills transferring to Thailand in 2018 were Capital Market, Auditing, and Financial Accounting.
Thailand was attracting labour with disruptive tech skills(defined by the WBG-LinkedIn datasets), including genetic engineering, development tools, human-computer interaction, artificial intelligence, data science, material science and others. And no outflow of disruptive tech skills was observed in 2018. There is a separate group of tech skills, where game development, animation, and web development are the most dominant ones contributing to skill inflows.
It is apparent that there is a huge demand for skills such as those under the group of disruptive tech, but can velocity of supply for those skills go faster than the one in the demand side? It will normally take four years for bachelor students to graduate, but industries are craning their neck for those skills now.
Can the traditional way of obtaining skills through formal education channels feed the growing demand for new skills? Can universities design programmes for skills cultivation rather than just a degree? Companies can use high paid salaries and benefits to allure talents as a short term solution. But it is not effectively tackling the problem rooted in the existing workforce pool. How can we keep reskilling ourselves and enable an active learning space to stay relevant?
Unlike hard skills which are normally related to technical expertise in certain domains, soft skills are more about personality, social engagement, and thinking. Soft skills such as time management, problem-solving, leadership, oral communication, negotiation, and active learning are emerging in Thailand, while globally, the top five soft skills are creativity, persuasion, collaboration, adaptability, and time management, according to LinkedIn.
(Let’s also ask) which skills are not included?
When the dataset of skill migration in Thailand is compared with the one in other countries (presented in the datasets) among Asia and the Pacific region, some records are missing in the Thai dataset. (These are shown in the visualization below.) According to the methodology in the report, to reduce bias caused by small sample size, the migration of skill is recorded with a value only if it has more than 50 observations.
So the missing data doesn’t mean migrant workers in Thailand are not equipped with these skills, but those skills are rare to find on LinkedIn for various reasons. It might be sending a weak signal to reveal the emerging skills with high demand, while supply is staggering to catch up. For example, the skill of data-driven decision making is not being captured in the dataset, but it is encouraged by the Thai Government to embed such capacity in civil service for Thailand 4.0. And according to LinkedIn, cloud computing is the top hard skills companies need most, but it is absent in the Thai dataset.
Some open questions for you-
The data exploration through the lens of the LinkedIn professional network provides a new angle to unveil the labour dynamics across countries, industries and skills. And most importantly, it helps surface the patterns from complexities, inspiring us to think deeper, immerse ourselves with the reality, and connect them to make sense of why it is or is not happening in a certain way so that we can make better decisions. Like a ripple brought by a drop of water, it opens up more questions for further exploration.
- Stretch wider. What is the signal of demand in the labour market? The labour dynamics is a concept of mutuality between supply and demand, mirrored by the relationship between employees and employers. What skills are increasingly demanded by the employers? Though labour getting a job and having a presence in the market may indirectly imply their abilities to meet the demand in the market, it relies on the assumption that employers can always hire the staff they want. However, 45% of employers said they couldn’t find the people with the right skills. There are gaps in skills between supply and demand requiring policymakers, education providers, employers, and employees to bridge them collectively. So we are also interested in sourcing data from various data sources such as job boards to understand the demand side and compare them with the supply side such as resume posts, university website and online training platforms to spot the gaps.
- Zoom in. How is the domestic labour dynamics in Thailand? Apart from making sense of the labour landscape at the international level, governments also have a strong interest and need in understanding the domestic labour dynamics. Where are the domestic data represented? We are thinking whether we can apply a similar approach to excavate data on the local professional networking platforms like LinkedIn and link it to official data in governments and the database from ILO, connecting the fragmented data residing in different systems.
- Dive deeper. How to harness data to generate actionable intelligence? First, we need more high-quality evidence. The LinkedIn data for this analysis has advantages in terms of volume, velocity, and value. But it’s limitations associated with its representativeness for the population and it’s tightness to specific policy questions such as the landscape of skills in domestic labour market manifests the need for more evidence from different sources to be truly informed. Thus, we should also unleash the qualitative data driven by people behind them, connect the data silos to map out a comprehensive picture, and set them as the foundation for generating actionable intelligence.
This is just the start of the journey of data exploration! And we are looking forward to having you on board. Having said that, we would love to know your thoughts and questions for further discussion and inspiration. Let us know your thoughts on twitter @ricap_undp
With special thanks to Giulio Quaggiotto, Alex Oprunenco, Courtney Savie Lawrence, and Diastika Rahwidiati for their inputs, suggestions, and encouragement.