Big data makes educational institutes more responsive
One of the biggest challenges for universities and colleges is to adjust their curricula and teaching methods to the quickly changing market demand. Those successfully responding to the challenges will allow their students to find good jobs upon graduation. However, there is some worrying evidence indicating the curricula taught in universities and colleges are outdated and not aligned with employers' demands.
According to TDRI research, only 24% of college and university graduates in science, technology, engineering and math (STEM) subjects are employed in the fields of their studies. In other words, three out of four graduates are not using the knowledge they have learnt from colleges to the full. Perhaps they are trained with outdated skill sets.
The main culprit here is the disconnection between the education sector and the job market. Lacking the information regarding market demand, most universities and colleges are virtually blind to the types of skill sets employers seek from their graduates. This problem greatly undermines the ability of the education sector to respond to the private sector's needs and therefore results in a considerable number of graduates with mismatched skill sets.
Nuthasid Rukkiatwong is a researcher at the Thailand Development Research Institute (TDRI). Policy analyses from the TDRI appear in the Bangkok Post on alternate Wednesdays.
The key to solve this skill mismatch is to have an effective system whereby accurate information regarding employers' demand for skills can be gathered and disseminated to educational institutes. However, the current practice of producing such information is largely inadequate. The National Statistical Office has long conducted labour demand surveys whereby questionnaires are sent to randomly selected firms. There are at least five shortcomings with this approach.
Firstly, the survey process is very costly. As a result, the sample size tends to be small and cannot represent the overall picture. Secondly, it has a very long lead time to conduct the survey, analyse and disseminate the result. By the time the result is disseminated, the market has moved on.
Thirdly, the survey is inherently inaccurate as it depends on the firms' willingness to respond. Academically speaking, even a well-designed random survey can find it hard to escape the problem of "selection bias". Fourthly, since the survey is based on small sampling, many responses have to be grouped together so that the data can be blown up to obtain the population estimates. For example, a demand survey in 2013 grouped web application developers and programmers into the same category even though their skill profiles are quite different.
Finally, obtaining data with detailed skill sets requires prohibitively lengthy questionnaires, rendering such attempts hopelessly impractical.
Instead of using surveys, we adopt an "online job-posting analysis" approach by making use of publicly available data from classified job ads posted online. This approach has several advantages over traditional survey methods.
Firstly, the data from online job posting can be collected in real-time with little cost. Secondly, information from job postings better reflects actual market demand as any employers seeking workers have to announce their job openings. This is especially true for white-collar jobs, which are mostly posted online.
Thirdly, as there is no sampling involved, it eliminates the need to group data into pre-defined categories. Finally, job postings contain very detailed descriptions of skill sets unobtainable using traditional survey.
In collaboration with the National Electronics and Computer Technology Centre (Nectec), our first prototype collected 100,000 job postings over last February from five online job boards. The data, which is basically text, is then transformed into a database format using natural language processing software, and is ready for analysis.
To demonstrate the idea, we analysed the required skill profiles for software/website developers from 2,712 unique job postings. We found that the necessary skill sets can be grouped into five categories, which are: 1) programming languages, among which Java and .NET are most common, 2) database, 3) client-side scripting languages, 4) knowledge in user interface/experience design, and 5) basic skills including English, management skill and communication skill.
Our further analysis reveals that while most employers require applicants with .NET programming skills, few computer science programs are training their students to use it. We also found that 88% of job postings require applicants to have previous work experiences. As a result, newly minted graduates will have difficulties filling the jobs unless their universities or colleges have solid apprenticeship schemes or extended traineeships that can be partially qualified as "work experience".
Educational institutes have long been blind to market demand. Now big data from the online job market has enabled them to be responsive to demand. Let us hope this opportunity is not wasted.