This post was originally published on this site.
The spread of the Internet, information and communications technology, and artificial intelligence (AI) has had a significant impact on job recruitment and job applications. These technologies were and are expected to reduce job posting and application costs, enhance the efficiency and quality of matching companies with job seekers, and improve employment and unemployment rates.
However, analyses using data from the United States, Germany, and other countries up to the mid-2000s found that online job search failed to reduce the unemployment rate more effectively than traditional job search methods. It was found that online job search was more effective in improving the quality of job posting matching for employees who wanted to change jobs than for the unemployed.
As online job search became more widespread and took root, studies began to show its positive impact on unemployment. Peter Kuhn at the University of California, Santa Barbara, and his colleagues used data from the United States in the late 2000s to analyze the difference between those who successfully used online job search methods and those who did not. The analysis showed that the average period of unemployment in the former group was about 25% shorter than in the latter.
Nicole GĂĽrzgen at the University of Regensburg in Germany and others used data from Germany in the late 2000s to focus on the impact that was seen on the use of online job searches based on differences in broadband adoption in different regions. Their research demonstrated that adopting the technology increased re-employment rates of unemployed people who were actively looking for jobs.
Among more recent studies is a 2023 paper by Manudeep Bhuller at the University of Oslo in Norway and others. Using data from Norway up to the mid-2010s, it showed that broadband penetration led to a 9% fall in the duration of vacancies and a 13% drop in unsuccessful job postings on the employer side. On the job seeker side, the employment rate increased by 2.4% and wages just after hiring increased by 6%. Using a general equilibrium model, the paper also showed that such broadband penetration reduced the equilibrium unemployment rate by 14%.
◆◆◆
While online job searching has clear benefits, there are also some clear challenges. One issue noted by David Auter at the Massachusetts Institute of Technology (MIT) in the early 2000s is the problem of “adverse selection.” This refers to a significant decrease in the cost of job search and application for job seekers leading to an increase in the number of low-quality job applicants, forcing employers to spend more time and labor to screen them.
According to a 2021 paper by GĂĽrzgen et al., online job search under broadband penetration in Germany increased both the number of job applications and the proportion of inappropriate applicants, while failing to improve the job search quality regarding the stability of matching and wages.
Recent studies have shifted focus from only examining the frequency of online job searches, to the effectiveness of personalized advice tailored to the individual characteristics of job seekers.
A 2022 paper by Michele Belot and colleagues at Cornell University found that long-term unemployed individuals in the UK are more likely to stably find jobs and gain wage increases above a certain level when they are given automatically-generated advice on alternative jobs when they search for jobs online. The likeliness was more pronounced for those who had remained unemployed longer.
They explained that the positive impacts were due to the relevance of the advice on alternative jobs to actual job postings, which facilitated specific understanding among job seekers. On the other hand, AĂŻcha Ben Dhia at MIT and colleagues used large-scale experimental data from French public job placement agencies and found that personalized job search advice was not necessarily effective. Belot et al. attributed this result to the fact that the personalized advice was overly abstract, even though it was personalized.
A 2022 paper by Steffen Altmann at the Danish University of Copenhagen and others used large-scale experimental data from public job placement agencies in Denmark. It found that advice on specific vacancies personalized for job seekers and on alternative jobs had an impact that was 8-9% more positive on subsequent working hours and wages, while exerting different effects on job seekers’ behavior.
Job seekers who were given information on specific vacancies alone looked for jobs in a narrow range of occupations, such as those in which they had experience. On the other hand, those who received advice on alternative occupations looked for jobs in other occupations that were different from those in which they had experience.
However, for those who simultaneously received advice on both specific vacancies and alternative occupations, there was no definitive effect because their effects were offset by each other. The paper also found a negative externality effect in which when the proportion of people receiving advice in the same area is quite high, competition increases for a single vacancy, making it more difficult to achieve positive outcomes.
◆◆◆
Needless to say, AI is used for personalized and automatically generated advice for job seekers, but is it also increasingly being utilized for almost all processes in matching job postings and job seekers.
A 2023 report by the Organization for Economic Cooperation and Development (OECD) provides a comprehensive look at how AI is used by recruiting companies which are involved in such labor market matching, including by job placement agencies that mediate job postings and applications, and by online job boards and platforms.
Specifically, areas where AI is applied range from creating job descriptions, identifying job applicants, analyzing resumes, as chatbots, for interview scheduling, and in shortlist creation tools to analyzing faces and voices during interviews. It has been pointed out that many AI tools have helped to improve efficiency and reduce costs.
On the other hand, some barriers to the expansion of the use of AI tools have been noted, including a lack of organizational and personnel readiness to use AI and concerns about the AI predictions and recommendations (robustness, bias, explainability). Problems for job posters are those related to their privacy and transparency about AI use, which arise when information is collected, particularly through social networking services. While these problems need to be tackled carefully, they should not be used as an excuse for avoiding the introduction and utilization of AI tools.
The OECD report cites VDAB, a public job placement agency in Flanders, Belgium, as the world’s most advanced AI user. Most notably, VDAB AI uses information about individual job seekers to predict the likelihood of long-term unemployment, helping caseworkers to select job seekers who should be given priority (see table).
Sources: Prepared by the author from OECD (2023) “Artificial Intelligence and Labour Market Matching”, OECD SOCIAL, EMPLOYMENT AND MIGRATION WORKING PAPERS No. 284
Finally, how do the Hello Work public job placement offices in Japan compare? In recent years, they have enhanced their systems to support online services, allowing job seekers and employers to use the Hello Work Internet Service for online job postings, applications, and placement. Moving forward, AI utilization can be seen as a major future challenge and area of concentration.
>> Original text in Japanese
* Translated by RIETI.
September 17, 2024 Nihon Keizai Shimbun