The Great Hesitation in Singapore Labor Markets

Sir, my contribution to today’s debate is straightforward: I argue that, if our objective is to protect our workforce from job losses that could result from an economywide embrace of artificial intelligence (AI), then our efforts today should mainly be directed toward policies that prom0te new hiring, rather than those that focus on either reducing displacement, or pushing for retraining.

The explanation is simple: the evidence shows that, to date, job displacement of current workers due to AI has been modest and localized, whereas the hiring slowdown of new workers is already evident, and even likely to accelerate.

Hiring slowdowns dominate job displacement in the data

In principle, the consequences of AI could bite on two ends of the labor market. It could reduce the incentives for firms to hire, thereby lowering the number of new job openings. Or it could raise the frequency of firing by companies, or induce workers to quit.

Hiring slowdowns result when AI tools allow a company to cheaply and effectively replace job functions that they previously needed to hire a human worker for. This is especially the case for the entry level, since the lack of experience among new graduates, and the relatively straightforward grunt work new hires are generally tasked to do, make this group less valuable and more replaceable by a machine.

But as many observers—including members of this House—have already pointed out, this is a chicken-and-egg problem. If we do not absorb new workers into our corporations, we surely cannot expect them to gain the necessary experience and job-specific skills that would make them valuable as mid-career professionals.

Job displacement occurs when AI tools reveal that certain roles are no longer needed, as they can be well-replicated by AI. Tasks that used to be done by a human are replaced, and if there is nowhere else in the firm to reassign the erstwhile worker to (or if the individual is too costly), the person is let go.

On the positive side, AI could open up new opportunities for displaced workers to pursue a career elsewhere, either because they gain AI-related skills that make them more valuable in the marketplace, or because they can now go into business for themselves.

For now, however, the message from studies worldwide is clear: while there has been little evidence of displacement thus far, there are more ample signs that there has been a decline in new hiring. This has been the case for AI-exposed sectors following the advent of generative AI, such as ChatGPT, and is likely to accelerate as agentic AI matures. Prospects are especially perilous for early-career, entry-level workers. Even when hired, such workers tend to receive lower salaries.

The reasons for this are intuitive. AI mainly substitutes for mechanical, repeatable, and well-defined tasks, which are mostly performed by junior employees. Firms still value the maturity and experience of senior employees, and by and large, would rather skimp on hiring and reallocate their otherwise loyal staff, rather than giving them the boot.

This is the case in Singapore, too

These trends are also visible in our local labor markets. Thus far, AI does yet to contribute much to job displacement here. MOM’s latest Labor Market report reveals that, since 2023, overall retrenchments have remained stable, and the unemployment rate of around 2 percent hasn’t budged much since 2022. In response to questions about the PMET sector, SPS for Manpower Shawn Huang also pointed out that retrenchments among PMETs in AI-exposed sectors—such as finance and infocomm—remained low, numbering 960 in the final quarter of last year.

Moreover, SPS Huang also pointed to the large number of vacancies—around 10 times higher—in these sectors over the same period. This could be interpreted as a healthy, robust labor market for new hires, but I will caution against it. This is because, as any jobseeker will tell you, an opening does not a job make. These jobs have to be filled, ideally by Singaporeans who are looking for jobs.

Here is where the picture is less encouraging. The latest Graduate Employment Survey shows a drop in the share of graduates that managed to land a job in almost every single field of study, with around 1 out of every 4 graduates unable to secure full-time employment. In response, Minister Desmond Lee pointed to how the decline was due to the post-pandemic hiring surge, and that what we’re seeing is simply a mean reversion to the earlier trend.

I am less sanguine. Based on my calculations, while it is true that pre-pandemic permanent employment among graduates averaged around 70 percent in 2020, it was closer to 85 percent a decade ago, which is significantly better. The youth unemployment rate, while lower in the 2020s than in the prior two decades—and, admittedly, elsewhere in the world—has also steadily inched up since 2022, by around a percentage point. On their part, some employers also appear more tentative and reluctant to onboard workers, although these appear to be the minority for now.

Moreover, this pessimistic picture also masks more troubling pathologies. Many Singaporean workers may have had to content themselves with unattractive opportunities, just to make ends meet. Such mismatches are not well-captured in the aggregate data.

Think of the graduate with an advanced degree from a local university, but nevertheless still felt compelled to work in food delivery. Or the student who spent years abroad at a top university, but has been repeatedly rejected by employers after coming home. Or the 50-year-old manager who got retrenched, floundered unsuccessfully despite applying to every job he was qualified for, before eventually resorting to driving a private hire car, just so that he can feed his family.

Coupled with the global climate of economic policy uncertainty, we appear set for a so-called “Great Hesitation” in hiring in our local labor markets, similar to what has been observed elsewhere in the world.

Labor policies targeting jobs should focus on the hiring end of the market

Now, if our goal is, as the title of the motion suggests, to avoid jobless growth, then it follows that we should prioritize policies that target the hiring end of the labor market. Let me suggest a few.

First, we can improve the incentives for companies to hire fresh graduates. As I shared in a long cut for this year’s Committee of Supply, this would call for expanding the existing GRaduate Industry Traineeship (GRIT) program, to national-level, cross-sectoral national internship initiative. Young workers should be free to apply their SkilllsFuture credits toward paid apprenticeship and internship programs with companies willing to take them on. Corporations—especially small and medium enterprises (SMEs)—should also be able to submit credible proposals for in-house, on-the-job training to MOM, which will offset their costs of taking on these trainees, drawing on the SkillsFuture Enterprise Credit and other subsidies already earmarked for businesses. My honorable friend, Gerald Giam, has also proposed an AI Mastery Fund for this purpose, which is complementary to what I am suggesting here.

Second, such short-term—by which I mean 6 months to a year—apprenticeships and internships should also embed an employment pathway, conditional on reasonable performance on the part of the employee, unless a waiver is granted to employers due to changed economic circumstances. These trainees should be treated as employees under the Employment Act, and receive the same legal protections and entitlements, including a minimum period of annual leave, which GRIT trainees currently do not receive.

Third, we can ramp up the delivery of social skills training—communication, empathy, judgment, networking, and vision—in the final year of their tertiary education, prior to workforce entry. Research has shown that AI is most complementary to workers when the job demands require the fulfillment of not only cognitive tasks, but also iterative collaboration between humans and AI. But our graduates often load their school time with the pursuit of academic competencies, leaving them woefully underprepared for such interfacing functions.

Fourth, if indeed we stand by our belief that we want our graduates to focus on acquiring competencies rather than certifications—as MOE has made clear in its support for stackable microcredentials pathways in our AUs, and as corroborated by recent research—then we should put our money where our mouth is, and end hiring requirements that insist on a diploma or a degree in the public sector, if the competency can be demonstrated otherwise. This can occur with proof of skills via a series of microcredentials, or when candidates pass a live demonstration during the interview stage.

Conclusion

Artificial intelligence is a general purpose technology. Like all general purpose technologies before it, AI will destroy as many jobs as it creates. But as we confront the bleeding edge of the transition, we must set the stage for those most affected by the rollout, which are our young, entry-level workers. This is how we best ensure that the growth promises of AI are not overshadowed by fears of millions of missing jobs.

‍