Singapore runs the world’s best-funded lifelong learning system and a dedicated scheme for workers over 40, yet it publishes a re-employment rate for only a niche of about 4,300 people. For its flagship programme the government declined in Parliament to give placement rates or before-and-after wages. It is a measurement problem for anyone promising to reskill workers ahead of AI.
In May 2024 Singapore’s government credited S$4,000 to the training account of every citizen aged 40 and over. Not cash: a credit spendable only on courses, stacked on top of subsidies that can cover up to 90% of course fees. The scheme is called SkillsFuture Level-Up, and it is the most ambitious answer any government has yet given to the question everyone is asking: what happens to a fifty-year-old office worker when the job gets automated.
Inside that catalogue — more than 7,000 courses covered by the mid-career credit — artificial intelligence appears in two distinct ways, and it is worth keeping them apart. First as subject matter: in 2025 the government launched over a hundred new “train-and-place” pathways in high-demand areas including AI, teaching people to use and integrate generative models and machine learning systems into business workflows. Second as the underlying reason a forty-five-year-old signs up at all: the fear that a system trained on the historical data of their own industry will end up doing, at near-zero marginal cost, the predictable part of their job. Singapore has built the infrastructure to answer that fear. What it has not built, or at least not published, is the means to know whether it works.
The only outcome number that exists
The system is vast. Participants in training supported by SkillsFuture Singapore, the statutory agency that runs the scheme, rose from 418,000 in 2016 to over 600,000 in 2025 — roughly a fifth of the resident workforce, every year. In 2024 alone, 260,000 people used their SkillsFuture Credit, and 1.05 million have used it at least once since the programme began.
Out of all that volume, there is exactly one public employment-outcome figure, and it covers a far smaller programme: the SkillsFuture Career Transition Programme (SCTP), the pathways that pair training with placement. According to a Ministry of Education parliamentary reply in September 2025, between the programme’s launch and December 2024 roughly 55% of nearly 4,300 previously unemployed SCTP participants found a new job within six months of finishing their course.
Four thousand three hundred people. Against six hundred thousand trained a year. That is under one percent of a single annual cohort — and the ministry itself has acknowledged it lacks complete data on how many SCTP participants were unemployed when they enrolled, which makes the denominator itself an estimate.
54 and 55 are not the same series
An earlier figure often circulates alongside it: in August 2024 the same ministry stated that, as of December 2023, over 54% of SCTP participants had found a new job within six months of completing their course. Several outlets have treated 54% and 55% as two points on one curve, the second marginally better than the first.
They are not. The 54% covered all participants; the 55% covered only those previously unemployed. Different denominators, different periods. Lining them up manufactures the illusion of a trend where there are two different measurements of the same programme. Worth adding: the two ministerial documents date the SCTP’s launch inconsistently — April 2022 in one, June 2022 in the other — which says something about how carefully these numbers are handled even at source.
The answer Parliament did not get
The most telling item is not a number but a refusal. A Member of Parliament asked the government for the share of Level-Up participants employed within nine months, broken down by age band 40-49 and 50-59, and for median wages before and after training. The reply contains none of those figures. What it contains instead is an argument: the programme’s success cannot be measured by employment outcomes alone, because training by itself cannot guarantee a job.
That is a defensible claim, and partly a true one. It is also exactly what a ministry says when the data either does not exist or is not encouraging. From the outside, the two situations look identical.
In their place, perception indicators circulate: in 2024, 69% of trainees reported improved job performance (up from 65% in 2023), and 64% linked their training to a career advancement. These are survey answers, not outcomes verified against employment records. People who have just invested months in a course tend to tell themselves a good story about it — a well-known effect in every evaluation of training policy.
What is actually missing
None of the following appears in the public material:
- a control group — a comparison with similar workers who did not take the course. Without it, the programme’s effect cannot be separated from the self-selection of who enrols;
- measured wage effects: finding a job at half the previous salary still counts as a success in a re-employment rate;
- job retention beyond six months, which is the horizon at which you learn whether a transition held;
- the share who actually moved into the new sector they trained for, rather than back into the old one.
This is the hard part, and it is hard everywhere. It is also the part that matters if the stated goal is managing a technological transition rather than delivering courses.
How many actually use the credit
Level-Up’s first year (May 2024 to April 2025) tells a story of real but contained uptake: over 36,000 citizens used the mid-career credit, with more than S$24 million reimbursed. SCTP enrolments rose from around 1,500 to around 8,500, nearly six-fold. Those using the credit for full qualifications went from about 3,000 to about 7,000.
Yet in 2024, the 28,000 people who tapped the mid-career top-up between May and December amounted to roughly 1.3% of the 2.1 million eligible. A free credit with subsidies up to 90% reached, in its opening stretch, barely one eligible person in a hundred. That number explains why a training allowance arrived in May 2025: 50% of average income over the previous twelve months, capped at S$3,000 a month, for up to 24 months over a lifetime. The binding constraint was never the price of the course; it was the income foregone while taking it. In the first three months, over 3,200 citizens received the allowance for full-time training, with more than S$30 million committed by end of July 2025.
An apparatus has grown around it. SkillsFuture Jobseeker Support, open for applications since 15 April 2025, offers up to S$6,000 over six months to those involuntarily made redundant, with over S$200 million set aside. Career Conversion Programmes fund employers reskilling mid-career hires, with salary support of up to 90% across some thirty sectors. And at Budget 2026 the prime minister announced the merger of SkillsFuture Singapore and Workforce Singapore into a new Skills and Workforce Development Agency, established by legislation passed in May 2026 and operational from 1 July.
Why this matters outside Singapore
The most encouraging precedent remains the pandemic one: over 60% of participants in the SGUnited programmes between 2020 and 2022 found new employment within six months. But that market was rebounding from a contraction, and a high placement rate also happens without any programme at all when hiring restarts across the board. With no counterfactual, we cannot know how much of that 60% the training earned.
On this front Singapore does more than almost anyone: the best-funded system, the most transparent participation figures, a chain of instruments running from credit to allowance to jobseeker support. Which is precisely why the gap matters. If the country that has invested the most publishes a re-employment rate for 4,300 people and stops there, then anyone else announcing reskilling plans for the generative-model era is promising a result nobody can currently show they achieved.
One detail sums it up. Of the 7,000 courses covered by the mid-career credit, only around 700 are full-time and therefore eligible for the allowance: one in ten. And the SCTPs, the only pathways with any outcome data at all, numbered roughly 330 across 21 sectors as of July 2025. The part of the system that measures itself is also the smallest part — and the only one that can be defended with a number.