Two-fifths of your skills go stale in five years, and that is an improvement

The figure is a survey of expectation, not an audit of skills. Employers are asked what proportion of skills they think will change, and the answer moves with how uncertain they feel. That makes it a sentiment series as much as a labour one, which is exactly why it peaked in 2020 when nobody could see six months ahead.

Quoted alone, 39% reads as a warning. Quoted in series, it reads as a recovery. Both use the same number from the same report. This is the clearest case in the corpus of a figure whose meaning is set entirely by whether the previous readings travel with it, and they usually do not.

Skill instability is not the same as job loss. A skill being transformed can mean the job persists and the tools change, which is the ordinary condition of most trades. Reading two-fifths of skills going stale as two-fifths of work disappearing conflates a change in how work is done with a change in whether it exists.

Quote the series, not the point: 39% for 2025 to 2030, against 44% in the 2023 edition and 57% in 2020. One number alone cannot show a direction.

Name it as employer expectation. It measures what companies think will happen to skills, not an inventory of skills that have actually gone obsolete.

Keep it apart from job counts. Skill change and job loss are different claims, and this figure only supports the first.

If you want the direction in a sentence, say skill instability has fallen since 2020 rather than saying skills are changing faster than ever, which the data does not support.

Presenting 39% as evidence that change is accelerating. On this measure it has been decelerating since 2020.

Reading it as two-fifths of jobs going away. It describes skills within jobs being transformed or outdated, which is a statement about how work is done.

Comparing it against a skill-obsolescence figure from another source. The number is defined by this survey's own question wording, so it is only comparable with its own earlier editions.

Treating the 2020 peak as the baseline. That reading was taken in the most uncertain labour market in living memory, and using it as the normal case makes every later figure look like relief.

All three readings come from a single passage of the WEF report, which states the current figure and both predecessors together. Quoting them as a series therefore takes nothing out of context; the source presents them as one comparison.

We checked the report's own characterisation before adopting it. It describes the measure as having slowed relative to previous editions, so the falling trend on this page is the source's reading and not our inference from three data points.

The figures rate High because each is printed explicitly and the comparison is the report's own. That rating covers how safely you can cite what the survey found, which is separate from how much weight a forecast of skills deserves.

We do not convert the percentage into a number of workers or skills. The survey reports a share of skill sets, and putting a headcount behind it would manufacture a precision the question never had.