Korea's smallest firms lose workers 1.6x as fast as its largest. The pay ladder runs the other way.

The Korean labour market’s most one-sided trade is small versus large. The big firm pays more and churns less. Nothing about that is balanced, and the pension register shows both halves of it in the same file.

Churn falls as firms grow

Divide each month’s lost pension members by enrolled members and you get a separation rate — how fast a workplace turns over. By size:

• 2–4 workers — 4.29% a month.
• 5–9 — 4.46% (the peak).
• 10–29 — 4.21%.
• 30–49 — 3.97%.
• 50–99 — 3.71%.
• 100–299 — 3.26%.
• 300–999 — 2.95%.
• 1,000+ — 2.66%.

The smallest firms lose workers about 1.6 times as fast as the largest. The gradient is not perfectly smooth — churn actually peaks in the 5–9 band before falling — but from there down the size ladder it drops at every step. Big firms hold people; small firms cycle them.

Hiring tells the same story from the other side: the smallest firms enrol new members at 5.26% a month, the largest at 2.50%. Small workplaces are busier at both doors — more people in, more people out — because that is what a small, less permanent payroll looks like.

The two ladders point the same way

Read this against the companion cut on pay and the asymmetry is stark. The national-pension bill per worker — our pay proxy, capped at the pension ceiling — rises with size, from 291,760 won at the smallest firms to 456,315 at the largest. The churn ladder falls with size. So the large Korean firm wins on both counts at once: it pays more and it keeps you longer. There is no size band where you trade one for the other.

That is the economic engine behind Korea’s fixation on the big-company job. It is not one advantage, it is two, stacked — and the small firm has no offsetting edge in this data to put against them.

What this rate is, and isn’t

A separation here is a member leaving a workplace’s enrolment. It does not say why. A worker who quit for a better job at a larger firm counts as a separation at the small one and a hire at the large one — which means some of what looks like small-firm instability is really the pipeline feeding the big firms. That reading is consistent with the numbers but not proven by them. What the numbers prove is the pattern: attachment to the payroll strengthens, monotonically enough, as the payroll grows.

What we did not claim

We have not seasonally adjusted, and June is one month. We have not separated voluntary from involuntary exits, because the register cannot. And we have not controlled for industry — construction, which is small-firm-heavy, churns for reasons of its own that the companion industry cut takes up. The size gradient survives all of that as a raw fact; the causes behind it are a longer argument.