Decoding RaymondsIndex: What an 85.9% Order-Agreement Rate Can and Cannot Tell You

Two control contests reignited in Seoul in the first week of September, and the market priced them in opposite directions.

At Hanjin KAL, the holding company for Korean Air and Asiana, Hoban Construction lifted its stake to 20.15% in July. Chairman Cho Won-tae and related parties hold 20.57%. Hankyung reports the gap as 0.42 of a percentage point; Money Today reports 0.41. The stock closed at 146,000 won on September 4, up 9.12% on the day and 21.3% over the month. At Korea Zinc, where Young Poong and MBK Partners opened a hostile approach two years ago, accumulation is largely complete and the contest has moved to an audit-committee election at an extraordinary meeting on September 9. That stock fell close to 10% across two sessions.

Same category of event. Opposite reaction. The variable is not size.

One number, read correctly

The headline validation figure for RaymondsIndex is an 85.9% concordance across a sample of 3,109 Korean listed companies. That figure is routinely over-read, so the derivation matters: concordance is an order-agreement rate. It says the ranking placed companies in broadly the right sequence. It is not a magnitude, and it is not an accuracy rate — it does not claim that 85.9% of predictions were correct, and it says nothing about how far apart two ranked companies sit.

Last time we published this, the point was what the number is not. This time the useful question is what follows from it. If a ranking is good at order and silent on distance, then it is built for triage — deciding which names to open first out of three thousand — and it is not built for scoring, thresholds, or cut-offs. Those uses need distance, and distance is exactly the part the statistic withholds.

Why distance is the fragile number

A rank does not flip on a large move. It flips on a move larger than the gap. At 0.42 of a percentage point, almost nothing has to happen for first and second place to swap; the reporting even notes that a one-point increase by the same holder in the second quarter of last year sent the stock limit-up on consecutive days. Meanwhile the Korea Development Bank holds 10.58% and is expected by some analysts to consider a sale after the Korean Air–Asiana integration completes — a block several times the size of the gap it would be crossing.

Korea Zinc shows the mirror image. There, ownership order has stopped being the live question. The next step is counting votes on a board, which is why the market found no premium in it. The distance that matters has moved from a shareholder register to a ballot.

The academic frame

This distinction is old and well-established. Kendall (1938) introduced rank correlation precisely as a measure of agreement in ordering, deliberately discarding information about the size of differences between observations. Harrell, Lee and Mark (1996) formalised the concordance index for prognostic models on the same basis: it evaluates whether pairs are correctly ordered, not whether the predicted quantities are correct. A model can score well on concordance while being badly calibrated on levels. That is a known property, not a defect — but it does dictate how the output may honestly be used.

What this means for an individual investor

The practical translation is unglamorous. A ranking earns you a reading order, not a verdict. When you see a name high on a risk ordering, the second question is not "how bad is the score" but "what is standing between this company and the one ranked next to it, and how easily does that thing move?" In an ownership fight, that thing is 0.42 of a percentage point. In a balance sheet, it is usually a covenant, a maturity, or a review date.

If order is the part we can validate and distance is the part that decides outcomes, which of the two is your own watchlist actually sorted by?

Companies are named here as public reporting. Nothing above is investment advice or a recommendation to buy, hold, or sell any security.

#RaymondsRisk #RelationalRisk #CorporateGovernance #OwnershipGap #ModelValidation