Atlas / Free reports / Risk analysts

The Classification Problem in Country Risk

Why analyst notes get argued with, what a coded position does that a label cannot, and how to build a house vocabulary two analysts can share.

The problem is not that your classification is wrong

It is that a label cannot be defended. When a client disputes the characterisation in a note, there is nothing to point at except the analyst who wrote it. The disagreement becomes personal, because nothing else is available for it to be about.

The underlying issue is compression. A phrase like right-wing populist is carrying at least six independent claims at once. A client who disagrees with one of them appears to be disagreeing with all six, and the analyst who wrote it has no way of establishing which one is actually in dispute.

Six questions, not one

Political organisations differ on at least six things that move independently: who owns productive assets, how social life should be ordered, whether power can be checked, at what level decisions are taken, who counts as a member, and by what means change is pursued.

Compress those into a single left-to-right position and the information does not blur. It disappears — and the axis it disappears on is usually the one the client cares about.

Six parties routinely grouped as one family. They span six points on economics and five on the organisation of power. A note that treats them as interchangeable is wrong about at least one client-relevant question.

Worked position: Fidesz

Hungary’s governing party is the cleanest demonstration available.

It scores +04 on economic organisation — to the right of most European social democracy, and a long way from the interventionist position its critics often assume. On the organisation of power it scores +07: elections continue, are genuinely contested, and are counted accurately, while the institutions that could reverse their outcome have been restructured by the incumbent.

Both figures are true simultaneously. A single label cannot carry them, which is why a client who disputes the label has nothing to argue with except you. Two numbers with their evidence give the disagreement somewhere else to go.

The trajectory matters as much as the position

Fidesz was a member of the Liberal International in 1990, with an age limit of thirty-five for membership and a classically liberal programme. An organisation that has travelled a long way is a different analytical object from one that has always been where it is, and only one of those two facts is visible in a label.

The movement is concentrated on two axes, and one of them is form. Economics moved comparatively little. The transformation was not about what the Hungarian economy should look like; it was about how power is held and who belongs.

The pattern that recurs across the region

Run the same reading across the cases a regional note is most likely to cover, and a consistent shape appears: convergence on culture and membership, divergence on economics and on how far the constraints on power have actually been removed.

The economic spread here is thirteen points. Any note that characterises these as a single economic tendency is making a claim the record does not support.

The real purchase is consistency

Speed is the obvious benefit and it is the smaller one. The larger problem in most research teams is that two analysts characterise the same party differently, so notes are inconsistent across engagements — and a client who reads two of them notices.

A published codebook with fixed anchor definitions is what fixes that, and it is the thing an individual researcher cannot supply. Anchors work by writing down what each point on the scale means before anyone codes anything, which is how the Chapel Hill expert surveys keep hundreds of separate academics on one scale.

Here is the economic axis as an example. The same treatment exists for all six.

Four checks before a characterisation goes in a note

These run on paper, in about ten minutes, with no dataset at all.

1. Name the axis. Which of the six is the claim about? Extreme, hardline and far describe intensity, not position. Most disputed characterisations collapse at this step.
2. Test the name against the record. What did the organisation legislate, expropriate, ban and build? Those four outrank anything it calls itself, and they are the four a client can check.
3. Read the composition, not the distance. If one axis carries most of the gap, the claim is about that axis and should be written as such. If the gap is even, the comparison is between broad types and no single axis is doing the work.
4. Check the frame. Is the claim about absolute position or relative position? A party can appear to have moved when the centre moved and it stood still.

What to check before relying on any coding scheme

Four questions, which apply to this instrument as much as any other.

• Is the scale fixed? If nobody has written down what +7 means, two coders will not agree on it.
• Is there evidence per cell? A score without an artefact behind it is an assertion in monospace.
• Does it publish anything against itself? An instrument that only ever reports findings in its own favour is telling you something.
• Can you test it? This atlas has a blind coding instrument that shows a cell’s evidence with the score withheld and asks you to assign one.

On the third: dropping any single one of these six axes leaves a distance matrix still correlating with the full model above 0.95. The six-axis model is therefore not defensible on aggregate statistics, and is justified instead by a named twenty entries a smaller model would misplace. That is published on the method page because it is true.

What this instrument cannot do

The other three reportsMore for this role