Methodology

Every item is a balanced-alternative pair. The respondent is not asked whether they agree with a one-sided claim — that produces acquiescence bias and lets the author smuggle in a thesis — but which of two statements comes closer to their view. Both are written to be endorsable by a reasonable, informed person; neither is a strawman.

Which option appears first is randomised per respondent per item, so position confers no advantage. Answers are stored un-flipped.

Polarity balance

A self-selected web poll is not a probability sample. Within each nation-state, respondents are raked (iterative proportional fitting) to that country's adult population marginals on age, gender and education, with weights trimmed to a 5× band so rare cells can't dominate. Across nations, each country's contribution to any global figure is post-stratified by its share of world adult population.

Every published figure carries a Kish effective sample size and a margin of error derived from it, not from the raw count — raking always costs precision, and the number shown should admit that.

Each item's midpoint is recomputed from the world's weighted answers. The centre of an axis is therefore where the world's population actually sits, not the arithmetic zero a designer picked. A consequence worth stating: you can move without changing a single answer, because the world moved around you.

A direct consequence, stated so it isn't mistaken for a fault: once every item is re-centred, the world's mean position on each axis is zero by construction. There is no absolute origin in opinion space, so we don't pretend to one. Every published score — yours, your nation's — is a distance from the world's actual centre of gravity, and that centre moves as the data does.

Item-rest correlation is computed per question. Core items are never silently reweighted — comparability across years is worth more than a marginal gain in fit — but weak ones are flagged below. Pulse items, which carry no longitudinal promise, are attenuated when they measure poorly.

• Nothing about countries where internet access is restricted, low, or surveilled. Silence there is not moderation, and we do not impute it.
• Nothing causal. This is a cross-section of opinion, not an explanation of it.
• Nothing about people who won't answer political questions online — a group that is almost certainly not politically random.
• Weighting corrects for observed composition on three variables. It cannot correct for the fact that the politically disengaged are underrepresented on a site about politics.