Every regional election map on this site required someone to answer a question that sounds trivial and is not: what shape is a constituency?
Countries publish administrative boundaries — provinces, municipalities, counties — because those are what statistical offices and mapping agencies need. Electoral districts are drawn by a different authority for a different purpose, and they routinely fail to line up with anything published. So a constituency map is almost never a download. It is a construction, and the construction leaves marks worth being honest about.
Case 1: merging upward — Malta
Malta elects its parliament by single transferable vote in 13 districts, each returning five members. No open boundary dataset contains those 13 districts. What is available is a file of 68 localities — town-level units.
The 13 electoral districts are built from those localities, so the map is a lookup: each locality is labelled with the district it belongs to, and localities sharing a district share a colour and a result. The districts are never dissolved into single polygons, which means the map draws 68 shapes to show 13 results. Visually the outcome is correct; structurally it is a compromise that keeps the source data intact and traceable.
Case 2: splitting downward — Poland
Poland's Sejm is elected in 41 constituencies whose boundaries follow administrative lines but do not correspond to any single published administrative layer. They can, however, be reconstructed: the Senate's smaller districts nest inside them, so dissolving the Senate geometry along the right groupings reproduces the 41 Sejm constituencies.
This is the ideal case. Nothing is approximated — one published geometry is combined into another by a rule the electoral law itself specifies. When it works, it is exact, and the resulting map can be checked constituency by constituency against the commission's own list.
Case 3: the boundary that cannot be drawn — Iceland
Iceland has six parliamentary constituencies, and its results are published for all six. Two of them — Reykjavík North and Reykjavík South — divide the capital between them along a line that no open boundary dataset contains.
There are three options here and only one of them is honest. You can invent a plausible dividing line, which produces a confident map that is wrong in the exact place readers will scrutinise most closely. You can leave the capital blank, discarding real data. Or you can merge the two into a single Reykjavík polygon and show the mean of their two shares, stating plainly that this is what you did.
The Iceland referendum map takes the third option: five polygons for six constituencies. The map is slightly less granular than the data. It is not misleading about where a line falls, which matters more.
Case 4: aggregating upward, when the electoral unit is too small
The opposite problem is just as common: the electoral geography is far finer than anything worth drawing. Denmark counts municipally; New Zealand counts by electorate. Neither makes a readable national map — one produces a hundred slivers, the other seventy-two.
Both are handled by aggregating upward, and the two are not equally easy. Denmark's municipalities nest cleanly inside its five regions, so summing votes municipality by municipality into regional totals is exact arithmetic. New Zealand's electorates do not nest inside its sixteen regions — an electorate can straddle a regional boundary, because the two boundary sets are drawn by different bodies for different purposes. Aggregating there requires a lookup that assigns each electorate to a region, and that assignment is a judgement, not a calculation.
Where the aggregation is exact, the map carries the same authority as the table. Where it rests on a lookup, the page says so. The distinction is invisible in the rendered image, which is precisely why it has to be written down somewhere.
When there is no regional breakdown at all
Sometimes the geometry is fine and the data is missing. A number of countries publish only a national result — no per-region breakdown at any level. In those cases the election page carries no map and says why, rather than shading a map with something plausible.
This is a more common situation than the polished maps around the internet suggest, and it is where invented data most often enters circulation: the map is the expected format, so something gets drawn. A blank space with an explanation is a better artefact than a filled one with an asterisk.
Why we build them anyway
Given all this, it is worth asking whether the regional map earns its cost. I think it does, for one reason: the national percentage tells you the outcome, and the regional pattern tells you what happened. Almost every election on this site is more legible geographically than it is as a single number — an urban-rural split, a linguistic border, a region moving against the national swing.
But the map is a summary of the table, not a replacement for it, which is why every page shows both, and why the area-versus-population problem is worth reading before drawing conclusions from any of them, including ours.