The Alaska County Map Is a Reconstruction

There is a version of the 2024 presidential map in which Alaska is shaded in thirty pieces — nineteen boroughs and eleven census areas — each carrying its own margin, each rendered with the same confidence as a county in Ohio. It is a handsome map. Nobody certified it. The Alaska Division of Elections does not report presidential results by borough or census area. It certifies them by state house district.

That means every borough-level Alaska election map you have ever looked at is a reconstruction. Someone took district returns and pushed them down onto boundaries the returns were never counted against, using assumptions that are almost never printed next to the map. The assumptions may be reasonable. They are still assumptions, and they are load-bearing.

Politigoat stores Alaska at the grain the state actually certifies: 40 house districts across 7 cycles, plus the District 99 bookkeeping pile that reports in three of them — 283 records in total. We publish six of those seven cycles. MEDSL's Alaska rows for 2004 sum to 488,693 votes, roughly half again the total of every neighbouring cycle in the same file, and the file's own totals column agrees with the inflated figure, so it is wrong at source rather than in our reading of it. Every other state's 2004 total there reconciles to within a third of a percent. We drop the cycle and say so on the page. We publish no borough figures at all. Not estimated ones, not modeled ones, not ones with an asterisk. The cell is empty because the number does not exist, and inventing it would make our map look more complete than our evidence is.

What Alaska Certifies

Alaska carries two entirely separate sets of geography, and the mismatch is the whole story. The Census Bureau maintains 30 county-equivalents. The Division of Elections maintains 40 house districts, plus a District 99. Neither system was built to translate into the other. House districts are drawn to equal population; boroughs and census areas are drawn to everything else.

Thirty into forty does not go, and it does not go in a way no amount of care can fix. Any borough number is therefore the output of an allocation rule someone chose — split by precinct, by population, by registered voters, by area. Different rules produce different maps. All of them will look equally authoritative once they are colored in.

The Number That Makes Allocation Guesswork

If the allocation problem were only about boundaries, it would be tractable: assign each precinct to a borough and add. But a large share of Alaska's ballots are never assigned to a precinct in the first place. They are tallied at the district level, and the district is where the trail stops.

Read that 2020 row again. More than half the ballots in the state carried no precinct assignment at all. In House District 33, it was 73.2 percent. Anyone reconstructing boroughs from precincts in 2020 was, at best, allocating a minority of the vote by geography and spreading the majority around by rule. The rule is doing more work than the data.

In 2020, 56.5 percent of Alaska's ballots carried no precinct assignment. A borough map built on the other 43.5 percent is not a measurement.

Codes That Look Like Counties

Here is where our own pipeline nearly walked into it. MEDSL, a standard academic source for precinct and county returns, ships Alaska under pseudo-FIPS codes in the band 02001 through 02050 — a fifty-slot range that is numerically indistinguishable from real county FIPS codes. Nothing about the string 02017 announces that it is a house district rather than a borough. Feed it to a county-grain loader and it will be accepted without complaint, join cleanly against a shapefile, and render.

What caught it was arithmetic, not intuition. Alaska has 30 county-equivalents. The file had 41 Alaska units. A state cannot have more counties than it has counties, so the loader stopped and asked.

• Unit count exceeded the known county-equivalent count for the state (41 against 30).
• The codes occupied a contiguous low band, 02001–02050, rather than the scattered, alphabetically assigned pattern real county FIPS codes follow.
• The genuine Alaska county codes that turn up in the FIPS change record — 02158 Kusilvak, 02195 Petersburg, and the retired 02270 Wade Hampton — all sit above 02050. That is luck, not a guarantee, and we do not treat it as a safety margin.

Alaska's real county codes are not stable enough to lean on anyway. Wade Hampton Census Area became Kusilvak Census Area on 2015-07-01, moving from 02270 to 02158. Petersburg Borough changed territory on 2013-01-03 — a population transfer of about 613 — while keeping FIPS 02195 throughout. A state whose actual county identifiers churn that much is not a state where a plausible-looking 02xxx code should be trusted on sight.

District 99

Then there is House District 99, which has 806 registered voters, 401 ballots, and no territory. It is a bookkeeping district — a place for votes that belong to Alaska but not to any of the forty geographic districts. It is entirely real as an administrative object and entirely unmappable as a place.

District 99 is also a small, clean demonstration of why the borough reconstruction fails. There is no honest borough to put those 401 ballots in. Any allocation rule must either drop them, which changes the state total, or distribute them somewhere they demonstrably did not come from.

What We Publish, and the Cell We Leave Empty

So Politigoat's Alaska page shows house districts. It does not offer a borough toggle, because a borough toggle would be a rendering of our assumptions rather than of Alaska's returns. Statewide totals are unaffected — the ambiguity lives entirely below the state line — and district-level series are directly comparable cycle to cycle, which is more than a reconstructed borough series can claim.

This is the same instinct that made our build stop on Connecticut, where 8 county-equivalents retired and 9 planning regions were adopted between the 2020 and 2024 Census vintages and the national count moved from 3,143 to 3,144. In both cases the pipeline's job was not to smooth the discontinuity over. It was to refuse to pretend the discontinuity wasn't there.

A missing number is a worse-looking product and a better honest one. If you need Alaska by borough, you will have to choose an allocation rule and own it — and if you do, publish the rule next to the map. The map without the rule is the part we object to.