Spain
The Congress of Deputies, and what its two-seat provincial floor costs.
Same shape as Brazil, different numbers and a much stranger law. The Congress of Deputies has 350 seats. Every province gets an initial two, Ceuta and Melilla one each, and the remaining 248 are shared out in proportion to population.
A resident of Soria weighs 4.13 Madrid residents: 44,189 people per seat against 182,442. And the ratio has widened at every election since 2008, from 3.71 to 4.13, without a single change to the rule. Depopulation of the interior does the work — the floor is fixed at two seats and the provinces sitting on it keep losing people.
| Scenario | Vote-weight ratio | Loosemore–Hanby | Seats moved |
|---|---|---|---|
| The law as written | 4.13× | 10.52 | — |
| Provincial floor cut from 2 seats to 1 | 3.50× | 5.46 | 18 |
| No floor at all | 2.35× | 1.97 | 36 |
| Sainte-Laguë instead of quota and remainders | 4.13× | 10.52 | 0 |
| One national 350-seat district | 1.00× | 0.00 | — |
Cutting the floor from two seats to one halves the disproportionality and moves 18 seats: Madrid goes from 37 to 44 and Barcelona from 32 to 37, paid for by one seat each from eighteen interior provinces. Removing the floor altogether moves 36 and takes Madrid to 50.
Changing the method instead does almost nothing. Replacing quota-and-remainders with the Sainte-Laguë divisors moves one seat in two of the seven elections and none in the other five — and in 2008 the one seat it moves makes the vote-weight ratio worse, handing Soria a third seat. Spain's malapportionment lives in the floor, not in the arithmetic.
Two caveats worth stating. The vote-weight ratio falls only from 4.13 to 3.50 when the floor is halved, because Ceuta and Melilla still hold a seat each on about 85,000 people and they set the ratio once Soria stops doing so; the ratio is an extreme and the index is a total, and they are measuring different things. And none of this touches D'Hondt, which in Spain is not the apportionment rule at all but the rule for turning votes into seats within a province — where, in a five-seat province, it does more damage than the floor does. Testing that needs the province-level party votes, which are published as fixed-width records with no independent total to check a column offset against, so they are deliberately left alone rather than guessed at.
Checked, and this one was unexpectedly interesting: the law says the basis is the población de derecho but neither it nor any decree says which edition of the population register. So rather than assume, the rule is run against every edition the statistical institute publishes, and the ones reproducing the official seat table exactly are recorded. All seven elections reproduce 52 provinces out of 52, and the edition that works is in every case the most recent one published before the decree.