It’s not D’Hondt fault

In this article, I will examine the results from the last Spanish general election, which took place in November of 2019. In Spanish general elections, the seats from the Congress of Deputies assigned to each political party are not always representative of the popular vote. And contrary to what many people here in Spain seem to believe, that is not due to the use of the D’Hondt method. He is innocent. The main reason for the misrepresentation of the popular vote in the Congress of Deputies, as I will try to explain, is that Spanish general elections are not really one national election, but a set of local elections. Each of the 50 provinces of Spain elects a minimum of two seats, and the remaining 248 (there are a total of 350 seats) are distributed in proportion to their populations. The only exceptions are Ceuta and Melilla, which only elect one seat each, although they are not provinces of Spain (they are “autonomous cities”).

How big is this misrepresentation between the Congress of Deputies and the popular vote? Pretty big if you ask me, especially for the two biggest parties: PSOE earned 28.3 % of the vote but got 34.3 % of the seats, while PP earned 20.1 % of the vote but got 25.4 % of the seats. The overrepresentation of these two parties means that, in general, smaller political parties are underrepresented. ERC earned 3.6 % of the vote and got 3.7 % percent of the seats, which is fair; but Ciudadanos, with 6.9 % of the vote, only got 2.9 % of the seats in Congress. And here is another pair of misrepresented parties: ¡Teruel Existe! got one seat (which is 0.29 % of the seats) with just 0.08 % of the votes; while PACMA, with 0.95 % of the votes, got no seats in Congress.

Why does this happen? As I have said already, it’s because Spanish general elections are not really one national election, as the name would suggest, but the sum of 52 local elections (the 50 provinces and the 2 autonomous cities). It is easier to see the problem with the autonomous cities of Ceuta and Melilla. In these two cities, whichever party gets the most votes, gets the local seat. Last year, the one seat from Ceuta went to Vox, which got 35.5 % of the votes from Ceuta. But what this means is that the rest of the political parties, which in total got 64.5 % of the votes from Ceuta, got no representation in Congress. Likewise, the one seat from Melilla went to PP, even though they only got 29.7 % of the votes.

This is a common problem in science: categorization of a continuous variable means loss of information. In every one of these 52 local elections, each city must divide their votes into a limited number of seats. When you do this, you toss away valuable information.

Let’s imagine the next hypothetical scenario, let’s say that each of the 52 Spanish constituencies elects only 3 seats. And we are going to assume that in every one of these cities, party A earns 100 votes, party B earns 90 votes, party C earns 80 votes, party D earns 70 votes and party E earns 60 votes. That makes for a total sum of 400 votes. Therefore, we can say that party A got 25 % of the votes, party B got 22.5 %, party C 20 %, party D 17.5 % and party E 15 %. However, following the D'Hondt method (which is a proportional and therefore fair method to allocate seats), we get that party A, party B and party C all get one seat each in every one of the 52 Spanish constituencies. Thus, in the Spanish Congress of Deputies, party A, party B and party C would each have 33.3 % of the seats; while party D and party E would get no seats.

Do you understand the issue now? Cities which elect few seats will usually not be well represented in Congress. In total, 21 of the 52 Spanish constituencies elect 4 or fewer seats. That means that even if you are the fifth most voted party in every city of Spain, you will get no seats in any of these 21 Spanish constituencies.

And that is almost exactly what happened to Ciudadanos. They were the fifth most voted party in Spain, with almost twice as many votes as ERC. However, ERC still got 3 seats more than Ciudadanos. Why did this happen? It’s easy to explain: Ciudadanos received votes from all around Spain but didn’t get any seat from any constituency with 7 seats or less. Therefore, all the votes Ciudadanos earned in these constituencies were not taken into consideration, rendering them essentially useless. On the other hand, ERC only received votes from the 4 Catalonian provinces, but got seats from all these 4 provinces. Thus, we can see that this system is disadvantageous to political parties with no concentrated voting base.

And Ciudadanos is not the only party with many votes but a small representation in the Congress of Deputies. Similarly, Más País had 2.33 % of the popular vote but only 0.86 % of the seats. And PACMA earned more than 11 times as many votes as ¡Teruel Existe!, yet PACMA got no seats while ¡Teruel Existe! got one seat. Why? You guessed it: both Más País and PACMA received votes from almost all Spanish constituencies, yet ¡Teruel Existe! got many votes from just one province (Teruel).

So, what do you think? I believe that we should fix this. There are too many votes not taken into consideration from having 52 small local elections instead of one bigger national election. This system was put in place to ensure local representation from every constituency (the 50 provinces and 2 autonomous cities) in the Congress of Deputies. However, the net result is that –for the most part– the only political parties that benefit from this are PSOE and PP, the two biggest parties in Spain. 

This has many implications, and we could keep talking about this for days. For example, people might be more likely to vote for PSOE or PP, not because those are the parties that represent them best, but because they realize that their vote is better strategically placed when they vote for the lesser of the two evils. Maybe few people want to vote for PACMA because they know that it will probably have no representation in Congress, even if it gets more than 1 percent of the popular vote.

But, from a purely scientific point of view, my conclusion is this: never categorize a continuous variable unless you really have to. Otherwise, you will be losing precious information!