Copyright: Jo Devriendt, Nonfiction Software. Created: 2024-07-03. Last updated: 2024-09-10. Comments or enquiries are welcome via our contact form.

An analysis of different electoral systems applied to Belgium's 2024 elections

When we think of the word democracy, we imagine a system of governance where every person in a country participates by casting a vote for a certain person or party at certain points in time. This indeed is the basis of democracy, but the electoral system - how votes are tallied and converted to elected representation - strongly influences whether the concerns of the people are properly reflected in the resulting governing body.

These electoral systems are not made equal. All electoral systems have some percentage of validly cast votes that are ignored in the final representation, but some will do so with abundance while others strive to minimize this effect. The good news is that we, the people, as participants in and maintainers of our democracies, can change our electoral systems for the better. To do so, we need a proper analysis of possible alternatives.

On this page, we have a look at the electoral system for the Belgian federal Chamber of Representatives. We analyze it in terms of its democratic deficit: the percentage of people that cast a valid non-blanco vote, but whose party of choice ends up with zero seats in the legislative body. Other fairness metrics of electoral systems exist, but because it is both simple and intuitively clear that we want to keep the democratic deficit low, we focus on this measure. We analyze the democratic deficit of the 2024 Belgian elections, estimate the worst case democratic deficit for the current Belgian electoral system, and compare this with the electoral systems of The Netherlands and the United Kingdom.

Current democratic deficit in Belgium

Firstly, we calculate the democratic deficit of the 2024 Belgian federal elections. It simply is the total number of people that voted for a party that ended up with zero seats in the federal parliament, divided by the the total number of votes for all parties (so we ignore invalid and blanco votes and people that did not vote). This yields 299 027 / 6 984 906, for a final percentage of 4.3%. More than 1 in 25 people that cast a valid, non-blanco vote do not have any representation amongst the 150 members of parliament. If these unrepresented three hundred thousand people would get a proportional representation, they would fill at least six seats in parliament.

Worst case democratic deficit in Belgium

Is a democratic deficit of 4.3% a high or low amount? Is it possible that in future elections we have an even bigger democratic deficit? How bad can this democratic deficit ever get? This all depends on the rules the Belgians decided to use in their elections.

Belgium is divided into eleven electoral districts, one for each province plus the Brussels capital region. Relative to the population size, each district sends a number of representatives to the Chamber of Representatives, the country's national legislative body. These representatives are distributed to political parties based on the number of votes they get in each electoral district. Parties that do not reach an electoral threshold of 5% in an electoral district do not get to send representatives for that district. For those parties that do reach the threshold, representatives are distributed according to the D'Hondt method. Many parties identify as Flemish or Walloon, and hence do not participate in all electoral districts.

To perform our analysis, we use the 2024 Belgian federal election as input data. As such, we use twelve parties - the current number of parties with elected representatives at the national level. These twelve parties map to the largest existing Belgian parties, and we fix their participation in the electoral districts to be the same as in the 2024 election. In other words, in our simulation, parties participate in the same districts as in 2024. Similarly, the number of representatives for each district is identical to the 2024 election, as is the number of valid non-blanco votes cast for each district. Finally, to reduce the computational burden, we divide all vote and population counts by a thousand. While this reduces numerical precision a little, it does not affect the overall conclusion.

We modeled these rules and this data as a ManyWorlds program and ran it on a home computer to calculate an extreme case of democratic deficit. We find a a worst case democratic deficit of 31.9%, or 2 231 000 votes, way more than 2024's blanco/invalid votes (416 577) and non-voters (966 546) combined. So in the current electoral system, we can end up with almost one in three valid non-blanco votes completely ignored.

This democratic deficit arises when one party (one that participates in all electoral districts) has just enough votes to grab all the seats. All other parties end up either slightly below the electoral threshold, or just short of enough votes for the D'Hondt system to assign them a seat. Of course, this is an extreme case that will not occur in practice. But it does provide another data point to compare with. Let's have a look how the Belgian elections would have turned out under the Dutch electoral system.

The Dutch electoral system applied to Belgium

The Dutch electoral system is much simpler. There is one electoral district encompassing the whole country. For every 0.67% of the vote a party gets, they get one of the 150 seats in the national parliament. If a party scores below 0.67%, they get no seats, making this the electoral threshold. Remaining seats are allocated one by one to the party with the highest average number of voters per seat. And that's it!

Before applying the Dutch electoral system to the 2024 Belgian vote results, we have to correct for the fact that Walloon electoral districts have more blanco/invalid votes and a higher non-voter percentage than Flemish ones. As a result, the total number of votes for Walloon parties will be relatively low compared to Flemish parties, because they participate mainly in the Walloon electoral districts. So instead of using the raw 2024 national vote counts, we normalize these votes along the number of total eligible voters for each electoral district. After doing this, we model the Dutch electoral system in ManyWorlds and calculate the allocation of representatives:

There is a clear trend: large Walloon parties (MR, PS, Les Engagés) lose a lot of seats, which go to small parties as well as large Flemish parties. This happens even after the above correction for invalid/blanco votes and non-voters. The reason is three-fold. First, the Dutch system is less advantageous to large parties than the Belgian one, due to a low electoral threshold and a more proportional distribution of seats. This explains why small parties gained seats and large parties lost seats. Second, there exist quite a lot of small Walloon parties that do not reach the electoral threshold, but do take away votes from the other Walloon parties. Third, the Brussels electoral region has a higher ratio of representatives to eligible voters, because it has fewer eligible voters per inhabitant. Since almost all seats in Brussels go to Walloon parties, these last two reasons explain why only Walloon parties lose seats.

Democratic deficit in the Dutch electoral system applied to Belgium

The democratic deficit in the Dutch electoral system applied to Belgium's 2024 elections yields 2.4%. This is a little more than half of the 4.3% democratic deficit in the Belgian electoral system. In absolute vote numbers, this is an improvement of more than 130 000 valid votes that now have representation in a Dutch-style seat allocation.

To calculate the worst case democratic deficit for the Dutch system applied to Belgium, we use the same data as before, ending up with a new ManyWorlds specification. Running this yields a democratic deficit of 7.2%, arising again when one party grabs all the seats, and all other parties end up slightly below the electoral threshold. The important thing to notice here is that by sensibly adapting our electoral system we could improve the fairness of our elections, both in extreme worst cases and in sensible circumstances. And the effects would not be minor: we are talking about hundreds of thousand of unrepresented voters!

The United Kingdom's electoral system applied to Belgium

While the Dutch electoral system has one big electoral district where all representatives are elected, the United Kingdom (UK) uses the opposite approach: each individual representative is elected in a tiny electoral district called a constituency. The candidate with the most votes in a constituency gets that constituency's seat. Such a winner-takes-all system does not require that the candidate has 50% of the votes of the constituency. Indeed, if there are a handful of parties competing in one constituency, often only 30% of the vote share is enough to claim the seat.

Again, we apply Belgium's 2024 vote results to the UK electoral system. For this, we use the Belgian cantons: small districts where vote counts are collected before they are aggregated on the provincial electoral district level. We assign to each canton a fractional seat proportional to the number of inhabitants in the canton, and each party gets the sum of the fractional seats of the cantons where they get the most votes. The result looks like this:

These results wildly differ from the actual election result! E.g., Vooruit, cd&v and Open Vld, the classic Flemish broad center parties, together have 0.1 seats. The reason is simple: in (almost) no single canton do they win, and winner-takes-all implies loser-gets-none. To visualize this: the colors of this map represent the biggest party in each canton. And indeed, in Flanders (the north of Belgium), the colors red (Vooruit), dark orange (cd&v) and blue (Open Vld) are firmly lacking. Only the small Zwalm and Horebeeke cantons color blue but their combined population merits about 1/7th of a seat. Surprisingly, the small Groen party (green) does conquer a couple of seats as they are the winner in the Ghent canton, the third largest city of Belgium.

Democratic deficit in the UK electoral system applied to Belgium

As even mid-size parties get no seats, it is to be expected that the democratic deficit of a winner-takes-all simulation will exceed the previous ones. Indeed, assuming Open Vld's 0.1 seat counts as no representation, we get a current democratic deficit of 29.9%, or almost 2 100 000 valid non-blanco votes that would not get any representation. Though not uncommon for multi-party winner-takes-all, it is comparable to the simulated worst case democratic deficit of Belgium's current electoral system, and is more than four times as high as the worst case democratic deficit using the Dutch approach.

To estimate the worst case democratic deficit for the UK system applied to Belgium, we assume there are 150 constituencies in Belgium which all have 47 000 eligible voters. We partition these in Walloon, Flemish and Brussels constituencies, and let only those parties participate in a constituency if they participated in any of the corresponding electoral districts. Running the resulting ManyWorlds program yields a worst case democratic deficit of 83.2%. This democratic deficit again happens when one party wins all constituencies, and all other parties are close to this winner. This worst case again is not realistic in practice, but it is a symptom of the unexpected electoral dynamics in a winner-takes-all electoral system. Notably, in such a system, there is pressure on smaller parties to join the larger ones, thereby diminishing the diversity presented to voters and the diversity amongst the parliamentary representatives. A clear example of this evolution towards a two-party state is the United States, where two parties control almost all representatives, and other voices only have a chance of being heard as subordinate members of one of the two collectives.

Given this comparatively high democratic deficit of winner-takes-all, surely nobody in Belgium is thinking of switching the current setup to a UK-style system. Right?

From N-VA's 2024 manifesto:

We [envision] introducing a new electoral system in which the majority of the Flemish parliamentary seats are divided on the basis of smaller local electoral districts in which one candidate ('winner takes all') is elected each time.

It definitely makes sense for N-VA to propose this: as the largest party in the country, they stand to gain a huge amount of seats, as already shown in the above table. But our analysis is a dire warning to small and even mid-size parties: in a winner-takes-all electoral system, they are doomed to obscurity.

Conclusion

When working with Belgian data, the Dutch electoral system showed a lower democratic deficit and the United Kingdom's a higher deficit than the Belgian approach, both in realistic simulations as well as in a worst case analysis. The main culprit seems to be the size of the electoral districts: the smaller the electoral district, the larger the democratic deficit.

While the lowest democratic deficit occurs in the Dutch approach with one big electoral district, it is possible to argue for electoral districts that are smaller than the national level when there are large cultural or demographic differences between certain parts of a country. In Belgium, a case could be made for three electoral districts based on the three different regions (Flanders, Wallonia, Brussels). Any smaller will incur an increasingly large electoral deficit, as showcased by the United Kingdom's extreme approach of tiny electoral districts for single candidates, leading to a non-proportional winner-takes-all result. Let's not go there...