Is German Politics Becoming More Polarised?

As a life-long observer of German politics, I have seen government being dominated by one party (and one leader), the collapse of a coalition that promised a new beginning, and the steep rise of extreme sentiments at the left and right of the political establishment. Pundits usually attribute these observations to the rise of polarisation, a phenomenon that is thought to be pervasive of politics in the 21st century. While a lot of quantitative analysis has been done on polarisation in the US two party system (Hare, Pool 2014), continental European coalition-style parliaments have received somewhat less attention. I’m setting out to explore what polarisation means in terms of the German Bundestag and whether it actually got worse over time.

Political polarisation describes the emergence of opposing parties struggling to find a common ground on any matter. Search interest in the term has surged over the last decade. But how can one quantify polarisation from voting behaviour in parliament? I scraped every vote cast in the Bundestag since 2005 from abgeordnetenwatch.de and built a metric to measure how similarly any two MP’s vote.

The dataset contains roll-call data from 6 legislatures ranging from 2005 to the current 21st Bundestag in 2026. Roll-calls are votes where each MP’s individual vote is recorded by name (yes, no, abstain). The majority of votes taken in the Bundestag are taken by show of hand - no data exists for these votes. Roll-calls only occur on request of either one faction or 5% of all MPs and are thus mainly used for politically contentious votes. The dataset contains 649 polls with 437k votes and 1,905 distinct MPs across the six periods. Absentees are removed from the data to obtain a clean set of yes/no/abstain votes in all 649 roll-calls.

As a proxy for agreement between MP’s, I calculate the pairwise Cohen’s $\kappa$ for all intra- and inter-party MP pairs. The $\kappa$ measures how much two MPs agree with each other beyond what one would expect from chance and is defined as

$$\kappa = \frac{p_o- p_e}{1-p_e}$$

where $p_o$ is the relative observed agreement between MPs and $p_e$ is the hypothetical agreement between MPs. The observed agreement is obtained by dividing the number of polls where both MPs $i$ and $j$ voted the same by the number of polls they voted in. The individual vote rates for $p_{i}$ and $p_{j}$ can then be combined to form the probability of hypothetical agreement

$$p_e = p_ip_j+(1-p_i)(1-p_j)$$

which is the level of agreement one would expect if $i$ and $j$ voted independently. Cohen’s $\kappa$ then gives us the perfect agreement beyond chance for $\kappa = 1$ and perfect disagreement $\kappa = -1$. For $\kappa = 0$, the two MPs vote perfectly independently.

A nice way to visualise this is by drawing a graph of all MPs as nodes grouped by their party affiliation where the edges between MPs reflect the positive $\kappa$ contribution of each MP pair. A corresponding plot of inter-party edges with $\kappa > 0.15$ of the Bundestag period of 2021–2025 is shown below. Note that the dashed hull across SPD, Die Grünen, and FDP indicates the government coalition of this period.

Each edge shows a cross-party MP pair that often votes together. I visualised the top forty cross-aisle MPs adjusted by party size. These are members of the parliament voting most often and most consistently with their cross-party partner. As expected, most cross-party MPs can be found within the coalition of SPD, Die Grünen, and FDP. CDU/CSU member share large amounts of edges with SPD and FDP, perhaps an effect of their long coalition history. The AfD proportionally has the least amount of cross-party votes in parliament – this leads to some interesting follow-up questions on party-specific political cooperation within the parliament. More on that later.

With forty MP pairs per party at most being visible, this graph is highly selective. We need to look at distributions rather than individual MPs to study the phenomenon of polarisation. Even when looking at distributions, we always expect to have strong intra-coalition cross-party cooperation. Framed differently: coalition systems always polarise since the government has an automatic majority and cross-aisle votes are not needed to decide most legislation.

Having split the MPs into coalition and opposition blocs, how cohesive is their voting behaviour? The government parties should vote together more ofen than an opposition made up of parties from a broader political spectrum. This figure shows the distribution of $\kappa$ values for both opposition and coalition groups. Governments clearly tend to pose as one coherent voting front with its distribution shifted to the right of the scale. Angela Merkel’s first government in the 16th Bundestag notably shows the least amount of measured cohesion with the peak of the distribution quite far from perfect agreement. This might be an artefact of the limited roll-call data for this parliament or an effect of the newly formed Grand Coalition after two legislative periods of SPD led governments before 2005. The oscillations visible in the distribution of MPs in the current government coalition is due to data sparsity at the time of writing.

In all parliamentary periods, coalition parties show a distinct bimodal $\kappa$ distribution with a bloc of MPs voting somewhat cohesively (righthand peak in $\kappa > 0$) and a bloc disagreeing ($\kappa < 0$) or voting without preference ($\kappa \approx 0$). This is typical of the Bundestag as the opposition groups parties that ideologically disagree with each other, as shown in this table:

Bundestag	Years	Opposition Parties
16th	2005–2009	FDP, Die Grünen, Die Linken
17th	2009–2013	SPD, Die Grünen, Die Linken
18th	2013–2017	Die Linken, Die Grünen
19th	2017–2021	FDP, Die Grünen, Die Linken, AfD
20th	2021–2025	CDU/CSU, AfD, Die Linken
21st	2025–	AfD, Die Grünen, Die Linken, BSW

The most coherent opposition with the rightmost left peak is the 2013-2017 pairing of Linken and Grünen - hardly surprising given their close political alignment. Despite containing three left leaning parties, the opposition of the current 21st Bundestag is still largely uncohesive due to the outsized impact of the right-wing AfD which took 21% of the vote. With the internal cohesion of the blocs clarified, it is time to ask the central question of this piece: is the Bundestag becoming more polarised?

Using the coalition-opposition structure established in the intra-bloc distribution plot, we now plot the level of agreement based on votes for each coalition and opposition MP pair. Each dot in this graphic represents such a pair, coloured by the party colour of the opposition MP.

Two observations stand out. Firstly, the largest part of these pairing show a negative $\kappa$-value. As expected, the opposition votes against the coalition and not with it. Secondly, the mean of the distribution is negative across all Bundestag periods but does not get larger in absolute terms. In fact, the mean of the two most recent parliaments in 2017-2021 and 2021-2025 is the lowest out of the five full parliaments studied, indicating a lower level of disagreement than in previous parliaments.

The 18th Bundestag of 2013 to 2017 shows the biggest polarisation with a mean $\bar{\kappa} = -0.39$. The ideologically tight opposition of Die Grünen and Linken took only 19% of the vote and was arithmetically unable to influence government policy. The least polarised parliament is the grand coalition of 2017-2021. The opposition of FDP, Grünen, Linken, and AfD only is in slight disagreement with the government at $\bar{\kappa} = -0.14$. The biggest effect on the agreement has the FDP which tends to agree with the government and - uncommonly for the German system - acts as a semi-partner outside of the actual coalition. Grüne only slightly disagree with the government whereas Die Linke and the AfD disagree more strongly.

The perceived colour of the distributions indicates a level of cooperation between the opposition parties. Strikingly, the Ampel coalition of 2021-2025 has a relatively cooperative main opposition with Linke, BSW, and CDU clustered in colourful peak close to the mean. The AfD inhabits a clearly blue island on the left of that peak. We can see that separation in the previous intra-opposition distribution as well. The same clear separation seems to develop in the current 21st Bundestag as well, with the AfD again being the party in highest disagreement with the government.

Is German politics getting more polarised? The data tends to disagree. Polarisation in the Bundestag - measured as the voting disagreement between government and opposition parties - has not monotonically increased since 2005. What has changed is the polarisation structure of the parliament. The entry of the AfD in 2017 introduced a bloc that systematically disagrees with everyone.

This conclusion comes with some caveats. Cohen’s $\kappa$ tells us how much coalition and opposition disagree but not why and on which matters. It treats each MP pair independently, ignoring that voting behaviour is influenced by party affiliation and perhaps direct cross-aisle contacts in other parties. In reality, there are standout MPs such as Bundeskanzler and Fraktionsvorsitzende (leaders of the faction) that might have a bigger impact on party votes than others. How do these figures influence the voting behaviour of parliament? Stay tuned for an analysis of political influence of MPs in the German Bundestag - polarised or not.

The code for this essay is available here.