This year marks a century since some women in the United Kingdom were awarded the right to vote in and stand for elections. Since then, 490 different women have represented their constituents in the House of Commons, bringing the Commons to 32% women. In this article, we follow their stories through data and identify their contributions with the help of machine learning.
Let's draw a line for each woman MP's stay in parliament. For example, these are the lines for the longest-serving woman MP.
Dame Margaret Beckett left parliament for a few years so she gets two separate lines.
Here are all the other women MPs. Many different women from across the political spectrum have served over the past century. Let's dive in and look at a few prominent ones.
Since 1918, 490 women have been elected to parliament and there are currently 209 women in the House of Commons.
The 1997 election saw a doubling in the number of women MPs, with 80 newly elected. Of these, 70 were Labour MPs (highlighted in the chart).
Half of these new Labour women MPs had been selected through All Women Shortlists (AWS, for short).
The controversial decision to use AWS (a process in which only women are allowed to stand for candidate selection in certain seats) faced opposition on the grounds that it was discriminatory. Legal challenges were later nullified by legislation passed in parliament.
While the merits and pitfalls of AWS are still being debated, there is no doubt that they have helped improve the representation of women in parliament.
Let's now look at how close we are to gender parity in parliament.
In 2018, there are 209 women MPs, which, while significant, is still a long way from gender parity. For every woman currently in the House of Commons, there are twice as many men.
The breakdown also varies by party. In 2018, 45% of Labour's MPs are women. In part due to all women shortlists, there are currently more women Labour MPs than all the other parties combined (and at a time when the Labour Party is in opposition).
But only 21% of current Conservative MPs are women. A Conservative landslide election could cause a significant step backwards in progress towards equal gender representation unless the party puts more women forward as parliamentary candidates in marginal seats.
For smaller parties such as the Liberal Democrats and the Scottish National Party, women make up about a third of their ranks.
While approximately a 32% of MPs in the UK are women, it’s important to place the imbalance in context of other countries. Is the UK progressive, or does it pale in comparison to other governments in gender representation?
Visualized is the percent of women elected to other national government bodies (among OECD member countries). For Norway, this is the Stortinget. For Mexico, it's the Cámara de Diputados. For the United States, it's the House of Representatives.
The Nordic countries lead in equal gender representation, although Mexico has also made great strides and is progressing much quickly than the UK despite the fact that both countries started with the same gender representation in 2002.
In fact, the UK is a rather middle of the road performer. None of these countries has made it to gender parity. In fact, the only two countries in the world to meet or exceed 50% women in governing bodies are Bolivia and Rwanda, both of which also have gender quotas.
With gender parity, constituents might hope that elected officials focus on more inclusive issues. To better understand how a gender shift in the House of Commons might affect the topics debated, consider the speeches made by MPs in the main chamber.
We compiled a collection of of 1.2 million speeches delivered by men and women who have served in the House of Commons since 1970, including monologues, debates, single-sentence replies, and acknowledgments.
We then used a probabilistic machine learning model1While there are a number of different machine learning techniques that can do this, including neural networks, the beauty of Latent Dirichlet Allocation (LDA) is that it is an unsupervised learning algorithm. Unlike most other algorithms, it does not require pre-labeled speeches to train the model, which would bias the inference process by selecting the topics in advance. The only necessary inputs are the corpus of speeches and a rough guess of the total number of topics within the corpus. to help us infer the topics within a speech, looking for words that often appear together and clustering them into the same group2For example, suppose we choose a small number of topics (e.g., 10), the model will identify the 10 most overarching themes in congressional speeches. We wanted to identify more subtle topics and themes, and after a bit of experimentation, we settled on 100 topics. As is often the case with LDA models, many of the topics identified were nonsensical, but the majority were well-defined. (eg. NHS & social care, energy policy, armed forces, education, etc.)
The result: a rough calculation for the fraction of time each representative has spent on an issue.
Here is the topic “economy.” Each circle represents one MP, with men on the left and women on the right. The higher the circle, the greater the proportion of their time was spent on the economy.
We can represent the difference between the median amount spent by men versus women with a line.
Men and women appear to devote roughly the same amount of their speeches to the economy. Next let’s take a look at another topic, one to which women tend to devote more parliamentary time.
While welfare reform is an important topic for all MPs, women spend more than twice the amount of their speeches discussing reform than men. Now let's examine a topic favored by men: parliamentary terminology.
Even when accounting for party, the result is almost exactly the same: Conservative women are as far apart from Conservative men as Labour women are from Labour men1View the breakdown by party here..
This time, we see that men spend twice as much time talking about parliamentary terminology.
Let's take the median difference between men and women for parliamentary terminology (depicted as a line) and plot it against all other topics.
A difference of +100% (eg. “education”) means that women are twice as likely as men to refer to the topic in a speech. A difference of -100% means the opposite: men are twice as likely as women to refer to the topic in a speech.
Women MPs focus more of their speeches on welfare reforms, child care, the NHS and social care.
Men spend more of their speeches on legislation, energy, the European Union and the armed forces, and the disparitities exist even after accounting for party affiliation.
Select a topic to examine its data more closely.
Lastly, here are the number of women from all parties who were selected as candidates for parliament since 1945.
Parliamentary candidates for the next general election are yet to be selected, so it is still too early to tell if the next election will see an improvement in representation. Ultimately, what matters is not how many women candidates stand for election, but rather where they stand. Candidates selected in marginal seats are much more likely to be elected than those who run in safe seats which do not change party hands.
In the 2017 general election, only 38% of candidates in winnable seats were women1Winnable seats are defined as those with less than 5% vote share between the top two parties.. In order to achieve gender parity, this number will need to be much greater than 50% to overcome the huge gender imbalance in non-marginal seats which do not change hands often, and are thus mostly occupied by veteran male MPs.
Academic research overwhelmingly finds that the lack of gender equality in parliament is a demand-side problem: parties are not selecting enough women, rather than not enough women standing. This is particularly true of the Conservative party, which must select significantly more women candidates in marginal constituencies.
You may find all the code for this project, along with much more detail on the methodology on Github.
Chronological data of women MPs from the House of Commons Library.
All speech and MP data from TheyWorkForYou.com.
Data on parliaments in the OECD from the Inter-Parliamentary Union and licensed under the Creative Commons International 4.0 license.
Data on parliamentary candidates over time from the House of Commons Library upon request.
Finally, this work wouldn't be possible without the help of various machine learning libraries such as spaCy and gensim, data libaries such as pandas, visualisation libraries such as D3 and the python and R ecosystems.