I’ve long been interested with how political parties position themselves on issues. It’s an exercise that requires balancing a number of different goals. Parties generally want to attract voters, so staking out a position closest to the most voters is a good idea. Parties also want to enact favored policies, which might conflict with the vote-getting goal. To do so, at least in countries with three or more parties represented (i.e. not the US), parties often have to join a coalition with other parties to gain a majority and enter the government. That usually requires setting policy stances that are compatible with potential coalition members.
Of course, these goals often conflict with each other. Parties that want to attract more votes can calibrate their policy stances to appeal to a greater number of voters. That, however, might alienate activists that parties rely upon to help them campaign. Other parties also want votes and might be setting up a similar position on the issues. The question then becomes how parties differentiate each other to voters. Policies that appeal to activists might not be broadly popular or might alienate potential coalition partners. Radical right parties in countries like Belgium and Sweden have been frozen out of coalitions by informal agreement (the so-called cordon sanitaire) by the other parties due to radical right parties’ position on immigration and foreigners.
Basically, there has been a ton of research on party competition over the decades (if you printed all of the academic books and articles on the topic, it would probably actually weigh several tons). Along with understanding why and how people vote, it’s about as core a topic in political science as there is. Of course, a starting point for understanding how parties compete against each other is understanding where parties stand on the issues.
Party Policy Documents as Data
There are a variety of ways to do this, but one of the most popular sources for party issue positioning is the Manifesto Project. The Manifesto Project takes party election manifestos and assigns each sentence an issue code based on that sentence’s content. For example, a sentence talking about the need for environmental protection gets assigned the ‘Environmental Protection – Positive’ code.4 The process can be painstaking – I coded election manifestos in Russian elections using the scheme for my dissertation and can confirm that it is not the most fun I’ve ever had doing research. The researchers running the project calculate the frequency of each issue as a percentage of the whole document. You can then aggregate combinations of the 50+ issues into policy dimensions like social policy or state involvement with the economy.
The data presented here come from the Manifesto Project dataset. The main dataset covers elections around the world, but I’ve decided to limit the data to elections in EU member states for now. You can choose the country, election and issues that you want to examine using the menus in the dashboard. Not all parties in all elections are recorded in the dataset, so if you are a fan of an obscure monarchist party in Sweden, you may be out of luck. You can read a bit more about the method used to create the policy dimensions here and more about the dashboard itself here.
I created the dashboard itself in Tableau and have a version of it created using Shiny in R. That proved more difficult to get online without either posting the original data (which I can’t per the data TOU) or paying to host the Shiny app. If you know of any easy way around those problems, please drop me a line. In addition, I welcome any comments or suggestions, particularly about features you would like me to add. I’ll do my best to accommodate good/practical ideas for improving the dashboard.
Finally, you may need to change browsers to view the embedded dashboard; I had problems viewing it in Firefox, but was able to do so in Chrome. You can also view the dashboard hosted on Tableau Public.
A Few Notes on the Data
The promise of the data is pretty self-evident. It provides a measure of political party issue attention that can be consistently applied to party documents across time and space. For example, you can explore how party systems change over time using the controls above. It also provides a more or less objective way of measuring party policy preferences, at least compared to surveys of experts or voters. By using a pre-established coding scheme to analyze what the parties themselves are saying, we can cut through some of the noise that surrounds party politics. Of course, parties sometimes (often? always?) misrepresent their policy preferences so no system is perfect.
And the project is not without its downsides. At a broad level, manifestos represent only one outlet for parties to communicate their policy stances. Politicians speak with the media, give speeches and create other ways of getting their message out like advertisements and pamphlets. Manifestos can also cover up different factions within a party, since they present a unified stance towards the public. Finally, the manifesto data, by their nature, don’t tell us much about how voters perceive the parties.
More narrowly, the Manifesto Project data has been critiqued on methodological grounds. One common critique is that the coding scheme rest on assumptions of something called saliency theory, which posits that parties don’t actually take contrasting positions on issues so much as they simply emphasize different issues. That means that the codes are not always ideal for assessing how parties position themselves on the same issue. Another common critique is that the data do not provide uncertainty estimates around the coding assignments.5 For non-social-scientists, this might be a head-scratcher. In a nutshell, errors can creep into the data in two ways. First, writers of any document are not 100% accurate or consistent in communicating their ideas. Second, human coders aren’t 100% accurate in matching issue codes to text, particularly when sentence meanings aren’t clear. Without uncertainty measures (e.g. the margin of error in an opinion poll), we might end up overestimating differences between parties. Finally, the included aggregate measures included in the dataset for measuring right-left party positions are controversial.6 Type in ‘manifesto analysis and right-left scale’ to Google Scholar and you’ll see
Of course, the alternatives to the Manifesto Project data – expert surveys, surveys of voters, media analyses – have their own problems. Rather than dive further into the rabbit hole – you can do that yourself with some strategic reference-searching and Google Scholaring – I’ll leave you with the dashboard. I hope you enjoy.