← Back to portfolio

Complex Network Analysis Design

Published on 4th January 2018

This data analysis design focuses on the Slovenian National Assembly. It builds on the dataset and analyses from Parlameter project conceived by Danes je nov dan, Inštitut za druga vprašanja (Today is a new day, Institute for other studies) from Ljubljana, Slovenia. Parlameter is an open source project focused on providing data for political journalists, researchers and interested publics or citizens. Despite the fact this data is not digitally native, the proposal relates to digital society as it seeks to analyse digitised parliamentary activity, set an example for increased transparency of national governments and other political institutions while opening up their data and making it machine-readable.

The proposed analysis was presented to 30-50 researchers from the field of complex network analysis from educational and research institutes in Slovenia. As it will be presented at their conference in January 2018, the data analysis challenge is adapted to their perceived skills.

In order to understand more about the field of complex network systems, I read several articles (Baur, et.al., 2009; Dal Maso, et. al., 2014) researching the applications of this field's advances in the realm of representative political systems. I have been inspired by an article (Lansdall-Welfare, et. al., 2017) proposing a community detection analysis that came out of a historical study enabling the researchers to unveil the institutions of power in British society between 1800 and 1950. Reading about network analyses of cosponsoring legislation in parliaments (Tam Cho, et.al., 2010; Briatte 2016) and other data mining analyses (Jakulin, et.al., 2004) applied to various political environments, I decided to devise a challenge that I have not encountered in the literature I found.

The challenge is the following:

  1. How does the chain of influence form between agents in the political sphere (influencers towards MPs who then ask questions to ministries and the government)?
  2. What (or whom) do the questions reference and are those references deployed strategically towards specific ministries or does the network analysis suggest MPs reference things close to them regardless of the institution they are addressing?

Use case description

The 7th parliamentary term of the Slovenian National Assembly began in 2014. The next election will, normally after 4 years in office, take place in 2018. In case the data challenge gets finished soon enough, the preliminary analyses and visualisations presented to the interested public(s) could have an influence on the outcome of the next elections.

In recent years, plenty of network analyses of parliamentary and Congress activity focused on party divisions, bill co-sponsorship, communities surrounded around specific topics, etc. Despite the important insights of those analyses, such research often does not have a direct impact on citizens’ voting decisions. Our data analysis proposal seeks to inquire into more local, grounded topics or ways of exerting power. In Slovenian parliamentary system, citizens vote for individual representatives in their area and not the MPs' parties. That means voters are often inclined towards candidates because of the specific opinions or topics they bring up, or because they represent their local area well.

Moreover, legislation is predominately proposed by parties or the government, and only a minority of law proposals are sponsored by MPs themselves, which makes bill co-sponsorship analyses less important for the Slovenian political system. Another factor affecting my decision to focus on researching MPs’ questions is the fact that by joining the EU, national states lost sovereignty in many areas, which means national parliaments are more effective or influential when tackling more local issues or issues more directly important to citizens.

Each MP can pose a question (oral or written) or propose an initiative (written) to the Government, a Minister of the Government or the Secretary General of the Government. By doing that, an MP can push for a regulation of individual cases or adoption of certain measures, or critique the (non)existence of certain measures, as well as intervene in the (under)representation of certain topics in the National Assembly. Questions and initiatives can be proposed to MPs by different stakeholders, such as mayors or other local entities (Zakonodaja.com).

In some cases, MPs are – formally or nominally - the sole creators of questions and initiatives. But as they often collaborate with professional advisors, other MPs, their respective parties and lobbyists, that is only partly true. In other cases, MPs refer to other entities, for example political actors in other countries (such as Head of the German Office for the Protection of the Constitution - BfV) or companies (such as the successful Slovenian businessman Igor Akrapovič).

The prescribed and obligatory reporting of lobbying contacts by The Commission for the Prevention of Corruption of the Republic of Slovenia, in force since 2010 (Komisija za preprečevanje korupcije), is known for its inefficiency. With the network analysis I am proposing, we could track the activity of MPs back to specific stakeholders and local areas, which would resolve some of the ambiguities regarding the power or influence of local entities/businessmen in the National Assembly. Such analyses enable us to track lobbying otherwise, presenting publicly available albeit dispersed data in an understandable manner by visualising it. 


Data

As I worked as data accuracy manager and analyst for the project Parlameter, I understand the dataset well. For a year, I have been responsible for the static data input (biographical information about MPs, their memberships and positions in parliamentary groups and bodies) and update of membership changes, as well as tagging motions (legislation). I therefore vouch for the reliability of the data - as far as the government did not interfere with it prior publishing it on their website, which must be always taken into consideration when working with open data published by governments.


The main building blocks of Parlameter website are Profiles of MPs and parties, voting records and session transcripts – all parsed from the website of the National Assembly. Moreover, Parlameter offers three open-sourced APIs for datasets gathered from the website of the National Assembly. The open-source documentation posted on GitHub is prepared in English, which makes Parlameter applicable to countries with similar political systems.


In the beginning of 2018, Parlameter will get an upgrade: the dataset will be richer for summaries of motions passed in the National Assembly, which will enable users a more in-depth insight into the legislative changes affecting citizens. It will also bind legislation to votes that happen in the parliament, something which wasn’t possible until now. This allows researchers to analyse a specific piece of legislation among the multitude of voting events that happen in its name.


Parladata API

Is an exact copy of the data parsed from the website of the National Assembly. Returning to the remark about data reliability - it is important to point out that to believe in and maintain legitimacy of a representative parliamentary system, we must question the dataset’s complete accuracy only if we encounter unexpected, missing or surprising data. Sometimes, a speech is assigned to wrong people or people bearing the same names. In order to have a dataset as accurate as possible, we pursue manual reliability checks by checking Speeches by MPs, other political representatives and guests in the parliament whose names are similar. Moreover, MPs can request changes of their speeches´ transcripts. That is why Parladata saves all the versions.

In a few cases where the parsed data was not identical to the information found on the website of National Assembly (for example, number of proposals or motions by MPs with insufficient data in the title) we realised it was because of administrative mistakes in naming documents, so both sides, the National Assembly and Parlameter, benefit from the opening of the data, analyses and friendly collaboration.

Parlalize API

The Parlalize API serves calculated and organized data and metrics. The calculations are performed on data accessible at the Parladata API endpoint: most of the metrics are calculated every 24 hours or less.

Apart from open/publicly available data it would be interesting to inquire into the content of proprietary data, such as e-mails, social accounts or telephone conversations of MPs. That would enable us to track the lobbyists, local entities or party influencers, as well as members of the Government who influence the questions proposed by MPs. As it is often heard in Slovenia, the political system is being led by “uncles from the background.” That analysis would be interesting, but it is very unlikely surveillant capacities will be ever given in the hands of individuals, NGOs or institutes researching political networks.

Data exploration and analysis

Using the terms of network analysis: influencers, actors and MPs, as well as MPs' questions or proposals, will be visualised as nodes. Sponsorship of questions or proposals will stand for directed edges. We took the idea from measuring influence in social networks, where more influential nodes are visualised as bigger nodes. In this case, more influential nodes will be MPs who propose and address a higher number of issues. Ranking of MPs by the number of questions and proposals is already part of Parlameter’s analyse

In some cases, the MP is solely responsible for the proposal, which makes them a “double node.” This should be evident from the visualisation itself, possibly using a different colour.


For the network analysis, we need to search the non-machine-readable PDFs and:

- connect questions and proposals to the entities or nodes who proposed them to the MPs;

- connect questions and proposals to the entities or nodes the MPs refer to (could be books, media, etc.);

- connect questions and proposals to the entity or node they address (Government, a Minister of the Government or the Secretary General of the Government)

The data needed to begin the analysis is almost ready, and can be called from Parladata API. The element that needs to be further organised are labels on questions or proposals that would specify them according to the topic, which would present another layer of the visualisation, one that users could enable or disable with a click. Recent examples of questions and proposals by MPs include: inquiring into the dedicated budget for sailing tuition in elementary schools; “migrant” tents in a town called Lendava; role of the anti-vaccination movement. A useful categorisation can be found in Legislative Explorer, a website tracking legislation co-sponsoring in the American Congress, where legislation is grouped under “Topics” and “Minor topics”, for example Public Lands and Water Management, Macroeconomics, Environment (Topics) and National Budget and Debt, Freedom of Speech & Religion, Research and Development (Minor topics).

Legislation in Parlameter that comes from the Government is tagged or labelled with references to specific working bodies that discuss them before being voted on in the National Assembly – in the case of questions and proposals that are more specific, it would be sensible to choose different labelling.

Viewers of the network visualisation could change from “MP View” to “topic view” or “geographical view”. This could be made possible by the previously proposed labeling of questions and proposals by topics, which would be very time consuming and it would need help from political scientists. The end result, however, would be useful not only to voters, but also to researchers of political culture who could delve deeper into the issues held dear by MPs and so-called “minor” lobbyists. Furthermore, this analysis would enable us to make a visualisation of actors influencing MPs, for example mayors. It would enable us to identify influential nodes - actors whom more MPs represent or actors or entities (different ministries) who are being more questioned than others.

Another possibility for an analysis would be focusing on time-series. A chronological visualisation would demonstrate the number of questions or proposals raised at a specific date, as well as what topics were raised at specific moments in the parliamentary convening. That could be compared to the legislation, which was in the voting process at the time.

As

Parlameter includes an analysis of words that make MPs special or are

characteristic of their corpus versus the whole corpus, we could add

another layer to the network analysis. This analysis would tell us

whether there is a correlation between “fancy” speech

and

addressing more macro-economical issues, simple speech

and

addressing more local issues, as well as excessive speech and more

ideologically polarized inquiries.


Our network analysis could be visualised using Gephi at first and then ported into browser-based technologies to make them easy to distribute and share. A lot of effort should be put into visualising different types of active and inactive MPs and stakeholders (or nodes) in a comprehensible manner. A useful guide for simplifying network visualisations is found in an article on motif simplification by Dunne and Shneiderman (2013). Moreover, to visualise the local efforts or networks of MPs, geographical areas that are being mentioned or referred to in their questions or proposals could be mapped using QGIS soft

Discussing questions and proposals by MPs, as it is written in The Rules of Procedure of the National Assembly, happens once a month (Zakonodaja.com). The Prime Minister, ministers and the Secretary General of the Government must attend the session, but there are often many political representatives - predominately MPs – who do not attend the session. Moreover, not all questions and proposals by MPs get an answer: after establishing a certain “waiting time”, the analysts could mark the issues that did not get the attention from the Government.


Conclusion

This data analysis proposal departed from Parlameter database and analyses. Firstly, I thought about the data available and the analyses that are already part of the website. Reading about the current applications of network analysis in parliamentary data research and about the application of that research in other fields, I tried to find a question that would enable us to visualise a more bottom-up flow in a representative democratic system.


Questions and proposals are a mechanism predominately used by opposition parties and are an important mechanism for demonstrating lacks of the coalition. On the other hand, however, posing a lot of questions and issues is also a strategy for slowing down the political process, which can be used to make the ruling party or coalition less efficient. In this manner, network analysis would enable individual voters to judge for themselves whether the questions could be justified by the importance of issues raised, or are they just a tool for slowing down the work.


I suppose various stakeholders will profit from a network analysis taking into account the MPs' more individual inquiries and their collaboration with non-governmental entities. It is the MPs questions and most importantly, proposals by various stakeholders that lead us to the core of their role in the National Assembly. Their main function is to be representatives of their voters and to exert pressure towards Government that often forgets who it is serving.




References

Baur, M., Brandes, U., Lerner, J. and Wagner, D., 2009. Group-Level Analysis and Visualization of Social Networks. In: Lerner, J., Wagner, D. and Zweig, K.A., eds. Algorithmics of Large and Complex Networks. Design, Analysis, and Simulation. Berlin and Heidelberg: Springer-Verlag.


Briatte, F. 2016. Network Patterns of Legislative Collaboration in Twenty Parliaments. Network Science, 4 (2). Available at:

> [Accessed 15 December 2017].


Commission for the Prevention of Corruption. 2011-2013. Lobbying [website] Available at: [Accessed 3 January 2018].


Dal Maso, C., Pompa, G., Puliga, M., Riotta, G. and Chessa, A. 2014. Voting Behavior, Coalitions and Government Strength through a Complex Network Analysis. PLoS ONE, 9 (12). Available at: [Accessed 15 December 2017].


Danes je nov dan, Inštitut za druga vprašanja, 2017. Parlameter for developers. [website] Available at: [Accessed 3 January 2018].


Dunne, C. and Shneiderman, B. 2013. Motif simplification: Improving network visualization readability with fan and parallel glyphs. Human-Computer Interaction Lab Tech Report. Available at: [Accessed 23 December 2017].


Jakulin, A., Buntine, W., La Pira, T.M. and Brasher, H. 2009. Analyzing the U.S. Senate in 2003: Similarities, Clusters, and Blocs. Political Analysis 17 (3). Available at: [Accessed 20 December 2017].


Lansdall-Welfare, T., Sudhahar, S., Thompson, J. and Cristianini, N., 2017. The Actors of History: Narrative Network Analysis Reveals the Institutions of Power in British Society Between 1800-1950. In: Advances in Intelligent Data Analysis XVI, 16th International Symposium onIntelligent Data Analysis. London, United Kingdom 26-28 October 2017. Cham: Springer Nature.


Legislative Explorer [website] Available at: [Accessed 23 December 2017].


Tam Cho, W. and Fowler, J.H. 2010. Legislative Success in a Small World: Social Network Analysis and the Dynamics of Congressional Legislation. The Journal of Politics, 72 (1), pp.124-135.


Zakonodaja.com. Poslanska vprašanja in pobude. [website] Available at: [Accessed 2 January 2018].





0 Comments Add a Comment?

Add a comment
You can use markdown for links, quotes, bold, italics and lists. View a guide to Markdown

You will need to verify your email to approve this comment. All comments are moderated before publication.

Close