ASPECT: Digital Election Ephemera to Support E-Democracy in Scotland

Andrew Williamson, Alan Dawson and Jane Barton

Centre for Digital Library Research, University of Strathclyde, Glasgow G1 1XH

2004

This article appears in Proceedings of the International Conference on Politics and Information Systems: Technologies and Applications (PISTA '04)

Abstract

ASPECT provides web-based access to a digital archive of ephemera (leaflets, flyers, postcards and newsletters) produced by political parties and candidates for the 1999 Scottish parliamentary election. This paper describes the background to and initial development of ASPECT, reports on current work to develop a multi-election archive with enhanced user services, and discusses the potential for ASPECT to play an active role in supporting e-democracy in Scotland.

Background

In 1707, the Act of Union merged the parliaments of Scotland and England. Government of the United Kingdom was centralised in London, with Scotland's interests administered by the Secretary of State for Scotland or the UK Home Secretary through the Lord Advocate.

A separate government department (the Scottish Office) was established in 1885 to handle administrative functions in Scotland in areas such as agriculture, education and health. Despite this, demands for greater self-governance continued, ebbing and flowing throughout the twentieth century, and grew substantially from the 1960s onward. The first legislative attempt to respond to this pressure by creating a devolved Scottish Parliament failed after a popular referendum in 1979, precipitating the defeat of the Labour government in that year's general election.

The Labour Party returned to government in 1997, with a manifesto commitment to legislate for the creation of a devolved parliament for Scotland. A popular referendum in 1997 established that devolution was the expressed will of the population, with 74.3% voting for the policy. The resulting Scotland Act [1] finally became law in November 1998.

The first election of a Scottish Parliament in nearly 300 years took place in May 1999. Recognising the significance of this event, the Andersonian Library of the University of Strathclyde undertook to collect the newspapers, leaflets, postcards and other ephemera produced by the parties and candidates ahead of the election. As a result of this and subsequent efforts, the Library now holds a substantial and unique archive of ephemera from the 1999 and 2003 elections and intervening by-elections.

Elections to the Scottish Parliament are conducted using a combination of the traditional first-past-the-post (FPTP) system and an additional member system of proportional representation to elect a total of 129 Members of the Scottish Parliament (MSPs).

Each voter has two votes:

The system produces a result much more proportional to votes cast than the traditional FPTP system, and has allowed small parties which are penalised by FPTP to gain parliamentary representation for the first time, breaking the traditional four-party model in Scotland. Six parties and one independent candidate were elected in 1999, and seven parties and three independents in 2003.

The two-tier electoral system has encouraged many smaller parties to participate in the electoral process. This has increased the amount and variety of ephemera produced, and made the collection more complex, particularly given the mixture of regional and constituency level campaigning.

Developing the digital archive

ASPECT 1999

The paper archive poses a number of problems for storage, retrieval and usage:

Clearly, the paper archive (and the library staff who maintain it) cannot support anything other than occasional usage.

These problems were addressed by the creation of a digital archive based on the paper collection. The ASPECT (Access to Scottish Parliamentary Election Candidate Materials) project was established to create a full set of digital images, selected machine-readable transcriptions and item-level catalogue records for the ephemera, and to integrate these with detailed results from the election itself. ASPECT is now available on the web [2] and currently includes over 1800 digital images (excluding thumbnails).

Usage of ASPECT

In addition to conserving the original materials and improving retrieval and ease of use, the digitisation has made the archive readily accessible to a much wider audience than would otherwise have been the case. Although available to everyone, it is of particular relevance to students and researchers of political science, political history, Scottish affairs and communication, as well as to politicians, civil servants, journalists and librarians.

Usage figures for the ASPECT service are high (see Figure 1), with trends in usage broadly mirroring the usage pattern of UK academic libraries - higher during term time and lower over the summer vacation - but there is a very noticeable peak in March, April and May 2003. This period covers the run-up to, and aftermath of, the Scottish parliamentary election on 1 May 2003.

Figure 1: Aspect visits and page downloads 2003-04

Figure 1: Aspect visits and page downloads 2003-04

At present there is little information on how, why and to what extent ASPECT is being used. A consultation exercise is currently being planned, the aim of which will be to assess the information needs and behaviour of a number of key user groups, including political science researchers in the Department of Government at the University of Strathclyde, staff and users at the Scottish Parliament Information Centre and its partner libraries, journalists, and members of the general public. The outcomes of the consultation exercise will be used to inform the future development and promotion of the ASPECT service.

ASPECT 2003

The ASPECT project has recently been extended to include ephemera from the 2003 Scottish parliamentary election. Work is already underway on digitisation of materials and on the design and implementation of a multi-election archive with its additional requirements:

The limited resources available means that the websites have to be created automatically.

The future: Aspect and e-democracy

ASPECT already represents a unique resource for Scottish political researchers and has the potential to make a valuable contribution to the electoral process in Scotland and to democracy itself. This section of the paper outlines a number of areas where some further research and development would allow ASPECT to begin to fulfil this potential.

Prior deposit of ephemera

An inherent problem with ASPECT in its current form is the time-lag between an election taking place and the associated materials being made available on the web. While the materials retain a degree of historical interest, their potential value in supporting and informing the democratic process is diminished.

The current ASPECT model can be broadly summarised as:

i. Collection of materials during the election campaign.

Before the 1999 election, staff of the Andersonian Library were encouraged to collect ephemera delivered to their homes, and to ask friends and family to do likewise. This approach has obvious limitations regarding coverage, so for subsequent elections the Library enlisted the help of the Scottish Parliament Information Centre's partner libraries, of which there is one in every parliamentary constituency.

ii. Sorting and cataloguing

In the weeks and months after the election, the collected materials are passed to the Special Collections department at the Andersonian Library. The ephemera must then be sorted by political party and arranged by constituency and region. Any duplicate items must be removed, and the material must be catalogued. This stage of the process can take some months to complete, depending on the availability of staff time.

iii. Copyright clearance

In tandem with the sorting and cataloguing of the materials, the political parties and independent candidates must be contacted to obtain permission for their materials to be digitised and made available.

iv. Digitisation

The paper collection is loaned to the CDLR for digitisation. Digital master images are created, and surrogate images for web delivery are processed. The text of election leaflets is scanned using OCR technology, and the database of candidates, parties and constituency results is updated.

v. Publication

Software tools developed for the Glasgow Digital Library [3] are used to generate the new website automatically from the ASPECT database.

In the case of the 1999 election the materials were not published on the web until May 2002, a full three years after the election. This delay was in part due to the need to find funding and develop the technical infrastructure for the project from scratch. The resulting methodology is transferable to future elections, which should significantly decrease future delays, but even so, the current model makes it unlikely that materials could be made available within a year of the election without a significant increase in investment.

Clearly, for the ASPECT materials to be of maximum value in supporting and informing the democratic process, it would be desirable to establish a system of prior deposit, whereby parties and candidates would supply materials in print or electronic form at the start of an election campaign.

Funding is now being sought to conduct a prior deposit pilot study in a small number of parliamentary constituencies at a suitable future election. The study would be timed to coincide with one of the following:

A system of prior deposit gives rise to a number of issues, some of which may have legal implications. For example:

ASPECT would be required to work closely with both the UK Electoral Commission and individual parties to clarify such issues.

Correlation of voter opinions to party preference

Textual transcriptions have been created for 30% of the election literature issued by constituency candidates in the 1999 election, and a much higher proportion for the 2003 election. Techniques are being considered for identification and categorisation of candidates' published positions on a range of issues, drawing on personal statements and party policies mentioned in the election literature. The aim of this categorisation would be to produce an online environment in which voters could assess the extent to which their views on specific issues correlated with any party political preferences, and whether this might influence their electoral behaviour. This online environment would have the potential to encourage the involvement of more people with the electoral process, as advocated by the UK government.

Current patterns of electoral behaviour

There is substantial evidence to suggest a decreasing level of interest in the political process in the UK in recent years, based on declining party membership and electoral turnout:

However, this pattern of falling interest in traditional party politics is counterbalanced by evidence of increasing interest in other avenues for political involvement:

This empirical evidence suggests that the UK electorate can still be engaged by specific issues, but has decreasing faith in career politicians and the political process. There is certainly no shortage of further anecdotal evidence of disenchantment, such as the views that are expressed by Mainwood:

'Mainstream politics is a problem for anyone - male or female - who, in this sophisticated age, does not relate to the tribal culture of the parliamentary political parties, whose methods of engagement have remained unchanged since the birth of the British democratic system. What we want is a consensual approach, based on the issues involved, where an honest assessment can be made - from which decisions flow that reflect a practical approach which is free of egotism and party loyalty.'[8]

The Electoral Commission, investigating the causes of the low turnout at the 2003 Scottish parliamentary election, identified two main factors:- a decline in the perceived importance of the Scottish Parliament, and a perception that differences between the major parties are not very significant:

'the main political parties in Scotland are converging at a point just to the right of centre... the parties with a chance of forming a government are all likely to offer the same solutions, which themselves are seen in the context of not changing anything'. [9]

Analysis of election literature

In view of the trends outlined above, the big question that arises from the ASPECT project is whether it might be possible to stimulate more interest in politics, and specifically in voting, by:

There are however huge difficulties in trying to achieve this. Election literature is messy and complex. It contains substantial duplication, sloganising and generalisation, a mixture of national and local issues, a focus on individual candidates as well as policies, and frequent criticism of other parties, while policies thought to be unpopular are rarely mentioned. The result is a remarkable similarity in content between literature published by competing parties. For example, all parties think education, health and jobs are important, and all want to protect Scottish interests, while all candidates pledge to work hard and would consider it an honour to serve the Scottish people, and so on. Policies may therefore have to be inferred by their absence from the literature. Candidates never promise to close schools and hospitals, but if one candidate pledges to keep a local hospital open, then the omission of such a pledge from a rival candidate might imply a policy of closure, or it could simply reflect a focus on other issues.

Another factor likely to have a strong influence on voting is the individual candidate. At constituency level, individual personalities, reputation, track record and even physical appearance may have a significant impact. This is well illustrated by the case of Dennis Canavan, a former member of the Labour Party, who stood as an independent candidate for the 1999 Scottish Parliamentary election, and was elected as MSP for Falkirk West with 55% of the vote, easily beating candidates from the four main parties. In this case issues and policies seemed largely irrelevant - Canavan was elected on the basis of previous reputation and achievement (he was re-elected in 2003).

A further layer of complexity is introduced by the governmental system in Scotland, which has four levels: local (councils), regional (Scotland), national (UK) and international (European Union). Voters are likely to be confused about the powers and responsibilities of their elected representatives at each level, yet those standing for election to the Scottish Parliament are free to mention any issue, regardless of whether it falls within the Parliament's remit. For example, the presence of Trident nuclear submarines in Scotland is mentioned by some candidates, even though responsibility for defence lies with the UK government not the Scottish parliament. Yet this is not necessarily irrelevant, as it is used to illustrate the broader issue of Scottish independence, or lack of it. Similarly, local issues are often mentioned even though they are the responsibility of local councils not the Scottish Parliament. Where the same party holds power at both local and parliamentary level, the perceived failings of the party at local level may be publicised in order to attract the attention of voters and to cast doubt on the general competence of the same party.

Identifying topics and issues

All this complexity makes it difficult to extract significant and relevant issues from election literature. In order to make the process feasible, there is a need to identify a controlled vocabulary of discrete issues that can be used as the basis for meaningful textual analysis.

The following 20 general topics are taken from the web site of the Scottish Executive, and may therefore be judged representative of its main areas of responsibility:

Agriculture Justice
Arts & Culture Older People
Business & Industry Planning & Building
Communities Research
Economy Rural Development
Education & Training Statistics
Environment Sport
Fisheries Tourism
Government Transport
Health & Community Care Young People

These areas do not correspond particularly well to the issues commonly highlighted in an election campaign. For example, the issue of fisheries is confined to a minority of constituencies, while 'research' and 'statistics' are rarely mentioned during elections, presumably because few voters are thought likely to be interested.

The list omits issues handled by the UK government, such as defence, international relations and social security, but which may still be relevant to a Scottish parliamentary election, especially where the same party is in power in Edinburgh and London (which has always been the case from 1999 to 2004). For example, the involvement of Britain in the Iraq war was almost certainly unhelpful to the Labour Party in the 2003 election in Scotland, even though it was a matter for the UK rather than the Scottish Parliament.

The list of 20 general topics is therefore known to be insufficient, but it does serve as a useful initial controlled vocabulary and an important element of the methodology discussed below.

Proposed methodology - content analysis

The mechanisms envisaged for the proposed online environment are still under development and are the subject of a current funding bid. The methodology outlined below is therefore illustrative rather than definitive. It will be refined as the proposed work progresses, but it is already clear that several steps will be required to translate printed leaflets to succinct statements:

  1. Carrying out optical character reading (OCR) to convert the literature to machine-readable form. This is being done by two methods; direct scanning from paper and by conversion of existing JPEG images to text files. In some cases pre-processing is needed to invert images, e.g. from white-on-black or yellow-on-blue to black-on-white.
  2. Removing duplication of content. This is time-consuming and requires some careful editing. Duplication exists both within and across leaflets.
  3. Removing slogans and other text devoid of policy issues. Election literature is full of exhortations to vote and general publicity slogans that have no policy content. These have no value for the proposed textual analysis, and may clutter the results from full-text searching, so have to be removed (they are still available on the digitised images).
  4. Separating personal statements from policy issues. This is relatively easy, as most literature separates the text relating to the individual from policy matters. A simple XML-style markup scheme allows separation of the two within a text file, though again this involves manual editing.
  5. Carrying out concordance analysis to create a ranking of importance of general topics based on frequency of occurrence of specific words. Although this would be an imprecise mechanism, the high volume of text that could be processed automatically should ensure validity of results.
  6. Identifying and marking-up specific policies and pledges. This is a difficult but essential component of the proposed analysis. An example of such a policy is: 'scrap student tuition fees'. The wording used is likely to vary between candidates, so automatic analysis is unlikely to be very accurate. By manually marking-up such issues with a consistent and systematic syntax, automatic extraction would become feasible, allowing such issues to be included in the proposed online environment. This would be a precise but labour-intensive process, and would be used initially in specific constituencies to assess feasibility of the method.
  7. If the above methods were successful, it would be worth mapping the specific issues, e.g. tuition fees, to the more general topics, e.g. education, in order to create a taxonomy of issues and policies that could be useful for further development of the online environment.

Proposed methodology - user studies

The aim of user studies would be to assess the value of the content analysis and try to assess any relationship between voting behaviour and voter preferences in terms of policies and issues.

A well-known technique in the wine industry is blind tasting, in which tasters express a preference, rating or rank ordering about several wines without knowing the colour, grape, country of origin or date of production. Only later do they discover the names of the wines they preferred. This is generally considered an effective means of removing preconceptions. Similarly, the ASPECT online voting environment would aim to simulate blind voting: prospective voters would choose topics, issues and personal profiles, rather than parties or candidates, before finding out which party or person best matched their preferences. The validity of this correlation is limited - electoral literature can be misleading, pledges and promises can be broken, and vital personal information can be omitted - but it still has the potential to enthuse a proportion of voters who are inclined not to vote for any particular party.

As with the content analysis, the precise methods would be determined by research, so the steps proposed here are illustrative rather than definitive.

  1. Map topics and issues to each party, so that it is possible to infer how a preference for policies and issues ought in theory to influence voting.
  2. Ask users to rank general topics in order of importance to them.
  3. Ask users to state whether they agree or disagree with promises and actions on several specific issues.
  4. Ask users to rank order candidates based on a personal profile without knowing their name or political affiliation.
  5. Ask users which has most influence on their voting behaviour - general topic, specific issue or candidate profile
  6. Ask users how they actually voted in 2003 and how they intend to vote at the next election.
  7. Measure actual voting behaviour against the party and candidate preferences implied by the responses.

An example of such a blind voting test is given in the Appendix.

Outcomes

If the blind voting exercise were successful it would provide a rich set of data, including:

Even if the blind voting exercise turned out to be unsuccessful, or produced uninteresting results, the research as a whole would assess the feasibility and validity of extracting topics and policy issues from electoral literature, and the potential of interactive online techniques for addressing voter apathy.

An Informational Tool to Support Voters

In recent years trials of electronic voting methods have been held in the UK. At the 2002 local elections in England, electronic voting was piloted in nine local authorities, and electronic counting of votes in 15 [10]. If a system of electronic voting were to be widely deployed (either at Scottish or UK level) then there would be a case for developing complementary electronic information systems to inform and enhance the democratic process. Robertson (2002) argues that the design of electronic voting systems should not be focused solely on the casting and recording of ballots, but that systems should be designed to support information gathering and sharing, deliberation, decision making and voting.

'Electronic voting can be structured so that it increases a sense of community. This can be done by making voting one of many acts within an ongoing, online political dialogue among voters, parties, special-interest groups, candidates, and other interested entities.'

ASPECT could become a useful component of such an electronic voting system, or indeed of the existing system of voting. Given a current rather than historic archive based on prior deposit, and the provision of retrieval and analytical tools such as described above, ASPECT would offer a holistic and independent view and could facilitate political deliberation and dialogue in a way that individual party and candidate websites would not.

References

[1] Scotland Act. London: HMSO, 1998.
http://www.hmso.gov.uk/acts/acts1998/19980046.htm

[2] ASPECT: http://gdl.cdlr.strath.ac.uk/aspect/

[3] Glasgow Digital Library: http://gdl.cdlr.strath.ac.uk/

[4] Political Parties, Elections and Referendums Act. London: HMSO, 2000.
http://www.hmso.gov.uk/acts/acts2000/20000041.htm

[5] Mair, P. & van Biezen, I, "Party membership in twenty European democracies 1980-2000". Party Politics, Vol. 7, No. 1, 2001, pp. 5-21.
http://www.bham.ac.uk/POLSIS/department/staff/publications/mair&biezen_pp2001.pdf

[6] Scottish elections 2003: the results, who voted, who didn't. London: Electoral Commission, 2003.
http://www.electoralcommission.gov.uk/files/dms/Scotlandelectionsreport_final2_11440-8843__S__.pdf

[7] Rosema, M. Turnout and the two-fold function of elections. Presented at: European Consortium for Political Research Joint Sessions, Uppsala, Sweden, 13-18 April 2004.
http://www.essex.ac.uk/ecpr/events/jointsessions/paperarchive/uppsala/ws9/Rosema.pdf

[8] Mainwood, V. Letter to The Guardian newspaper, 29 April 2004
http://www.guardian.co.uk/letters/story/0,3604,1205643,00.html

[9] Robertson, S. Electronic voting in context: a collaborative ballot for digital government. Presented at: ACM 2002 Conference on Computer Supported Collaborative Work, New Orleans, USA, 16-20 November 2002.
http://www.cis.drexel.edu/faculty/robertson/research/CSCW2002Abstract.pdf

[10] Modernising elections: a strategic evaluation of the 2002 electoral pilot schemes. London: Electoral Commission, 2002.
http://www.electoralcommission.org.uk/about-us/modernisingelections.cfm


Example blind voting exercise

1. General topics

The table below shows the main general topics emphasised in the election campaign by four political parties.

Rank Party1 Party2 Party3 Party4
1 Education & training Health Crime and justice Government
2 Health & community care Economy & jobs Economy & jobs Education & training
3 Economy & jobs Business & industry Rural development Health & community care
4 Transport Education & training Education & training Rural development
5 Environment Crime and justice Environment Tourism
6 Crime and justice Fisheries Health Environment
7 Young people Tourism Older people Arts & culture
8 Sport Rural development Rural development Economy & jobs
9 Agriculture Environment Business & industry Planning & building
10 Business & industry Young people Transport Research

Which of these parties best matches your own view of the importance of different topics:

First choice: Second choice: Third choice: Fourth choice:

2. Specific issues

Listed below are policies on several specific issues that have been mentioned during the election campaign. For each one, please indicate a) which policy best matches your own view, and b) how important each issue is to you (1 = not important, 5 = very important). For example, if you thought student tuition fees should be abolished, but it is not important to you, then you would enter c) in the first column and 1 in the second column.

Issue Options Your choice Importance
Student tuition fees: a) no stated policy b) keep fees c) abolish fees    
Minimum wage: a) no stated policy b) £4 an hour
c) 50% of average male earnings for all over 16
   
Scottish independence: a) no stated policy b) no referendum
c) referendum within four years
   
Housing: a) no stated policy b) build or renovate 100000 houses a year    

3. Personal profiles

Assume that the five candidates listed below are standing for election in your constituency. Choose which one you would be most likely to vote for based on their personal profile, regardless of their party or policies.

a) Female, lives in the constituency with her husband. She has worked as an administrator in the National Health Service and for the Scottish Trades Union Congress. Currently Scottish Officer with a political party.
b) Male, 35, born in Glasgow, educated at Shawlands Academy, has degrees from Glasgow, Aberdeen and Queen's universities. He is an environmental biologist working at Glasgow University Department of Research and Enterprise.
c) Female, born and educated in Glasgow, attended Bearsden Academy, trained as a nurse at Glasgow Caledonian University. She lives in the constituency, works at a local NHS hospital and has been a local councillor for four years.
d) Male, lives in the constituency. He is a social care worker and trade union steward.
e) Male, journalist and historian. He writes a weekly political column for a Scottish newspaper and is a frequent broadcaster on radio and television. He has held academic positions in Scotland, Europe and America, including Strathclyde University. An author or editor of five books on the history of modern Scotland, he is an officer of the Eighteenth-Century Scottish Studies Society and a member of the Society of Authors. He relaxes by cooking and eating.

Which of these candidates would you be most likely to vote for:

First choice: Second choice: Third choice: Fourth choice: Fifth choice:

4. Influence of topics, issues and profiles

Please state which of above three factors (general topics, specific issues, personal profiles) is most likely and least likely to influence your voting behaviour:

Biggest influence: Least influence: