In many parliamentary systems the timing of the next election is at the discretion of the current
government. Leaders need not wait until the end of their term, rather they are free to call
elections when it is advantageous to them-- when they expect to win. This proposal outlines a
theoretical and empirical research design to explore when governments call elections and how
the timing of elections influences the electoral result.
The decision to call an election is modeled using game theory. At each stage leaders must choose
between calling an election or waiting. If they wait then the voters get a further opportunity to
assess the government's performance. Thus, a popular government that waits risks having its
popularity undermined by poor outcomes.
The decision to call an election rather than wait depends upon a leader's beliefs about her
electoral future. Suppose that the incumbent leader is currently popular and trying to decide
whether to call an early election. If the leader believes that her party is competent and has the
appropriate policies to address her nation's problems then she risks little by waiting. Her party is
likely to perform well which enhances, rather than diminishes, her electoral fortune. Yet, for a
leader less confident in her abilities, waiting jeopardizes her electoral prospects. Anticipating
poor performance, she expects her future popularity is likely to decline if she waits. Hence it is
the least competent leaders, those who do not expect to perform well, that have the greatest
incentive to call an early election.
The inherent feature of the model is that leaders choose whether to call elections based on their
expectations about future performance. This modeling platform is flexible to a variety of
assumptions. One modeling choice is that a leader's expectation could be simply based on her
assessment of government competence. Alternatively, a leader's beliefs could stem from early
access to information. A particularly important form of this is when the government has
manipulated policy instruments for short term gains (political business cycles).
Although it is the least competent leaders that always have the strongest incentive to call an early
election, there are conditions under which even competent leaders do not want to delay. These
conditions, such as size of majority, unity of opposition, and level of public opinion, generate
hypotheses about when early elections are likely to occur. Whether these factors do indeed hasten
elections can be empirically tested using duration analysis.
Since the least competent leaders are the ones with the least incentive to wait, elections called
earlier than the public expects signal the incumbent's doubts about future performance. In terms
of the empirical model, when elections occur much earlier than anticipated, i.e. the estimated
hazard is low, then the government is likely to suffer a decline in popular support. In addition,
the theory also predicts that government performance is likely to be poor following unexpectedly
early elections. This project provides more than empirical estimates of when elections occur, it
develops and tests theoretical linkages between election timing, government performance and
This project has generated several conference paper, the most recent of which is “Election Timing in Majoritarian Parliaments.” (Presented at APSA meeting 2000, and the Wallis conference in Political Economy, University of Rochester, Rochester NY.) Available in pdf.
I have also produced a book manuscript, "Election Timing." Available
The following archives contain the data and code for the analysis in the above book (This data is an update of that used in the paper.) The ReadMe.txt file outline the basic structure of the data and files.
The archive workdata_2001.zip contains the main data file in STATA 7 format.
The archive Election_timing_data2001.zip contains all the individual data files required to assemble the main data. For example, there are separate data files for the key political events, change in seat shares, change in party allegiances, public opinion, files for various economic measures, and various stock data files. The archive also contains code which contructs these the main data file from these individual files. It is much easier to update these individual files and recompile the main data file rather than try and directly update the main file.
The archives newspaper_stories.zip and times_election_stories.zip contain the newspaper count data. These data are counts of the number of newspaper stories relating to next election. The first archive is compiled from three different newspapers (The Financial Time, The Guardian and The Independent) using Lexis/Nexis. The second archive is complied from "The Official Index to The Times, 1906-1980" which is available in an electronically searchable form at Historical Newspapers Online (http://historynews.chadwyck.com/).
The archive Election_timing_analysis_2001.zip contains STATA 7 do.files that will replicate all the analysis in the book manuscript.
If you use these data you should feel obligated to cite the original sources for the data (all contained in the manuscript above). Further, I would appreciate it if you provide me with any updates and/or corrections. The data come 'as is' with guarantees. Please keep me updated if you find any interesting results or glaring errors in the data.