Corrections and Refinements to the Database of Political Institutions’ yrcurnt Election Timing Variable

The yrcurnt variable in the Database of Political Institutions (Beck et al. 2001, updated in 2013)1–DPI–is a regularly used measure of government election timing. For example, Alt, Lassen, and Wehner (2014) use the variable in their recent study of fiscal gimmickry in Europe. They find that fiscal gimmickry–straying from accepted accounting standards–is more common directly before elections (and in countries with weak fiscal transparency).

Because the DPI’s yrcurnt variable is so regularly turned to for testing how election timing affects governments’ choices, it is especially important that it be reliable and valid. However, the variable in the current (2013) DPI release has a number of issues that this note aims to correct.2

Variable definition

Before looking at problems in the yrcurnt variable, let’s reiterate how the variable is defined. The DPI Codebook3 (p. 4) defines the yrcurnt variable as the years left in a country’s chief executive’s current term such that:

“a ‘0’ is scored in an election year, and n-1 in the year after an election, where n is the length of the term. In countries where early elections can be called, YRCURNT is set to the de jure term limit or schedule of elections, but resets in the case of early elections.”

Issues

The original variable has a number of issues that make it problematic for studying how election timing impacts government policymaking. The first set of concerns are clearly errors that make the variable a less reliable measure of the concept defined above. These errors can be straightforwardly corrected. The second set of concerns have to do with issues that bring into question the variable’s validity in a number of cases for studying how election timing shapes policymaking. I argue that simple refinements can be made to improve the variable’s validity in these cases.

In this note I focus on the EU-28 countries (all 28 current European Union member states). Problems with the variable may exist for a wider range of countries. Given the depth of problems with the yrcurnt variable discussed here, it would certainly be worthwhile in future work to reevaluate the variable for the full breadth of countries covered by the DPI.

I used the European Election Database4 to find correct election years. This information was cross-checked and supplemented with individual country-election articles on Wikipedia.5

Errors

There are many instances in the DPI where election years are not recorded as 0. In other cases non-election years were mis-coded as 0. In some cases incorrect statutory election timing was also given. The following table lists these errors:

Country Errors in the DPI yrcurnt variable for the EU-28
Belgium Missing the 2010 election.
Croatia Incorrect election timing for the 1995 and 2000 elections. Also, the DPI incorrectly classifies 1991 as 4 years from the next election. Croatia gained independence in 1991 and its first election as an independent country was scheduled for the following year. So, 1991 should be coded as 1 year from the next election.
Denmark Missing the 2001 and 2007 elections.
Estonia Incorrect election timing for the 1995, 1999, 2003, 2007, and 2011 elections. Also counting originally started at 4, but should start at 3 as there is a 4 year term limit (not 5).
Germany Missing the 2005 election.
Greece Missing the 2007, 2009, 2012 election years.
Ireland Missing the 2011 election.
Italy Missing the 2008 election.
Latvia Missing the 2006, 2010, 2011 election years.
Netherlands Missing the 2003 and 2006 elections.
Portugal Missing the 1979, 1999, and 2011 elections.
Slovakia Missing the 2012 election.
Spain Missing the 1989, 1996, and 2011 elections.
United Kingdom Missing the 2001 and 2005 elections.

Refinements

For a number of countries the elections recorded are for largely figurehead presidents. This can affect both when elections are recorded and how many years are given until the next election as figurehead presidents often have longer terms than parliaments. In these cases the current yrcurnt variable is not a valid measure of government election timing.

Some countries are less clear-cut because they have semi-presidential systems. Nonetheless, in a number of these cases the prime minister is the leader of the government and largely sets the domestic policy agenda. These powers are most relevant for studying topics such as public budgeting.

The following refinements should be made to create a more valid indicator for research on how election timing affects policymaking:

Country Refinements to make to the DPI yrcurnt variable for the EU-28
Austria Use parliamentary rather than (figurehead) presidential elections.
Lithuania Use parliamentary rather than presidential elections. Lithuania is a semi-presidential system where the president appoints the PM, the legislature’s approval is needed. The PM is more responsible for domestic policy.
Romania Romania is semi-presidential where the president appoints the PM, but the PM must be approved by the parliament. The PM is both the head of government and sets the legislative agenda. Before 2008, presidential and parliamentary elections occurred in the same year. Since then they have diverged, from which point the parliamentary elections should be used.
Slovenia Use parliamentary rather than (figurehead) presidential elections.

Available updated data

In sum, of the EU-28 countries there are clear errors for 14 countries’ election timing in the DPI yrcurnt variable. There are a further four countries where a more valid measure of election timing could be created by focusing on parliamentary rather than presidential elections.

A data set that implements these corrections and refinements for the EU-28 countries can be found on GitHub.6 The data (including both the original and corrected versions of the variable) can be downloaded directly into R using the repmis package (Gandrud 2015) with the following code:

yrcurnt_corrected <- repmis::source_data('http://bit.ly/1EM8EVE')

This file also has updated election timing data through 2013.

References

Alt, James, David Dreyer Lassen, and Joachim Wehner. 2014. “It Isn’t Just About Greece: Domestic Politics, Transparency and Fiscal Gimmickry in Europe.” British Journal of Political Science 44 (04): 707–16.

Beck, Thorsten, George Clarke, Alberto Groff, Philip Keefer, and Patrick Walsh. 2001. “New Tools in Comparative Political Economy: The Database of Political Institutions.” World Bank Economic Review 15 (1): 165–76.

Gandrud, Christopher. 2015. repmis: Miscellaneous Tools for Reproducible Research with R. http://cran.r-project.org/web/packages/repmis/.


  1. See: http://go.worldbank.org/2EAGGLRZ40. Accessed February 2015. All corrections and refinements discussed here refer to this version of the database.
  2. Note that Alt, Lassen, and Wehner’s (2014) substantive findings hold up when using the correct data presented here, though the estimated magnitudes of the effects are reduced. See the replication repository for more details: https://github.com/christophergandrud/Alt_et_al_2014_Replication.
  3. The DPI Codebook can be found at: http://siteresources.worldbank.org/INTRES/Resources/469232-1107449512766/DPI2012_Codebook2.pdf. Accessed February 2015.
  4. See http://www.nsd.uib.no/european_election_database/country/. Accessed February 2015.
  5. See https://www.wikipedia.org/. Accessed September 2014 and February 2015.
  6. See https://github.com/christophergandrud/yrcurnt_corrected/tree/master/data.

About christophergandrud

I am a lecturer at City University London and a post-doctoral fellow at the Hertie School of Governance
This entry was posted in Data, Replication and tagged , , . Bookmark the permalink.

One Response to Corrections and Refinements to the Database of Political Institutions’ yrcurnt Election Timing Variable

  1. christophergandrud says:

    I’ve now corrected the yrcurnt variable for most other OECD countries as well: http://christophergandrud.blogspot.de/2015/06/more-corrections-the-dpis-yrcurnt.html

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s