Awesome

European Opinion Polls as Open Data

This GitHub repository provides European opinion polls as open data.

Note: The poll files are being migrated to the APOPOD repository.

Read the article How “Simple Things” Quickly Become Complicated in Software Development to get some background information about the challenges to store opinion polls in tables.

Instructions for How to Contribute

In general, do a pull before you start editing a country poll file, so you get the most recent copy. Do a commit per country poll file, so that it's easier to merge and resolve conflicts, and consider to push or send in a pull request for each commit.

File Structure and Conventions

Each country poll file contains the following columns with the following conventions:

• Polling firm: name of the polling firm
  • If there are already other polls registered for that polling firm, reuse the name
  • If there aren't already other polls registered for that polling firm, try to find out the polling firm's proper name, i.e. the way they refer to themselves, e.g. on their web site or in their report
  • If more than one polling firm is involved in a poll, concatenate the names using commas (,) and and
• Commissionners: same as for the polling firms, or N/A if there's no commissionner for the poll
• Fieldwork Start: the start date of the fieldwork, formatted as YYYY-MM-DD
• Fieldwork End: the end date of the fieldwork, formatted as YYYY-MM-DD
  • Polls are sorted in descending order on Fieldwork End first, and Fieldwork Start after that
• Scope: the scope for the poll; the first letter is the code for the main scope of the poll, additional letters are the codes for the other scopes the poll can be used for; most polls will fall in the category NE, a national poll that can be used for European election projections too, some in the category EN, a European election poll that can be used for national projections too, and in case a poll has a separate national and European question, the national one should be labeled N, and the European one E; for Belgium, polls may be used for regional projections (R) too
• Sample size: the sample size of the poll:
  • An integer is the sample size was published by the polling firm or in the media
  • An integer in brackets if the sample size wasn't published, but the polling firm has a standard sample size, indicated a minimal sample size, or there have been other polls with published sample sizes by this polling firm, in which case the lowest number can be used
  • N/A if nothing is known about the sample size
• Participation: how many of the respondents were willing to reveal their voting intention for this poll, as a percentage, or N/A if it isn't known
  • If the sample size is a net sample size, use N/A, such that the actual sample size isn't further reduced
• Precision: the precision for the results; typically 1, 0.5 or 0.1; if the results are only expressed in terms of seats (like in the Netherlands), use S
• Results per party, as percentages, or N/A if the party wasn't polled or didn't exist at the time of polling
  • If you need a new column for a new party or electoral alliance, don't just add it, but request it to be set up. Otherwise the scripts producing the CSV files will start to fail.