Datasets I have released, and interactive charts built from the data behind my research. Sources and caveats for each chart are under Sources and notes below it.
Datasets
Frankenstein’s Bills: Replication Data
Zenodo, 2026 · CC BY 4.0
Raw and processed legislative text with section-level disjointedness measures for roughly 1.75 million legislative sections from the United States, the European Union and France, merged with lobby spending by sector and issue salience. Accompanies Frankenstein’s Bills.
Dataset (DOI: 10.5281/zenodo.20749943) · Code
Legislation
The Frankenstein Index
Which laws stitch together the most unrelated policy areas? Every section in the Frankenstein’s Bills data is classified into one of 17 policy topics. Each bar is one law, split by its sections’ topics. Hover over a segment for details.
Laws with at least 10 classified sections. Sections with no identifiable topic are left out. French texts come from English translations and are grouped by their official number, so French texts without a number in their title are not ranked. EU decisions ratifying broad international agreements naturally top the EU list. Hover over a name for the full title.
Laws, section by section In progress
I am extending the Frankenstein’s Bills data to more countries, splitting every enacted law into its sections. The map shows how far collection has got and how differently legislatures package their laws. Pick a measure, then click a jurisdiction for its laws per year.
Work in progress: collection is still running for several countries, so the numbers will change. Only laws with at least one section are counted. “Laws collected” covers every year collected; all other measures use a comparable subset in every country: laws passed by the legislature (no executive decree-laws, legislative decrees or ordonnances; EU acts only when adopted with the European Parliament), in their text as enacted, from 2008 to 2024. Canada has only consolidated (current) text, so it has no comparable laws. “Amending acts” is read from titles, so it slightly understates the share. Divergence compares each section with the law’s title using overlapping four-letter chunks of words, so inflected and compound-word languages (Finnish, German, Estonian) are scored on the same footing as English; “within law” compares each section with the rest of its own law instead of the title.
Public opinion
What Americans think is the most important problem, 1939–2018
Since 1939, pollsters have asked Americans some version of “What is the most important problem facing this country today?” The Most Important Problem Dataset (Heffington, Park and Williams, 2019) harmonises the answers into one annual series, which I use as the salience measure in Frankenstein’s Bills. Hover over any year for the full breakdown.
Shares of respondents per year, rescaled to sum to 100%. “Everything else” combines morality, youth issues, international economic relations, other problems, and don’t know/refused.
Who do people trust less?
Euroscepticism usually gets read as distrust of Brussels. But in much of Europe people trust the European Parliament more than their own parliament, including in countries that vote heavily for anti-EU parties. Each row is a country. The line runs from trust in the national parliament to trust in the European Parliament, and the bar on the right is the anti-EU vote at the European Parliament election.
Trust: weighted country means from the European Social Survey, rounds 4–6 (2008–12) and 7–9 (2014–18), on a 0–10 scale. Only countries with at least 300 respondents and both questions are shown. The ESS asks about trust in parliament, not in the government of the day. Anti-EU vote: the 2009 and 2019 European Parliament elections from EU-NED, counting parties that the Chapel Hill Expert Survey rates below 3.5 on its 1–7 EU integration scale. This includes governing parties such as Fidesz and PiS, so in Hungary and Poland an anti-EU vote is partly a vote for the national government. In the regional scatter under the EU spending map, the gap between the two trust measures can be plotted against the quality of local government.
Public money
Where the EU’s money goes
How much does the EU spend in each region, per resident, and do the regions that receive more like the EU more? Choose cohesion policy, the CAP’s rural development pillar or farm subsidies, then press play to follow the money from 1989 to 2022 as the EU enlarges. Below the map, compare each region’s spending, or the quality of its local government, with its anti-EU vote, its support for EU integration, and its trust in national and European institutions.
Cohesion and rural development: EU payments by region and year from the European Commission’s historic regionalised payments data, smoothed with a three-year moving average because payments arrive in lumps as programmes open and close. The data starts with the 1989–93 programmes and ends in 2022. Farm subsidies: subsidies in Eurostat’s regional agricultural accounts, which also include national aid and some rural development payments, so they overlap with the rural development layer and should not be added to it. Where a country has no regional series, its national average is shown. Per resident figures use Eurostat population (the earliest available year stands in for earlier years), and regions are shown only while their country is a member. Regions use the 2016 boundaries, and when a region’s boundaries changed later, each successor region gets its parent’s value per resident.
Attitudes and voting (scatter). Anti-EU vote: European Parliament election results by region from EU-NED (Schraff, Vergioglou and Demirci), 2009 for 2007–13 and 2019 for 2014–20. A party counts as anti-EU if the Chapel Hill Expert Survey places it below 3.5 on its 1–7 EU integration scale, using the wave nearest the election and linking parties through Party Facts. Parties CHES doesn’t rate (about 13% of the vote in the median region) count as not anti-EU. Ireland and Slovenia only report national results. Support for EU integration and trust in the national and European parliaments: weighted regional means from the European Social Survey, rounds 4–6 (2008–12) and 7–9 (2014–18), for regions with at least 50 respondents. Germany and the UK are only coded by larger NUTS-1 region. Local government: the European Quality of Government Index (2010 and 2017), which measures how people rate the quality, impartiality and honesty of their regional and local public services. It is not a direct measure of trust, and it has no UK scores. Hollow dots mark values that only exist for a wider area. These are left out of the within-country view, which subtracts each country’s average from both axes. All correlations are descriptive: poorer regions receive more money and differ in many other ways.