Read on Twitter
All of them! But here's some examples from my work and others https://twitter.com/ethanbdm/status/1267494825957429253
Everything to do with Polity https://twitter.com/benwansell/status/1229386394658836480?s=21
Aggregate electoral turnout (official turnout in the UK is underestimated by about 12 percentage points) and this error varies systematically across countries and over time https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3098436
The main post-materialism question in the World Values Survey is mistranslated in Chinese, Polish, Russian and possibly Arabic
https://www.nuffield.ox.ac.uk/politics/papers/2011/Jon%20Mellon_working%20paper%202011_08.pdf
https://www.nuffield.ox.ac.uk/politics/papers/2011/Jon%20Mellon_working%20paper%202011_08.pdf
The Barro-Lee cross-national educational attainment database has major discontinuities and inaccuracies https://twitter.com/jon_mellon/status/1095623453997232128?s=20
Promise keeping by governments as a measure of representation. It's typically measured as % of manifesto promises fulfilled. But this ignores that some promises are trivial and others central to the agenda. Accounting for this can completely change results
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3283813
https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3283813
99% of studies published using Google Trends. It usually doesn't validate against reliable source, but people just use it anyway https://jonmellon.com/2018/01/19/everybody-lies-but-google-trends-is-not-a-panacea/
Similarly, 99% of studies using social media data to draw inferences about the general population https://journals.sagepub.com/doi/full/10.1177/2053168017720008
Any studies of low frequency group or behaviour within surveys. This is a good example, but it's a wider phenomena: https://www.sciencedirect.com/science/article/pii/S0261379415001420
I'll add everything related to turnout in surveys. It's plagued by both overreporting and being particularly strongly correlated with non-response. This is further exacerbated by panel surveys.
Also, turns out the UK was mis-measuring net migration for decades. I suspect errors in migration statistics are rampant everywhere, but people just plug the numbers into models https://www.ft.com/content/56a3687e-c410-11e9-a8e9-296ca66511c9
In a completely different field, cosmologists aren't actually quite sure how old the universe is any more. Likeliest culprit is that the "standard candle" supernovas turned out not to be quite as standard as thought
Turns out we don't know how good different countries' education systems are because they differ wildly in how many students they exclude from taking PISA tests (Canada excludes 50% of its 15 year olds from the test whereas Finland and Japan exclude 10%)
https://link.springer.com/epdf/10.1007/s11092-020-09329-5?sharing_token=yt9UDjdn_Q8u2d8g4QDofPe4RwlQNchNByi7wbcMAY44Jf9Q2GsaUKlnv9LO8-ZIES5-K-PQS7l7YyxQXfz2pwFE0Ej4zuztSBM6UYRnCsahg4XMZZcBAJCtPcm7xCc6VCb_jzUit6gmQtN0TbEDW5816JKxuMNxQzYxAOs6M0w%3D
https://link.springer.com/epdf/10.1007/s11092-020-09329-5?sharing_token=yt9UDjdn_Q8u2d8g4QDofPe4RwlQNchNByi7wbcMAY44Jf9Q2GsaUKlnv9LO8-ZIES5-K-PQS7l7YyxQXfz2pwFE0Ej4zuztSBM6UYRnCsahg4XMZZcBAJCtPcm7xCc6VCb_jzUit6gmQtN0TbEDW5816JKxuMNxQzYxAOs6M0w%3D
