@lsegovernment.bsky.social

For help with this work, thanks especially to @thaliagerzso.bsky.social @gofosu.bsky.social @elliottdgreen.bsky.social @jeroengunning.bsky.social Adam Harris, Andy Harris, Richard Klein, Amanda Domingues, Michael Wahman, Nandini Patel, our late friend Francisco Cantú, and many others.

We conclude by emphasizing the administrative side of election management. In a time of declining democracy assistance and electoral trust, we shouldn’t forget the crucial role that human capacity and training play in making democracy work and ensuring election integrity.

We identify several exogenous sources of tallying complexity. Leveraging arbitrary registration cutoffs, we show that increases in polling station streams substantially increased rates of edits. Similarly, more complex ballot reconciliation procedures increased edits.

Alternatively, if human error and administrative capacity were major causes of edits (as claimed by many presiding officers), we would expect edits to be more common when the complexity of the tallying was high.

We start by noting that patterns of irregularities do not support most fraud claims. We code edits in 5,002 polling station results sheets and compare to candidate totals in official and independent tallies. We reject plausible hypotheses about how edits might have facilitated fraud.

Most election tallies have error, often fuelling accusations of fraud. So how do we distinguish between error and fraud? Johan Ahlbäck and I have a new @BJPS paper examining this question in the context of the famously disputed and annulled 2019 Malawi election.