The Model Challenges: Gathering, evaluating, and aggregating Covid mortality models

Co-authored with Tara Slough et al.

In 2020, shortly after the onset of the global Covid-19 pandemic, we devised a set of Model Challenges to explore what our discipline’s inherited knowledge had to offer to accurately model (and understand) how governments would respond to the pandemic. Set up as a crowdsourced tournament, we first gathered models of political drivers of Covid mortality and then asked experts to assess which models would perform well or poorly. In this paper, we describe and take stock of this large scale collective effort. Our analysis suggests three main conclusions. First, the ability of political scientists to predict or explain which polities will react effectively to the pandemic and which not appears very limited. Second, even when our models are relatively successful, as a discipline we are not good at telling apart effective and ineffective models. Third, the best results are generated when models are combined. On the basis of these findings, we suggest that our discipline would benefit from the establishment of continued structured interactions that encourage multiple perspectives on phenomena of common interest and that aggregate ideas across researchers. As a discipline, we should sharpen our theories by seeking to predict future events instead of predicting the past.

Covid_2025.pdf

Download PDF • 4.54MB

Biometric Identification Machine Failure and Electoral Fraud in a Competitive Democracy

Co-authored with Eric Kramon, George Ofosu, and Luke Sonnet

This paper studies election fraud in the competitive but not fully consolidated multiparty democracy of Ghana. Results from a randomized field experiment are used to investigate the effectiveness of newly-introduced biometric identification machines in reducing election fraud in Ghana's December 2012 national elections. We uncover a non-random pattern to the frequent breakdowns of the equipment. In polling stations with a randomly assigned domestic election observer, machines were about 50 percent less likely to experience breakdown as they were in polling stations without observers. We also find that electoral competition in the parliamentary race is strongly associated with greater machine breakdown.

Machine malfunction in turn facilitated election fraud, including overvoting, registry discrepancies, and ballot stuffing, especially where election observers were not present. Our results substantiate that partisan competition may promote election fraud in a newly-established competitive democracy. They also show that domestic election observers improve election integrity through direct observation and also thanks to their second-order effects on election administration.

Recently discussed in The Monkey Cage.

Criminal Candidate Selection for the Indian National Legislature

Co-authored with Toke Aidt and Devesh Tiwari

Utilizing data on self-reported criminal charges lodged against candidates to the Fourteenth and Fifteenth Lok Sabha, India's lower house of representatives, in this paper we study the patterns of criminal candidate selection in 2004 and 2009 by India's political parties. Indian political parties are more likely to select candidates charged with criminal wrongdoing when confronting greater electoral uncertainty and in parliamentary constituencies whose populations exhibit lower levels of literacy. We model these findings formally to interpret them and we discuss the mechanisms that might underlie the patterns we uncover.

Recently discussed in The Economist, in The Wall Street Journal (India Edition), and on the BBC India.