Candidate Scoring System, Third Release

bykbog17d2nd Apr 20194 comments


Link to PDF report

Link to Excel model


The Candidate Scoring System (CSS) is a method for selecting preferred candidates in elections. It is based on Effective Altruist (EA) ethics and methodology. Of course, opposing political positions are still valid in the EA movement and there is room for respectable disagreement. Other people in the EA movement may have different understandings of the factual impacts of various political actions, and they may have different values regarding the appropriate goals of government. But we approach the central, most important policy question – how to maximize global well-being – by gathering opinions and research from authorities in a wide range of domains, then modeling them together with our own careful judgment to fill in the gaps.

CSS1 was released on March 5, 2019, establishing basic policy preferences and providing tentative scoring of presidential candidates. CSS2 was released on March 18 with deepened analysis of policy questions, more information about political candidates, more candidates under consideration, calculations of election probabilities and counterfactuals, and simplification into a single report with an accompanying Excel model. CSS3 contains a variety of minor improvements, including the addition of an “EA Lite” profile for people who want a restricted evaluation of candidates’ views on common EA issues.

The conclusions have remained mostly similar throughout this process, so we intend to stop working on CSS as further improvements seem unlikely to change the conclusions. It may be beneficial for someone else to take the lead in producing CSS4 in case a change of perspective could illuminate new issues. Otherwise, we are just going to wait until major new developments suggest a need for revisions.

This project is limited by the constraints of time and manpower against the vast breadth, depth and complexity of the problems that it tackles. Therefore, many arguments and evidence will be missing. This does not mean the project is necessarily wrong or biased, it just means we haven’t yet included as much content and research as we would like to. It is a work in progress and open to input from others. We are uncertain about much of this content, but we minimize hedging language for the sake of readability. If some relevant information is missing, please submit ideas and content to improve the next version – everything here is subject to revision and elaboration.

CSS is an independent volunteer project. Thanks to all who have helped.

Summary for Voters and Activists

CSS3 makes the following recommendations:

  • John Delaney should be supported if there are tractable opportunities to do so, particularly in Iowa. Everyone should make a $1 donation to his campaign.
  • Cory Booker should be supported if his campaign is a significantly more tractable target, if one gives higher priority to qualifications rather than long-run issues, if one ignores our assessment of candidates' personal character, or if one doubts Delaney’s electability more than we do.
  • Potential Republican challengers to President Trump should be encouraged and supported if a real chance appears, especially John Kasich.

Our recommendations are based on estimates of the expected value of changing the outcomes of the primary races. We approach this question by first estimating the desirability of each candidate as a potential president, yielding presidency scores. We then factor in the nomination and election chances of all their competitors to produce nomination scores representing the difference in the expected election outcome based on whether the candidate wins or loses in the primaries.