Candidate Scoring System, Third Release

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.