@freesystemslab The swarm-based incentive scheme reduced the share of swarms converging on the false belief from over 90% down to roughly 65%.

Follow-up to our @freesystemslab experiment this week: Rewarding AI agents for the swarm’s performance rather than their own makes the swarm less likely to converge on a false belief. But even with the right incentives, agents don’t always figure out how to avoid getting it wrong

@jasonmfletcher I ran something similar a few years back comparing the age distribution of members of congress to the U.S. adult age distribution in each census year. That is essentially flat with a small decline in relative member age in the late 20th century that recovers by 2000.

@DjsokeSpeaking @masonreece13 You can see the pattern in bethel and trump vote share across counties—bethel underperformed in heavily pro-Trump counties and over-performed in heavily pro-Harris counties.

After selecting weights that put counties that did not apply for the grants on the same trajectory as recipients before 2020, we find that the grants did not substantially increase turnout or Democratic vote share in 2020. (We're using synthetic diff-in-diff weights.)

Using administrative data on which counties applied for and received the grants, we document that counties that supported Clinton in 2016 were substantially more likely to apply for and receive a grant.

The largest private election administration grant program in 2020 (~$350M) was run by the Center for Tech and Civic Life and funded by Mark Zuckerberg and Priscilla Chan. This funding has been highly controversial with 24 states banning future grants.

Private donors gave $350M+ to support election administration in 2020. Many critics argued that these grants swung the election to Biden. In a new paper, @Apoorva__Lal and I find that these grants did not provide Democrats a noticeable advantage.

bit.ly/lal_thompson_e…

@mattyglesias Gotta select 2011 Hillary Clinton

@samesfandiari it all goes back to this experiment cambridge.org/core/journals/…

@LachlanMcNamee @cesicruz

Our estimate of the effect in 2020 is similar to all earlier years. Still, it's important to note that this analysis is not a prediction. Election officials may change their behavior in the future.

We also find that Democratic and Republican officials oversee elections with similar turnout rates and choose similar policies.

Our result is not limited to close elections. Counties that voted for the D presidential candidate at about the same rate before the last clerk election continue to vote for D presidential candidates at the same rate whether they elected a D or R election official.

We estimate that narrowly electing a Democratic clerk increases Democratic presidential vote share by 0.3pp or less across our regression discontinuinty specifications. Our preferred estimator has 80% power to detect effects as small as 1.7pp.

To estimate how much partisan election officials advantage their party, we collected a new dataset of over 5,800 local election official elections.

How much do local election officials tilt elections in their party's favor? In a new paper, @Jferrer505, @GeynIgor, and I find that these officials improve their party's vote share by a small amount at most.

Comments welcome!

bit.ly/elec_admin

@matt_blackwell I'm just happy to be here

@joshmccrain One of Frisell youtube.com/watch?v=uBtjKT…
Another of Brian Blade youtube.com/watch?v=Z_orJU…

Columbus used the same system

I'm not sure who needs to hear this, but Cleveland had ranked-choice voting in 1915

@mattyglesias Wait til you hear about ketchup youtube.com/watch?v=9iYwUh…

If you thought American politics was weird, see H'Angus the Monkey for mayor. en.wikipedia.org/wiki/Stuart_Dr…

May all of your abstracts be this perfect journals.uchicago.edu/doi/abs/10.108…

The employee who designed the 1993-1994 South Carolina election report deserves an award for their clipart craftspersonship.

Cats like boxes, exhibit 531b

@FeyaadAllie perhaps it's theme_dubois? mentalfloss.com/article/92460/…

@prof_mirya @mlatner @emayfarris @MichaelZoorob My data covers almost all counties with partisan sheriff elections and more than 100k residents. I also gathered results for all counties in 15 states with sheriff elections. 55% of the races had two or more candidates. 37% had a Rep and Dem. More at bit.ly/2mFnZ9x

Hudson County, NJ threw out their precinct-level election results from 1980. Jersey City lost them in Hurricane Sandy. The State Archive? Got 'em on microfiche. And who has an hour to kill during a trip to Trenton? Somehow, my father from North Carolina.

In a new working paper, I find that the local elected officials most responsible for determining local immigration enforcement--sheriffs--make similar choices regardless of party.

Comments welcome!

bit.ly/2mFnZ9x

I'm about four years late on this, but this is a must read from @soashworth, Berry, and @ethanbdm clarifying the link between causal inference and formal theory.
home.uchicago.edu/~sashwort/bigd…

There were 3,468 people in the 2010 census with the last name Casias. Two of them, not in the same nuclear family, ran against one another for Colfax County Sheriff in 2010.

@sophie_e_hill Meanwhile, David Card was making some beautiful tables in the 80s.

adolescence in four acts

@annascodel @AnnHollingshead Apparently, A. Craig thought it was totally cool in 1609.

@AnnHollingshead @annascodel They drop the hammer with this Jefferson example

Is the flu going around? I think San Francisco has a fever.

I'm learning from @JohnFPfaff that getting saucy in your endnotes is a good way to keep people coming back to read the sources.

Found you, Mr. Bailey and Mr. Rigney! You'll make quite a nice point in my RD.

I don't think Clark understands how scrabble works

@Spotify Is Hank Williams Jr really more relevant to "Hank Williams" than Hank Williams?