Closed mzhao80 closed 1 year ago
Validation plots for additional constraints:
Hispanic majority-minority districts: Aimed for two (the number under the enacted plan.) Only achieves two on some of the simulated plans; finding quite a bit of difficulty getting any further.
Partisanship by HVAP
Competitiveness: Secondary constraint only to be considered after all others.
Analysis looks great.
Tagging @christopherkenny to see about VRA performance—cf. #91 for the 2020 comparison, which does look a tad better
This might be sufficient, as there seems to be some shift between 7 and 8. Can we get a count here?
In Arizona, districts must:
We enforce a maximum population deviation of 0.5%. We add a county/municipality constraint, as described below. We add a hinge Gibbs constraint targeting two majority-HVAP districts, one within Maricopa County and one outside of it (as exist in the enacted plan.) However, not all plans are guaranteed to have two majority-HVAP districts.
Data for Arizona comes from the ALARM Project's Redistricting Data Files.
No manual pre-processing decisions were necessary.
We sample 32,000 districting plans for Arizona across four independent runs of the SMC algorithm, and then thin the sample to down to 5,000 plans. To satisfy the Voting Rights Act constraint, we run the simulation in two steps.
We target a Hispanic-majority district outside of Maricopa County (HVAP 50-55%). We avoid splitting municipalities in this region.
We target 1 Hispanic-majority district in Maricopa County (HVAP 50-55%), and only keep plans where the district with the second-highest HVAP exceeds 30% (including the districts outside Maricopa County).
To balance county and municipality splits, we create pseudocounties for use in the county constraint. These are counties outside Maricopa County and Pima County, which are larger than a congressional district in population. Within Maricopa County and Pima County, municipalities are each their own pseudocounty as well. Overall, this approach leads to much fewer county and municipality splits than using either a county or county/municipality constraint.
SMC: 5,000 sampled plans of 9 districts on 2,224 units
`adapt_k_thresh`=0.985 • `seq_alpha`=0.5
`est_label_mult`=1 • `pop_temper`=0.05
Plan diversity 80% range: 0.58 to 0.86
R-hat values for summary statistics:
second_hisp pop_overlap total_vap plan_dev comp_edge comp_polsby pop_white pop_black
1.006106 1.004245 1.005938 1.009820 1.007458 1.008915 1.011136 1.026487
pop_hisp pop_aian pop_asian pop_nhpi pop_other pop_two vap_white vap_black
1.034838 1.003699 1.008666 1.015902 1.014525 1.004198 1.011367 1.031177
vap_hisp vap_aian vap_asian vap_nhpi vap_other vap_two pre_16_rep_tru pre_16_dem_cli
1.036375 1.003935 1.008228 1.013280 1.010103 1.009179 1.008768 1.009582
pre_20_dem_bid pre_20_rep_tru uss_16_rep_mcc uss_16_dem_kir uss_18_rep_mcs uss_18_dem_sin uss_20_dem_kel uss_20_rep_mcs
1.003445 1.005762 1.009399 1.017345 1.009997 1.006318 1.003671 1.009825
gov_18_rep_duc gov_18_dem_gar atg_18_rep_brn atg_18_dem_con sos_18_rep_gay sos_18_dem_hob adv_16 adv_18
1.010000 1.011501 1.011676 1.009417 1.012213 1.007467 1.014971 1.008400
adv_20 arv_16 arv_18 arv_20 county_splits muni_splits ndv nrv
1.003630 1.009062 1.012291 1.006772 1.003734 1.004238 1.008126 1.008690
ndshare e_dvs pr_dem e_dem pbias egap
1.002197 1.002333 1.001297 1.007189 1.015744 1.005142
Sampling diagnostics for SMC run 1 of 4 (8,000 samples)
Eff. samples (%) Acc. rate Log wgt. sd Max. unique Est. k
Split 1 5,498 (68.7%) 21.4% 0.46 3,734 ( 74%) 1
Split 2 5,045 (63.1%) 35.0% 0.82 4,416 ( 87%) 1
Split 3 4,702 (58.8%) 17.0% 0.91 4,194 ( 83%) 3
Split 4 4,907 (61.3%) 13.3% 0.90 4,032 ( 80%) 3
Split 5 4,253 (53.2%) 4.5% 0.85 3,650 ( 72%) 3
Resample 1,656 (20.7%) NA% 2.17 3,477 ( 69%) NA
Sampling diagnostics for SMC run 2 of 4 (8,000 samples)
Eff. samples (%) Acc. rate Log wgt. sd Max. unique Est. k
Split 1 5,496 (68.7%) 21.3% 0.45 3,763 ( 74%) 1
Split 2 4,910 (61.4%) 35.7% 0.82 4,389 ( 87%) 1
Split 3 4,881 (61.0%) 17.0% 0.92 4,151 ( 82%) 3
Split 4 5,174 (64.7%) 18.5% 0.86 4,097 ( 81%) 2
Split 5 4,315 (53.9%) 6.5% 0.83 3,699 ( 73%) 2
Resample 1,518 (19.0%) NA% 2.14 3,629 ( 72%) NA
Sampling diagnostics for SMC run 3 of 4 (8,000 samples)
Eff. samples (%) Acc. rate Log wgt. sd Max. unique Est. k
Split 1 5,560 (69.5%) 21.1% 0.46 3,707 ( 73%) 1
Split 2 5,134 (64.2%) 18.9% 0.79 4,333 ( 86%) 3
Split 3 4,928 (61.6%) 23.5% 0.89 4,191 ( 83%) 2
Split 4 4,634 (57.9%) 10.1% 0.88 4,075 ( 81%) 4
Split 5 4,180 (52.3%) 4.8% 0.92 3,665 ( 72%) 3
Resample 1,238 (15.5%) NA% 2.17 3,416 ( 68%) NA
Sampling diagnostics for SMC run 4 of 4 (8,000 samples)
Eff. samples (%) Acc. rate Log wgt. sd Max. unique Est. k
Split 1 5,523 (69.0%) 11.8% 0.45 3,731 ( 74%) 3
Split 2 4,979 (62.2%) 26.3% 0.81 4,335 ( 86%) 2
Split 3 4,760 (59.5%) 17.3% 0.94 4,174 ( 83%) 3
Split 4 4,977 (62.2%) 13.3% 0.89 4,048 ( 80%) 3
Split 5 3,770 (47.1%) 6.5% 0.86 3,681 ( 73%) 2
Resample 1,057 (13.2%) NA% 2.14 3,283 ( 65%) NA
• Watch out for low effective samples, very low acceptance rates (less than 1%), large std. devs. of the log weights (more than 3 or so), and low numbers of unique plans. R-hat values for summary statistics should be between 1 and 1.05.
TODO
lines from the template code have been removedenforce_style()
to format my coderedist_map
and redist_plans
objects, and summary statistics) have been edited@CoryMcCartan @christopherkenny
Validation plots for additional constraints:
Hispanic majority-minority districts: Aimed for two (the number under the enacted plan.) Only achieves two on some of the simulated plans.
Two-party vote by HVAP:
The districts with the highest HVAP clearly perform in electing. In the districts with the second-highest HVAP, all had greater than 30% HVAP (by construction) and all had a greater than 30% Democratic share of the two-party vote. 74.9% saw the Democratic share of the vote exceed the Republican share of the vote.
Competitiveness: Secondary constraint only to be considered after all others.
Looks better! @christopherkenny what do you think?
Looks good to me!
Redistricting requirements
In Arizona, districts must:
Algorithmic Constraints
We enforce a maximum population deviation of 0.5%. We add a county/municipality constraint, as described below. We add a hinge Gibbs constraint targeting two majority-HVAP districts, one within Maricopa County and one outside of it (as exist in the enacted plan.) However, not all plans are guaranteed to have two majority-HVAP districts.
Data Sources
Data for Arizona comes from the ALARM Project's Redistricting Data Files.
Pre-processing Notes
No manual pre-processing decisions were necessary.
Simulation Notes
We sample 32,000 districting plans for Arizona across four independent runs of the SMC algorithm, and then thin the sample to down to 5,000 plans. To satisfy the Voting Rights Act constraint, we run the simulation in two steps.
1. Simulate three districts outside of Maricopa County
We target a Hispanic-majority district outside of Maricopa County (HVAP 50-55%). We avoid splitting municipalities in this region.
2. Simulate six more districts in the remainder of the map
We target 1 Hispanic-majority district in Maricopa County (HVAP 50-55%).
To balance county and municipality splits, we create pseudocounties for use in the county constraint. These are counties outside Maricopa County and Pima County, which are larger than a congressional district in population. Within Maricopa County and Pima County, municipalities are each their own pseudocounty as well. Overall, this approach leads to much fewer county and municipality splits than using either a county or county/municipality constraint.
Validation
Checklist
TODO
lines from the template code have been removedenforce_style()
to format my coderedist_map
andredist_plans
objects, and summary statistics) have been edited@CoryMcCartan