Maryland's congressional districts are widely considered to have been gerrymandered to favor the Democratic Party. This talk will examine the validity of this claim by creating an ensemble of nonpartisan maps that will serve as a baseline with which to compare the currently enacted plan. The
ensemble of maps will comply with historical and legal criteria for redistricting plans in Maryland. The ensemble, made up of 10,000 redistricting plans, is generated with a Markov Chain Monte Carlo algorithm. To compare the ensemble with the enacted plan, we simulate elections using
historical voting data and analyze election outcomes from the enacted plan against the ensemble's election outcomes. Our results suggest that the number of elected congressional representatives currently representing Maryland is typical. However, the structure of elections on the enacted map
is atypical of the ensemble and serves to make races less competitive.
1. Evaluating the Extent of
Gerrymandering in Maryland
Lisa Lebovici
Master’s in Statistical Science ‘19
Duke University
with Jonathan Mattingly, Greg Herschlag, Rob Ravier, Sam Eure, Rahul Ramesh
2. Maryland’s Congressional Districts, 1973-present
1973-1982 1983-1992 1993-2002
2003-2013 2013-present
https://en.wikipedia.org/wiki/Maryland%27s_congressional_districts
# Seats Won by Democrats
• 1973 - 1982: 4 to 7
• 1983 - 1992: 5 to 7
• 1993 - 2002: 4
• 2003 - 2013: 6 to 7
• 2013 - present: 7
3. “Part of my intent was to create a map
that, all things being legal and equal,
would, nonetheless, be more likely to
elect more Democrats rather than less.”
https://www.theatlantic.com/politics/archive/2017/06/how-deep-blue-maryland-shows-redistricting-is-broken/531492/
— Martin O’Malley (D), former Gov.
of Maryland (2007-2015)
4. Did the Democratic Party gerrymander the
state of Maryland?
How “well” did they do?
5. What information do we need to measure this?
● A non-partisan benchmark against which we can compare the currently
enacted plan
○ We can build an ensemble of legal districting plans
● Criteria for a compliant plan:
○ Equal population: “one person, one vote”
○ Satisfies the Voting Rights Act: minorities have the opportunity to elect a
representative of their choice
○ Contiguity
6. Is a district contiguous if it
crosses the Chesapeake Bay?
7. Given criteria, how do we build our ensemble?
● Using the aforementioned criteria, we can define a distribution on maps from
which we can subsequently sample compliant districting plans:
● Define , where are the districting criteria, is the
score for a particular criteria and is its weight.
● Use Markov Chain Monte Carlo to sample 10,000 plans from this distribution.
8.
9. 1. Select a districting
map from the ensemble
of sampled plans.
2. Obtain Dem. and Rep.
vote counts by precinct
from historical MD
elections (above: Pres 08)
3. Calculate number of
congressional seats
hypothetically won by
each party.
What do we do with these maps?
10. # Congressional Democrats
Elected
Election Outcomes
# Congressional Democrats
Elected
FractionofTotal
Ensemble
Democrats Elected based on 2008 Presidential
Votes
StatewideDemocraticVoteFraction(per
election)
Gov
2014
House
2014
Pres 2016
Pres 2012
Pres 2018
Senate
2012
House
2008
2011
11. Election Outcomes (cont.)
Ordered District (Least to Most
Democratic)
FractionofDemocratic
Vote Democratic Vote Fractions by
District
Pres 08 Gov 14
12. 0.50
0.55
0.60
0.65
0.70
0 1 2 3 4 5 6 7 8 9
Election Outcomes (cont.)
# Congressional Democrats Elected
StatewideDemocraticVote
Fraction
Gov 2014
House
2014
Pres 2016
Pres 2012
Pres 2018
Senate
2012
House
2008
Ordered District (Least to Most
Democratic)
FractionofDemocratic
Vote
House
2008
Pres 2016
Ordered District (Least to Most
Democratic)
FractionofDemocratic
Vote
13. Concerning Maryland’s map
“Behind closed doors, Democratic insiders and high-ranking aides referred to it
as ‘the 7-1 map.’ Hawkins… not only made it happen, but imagined an 8-0 map
that might have shut Republicans out of power altogether. That, however,
would have required spreading Democratic voters a little too thin and made
some incumbents slightly less safe; these congressmen were partisans, sure,
but they were also reluctant to risk their own seats.”
https://www.theatlantic.com/politics/archive/2017/06/how-deep-blue-maryland-shows-redistricting-is-broken/531492/
— Dave Daley, The Atlantic
14. How secure was the Democrats’ 7-1 Victory in
2012?
● We can define a map X to be more “stable” than a map Y if a smaller
Democratic vote fraction is needed to achieve the same election result.
○ 11.1% of plans in ensemble are more stable than the enacted plan for at least 7 of 14
elections.
○ 0.6% of plans in ensemble are more stable than the enacted plan for all 14 elections.
15. Nate Silver: Envisioning the 8-0 Map
● In the 2nd least Democratic
district, the 538 plan has
higher vote margins than the
enacted plan across all
elections, giving it a greater
chance of electing a
Democratic candidate.
House
08
Pres 16
538
PlanEnacted
Plan
Ordered District (Least to Most
Democratic)
FractionofDemocratic
Vote
Democratic Vote Fractions by
District
Pres 08 Gov 14
16. So, did the Democrats gerrymander Maryland?
● It’s a mixed bag:
○ In terms of election outcomes, not a whole lot changed. It’s not unreasonable to
assert that the Democrats would have won 7 out of 8 seats on many other maps.
○ But in terms of the structure of the map, the Democrats seem to have created a
cushion for that 7th seat, and in doing so made some districts less competitive.
17. Jonathan Mattingly
Thanks for listening!
Collaborators
Greg Herschlag Rob Ravier Lisa Lebovici Sam Eure Rahul Ramesh