I envisioned this column as a means for doing for Arkansas politics what the Indivisible Guide did for the resistance: to demystify how it works and to arm Arkansas Times readers with the information and tools they’d need to remake the system as they saw fit. Distinct projects with shared ideals. With the help of data scientist John Ray and digital strategist Jesse Bacon, the column’s first and second editions parsed a sea of electoral, demographic and polling data to build a foundation for understanding if and how the revolt in red states might spread to Arkansas.
Even to us, those initial results proved surprising — we didn’t expect to see so many opportunities for Democratic Party gains in Arkansas. But we also realized that we weren’t living up to our goal of making Arkansas politics and elections accessible to folks beyond the state’s cynical pundit class. To do so, we needed to do a better job of explaining how we built our model and how we plan to update it throughout the year.
In both our House and Senate models, we used eight variables to simulate what might happen if the 2018 election were held today. Only three variables — district-level Democratic vote share from the previous contested election, statewide Clinton share of the 2016 general election and statewide presidential approval — proved to be statistically significant. The others — district-level shares of the black voting-age population, Latino voting age population, residents holding a bachelor’s degree or higher, median household income and population density — all failed tests for statistical significance. The output our model generates is a simple win probability, or the odds of a Democrat winning a particular state legislative seat. This doesn’t tell the whole story, however.
Failing a statistical test for significance doesn’t necessarily render a variable insignificant. Indeed, the racial, education and income measures included are all cornerstones of electoral modeling in the political science literature. We know they matter and that’s why we’ll continue to include them. We used the 96 state House and 52 state Senate races that have transpired since Election Day 2016 as data points to calibrate our model — meaning we ran about 10,000 iterations or simulations to generate a modal-rounded seat prediction. That’s the number you’ve seen beside each state legislative district in our results (e.g. HD-84 [96 percent]). As more state legislative races are held over the course of this year, we’ll continue to treat them as data points against which our model can be tested.
For our next simulation — which we’ll run after candidate filing closes on March 1 — we’re introducing important changes to our Clinton vote share and presidential approval variables. Rather than statewide approval or vote shares, we’ll have district-level estimates for both variables. As far as we know, we’ll be the only ones with that information. We’ll also introduce a variable for candidate quality. Expect our projections to change a lot over the next nine months.
But we also know that there’s plenty we can’t or won’t measure in our model that could make a big difference on Election Day. Chief among those is campaign infrastructure, the grassroots network of people who knock the doors, make the calls and raise the money that wins elections. A central theme of state and national political coverage since President Trump’s election has been the growing resistance to his agenda, one that has already carried Sen. Doug Jones to victory in Alabama, flipped 15 Virginia House of Delegates seats and helped Democrats running for state legislative seats in 2017 outperform their predecessors by an average of 24 points. Those who insist that a wave isn’t building have not been paying attention.
To get a sense of how all this is playing out in Arkansas, I asked progressive leaders about their electoral plans. Democratic Party of Arkansas Chairman Michael John Gray remarked that, “One of the reasons I ran for chair was to rebuild our [candidate] support system. We’ve been reaching out to issues-based organizations who’ve already built strong grassroots networks across the state to try and mobilize them alongside us. … It would be arrogant of me to sit in Augusta … and say that I know what it takes to win everywhere — most of these groups know better than I do, so we’re just trying to support their work.” This is an important point, as no one could rebuild a statewide organizing infrastructure from scratch in a single cycle.
The good news is that the party won’t have to. All of the Indivisible group leaders I spoke with are working on voter registration, canvassing trainings and other get-out-the-vote strategies aimed at ensuring their electoral expertise is as great as their enthusiasm. Caitlynn Moses of Ozark Indivisible said, “Last year was just preparation for this election. We have 50 to 100 members who come to every meeting and are ready to get to work.” A few counties east, Boone County Indivisible’s Daniella Scott said, “There are so many more Democrats in Boone County than people think — everywhere I go, I run into people who voted for Congressman Womack because there was no other option and they’re fed up … and hungry for another option. We’ve picked up so many new tools over the last year and I can’t wait to apply them against [Womack] in November.” That relationship between establishment folks and progressive activists will be the thing to watch this year. Can the state’s Indivisible groups and progressive activists sharpen their campaign skills in time for November? Can the Democratic Party truly share power and let their grassroots lead? If so, there’s no limit to how big Arkansas’s blue wave can grow.