Case Study 15.2: The 2017 Alabama Senate Race — Campaign Effects at the Margin

Background

The December 2017 Alabama Senate special election between Democrat Doug Jones and Republican Roy Moore is one of the most analytically interesting elections in recent American history for a simple reason: Jones won in one of the most Republican states in the country, in a special election (with low baseline turnout), during a Republican presidential administration. If there was ever a case where campaign effects appeared to dominate structural factors, this is a candidate.

Alabama's partisan landscape in 2017 was extreme by any measure. Donald Trump had won the state by 28 percentage points in 2016. No Democrat had won a Senate seat in Alabama since 1992. The structural model for this race, run without reference to the specific candidates, would have predicted an easy Republican victory by 15–25 percentage points.

Jones won by 1.7 percentage points.

What Happened: The Components of Surprise

To understand whether this was a campaign effects story or a candidate quality story or something else, we need to decompose the swing.

Roy Moore's disqualifying liabilities. Moore had been removed from the Alabama Supreme Court twice for defying federal court orders. During the campaign, multiple women accused him of sexual misconduct with teenage girls when Moore was in his 30s. The accusations were credible and corroborated. Moore denied them and had the support of the Trump White House. But the allegations created a context in which many traditional Republican voters faced a genuine dilemma: vote for a deeply flawed candidate or cross party lines.

Doug Jones's campaign organization. Jones ran an aggressive campaign that invested heavily in Black voter turnout infrastructure — a demographic that represents approximately 27 percent of Alabama's registered voters but that typically turns out at much lower rates in off-year special elections. Jones's campaign partnered with Black churches, activist organizations, and prominent figures (including historically Black colleges and universities) to build turnout operations in the state's Black Belt counties.

Demographic decomposition. Exit polls and voter file analysis by Nate Cohn and others at the New York Times Upshot showed: - Black voters composed approximately 28% of the actual special election electorate, up from 25% in the 2016 Senate race and substantially above typical special election baselines - Jones won Black voters by approximately 96% to 4% - Among white voters, Jones held Moore to 68% — meaning roughly 32% of white voters either crossed over (voted for Jones), voted third-party, or declined to vote in the Senate race - The crossover rate among college-educated white voters and white women was notably higher than their typical Republican support levels

The Campaign Effects Analysis

How much of the outcome was Moore's liabilities (a candidate quality effect) versus Jones's mobilization campaign (a mobilization effect)?

The candidate quality effect is substantial and dominant. In a normal Republican vs. Democrat Senate race in Alabama, Jones would have lost by 15+ points. Moore's specific disqualifications — not just ideological extremism, which Alabama voters might have tolerated, but credible personal misconduct allegations — drove Republican crossover and abstention that no campaign could have manufactured. This is a reminder that campaign effects operate on margins; Jones's campaign could not have converted 30+ points of structural Republican advantage, but it didn't need to. Moore converted much of it himself.

The mobilization effect is where Jones's campaign can claim credit. The Black voter turnout story is analytically important. Black registered voters in Alabama had a modeled turnout propensity (for an off-year special election in December) in the range of 40–50 percent. The actual turnout among Black voters in the December special election, in an environment where Jones's campaign was running an intensive GOTV operation, was approximately 60 percent. This 10–15 percentage point uplift above baseline, across a population that constituted roughly 27% of the registered electorate, represents a meaningful campaign mobilization effect.

Back-of-envelope calculation: If the Black electorate was 27% of registered voters and Jones mobilized 10 extra points of turnout across that group, and Jones won 96% of those votes, the net vote contribution of the mobilization uplift above baseline was approximately: - Extra votes from mobilization: 0.27 × 0.10 = 2.7% of registered voters - Net Democratic votes from these extra voters: 2.7% × (0.96 - 0.04) = approximately 2.5% net margin contribution

In an election decided by 1.7 percentage points, this back-of-envelope estimate suggests that Jones's mobilization of the Black electorate was decisive, in the literal sense of making the difference between winning and losing.

Lessons for Campaign Effects Theory

Structural baselines are not ceilings. The structural model would have predicted Jones losing badly. He won. This doesn't mean structural models are useless — Moore's specific liabilities created a unique environment outside the historical training data for most structural models. But it illustrates that structural predictions are averages over normal candidate quality and campaign intensity. When either of those deviates dramatically, outcomes can deviate too.

Candidate quality is a campaign effect. The distinction between "structural factors" and "campaign effects" is sometimes drawn too sharply. The decision of which candidates enter a race, how they are vetted, and what liabilities they carry is partly a product of the political and organizational process that campaigns are part of. Moore cleared a crowded Republican primary in part because of factors his campaign managed; his liabilities emerged partly because a national media environment (shaped by campaign communications and opposition research) surfaced the allegations. This is a form of campaign effect even if it's not the kind that shows up in a field experiment.

Mobilization effects are largest in low-turnout environments. A special election in December is a low-participation event by definition. The baseline is low, which means a mobilization campaign can produce large percentage-point gains over that baseline. The same organizational investment in a high-turnout presidential election year, when the baseline is already high, would produce smaller marginal gains. This is precisely why campaigns and advocacy organizations sometimes prioritize off-year and special elections for their mobilization investments.

The counterfactual is everything. Did Jones's campaign win the race? The answer depends entirely on what you assume the counterfactual looks like. If you assume the counterfactual is a world where neither side ran a campaign (pure structural effects), then campaign effects were large. If you assume the counterfactual is a world where Jones didn't run his Black voter mobilization program but everything else stayed the same, the answer is probably yes — Jones's mobilization was likely decisive at the margin.

The Garza Connection

The Alabama 2017 case is directly instructive for the Garza-Whitfield race in several ways:

First, it demonstrates that demographic mobilization can swing a race even in a disadvantaged structural environment. Garza's state is more competitive than Alabama was in 2017, but the principle holds: if Garza's campaign can produce a 10-point mobilization uplift among its strongest-supporting demographic groups (young Latino voters, Black voters in urban counties), the vote contribution could be decisive in a close race.

Second, it demonstrates that structural baselines are not destiny. Garza is trailing by two in the structural baseline; the 2017 Alabama race illustrates that two-point structural deficits are fully within the range of campaign effects that well-organized operations have delivered historically.

Third, it highlights the importance of candidate quality on the other side. Whitfield is a populist Republican with his own liabilities — a history of inflammatory rhetoric, a thin policy record, and some vulnerabilities among college-educated suburban voters in the state's growing tech corridor. Nadia's model tracks whether Whitfield's liabilities are generating the kind of Republican crossover and abstention that Moore's did in Alabama.

Discussion Questions

1. How much credit should Doug Jones's campaign take for the Alabama win, given that Roy Moore's liabilities were so severe? At what point does "campaign effects" shade into "candidate quality effects"?

2. The Jones campaign's Black voter mobilization was decisive in a race with a low structural baseline. How would you expect the same organizational effort to perform in a higher-baseline presidential election year? What does this imply about optimal resource allocation across election cycles?

3. Structural models of elections are trained on historical data. What happens when a race has features (like credible misconduct allegations against an incumbent) that have no close historical analogs? How should analysts communicate uncertainty in these situations?

4. Should campaigns plan explicitly for the scenario in which their opponent experiences a disqualifying event? How would "opponent vulnerability" planning change campaign resource allocation?