Chapter 42: Capstone 1 — The Battleground State Audit

"The job of the political analyst is not to pick winners. It is to illuminate the territory so that citizens and campaigns alike can navigate it honestly." — Adaeze Nwosu, OpenDemocracy Analytics

Section 1: Introduction and Audit Framework

The Assignment

It is a Tuesday morning in early August, seven weeks before Election Day. The open-plan office of OpenDemocracy Analytics hums with quiet urgency. Adaeze Nwosu, the organization's executive director, sets her coffee down on the conference table and looks at the assembled team: Sam Harding, ODA's senior data journalist, and a cohort of analyst interns fresh from graduate programs in political science, data science, and public policy.

"We've been asked," Adaeze begins, "to produce the definitive public-facing analytical audit of the Garza-Whitfield Senate race. Not a poll. Not a horserace piece. An audit." She pauses to let the word settle. "The kind of document that a voter, a journalist, a civic organization, or a graduate seminar can use to understand what is actually happening in this state and why."

The race between State Attorney General Maria Garza, the Democratic candidate, and former State Senator Tom Whitfield, the Republican, has become one of the most closely watched Senate contests in the country. The Sun Belt state they are competing to represent — with its 2.1 million registered voters, rapidly shifting demographics, and a recent election history that has swung from Republican to Democrat and back — sits at the intersection of every major story in American politics: the education realignment in its suburbs, the contested loyalty of its large Hispanic and Latino electorate, the geographic sorting of its exurban and rural precincts, and the tightening gap between the two parties' registration advantages.

Adaeze slides a one-page brief across the table. "The audit has six questions it must answer."

The Six Audit Questions

Question 1 — What does the polling evidence actually show? Not just the headline numbers, but a rigorous examination of polling quality, house effects, methodological variation, and trend lines. What does the best available evidence suggest about where the race stands, and how confident should we be in that assessment?

Question 2 — How are different demographic groups positioned, and what does the electoral geography tell us? Which communities are trending toward which candidate? How has the suburb-exurb-rural sorting evolved since 2016? What do we know — and not know — about the political behavior of the state's large Hispanic and Latino population?

Question 3 — What do turnout scenarios suggest about likely outcomes? Given registration data, early-vote banking, and historical patterns, what range of electorates is plausible on Election Day? Which scenarios favor which candidate?

Question 4 — How is each campaign shaping — and being shaped by — the media environment? What are the dominant advertising narratives? How is the race being framed by local and national outlets? Where are the fact-checking pressure points?

Question 5 — What does campaign finance tell us about strategic priorities and resource advantages? Who has more money, who is spending more efficiently, and where is outside money flowing? What do spending patterns reveal about each campaign's theory of victory?

Question 6 — Is this race being analyzed equitably? Are the polling samples adequate across demographic groups? Are there communities that are systematically underrepresented in the data? What are the ethical obligations of a public analytics organization when gaps exist?

Analytical Framework and Data Sources

The ODA team will draw on five categories of data, each corresponding to a set of methods developed over the preceding chapters of this textbook.

Polling data comes from a compilation of all publicly available surveys conducted in the state during the 90-day pre-election window (approximately early August through late October). This includes partisan and nonpartisan polls, automated IVR surveys, live-caller telephone surveys, online panel surveys, and text-to-web surveys. The team will apply quality-weighting methods to construct a defensible polling average.

Electoral geography data comes from the state's official voter registration file, county-level results from the 2016, 2018, 2020, and 2022 elections, and U.S. Census Bureau demographic estimates at the county and census tract level (ACS 5-year estimates).

Early voting and absentee data comes from the state's official early voting portal, updated daily. The team tracks early vote requests and returns by party registration, county, and (where available) race/ethnicity.

Advertising and media data comes from FCC public inspection files for broadcast television, AdImpact tracking for cable and digital advertising, and a systematic content analysis of coverage from the state's three largest daily newspapers, two major television market affiliates, and a selection of digital outlets.

Campaign finance data comes from FEC filings through the most recent reporting deadline (15 days before Election Day), supplemented by the state's own campaign finance disclosure portal.

The Audit Timeline

The 90-day window begins ninety days before Election Day — roughly the period during which public polling, advertising, and grassroots organizing reach their peak intensity. This is the window when persuadable voters are most likely to be engaged, when early voting begins, and when the decisions made by campaign managers have the most consequential downstream effects. Understanding this window analytically is the central task of modern political data science.

💡 Intuition Think of the 90-day window not as a single "race" but as a series of overlapping strategic games. The polling environment in Week 1 of August is shaped by fundamentals — the economy, incumbency, national environment. By Week 12 (late October), it is shaped by advertising saturation, early-vote patterns, and late-breaking news. Good auditing requires treating each phase on its own terms.

The audit proceeds from the broadest structural view — polling averages and demographic composition — down to the most granular operational decisions — precinct-level GOTV targeting, ad creative strategy, and small-dollar fundraising email cadence. This movement from macro to micro mirrors how professional campaign analytics actually functions, and it is the structure you will replicate in your own analysis.

Sam Harding summarizes the team's mandate at the whiteboard: "We're not here to predict the winner. We're here to produce the most rigorous, honest, and publicly accountable picture of this race that is possible given available data. If we do our job right, people will be better equipped to understand what they're seeing on Election Day — and for weeks afterward, if the margins are close."

Let's begin.

Section 2: The Polling Landscape

Overview: A Close and Contested Race

The first task in any election audit is to assemble the full polling record and evaluate it systematically before drawing any conclusions. Individual polls are noisy instruments. Taken together, with appropriate weighting for quality and methodology, they tell a coherent story — but that story requires careful reading.

Over the 90-day window, ODA identified fourteen publicly released polls of the Garza-Whitfield race. These ranged from a well-funded university-affiliated survey using live callers across twenty-two days of field work to a one-night automated IVR poll with a paid subscription paywall and no detailed methodology disclosure. The spread across pollsters, methodologies, sample sizes, and likely voter models is substantial — as is the range of topline numbers.

Table 1: Garza-Whitfield Senate Race — All Public Polls, 90-Day Window

Note: "G" = Garza lead, "W" = Whitfield lead. Grade reflects ODA quality assessment (see below). "ED" = Election Day. LV = likely voters.

Identifying House Effects

A "house effect" is the systematic tendency of a particular pollster to show results that lean in one direction relative to the field average. House effects can arise from methodological choices (which phone numbers to call, how to weight, what likely voter screen to apply), but they can also arise from deliberate or inadvertent partisan lean.

Looking at Table 1, several patterns emerge immediately:

Partisan sponsors diverge sharply from the nonpartisan consensus. Polls 4 and 11 (Clarity Research for the Garza campaign and Progressive Strategy Group for a progressive PAC) show Garza leads of +7 and +6, respectively — substantially above the nonpartisan cluster. Polls 2, 5, and 9 (SunPoll IVR, Whitfield's internal poll, and the Republican Senate committee) show Whitfield leads of +5, +6, and +3. Partisan polls are not inherently worthless, but they must be treated with substantial skepticism and, in most quality-weighted averages, either down-weighted significantly or excluded entirely.

The nonpartisan cluster is remarkably consistent. Polls 1, 3, 6, 7, 10, 12, and 14 — all conducted by Meridian Research, Coastal University, or the National Political Survey — cluster tightly between Garza +1 and Garza +3, with the single exception of the DataPulse text-to-web survey (Poll 8) showing a tie. This consistency is the signal we're looking for.

The IVR poll (Poll 2) shows a sharp Whitfield advantage at the start of the window. SunPoll's automated IVR methodology, combined with the fact that it was conducted very early (Day -85), makes this an outlier that deserves careful scrutiny rather than dismissal — early polls sometimes capture a state of play that live-caller surveys miss. But given the subsequent consensus around Garza +2, the SunPoll result is more likely a house-effect artifact than a genuine early Whitfield lead.

⚠️ Common Pitfall Don't assume that a poll showing a large lead for your preferred candidate is "the real one" and that others are biased. Partisan confirmation bias operates in political analysis just as strongly as anywhere else. Always ask: what would I need to believe about methodology to credit this poll's result, and is that belief defensible?

Constructing the Quality-Weighted Polling Average

The ODA methodology for constructing a polling average has four steps, developed in the polling analysis chapters of this textbook.

Step 1: Grade each poll. Using AAPOR Transparency Initiative membership, disclosed likely voter screen methodology, disclosed weighting variables, disclosed field dates, and track record of past accuracy, each poll receives a grade (A, B+, B, C, or F). Partisan sponsors receive an automatic one-grade penalty before other factors are applied.

Step 2: Assign numeric weights. Grades translate to weights as follows: A = 1.0, B+ = 0.75, B = 0.50, C = 0.25, F = 0 (excluded). Partisan-sponsored polls receive the additional one-grade penalty before weighting.

Step 3: Apply a recency adjustment. Polls conducted more recently receive a recency multiplier. Polls within the last two weeks receive a 1.2x multiplier; polls from 30–14 days ago receive 1.0x; polls from 30–60 days ago receive 0.8x; polls from over 60 days ago receive 0.6x.

Step 4: Compute the weighted average. Each poll's result is multiplied by its combined weight (grade weight × recency multiplier), then the weighted results are summed and divided by the sum of all weights.

Applying this methodology to the fourteen polls in the window, with particular attention to the most recent seven polls (all of which have A or B+ grades and thus dominate the average), the ODA quality-weighted polling average for the final two weeks of the race is:

Garza: 48.3% | Whitfield: 46.9% | Undecided/Other: 4.8%

The margin is Garza +1.4 percentage points, with a 95% credible interval of approximately ±2.1 points. This places the true state of the race — conditional on our model assumptions and the polls being unbiased — between Whitfield +0.7 and Garza +3.5.

📊 Real-World Application This is structurally similar to how FiveThirtyEight, the New York Times/Siena polling aggregator, and the Economist model all construct their averages — though each uses different weighting schemes and treats recency and house effects differently. No single approach is definitively correct; what matters is that the methodology is transparent, defensible, and consistently applied.

When Did the Race Tighten?

Looking at the trend line, the race appears to have tightened meaningfully in the period around Day -50 to Day -45. The three nonpartisan polls conducted before that point (Polls 1, 3, and 6) averaged Garza +2.0. The DataPulse tie (Poll 8, Day -52) was the first signal of movement toward Whitfield. The Redfield & Partners poll (Poll 9, Day -49, R+3) — while partisan — corroborated the directional shift.

By the Day -30 to Day -20 window, the race had settled into a new equilibrium at approximately Garza +1 to +2, down from the earlier Garza +2 to +3. The tightening coincides with the launch of a major advertising blitz from outside Republican groups (documented in Section 5), suggesting a possible advertising effect — though correlation should not be mistaken for causation.

Assessing Poll Quality: AAPOR Transparency Standards

The American Association for Public Opinion Research's Transparency Initiative asks pollsters to disclose, at minimum: the full question wording and order, the methodology, the sample size and margin of error, the field dates, the weighting variables, and the likely voter screen methodology. These disclosures are the minimum floor for a poll to receive serious analytical treatment.

Of the fourteen polls in Table 1: - Polls 1, 3, 6, 7, 10, 12, 14 (Meridian Research, Coastal University, National Political Survey) met all AAPOR transparency standards. Grade: A. - Polls 4, 5 (the two campaign internals, Clarity and Vanguard) disclosed field dates, sample sizes, and methodology but did not disclose question wording or weighting variables. Grade: B (before partisan penalty applied). - Polls 8, 13 (DataPulse text-to-web) disclosed methodology and field dates but not full question wording. Grade: B+. - Poll 9 (Redfield & Partners) disclosed methodology and field dates, with partial question wording. Grade: B (before partisan penalty). - Poll 11 (Progressive Strategy Group) did not disclose question wording, weighting variables, or likely voter screen. Grade: C (before partisan penalty; final grade: F for weighting purposes). - Poll 2 (SunPoll IVR) disclosed only that an automated system was used. No question wording, no weighting details, no likely voter screen. Grade: C.

✅ Best Practice When evaluating polling for any purpose — academic, journalistic, or campaign — build a quality checklist before you look at the topline numbers. Train yourself to ask "how was this done?" before "what did it show?" The number is meaningless without the methodology.

The ODA Poll Quality Dashboard

The team's Python-based Poll Quality Dashboard — built using the skills from Chapter 10 — automates the quality-grading process. When a new poll is released, Sam enters the disclosed characteristics into a structured form. The dashboard applies the grading rubric, adds the new poll to the weighted average, and updates the trend visualization in real time.

The dashboard surfaces three key outputs: the current quality-weighted average with confidence interval; a "house effect" estimate for each pollster with more than one poll in the dataset; and a "transparency score" for each new poll, displayed on a 0–100 scale. The transparency scores for the most recent two-week window range from 94 (Meridian Research's final poll) to 31 (SunPoll IVR), providing an at-a-glance signal of which polls deserve weight.

One feature that Adaeze specifically requested: a "partisan disclosure flag" that identifies, for each poll, whether the sponsor has a documented political affiliation. The dashboard displays these prominently in orange, ensuring that consumers of ODA's public-facing materials can immediately identify which polls come from interested parties.

🔴 Critical Thinking Consider: would a poll from a campaign that showed the campaign losing ever be released publicly? Campaigns release internal polls when the results serve a strategic purpose — typically to generate favorable coverage, depress donor enthusiasm for the opponent, or shore up endorsements. A campaign poll showing a candidate leading by seven points is not evidence that the candidate is leading by seven points. It is evidence that the campaign believes releasing a poll showing a seven-point lead serves their interests.

Section 3: Demographic and Electoral Geography Analysis

The State in Demographic Transition

Understanding why this race is competitive — and what kinds of electoral coalitions could produce a winner — requires a clear-eyed view of the state's demographic and geographic structure. This is a state in accelerating transition: its Hispanic and Latino population has grown by nearly 40% since 2010, its suburban counties are shifting toward Democratic candidates in ways that mirror national education-polarization trends, and its rural precincts are becoming reliably Republican by margins that would have been unthinkable twenty years ago.

The registered voter universe of approximately 2.1 million breaks down as follows, based on the state's voter registration file and ACS demographic estimates:

Table 2: Registered Voter Demographics by County Group

The Democratic registration advantage of D+2 (40% vs. 38%) is, as Adaeze notes in the team briefing, "a fragile thing." Registration advantages translate to vote advantages only when registered voters of each party actually turn out at similar rates, and when non-affiliated voters split reasonably evenly. Neither of those conditions can be assumed in this state.

County-Level Political History

The four anchor counties — Riverside, Millbrook, Vega, and Redstone — provide the clearest signal of how this state has been changing.

Riverside County is the state's population center and the Democratic anchor. In 2020, Biden carried Riverside by 19.2 points; in 2018, the Democratic Senate candidate won it by 24.1 points; in 2022, the Republican Senate candidate held Riverside to just a D+12.4 margin — a significant erosion. The question for the Garza campaign is whether Riverside's 2022 underperformance was an anomaly driven by a weak candidate and an unfavorable national environment, or whether it reflects a structural shift in the county's political behavior.

Millbrook County is the state's critical swing county. A majority-white, college-educated suburban county with a growing Hispanic and Asian-American professional class, Millbrook voted for Trump by 8.1 points in 2016, then swung to Biden by 2.3 points in 2020 — a 10-point swing that mirrors the national suburban education realignment. In 2022, however, the Republican Senate candidate carried Millbrook by 5.8 points, a 8-point swing back toward Republicans in a midterm with a favorable Republican national environment. The 2024 Senate race will be fought heavily in Millbrook.

Vega County is the most analytically contested territory in the state. With a 68% Hispanic registered voter base and a history of split-ticket voting — it voted for Biden by 7.2 points but re-elected a Republican state legislator in the same cycle — Vega County does not behave according to simple demographic prediction. The county's political identity is shaped by a mix of long-established Mexican-American families, more recent Central American immigrants, and a substantial Cuban-American community in the county seat, each of which has distinct political tendencies.

Redstone County is Whitfield territory. It voted for Trump by 41.2 points in 2020 and has trended Republican in every cycle since 2008. The question here is not who wins but by how much — and whether Whitfield can replicate or exceed Trump's margins, which are the baseline assumption of the Republican path to victory.

The Education Realignment in This State's Suburbs

The national pattern of college-educated voters shifting toward Democrats while non-college voters shift toward Republicans has played out in this state with particular intensity. Millbrook County's four-year college attainment rate is 48%, one of the highest in the state, and its 10-point swing toward Biden in 2020 tracks almost perfectly with the national correlation between college attainment and Democratic vote shift.

But the state's exurban counties — a ring of communities beyond Millbrook that include substantial shares of voters with some college but not a four-year degree — tell a different story. These counties have swung hard toward Republicans since 2016, and the swing has not reversed even as Millbrook moved back toward Republicans in 2022. The "some college but no degree" segment appears to be the most volatile portion of the electorate, and its behavior in 2024 will significantly shape the outcome.

🔗 Connection This education realignment dynamic maps directly to the concepts developed in Chapter 18 (Demographic Sorting and the New Electoral Geography) and Chapter 22 (The Suburban Realignment). Review those chapters' regression models for college attainment as a predictor of partisan shift before you construct your own county-level demographic model.

Hispanic and Latino Voter Coalition Analysis: The Vega County Deep Dive

Vega County deserves extended analysis because it embodies the complexity of Hispanic and Latino political behavior at the national level. There is no single "Hispanic vote." There are Mexican-American voters whose families have lived in this state for generations. There are Guatemalan and Salvadoran immigrants who became citizens in the last decade. There are Cuban-American voters with distinct political identities shaped by their families' experience with the Castro government. There are Puerto Rican migrants from the Northeast. And there are second- and third-generation Latino voters who may have weak ethnic identification but strong local economic concerns.

The internal ODA analysis of Vega County — using the ACS and the state voter file, cross-referenced with census tract-level surname analysis — suggests three distinct sub-populations:

Established Mexican-American community (est. 38% of Latino registered voters in Vega): This population has the highest registration rates and turnout rates among the county's Latino voters. It has historically voted Democratic by wide margins — roughly 65-35 — but has shown some movement toward Republican candidates, particularly in 2022, likely related to economic concerns in agricultural and service-sector employment.

Recent immigrant-origin naturalized citizens (est. 29% of Latino registered voters in Vega): This population has lower turnout rates and shows higher rates of non-partisan registration. When it votes, it has trended Democratic, but by narrower margins — roughly 55-45 — and it is more susceptible to Spanish-language advertising from both parties.

Cuban-American and South American-origin voters (est. 19% of Latino registered voters in Vega): This population skews substantially Republican — historically around 40-60 Democrat-Republican — and has moved further toward Republicans in recent cycles. The Whitfield campaign has invested heavily in Spanish-language outreach specifically targeted at this segment.

Younger (under-35) Latino voters of mixed background (est. 14% of Latino registered voters in Vega): This population is the least predictable. It has high rates of non-partisan registration, low historical turnout, and limited polling coverage. It is the segment most likely to be affected by GOTV operations from progressive organizations active in Vega County.

⚖️ Ethical Analysis The practice of using surname analysis to estimate racial and ethnic composition of a voter file — sometimes called "Bayesian Improved Surname Geocoding" or BISG — is a standard analytical tool, but it has significant limitations and ethical implications. It systematically misclassifies Hispanic voters with common Anglo surnames (Rodriguez is fine; Martinez-Smith is ambiguous; Jennifer Williams who is Mexican-American is invisible). It treats race/ethnicity as a stable attribute when political identity is complex and contextual. Any analysis that relies on BISG estimates should clearly disclose its limitations and avoid making strong causal claims about "how Latinos voted."

Black Voter Turnout and Its Potential Swing Impact

The state's Black voter population — 18% of registered voters — is concentrated most heavily in Riverside County (22% of registered voters) and in Metro 2 (20%). Black voters have been the most consistently Democratic demographic in the state, voting for Democratic candidates in Senate races by margins of roughly 85-12 across multiple cycles.

The swing impact of Black voter turnout is therefore less about persuasion than about mobilization. If Black voter turnout in Riverside County reaches 65% of registered voters, the margin contribution from this population is approximately 52,000 net Democratic votes. If turnout falls to 55%, the net contribution drops to approximately 44,000 — a swing of 8,000 votes in a race that could be decided by fewer than 20,000 statewide.

The Garza campaign has invested heavily in field organizing in Riverside County's majority-Black neighborhoods. The question — which Nadia Osei, the campaign's analytics director, tracks obsessively — is whether the enthusiasm generated by a Hispanic Democratic candidate leading a statewide race translates into cross-racial coalition mobilization, or whether turnout in Black communities tracks closer to the 2022 low.

Urban-Suburban-Rural Polarization

The geographic sorting of this state's electorate has accelerated since 2016 in ways that make aggregate registration numbers increasingly misleading. The two-point Democratic registration advantage obscures a county-level landscape in which:

• The three most urban county areas (Riverside, Metro 2, and the urban core of Millbrook) have a combined Democratic registration advantage of approximately 8 points.
• The four exurban counties have a combined Republican registration advantage of approximately 14 points.
• Redstone and the other rural counties have a combined Republican registration advantage of approximately 27 points.

The practical implication: Garza needs to run up large margins in urban areas and limit her losses in Millbrook County's suburban ring. Whitfield needs to replicate or exceed recent Republican rural margins and hold enough of Millbrook to keep his overall deficit within range of his rural advantage.

Identifying the Swing Universe

The "swing universe" — voters whose behavior is genuinely persuadable rather than determined by party identity — is smaller than most campaigns acknowledge. Based on cross-tabulation of party registration, past voter history, and self-reported party identification in available polls, ODA estimates the persuadable universe at approximately 180,000–210,000 registered voters statewide, concentrated in:

• Millbrook County non-affiliated voters (est. 55,000 persuadable): High-education, moderate-income suburban voters with ticket-splitting histories.
• Vega County non-affiliated and weak-partisan voters (est. 42,000 persuadable): The most targeted segment by both campaigns.
• Exurban "soft Republican" voters (est. 38,000 persuadable): Economically anxious, college-educated women in exurban counties who supported Biden in 2020 but voted Republican down-ballot.
• Riverside County non-participating registered voters (est. 35,000 persuadable via GOTV): Not persuasion targets per se, but registered Democrats who have not voted in recent cycles and could be mobilized.
• Rural non-affiliated voters (est. 18,000 persuadable): Small in number but relevant in a tight race.

🧪 Try This Take the swing-universe estimates above and apply a simple sensitivity analysis. If the Garza campaign converts 60% of Millbrook non-affiliated voters (rather than the baseline 50%), how much does that shift her overall margin? What about if Vega County persuadable voters split 55-45 toward Garza instead of 50-50? Work through the math using the registration numbers in Table 2 and the county-level turnout assumptions developed in Section 4.

Building a Demographic Scenario Model: What If Hispanic Turnout Increases 5%?

The Garza campaign's most optimistic internal scenario involves a surge in Hispanic and Latino turnout — particularly in Vega County and in Riverside County's growing Latino neighborhoods. This scenario is plausible: Garza is the first Latina to appear on a major-party Senate ballot in the state's history, a fact that has generated significant grassroots organizing energy in Latino communities.

The baseline ODA assumption: Hispanic registered voters turn out at 54% in Vega County and 51% in Riverside County, consistent with 2020 rates adjusted downward for midterm-style drop-off in a non-presidential year.

Scenario: Hispanic/Latino Turnout +5 Percentage Points in Vega and Riverside

If Hispanic and Latino turnout increases by 5 percentage points from the baseline in both Vega County and Riverside County — a substantial but historically plausible surge given the candidate's profile — the additional votes break approximately 57-43 for Garza based on historical patterns for newly mobilized Latino voters.

In Vega County: 5% of 187,000 registered voters × 68% Hispanic registration rate = approximately 6,358 additional voters. At 57-43 Garza, that's 3,624 Garza votes and 2,734 Whitfield votes — a net Garza gain of approximately 890 votes in Vega County.

In Riverside County: 5% of 712,000 registered voters × 29% Hispanic registration rate = approximately 10,324 additional voters. At 57-43 Garza, net Garza gain of approximately 1,445 votes.

Total net Garza gain from the Hispanic turnout surge scenario: approximately 2,335 votes statewide. In a race that could be decided by 10,000–25,000 votes, this is meaningful but not decisive — it represents roughly one-tenth to one-fourth of a typical winning margin in a race this close.

The lesson is not that Hispanic turnout increases are unimportant. It is that the scenario analysis reveals just how hard it is for any single demographic strategy to be decisive. Winning requires stacking multiple successful scenarios simultaneously.

Section 4: Turnout Modeling and Scenario Analysis

Baseline Turnout Prediction

The foundation of the turnout model is the historical record combined with current registration data. ODA's baseline projection assumes turnout patterns consistent with the average of the 2018 and 2022 Senate elections — two midterm cycles that bracket the plausible range for a competitive 2024 Senate race — adjusted for registration changes, early voting data, and current enthusiasm indicators from available surveys.

Table 3: Baseline Turnout Projection by County

Note: Shares do not add to 100% due to third-party/write-in votes. Net Garza figures are rounded.

The baseline scenario produces a Garza win of approximately 9,000 votes — within the range suggested by the quality-weighted polling average, and therefore internally consistent. But as the scenario analysis below illustrates, 9,000 votes is a razor-thin margin that can be erased by fairly small changes in any of several assumptions.

Three Scenarios: Low, Medium, and High Turnout

Low Turnout Scenario (est. 55% overall; 1,155,000 total votes)

A low-turnout scenario in this state historically disadvantages Democrats more than Republicans, because Democratic registration advantages are concentrated in urban areas where "soft" registrants — those with weaker partisan attachment or lower voting histories — are more numerous. In a low-enthusiasm environment, these voters are less likely to participate.

In the low-turnout scenario, Riverside County's turnout falls to 56% (rather than 61%), and Millbrook County's turnout falls to 62%. Both changes disproportionately reduce Democratic net votes. The result: the Garza margin narrows to approximately 2,500 votes — well within any reasonable margin-of-error band, and a margin at which concerns about counting errors, provisional ballot adjudication, and late returns become determinative.

Under the low-turnout scenario, Whitfield's path to victory becomes clear: if he can hold Riverside County to D+14 or narrower (achievable if soft Democrats stay home and his Spanish-language advertising in Riverside's Latino neighborhoods has eroded Garza's coalition) while running up his rural margins, he wins by approximately 8,000 votes.

Medium Turnout Scenario (est. 60% overall; baseline)

As described in Table 3, the medium scenario produces a Garza win of approximately 9,000 votes. This is the ODA central estimate, consistent with the polling average, and the scenario the team will publish as its headline finding.

High Turnout Scenario (est. 65% overall; 1,365,000 total votes)

A high-turnout environment in this state, driven by competitive mobilization from both campaigns and high voter interest, has historically benefited Democrats because of the depth of their registration advantage in Riverside County and Metro 2. In the high-turnout scenario, overall turnout rises to 65%, with the largest gains in Riverside County (reaching 66%) and in Vega County (reaching 60%).

The high-turnout scenario produces a Garza margin of approximately 22,000 votes — a more comfortable win, though still well within a contested range if the result triggers an automatic recount under state law (typically invoked when the margin is below 0.5%, which at this turnout level would be approximately 6,800 votes).

Registration Changes Since 2022: The New Voter Factor

Since the 2022 election, the state's voter rolls have changed substantially. Approximately 187,000 new registrations have been added, with the following broad characteristics:

• Vega County: Net +18,400 registrants, with new registrations skewing 48% non-affiliated, 32% Democratic, 20% Republican. This represents a meaningful expansion of the persuadable universe in the county.
• Riverside County: Net +41,200 registrants, skewing heavily Democratic (52%) and younger. If these voters participate at reasonable rates, they represent a structural advantage for Garza.
• Millbrook County: Net +12,600 registrants, more evenly split: 35% Democratic, 39% Republican, 26% non-affiliated.
• Exurban counties: Net +22,800 registrants, skewing Republican (43%) with a substantial non-affiliated share (38%).

New registrants, as a general rule, vote at rates 10–15 percentage points lower than established voters in their first election. Incorporating this adjustment into the baseline model slightly reduces the advantage that the Riverside County registration growth would otherwise confer on Garza.

The Early Voting Situation

With three weeks remaining before Election Day, the early voting data provides a real-time signal of turnout trajectory. The state's early voting portal (updated nightly) shows the following cumulative pattern as of Day -21:

• Total early votes cast: 312,440
• By party registration: D 44%, R 39%, NPA/Other 17%
• By county: Riverside County leads with 31% of all early votes cast, followed by Millbrook County (19%), Metro 2 (16%), and Vega County (9%).
• Comparison to 2022 pace: Overall early voting is running 18% ahead of 2022, with the largest increases in Riverside County (+27%) and Vega County (+24%).

The Democratic party-registration share of early votes (44%) slightly exceeds the Democratic registration share of the overall electorate (40%), suggesting that Democratic-leaning voters are slightly outperforming their registration share in early voting. However, Republicans typically outperform their early vote registration share on Election Day — the so-called "Red Mirage" phenomenon — which means the early vote party-registration data does not directly translate into a 5-point Garza early-vote advantage.

⚠️ Common Pitfall Many analysts in 2020 made the error of interpreting early-vote registration data as vote tallies. They are not. A registered Democrat who casts an early ballot may have voted for Whitfield. A registered Republican who shows up on Election Day may have voted for Garza. Registration is a proxy for intended behavior, not a measurement of actual votes.

Identifying High-Priority GOTV Opportunities

Based on the baseline model, the turnout scenario analysis, and the early-vote data, ODA identifies four high-priority GOTV opportunity areas for the Garza campaign:

1. Riverside County's majority-Black precincts with historically variable turnout — Where a 5-percentage-point turnout increase yields approximately 4,200 net Garza votes.
2. Vega County's younger Latino non-affiliated registered voters — Where mobilization of 30% of the untapped persuadable pool yields approximately 3,800 net Garza votes (assuming 57-43 break).
3. Millbrook County's Democratic-registered but infrequent-voter population — Where targeting the "low-propensity Dem" universe yields modest but potentially margin-relevant returns.
4. Metro 2's suburban progressive neighborhoods with growing millennial and Gen Z populations — Where enthusiasm-driven organizing has already produced above-average early vote returns.

Comparing Internal Models: Nadia's Assumptions vs. ODA's

Nadia Osei, the Garza campaign's analytics director, has shared (through background conversations with Sam Harding) some high-level parameters of the campaign's internal turnout model. The comparison is instructive.

The campaign's internal model assumes Riverside County turnout at 64% — three points higher than ODA's 61% baseline. The campaign also assumes Vega County turnout at 57%, compared to ODA's 54%. Both assumptions are more optimistic than ODA's from a Garza perspective.

Nadia's justification: "We have 450 field organizers in the state. We know what our volunteers are seeing on the doors. Our early vote targets are being hit. When you have ground truth from the field, you update your model accordingly."

This is a legitimate argument. Campaign field data is, in some respects, superior to historical-average modeling. The risk is that campaign models are subject to motivated reasoning — the same enthusiasm that drives organizers' optimism can bias their reporting of door-contact conversion rates. ODA uses a conservative prior and updates toward campaign field data only when the early-vote numbers corroborate the campaign's claims.

As of Day -21, the early-vote data in Riverside County and Vega County is running above ODA's baseline but below the campaign's optimistic projection. ODA adjusts its baseline upward by 1.5 percentage points in both counties — a Bayesian update toward the campaign's position, but not a full adoption of it.

Section 5: Media and Advertising Audit

Ad Spending Overview

The advertising race in the Garza-Whitfield contest has been among the most expensive Senate races in the country. ODA's analysis of FCC public inspection files and AdImpact tracking data shows the following cumulative spending through Day -21:

Table 4: Advertising Spending by Entity (through Day -21)

The pro-Whitfield advertising ecosystem holds a modest overall spending advantage of approximately $1.1M through Day -21 — a gap that has narrowed from the approximately $3M advantage that outside Republican groups established in the Day -60 to Day -45 window. Senate Majority PAC's late entry into the state has partially closed the gap.

The geographic distribution of broadcast spending reveals strategic priorities. The Whitfield campaign and its allied groups have concentrated roughly 42% of their broadcast spending in the Millbrook County media market — an expensive, suburban market where the Republican campaign is trying to hold suburban ticket-splitters. The Garza campaign has concentrated its spending 38% in the Riverside County market (base mobilization and margin-building) and 28% in the Vega County media market (persuasion and mobilization of Latino voters).

Message Analysis: The Core Narratives

The Garza Campaign Narrative

Maria Garza's advertising has centered on three interlocking themes:

Economic security for working families. The campaign's highest-rotation ad features Garza in a kitchen with a family discussing grocery prices and healthcare costs. The closing line — "I've been fighting for families like yours my whole career. Tom Whitfield has been fighting for the people who write the checks" — is a direct contrast draw that positions Garza as an economic populist against a candidate linked to corporate donors.

Border security and her record as AG. Recognizing the vulnerability that immigration policy creates for Democratic candidates in Sun Belt states, the Garza campaign has run a substantial rotation of ads emphasizing her record as state attorney general: prosecuting drug trafficking organizations, securing deportations of violent offenders, and working with Border Patrol on joint operations. This is an explicit effort to neutralize Whitfield's attacks on border security.

Representation and the historic nature of the race. A Spanish-language ad running heavily in Vega County media markets features Garza speaking directly to camera about her parents' immigration story and her identity as a first-generation Latina. The ad has generated significant earned media and is widely credited with energizing volunteer organizing in predominantly Latino neighborhoods.

The Whitfield Campaign Narrative

Tom Whitfield's advertising has been similarly structured around three themes:

Crime and public safety. Whitfield's highest-rotation ad ties Garza to a single prosecutorial decision from 2019 in which her office declined to seek the death penalty in a high-profile murder case. The ad is the most sharply negative in the race and has been the subject of multiple fact-checks (see below). It has run approximately 4,200 GRPs in the Riverside County market — a heavy rotation clearly aimed at suburban voters who may be uncomfortable with Garza's record on criminal justice.

Inflation and fiscal responsibility. Whitfield's economic message ties Garza to the incumbent Democratic president and to federal spending legislation, arguing that her support for "trillions in new spending" caused inflation. This is a straightforward partisanship-nationalization message.

Immigration and border security. Whitfield's border security advertising is the highest-volume message in his rotation. A series of ads feature images of the southern border and statistics on border apprehensions, with the tagline "Maria Garza calls it 'managed.' I call it a crisis." This message has tested strongly in the exurban and rural counties that form his base.

📊 Real-World Application Message analysis of the type ODA is conducting here maps directly to the media content analysis methods from Chapter 29 (Advertising and Persuasion) and Chapter 31 (Political Communication in the Digital Age). Note that "analyzing the message" is distinct from "evaluating the truth of the message." Both are important and both appear in this audit — but they serve different analytical functions.

Fact-Check Tracker

ODA maintains a running tracker of claims made in advertising and on the campaign trail that have been evaluated by nonpartisan fact-checkers. The most significant in this race:

Whitfield ad on Garza's prosecutorial record (the death penalty claim): Rated "Mostly False" by the state's largest newspaper's fact-check desk and "Misleading" by PolitiFact. The ad implies that Garza personally made the decision not to seek the death penalty; in fact, the decision was made by a line prosecutor and Garza's office has stated she was not consulted. The ad has run continuously despite the fact-check rulings, and the Whitfield campaign has not modified it.

Garza ad on Whitfield's healthcare position: A Garza ad claims Whitfield "voted to end Medicare as we know it." Rated "Needs Context" by the state newspaper. Whitfield did support a budget framework in the state legislature that included long-term Medicare reform proposals, but the characterization of the vote as supporting elimination of Medicare is an overstatement. The Garza campaign modified this ad's script after the fact-check, substituting "cut Medicare" for "end Medicare."

Garza claim on border prosecutions: Garza has claimed in stump speeches and a debate that her office "secured more drug trafficking prosecutions than any AG in state history." This claim is rated "True" by fact-checkers, based on official state Department of Justice records.

Whitfield claim on inflation: Whitfield's ads cite the statistic that "inflation cost the average family $14,000 under Biden." This figure comes from a Heritage Foundation analysis and is rated "Misleading" — it measures the cumulative price level increase since 2021 rather than additional costs, and assumes consumption patterns that don't apply to lower-income families.

Sam's ODA Media Dashboard

Sam Harding's media tracking dashboard, built on the Chapter 31 methods, monitors coverage across twelve major outlets: three broadcast television affiliates, two daily newspapers (the Riverside Courier and the Metro Tribune), the state's dominant digital news aggregator, two Spanish-language television stations, four digital-native outlets, and one podcast network with substantial reach among young voters.

The dashboard's sentiment analysis — applying VADER to a corpus of approximately 4,800 news articles mentioning Garza or Whitfield — shows a slight negative sentiment advantage for Whitfield in recent weeks, driven primarily by a cluster of investigative stories about his campaign's donor relationships. Coverage of Garza has been more neutral on average, though a late-October story about a staff departure generated a brief negative spike.

The Spanish-language media landscape deserves separate treatment. The two Spanish-language television stations — Univisión affiliate KUVS and Telemundo affiliate KTLV — have both provided substantially more Garza-favorable coverage over the course of the race, reflecting both her historic candidacy and her active engagement with Spanish-language press. The Spanish-language media audience overlaps significantly with Vega County and the Latino neighborhoods of Riverside County, making this a strategically important media environment.

🔗 Connection The framing analysis of media coverage here draws directly on Chapter 32's treatment of framing theory. Note how the same set of facts — Garza's prosecutorial record — is framed completely differently depending on the outlet and its audience. This is not (necessarily) bias; it reflects genuine differences in what different audiences find salient.

Section 6: Campaign Finance and Outside Spending

Fundraising Totals and Burn Rates

Through the most recent FEC filing deadline (approximately 15 days before Election Day), the two campaigns have reported the following aggregate fundraising:

Maria Garza for Senate: - Total raised (cycle): $24.7 million - Cash on hand: $3.1 million - Total spent: $21.6 million - Burn rate (past 30 days): 94% of monthly receipts

Tom Whitfield for Senate: - Total raised (cycle): $21.3 million - Cash on hand: $2.8 million - Total spent: $18.5 million - Burn rate (past 30 days): 91% of monthly receipts

Garza holds a fundraising advantage of approximately $3.4 million for the cycle — a substantial gap that reflects both the financial advantages of a Democratic candidate in a competitive Sun Belt race and the national Democratic donor community's investment in a historic candidacy. However, as the advertising analysis in Section 5 shows, the overall pro-Whitfield advertising ecosystem (campaign plus allied groups) holds a modest spending advantage, meaning that Whitfield's relative campaign disadvantage is more than offset by outside group activity.

Small-Dollar vs. Large-Dollar Donor Breakdown

Garza campaign: 61% of total raised from donors giving $200 or less (small-dollar); 21% from donors giving $201–$2,900 (mid-range); 18% from donors giving $2,900 (the individual maximum). Average donation: $47.

Whitfield campaign: 38% of total raised from donors giving $200 or less; 28% from mid-range donors; 34% from maximum donors. Average donation: $89.

The contrast in small-dollar donor profiles reflects a structural difference in donor coalition. Garza's campaign has been heavily fueled by online grassroots fundraising — primarily through ActBlue, with a particularly strong surge following a viral moment in the lone debate when she challenged Whitfield on healthcare. Whitfield's campaign relies more heavily on large-dollar donors and traditional Republican fundraising networks, which has historically been more efficient per dollar raised but less scalable in terms of donor count.

The practical implication for the campaign's closing weeks: Garza has a larger pool of existing donors who can be solicited for additional contributions (multiple giving is common among small-dollar online donors), while Whitfield is more dependent on maxed-out donors and therefore has a structurally more limited end-of-campaign fundraising ceiling.

📊 Real-World Application The rise of small-dollar online fundraising — through platforms like ActBlue (Democratic) and WinRed (Republican) — has transformed the campaign finance landscape over the last decade. Small-dollar fundraising data is sometimes used as a measure of grassroots enthusiasm, but this interpretation requires care: viral fundraising spikes are often triggered by outrage or media moments rather than sustained organizational strength.

Super PAC and Dark Money Involvement

The outside spending ecosystem in this race is complex. Senate Majority PAC, the Democratic Senate campaign's aligned super PAC, has made an independent expenditure of $6.4 million — publicly disclosed and coordinated in timing (though not in content) with the campaign. American Leadership Fund, the analogous Republican super PAC, has spent $8.1 million.

Beyond these disclosed super PAC expenditures, ODA has identified approximately $5.4 million in what is classified as "dark money" spending — primarily 501(c)(4) social welfare organizations that are not required to disclose their donors. Progress Now (progressive) and Heritage Alliance (conservative) are both 501(c)(4)s, and the donor networks behind both organizations are not publicly disclosed.

The Heritage Alliance, in particular, warrants scrutiny. Its $3.1 million in spending represents a significant investment in a state where the organization has limited historical presence. FEC filings indicate that Heritage Alliance has received transfers from at least two other 501(c)(4) organizations, both of which have connections to a network of donors associated with fossil fuel industry interests — a relevant fact given that one of the race's subsidiary debates involves the state's energy development policies.

⚖️ Ethical Analysis Dark money — spending by organizations that do not disclose their donors — represents one of the most significant transparency challenges in contemporary campaign finance. ODA's ethical commitment is to document what is known, clearly label what is unknown, and resist the temptation to speculate about donor identities beyond what the documentary record supports. When ODA says "connected to fossil fuel industry interests," it means: FEC records show organizational transfers; the receiving organizations' stated missions and activities align with fossil fuel policy positions; and board members of the organizations have documented professional relationships with industry actors. This is suggestive, not conclusive, and should be reported as such.

Outside Spending and Air Time Implications

The combined pro-Whitfield outside spending advantage in broadcast television — approximately $1.5 million more in broadcast GRPs than the pro-Garza ecosystem — translates to a meaningful difference in total air time. In the Millbrook County market, which is both expensive and strategically critical, the pro-Whitfield ecosystem has purchased approximately 19% more gross rating points than the pro-Garza ecosystem over the final 30 days.

This advertising advantage is almost certainly a factor in the race's tightening in the Day -50 to Day -40 window. Advertising effects are typically modest (3–5 points of movement in persuadable voters per 1,000 GRPs in competitive conditions), but in a race this close, a 3-point swing in Millbrook County among persuadable voters is approximately 2,900 net votes — the difference between a 9,000-vote Garza win and a 6,100-vote Garza win.

Section 7: Forecasting the Race

Integrating All Evidence

Forecasting a competitive election requires integrating four categories of evidence: the polling record, fundamental economic and political conditions, demographic and structural factors, and campaign-specific inputs. Each category carries different weight depending on how far out from Election Day the forecast is generated; in the final two weeks, polling dominates.

ODA's forecast model proceeds as follows:

Polling component (weight: 55%): The quality-weighted polling average of Garza +1.4 points, with a standard error of approximately 1.1 points based on the historical accuracy of state-level Senate polling in similar cycles.

Fundamentals component (weight: 20%): Based on presidential approval rating, generic ballot, and economic indicators, the fundamentals model suggests a slight Republican-favorable environment nationally — consistent with a Democratic candidate running roughly at parity with the national environment rather than above it. This contributes approximately +0.3 points to Whitfield's fundamentals adjustment.

Demographic and structural component (weight: 15%): The demographic trend analysis — particularly the continued growth of the Hispanic and Latino electorate and the ongoing suburban education realignment — is slightly favorable to Garza beyond the registered voter baseline. This contributes approximately +0.4 points to Garza.

Campaign-specific component (weight: 10%): The small Whitfield outside-spending advantage, partially offset by Garza's ground game advantage and small-dollar fundraising momentum, is assessed as roughly neutral — contributing less than 0.1 points in either direction.

ODA Final Forecast: Garza +1.7 points (approximately 49.0% to 47.3%), with a 95% credible interval of ±2.3 points.

Probability Distribution

Converting the point estimate and credible interval to a win probability requires assumptions about the shape of the distribution of possible outcomes. ODA uses a t-distribution with heavier tails than a normal distribution, reflecting the reality that election polling contains correlated errors (pollsters systematically missing the electorate in the same direction) and non-polling uncertainty.

ODA win probability: Garza 64% | Whitfield 36%

This is a competitive race that Garza is modestly favored to win. The 64-36 probability reflects both the direction of the evidence (Garza leads in most nonpartisan polls) and the substantial remaining uncertainty (a uniform 2-point polling error in either direction is historically common in competitive Senate races).

💡 Intuition A 64% win probability means roughly two-out-of-three election instances produce a Garza win. In baseball terms, a team that wins 64% of its games goes 104-58 over a full season — a good team, clearly better than its opponent, but by no means invincible. Political analysts should communicate probabilities in ways that preserve, rather than collapse, the genuine uncertainty.

Sensitivity Analysis: What Would It Take for Each Side to Win?

For Whitfield to win from the current position:

1. A uniform 1.7-point polling error in his favor — historically uncommon but entirely within the range of actual polling error distributions.
2. Millbrook County breaking R+4 or better (consistent with 2022 and not implausible).
3. Black voter turnout in Riverside County reaching only 57% of registered voters (two points below ODA's baseline).
4. Vega County's Latino voters splitting 48-49 for Whitfield rather than the baseline 46-51 for Garza.

Any one of these scenarios is plausible. A combination of two or three, simultaneous, produces a comfortable Whitfield win of 30,000+ votes.

For Garza to win more comfortably than the base case:

1. Riverside County replicates 2020 rather than 2022 Black voter turnout patterns (+5 percentage points).
2. Millbrook County's college-educated women break 56-44 for Garza, up from the baseline 53-47.
3. The Hispanic turnout surge scenario described in Section 3 materializes (+2,335 net Garza votes).
4. Statewide polling average understates Garza by 1 point (a common pattern when women candidates face a "shy voter" dynamic).

Again, any one of these is plausible. All four together produce a Garza win of 35,000–45,000 votes, approaching 2018 Democratic margins in the state.

Path to Victory Scenarios

Garza's path: Win Riverside County by 16+ points (net approximately 65,000 votes), hold Millbrook County to R+1 or better (net loss less than 3,000 votes), win Vega County by 5+ points (net approximately 5,000 votes), and limit Redstone County to R+38 or narrower. This produces approximately 5,000–15,000 vote Garza statewide margin.

Whitfield's path: Win Millbrook County by R+5 or better (net approximately 15,000 votes), hold Riverside County to D+13 or narrower, win Vega County or hold it to single digits, and run up 42+ point margins in rural precincts. This produces approximately 5,000–20,000 vote Whitfield statewide margin.

Communicating Uncertainty

One of the most important skills in political analytics is communicating uncertainty to a public audience without either (a) false precision that implies more predictability than exists, or (b) excessive hedging that leaves readers with nothing useful.

ODA's public-facing language for the forecast: "Based on available polling and analytical evidence, Maria Garza is a modest favorite in this race. Our forecast gives her approximately a two-in-three chance of winning — reflecting both the consistent (if narrow) leads she holds in nonpartisan polling and the genuine uncertainty that characterizes any competitive Senate race. A Whitfield win would not be a polling failure; it would be a fully expected outcome given the current distribution of possibilities."

Section 8: Equity and Representation Audit

Demographic Representation in the Polling Record

One of ODA's organizational commitments is to evaluate not just what polls say, but who they are hearing from. In a state where 32% of registered voters identify as Hispanic or Latino and 18% identify as Black, the polling record must be evaluated for whether these communities are adequately represented.

Analysis of the methodology disclosures for the seven A-grade polls in the race reveals:

Hispanic/Latino representation: Five of the seven A-grade polls report sample demographics; of these, the average share of Hispanic/Latino respondents is 24%, meaningfully below the 32% share in the registered voter file. Two polls explicitly report weighting for race/ethnicity; the others do not disclose this. This underrepresentation is not unusual — Hispanic and Latino respondents have lower survey response rates for structural reasons (cell-phone-only status, language, trust in institutions) — but it means that polls are relying on extrapolation from a smaller-than-representative Latino sample, with corresponding uncertainty in estimates of candidate support among that community.

Black voter representation: Four of the seven A-grade polls report sample demographics. The average Black respondent share is 15%, below the 18% registration-file share. Again, this reflects structural survey participation gaps rather than deliberate exclusion, but the consequence is that estimates of Black candidate support carry larger margins of error than the headline MOE would suggest.

Young voter representation (under-35): This is where the representation gap is most severe. All five polls that report age demographics show significant underrepresentation of voters under 35, with shares ranging from 12% to 18% compared to a registration-file estimate of approximately 23%. Polls weighted to correct for this underrepresentation would need to apply substantial upweighting to young respondents, introducing additional variance.

⚖️ Ethical Analysis Underrepresentation of certain demographic groups in polling is not a neutral technical problem. It has political implications: if Latino voters are systematically underrepresented in polling, candidate and campaign decisions made on the basis of those polls may underinvest in Latino outreach. If young voters are routinely underrepresented, campaigns may systematically underestimate their potential impact. The equity audit of polling methodology is not just an academic exercise — it shapes resource allocation decisions with real democratic consequences.

Voter Suppression Concerns: Registration and Access

ODA's audit flags three structural concerns about voting access in this race:

Polling place consolidations in Vega County: Following a 2019 decision by the county election commission — whose Republican majority voted along party lines — the number of polling places in Vega County was reduced from 47 to 31. The eliminated polling places were disproportionately in the county's higher-density Latino neighborhoods. The county claims the consolidation was a cost-saving measure; civil rights organizations have challenged it in federal court. The litigation is pending as of Election Day.

Registration purge controversy in Riverside County: In 2023, the state's Republican-controlled legislature passed a voter list maintenance law requiring the removal of any registered voter who did not cast a ballot in two consecutive general elections unless they responded to a mailed confirmation notice within 30 days. Riverside County civil rights groups estimate that approximately 14,000 registered voters — disproportionately renters and low-income residents who move frequently — were removed under this law between August 2023 and June 2024. Some of these voters may attempt to vote and discover they are no longer registered; provisional ballot adjudication procedures in the state are complex.

Language access gaps: The state's Voting Rights Act obligations under Section 203 require Spanish-language ballot materials and poll worker assistance in counties exceeding certain Hispanic population thresholds. Vega County is clearly covered. Two exurban counties with growing Hispanic populations may meet the threshold but have not yet implemented Spanish-language materials. ODA recommends further investigation.

Data Gaps: Communities Hardest to Reach

The communities hardest to reach in polling, advertising, and analytical modeling tend to be:

• Mobile populations with high rates of residential mobility (renters, seasonal workers)
• Communities with lower rates of English-language media consumption
• Populations with institutional distrust of surveys and data collection
• Communities with lower rates of voter registration (and therefore outside the voter file entirely)

In this state, these characteristics overlap substantially with undocumented and mixed-status families in Vega County, with Indigenous-identifying communities in two rural counties, and with returning citizens (individuals recently released from incarceration who have had their voting rights restored). The combined estimated size of these hard-to-reach communities is difficult to estimate precisely, but ODA places it at 80,000–120,000 adults of voting age — a non-trivial share of the total electorate.

Adaeze's Equity Checklist

Before publishing any analytical work, ODA applies an internal equity checklist developed by Adaeze Nwosu and the organization's equity director. Applied to the battleground state audit:

1. Have we clearly disclosed the demographic limitations of our polling data? YES — see the representation analysis above and accompanying disclosures in the public document.

2. Have we avoided using aggregate group statistics to make individual-level predictions ("ecological fallacy")? YES — all analysis is conducted at the appropriate level of aggregation.

3. Have we sought community input on our framing of underrepresented groups? PARTIAL — ODA consulted with two Vega County civic organizations and one Riverside County civil rights group. Additional community review was limited by timeline.

4. Have we clearly identified voter access concerns and refrained from treating them as background noise? YES — Section 8 of this document treats voter access as a primary analytical concern, not a footnote.

5. Have we been transparent about what we don't know? YES — data gaps are identified explicitly throughout this document.

6. Could this analysis, as published, be used to suppress or discourage voting? REVIEWED — ODA has determined that nothing in this document provides tactical information that could facilitate voter suppression. The voter access concerns are documented as threats to democratic participation, not as strategic vulnerabilities for exploitation.

Section 9: Conclusions and Recommendations

The Six Audit Questions Answered

Question 1 — What does the polling evidence actually show?

The best available polling evidence shows a closely contested Senate race in which Maria Garza holds a narrow but consistent lead. The ODA quality-weighted average, based on fourteen polls with methodological quality weighting applied, places Garza at approximately 48.3% to Whitfield's 46.9% — a 1.4-point lead with a credible interval of ±2.1 points. The race appears to have tightened modestly (by approximately 0.8–1.0 points) in the Day -50 to Day -40 window, coinciding with the peak of outside Republican group advertising activity. Polling quality in this race has been adequate for nonpartisan organizations but inconsistent across the full pool of public surveys; consumers of polling data should give substantial weight only to the A and B+ grade polls.

Question 2 — How are different demographic groups positioned, and what does the electoral geography tell us?

The state's electoral geography shows an accelerating urban-suburban-rural polarization that mirrors national trends. Riverside County and Metro 2 are Garza's banking territory; Redstone and rural counties are Whitfield's bank. Millbrook County is the primary contested terrain, with recent Republican gains among college-educated suburbanites at risk of erosion from Garza's education-realignment appeal. The Hispanic and Latino electorate is genuinely contested — not the monolithic Democratic bloc assumed by some Democratic strategists — with meaningful variation between established Mexican-American communities (more reliably Democratic) and Cuban-American and South American-origin voters (substantially more Republican). Black voter turnout in Riverside County is the single highest-leverage structural variable in the race.

Question 3 — What do turnout scenarios suggest about likely outcomes?

The medium (baseline) turnout scenario, with 60% overall turnout consistent with competitive 2018 and 2022 patterns, produces a Garza win of approximately 9,000 votes. A low-turnout environment (55%) narrowing to approximately 2,500 votes creates conditions in which Whitfield's path to victory is plausible. A high-turnout environment (65%) would likely widen Garza's margin to 20,000+ votes. Early vote data, running approximately 18% above 2022 pace, is slightly more consistent with the medium-to-high scenario. The registration changes since 2022, particularly the addition of younger and Latino-origin registrants in Riverside and Vega counties, provide a structural advantage for Garza in higher-turnout environments.

Question 4 — How is each campaign shaping — and being shaped by — the media environment?

Both campaigns are running well-structured advertising strategies. Garza's most effective advertising integrates her personal narrative with border security credibility and economic populism — a combination that addresses her structural vulnerabilities directly. Whitfield's most contested advertising (the death penalty ad) has been widely fact-checked as misleading but has run continuously, reflecting a calculation that emotional resonance outweighs fact-check blowback among the targeted suburban audience. The pro-Whitfield outside spending advantage in broadcast television has been a material factor in the race's tightening. Spanish-language media coverage has been substantially more favorable to Garza, a structural advantage in Vega County and Riverside's Latino neighborhoods.

Question 5 — What does campaign finance tell us about strategic priorities and resource advantages?

Garza holds a campaign-level fundraising advantage of approximately $3.4 million (cycle total), driven by a large and highly motivated small-dollar donor base. The pro-Whitfield outside spending ecosystem, however, holds a modest overall advantage in total dollars deployed — approximately $1.1 million. The most strategically significant financial finding is the geographic concentration of outside spending: the pro-Whitfield ecosystem is outspending the pro-Garza ecosystem in the Millbrook County broadcast market by approximately 19%, a difference that may be driving the race's suburban tightening.

Question 6 — Is this race being analyzed equitably?

Not entirely. Hispanic and Latino voters are underrepresented in the A-grade polling sample by approximately 8 percentage points; young voters are underrepresented by 5–11 points. Voter access concerns in Vega County (polling place consolidation) and Riverside County (registration purge litigation) are genuine threats to equitable participation that are underreported in mainstream political coverage. Dark money spending — particularly the Heritage Alliance's $3.1 million in non-disclosed-donor spending — represents a transparency gap that undermines the accountability function of campaign finance disclosure. ODA recommends that these equity concerns receive prominent placement in all public-facing coverage of this race.

Key Findings and Their Implications for Democratic Accountability

The Garza-Whitfield audit yields six key findings with implications beyond this individual race:

1. The most reliable polling signal is the nonpartisan consensus, not the range. The 8-poll nonpartisan consensus of Garza +1 to +3 is more informative than the 6-point spread between partisan polls in either direction. Journalism that treats partisan polls as equally credible to nonpartisan polls does a disservice to its audience.

2. Demographic complexity demands demographic humility. The Hispanic and Latino electorate in this state is not a monolithic bloc; it is a coalition of communities with distinct political identities. Any analysis that relies on a single "Latino vote" variable is suppressing information, not summarizing it.

3. Turnout modeling is the highest-leverage analytical activity in a close race. The difference between the low and high turnout scenarios is approximately 20,000 net Garza votes — more than twice the base-case margin. Campaigns and analysts who model turnout carefully are playing a different game than those who use only registration and polling.

4. Outside spending transparency is inadequate. The combination of super PAC disclosures and dark money non-disclosure creates a fragmented accountability record that makes it difficult for the public to understand who is paying for the advertising shaping this race.

5. Voter access concerns are not peripheral to the analytical picture. Polling place consolidations, registration purges, and language access gaps are not separate policy debates — they are factors that shape the electorate being analyzed. An audit that ignores them is analytically incomplete.

6. Probability distributions are more honest than point predictions. The ODA forecast of Garza 64% / Whitfield 36% communicates both the direction of the evidence and the genuine uncertainty more honestly than a headline declaring "Garza leads."

What ODA Will Publish and With What Caveats

The public version of this audit will include all sections of this document, with the following caveats prominently displayed:

• All polling data reflects surveys conducted through Day -9; last-minute developments are not captured.
• The demographic analysis is based on registered voter estimates, not actual voter behavior; results will depend on who actually votes.
• The campaign finance analysis is current through FEC reporting deadlines; spending in the final 10 days may differ from patterns described here.
• The equity audit reflects ODA's best assessment of publicly available information about voter access; conditions may have changed and community-level reporting from partner organizations may surface additional concerns.

Ethical Considerations in Public-Facing Political Analytics

Adaeze returns to the whiteboard on the morning before the audit goes live. "Before we publish," she says, "I want us to be clear-eyed about what we're doing and what we're not doing."

What ODA is doing: providing the most rigorous, transparent, and accountable analysis of available evidence, so that voters, journalists, and civic organizations have better tools for understanding this race.

What ODA is not doing: predicting the winner, endorsing a candidate, generating data that campaigns can use for targeting (ODA's methodology is published so that both campaigns have equal access to ODA's analytical logic), or suppressing uncertainty in order to generate a clean narrative.

Sam Harding, who has spent six weeks living inside this dataset, adds a final observation: "The most important thing a political analyst can do is remain genuinely uncertain. Not performatively uncertain — where you hedge everything so you can never be wrong — but genuinely open to being surprised. Because democracy should surprise you. The whole point is that we don't know, in advance, what millions of people will decide when they walk into that booth."

This audit is a contribution to democratic accountability. It is not a substitute for voting, organizing, or the messy, irreducible human process by which political power is contested and conferred.

Capstone Reference Tables

Table 5: ODA Quality-Weighted Poll Average — Calculation Worksheet

Weighted Garza average: (0.60×47 + 0.15×44 + 0.60×48 + 0.20×51 + 0.20×44 + 0.80×48 + 0.80×49 + 0.60×47 + 0.25×45 + 1.00×49 + 1.20×48 + 0.90×48 + 1.20×49) ÷ 7.50 = 48.3%**

Weighted Whitfield average: (0.60×45 + 0.15×49 + 0.60×46 + 0.20×44 + 0.20×50 + 0.80×46 + 0.80×46 + 0.60×47 + 0.25×48 + 1.00×47 + 1.20×47 + 0.90×47 + 1.20×47) ÷ 7.50 = 46.9%**

Table 6: Turnout Scenario Comparison

The Whitfield win scenario requires low overall turnout AND Riverside County underperforming by 3+ points vs. baseline AND Millbrook County breaking R+5 or better.

This capstone document was produced by the OpenDemocracy Analytics team under the direction of Adaeze Nwosu. All characters, campaigns, and state data are fictional composites created for instructional purposes. The analytical methods described are drawn from real-world political analytics practice and the preceding chapters of this textbook.

End of Chapter 42 Capstone Document