Capstone 1 Student Guide: The Battleground State Audit

What This Capstone Is

Capstone 1 is the first of three extended project assignments that form Part IX of Political Analytics: From Populism to Polling. Unlike chapter exercises and case studies — which ask you to apply a specific method to a defined problem — capstone projects ask you to integrate methods from across the textbook to produce a coherent, publication-quality analytical document.

The model document you just read — the OpenDemocracy Analytics "Sun Belt State Political Audit" — is both an example and a scaffold. It shows you the kind of work the capstone expects. It also provides a structure you can follow, adapt, or argue against as you construct your own audit.

This guide explains: what you're being asked to do, what resources you have, how to organize your work, what the evaluation criteria are, and how to avoid the most common errors students make on this project.

Your Task

You will produce a comprehensive analytical audit of a competitive election — real or simulated — organized around the six audit questions developed in the capstone document:

1. What does the polling evidence actually show?
2. How are different demographic groups positioned, and what does the electoral geography tell us?
3. What do turnout scenarios suggest about likely outcomes?
4. How is each campaign shaping — and being shaped by — the media environment?
5. What does campaign finance tell us about strategic priorities and resource advantages?
6. Is this race being analyzed equitably?

Your audit must be a standalone document that a knowledgeable reader with no prior exposure to your race could use to understand what is happening and why.

Option A: Use the Garza-Whitfield Framework (Recommended for Most Students)

You analyze the Garza-Whitfield Sun Belt race as described in the capstone document. You use the data provided — the poll table, the demographic tables, the campaign finance figures, the advertising data — as your empirical foundation. Your task is to conduct your own analysis of this data: construct your own quality-weighted polling average (showing your work), build your own turnout scenario model, develop your own forecast with probability distribution, and produce your own equity audit.

You are not reproducing the capstone document. You are using the same data to conduct your own analysis, which may reach similar or different conclusions. If your polling average differs from ODA's, explain why. If your turnout model produces different baseline assumptions, justify them. Intellectual disagreement with the capstone document — when well-reasoned — is evaluated positively.

Option B: Analyze a Real Recent Election (For Students with Research Access)

You identify a real, recently completed competitive Senate, gubernatorial, or major local election for which public polling, campaign finance (FEC filings), and media data are available. You collect and organize this data yourself, construct your own analytical framework, and produce an audit structured around the six questions above.

Note: Option B requires substantially more upfront data collection work. Before choosing Option B, confirm with your instructor that your chosen race has sufficient public data available across all six audit dimensions. Races with fewer than five public polls and limited campaign finance disclosure are not suitable.

Option C: Design and Analyze a Simulated Race (Advanced Option)

You design a fictional competitive election — defining the candidates, state demographics, electoral history, polling environment, and campaign context — and then conduct an analytical audit of your own creation. This option requires the most imagination and analytical rigor, because you are responsible for ensuring that your simulated data is internally consistent and realistic.

Option C is appropriate for students who wish to focus on the methodological dimensions of the audit — the construction of quality-weighted averages, turnout models, and probability distributions — rather than data collection and political context.

Scope and Length Requirements

Your final audit document must be a minimum of 8,000 words and a maximum of 15,000 words (not counting tables, figures, and reference lists). All six audit questions must be addressed, but you have flexibility to weight them according to the strengths of your data and analysis.

The document must include at minimum: - An introduction establishing the analytical framework and data sources (500–800 words) - A polling analysis section with a quality-weighted polling average (your own calculation, shown step-by-step) - A demographic and geographic analysis section - A turnout modeling section with at least three scenarios (low, medium, high) - A media and advertising analysis section - A campaign finance analysis section - A forecasting section with a point estimate, probability distribution, and sensitivity analysis - An equity and representation audit - A conclusions section answering all six audit questions

How to Organize Your Work

The capstone is designed to take 15–20 hours of sustained analytical work over approximately two to three weeks. Students who begin the capstone one week before the deadline consistently produce work of lower quality than students who spread their work across the full window. Political analysis is iterative: your demographic model will inform your turnout model; your turnout model will inform your forecast; your forecast will raise questions that send you back to the polling analysis. This recursion takes time.

Suggested Work Schedule

Days 1–3: Data assembly and race orientation

Collect all polling data for your race and organize it in a spreadsheet. Collect all relevant campaign finance data (FEC filings, if using a real race). Download or compile county-level election results from at least the three most recent comparable elections. Identify at least three major news outlets that have covered your race and compile a representative sample of articles. If using Option A, re-read the relevant sections of the capstone document and identify where you might conduct an independent analysis that differs from ODA's.

Days 4–6: Polling analysis

Construct your poll quality grading rubric (drawing on Chapter 10's methodology). Apply the grading criteria to each poll in your dataset. Show the grade calculations for at least three polls in detail in your document. Compute your quality-weighted polling average step-by-step, matching the structure of Table 5 in the capstone document. Construct a trend line. Write your polling analysis section in full.

Days 7–9: Demographic and turnout modeling

Build a county-level or district-level demographic table analogous to Tables 2 and 3 in the capstone document. Develop three turnout scenarios (low, medium, high) with explicit assumptions for each county and demographic segment. Run at least one counterfactual scenario (e.g., "what if Black voter turnout increases 5%?"). Write these sections in full.

Days 10–12: Media, finance, and equity analysis

Analyze advertising (for real races: FCC filings and AdImpact or similar trackers; for Option A: the data in the capstone document). Identify and analyze the core campaign narratives. Review fact-check coverage. Analyze FEC data for fundraising breakdown. Apply the equity checklist to your own analysis — what are the data gaps, and who is hardest to reach? Write these sections.

Days 13–14: Forecasting and conclusions

Integrate your analyses into a forecast. Compute a point estimate and probability estimate. Conduct sensitivity analysis. Write the conclusions section answering all six audit questions explicitly. Draft the executive summary or introduction last.

Day 15+: Revision and formatting

Read your document as if you've never seen the race before. Does the logic flow? Does the evidence support the conclusions? Are all data sources cited? Is the equity audit thorough? Are limitations clearly disclosed?

Using Chapter Methods

The capstone integrates methods from the following chapters. Where the assignment asks you to use a method, the chapter it comes from is indicated.

Collaboration Policy

Capstone 1 is an individual project. You may discuss your approach with classmates and seek feedback on drafts, but all analysis and writing must be your own. The polling average calculation, turnout model, and forecast must be your independent work — group-constructed tables are not permitted even if the underlying data is shared.

Students using Option A may consult each other about the data (since all Option A students share the same dataset), but your analysis, interpretation, and written document must differ substantively from your classmates' work. Identical or near-identical polling averages or turnout tables (indicating copied calculations) will be treated as academic integrity violations.

Common Errors and How to Avoid Them

Error 1: Treating the poll table as the analysis. Many students spend the bulk of their polling section describing what each poll said, rather than analyzing the polls as a data ecosystem. The analysis is not "Poll 7 showed Garza ahead by 3 points." The analysis is "When we weight Poll 7 by its grade and recency and compare it to the seven other nonpartisan polls in the window, it contributes [X] to the weighted average, consistent with the consensus signal."

Error 2: Ignoring the equity audit. Students who are most comfortable with quantitative analysis frequently underinvest in Section 8 (Equity and Representation). The equity audit is not a box-checking exercise. It is a substantive analysis of who is and is not represented in the data, what structural factors are limiting democratic participation, and what an honest analyst should disclose about these gaps. A weak equity audit will significantly affect your grade.

Error 3: Presenting a probability as a fact. A forecast probability of 64% Garza is not a prediction that Garza will win. It is a statement about the distribution of possible outcomes given available evidence and model assumptions. Your document should make this distinction clearly, and your language throughout the forecasting section should reflect genuine uncertainty rather than false precision.

Error 4: Separating the financial analysis from the strategic analysis. Campaign finance data is only interesting analytically when it is connected to strategic behavior. The question is not "how much did each campaign raise?" The question is "what does spending pattern tell us about each campaign's theory of victory, and does that theory appear consistent with the demographic and geographic analysis?"

Error 5: Describing advertising rather than analyzing it. Message analysis means more than summarizing what ads say. It means situating the messaging within the campaign's overall strategy, assessing its targeting logic, evaluating its consistency with the polling and demographic evidence, and identifying where it is being deployed geographically.

What a Strong Capstone Looks Like

A strong capstone integrates its sections rather than stacking them. The polling analysis informs the forecasting section; the demographic analysis shapes the turnout model; the media analysis contextualizes the campaign finance patterns; and the equity audit runs as a thread through all of them.

A strong capstone is honest about what it doesn't know. Where data is unavailable, a strong capstone says so clearly and explains what the gap means for analytical confidence. Where assumptions are necessary, a strong capstone states them explicitly and explores how the analysis changes when assumptions change.

A strong capstone is written for a reader who is not you. It doesn't assume familiarity with the data; it builds the reader's understanding methodically. It doesn't skip steps in the calculation; it shows its work. It doesn't collapse uncertainty into a confident prediction; it preserves the honest range of outcomes.

A strong capstone is positioned as a contribution to democratic accountability. It isn't just a data exercise — it is an act of civic responsibility. The best student audits will, in the course of rigorous analysis, surface something the political media missed: a polling house effect nobody reported, a demographic trend nobody noticed, a campaign finance pattern nobody connected to an advertising strategy.

Submission Format

Submit your capstone as a single document in Markdown (.md) format, with YAML frontmatter. Include all tables within the document (not as separate files). If you produce any supplementary code (for the polling average calculation, turnout model, or other analyses), include it as clearly labeled code blocks within the document.

Your document should include: - YAML frontmatter with: title, your name, option (A/B/C), race analyzed, word count - All sections in order (Introduction through Conclusions) - All calculation tables (polling average worksheet, turnout scenarios, etc.) - A References section listing all polls, data sources, news articles, and FEC filings cited

Submit to your course management system by the deadline specified by your instructor.

A Final Note from Adaeze (and Your Authors)

In the capstone document, Adaeze Nwosu tells her team: "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."

That sentence should guide your work on this project. Political analysis at its best is a service to democratic culture — it gives citizens, journalists, and civic organizations the analytical tools to hold power accountable. That purpose doesn't require you to be neutral about the value of democracy, or about the importance of voting access, or about the obligation of campaigns to tell the truth in their advertising. It does require you to be rigorous about evidence, transparent about methods, and honest about uncertainty.

That combination — rigor, transparency, and honest uncertainty — is what this capstone is designed to develop. We look forward to reading your audits.