Data Analysis for PolicyMaking in Georgia
Module 5 · Building Your Policy Finding › 5.1 The anatomy of a policy claim
Subsection 5.1

The anatomy of a policy claim

~3 min

Reading

A policy finding is not just a statistic. It is a structured argument that moves from a data observation to a defensible conclusion. Understanding the parts of this argument—and where each part can go wrong—is the difference between a finding that holds up under scrutiny and one that collapses when challenged.

The four-part anatomy

1. Research question

A one-sentence question that is specific, answerable with available data, and relevant to a policy decision. Bad: "Is Georgia's voting system fair?" (too broad, not answerable with a single dataset). Better: "Do Georgia counties with higher Black voter share have higher mail ballot rejection rates, controlling for county size?"

2. Data finding

A factual statement about what the data shows, in plain language. No interpretation yet. The finding should be specific (include a number), honest about its limitations (sample size, data vintage, exclusions), and reproducible. Example: "In the 2020 Georgia general election, counties where Black voters made up more than 40% of active registrants had a mail ballot rejection rate of 2.1%, compared to 0.9% in counties where Black voters made up less than 20% of active registrants."

3. Interpretation

What does the finding mean? This is where you introduce context, comparison, and cause. Interpretation requires intellectual honesty: what does the finding support, and what does it not support? Example: "This 2.3× disparity is consistent with disparate impact—but it does not by itself establish that any specific county official acted with discriminatory intent." The post-Callais distinction between effects and intent belongs in the interpretation.

4. Recommendation

What should a policymaker do? A good recommendation is: specific (naming an action, not a vague hope); actionable (something a legislature, agency, or court can actually do); scoped to what the data actually supports; and honest about what it would not fix.

The most common error: overclaiming

Overclaiming means drawing a stronger conclusion than the data supports. Common forms: treating correlation as causation; inferring intent from disparate effects alone; generalizing from a single county to the whole state; or ignoring alternative explanations. After Callais, overclaiming in redistricting contexts carries extra risk because intent claims require a higher evidentiary standard.

The Anatomy of a Policy Claim Research Question Specific, answerable, data-grounded Output: One focused research question Pitfall: Too broad Data Finding Factual, specific, reproducible, numbered Output: "County X had Y% rejection rate" Pitfall: Vague claims Interpretation Context, comparison, honest about limits Output: "Consistent with disparate impact" Pitfall: Overclaiming Recommendation Specific, actionable, scoped to the data Output: "Legislature should require auto notice-and-cure" Pitfall: Vague hope Most common error: overclaiming — drawing stronger conclusions than the data supports After Callais: explicitly distinguish effects from intent in your interpretation.
Diagram 5.1 · The anatomy of a policy claim. Four parts, each with a specific output and a characteristic failure mode.