Data Analysis for PolicyMaking in Georgia
Module 2 · Reading Voting & Registration Data › 2.4 Spotting data quality issues
Subsection 2.4

Spotting data quality issues

~3 min

Reading

Every voter file has data quality problems. Treating raw voter data as clean is the most common analytical mistake. Before you calculate any rates or draw any conclusions, you need to know what is wrong with your data—and whether it is wrong in ways that affect your specific analysis.

Common data quality issues in Georgia's voter file

1. Missing registration dates

Some older records in the Georgia voter file carry null or implausibly early registration dates (e.g., 1900-01-01, which indicates a data migration artifact, not a 125-year-old voter). If your analysis involves registration recency—"how many voters registered in the last 30 days?"—you must either exclude null-date records from the denominator or document that exclusion clearly.

2. County code mismatches

Georgia uses numeric county codes internally (1–159). The file may include a FIPS code field that uses a different numeric scheme. If you join voter file records to census data using county codes without verifying the coding scheme, you will mismatch counties. Always check: are you joining on GA county code or FIPS code? These are different.

3. Duplicate entries

Voters who moved within Georgia and re-registered may appear twice—once with a canceled status at the old address and once as active at the new address. If you are counting registered voters and include inactive/canceled records, you will overcount. Always filter to status = "Active" before counting.

4. Race/ethnicity coded as "Unknown"

Approximately 15–20% of Georgia voter records have race/ethnicity coded as "Unknown." This is not random—it correlates with registration method and county. Any racial demographic analysis must account for the Unknown category. You cannot simply ignore those rows without checking whether they are disproportionately from specific counties or demographic groups.

5. Precinct assignment lag

After a redistricting cycle, precinct assignments in the voter file may lag behind actual boundary changes for several months. A voter who should be in the new Precinct 14-B may still show Precinct 14-A in the file. Cross-check precinct assignments against official boundary shapefiles if your analysis depends on current precinct-level geography.

Best practice: document your exclusions

Every time you filter records, note: what you filtered, how many records you excluded, and why. Your Policy Data Brief should include a methodology note that describes these decisions.

Georgia Voter File — Common Data Quality Issues Issue What to look for Fix / mitigation Missing reg. dates Null or 1900-01-01 in registration_date field Exclude from recency analyses; document count County code mismatch GA county code (1–159) ≠ FIPS code Use a crosswalk table; verify join key before merging Duplicate entries Same person, two rows (old + new address) Filter to status = "Active" before counting Race coded "Unknown" ~15–20% of records; not random by county Analyze Unknown distribution; disclose in findings Precinct assignment lag Post-redistricting, file lags boundary changes Cross-check against official boundary shapefiles Document every exclusion decision in your methodology note. Never silently drop records.
Diagram 2.4 · Common Georgia voter file data quality issues. Each issue requires a deliberate fix. Document every exclusion in your methodology note.