Wobbly Geek

Outlier Detection in Election Data Using Geospatial Analysis

Wobbly Geek's photo
Wobbly Geek
·

3 min read

Case Study: Ensuring Election Integrity

Introduction

In the recently concluded election, the Independent National Electoral Commission (INEC) has faced multiple legal challenges concerning the integrity and accuracy of the election results. Allegations of vote manipulation and irregularities have been widespread, prompting a thorough investigation into the matter. This report aims to uncover potential voting irregularities by identifying outlier polling units where the voting results deviate significantly from neighbouring units. We shall use Kaduna State for this exercise.

Task Overview

The analysis involves the following stages:

  1. Dataset Preparation

  2. Neighbour Identification

  3. Outlier Score Calculation

  4. Sorting and Reporting

Dataset Preparation

Steps:

  1. Kaduna state was selected for the study.

  2. Since the latitude and longitudes were not present, I used the Awesome Table to impute the data.

  3. Some missing latitude and longitude values were dropped.

Neighbour Identification

Geodesic distance was calculated to determine the distance between polling units. Units within a 1 km radius were considered neighbours.

Outlier Score Calculation

For each polling unit, the difference in votes for each party compared to the votes of its neighbouring units was calculated. The outlier score for each party was computed as the absolute difference between the unit’s votes and the votes of its neighbours.

Findings

Top 3 Outliers

  1. UNG MAL SANI/UPE KWANGILA III (APC)

    • Score: 5.9578

    • Closest Units: MAKARANTA/ K/G ALH AHMADU, JIM HARRISON HOTEL/ HARRISON HOTEL GATE, etc.

    • Explanation: Significant deviation in APC votes compared to neighbouring units.

  2. UNG. LIMAN/ K/GLIMAN (APC)

    • Score: 5.1351

    • Closest Units: UNG. KAJURU/ K/GIDAN DAKACHE, UNG. LIMAN/ K/GLIMAN, etc.

    • Explanation: Notable difference in APC votes from nearby units.

  3. KWARIN KUBANNI/ MAI ANGWAN HOUSE (APC)

    • Score: 4.9911

    • Closest Units: LAYIN SARKIN I, II M/YAKAWADA/ L.E.A PRIMARY SCHOOL, U/MAIDOKI/ U/MAIDOKI, etc.

    • Explanation: APC votes at this unit deviate significantly from neighbours.

  4. GRACE LAND/BUKS INTER SEC SCHOOL (LP)

    • Score: 9.7309

    • Closest Units: MAKARANTA/ K/G ALH AHMADU, JIM HARRISON HOTEL/ HARRISON HOTEL GATE, etc.

    • Explanation: LP votes at this unit are substantially higher than surrounding units.

  5. BIRNIN GWARI ST. BY BODA (LP)

    • Score: 9.6359

    • Closest Units: AREA COURT, MAIGWARI PRY. SCH., etc.

    • Explanation: Significant deviation in LP votes compared to neighbours.

  6. GASKIYA I/ GASKIYA VILLAGE K/G H/SALE (LP)

    • Score: 8.6346

    • Closest Units: UNG. S. ADAMU II/ PRI.SCH., MARJI/UNG. ADAMU/ K/GIDAN MAIUNGUWA, etc.

    • Explanation: Notable difference in LP votes from neighbouring units.

  7. YUSUF NA KYAUTA/ K. YUSUF N/KYAUTA (PDP)

    • Score: 6.8728

    • Closest Units: GWAKUN K/G SARKI, GALADIMAWA II/ KOFAR GIDAN SARKI, etc.

    • Explanation: PDP votes at this unit deviate significantly from neighbours.

  8. ANG. S. FADA I/ T.V. CENTRE (PDP)

    • Score: 5.4651

    • Closest Units: YAMUSA ROAD/ NEAR ISLAMIYA SCHOOL, ANG. S. FADA I/ T.V. CENTRE, etc.

    • Explanation: Significant deviation in PDP votes compared to neighbouring units.

  9. KOFAR FADA/ KOFAR FADA (PDP)

    • Score: 5.2891

    • Closest Units: UNG. NASARAWA/ NASARAWA, KOFAR FADA/ KOFAR FADA, etc.

    • Explanation: PDP votes at this unit are notably higher than surrounding units.

  10. DANDAURA I L.E.A. PRY SCHOOL (NNPP)

    • Score: 13.2472

    • Closest Units: BAKIN KASUWA PRY. SCH., WUYA K/G SARKI, etc.

    • Explanation: Significant deviation in NNPP votes compared to neighbours.

  11. DOKA I/K.MAIUNGUWA (NNPP)

    • Score: 10.8464

    • Closest Units: AKILIBU I/ PRI.SCH. AKILIBU, AKILIBU II/ KOFAR G/SARKI, etc.

    • Explanation: NNPP votes at this unit deviate substantially from surrounding units.

  12. UNG. WAZIRI II/ K/M WAZIRI (NNPP)

    • Score: 8.1246

    • Closest Units: GDSS KAWO I, SABON BIRNIN STREET/ NEW EXTENSION JUNCTION, etc.

    • Explanation: Notable difference in NNPP votes from neighbouring units.

Conclusion

The analysis identified significant outliers in the voting patterns at various polling units, suggesting potential irregularities. These findings can aid in ensuring the transparency and integrity of the election results. Enhanced monitoring and further investigation of these outliers are recommended.