
Degrees without dividends: The pay gap for black graduates
December 26, 2024
Black Community Resources for Data-Driven Decision-Making
January 16, 2025In June 2022, the Canadian firm of Dunn Pierre predicted that the New Democratic Congress (NDC) would be victorious in the upcoming elections. This article dives into the science behind Dunn Pierre Barnett and Company Canada Ltd.’s (DPBA) accurate predictions of the Grenada 2022 Elections.
ccording to the Director of research and Dr. Justine Pierre, a Grenadian by birth, indicating that the prediction was based on using data analytics, data from the Grenada telephone directories, psychographic profiling, data from the Grenada diaspora, data on the Grenada Voter’s universe, and behavioural microtargeting to influence public opinion and voter behaviour.
He added that the company had over 17 different focus group discussions with potential voters and applied data science, psychological research, and artificial intelligence to analyze large datasets on the Grenadian voting population and monitored targeted marketing campaigns from both political parties through the use of social media, rallies and other outlets which were used to inform voters.
In a recent interview with Dr. Pierre, he broke down the core sciences and techniques behind their operations, which followed 8 scientific steps. They were as follows
1. Establishing the Grenadian voters universe:
Initially, the team stratified the Grenadian Voter Universe (VU) into 18 main criteria, and a sample was taken from each sample frame. Thus, the findings were more accurate and reliable for predicting the election results.
The voters were stratified according to the following.
- Geographic Location
in addition to urban and rural areas, the voters were stratified according to parishes, constituency polling stations, and cell phone tower locations. - Political Affiliation
Stratify by party registration or historical voting behaviour. - Age Groups
To capture generational differences in political priorities and turnout rates.
18–25, 26–40, 41–60, 61-70, 71+. This also included children’s views on how they believe their children would vote. - Gender
Male, Female - Socioeconomic Status
Group by income, education, and employment type to capture low-income vs. high-income and college-educated vs. non-college-educated. - Ethnicity or Race
Stratified by racial and ethnic identity. - Religion
Stratify based on religious or cultural identity, such as Christian, Muslim, Hindu, Indigenous, or secular. - Sporting Affiliation
Stratify based on sporting affiliations, cricket, football, dominoes, track and field, and netball. Lawn tennis, swimming, and others. - Union/ professional representation
Persons were stratified based on their affiliation to unions and other professional organisations, such as the medical and legal. Teaching, nursing and other professional organisations. - Voter Turnout History
Categorize based on past voting frequency, for example, frequency, infrequent, and newly registered voters. This was crucial to distinguish between reliable, occasional, and first-time voters. - Employment Sector
Group based on work industry or occupation. For example, in the public vs. private sector, self-employed vs. salaried. And ISCO and ISIC codes. - Education Level
Stratify by the highest level of education achieved, for example, High school, college, and postgraduate. - Household Size
Stratify by the number of people in a household and who heads the household. For example, Single-person, nuclear family, extended family households, and if a woman heads the home. - Media Consumption
Categorize voters by their primary sources of information. This was used to identify the role of media influence on political opinions, for example, how Social media, traditional news (TV, radio), or online platforms affect their views. - Urbanization Level
Stratified by levels of urbanization (city, town, village). - Language or Linguistic Groups
Group by native or spoken languages. - Civic Engagement
Stratify based on participation in civic or political activities, such as community not-for-profit or volunteerism. - Support for Maurice Bishop/ Bernard Code/ Eric Gairy/ Innocent Belma
The voters were stratified based on the support of Grenadian leaders and other political figures.
2. Psychographic Profiling:
Interestingly enough, psychographic profiling went beyond demographic data (age, gender, income). The team analyzed personality traits, values, attitudes, and lifestyle to predict behavior.
The company used the “Big Five” personality model (OCEAN model) to assess:
- Openness to experience
- Conscientiousness
- Extraversion
- Agreeableness
- Neuroticism
2) By profiling users based on these traits, the team could predict voting behaviour with over 95% accuracy.
3. Data Collection and Mining:
DPBA mined massive amounts of personal data, including from social media platforms like Facebook, the Grenada telephone directory, the Grenada election database, and other available public data. Additionally, the team leveraged quizzes, apps, and surveys to harvest direct user data and data from users’ social connections. This provided insights into users’ preferences, interests, and psychological makeup.
4. Like and share ratios
The like and share ratios are more than just numbers; they are indicators of content effectiveness and audience connection in social media. By understanding and leveraging these metrics, the team could predict the reach, effectiveness of political ads and surveys, credibility, and influence.
5. Behavioral Microtargeting:
Microtargeting involves crafting personalized messages or advertisements tailored to very specific population segments, such as voters in a constituency or persons in a parish, city, province or state and country. Leveraging psychographic data, political ads and campaigns can be designed and sent to any specific subsector via emails or social media campaigns that are emotionally and cognitively aligned with users’ psychological profiles and cognitive biases. For instance, someone scoring high in neuroticism might receive fear-driven political ads, while someone high in openness might see more progressive, idealistic messages.
6. Machine Learning and Predictive Analytics:
To begin, DPBA used machine learning algorithms to detect patterns in massive data sets and predict how individuals or groups might respond to different messages. By feeding in demographic, psychographic, and behavioral data, it was able to predict voting patterns and consumer decisions with high precision.
7. A/B Testing and Message Optimization:
The company used A/B testing to see which types of messages resonated best with different audience groups. They constantly refined messaging strategies based on real-time feedback from social media interactions.
8. Political and Marketing Applications:
DPBA has been involved in many other campaigns in various countries, however in Grenada it was the first time we predicted the elections with such accuracy. This was done by recording all political speeches, capturing and using facial recognition data, and using artificial intelligence. The issue of Artificial Intelligence can also be used to leverage emotional triggers like fear, anger, and hope to shape perceptions and influence decision-making.
Dr. Pierre emphasizes the evolving landscape where data plays a pivotal role in modern decision-making processes, particularly in forecasting election results and evaluating voter satisfaction. He advocates for political parties to transition towards data-driven decision-making rather than relying solely on intuition. Utilizing data, including surveys and polls, offers valuable insights into voter preferences, demographic patterns, and behavioural trends, enhancing the ability to forecast potential election outcomes accurately.
For further details, individuals can explore the company website or access the report on Grenada’s Election Predictions for 2022.
For inquiries, reach out to Dr. Pierre via email at info@dpbglobal.com