Your cart is currently empty!
*Disclaimer: The values and predictions presented in this blog post are entirely random and do not reflect actual findings. Additionally, this project was developed independently out of personal curiosity and was not conducted on behalf of the PvdA.
Optimizing Campaign Strategy for the 2026 Municipal Elections
With the 2026 municipal elections coming up, efficient campaigning is crucial, given that there are only so many volunteers. This model helps identify key neighborhoods with potential for vote growth—places where campaigning can truly make a difference.
To achieve this, it is crucial to first predict the expected voting behavior in each neighborhood for the upcoming elections. For this purpose, I developed a machine learning model trained on approximately 500 characteristics per neighborhood. These features range from factors such as past voting behavior to the average WOZ property value.
Relying solely on previous election results for each neighborhood can be unreliable. This is evident in the comparison below (each of the 511 blue dots represents a neighborhood in Amsterdam).
- The left plot shows the model’s predictions (x-axis) versus the actual results of the 2022 municipal elections (y-axis).
- The right plot compares the 2021 national election results (x-axis) with the 2022 municipal election outcomes (y-axis).
As shown, the model achieves a much higher level of accuracy, with an average error of only 1.37% per neighborhood.
*The red dotted line indicate perfect predictions.

Neighbourhood predictions

This map of Amsterdam shows the model’s predictions for the 2026 municipal elections in each neighborhood. However, as previously mentioned, the values in this map are randomly generated.
Election Results

This plot displays for each specific neighbourhood the PvdA election results for the past 3 election cycles, together with the prediction for the municipality elections in 2026. The error bar around the prediction for the ‘TweedeKamer23‘ reflects degree of consensus between the voting locations within one neighbourhood concerning PvdA.
Beneath the plot you see the ‘opkomst %’, voter-turnout. This is calculated as the amount of people who voted in that neighbourhood divided by the amount of people who were invited to vote in that neighbourhood.
Key Characteristics

This plot displays the characteristics from which this specific neighbourhood deviates from Amsterdam as a whole. The red bars indicate that, for that specific characteristic, the neighbourhood scores below average, i.e., Overtoomse Veld-Noord has below average drinking problems. The blue bars indicate on which characteristics the neighbourhoods score above average.
Top Potential and Top Contenders

The plot on the left displays the top 9 neighbourhoods that are estimated to have the most potential for voter growth given campaigning efforts. The plot on the right displays the 4 political parties who received the most votes last election cycle, i.e., the top 3 or 4 contenders.
BartAmin
Leave a Reply