A multidimensional political map from 977 survey answers
A proof of concept that places each respondent on three axes (economy, society and institutions) and compares how people label themselves, what they answer and which leaders they identify with.
Matías Gaglio ·
Leer en españolHow people label themselves does not always match what they answer.
Political labels summarize a lot but hide differences. I built a proof of concept to study three things that do not necessarily line up: the identity each person declares, the positions they take on concrete questions and the leaders they identify with.
From a form to a three-axis space.
Collect
Responses land in Google Sheets; a GitHub Actions workflow downloads them hourly as CSV.
Score
An audited key maps each option to a position on economy, society and institutional politics.
Analyze
Python with pandas, SciPy and scikit-learn computes associations, principal components and per-group dispersion in an isolated environment with no network access.
Publish
Plotly and JavaScript draw the charts; only aggregates and coordinates without personal identifiers are published.
A group's shape says as much as its center.
Besides where each identity sits on average, I measure how spread out its members are. A compact cloud shows agreement on the measured dimensions; a wide cloud gathers diverse positions under the same label. That describes the distribution and does not imply more openness or rigidity.
| Identity | People | Spread |
|---|---|---|
| Social democracy | 57 | 11.0 |
| Center | 42 | 13.7 |
| Peronism | 159 | 17.1 |
| Liberalism | 154 | 22.2 |
| Libertarianism | 103 | 25.1 |
| Left / socialism | 83 | 25.1 |
| Right / center-right | 129 | 31.0 |
| No declared identity | 53 | 47.0 |
- People with no declared identity are the most spread-out group: they do not form a bloc, and the missing label groups different phenomena that should not be merged.
- The self-declared right is more heterogeneous than Peronism or social democracy on the measured dimensions.
- The questions that best separate identities are about modes of production, markets and banks (bias-corrected Cramér’s V between 0.62 and 0.74).
- The outlet people get their news from is strongly associated with the problem they rank first (V = 0.52).
What this analysis does not say.
- Not representative: it is a convenience sample without population weighting; it describes this dataset, not the country.
- Association is not causation: relationships are reported with effect sizes and a correction for multiple comparisons.
- Small cells: some cross-tabs have low expected frequencies, so those values are read with caution.
- Privacy: no table linking demographics to individual political answers is published.