Participants could agree, disagree or pass. Over a period of three weeks, 1,398 people participated; 877 statements were submitted and voted on nearly 124,000 times. All participation was anonymous. The organizers exercised some light moderation to ensure that submitted statements were clear, unique and civil.
Because participants self-selected, the survey is not statistically representative of the views of Louisville-area residents. However, because Polis tracks patterns of voting among subgroups, we can see how lines of agreement and disagreement emerge within the respondent community, and how those lines shift as respondents explore issues through the statement process.
Polis clusters participants into larger voting blocs based on their responses to the statements. In the Louisville voting community, this process distinguished two broad groups: a small Group A (128 people) with generally conservative views and a much larger Group B (1,037 people) with more liberal views. People who voted fewer than 7 times count in the overall results but not in the groupings. So while 1,398 people voted overall, only 1,165 were grouped.
Here is what that grouping looks like to Polis:
There is a lot of information compressed into this graph, but the main point is that the statements (each represented by a number) that receive the most support across all participants cluster in the middle, while the more divisive statements are placed toward the outer edges. The concentration of statements in the middle indicates a high degree of consensus across most of the submitted statements. The consensus blob defines Group B, while Group A is defined mostly by its distance from this central consensus. (If you want to explore this in more detail, see the full Polis technical report for the conversation.) Because the conversation passed through user-submitted statements, there was no effort to introduce statements that would organize votes into conventional ‘liberal’ and ‘conservative’ voting blocs. Although it is easy to characterize Group B as more liberal most of the time, the most interesting Polis results are those where liberal/conservative labels don’t adequately explain opinion. And there are many of those results. It’s worth noting that the self-identified political leanings of the Louisville Public Media audience generally mirror voter registration in Jefferson County.
The same strong tendency toward agreement can be visualized along a single axis, with the high consensus statements on the left and the divisive ones on the right:
Here’s the most divisive statement, marked by sharp differences between Groups A and B:
And here’s one on which nearly everyone agreed:
How Should I Read These Bar Charts?
Polis serves up statements in semi-random order, favoring statements that have received the fewest “passes.” In practice, not everyone will see every statement—especially those statements that arrive late in the process. In a Polis result like the one above, the white section on the right side of the bar represents the number of people who didn’t see the statement. The gray numbers in parentheses are the total number of people in each group for that statement.
Because the Polis survey is generated by its participants, it can go in any direction. There are no fixed questions or topics. There is, however, loose path dependency as participants build up an impression of the process based on the statements they initially encounter. For our part, we provided a framing question designed to encourage respondents to think locally. Nearly all of the statements referenced local community issues such as traffic, development, zoning, schools and municipal governance. Some provided local inflections of issues with wider state and national profiles—statements about climate change, for example. Only a few referenced decision-making or policy at the state or federal levels.