Who votes with whom

Three ways to explore how council members vote together. All of them count only the 73 recorded votes that were not unanimous, from 42 meetings.

Voting Alliance Network

Councillors are connected based on how frequently they voted the same way on votes both attended. Thicker lines indicate stronger agreement, while colours represent groups identified through voting-pattern analysis. These groups reflect observed voting similarities, not necessarily political affiliations.

Show agreement of at least 75% Councillor Node size Community colours

How this network is built

Lines and percentages come straight from the votes: of the 73 non-unanimous recorded votes, the share both people attended on which they voted the same way. Pairs who never voted together have no line, because that is unknown rather than 0%.

Positions: every known pair pulls toward a distance set by its agreement, including pairs below the slider, so moving the slider never rearranges the network. Layout starts from fixed positions, so it comes out the same on every visit.

Communities use the Louvain method on the full network, whatever the slider says. On average, any two members here agree 65% of the time, and with raw percentages those moderate links would blur everything into one group. So only agreement above that average counts as a link, and an unknown pair counts as no evidence either way. Modularity is 0.25, where 0 means no community structure, so these communities are real but overlapping.

How firm they are: on 300 random resamples of the votes, 3 communities came out 63% of the time. Each person's detail panel shows how often they stayed with their community. Bridging marks someone whose average agreement within their community is less than 10 points above their average with everyone else; that cutoff is a judgement call.

These communities can differ from the voting groups further down the page, which use a different method. Comparing the two shows which memberships are firm: people in the same group under both are the most consistent allies.

Agreement circle

The thicker the line, the larger the share of votes on which two council members voted the same way.

32%64%95%of votes in agreement

Select a name to see only their lines, or select a line for its count.

Agreed most

Leanne Caron & Katherine Hauser95% (20 of 21)

Katherine Hauser & Carly Klassen95% (20 of 21)

Cathy Downer & Katherine Hauser91% (20 of 22)

Cathy Downer & Carly Klassen91% (60 of 66)

Phil Allt & Linda Busuttil88% (60 of 68)

Agreed least

Erin Caton & Dan Gibson32% (21 of 65)

Christine Billings & Erin Caton35% (24 of 68)

Phil Allt & Christine Billings35% (23 of 65)

Linda Busuttil & Dan Gibson38% (24 of 64)

Phil Allt & Dan Gibson39% (24 of 62)

Voting groups

Each portrait is a council member. The closer two portraits are, the more often those two voted the same way. Hover over or select someone to see how often they agreed with each colleague, as a percentage of these votes that both attended. The groups agree with each other on only 55% of these votes.

Group A

Agree with each other on 80% of these votes. Dominique O'Rourke fits this group least firmly.

Phil Allt

Linda Busuttil

Erin Caton

Leanne Caron

Katherine Hauser

Cathy Downer

Carly Klassen

Dominique O'Rourke

Group B

Agree with each other on 76% of these votes.

Christine Billings

Ken Yee Chew

Dan Gibson

Michele Richardson

Mayor Cam Guthrie

Between the groups

About as close to one group as the other.

Rodrigo Goller

How the groups were found

Starting from how often each pair disagreed, multidimensional scaling placed everyone on this plane; two dimensions hold about 76% of the differences, so distances are close but not exact. Hierarchical clustering (average linkage) then formed the groups, and the silhouette score chose how many: 2 groups 0.44, 3 groups 0.42, 4 groups 0.33, 5 groups 0.33, where higher means better separated. Clustering used the full data, not the 2D positions.

To test this, we re-ran it on 1,000 random resamples of the 73 votes. 2 groups came out best in 72% of them, and members of a group stayed together 92% to 94% of the time. Rodrigo Goller is shown between them because their silhouette score is 0.03, below our 0.1 cutoff; in the resamples they land with Group A 51% of the time and B 38% of the time. Dominique O'Rourke is the loosest fit, with a silhouette of 0.29 (everyone else in a group is 0.34 or higher), and lands with Group A in 81% of resamples.

People who never voted together have no real distance, so theirs is estimated from how alike they are with everyone else. With 73 votes these are working patterns, not fixed blocs. Left and right on the map come from the maths and don't mean political left and right. Portraits with near-identical records are eased apart so each can be seen, so very small gaps mean nothing. Group names are arbitrary.

The numbers

Every pair, as a table

Percent of the votes both attended on which they agreed.

	Mayor Cam Guthrie	Michele Richardson	Dominique O'Rourke	Erin Caton	Phil Allt	Linda Busuttil	Katherine Hauser	Leanne Caron	Cathy Downer	Carly Klassen	Rodrigo Goller	Ken Yee Chew	Dan Gibson	Christine Billings
Mayor Cam Guthrie		72	61	42	46	48	73	56	72	69	51	75	66	75
Michele Richardson	72		82	48	54	53	59	62	70	68	62	82	72	75
Dominique O'Rourke	61	82		74	72	70		85	82	82	62	80	54	60
Erin Caton	42	48	74		83	83	77	79	70	67	45	44	32	35
Phil Allt	46	54	72	83		88	85	78	71	69	49	47	39	35
Linda Busuttil	48	53	70	83	88		86	76	75	71	50	49	38	39
Katherine Hauser	73	59		77	85	86		95	91	95	50	57	41	64
Leanne Caron	56	62	85	79	78	76	95		85	83	65	61	47	51
Cathy Downer	72	70	82	70	71	75	91	85		91	70	72	55	65
Carly Klassen	69	68	82	67	69	71	95	83	91		74	76	52	63
Rodrigo Goller	51	62	62	45	49	50	50	65	70	74		66	46	59
Ken Yee Chew	75	82	80	44	47	49	57	61	72	76	66		80	86
Dan Gibson	66	72	54	32	39	38	41	47	55	52	46	80		79
Christine Billings	75	75	60	35	35	39	64	51	65	63	59	86	79