Some reflections on trying to find out information about local government elections.
Despite the timing, this post is not about the recent sorry excuse for democracy in North America. I’ve been looking into the results of the recent local body elections in New Zealand that used Single Transferable Vote (STV), and have some reflections on the transparency of results.
First Past the Post (FPP) elections are pretty easy to report on — everybody makes a certain number of votes (depending on the number of seats available) and you can just list the number of votes each candidate received. The official results are brief and complete, and it’s completely clear to members of the public who got elected and why.
STV elections, as I’ve recently discovered, are not so straightforward. The councils and DHBs appear to be mandated to report the iteration number and count of votes when each candidate was excluded or elected1, but that doesn’t provide any explanation or insight into what votes/preferences were cast, and how the vote counting process proceeded.
Understanding how the election results were determined is important for giving the public a sense of ownership and engagement with the voting process, which in turn seems like an important issue given the low voter turnout in local body elections. And the transparency this brings is important for making it clear that the elections are free and fair.
STV is arguably a fairer system, but it won’t keep and gain support if voters don’t understand it. In my view, the only way to do this is to provide the votes for every candidate at every iteration of the counting process, thus making it clear how and why each candidate was elected or excluded. The STV iteration report provides this.
Six of the eight councils proactively provided STV iteration reports, as did two of the 19 DHBs:
Only one of these was completely machine readable, in that I could input it directly into R. With the RTF and a few of the PDFs I could copy the data directly into a TSV, but the others would have required manually retyping if I hadn’t recently discovered tabulizer.
I wanted to look into the rest of the elections too, so I sent off 17 requests2 under the Official Information Act (OIA) 1982 or Local Government Official Information and Meetings Act (LGOIMA) 1987 through FYI.
The responses were … variable.
The Acts say that a request must be responded to “as soon as reasonably practicable” with a hard deadline of 20 working days. A request like this, which is for an existing file that should be held by the electoral officer, could be completed the day it’s received (thanks Waikato DHB!), but it seems reasonable to allow up to a week.
Ten of the local bodies gave me the file within a week, but six of them appeared to take the 20 day deadline as more of a target. Hawke’s Bay DHB on the other hand didn’t seem at all bothered by their legal responsibility, taking four days after the deadline to respond.
The following graph shows the time taken by each local body to respond, and each of the main communications back and forth. The icons give more detail when the mouse hovers, but if you’re on a mobile device you should be able to tap to display.
Note that I have counted the day of request as Day 1, for nicer graphing. To be fairer, this makes the legal deadline Day 21.
There are various things to comment on. First and foremost, none of this would be necessary if the STV iteration reports were all proactively published with the election results. Hopefully that will be the case universally in 2019.
One notable variation between the agencies is whether an initial acknowledgement is sent. Given that BOP DHB “lost” my original email I think it would be good practice to do this for every request so that there is no uncertainty. It’s frustrating to wait with no response for two weeks not knowing whether it’s legitimately in a queue or has just been forgotten about.
I was surprised that there were a few agencies that sent me or directed me to something other than the full iteration report. I realise that it’s a bit technical and confusing if you don’t know anything about it, but I was specific about the information I wanted and I would have thought it wise to check the wording of the request before sending a response.
Somewhat related, many of the elections were contracted to a third party election company. In some of these cases, the report was sourced from them and passed on to me, in others my request was forwarded and they responded directly to me. In a couple of cases there was an attempt to say “we don’t have the file — you’ll have to contact [the election company] yourself”. This demonstrates a lack of understanding of the OIA/LGOIMA: this is still deemed to be held by the council/board,3 so shouldn’t be my responsibility to chase up. And either way the Acts’ duty of “reasonable assistance” requires a better response than ¯\_(ツ)_/¯. The Ombudsman has some useful guides.
Frustrating as it was to wait days and weeks for many responses, it was another level again to receive a PDF of images indicating that a machine-readable file had been printed and then scanned. With this, anything non-trivial on my part would require manually retyping pages and pages of names and numbers (hopefully mistake-free). Thankfully, I did manage to get the “original” file in each case after explaining the problem, but there didn’t seem to be an understanding of how and why data might be used, as opposed to text. More broadly, the variety of file formats were often less than ideal for extracting data from, but “they were all that we were given”. In hindsight I think I could have pushed for XML files from everyone (as above), but I managed to make them work. There are actually very good government guidelines from NZGOAL about file formats for releasing data; it would be great if these were more widely known.
Having collected all of the STV reports, I’m planning to poke around and do some more analysis and visualisation. (Especially if my kids will stay out of hospital.) In the meantime, I’ve made the files, original and processed, available in a Github repository for anyone who’s interested: http://github.com/dakvid/stvnz
I was keen to try and get and get hold of the more detailed data of all preferences, to be able to look at other counting methods, such as Condorcet, but it seems that that information is not stored for long. But there’s enough to work with for now.
The CSV file of event data.
The R code to generate the plot:
# This graph is constructed in parts:
# - a bar graph of total durations
# - a scatter plot of events, with hover info
# - line and text annotations
# - image annotations to sit over the scatter dots.
library(magrittr)
library(readr)
library(dplyr)
library(purrr)
library(forcats)
library(plotly)
oia <- read_csv("data/stv_oia_events.csv", quote = "|")
oia_days <-
oia %>%
group_by(body) %>%
summarise(duration = max(day)) %>%
ungroup()
oia_days %<>% mutate(body = fct_reorder(body, duration))
oia %<>% mutate(body = factor(body, levels = levels(oia_days$body)))
num_bodies <- oia_days %>% nrow()
make_image_list <- function(my_event) {
oia %>%
filter(event == my_event) %>%
transpose() %>%
map(~ list(source = paste0("images/",my_event,".png"),
xref = "x", yref = "y",
x = .x$day %>% subtract(0.5) %>% add(.x$adjust),
y = num_bodies - (.x$body %>% as.integer() %>% subtract(0.4)),
sizex = 1, sizey = 1,
opacity = 1,
layer = "above"))
}
plot_ly() %>%
add_bars(data = oia_days,
y = ~body %>% fct_rev(),
x = ~duration,
orientation = "h",
showlegend = FALSE,
name = "Duration",
hoverinfo = "none") %>%
add_trace(data = oia,
y = ~body,
x = ~(day + adjust),
type = "scatter",
name = "Events",
text = ~paste0("<b>Day ", day, ": ",status,"</b><br> ", comment),
hoverinfo = "text",
showlegend = FALSE) %>%
add_annotations(x = 8,
y = num_bodies - 1,
text = "reasonable<br>deadline?",
showarrow = FALSE,
font = list(color = "black", align = "left")) %>%
add_annotations(x = 23,
y = 1.5,
text = "LEGAL<br>DEADLINE",
showarrow = FALSE,
font = list(color = "red")) %>%
layout(title = "Response Times to Official Information Requests for Election Reports",
xaxis = list(title = "Working Days"),
yaxis = list(title = ""),
margin = list(l = 125),
shapes = list(
list(type = "line",
line = list(color = "black"),
y0 = -0.5, y1 = num_bodies, x0 = 6.5, x1 = 6.5),
list(type = "Line",
line = list(color = "red"),
y0 = -0.5, y1 = num_bodies,
x0 = 21.5, x1 = 21.5)
),
images = oia %>% use_series(event) %>% unique() %>% map(make_image_list) %>% flatten()
)I haven’t come across the exact requirements, but this was the minimum that was reported for each election. See eg the Electionz results.↩
If that number doesn’t seem to add up, it’s because I asked Auckland Council about 3 DHBs, and asked Southern DHB, not realising they didn’t have an election this year. I also asked Whanganui District Council about Whanganui DHB, as they appeared to have run the election for them.↩
Any information held by an independent contractor engaged by a body in his or her capacity as such contractor shall, for the purposes of this Act, be deemed to be held by the body. (OIA) (LGOIMA)↩
If you see mistakes or want to suggest changes, please create an issue on the source repository.
Text and figures are licensed under Creative Commons Attribution CC BY 4.0. Source code is available at https://github.com/dakvid/dakvid.github.io, unless otherwise noted. The figures that have been reused from other sources don't fall under this license and can be recognized by a note in their caption: "Figure from ...".