Tl;dr: Through spring 2021, the poorest areas in most states are 20% behind the least-poor areas in the vaccination rollout.
Sorry, updates to Mappage have broken the visualisations for this article.
The pandemic is one of those news topics with a lot of numbers that gives stats nerds like me things to follow and discuss. I begin by noting how lucky we have been so far in Australia and especially in South Australia with only 4 deaths in 20 months1One thousandth the per capita death rate of the US, so I can spend more time looking at vaccination statistics rather than deaths and hospital cases. Most of the article is heavy-duty stats with pretty graphs; jump to the last few paragraphs for my non-expert thoughts on why the poor are less vaxed.
I wrote this first in late October with a big revision in early December. Things have continued to change since then, with more children getting vaccinated, the statistics including kids and the advent of the omicron variant. I’ve added more data since, but this article will remain as discussing how the numbers stood through spring 2021 to 6th Dec.
This article is viewable both on this site and on Mappage, my webapp for looking at statistics. If you’re reading this on a large screen, it’s best to read this article in Mappage, where I have set it up with maps and charts. If you read this version, you will see references to the charts you can’t see here.
This article is viewable both on this page and in plain form on the Mappage WordPress site. The one-letter links here will show maps and charts. This is not so much an article with a Google Map embedded; it is my statistics & mapping web app (called Mappage) which I have set up to accommodate articles like this. Clicking those links just changes the settings on the right. You’re not restricted to looking at what the links provide; you can look around for yourself at any other stats.
The Department of Health releases data weekly with the proportion of people aged 15+ vaccinated in each LGA, SA32In Adelaide, SA3s often approximate Local Government Areas; in country SA, each SA3 takes in a few LGAs. and SA43The first level below state/territory, eg Northern Adelaide.. I’ll start with my main exhibit. O fnpox9lm The circles are the SA3s in South Australia. The y-axis is the proportion of people aged 15+ who’ve had their 2nd dose, with that number rising as the weeks pass. The x-axis is the Index of Disadvantage4A composite measure of disadvantage, from the 2016 census. 1000 is the national average.. The animation starts at 28 August and shows every week up to 22nd November.
Vaccination v Disadvantage in SA
The poorest areas have been well behind the least disadvantaged for the whole rollout. Playford, the poorest SA3 in the state, has been as much as 15% behind the average, while Burnside, the most well-off, has been 15% ahead of the average. I wasn’t surprised that poor areas were behind, but it was interesting that there was such a difference and the correlation for both 1st and 2nd dose levels with disadvantage was over 90% for many weeks, meaning that you could predict an area’s vax level within a few percent just from its Index of Disadvantage5I tried out other similar indices – Advantage & Disadvantage (very similar to Disadvantage) and Education & Occupation. Disadvantage had the greatest correlation, so I’ve used that here.. The slope of the line was greatest at 1st November, where a 100 point difference in the index made a difference of 11.7% of 2nd doses.
There are lots of settings you can change to explore the data. You can click “1st/2nd” and click “1st dose”. The 1st dose numbers are a month or more ahead of those for 2nd dose. You can click “Date” and click any date in the list, the chart will animate to represent the stats at that date (clicking the box doesn’t animate).
You can click “Residence” and click another state to see the SA3s in that state. You can click “Each circle is a…” and change it to LGA (most of the options in that list don’t work).
O 6lfw7eyr Looking at SA at LGA level, which the media uses more often than SA3 or SA4, the correlation is not so strong. In Adelaide the SA3s approximate the LGAs, so we would expect similar results. But outside Adelaide, LGAs that are socio-economically similar have different vax levels, perhaps due to age profile or varying availability in the early stages.
Elizabeth
Playford is the poorest metropolitan SA3 in Australia, scoring 853 on the Index of Disadvantage6The other poorest SA3s are mostly indigenous areas.
ABS recommend using SEIFA only for ranking areas and not relying on the value for anything, but given the strong linear relationship I’ve found, I think I’m OK to do so.. Within Playford is the Elizabeth area, consisting of the ten poorest suburbs in Adelaide7I’m counting all the suburbs there built by the Housing Trust around 1955-65, not just the former Elizabeth LGA. The 8 suburbs with Elizabeth in their name plus Davoren Park and Smithfield Plains., averaging 713. The rest of Playford averages around 950. According to our model I 4m5o6dpq, the Elizabeth area (my home when I originally wrote this) may be 10% less vaccinated than Playford as a whole, which may be concerning now that we’ve opened to the eastern states and the virus is circulating again.
Vaccination v Disadvantage in other states
Looking at the other states comparing Disadvantage and 2nd dose: In Western Australia O 46tdmj9p, the slope peaked at 16.3%/100 (70% corelation) at 18th Oct, with the wealthy Cottesloe-Claremont nearly 22% ahead of the average. In Queensland (click “Residence” and “Queensland” to change state), slope peaked at 8.9%/100 (64% correlation) at 25th Oct, with the well-off Gold Coast less vaccinated.
O ot7gup1i In NSW there was an 82% correlation and a 8.3%/100 slope between Disadvantage and 2nd dose in the first week of our data. Fairfield, the poorest SA3, was originally among the least-vaccinated, but had the worst of the case numbers in July, leading to the rollout being sped up there, so that they were up to the state average for 1st dose by 28th Aug and for 2nd dose by 27th Sep, even as the whole state sped ahead. Likewise in Melbourne had a wave of infections that hit the disadvantaged parts of Melbourne hardest, so vaccines were pushed hard in those places.
C 3lo5bi4n This chart shows how the states 2nd dose levels compared from 28th Aug to 6th Dec. Sydney surged ahead in September, having done so on 1st dose in August due to the high case numbers there in July-August. The rest of NSW and Victoria surged a month later in response to the situations there. Queensland, WA and SA had relatively few cases so did not show the same urgency.
O 8sn0v6lz This chart looks at all of Australia at SA4 level, coloured by state, with 2nd dose v Disadvatange. You can click “Each circle is a” and change it to SA3 to see more detail. The SA4s with the lowest levels are outback areas with indigenous populations and this has been an increasing concern. M usi0cv35 The same as a map.
A note about the percentages: They’re calculated from 2019 population estimates, which may no longer be accurate. For this reason any levels over 95% are listed as “>95%”, because where the number tops out may be as much about the inaccuracy of the population number used as the proportion remaining unvaccinated. In particular, the inner parts of Sydney and Melbourne likely have less population while (I think) parts of Tasmania have more. As a state gets higher and more of its areas hit that 95%, the calculated slopes will diminish. Playford has a growing population, so its true vaccinated % may be lower than quoted.
Vaccination v Age profile
Having found one variable that explains a lot of the variation, I now continue the analyst’s quest for a further variable that accounts for more of it. We might also expect areas with an older age profile to be more vaccinated, since we know older age groups have higher vaccination rates. In September, those aged 60+ were 25-30% ahead. O 8b054cxe This scatter plot shows vax levels in SA against the proportion of residents aged 60+, excluding those under 15. We might expect a slope of 0.3 in September. Remarkably, for SA3s in SA, 1st dose at 28th Aug the slope is 0.79, meaning that places with more older people are more vaccinated by a much higher amount than would be expected just by the vax levels of their age groups, ie in places with more 60+, both older and younger people are more vaccinated than their respective age groups elsewhere.
This slope gradually reduces to 0.23 at 22nd Nov. On 2nd dose, the slope was 0.48 at 28th Aug, 0.61 at 27th Sep and 0.37 at 22nd Nov. In WA, slope of 1st dose was 0.98 at 20th Sep, three times what we should expect. Queensland shows less slope. O 2jfgs7xq In Victoria, 1st dose 28th Aug, there is slope of 0.85 and 74% correlation, with the retiree-rich areas ahead of the well-off ones.
O bml5ez9t New South Wales is different on this score. The virus spread in the poorer parts of Sydney, which have fewer aged 60+, so the vaccine was prioritised in those places. On 1st dose at 6th Sep, the slope was -0.29.
O 5z9bv1su Australia at SA4 level. As before, you can change or SA3 or LGA level or 1st dose.
Vaccination v Disadvantage and Age profile
(very heavy duty stats; feel free to skip8and it’s my first time doing this with three variables, so I could be getting it wrong)
Having found two useful explanatory variables, I’ll have a go at combining them to see how much variation they account for in different states and points in time. As before I’ll describe the slope as %/100 for 2nd dose v Disadvantage and as a decimal comparing with the 60+ proportion. Spreadsheet with all the outputs of my analysis (O uofhz02c9Loads up all three variables in preparation to analyse. My discussion uses age 60+ ex <15; the link here users <40 ex <15; the spreadsheet includes both. Link is here so I don’t lose it.).
NSW went from 9%/100 at 6th Sep to below 0 by 22th Nov, following 5 weeks behind the 1st dose. The age slope started at 0.19, fell to -0.24 at 4th Oct and went back up to 0.33. In Greater Sydney10includes Central Coast the disadvantage slope is similar but the age slope went from 0.27 to 0.21 up to 0.65. Blacktown North was very high; was there an early cluster there, or is population much higher than 2019?
In Victoria, the disadvantage slope is 8.4%/100 at 4th Oct but then drops to -2.1%/100. The age slope is always higher (inflated by a population overestimate of youthful inner Melbourne). It peaks at 0.89 at 20th Sep and goes down to 0.43. Looking just at Melbourne, the disadvantage slope peaks at 10.2%/100 (despite the average by then of 81.3%) and drops to -2.9%/10011again artificially low because of inner Melbourne by 29th Nov.
Queensland: Disadvantage slope peaked at 10.9%/100 at 25th Oct, down to 7.8%/100 by 29th Nov. Age slope varies little and ends at 0.41.
SA: Disadvantage slope peaked at 11.2%/100 at 11th Oct. Age slope peaked at 0.38 at 27th Sep. In Adelaide, a bit less age slope and R2 was 0.90 at 22th Nov.12The high correlation with disadvantage in SA is how I came to write this. In analysing everything, I found that my state was the anomaly, but thought it was still worth writing.
WA: Disadvantage slope peaked at 15.4%/100 at 18th Oct, the highest for any state. Age slope peaked at 0.76, also at 18th Oct. In Perth disadvantage slope reached 18%/100. Perhaps the rich wanted to travel while the rest remained content behind a closed border. The northern outback remains way behind.
Tasmania: Disadvantage slope peaks at 13%/100 at 10th Oct, but disappearing as the overall total approaches 90%. Age slope is surprisingly negative, -0.53 at 15th Nov (and -0.50 for 1st dose at 25th Oct, so, as elsewhere, the longer gap between AstraZeneca doses that went mostly to the 60+ doesn’t make much difference in the calculated slopes for 1st v 2nd dose).
I could have used age data better: I could have tried breaking down using different or multiple age bands. I could have made use of the vax numbers available daily broken down by state, gender and 5-year age group. There’s plenty more one could analyse.
Why are the poor less vaccinated?
Having lived two years in the Elizabeth area (but not suffered the same disadvantage as many there), these are my thoughts on why poorer people have been less vaxed through spring 2021. Many poorer people are generally worse at decision-making (many people gamble13Some places have vaccine lotteries to appeal to those bad at probability-based decision-making.) and make poor health choices (more people smoke, more people have bad teeth). They have other things to worry about and don’t follow the news as much. Many don’t know or don’t care about rules (I see people riding bikes without the mandatory helmet; a friend came over in a lockdown week, apparently unaware that he wasn’t meant to). These traits may create higher risk of Covid spreading even apart from vaccine levels.
Poorer people trust “the system” less, having suffered under it. Educated people may know something of how medical research and governments work; they could see themselves in such jobs and assume most in them act in good faith. But it goes beyond the education of individuals. For many people in places like Elizabeth, they don’t have many educated friends and the whole medical sector, universities, the government and media is something “other” to them and therefore easier to distrust. The opposite to this is the ACT, where many people work for government departments. They trust the system because they know the system – they are the system – and their vax levels are among the highest in the world.
All of these traits can be found in other demographics. There is distrust of government, media and the medical sector among my educated friends and system insiders. There are people in all areas afraid of needles or side effects. But so far Playford is well behind the pack, Elizabeth probably much more so. The experience of Fairfield in Sydney suggests they may catch up quickly if the virus spreads, but there may be a difference between hurrying up the majority and convincing the last 20%.