Interpreting Census data at the local level
Seven months out from first the release of 2021 Census data, we’re already working to ensure insights from this key demographic dataset are available to our partners in local government and related organisations as early as possible. In this blog Daniel shares how we’re working with our partners to review how they define geographic areas within their LGAs and how this informs our Census preparation work.
The Census is Australia’s largest peacetime logistical exercise, producing our most robust and meaningful set of demographic data. One of it’s key strengths is the fact that it’s a national set of data available down to very small geographies. This allows for a great understanding of local demographic drivers and how these compare to other areas, be it neighouring suburbs right up to the state and national benchmarks.
Census data is provided at pre-defined geographies, from the national set right down to very small local areas. One of the benefits of our demographic information tool, profile.id, is that our local government partners are able to work with us to customise the small area geographies used on these sites to suit their own needs. We’re working with subscribing councils to review and, if necessary, update their small areas. This is a key phase in our plan to deliver Census data and insights as quickly as possible when it released in June next year.
At what geographic areas is Census data available?
To give a bit of background on this topic, we need to start with the Australian Statistical Geography Standard (ASGS). This standard classifies Australia into a hierarchy of statistical areas. It is a “social geography” designed specifically for the publication and analysis of statistical data. The ABS divides the ASGS into “ABS structures” and “non-ABS structures”.
Here’s a quick lesson useful to anyone working with ABS data.
ABS structures are a set of nested geographies. The majority of them are called Statistical Areas. Statistical Areas Level 1 (SA1s) are “designed to maximise the geographic detail available for Census of Population and Housing data while maintaining confidentiality”. Most SA1s have a population somewhere from 200 to 800 people. There are four levels of Statistical Areas. Each level consists of progressively larger geographic shapes designed with a particular purpose. SA4s are the largest. A typical SA4 has population greater than 100,000 and represent broad regions within a city or state.
Above SA4s in the ABS structure are States and Territories, and Australia. Beneath Statistical Areas are Mesh Blocks, the smallest geographic areas defined by the ABS. These are the building blocks of all other area definitions in the ASGS. According to the ABS, most Mesh Blocks contain 30 to 60 dwellings; they often literally represent a “block” in urban areas.
Other ABS structures include Indigenous Structures, Urban Centres and Localities, Significant Urban Areas and Greater Capital City Statistical Area Stucture (GCCSAs). These structures are all built from Mesh Blocks and have a specific purpose they are designed to measure (eg. the urban extent of a town).
Non-ABS structures include geographies such as Local Government Areas (LGAs), State and Commonwealth Electoral Divisions, Suburbs and Localities and even Drainage Divisions. These boundaries are maintained by other levels of government, not by the ABS directly. The ABS matches them as closely as possible using Mesh Blocks as a basis.
Changes to ASGS geographies
ASGS structures – particularly Statistical Areas – are reviewed regularly by the ABS to ensure they continue to meet their intended purpose. When these geographies are updated they can lose comparability over time. Statistical Areas are mostly divided into two or more areas to account for population growth and maintain the population size ranges of the areas. This means they can be recombined to form the old outer boundary. (This is simple enough to do, but still needs to be done!) In some cases it is necessary for the ABS to adjust the boundary itself, making the areas truly non-comparable. At .id, we have established methodologies to maintains comparability over time; I’ll get to that shortly.
Can I get Census data for a custom geography?
A lot of thought goes into ASGS geographies to ensure they’re as useful and relevant as possible to a wide range of consumers. In some cases, though, local governments need Census data for geographic areas that don’t align to the ASGS. This is where .id is able to help.
When a local government subscribes to profile.id, we spend time with them working out the best way to divide up the LGA. The goal is to determine the best areas that meet the statistical requirements and represent logical communities of interest that the council works with in the real world. The areas need match how a Council thinks about, plans for and supports different areas within the LGA.
Sometimes we land on areas that line up perfectly with ASGS definitions, sometimes we create custom geographies based on the Council’s needs. Often it’s a mix of the two.
We use a range of smaller areas and adjustments of datasets over past Census years to match any custom geographic area with a minimum of 1,000 people. This methodology can be used to create the small areas that make up the LGA, as well as what we call “overlays” areas, based on a particular focus such as population growth corridor or an activity zone.
The challenge of confidentiality
From the explanation of ASGS above, you might be thinking, “Surely I can make up any geography I want from Meshblocks”. (If you are thinking that, thanks for playing along!) There are a couple of issues with this. One is how to handle the case where the geography you want to work with doesn’t line up with Meshblocks. The other is the way the ABS deals with confidentiality.
Protecting individuals’ privacy is of utmost concern to the ABS. SA1s are the smallest geographic detail available that maintains confidentiality. What does that mean for MeshBlocks, which are smaller than SA1s? Around the 2016 Census we wrote about how the ABS processes data to avoid being able to detect an individuals’ data – and the challenges this process can create. In short, small random adjustments (“pertubation”) are made to data to ensure no individual’s characteristics can be determined. This means if you were to look at data for an SA1 and compare it to the total of the same data of all the MeshBlocks that make up that SA1, you would always get slightly different figures. (Also, you will never see a 1 or 2 in the Census data.)
Ensuring pertubation doesn’t render data meaningless or misleading is core to .id’s methodology for working with small area geographies.
How can you compare Census data over time when geographies change?
Understanding change over time is a core facet of how we convert data into knowledge. profile.id shows data from previous Censuses (right back to 1991 in some cases) to demonstrate how an area’s demographic story is changing over time.
Even where a local government uses geographies based solely on ASGS, .id’s methodologies are essential for building this comparability over time. This is because ASGS geographies can and do change. We adjust past Census datasets to match current geographic areas to create a consistent time series. An area might have the same name but have had a significantly different geographic definition in 2011, 2016 and 2021. Comparing this area without making the relevant adjustments can lead to incorrect and misleading data. Enforcing a minimum population size for small areas (~1,000 people) ensures this process is robust.
All our datasets are set up to show change over time as well as geographic benchmarks. Understanding how an area is changing and how it compares to other areas are the most important concepts in telling the demographic story of an area.
How does updating geographies help with delivering Census data?
For the past month or two we’ve been reaching out to our existing subscribers and asking them to review the small areas used on their community profile site. This is an important exercise to undertake semi-regularly – the way councils think about an area can change over time – but the upcoming Census data release is the main driver for us starting these conversations now.
Delivering Census insights as efficiently as possible once the data is released is our “true north”. Taking the opportunity to review and refresh these geographic definitions now puts us in the best possible position to do so.
There are a couple of ways changing geographies can impact Census data rollout.
- How an LGA defines its small areas can affect how we pre-order Census data for them. Changing small areas during or just after the data is released means we might need to re-extract the data, delaying how quickly we can get it up on the site.
- Where boundaries align to suburbs, wards etc. it’s important to ensure these are up-to-date and reflect the latest boundaries (which do change) – and do the work to ensure they are comporable over time.
- More importantly, ensuring small areas on profile.id best match Council’s needs now mean that when the new Census data is updated, it will be relevant to Councils needs immediately.
We’ve reached out to all 220+ Councils who subscribe to a small-area profile.id site. Many have confirmed they’re happy with the existing geographies, while other are working with us to update the geographic definitions now.
If you work at a Council with profile.id and want to discuss reviewing your small areas, contact us at firstname.lastname@example.org.
Are you prepared for the Census data roll out?
This year’s Census will form an invaluable evidence base for understanding the impacts of the COVID-19 pandemic and planning for local recovery. If you work at a Council or related organisation, and have questions aobut making the most of the new data when it comes our, we’d love to hear from you.