- For evaluators’ eyes only (21/07/2018)
- Reconciliation Week and findings from an Aboriginal health evaluation (04/06/2016)
- Evaluation amidst complexity: 8 questions evaluators should ask (04/12/2015)
- To count or not to count: Australian population data (20/02/2015)
- My pick of readings on scaling up health interventions amid complexity (12/12/2014)
To count or not to count: Australian population data
Friday, 20th February
Photo: Australian Bureau of Statistics
I am a card-carrying demographer—a member of the Australian, Asian, American and international population associations and a PhD in the subject from the Australian National University. I use population data and demographic concepts in my research and consultancies and to think about social issues. But it has been a long time since I made my living as a demographer and no one in the ABS would be interested in hiring me for my technical skills.
So, I can approach the current debate about the value of Australia’s census from the perspective of an interested and reasonably informed consumer of population data and not as a demographic whiz who depends on the census for my livelihood. I am using that perspective to think about two related questions. Should the census be taken every five years or every ten (I will not even consider the question of whether Australia should stop conducting censuses)? and Should there be a regular program of large population surveys either nested within the census or taken between censuses?
The British settler societies of Australia, New Zealand, Canada and the United States are unusual in many ways but demographically two things are striking and that affects how often you need to conduct population censuses. The first is that by reason of geography and politics they have fairly high overseas migration levels which fluctuate from year to year as a result of the global and regional economy and government policies. Second, because of preferences and the housing market, people in these countries move a lot and some people even have more than one place of residence. In Australia’s 2011 census 39% of people lived at another address five years earlier. That is a lot of mobility.
All of this means that even though populations do not change very quickly, international and domestic migration and the changes in age structure they bring means that five years is a long time for some communities.
Or at least there is the perception of a lot of change. Boosterism is a problem. I recall one local leader being so sure that our town would grow and that businesses would flock here that he said that the main impediment to development was the state’s Chief Demographer because official projections forecasted only modest growth.
The census is a reality check to city suburbs, regional cities and small towns about how their communities are really tracking. Did that big construction project actually bring in more people? Does population growth explain the shortages of beds at the local hospital?
The only way to estimate population size between censuses is to project on past trends, informed guesses (often from local boosters) of triggers for change and supplemented by some information from changes of address in Medicare, electoral rolls and Australia Post. All of these are crude measures when there is rapid change. How crude is an empirical question. How wrong were estimates of the 2011 population based on the 2006 census? What would have been the estimates for 2016 population if we didn’t have the 2011 census to recalibrate the projections? Are those errors ones that policy makers, businesses and service providers can live with or would they have resulted in too many misdirected efforts? I don’t know the answer but someone should do the calculations. The results will tell us which localities or subgroups will be disadvantaged by a move to a decennial census.
Although censuses are essential for describing the population, they are blunt instruments for describing how society is functioning. That is the job of surveys. As a user of census data I have been spoiled by my use of survey data. In survey data you know a great deal about your respondents. A survey I did on Geraldton residents’ use of the internet gave me (and the City of Greater Geraldton, I hope) rich insights into how people connect and what they did once they were there. I learned about the subtle differences by age, presence of children in the household and the use of the internet at work. But I could not have convinced anyone to believe my findings if I didn’t have a census to tell me the age, sex, geographic distribution and other aspects of my town so I could show my respondents were representative.
My efforts to get meaningful data on social issues from the census have not been satisfying. The question on internet access was woefully out-of-date before the results were published, because of rapid changes in how people connected. It is almost impossible to make sense of what types of post school education people have had based on the census and the broad income categories makes it impossible to discern economic benefits from employment, education or housing.
In short, the current mode of collecting census data by asking everyone a very large number of questions that are still not detailed enough is not producing the richness we need. A smaller core set of questions with surveys of a sample of the population would be more informative, even for small areas.
A census (and the insightful population surveys that can be done with an accurate census) might not be necessary for some northern European countries whose staid populations can be tracked with a population register. Other countries that haven’t had regular censuses include Lebanon, Nigeria, Afghanistan and Myanmar. Is that the company Australia wants to keep?