Is the Census really useful?

Sorry Sir, we understand that you're about to die from a heart attack, but the Census 5 years ago estimated that there would only be 436 deadly heart attacks in this hospital and you, as it just so happens are number 437. So the security guard here is going to escort you out of the hospital so you can prepare for your demise.

Sound ridiculous and Monty-Pythonesque? To understand the complexity and real world nature of service delivery is to understand why the Census is nothing more than a ritualistic waste of money.

The Census' own website claims "The information you provide in the Census helps estimate Australia's population, which is used to distribute government funds and plan services for your community – housing, transport, education, industry, hospitals and the environment"

Firstly, as the ABS concedes, Census data is always a proxy to estimate a FUTURE demand for a service. So we already have a loss of fidelity in using the Census data for service provision. And we should remember that the actual name of the Census is the Census of Population and Housing. No matter how many supplemental questions of income, religious beliefs etc are asked, the information could never be enough to predict demand for services; a demand which is influenced by a prohibitively complex array of biological, psychological and social factors that could never be predicted with the Census data.

Therefore, while the example above is obviously tongue-in-cheek, the principle is the same for any service. A five year estimation of the likelihood of needing a service is never going to trump the actual need for that service as people present with their problems.

Let's take health as an example. The likelihood of having a disease is determined by an impossibly complex interaction between genetic, environmental and behavioural phenomenon. Academics make their careers out of trying to predict disease incidence based on risk factors. And of course, we never know for sure and have to deal with all the patients that come through the door. It is preposterous to believe that the information contained in the Census is enough to predict disease incidence into the future. If you were wondering why the Australian Bureau of Statistics is sending a syringe, blood testing kit and a urine specimen jar, now you know; it's so they can tell us where to build new hospitals.

So if not for Census data, how do we know where to build hospitals? What happens in the real world is that a health service constantly monitors demand for service, I.e. how many patients and with what diseases are actually coming to the hospital. Changes in population and disease prevalence will see the data move. Of course, if there is a whopping big new housing estate in the area, this change in population will be factored into future planning, with the resultant estimates of service demand being continually updated in real time. More importantly, unlike the case for the poor bugger mentioned at the start of the article, hospitals have to have provisions to treat more or less patients than predicted to meet actual demand.

So, as a general rule, the estimates of service demand based on any historical Census date will never, ever have better fidelity than that which is generated by constantly changing signals on the ground.

Believing that grand data collection, deciphered and analyzed by experts, can better organize the allocation of goods and services than the almost infinite market signals happening on the ground is reminiscent of the debate last century between command economies and market based economies; and we know how that one turned out. Certainly we are seeing a change in medicine with regard to how we interpret data. For even the most simple drug A versus drug B comparisons, the profession very often gets it wrong or the data quickly becomes obsolete in a changing environment. In response to this understanding of the perils of data interpretation, ever more stringent systems have been put in place. And new trial designs have arisen such as adaptive clinical trials, platform trials and the like to deal with data interpretation in dynamic complex systems. In this regard, the prediction of service demand based on a Census of Population and Housing is so 1920's.

Unfortunately our politicians never demonstrate the courage to stand up and challenge the uber-simplistic, motherhood statements that are used to justify the waste of money which is the Census.