The CSIRO's lesson? Economists and scientists must learn from each other

A CSIRO team has spent ten years, it seems, building a model of the "physical economy" to consider the environmental impact of alternative population scenarios. In particular it considers the impact on the availability of natural resources like water, fossil fuels, forests and fish stocks, on biodiversity, on demand for food and housing, and on a host of other "physical" inputs in economic activity. It does this by trying to "keep track of all physically significant stocks" like "people, livestock, trees, buildings, vehicles, capital machinery, infrastructure, land, air, water, energy and mineral resources". The main constraint in solving the model is that the economy must "obey the known physical laws". The authors conclude, for example, that the Australian economy could sustain a population of 50 million by 2100, although it doesn’t nominate an optimal population target.

Is this a useful exercise? Consider this analogy. The CSIRO model treats the economy like a pinball machine. The only role for human decision-making is to put the ball in play. Once in play the ball bounces around according to automatic, preset mechanisms. If the ball threatens to disappear down the centre chute – in the language of the CSIRO model, the economy suffers "tensions" - the controller (policy-maker) can flip it back into play on a different trajectory and watch it bounce around again.

Economists also use simulation models to consider the impact of future demographic scenarios. But their models are more like a soccer game. The referee (policy-maker) puts the ball into play but, once in play, it doesn’t bounce around according to the predetermined forces of the physical world. Rather, players (consumers and firms) interact, make plans according to their goals and targets, and adjust their plans in response to the actions of the other players. The path of the ball is not preset – it is continually altered by purposeful human interaction.

Regrettably, the CSIRO model completely misses this fundamental idea. Consider some examples of how their model can get the effects of demographic change seriously wrong.

1. We know that as the population gets older and the workforce shrinks as a share of the population, the ratio of physical capital goods to labour will rise and hence the productivity of additional capital goods will fall. This will make additional investment less profitable and so firms will invest less. As a result the call on the nation’s resources will be commensurately lower. The CSIRO model cannot capture this because it ignores the basic idea that firms respond to changes in profitability by making adjustments to their inputs and outputs.
2. The above is an example of a more general problem in that the model assumes no changes in relative prices as a result of demographic change. The failure to allow relative prices to change means that the model ignores the fundamental principle that people respond to incentives to look for cheaper ways of doing things when prices go up. We know that when oil prices went up in the 1970s people found that natural gas was a substitute in some uses, and they started to drive smaller cars. We know that in places where land prices are high, people economise by living in smaller houses. Similarly, if in the future the price of water were to rise, people would economise on water use. If medical procedures become more expensive, people will tend to look for preventative treatments and perhaps alter their lifestyle. And so on. None of this is acknowledged in the CSIRO model.
3. The CSIRO’s model of international trade assumes that exports are the "quid pro quo of imports". The documentation is a bit vague about what this means, but the suggestion is that somehow exports of physical goods must be found to more or less match imports of physical goods. Assumptions are then made about the value of exports and imports of services and the interest rate in order to calculate a current account balance (CAB). Again the documentation is short on detail here. But on the face of it, their method of determining the CAB appears to bear no relation to the way the CAB is actually determined. For a start, exports are not the "quid pro quo of imports". That is simply wrong. We do not need exports to match our imports. Imports can be greater than exports if foreigners are willing to make a net investment in Australia. The CAB is in fact equal to the difference between domestic saving and domestic investment, which are in turn determined by forward-looking behaviour of households and firms. These decisions are in turn based on the weighing up of costs and future benefits. The consequences of ignoring all this is that the model could get the future size of the CAB seriously wrong which in turn means that Australia’s international debt will be wrong. That can be important because there are constraints on the feasible value of Australia’s future international debt. In a properly specified economic model these constraints cannot be violated.

Essentially, in trying to keep track of all physical stocks the model ignores the basic truth that human decisions move these stocks around according to economic forces. It is not good enough to treat the "controller" as a pinball player who puts the ball into play and watches it bounce around according to physical laws. It must bounce around according to economic laws that are based on what we know about human behaviour. Economic models try to do this.

Notwithstanding all of the above, there is value in the CSIRO’s exercise. Economic models could certainly be improved by accounting for the demands on the physical environment and ensuring that the allocation of resources does not violate known physical laws. A weakness of existing economic models is that they tend to ignore these physical constraints.

Most economists would probably have no problem with the CSIRO’s conclusion that Australia can support a population of 50 million by 2100 – but the economy is likely to look a lot different to that projected by the CSIRO model. Some other conclusions are very difficult to sustain given the deficiencies in the modelling approach. For example, the conclusion that the low-population scenario will result in a higher physical trade balance cannot be relied upon given the flawed model of the international trade. The conclusion that without substantial structural change it could be hard to maintain economic growth also suffers from the model’s weak economic foundations. Economic growth is driven primarily by productivity growth. The effect of population on productivity growth is in fact an issue over which there is considerable economic debate. Suffice to say that there are plausible arguments that the effect could go either way. That is, a lower population could either boost or lower productivity growth. We don’t know. A reasonable position to take would be to assume a zero net effect, in which case economic growth would be maintained just as well under a low population as under a high one.

It is a pity that the CSIRO team did not seek input from economists earlier on in this project. Apparently there were some economists on the "reference group", and they were not impressed with the violation of economic laws described above. It would have been better if they were brought into the team at the model design stage. Perhaps a positive to come out of this exercise will be a much closer collaboration between scientists and economists who are both interested in modelling the future impact of demographic change on people’s lives.

The report can be downloaded here (pdf, 343Kb).