If this blog were the BBC, in an effort for impartiality, I would give equal time to 2 ideas:
- Climate change is completely true
Agreed by all reputable scientists and means drastic action needs to be taken immediately
- Climate change is a hoax
Devised by the Chinese to limit the US economy and we need do nothing
However, I’m not bound by any such desire for manufactured impartiality so would like to ignore climate denial theories. I want effective action to be taken and I’m interested in why the first argument is failing to find broad enough support against those who want to dismantle the Paris Agreement.
I think I can reasonably simplify the scientific arguments to this:
- There is an empirical relationship between carbon and temperature
- There is a causal theory relating the two, the greenhouse effect
- There is no better theory, in fact no other credible theory
The criticisms for each item can be summarised like this:
- Poor and incomplete data
- The underlying system is dynamic and complex
- The models are not very precise with large amounts of uncertainty.
The responses I have seen to these criticism from climate change scientists are:
- We are collecting more data all the time. We are cleaning up the old data sets.
- We are building more complex models which incorporate more variables
- We are working to improve our accuracy with better models to reduce the level of uncertainty.
Coming from a macroeconomics background, the criticisms wouldn’t bother me much. We face them all the time. The responses from the climate change scientists do bother me however. If this reflects their research programmes, I fear they are heading in the wrong direction, sharing many of the methodological problems we see in macro and likely making the same mistakes.
My preferred responses to the criticism would be
- Yes, the data is poor and incomplete. Not much we can do about that. We will not clean up data to pretend it is better than it is.
- Yes, it is dynamic and complex. Attempting to making models more complex would mean data-mining the limited data sets to produce models which are pretty but have no validity. Simple models of complex systems often work much better. Read the Borges short story (https://appliedmacro.com/2017/05/11/models/) for an elegant refutation of the idea that seeking perfection in models leads to good outcomes.
- Accurate forecasts are not possible because we do not have ‘out of sample’ data and associated feedback. There is little new data to use so we cannot properly test hypotheses in the way we can with weather forecasting.
The good news is that we do not need accurate forecasts because:
- If our model provides an unbiased best estimate and uncertainty is two-sided, then the level of uncertainly around it does not affect the policy response. Essentially the policy response will be versus the expected increase in temperature.
- We are providing conditional, not unconditional forecasts.
To take an analogy, I am thinking of running the London marathon next year. Please estimate how long it will take me to run it i) in running kit ii) wearing a gorilla costume. I would strongly expect that your confidence in both of your answers is very low. However, I bet you are very confident that ii) will take longer than i).
Financial markets can be a very good training ground for learning about practical model building and their methodical dangers. Most of us have been seduced by beautiful models with great back tests with desirable correlations and high Sharpe ratios. But then trying to use them to make money, we find they had no predictive qualities at all. After enough painful experiences, we learn to be highly suspicious of any model that fits the data too well – it is the obvious symptom of data-mining. The models that work in practice are the ones that are intuitive, simple and accept that the world is a messy complicated place.
What climate change scientists and macro-economists can actually do have a lot of similarities. We do not know what the temperature will be on October 23rd 2087 but we have a good guess it will be higher if there is more carbon in the atmosphere.