Friday, June 15, 2007

Learning from the past

Why are economists so bad at forecasting?

As I mentioned before, economists need to make assumptions about the future in order to forecast.

Unfortunately, these assumptions tend to be wrong, as do their forecasts, which is why forecasts tend to be revised on a regular basis.

But I've come to realise that bad forecasting is also the result of looking at what happened in the past and failing to learn from it in a way that will help and not hinder predicting what could happen in the future.

While we are used to hearing that "the past is no indication of future performance", it is still possible to glean from past data critical information that will help in making a forecast more accurate and less prone to revisions.

Economists fail to take away the right information from past observations and keep making the same mistakes over and over again.


So, how can the past help us to predict the future? It depends on how you look at what happened.

Imagine you travelled down a road at 8am and it becomes very heavily congested. Based on what you have observed, you might also expect that if you travel down the same road tomorrow at 11am then it will be congested again.

Of course, this fails to take account of the fact that before 9am the road is unusually congested, since it's in the middle of the rush hour.

Economists make a similar mistake when using past observations to help forecast the future path of the economy.

They tend to look back at what actually happened rather than at what caused it to happen. In other words, economists look back at backward-looking data to help predict the future!


In trying to predict a recession it would be better to look at what leading indicators were doing in the run up to previous recessions and compare this to what they are doing now.

Instead, economists look at what happened to GDP, unemployment and inflation, which give no indication about the future, to predict what will happen to GDP, unemployment and inflation!

Part of the reason why economists don't use leading indicators is because they don't trust their accuracy. Currently, the OECD and Conference Board produce the most widely-followed leading indexes but these have had a poor record at predicting recessions.

Another reason why economists are sceptical about leading indexes is because they don't fully understand how to accurately construct one of their own.

A lot of work has been done in this area by my former employers at ECRI and you can follow some of their leading indexes in the press. However, since they are a private firm only a small amount of information is made public. This is unfortunate, since they have a pretty decent record at predicting turning points and in particular, the timing of recessions.


The fact that economists make bad forecasts is not necessarily a problem (they can be useful in other ways too!). However, financial markets look to economists, either in the private sector, government or at central banks to guide expecations about the future. If these expecations prove to be inaccurate or plain wrong then stock and bond markets will be unecessarily volatile. In reality, this is what we observe.

Trying to understand what caused something to happen in the past can help predict what is likely to happen in the future.

It would be like saying that a road (economy) that becomes heavily congested (has very high gasoline prices) at 11am (when leading indicators are stronger) will be better able to deal with that congestion (avoid a recession) than a road (economy) that becomes heavily congested (has high gasoline prices) at 8am (when leading indicators are weak).

Wouldn't it be better if we could assess the future of the economy with as much clarity as congestion on the roads?

No comments: