Last 5 posts

Archive

12 May 2015

Pollsters and Forecasters




Spare a thought for the analysts and forecasters. I know I’m often critical, particularly when the modus operandi seems to involve changing the forecast every other week - let me change my forecast as often as I like and I’ll be the most accurate one out there, by the way - but they certainly came in for a drubbing in the wake of the UK general election. For those who weren’t following, there wasn’t a single pollster who predicted the result correctly; admittedly, with the numbers so close to the win/lose demarcation line, the stakes were high, but nevertheless it was a pretty lamentable exhibition. They had a tough task, though, perhaps tougher than the market analysts who are looking at prices.

A tough task

 Why do I say that? Well, the obvious thing that comes to mind is that they have no way of knowing if the answers they are getting are honest; and that’s quite apart from the initial problem of establishing what questions to ask in the first place. If you step back for a moment and look at what they are trying to do, you probably wouldn’t give them much chance of success. After all, they are trying to find a sample - something like eleven hundred out of a total electorate of maybe thirty million - and extrapolate from the answers given by that sample to a series of questions what the intentions of the whole electorate are. Put like that, it’s a wonder anybody takes up the challenge in the first place. 

(Incidentally, the evening before polling day, I was at a dinner where the head of one of the major polling companies gave us an idea of how their business works; it’s broadly attempting to apply a scientific methodology to something which is a subjective matter to each of the individuals sampled. If I had to do that, given the number of variables, I would end with a scrambled brain; so although I may question the success rate, I’m very impressed with the techniques. That’s probably the wrong way round, though.)

Market analysts

How does that task compare with the one undertaken by market analysts? It’s a fair question, because both are trying to predict a picture of the future based on the information available to them in the present. I do understand the ‘snapshot not prediction’ argument, but I’m afraid it’s not really that useful; it may be the purist view, but in reality, polls and forecasts are used in the same way by those who read them - to gain an insight into the future. 

My instinctive reaction is to say that the pollsters have the more difficult task, since they have a far wider pool to try and model. After all, market price forecasters are really just focussing on one thing - the balance between the supply and demand of the commodity in question. That should mean a simpler operation; check out mine supply, have a look at growth in the major consuming economies, compare the expected rates of change, and draw your conclusions. Who could get that wrong? 

Multiple variables

On reflection, though, I think that simplistic view is wrong. Once you start looking at the detail, problems immediately raise their head. Even the mining companies don’t really know what they will produce - ore grades are uncertain until you actually get the rock out of the ground, natural or man-made events (like rockfalls or strikes) can throw the best estimate wildly out, processing facilities can break down. There are many possibilities to throw expectations off course. And on the demand side, it’s probably even more difficult. The days when demand was the amount consumers bought have largely gone - financial demand is now an integral part of the equation, and that is subject to an enormous number of extraneous forces. In the same chaotic way that the butterfly fluttering in Brazil can set off a tornado in Texas, so a loss of confidence in equity markets can launch a tsunami of hot money at the copper market. Technological innovation also creates an uncertainty on the demand side; look at the yo-yoing of whichever minor metals are currently in vogue for batteries, for example. 

Make order out of chaos...

So we have two sets of people doing a similar sort of job; trying to tell us what the indicators they can grasp are suggesting the future may look like. We all rely on them, to one extent or another, we (all?) criticise them frequently for being wrong; probably not really fair. It’s a difficult task, trying to bring order to chaos.









Enter text here]

comments powered by Disqus