The Olympics finally finished, and the Paralympics are about to begin. I managed to go to some of the Olympic football matches in both Lyon (in the photo above) and Nice. The vibe in general was very good, and you could see there was a lot excitement about the Olympics in both cities. Thinking back to the Olympics in London, there’s always a bit of scepticism in the run up, but as soon as the event starts, a sudden wave of enthusiasm appears, and I suspect it was similar in the case of Paris 2024.
There was some poll going around on Twitter (how scientific it was, I have no idea), which seemed to suggest that most folks felt they had a good chanc e of being to compete at the next Olympics. I suspect that the only thing this proves is that the set of folks who think they will be able to compete at the Olympics and the set of those who misjudge probabilities overlaps substantially. Joking aside, to compete and in particular to win an Olympic event is exceptionally difficult, and the margins of victory can be very small. The time of the 100 metres men’s final winner, Noah Lyles, was 9.79 seconds and the last placed finalist, had a time of 9.91 seconds (see World Athletics).
When it comes to financial markets, to be right the margins to be “right” can vary. If you take a release such as US CPI, the equivalent maybe of the 100 metres final (!), if we collect the forecasts submitted to Bloomberg’s survey few days beforehand, the mean absolute error can be under 10bps (for the year on year number). To be closer than the consensus is difficult. There are lots of forecasters participating in the survey. Also by the time US CPI is released, it will be after the reference month has passed. For example, this month the US CPI for reference month of July was released on 14 August. Hence, it becomes more a matter of measuring (ie. nowcasting) what has happened to prices over the previous month, as opposed to forecasting. Just a few days before the release, the competition for getting this number is intense and furthermore there’s less uncertainty because we “know” how prices moved in the previous month. So you margin to be better than the market is extremely small. Of course, the consensus can still get it wrong. It also depends precisely what you mean by “right”. It is one thing to get a smaller mean squared error, it is another to guess the direction of both the surprise and the directional move in inflation.
If we instead forecast inflation slightly further out, say 1 month. At least for US CPI, we theoretically have half a month of realised prices, but there’s also half of month where the prices are unknown, which require forecasting. Obviously, if we extend our forecast horizon to 2 months, 3 months etc. the uncertainty increases for any forecasting model. This might sound like bad news, if we are forecasting inflation. But obviously the uncertainty is something that impacts everyone, and on top of that there are fewer folks forecasting further out. Hence, our margin to get a good forecast further out isn’t going to be as tight, as it is with the nowcast.
So what should we do? For obvious reasons, short term forecasts (or nowcasts) garner a lot of market interest, but if we look further out in the curve there can also be trading opportunities, where we are also less likely to face less crowding as well. There can also be information content in understanding the shape of the curve of inflation, to understanding turning points in the future, which you would not capture purely with a nowcast. Indeed, if inflation might appear flat in the nowcast, is the same true in the coming months? Ideally, being able to forecast not only nowcasts, but also the whole curve of forecasts should give us insights that can be useful for trading. Capturing only one part of the curve, and nothing else could potentially give us a blind spot, we should try to avoid that.
I guess I have 4 years to train for the Olympics… the question is for which event?