by Calculated Risk on 2/07/2019 09:46:00 AM
Thursday, February 07, 2019
Because of my track record, I'm frequently asked about forecasting methods, my current views and what models I use. So here are a few tips.
First, why should someone care about the business cycle? If you are in a cyclical industry, forecasting is helpful with planning. Of course, if you forecast wrong - as people frequently do - then a company might hold back on investments and miss out on some opportunities. Another possible reason to forecast the business cycle is to improve investing performance. If someone could predict recessions and recoveries with reasonable accuracy, they could significantly improve their returns - but, once again, the track record for forecasters is pretty poor. Also forecasting is important for policymakers too (but most of us aren't policymakers).
I've successfully used forecasting as part of a cyclical business planning and also for my own investing. But it only helps if you are correct - so I don't recommend it for most investors - in general, for investing, it is better to just ride out the downturns then to try to time the business cycle. As an example, imagine if you had used ECRI forecasts for investing. You would have been late selling prior to the 2007 recession, and you would have missed a large percentage of the 2009 - 2019 bull market. Not a good result.
Second, many forecasters have an agenda, so be careful. For example, many newsletter writers and bloggers make frequent bearish forecasts. Perhaps they really believe their forecasts, or perhaps they are blinded by the incentives since being bearish garners publicity (and generates clicks and newsletter subscriptions).
Tip #1: Don't bother forecasting the business cycle (for most people). And don't rely on me - I could miss the next down turn (although I'll try to forecast the next recession).
If you are still reading, rely on history. Any reasonably intelligent person can almost always make a cogent argument that a recession is imminent. And yet, most of the time the economy is in expansion and stocks are in a bull market. This means forecasters have to fight through the tendency to be bearish. The economy and market don't care that you have a strong argument for a recession. Expansion is the norm.
Tip #2: Have a bias towards expansion.
Many forecasters start with the results of a model, and then based on the model, start warning that the sky is falling. A recent example was the focus on the yield curve inverting late last year.
This is backwards. My view is a forecaster should start with a story. For example, the Fed is overtightening, or the stock bubble burst will lead to less business investment, or the housing bust will lead to a collapse in the residential investment and a financial crisis. Those were some key stories prior to previous recessions.
Examples of possible stories now: If the Fed kept raising 25 bps a quarter (like last year), maybe rates would be too tight later this year or next year - and a recession might follow. But that doesn't seem to be the current course. Another possible story is the trade war might lead the economy into recession (but the negative impact doesn't seem large enough). Or perhaps the huge fiscal deficit - due to the new tax policy - will lead to much higher rates. Or there is too much corporate debt … and on and on.
This is why I track so many pieces of economic data - it helps me think about possible stories. And when data surprises, I try to take a closer look.
As I noted above, it is always possible to make an argument that a recession is coming, but it is usually wrong.
Tip #3: Start with a story (and try to see if the story makes sense - and the impact large enough to derail the expansion).
When a story makes sense, then it is time to start looking at some leading models. My favorite model uses residential investment as a leading indicator (not perfect, but pretty good). I also follow the yield curve (but it isn't currently as useful in a low interest rate environment).
When I'm on recession watch (due to a strong story and some modeling), then I start looking at some coincident indicators and models that are useful in timing a recession. For example, I start looking at unemployment claims, and some models based on the work of Hamilton, Chauvet, Piger and others.
Tip #4: Use models, but don't be a slave to the models.
Currently I'm not on recession watch. But I could be wrong!