The much anticipated Conference Board LEI revision is out. Many people were fearing that the removal of M2 from the composite would plunge it into recession territory, but that is not the case. Our own investigations into this matter in the lead up to the announcement revealed as such, but it is comforting to receive hard confirmation now.
Composite LEI’s such as the e-forecasting.com eLEI, the Conference Board US-LEI, the OECD US-LEI and the ECRI Weekly Leading Index (WLI) are generally composed of a small set (no more than 10) of well selected leading economic indicators that as a composite are subjected to rigorous statistical and out-of-sample testing. You can see a perfect example of the rigors of such statistical selection in the white paper produced by the Conference Board on the construction of their new LEI. For that reason we lean toward the use of these composites for recession forecasting purposes rather than large sets of seemingly unrelated economic leading indicators cobbled together. There is the well-known saying that if you torture large arbitrary data series enough, they will tell you any story you wish them to.
We do not wish to debate the revisions or which LEI is the best (more on that in future articles). Instead, we will have a “first look” at the suitability of the new LEI for recession dating. For differences between the new and old LEI, see Doug Short’s commentary here. Using a 6-months smoothed compound annualized growth rate of the monthly LEI (as defined by Geoffrey H. Moore) into Probit recession probability models, the new Conference Board LEI gives very accurate leading probabilities of recession with zero false positives since 1967 as shown below:
The old LEI was more optimistic than the new one, but as you can see, the new LEI is still well below the trigger level for a recession call with probability of being in a recession now sitting at 0.7% and of being in recession within the next 6 months sitting at 6.2%. The one thing we notice is that once the probability of recession now rises above 13% this has ALWAYS led to recession within 2, 3, 10, 7, 0, 5 and 4 months respectively since 1967 with zero false alarms. This gives an average lead of 4.43 months, or 3.43 months to the real-time observer (since the LEI is one month post-dated).
If we prefer to observe the actual 6-month smoothed growth rates (annualised) themselves instead of the probabilities of recession derived therefrom, the chart below shows the average approach path of the LEI growth rate into NBER recession:
We note that the growth rate needs to fall below zero to give us an average 6-months (5 month to real-time observer) warning of recession. The growth rate over the last 12 months is now highly correlated with an average recession approach path meaning we need to keep on our toes despite it being above zero. However, a small consolation is that the last 12 months is tracking quite a bit above the average peak approach path.
In similar fashion to that which we described above, neither the e-forecasting.com eLEI, the Conference Board LEI, nor the OECD US-LEI is forecasting recession in the next 3-4 months. The only other composite LEI flagging recession is the ECRI Weekly Leading Index (WLI), but as we stated in our previous posting (Further Improving on the use of the WLI) this is a weekly leading index and in the original spirit of the construction of said index, it is more timely and frequent but also subject to far less accuracy (more false positives).
This by no means says that recession within the next 6-12 months is out of the question. But it says we are neither in recession now nor should we be in one for the next 3 months at least. There is no doubt the US economy faces significant external risk. Coupled with a sub-par recovery, this means the smallest external shock can plunge us into recession. These external risks are not fully discounted into all the components of these composite LEI’s namely a possible steep and protracted Euro-area recession, an EU member sovereign default or credit event, hard landing in China, failure by congress to reach agreement on deficit reductions, further possible US credit rating downgrades, Iran tensions/conflicts and oil-price shocks. The probability models just look at what we are seeing in the US economic leading indicators and do not take into account these external factors (since in our view, this would be speculation and we just want to deal with hard real numbers). Should these external shocks come to light, they will certainly show up in the US economy somewhere (in “short-order”) and then be detected by models, but until then one cannot speculate on the probability of these external risks. We can only acknowledge they exist, and are many. Our favoured approach is to watch short leading probabilities (4-6 months) and action stock market positions accordingly as having to action long-leading predictions that can vary anywhere between 6 to 18 months with wide standard deviation is unlikely to produce productive results in the long run.
Conference Board: LEI to be Revised by Lance Roberts (GEI News, 19 Jan 2012)
About the Author
Dwaine van Vuuren is CEO of PowerStocks Investment Research, a South African-based provider of investment research. If you would like to receive the next 4 weeks SuperIndex Recession Reports for free, just email us at firstname.lastname@example.org with FREE SUPERINDEX in the subject line.