Probability of US recession still appears overblown

Not even a week went by from the release of our blog Probability of US recession appears overblown before the Philadelphia Fed Business Outlook crashed to -30.7, the lowest reading since February 2009, and intensified speculation of a recession in the world's largest economy. Since then we have also received news of a disappointing Employment report and a better-than-expected, but weakening, ISM Manufacturing Index. In light of the news received in August we will revisit our story, though from a slightly diferent angle.

U.S. GDP vs. Recession
Our first piece concentrated on GDP explicitly and proceeded to suggest that while we were likely to see weak growth in the coming quarters in the U.S. it would not likely be negative growth and hence our headline read Probability of US recession appears overblown.

Chart 1: U.S. GDP and recessions since 1959


The National Bureau of Economic Research (NBER), however, does not use GDP to judge whether the economy is in a recession or not, but rather a wide range of economic time series. Four of those are part of the Conference Board's Coincident Indicators Index*. Therefore we will deal with the monthly recession index this time rather than the (real) GDP series in judging whether a recession is close by - or perhaps already here.

Is the U.S. economy already in recession?
The string of weak data in August caused many to exclaim that not only was a recession probably around the corner, but it was perhaps indeed already here. The NBER has not issued any stance on the first seven months of 2011, but given the reliance of the NBER on the four indices in the Coincident Indicators Index in determining recessions, we proceed to model the probability of a recession in the first seven months of this year for which we have data on the Coincident Indicators Index using logistic regression.

Table 1: Modelling recession in contemporary month

Model estimation (Apr 1959 to Dec 2010)
Variable	Value	Std. Err.
Intercept	-1.28	0.19
CB Coincident Indicators, MoM	-6.38	0.88
CB Coincident Indicators, MoM (lagged 1 month)	-4.29	0.79
CB Coincident Indicators, MoM (lagged 2 months)	-2.49	0.67
Note: 621 monthly observations, CB = Conference Board

Chart 2a: Fitted probability of a recession 1959-2010


Except for three false positive months in mid-to-late-1959 due to a 2 percent five-month decline in the Coincident Indicators Index, which did not cause the NBER to declare a mini-recession, and a one month spike in September 2005, the model does well to capture the official NBER recessions. Using the seven month's worth of data on the Coincident Indicators Index from January to July of this year we can model whether the U.S. has been in a recession in this time span.

Chart 2b: Nowcast probability of a recession in Jan-Jul 2011


The 'nowcast' probabilities (red shared area) point overwhelmingly to a no-recession vote in the first seven months of 2011 though we have to keep in mind that recent 'hard' data can still be revised downward to show more of the weakness seen in surveys.

The Conference Board's Leading Indicators Index does not point to recession...
So far so good. It seems unlikely - according to the model above, at least - that the data for the first seven months of 2011 will make the NBER rule in favour of a recession in any of these seven months, but what about the future? We already have plenty of disappointing data for August and if it continues into the fall months surely this must mean a recession is imminent?
Before turning to some of the more timely monthly indicators let us examine what the Conference Board's Coincident Indicators Index (COI) and Leading Indicators Index (LEI) have to say on the matter. Details on the five-month's ahead regression can be seen in the table below:

Table 2: Modelling recession five months ahead

Model estimation (Jun 1960 to July 2011)
Variable	Value	Std. Err.
Intercept	-1.68	0.15
CB Leading Indicators, 3Mo3M (lagged 5 months)	-0.79	0.11
CB Leading Indicators, 9Mo9M (lagged 8 months)	-0.24	0.04
Note: 614 monthly observations, CB = Conference Board

Chart 3a: Forecast probability of a recession in Aug-Dec 2011 using the Conference Board's Leading Indicators Index


No recession in the last five months of 2011 either (green shared area), according to a model based on the COI and LEI. Perhaps not surprising considering that the COI is up 2 percent year-on-year while the forward-looking LEI is up 6.2 percent. Furthermore, according to the model presented in table 2 the 3Mo3M LEI growth rate needs to decline approximately 3.7 percent in the coming three months (with the 9Mo9M growth rate standing at 1.58 percent in July) for the probability of a recession in March 2012 to reach 50 percent. Similar results are found for the models looking 1 to 4 months ahead.

Remember though that data for these indices only goes as far back as July and hence does not capture the weakness seen in August. Let us thus proceed instead with some of the most up-to-date data available, namely 1) Nonfarm Payrolls, 2) ISM Manufacturing, 3) Initial Jobless Claims and 4) Conference Board's Consumer Conference. These are available for August (Initial Jobless Claims through August 26), which make them timely, and in addition they are also among the economic data often used to justify that another recession is not far off.

...nor does a leading indicators index based on timely data point to recession
The first Principal Component (PCA) of the four time series mentioned above explains 71 percent of the total variation and hence allows us to use this single variable (and lagged versions of) as input to our logisitic regression; but even using these economic series a recession is not on the cards. While this regression does suggest a higher probability of recession in the coming months it remains below 20 percent.

Chart 3b: Forecast probability of a recession in Aug-Dec 2011 using timely data


In conclusion, we maintain our stance that the U.S. economy is weak, but it is not yet in recession nor do we expect the economy to slide into a recession in the coming months based on available data. Mind you though that this exercise does not include adverse impacts from (a lack of) government spending or the Eurozone sovereign debt crisis. Surely with the latter plastered all over the newspapers' front pages at the moment the risk of a recession is bound to be higher than what this exercise suggests; but the probability is not high enough for us to make a recession call.

Appendix: Nonfarm Payrolls as explanatory variable
Even when we use only the much bandied about Nonfarm Payrolls series, which printed a net change of 0 in August, we remain unable to call a recession with the model suggesting a 16 percent probability of a recession last month. This very simple model suggests that the year-on-year growth rate in payrolls needs to print -2.25 to -2.50 percent (or the quarterly growth rate needs to print -0.30 to -0.40 percent) for the probability of a recession to reach 50 percent. Surely this model is too simple to explain the business cycle as determined by the NBER, but it is nevertheless instructive that despite the loud noises surrounding the (admittedly poor) Employment report last Friday, payrolls are still growing almost 1 percent a year (0.97) and 0.08 percent quarter-on-quarter, and hence do not as such point to a recession.


* 1) Nonfarm payrolls, 2) Personal income less transfer payments, 3) Industrial production, and 4) Manufacturing and trade sales.