Earnings as leading indicator

U.S. equity earnings appear to have declined, if ever so slightly, in 2011:Q4 from 2011:Q3. (Click on "Index earnings" to see the data, in xls format.)

I was curious to see whether stock earnings are leading, lagging, or coincident with the business cycle, so I put together some data. Below are the dates for the cyclical peaks of the economy, as determined by the NBER, the peak of the inflation-adjusted Shiller earnings, and the lead or lag between earnings and the economy:

Economy peak	Real earnings peak	Lead (+) / Lag (-), in months
Oct. 1873	Nov. 1873	-1
Mar. 1882	Jan. 1881	14
Mar. 1887	Dec. 1886	3
Jul. 1890	Jan. 1890	6
Jan. 1893	Jun. 1892	7
Dec. 1895	Jan. 1896	-1
Jun. 1899	Dec. 1899	-6
Sep. 1902	Dec. 1902	-3
May. 1907	Jul. 1906	10
Jan. 1910	Jan. 1910	0
Jan. 1913	Dec. 1912	1
Aug. 1918	Dec. 1916	20
Jan. 1920	Dec. 1916
May. 1923	Dec. 1923	-7
Oct. 1926	Aug. 1926	2
Aug. 1929	Dec. 1929	-4
May. 1937	Sep. 1937	-5
Feb. 1945	Jun. 1945	-4
Nov. 1948	Jul. 1949	-8
Jul. 1953	Sep. 1953	-2
Aug. 1957	Mar. 1956	17
Apr. 1960	Aug. 1959	8
Dec. 1969	Jan. 1969	11
Nov. 1973	Dec. 1973	-1
Jan. 1980	Sep. 1979	4
Jul. 1981	Sep. 1979
Jul. 1990	Mar. 1989	16
Mar. 2001	Sep. 2000	6
Dec. 2007	Jun. 2007	6
	Avg.	3.3
	Median	2.0

The average lead time is 3.3 months, and the median lead is 2 months. There is substantial dispersion, though: from a maximum of 20 months, to a minimum of -6 (i.e. the earnings peak occurred six months after the economy peaked). For the last three recessions, earnings crested before the economy did.

That is not to say, of course, that every earnings peak is a foreboding of recession. In part the trick lies in defining a "peak" for earnings. There have been multiple local maximums over the decades, with no ensuing recession. Still, a possible peak in earnings adds to the body of evidence that suggests that the odds of a recession in the near term are high.

A composite of leading indices

Following up on yesterday's post, here's a chart of a composite of leading indices for the U.S.:

When the super-index is below zero, there is a high probability that the economy enters a recession within three to six months. The lowest value of the super-index with a false positive signal of recession since 1967 has been -3.3 (in 2002). As of November 2011, the super-index was at -5.7. November was the fourth month in a row with a negative value. All three components of the super-index were negative as of November.(By the way, I’m using the brand new, improved index from the Conference Board, not the old, useless one.)


The December’11 value for the OECD index is not available yet. The other two indices were higher in December than in November, indicating a weaker signal of recession than in November. However, in all likelihood the super-index will still be around -5.

Statistical models like this do not “guarantee” anything. It’s still possible that a recession doesn’t happen.

One of two things may be happening:
1) The model is correctly pointing to a recession.
2) The model is “breaking down”: it is not able to capture the leading business cycle dynamics at present.

In the absence of any arguments supporting #2, and in light of the broad evidence from domestic and international macro data, it is prudent to say that the risk of recession is high.

It is also prudent considering the large divergence of outcomes. If the model is wrong, and no recession occurs, the best we can hope for is a mediocre recovery. This is what the stock market seems to be pricing in at present. If the model is right, and we do have a recession, sales and earnings will fall way short of “consensus expectations,” macro data will surprise on the downside, and risk-aversion will kick in, in which case stock prices are likely to dip. This is definitely not priced in by the market at present.

Construction of the super-index:

1. Calculate the three-month moving average of each of the following indices: the Conference Board’s Leading Economic Indicator (LEI) index, the OECD composite of leading indicators, and the ECRI weekly leading index. (For the latter, I start with the monthly figure, which is itself a monthly average of the weekly values.)

2. Calculate the six-month % change, at an annual rate, of each of the moving averages from step 1.

3. Average the three % changes.

A less-brave, new LEI

The Conference Board updated this month the composition of its Leading Economic Indicators index. The most important change was the substitution of M2 for a proprietary index of credit conditions. Over the last year, especially, it had become embarrassingly evident that M2 was pushing the LEI up, and that this was not justified by economic conditions. M2 accounts for almost a quarter of the index. Now that the Conference Board has replaced M2 with a more appropriate proxy for credit conditions, the LEI is a lot closer to pointing to a near-term recession than it was before.

Source: John Hussman.

It is not yet clear whether the LEI has reached a cyclical peak, or whether it's just taking a breather. (I, of course, think it's the former.) The six-month change of the index, for instance, is now in negative territory. Historically recessions have always been preceded by a negative six-month change of the index, although there have been instances where such negative change has not been followed by a recession. I have taken the three-month moving average of the LEI (LEI-3MA), in order to remove some of the short-term noise, and then calculated the six-month percent change, annualized, of that moving average. These are the true and false positives of this signal:

Aug. 69: True positive. Recession started in Jan. 70. Avg. value in the three months before recession start: -4.2%

Aug. 73: True positive. Recession started in Dec. 73. Avg. value in the three months before recession start: -5.1%

Mar. 79: True positive. Recession started in Feb. 80. Avg. value in the three months before recession start: -8.2%

Mar. 81: True positive. Recession started in Aug. 81. Avg. value in the three months before recession start: -4.3%

May. 89: True positive. Recession started in Aug. 90. Avg. value in the three months before recession start: -1.7%

Mar. 96: False positive. Recession did not occur. (Only one month when signal was present. Value = -0.77%)

Nov. 98: False positive. Recession did not occur. (Three consecutive months when the signal was present. Values: -0.53%, -0.30%, -0.15%)

Sep. 00: True positive. Recession started in Apr. 01. Avg. value in the three months before recession start: -8.9%

Jul. 06: True positive. Recession started in Jan. 08. Avg. value in the three months before recession start: -5.5% (Warning: in this instance, the six-month change of the LEI-3MA was negative between Jul. 2006 and Feb. 2007, then turned positive for five months, till Jul. 07, and dipped again below zero from Aug. 07 on. In those five months, the six-month percent change of the LEI-MA stayed below 1%.)

Nov. 11: ??? The six-month change of the LEI-3MA was -0.64%, and then it December it was -0.14%. The LEI-3MA does not offer strong-enough evidence, yet, of an upcoming recession.

Which EM economies are most likely to be in trouble?

The Economist has constructed a policy room index. It is intended to measure how much monetary and fiscal space economies have to conduct policy. The index has six components: inflation, excess credit (the growth in bank lending minus the growth in nominal GDP), real interest rates, currency movements, current-account balances, and a "fiscal flexibility index." The combine those inputs to produce an overall "wiggle-room index."

This reminds me a lot of another index The Economist put together a few months ago: the overheading index.

I combined the two indices to find which countries are in "most trouble" (i.e. high risk of overheating and little wiggle room). I set the threshold at 70, for both indices. I find that the trouble spots are: Brazil, Argentina, India, Vietnam, and Turkey.

Accelerating, or decelerating? GDP, or GDI?

We have had evidence for quite some time that "GDP(I) growth is better than GDP(E) growth at tracking fluctuations in true output growth." GDP(I) is commonly called Gross Domestic Income, and GDP(E) is Gross Domestic Product. Conceptually, both GDP(I) and GDP(E) measure the exact same thing, using different methodologies. The main reason, I think, why GDP(E) gets all the attention is that it is released one month earlier than GDP(I). It's also possible that a lot of people still don't know why and how GDP(I) is useful.

In 2011:Q3 GDP(I) grew just 0.22%. Perhaps more importantly for the short-term outlook, the q/q growth rate of GDP(I) has been diminishing during 2011: 2.68% in Q1, 0.26% in Q2, 0.22% in Q3. Far from improving, the U.S. economy seems to have been slowing down.

The growth rate of GDP(E), on the other hand, has been increasing: 0.36% in Q1, 1.33% in Q2, 1.81% in Q3. I do not know of any theory of why the path of the two growth rates is diverging. Nonetheless, an implication of the research mentioned above, by economist Jeremy Nalewaik at the Federal Reserve, is that output growth is more likely to have decelerated than to have accelerated through 2011:Q3. The naive approach of averaging the two measures yields the conclusion that output growth accelerated from 0.8% to 1.01% between Q2 and Q3. Still, a growth rate of 1.01% is nothing to be too cheerful about.

How fast is the economy really growing? We may not know till March 2012 (for Q4 the estimate for GDP(I) will probably be released with a three-month lag, rather than the customary two-month lag).