Recession Watch: January 2018

by: John Early


Seven factor model estimates 2.6% Q3 GDP, splitting the Atlanta Fed GDPNow forecast of 2.7% and the Blue Chip Consensus at 2.4%.

With 58% of the data in to estimate Q4 the model suggests 1.4%.

Still expecting a Recession, the watch now is on Q1 2018.

The yield curve and low unemployment claims do not reduce the chance of recession in Q1.

Stocks will likely trend up until recession, but face great risk when economy turns down.

GDP model and factors
The Model attempts to find the combination of seven indicators or influences that work together to best predict quarterly GDP growth. It estimates the relationship with GDP of these seven factors over rolling eleven year periods. The lead time of the factor with the highest correlation is used in the model. As time moves on factors may come or go from the model. There were no factor changes or significant weighting changes in the last quarter.

The Seven Factors

The S&P 500 (purple line) The 13 weeks 7/19/17 through 10/18/17 associated with Q3 GDP rose at an annualized rate of 13.2% from the average of the prior 13 weeks. For purposes of estimating Q4 our guess is the next 13 weeks will show a similar increase.

Real retail sales (Orange Line) grew at an annualized rate of 1.7% just one tick higher than Q2. A strong September number and an upward revision to August pushed this number much higher than estimated a month ago. I am guessing an identical increase to estimate Q4 GDP.

Cass Freight index TM From Cass Information Systems, Inc. (Black line) Shipments corresponding with Q3 growth were up at a 24.2% annualized rate. The leading point corresponding with Q4 suggests a slight decline. Seasonality should also send Q1 2018 down. Curiously leaving the seasonality in this indicator improves the GDP model.

Industrial production (green line and dots): Months indicating Q3 matched the growth of the prior quarter. The two leading months suggest Q4 will be negative.

Real 10 Year Treasury Yield (brown line and dots) indicates a declining rate of growth for all of 2017. The indicated decline may bottom in Q1 2018, but it is too soon to call it a low since the last two months showed declines and future months could go lower than they indicate for Q1 2018.

Real private inventory (blue line) Inventory associated with Q3 and Q4 growth suggests weak growth, but not enough to lower the GDP estimate much. It suggests a nice bump in Q1 2018, but then even weaker growth for the two quarters after that.

Oil (Gold line and dots) The falling oil prices 21 months ago (rise on graph since scale is inverted) suggest modest growth for Q3 and Q4. The 26% price spike (251% annualized rate) back in 2016 suggests Q1 2018 could be very weak.

Q2 recap: GDP grew at 3.1%. Our original estimate was for 2.5% revisions to the data in the model moved the current model value up to 2.7%.

Recession Watch

The current expansion is 99 months old and the 3rd longest in US history. If it lasts 7 more months it displaces the one primarily under Kennedy/Johnson as number two.

The popular belief that the positive sloping yield curve means a recession is at least a year or two off probably is not valid since that indicator has historically not worked when the 3-month T-Bill yield is below about 2%.

yield curve An inverted yield curve correctly gave warning for 13 of the last 17 recessions. The 4 times it failed to give warning between 1935 and 1955 the 3-month T-Bill yield was below 2% as it is today. The low yield associated with an elevated monetary base probably renders this indicator useless at present.

The yield curve is about where it was when the 1949 recession started and is flatter than when the two recessions prior to that started.

Initial unemployment claims are a lagging indicator so the lowest claims since 1973 reported last week do not show what is coming. Unemployment claims lag job growth by about 17 weeks.

initial claims & jobs

The trend in unemployment claims disconnected from the trend in job growth in 2014. Claims are probably suppressed since about 10,000 baby boomers are retiring or changing to part time work every day. This makes it easier for businesses to manage their payroll through attrition without having to lay people off.

Job growth has been trending down since 2014 or 2015 depending on how you measure it. Job growth trending down for two or three years is consistent with a recession starting anytime, but since it too is a lagging indicator it does not indicate a recession coming.

gdp & jobs

Since job growth lags quarterly GDP about five months and the underlying trend in initial claims lag job growth 17 weeks. The low claims number meant the economy was strong back around April. It does not mean there is no chance of recession in January.

It is too soon to actually forecast a recession. If retail sales, industrial production and the stock market were to roll over it could mean a recession starts in December. If they remain strong through early 2018 this recession watch will have been just a watch.

The point is the chance of recession early next year is significantly greater than the models which rely on the yield curve or unemployment claims suggest. All the models I have seen in Seeking Alpha articles that say the chance of recession is remote are based on indicators that will likely not give a warning this time.

Shaky Market implications

Last quarter I said the stock market (SPY) and economy were all clear for another 3 months. And so it was. Now I am less certain. Momentum suggests stocks will continue trending up. It is unlikely we have had the top in stock prices and even if we have there would probably need to be at least 3 months of dip buying failing to be rewarded with new highs before there was a significant decline. However, the weaker expected growth in Q4 and potentially shrinking Q1 2018 mean more risk for stocks.

If a recession does start overvaluation of stocks makes for a potentially devastating bear market. The valuation measure I use, PEses, is similar to Robert Shiller's CAPM, but I use single exponential smoothing for real earnings as suggested by Bob Bronson instead of a 10 year moving average.


The current reading suggests greater overvaluation than anytime except for 3 years from early 1998 to early 2001.

Shiller and John Hussman typically look at valuations impact on return over the following 10 years. I examined return over 1 year periods 2 year periods etcetera out to 25 year periods. Valuation appears to have its strongest correlation with return over 19 year periods.
peses 19 year returns The chart shows the relationship between the PEses at the beginning of a 19 year period (green inverted line on the right part of the chart) and the rate of return after inflation and reinvested dividends over the next 19 years (dark gray line on right side of chart). The left and right side of the chart show the same data. The scatter plot on the left side of the chart shows the same data as time-series plot on the right.

The last 19 years annualized a 4.5% return. The PEses was 50.4 back in October 1998. Using the historical correlation this would have been consistent with stocks loosing -1.7% a year the last 19 years.

The chart suggests the stock market will annualize a loss of 6% after inflation and dividends over the 19 year period ending in December 2018 based on the PEses peaking at 66.1 in December 1999. A 6% loss annualized for 19 years would mean a 70% loss from December of 1999.

While I don't think there is any chance of that, the chart still has useful implications.

In the last 20 years this valuation measure along with Shiller's CAPM have averaged about twice as high as they did in the previous hundred years. Hussman has also had trouble with his valuation measures.

With valuation measures staying at elevated levels the last 20 years the gap between estimated return and actual 19 year returns began to grow to unprecedented levels in early 2016. So far, the gap reached a peak in April 2017 and has moderated the last few months.

The gap between 4.5% and -1.7% for the most recent period was 6.2%. If this gap persisted the chart would be consistent with a bear market taking prices down to about where they were in 1999 or about a 44% loss, perhaps by the end of 2018.

If we were to get a recession in 2018 a 44% loss would be comparable to the stock market decline that came with the 2001 recession.

While not ready to forecast recession and certainly not a market top, market risk may be substantially higher than perceived. If the gap of 6.2% persisted over the next 19 years it would suggest stocks return about 3.8% a year after inflation and dividends. If valuation levels returned to what was normal prior to 1996 the implication would be a loss averaging 2.4% for 19 years.

While estimating value is more difficult these days, it will remain true that if you overpay, your return in the future will be weaker and perhaps negative.

If/when we go two or three months without a new stock market high and if new data suggests a negative quarter of GDP (perhaps Q1 2018), exits to the stock market could get crowded and result in rapid drops. Stay tuned for next quarter's update.

Disclosure: I/we have no positions in any stocks mentioned, and no plans to initiate any positions within the next 72 hours.

I wrote this article myself, and it expresses my own opinions. I am not receiving compensation for it (other than from Seeking Alpha). I have no business relationship with any company whose stock is mentioned in this article.

Additional disclosure: There is no guarantee analysis of historical data their trends and correlations enable accurate forecasts. The data presented is from sources believed to be reliable, but its accuracy cannot be guaranteed. Past performance does not indicate future results. This is not a recommendation to buy or sell specific securities. This is not an offer to manage money.