For a few weeks I have been examining the implications of monthly data points on quarterly GDP growth. Specifically I am trying to see which months affect which quarters. The results were put in a model that appears to look just a bit ahead.
The model above shows an estimate of 4.3% for second quarter GDP. In the last 4 years the model has averaged estimating high by about 0.4%, so a 3% something quarter might be a better estimate. Building a model to estimate the future is a bit like putting a jigsaw puzzle together where you don't have all the pieces, where some of the pieces are to a different puzzle and where putting a piece in place changes the way the other pieces fit together. The effort often gives a better view of the future, but can also turn out the other way.
Below are the five inputs to the model in the order that the regression analysis suggest have the highest signal to noise ratio: manufacturing hours, stock market, housing starts, GDP auto correlation and real retail sales.
Manufacturing hours are the most important factor in the model and by itself, based on the correlation in the last eight years, suggests Q2 will come in about 3.5% which is the strongest reading in about 4 years. Since this indicator leads by one month the three months that go into the Q2 estimate are March, April and May 2014. These three months are highlighted in yellow in the chart below.
The three months that correspond with the -2.9% Q1 were all negative and are highlighted in pink. The 0.21% drop in June, highlighted in green suggests Q3 is off to a weak start.
The next three variables have almost the same weight in the model, but the stock market is slightly more important.
While historically the stock market has a several month lead on the economy, in the last 8 years the strongest correlation is coincident. No lead! So, stock market performance in April, May and June are likely the three months with the best indication for Q2 growth. The average daily close for the S&P 500 in Q2 was up at an annualized rate of 15.2% from the average price in Q1. In the context of the last 8 years this would be consistent with about 1.6% GDP growth in Q2.
Housing starts maximize their correlation to quarterly GDP growth with a seven month lead. So the increase in housing starts for the months September, October and November 2013 indicate a strong Q2, in fact the third strongest in the last 8 years.
While Q2 should have a big bounce back, Q3 should be much weaker. The weak June starts number released Thursday should influence Q1 2015 given the 7 month lead time.
In choosing to put housing starts in the model I also examined building permits. In the last 8 years starts had a slightly stronger individual correlation with quarterly growth than permits. When both were included in the model the influence of building permits became insignificant. Even though I found starts to be the more robust indicator there are undoubtedly variable combinations where a model would show permits were better.
When growth is strong there is a modest tendency for it to be weak 3.75 years (11 quarters) later and vice versa. By itself it is a weak indicator, but in the context of the other indicators it appears to be more important.
So the red line in the chart above shows the same data as the grey bars, just inverted and pushed forward 11 quarters. This indicator indicates a rebound in Q2. However, the 4.9% growth in Q4 2011 now suggests Q3 2014 will be weak.
Real retail sales lead GDP by a month and suggest Q2 growth of about 3.6%.
Real sales in March, April and May highlighted in yellow below are behind the Q2 estimate.
The June data point, which will begin to tell the story for Q3, will be available July 22 when the CPI comes out. In the meantime, we can look at retail sales and get a slightly less accurate look at the beginning of Q3.
Since hitting a high in March the monthly growth rate of retail sales has declined for 3 straight months. In June it was up 0.24%. If CPI were to come in at 0.3% real retail sales would be negative for June.
I also tried adding the yield curve to the model, but its noise to signal ratio was quite weak.
The red line in the chart above shows the difference between the 10-year Treasury yield and the 1-year Treasury yield averaged for 3 month periods. Historically this leads GDP by about 20 months.
Normally this is one of the best indicators. An inverted yield curve (short term rate higher than long term rate) warned of the fall into The Great Depression and the nine recessions since 1955.
However, it did not warn of the four recessions between 1935 and 1955 when short term interest rates were below 2%. In 2008 the 1-year yield fell below 2%. The 20 month lead time means that the influence of the yield curve with an unusually low short rate began in Q2 2010. Since then the yield curve seems to have had little or no correlation with growth. For example, Q1 2011 and Q4 2014 were the two weakest quarters between the great recession and Q1 2014, yet the yield curve suggests these should have been some of the strongest quarters.
It is possible that the economy is in a state where the yield curve does not show any influence on economic growth. If so a recession could come without the expected year or more advance warning, as they did between 1935 and 1955. If the yield curve is still influencing growth it suggests 2014 will be the weakest year since 2009.
The two variables with complete data through the third quarter (housing starts and GDP auto correlation) suggest Q3 growth will be about 0.4%. Manufacturing hours and retail sales suggest the quarter is off to a weak start. While I am pretty confident Q3 will be weaker than Q2 it is too early to have much confidence in my expectation that it will be below 1%.
The growth in jobs, consumer credit, commercial and industrial loans along with the low initial unemployment claims do not indicate the economy will improve. They are all lagging indicators reflecting the afterglow of GDP expansion.
Problems with the L.E.I.
The Conference Board's index of leading economic indicators points to fairly strong growth over the next 6 months. However, it appears they do not correctly account for the actual lead time of their indicators. For example, they list Jobless claims as a leading indicator when it actually lags. They list the stock market as leading when it has been running at a concurrent pace for several years. Some of their indicators like manufacturing hours and new orders only lead one month. The credit spread between the 10-year Treasury yield and the Fed Funds rate is similar to the yield curve mentioned above; the lead time is long and recent improvement does not indicate what will happen in the next 6 months. Plus, as with the yield curve it may not currently be a valid indicator.
Given the problems and potential problems with the LEI and yield curve we should not rule out the possibility of a recession starting in the third quarter. Inventory accumulation may pose a serious threat to the economy.
Inventories have grown for ten straight quarters and for 16 of the last 17 quarters. If you back out the change in inventories above from GDP you get a bit less volatile view of growth.
For example, excluding inventory change Q1 only declined 1.3% rather than 2.9% and Q3 2013 grew a more modest 2.6% rather than 4.1%. The red line in the chart above marks the -1.3% level, which is worse than any of the quarters for the recessions in 1949, 1960 and 2001. The worst quarter in the 1970 recession was only a few basis points weaker. This is the first time a quarter has been this weak without being in a recession.
The annual growth rate for GDP excluding inventory changes has dropped for 5 straight quarters to 1.7%. The red line in the chart below shows that the last eleven times this measure of growth dropped this low the economy was already in a recession. In the 1960 recession this was the low point.
The $171 billion increase in inventory over the last 10 quarters was likely accumulated on the expectation for GDP growth of 3% or 4%, which for years economists have forecast was just around the corner. If growth does not improve businesses may feel the need to reduce inventory. The economy has been growing at stall speed while adding inventory, if businesses reduce orders to cut inventory it will likely feed through to a recession and job losses.
Don't party too much when strong growth is announced for the second quarter. The third quarter looks to be weak. Remembering the one month lead time the first look at the second month of Q3 will come when the July average manufacturing hours is reported on the same day as the jobs report. If the economy falls into recession the stock market (NYSEARCA:SPY) will be a bear.
Disclosure: The author is short SPY. The author wrote this article themselves, and it expresses their own opinions. The author is not receiving compensation for it (other than from Seeking Alpha). The author has 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.