Reflections [Expanded version]
Dealing with a runaway market
Those of you who have been following us since 2010 will identify us a perma-bulls. Even in the depths of the ECRI 2012 /Hussman recession calls we were firmly bullish on the US economy and stock market – quite contrary to the popular consensus at the time. Those subscribers who have been diligently following the RAVI SP500 forecasting model and its consistently accurate bullish forecasts will have noticed this year that all the targets we have set for 3Q2018 have been met on our Dashboard:
This means, for the first time since we have been running this model, that RAVI SP500 valuations have finally run ahead of themselves. This is not to say this exuberance will not continue for some time, but it is a warning for those who like to deploy valuation risk metrics in their asset allocation models.
Now for a long time, various valuation models have been at elevated levels. Here are a few below:
Despite these elevated levels, RAVI was forecasting bullish returns for 2016 and 2017. All these models above have very good correlations to 10-14 year ahead SP500 returns and in many instances of late were actually forecasting muted to negative 10-year ahead returns. But just because a valuation model is forecasting a negative 10-year return doesn’t mean that 1-3 year returns will be poor! This mistake is consistently made by forecasters! A case in point is shown below, using Warren Buffets famous valuation metric. Since 2014 this model has been persistently forecasting low to negative 10-year ahead returns – but that didn’t stop the SP500 roaring ahead!
Does this mean Warren Buffets indicator is useless? Of course not – it means it’s been interpreted incorrectly! In all likelihood, in 10 years’ time we come back to this chart and remark on how accurate it was since the black line (actual 10-yr returns) will closely track the green line (the forecasted 10-year returns.) As far as a long-run forecasting models go, the Buffet indicator, whilst not the best, is fairly respectable with an r-squared of 0.76 – so it would be surprising if that black line doesn’t hug the green line closely!
There is no model that can directly predict short-term SP500 returns with meaningful accuracy. But you can get pretty damn close (as these things go) by deriving short term returns from accurate long-run models like the Buffet model. It works like this:
1.Find a really robust and accurate long-run (assume 10 year) forecasting valuation model (there are many)
2.Get the 10-year forecast from this model from nine years ago
3. Work out how much the SP500 has grown since this date 9 years ago
4. Subtract (3) from (2) to work out how much upside is left in the year ahead.
In fact we can apply this short-run look-ahead method to any x-year horizon, and this is what we do with the RAVI to compute 1/3/5 year look-ahead forecasts:
The important thing to bear in mind is that the current reading of a long-term valuation model, be it Buffet, Tobin-Q, Shiller P/E, Hussmans Market-cap to Gross Value Add etc. means absolutely nothing to short-to-medium term forecasts. What is important, is what this model was saying 7-10 years ago and how much the SP500 has eaten into those forecasted gains since then.
Now the RAVI long-run model has the following 10-year forecasting profile, which at 0.89 r-squared is pretty respectable:
If we use the methodology described above to forecast 1-year ahead returns we get the following profile. I can tell you, an r-squared of 0.4 on such a short-run forecast is pretty decent as these things go. But quite obviously from the profile, you can see wide variances on occasion:
On a three year look-ahead things become rather respectable in terms of correlations for short run forecasts:
Five-year look-ahead accuracy is pretty remarkable given how close its correlation is to the accuracy of the 10-year forecast:
We can deduce that both the 5-year and the 3-year look-ahead models are forecasting around 4% annual growth for the SP500 from here on, which is pretty low. In fact it becomes pretty interesting to track the average annualized look-ahead forecast of all the models over time:
You can see that as the SP500 has been tracking upwards relentlessly, so the average annual forecasts have been declining of late. But more interesting is that when the average annual forecasts turn negative, it looks like, on the surface, we have a nice recession/bear market warning. Stock market valuation model as recessionary indicator – the idea is appealing!
So there you have it – we are in an era of weak forecasted SP500 returns with all the elevated risks associated with that. You need to decide if we are in a “new paradigm” of prolonged elevated valuations (its quite possible) and take your chances, or de-risk accordingly. Caveat Emptor
Like what you see? Please support us with a modest annual subscription over here http://recessionalert.com/sign-up/
Negative returns for SP500 in next decade
The RecessionALERT Valuation Index (RAVI) has been updated for 1Q17 and for the first time since 1999, is forecasting negative 10-year total returns for the SP500:
The chart on the right shows that the RAVI continues to forecast SP500 decade-ahead total returns with relative accuracy, especially when one considers that the green forecast line has data points that are seen 10 years before the black line depicting actual 10-year returns.
Now there have been a lot of valuation models predicting 10-year ahead negative returns but this does not mean one should be getting out the stock markets. One needs to review the short run (albeit less accurate) forecasts. As we can see below, these still hint at reasonable short run returns of the order of 11% per annum:
In March 2017, we used the 4Q16 RAVI forecast to predict returns for the SP500 to the end of 2017. Our low estimate was 2472, our median estimate was 2,565 and our high estimate was 2,718. As you can see from the RAVI daily tracker on top-right below, we have achieved the lower target and are 3.8% away from the median target and 10% away from the high target with some 5 months still to go:
Here is another interesting way to asses potential short-run returns for the SP500 – look at the US economy as per our Long Leading Index. When this signals recession then we have cause for concern for the stock market
25 most important US economic indicators animated in 1 minute
Here’s something fun we played with after just compiling our June Long leading US Index report for our subscribers
Horrific revisions to HWOL data
The Conference Board Help-Wanted-Online (HWOL) program is closely followed by us to get a feel for the labor market. It is one of over two dozen labor indicators we examine. The monthly HWOL data have been produced by the Conference Board since May 2005, replacing the Help Wanted Advertising Index of print ads, which was published from 1951 to 2008. HWOL data contain the universe count of all ads posted online during a month, with a mid-month survey reference period (e.g., data for October would be the sum of all posted ads from September 14th through October 13th). The HWOL program collects data from over 16,000 online job sources and removes duplicated ads.
The February release incorporated revisions with the following description “With the February 2017 press release, the HWOL program has incorporated its annual revision, which helps ensure the accuracy and consistency of the HWOL time series. This year’s annual revision includes updates to the job board coverage, a revision of the historical data from May 2005 forward, an update of the Metropolitan Statistical area definitions to 2015 Office of Management and Budget (OMB) county-based MSA definitions, and the annual update of the seasonal adjustment factors.”
“Great stuff” we thought. “Some solid revisions to keep us honest”. First up was “Total ads” which are all unduplicated ads appearing during the reference period. This includes ads from the previous months that have been reposted as well as new ads. We had to double-take when we looked at the data. It looked vastly different to the last time we peeked at it. Upon inspections the revisions were downright nasty:
We also look at the “New ads” which are all unduplicated ads which did not appear during the previous reference period. An online help wanted ad is counted as “New” only in the month it first appears. This is more volatile but provides useful comparison data to “Total Ads”. Again this is when that coffee you’re holding spills into your lap as you look in disbelief…ugly ugly ugly.
Both of these metrics are screaming recession from the hilltops – and this just after revisions. Given that the Conference Board is an organisation that’s being doing the data compilation and revisions thing for a while now, this has to be pause for thought. The labor market is not as well as we may think.
Let’s look at various labor indicators taken from our Long leading Index of US Economy:
So these HWOL revisions look rather alarming, but are not being collaborated by many other traditional metrics. Lets look at it again, but in the context of the last recession (grey shading). There’s no two ways about it – butt ugly.
The folks at The Conference Board are onto this though. From their website : “NOTE: Recently, the HWOL Data Series has experienced a declining trend in the number of online job ads that may not reflect broader trends in the U.S. labor market. Based on changes in how job postings appear online, The Conference Board is reviewing its HWOL methodology to ensure accuracy and alignment with market trends.”
The Board of Governers sniffed that something was up in June 2016 already, observing a substantial divergence since the end of 2012 between HWOL and the Job Openings and Labor Turnover Survey (JOLTS), administered by the Bureau of Labor Statistics. From their note : “All told, the average price for Craigslist job ads rose substantially, and roughly doubled since the end of 2012 (Figure 2), coinciding with the period when online vacancy posting as measured by HWOL noticeably underperformed the JOLTS vacancy growth.”
Now that the Conference Board is sanity-checking its HWOL methodology, it’s going to be really interesting to see how this pans out. But can this really be simply blamed on Craiglist prices? I’m not so sure. Surely its not that ridiculously simple?
Our house view is that we may have recently narrowly avoided recession, but are not close to it now. The lessons here are that reliance on single economic time-series, no matter how comprehensive they may appear (HWOL is fairly comprehensive in its number of datapoints), can be dangerous. It also highlights how revisions can wildly swing data with certain economic time series (as opposed to financial data such as interest rates, interest rate spreads, bond yields, credit spreads, corporate bond yields and the like.)
Mixed Signals from Labor Market
We keep getting good news about employment and the labor market. But we rarely see the less optimistic numbers.
Yellen’s Labor Dashboard (see here) is looking strong with all but 3 of the 9 components above pre-recession levels:
An index derived from the percentage of U.S states with rising unemployment looks worrying. Whilst the national unemployment figures seem fine, the population demographics of some large states seems to be masking underlying weakness at a state-level. Again, the broadness of this index makes for worrying numbers:
Here is another way of looking at the state-level unemployment numbers. The black line is derived from simply averaging up the unemployment rate from each individual state whilst the green line is the actual % of states with unemployment rates that are rising.
That means 75% of US States have unemployment levels higher than the best (lowest) numbers yet witnessed.
In summary, it would appear that several large-population states may be enjoying employment gains, but large swathes of the U.S are not.
NOTE : The state-level unemployment metrics shown above are excellent early-warning signals and are included in our highly comprehensive U.S Long-Leading Economic Index
U.S Economy remains vulnerable
It is true that some genuinely troubling signals are starting to make themselves known. Let’s look at some of them.
But by far the most worrying trend, is the labor market, where a broad-based, consistently increasing weakness among the 52 US States is being masked by national numbers being touted about that include high population density states:
The prior chart that shows the average unemployment rate among the 52 states, together with the % of 52 US States with rising unemployment (now at 50%) being in far worse shape than the national unemployment rate below implies:
These are just a few indicators in a battery of twenty-one that we examine, and whilst there are no alarm bells yet, the aggregate composite of all 21 indicators shows the US economy the most vulnerable to exogenous shock since this expansion started:
Unemployment more widespread than thought
The average state unemployment rate seems to be putting in a bottom, a less flattering picture than that painted by the nationwide unemployment rate. But the real surprise comes from the percentage of 52 US states that have rising unemployment, which shot to over 60% for the month of July 2016. The last time this rose to above 60%, was six months before the economy peaked in 2008.
The chart below shows all prior occasions the percentage of U.S States with rising unemployment went above 60%. We see that the percentage of US states with rising unemployment is an excellent proxy for the probability of US recession:
It is clear some close attention is going to have to be paid to this metric over ensuing months. Whilst the national rolled-up unemployment figures are not showing signs of stress, it is quite clear that stresses are building up when examining the individual state data.
The percentage of U.S states with rising unemployment is but but one of 21 indicators we use for our Long Leading US Economic Index.
NBER’s Big-4 Indicators had a narrow miss
Reading through all the positive press about jobs numbers and so forth, its hard to comprehend that the 4 main indicators used by the National Buro of Economic Research (NBER) to determine US recessions, had a narrow miss recently.
If you recall from our popular 2012 article, the NBER does not define a recession in terms of two consecutive quarters of decline in real GDP. Rather, a recession is a significant decline in economic activity spread across the economy, lasting more than a few months, normally visible in real GDP, real income, employment, industrial production, and wholesale-retail sales.
With regards to the monthly data, they examine 4 monthly co-incident indicators:
- Industrial Production
- Real personal income less transfers deflated by personal consumption expenditure
- Non-farm payrolls
- Real retail sales deflated by consumer price index
If one takes the month-on-month percentage change of these four data series, and combine them into a standardized average composite, you get the below, where we had negative prints for February and May:
We can also count the amount of times the non-smoothed month-on-month indicator in the first chart has dipped below zero in a rolling 4-month period. When this count reaches 2 we are in a “danger zone” and when it hits 3, we are most likely already in a recession. We hit the danger zone in May and June. Lets call this count reaching three Syndrome-Two.
Another interesting thing to note is how many consecutive negative prints the indicator in the first chart has made. When this hits three we are also most likely in a Recession. Lets call this event Syndrome-Three:
The below Syndrome count shows how many of the three syndromes are present at any time. If two or more syndromes are present, we are most likely already in recession.
We run a more sophisticated algorithm for the Big-4 for our subscribers, as described in this robust tested methodology, and here are the probabilities of recession from the model:
To conclude, looking at the individual co-incident monthly data used by the NBER shows a far more pessimistic view currently than when looking at a syndrome of conditions. But the co-incident data in this particular indicator and the recession probabilities we are registering are not as bullish as the employment data would have you think. In fact, taking our proprietary implementation of the Big-4 index, and comparing it to the last 8 expansions, shows just how meek this recovery has been:
There is one final interesting observation though – for the first time this expansion, the co-incident data is coinciding with the weak leading economic data made from 21 leading data series and first witnessed for February 2016. And that’s something worth pondering about.
Number of countries with back-to-back negative quarterly GDP prints is rising
The percentage of 41 OECD countries around the globe that have just posted a negative 1Q2016 on the back of a negative 4Q2015 (old fashion technical recession) has started to rise. Its nothing to be concerned about just yet but the rise itself, although shallow, is something worth watching as 2Q2016 numbers start coming out.
U.S Economy most vulnerable to any shock since 2008
The #Brexit vote caught the consensus view off-guard and stock markets, currencies and commodity prices have made large responses. This may be bringing up thoughts if Brexit could be the external shock that marks the decent of the U.S economy.
Whilst we will not entertain making predictions on this complex matter, we can however state that the U.S economy is the most vulnerable it has ever been since the 2008 financial crises. This is at least according to our composite of 21 leading indicators as at May 2016 shown below:
From the green diffusion line in the chart above we note that more than 50% (eleven) of these indicators are now in recession territory, with the overall (black) composite hovering near contraction (the red line). Eleven of the 21 indicators are long-leading and ten are short-to-medium leading. Six (60%) of the short-to-medium leading indicators are in recession and five (45%) of the long-leading indicators are in recession territory.
Visitations of the green diffusion index to the 50% mark and of the black composite to the red zero mark have been frequent in the past without leading to recession, but the fact of the matter is that these are periods of extreme vulnerability of the economy to internal or exogenous shocks or geopolitical events. With no margin of safety, or buffer, it doesn’t take a very large shock to tip many of the indicators past the point of no return, resulting in contagion to other indicators and finally the whole economy.
Recession Probability Roundup : Elevated levels
NOTE : This is a subscriber-only article that was made open for public viewership on 20 May 2016.
A probability of recession of 60% most certainly also implies probability of no recession of 40% and of course this is related directly to any false positives that may (and have) occurred in the past (see March 2003 in above chart.) You will note from the Headwinds chart on top of page 8 of the most recent SuperIndex report that we make a recession call only when the headwinds index exceeds 5, since this is the level that has produced the least false positives in the past. Only when this occurs, does the Headwinds index add one point to the Recession Forecast Ensemble (RFE)
Also, the Headwinds index, due to its stylized (as opposed to numeric) nature is very lumpy and thus recession probabilities can jump by large amounts moving from one headwinds level to the next. So whilst this model is showing elevated recession probabilities in 9-14 month time it does not constitute actionable information at this juncture.
The HeadWinds index is but one of the 6 models in the RFE and that is precisely why we designed the RFE – to allow for input of multiple robust diversified models into determining recession. Even if the Headwinds index was to move to 6 and the model were to call recession, the RFE would only rise by 1, which still does not constitute actionable information other than maybe some modest hedging or de-risking activity.
Another way to look at the RFE is to average the current recession probability showing on each of its six model components, which is currently showing a 14.6% probability of recession. This model appears to have served well in the past, with zero false positives above readings of 0.20.
It is therefore true that recession probabilities are elevated, the highest they have been since this expansion started in 2009. In fact, we have been witnessing this quite emphatically with non-RFE components, with both the Weekly Leading Economic Index (WLEI) and more importantly, the new and comprehensive US Long Leading Index.
But as the chart above shows, elevated probabilities only becomes concerning once the average RFE component recession probability is above 0.2.
It is our suspicion that the recession probabilities will abate in coming months, purely based on the recovery we are seeing in the WLEI and the March bounce we saw in the new Long leading Index report published last month to the USLLGI reports tab. This leading index has over 50% of its 21 indicators in recession territory (green line below).
Over 30% of States with rising unemployment
The useful thing with this breadth metric is that deterioration in unemployment is made visible long before it shows up in the average national unemployment rate. Whereas the national unemployment rate is at best a co-incident indicator for recession, the percentage of states with increasing employment acts as a reliable leading indicator – reliable enough to be added as one of the 21 components to our New Long leading Indicator.
The deterioration in this unemployment statistic is corroborated by several others we covered in a blog posting last week, titled “Labor market not as strong as you think“.
Labor market not as strong as you think
The strength of the labor market is constantly being trotted out in defense of the robust status of the US economy, but broad sets of labor data show this not to be the case.
This indicator needs to fall below -10 before the odds of recession skyrocket to a near certainty and so whilst there is no cause for immediate alarm, it is clear the indicator is not exactly shooting the lights out and has probably peaked. As it is a leading indicator, it is telling us that further gains in the labor market are likely to be muted going into Summer.
This indicator needs to fall below -5 before the odds of recession skyrocket to a near certainty and whilst there is no cause for immediate alarm, it is clear the indicator is not exactly shooting the lights out either and has definitely peaked. As it is a leading indicator, it is also telling us that further gains in the labor market are likely to be muted going into Summer.
The Conference Board also publish a “Help Wanted Online” (HWOL) index, that tracks Online advertised vacancies. Whilst the data has a much shorter history, it clearly peaked well before the 2008 Great Recession and has most certainly peaked in mid 2015:
This indicator has a long term history going back much further than that shown and is a reliable indicator of recession when it starts rising, although it is a co-incident and not a leading indicator. Whilst it appears to have leveled out recently, it certainly shows no cause for immediate alarm and cannot even be categorized as having peaked as with the other three indicators we showed.
Over 25% of the 52 US States now have rising unemployment, a record for the current expansion. Again, nothing to be alarmed about until this measure gets to over 50% (where it serves as a leading indicator of recession) but you have to agree it is far less flattering than the national unemployment rate measure would have you believe and has most certainly “peaked” in mid 2013.
So let’s have a bit of fun here, and create a “Super Composite” dynamic factor model combining all the Labor indicators discussed above bar the HWOL which has too short a history. So this is the 19-factor dynamic factor model plus the Conference Board ETI plus the U.S Civilian Unemployment rate plus the % of states with growing unemployment:
The above model has a very high 0.998 correlation (Area under the curve, or AUC) with US economic contractions and expansions when using -5 as the contraction/expansion trigger but is just as useful using the zero trigger level, which just happens to provide a 4-6 month lead to recession.
The bottom line is there are no alarm bells for recession according to the labor indicators just yet, but the labor market growth certainly looks to have peaked and is indicating far less strength than is commonly thought. In fact, leading indications are that labor market gains are likely to be muted going into Summer.
Animation : The incredible US employment recovery
Below is an animation of the annual average unemployment rate per U.S state from 2011 onward. It’s quite incredible to see how unemployment was erased state-by-state over the years:
However statewide improvements in employment have probably peaked-out as shown in the chart below, which depicts the aggregate (equal weighted) inverse 6-month unemployment rate growth for each of the 52 U.S States together with a diffusion showing the percentage of 52 U.S states with increasing unemployment. You can see that recently the diffusion has been rapidly rising showing that unemployment is starting to increase again in the 52 U.S states.
How can we forecast 30% upside for 2016 with RAVI?
The RecessionALERT Valuation Index (RAVI) is currently forecasting 30% growth for 2016 for the SP500 with its 1-year forecast model. “How the heck is this possible given current overvaluation of the market?” we can hear you say. Let us show you how this is calculated so we can put the forecast into context:
We can see that 10-year forecasts for the SP500 Total Return Index (including re-invested dividends) has a correlation of 0.92 to actual achievements. This is aptly shown by the chart on the right which plots the forecast (green line) and the actual achieved 10-year return (black line). Its important to note that the green line is seen 10-years before the black line!
The 10-year model is currently forecasting 121% growth from Dec 2015 to Dec 2025 or approximately 8.3% compound growth per annum. So one way we can forecast returns for the year ahead is to say that we are projected to grow 8.3% compound per annum so that means we can pencil in 8.3% growth for 2016. But this is a very broad brushstroke since we all know the market is highly unlikely to grow steadily by this amount for 10 years in a row.
The other way is to look at what the 10-year forecast was 9-years ago and then compare this to the actual growth we have achieved in the last 9 years and from this we should be able to work out how much upside is still left in the year ahead to make the 10-year forecast made 9 years ago true. In fact we can apply this logic to any look-ahead horizon:
Using the above table, if we set x to 9 we get the procedure used to forecast 1-year ahead returns. Similarly, setting x to 8 we can impute 2-year ahead returns.
Before we go ahead, let us just state that a correlation coefficient of 0.48 for a model forecasting 1-year ahead stock market returns is absolutely amazing, especially when considering that direct correlations between most commonly-used valuation metrics and 1-year Sp500 returns range from 0.1 to 0.15. But as you can see from the 1st chart above on the right, even this model can vary quite significantly from quarter to quarter (look at the vertical distances achieved between the green line and the black line). So you have to understand that these 1-year forecasts are rough guidelines and an achievement of 10% gains against a forecast of 30% gains is still “within target” of the model.
So let’s go through the logic in a real world, current example. At the end of December 2006, the RAVI model was forecasting 113.35% growth for the next 10 years to December 2016. As at the end of December 2015 the SP500 Total Return Index had grown 63.3% since this forecast was made. Using a mathematical formula we can thus deduce there must still be 30.6% upside left in the SP500 to bring the 10-year growth to 113.35%
The conclusion therefore, using this model, is that the stock market has not yet run way ahead of itself – contrary to what most people are trying to make you think.
The reason most people think the market has run ahead of itself is that everyone keeps trotting out valuation metrics like Shillers’ 10-year CAPE, Tobins’-Q, Market-Cap to GDP (Buffet Indicator) and so forth and these all look more or less the same and the messages being imparted are all more or less the same – the valuation metrics are at all time highs or way above long-run averages and therefore 10-year future returns are going to be poor and the market is in a dangerous period:
The fact of the matter is that these valuation metrics all have very little bearing on short-run returns and what is more important is:
1). What they were saying nine years ago and
2). How much the stock market has run since then.
This does not detract from the fact that the metrics are way above their long-run averages – they clearly are. But this is creating a problem for 10 years time, and not now.
Below is the best short-run correlation we could find directly from all the valuation metrics, obtained with the Market Cap to GDP ratio. As you can see, the correlation provided from using the metric directly is far short of the 0.48 achieved with the RAVI using the “look-back” methodology we described in this note.
Nonetheless, we now have several estimations we can use for 2016 returns for the SP500:
a). 8.3% if we annualize the 10-year forecast (2,213 target)
b). 12.4% if we annualize the 5-year RAVI forecast (2,297 target) <-most likely scenario
c). 20.7% if we annualize the 3-year RAVI forecast (2,467 target)
d). 27% if we annualize the 2-year RAVI forecast (2,595 target)
e). 30.6% if we use the 1-year RAVI forecast (2,669 target)
We tend to ignore the 1-year RAVI (due to its wide fluctuations) and use the low, average and high from (b),(c) and (d) above for 2016 targets on our daily updated dashboard:
Bear in mind that the 3 radial gauges change on a daily basis depending on the current level of the SP-500 and how far it needs to go to reach the LOW,AVG and HIGH targets.
Risk of U.S Economic Recession
A raft of analysts, perma-bears and bloggers are playing fast and loose with the R-word again. This is likely to reach a crescendo with the release today of the unexpected large drop in the ISM non-manufacturing survey. We recall a time in late 2011 when the mainstream perception was that we were headed for recession and we posted a widely read article that went against the mainstream, and attracted attention of some respected names:
Given the amount of attention looming recession calls are getting, we thought it prudent to throw our observations into the hat.
The main reasons for most parties claiming we are in recession are the SP-500 earnings recession, stock markets, the widening credit spreads, Weekly Leading Indicators and Industrial production/Manufacturing. Next week, see the articles about the ISM non-manufacturing PMI crawl out the woodwork. Let’s look at each of these in turn.
The SP-500 earnings are indeed in recession and this is normally a strong predictor of economic recession and bear markets:
However, if you dig under the hood, the earnings recession is mostly a function of the hammering taken by the Energy and the Materials sectors. It is debatable if these sectors make up enough of the economy to drag it into recession:
The stock markets are indeed pricing in recession, as shown with the Stock Market Composite of economically sensitive stocks which is one of the components of our Weekly Leading Economic Index (WLEI):
However, we all know the story about the stock market predicting 12 of the last 6 recessions with the false alarms shown above.
Third up is a favorite of the bears, namely widening Credit Spreads. Whilst corporate bond spreads (WAAA/WBBB, -WBBB and T-Bill/BBB) are indeed flagging emphatic recession (bottom chart in the pair below), when one views a very broad basket of over 35 credit spreads (top chart below), the picture is far less bearish (albeit at concerning levels). The top chart is updated weekly and is also a component of our Weekly Leading Economic Index (WLEI)
The widely followed ECRI Weekly Leading Index also is flagging recession, but it has done so three times already this expansion. Whilst our own WLEI has not flagged an alert before 21/08/2015, it now joins the ECRI WLI in an emphatic recession call.
Weekly leading indicators are much noisier than monthly ones so one would observe several weeks of recession signal before reading anything into it. This wait period has passed and so we need to determine the possibilities. Our Weekly Index, which is published for subscribers every Thursday evening, consists of 5 groups of composites covering corporate & treasury bonds, stock markets, employment and credit spreads and is made up of just over 100 weekly time series:
The WLEI is negative now because of the heavy weighting (60%) made up of the Corporate Bonds composite, the Treasury/Bond composite and the Stock Market composite. As we have warned subscribers in the past, we do not believe the 5 major categories above may be broad enough to properly encapsulate all manner of recessions and we are on the hunt for a sixth that also only contains weekly data of a short-leading nature (we do not believe in mixing long leading indicators with a short leading system nor mixing monthly time series with weekly ones.)
Also note from the first chart above that there have been 3 false alarms since 1973 with the depth of the current readings consistent with the 1984 false alarm. Thus, whilst the WLEI flags caution we do not use it for performing actual stock market actions at this point.
The fifth item hauled out the bag to argue for recession is Industrial Production which is indeed in recession. Unlike past declines in industrial production, the current decline has been driven primarily by the collapse in the utilities industry (weather) and mining (impacted by energy) which make up just 25% of Industrial Production.
During 1985/6, declining oil prices and a rapidly rising dollar also led to an Industrial Production slump. This did not turn into a recession. There is a great piece on this here.
Now we move onto Manufacturing which makes up about 70% of Industrial Production and 12% of US GDP. The output from U.S. factories has been little changed recently, although the main manufacturing indicator being watched, the ISM PMI composite, is indeed below the expansion level of 50. In the last few decades, it was not unusual for the ISM to signal recession as the broader economy remained in expansion.
The real rebuttal against the ISM Manufacturing PMI warning is highlighted by Tom Porcelli at RBC Capital Markets who stated “One of the reasons to believe the so-called manufacturing ‘recession’ is likely to be short-lived is that it has been limited to a handful of industries. In other words, it has not been a pervasive slowing.” There is nice coverage on this at Business Insider.
Lastly, lets cover the latest candidate to be trotted out in ensuing days. Non-manufacturing (a large chunk of the economy) is weakening, with an unexpected downturn on the latest reading, the biggest one-month decline since the recession. But it will require another two such drops to get to contraction territory:
We know that manufacturing is about 12% of the US economy (red line below). We also know that the manufacturing and the non-manufacturing sectors together make up 90% of the US economy. This means non-manufacturing (green line below) makes up 78% of the economy. We can combine these two according to their relative weights to give a representation of 90% of the US economy as at January 2016:
There was a disconcerting plunge in January, but no recession signal yet.
Now lets focus on the arguments AGAINST recession.
Secondly, there are precious few long-leading indicators flagging recession. Until we find long leading data (except of course corporate profits and corporate bond spreads) that concurs with future recession, we will have difficulty raising red flags on the economy. Our own long-leading indicator, USLLGI, tracks the growth of 8 reliable indicators which have consistently peaked 12-18 months before the onset of NBER defined recessions since the early 1950’s :
The National Buro for Economic Research (NBER) are the final arbiters of recession dating in the U.S. The NBER does not define a recession in terms of two consecutive quarters of decline in real GDP. Rather, a recession is a significant decline in economic activity spread across the economy, lasting more than a few months, normally visible in real GDP, real income, employment, industrial production, and wholesale-retail sales.
They will be examining 4 monthly co-incident indicators:
- Industrial Production
- Real personal income less transfers deflated by personal consumption expenditure
- Non-farm payrolls
- Real retail sales deflated by consumer price index
We used the above 4 indicators to create a “NBER Big-Four” composite index many moons ago. It is also showing disconcerting sluggishness, but no immediate signs of recession:
There are however some short-leading indicators not often mentioned that warrant concern. Here are some taken from our Monthly Long Leading Composite Index:
To summarize, despite many pockets of disconcerting co-incident indicator weakness, many brought on by the energy and commodities complex, the risk of near-term recession appears low. Although many short-leading indicators are showing recession there are not enough of them yet to warrant a confident recession call. What remains highly elusive is finding long-leading data that points to recession.
Recession odds are rising, as is commentary surrounding recession. Most of this commentary centers around “cherry picking” this or that indicator or sector to bolster an argument but rarely covers broad-based assessments. One thing is for certain – watching co-incident, short and long leading data closely is going to be back in vogue over the coming months:
Just like 2011/12 it is going to make for some interesting times.
Yellen Labor Dashboard reaches target
“Full Employment” target reached
Shortly after taking up office, Federal Reserve Chair Janet Yellen used her “jobs data dashboard” to justify the Fed’s easy money policies and to argue there’s still considerable slack in the labor market five years after the recession’s end. Seven of the nine gauges on this dashboard had not recovered to levels reached before the last recession, reinforcing her belief that the economy would need “extraordinary support” from the Federal Reserve for “some time to come.” It appeared at the time that the Federal Reserve had changed emphasis from their singular focus on the unemployment rate to a broader labor market approach, which was the right thing to do. The Federal Reserve went so far as to state that policy would remain accommodative well beyond the prior target threshold of 6.5% for the unemployment rate since the new guidance objectives are “Full Employment” and “2 % inflation.”
Whilst we have been regularly publishing a Labor Market Report for clients since October 2012 to gauge exactly what we thought the Federal Reserve ought to be looking at, the collection of indicators in Yellen’s dashboard had some interesting additions to our one. More specifically we felt there was a risk that several of the components she had chosen may paint a picture of more slack in the market than realistically existed, but maybe its was a good thing to look at the worst case scenario. Nonetheless, since we now had specific details on exactly which nine labor components the Federal Reserve would be examining on a monthly basis, we felt it would be prudent for our clients to keep their eyes on the exact same things the Federal Reserve was using to gauge the conditions of the economy.
We compiled a “Yellen Labor Dashboard” which we have been including as an addition to our standard monthly Labor Market Report for our subscribers. All the indices have been re-based to zero at the peak of the last business cycle and normalized so that they touch -100 at their worst point reached since then. About 12 months ago we began projecting when the labor composite would reach the prior business cycle peak (“full employment”) to give our clients an idea of when we should be expecting the Fed to be compelled to raise interest rates (assuming they stuck to their story-book, of course). This projection commenced around 1Q2015 and fluctuated slightly with each passing month, slowly rising to late 2Q2015. About 6 months ago it extended out to 3Q2015 as the rise of the labor composite waned, and about 3 months ago we began pinning this down to the end of September 2015 (when the July print of the index would reach its prior peak.)
We have updated this index with the relevant labor data today and can announce that the Yellen Labor Dashboard has now reached and surpassed the peak achieved in the prior business cycle, one month ahead of schedule (August), just in time for the Fed meeting in September:
Although the equally weighted Labor composite and 4 of its components have exceeded their prior peaks, we still have 5 components below (with the participation rate being well below due to certain demographic structural factors.) As we stated some 12 months ago, we fully expected the Labor composite to reach the prior peak sans 4 or 5 of its underlying components, but that we did not think that would have any bearing on the “full employment” status of this achievement.
So there is finally a tick in the box for the employment situation. The only missing piece is the 2% inflation target. If the Fed elects not to raise rates in their September meeting then the inflation outlook or other external global factors will likely be held responsible, not the labor market.
Things that go bump in the night
The U.S economy appears unstoppable right now. Just about every leading and co-incident indicator you can think of is pointing to positive growth. Among the hundreds of indicators we follow for our models on a daily basis, we have discovered a few that are displaying worrying trends and flagging a future recession. It should be pointed out that a handful of indicators flagging recession should not mean we have to push the panic button. A large raft of indicators all concurring is what you should be looking for, and right now a large raft of indicators is confirming economic growth. Nonetheless it is interesting to observe, in a plethora of positive economic indicators, those which are saying otherwise.
1. Corporate Profits
Corporate profits with inventory valuation adjustment (IVA) and capital consumption adjustment (CCA) relative to GDP has peaked over a year ago which is normally a reliable long leading indicator for economic recession.
2. Inventory to Sales Ratio
Inventories outpaced sales in a dramatic fashion in the last few month with inverted year-on-year growth plunging below zero which is normally also a long leading indicator of recession. An increase in inventory to sales ratio from one month to the next indicates that either inventory is growing more rapidly than sales or that sales are dropping. No matter which situation is causing the problem, an increase in the inventory to sales ratio signals an oncoming cash flow problem for corporate America:
5. Elevated Financial Stress levels
This composite, discussed over here represents a vast, broad array of financial metrics that measure financial stress. It is the broadness of this indicator that makes its rise particularly worrisome since in effect we have a large swathe of indicators more or less in consensus:
7. The ECRI Weekly Leading Index (WLI)
For the third time in this economic expansion, the ECRI WLIg growth metric is flagging an emphatic recession. This is most likely due to this indicator having a heavy commodities weighting. Whilst this bears careful observation, we also track our own weekly published leading indicator, the WLEI, which consists of over 50 weekly published time series. Whilst this is not flagging outright recession, it certainly has put in an emphatic peak.
8. Valuations are becoming stretched
Note the emphasis on “becoming”. As we show here, we do not believe the stock market is as overvalued as it currently would have you believe, but its certainly on its way there with a 5-year forecast of 52.2% or 8.8% compound growth upside. As you can see below, the 5-year forecast trend has peaked and is all downhill from here.
SP-500 Returns Forecast, 4Q2014
The RecessionALERT Valuation Index (RAVI) is a multifactor valuation model that examines cyclically adjusted trailing SP-500 earnings (various multi-decade horizons), the SP-500 total-return index level, total stock market capitalization, Gross Domestic Product, non-financial corporate equities and liabilities, non-financial corporate business net-worth and percentage of investors’ allocation to stocks versus cash and bonds to determine 10, 5, 3, 2 and 1 year forecasts for the SP-500 Total Return Index (dividends re-invested). The in-sample accuracy of the various forecast horizons since 1970 are shown below:
The RAVI model is updated quarterly and lags by one quarter. The data for the 4th quarter of 2014 (as at end of December) was released on 13 March 2015. Let us examine what it is telling us about ^SPXTR (SP500 Total Return Index) returns across the 2,3 and 5 year forecast horizons (we ignore the 1-year forecast due to the lower r-squared predictive ability)
The 2-year (8 quarter) forecast appears below. The first chart on the left shows the accuracy of the RAVI model since 1978 with the x-axis representing the model forecast and the y-axis representing the actual achieved ^SPXTR 2-year returns from the forecast date. An r-squared of 0.69 might not appear much, but considering the short forecast horizon, it is actually respectable as far as forecasting models go. In fact, the r-squared correlation over the last 10 years since 2005 has actually tightened to a respectable 0.79. The model is forecasting returns of 13.7% from end December 2014 through to end December 2016 which equates to 6.6% compound growth per annum.
The 3-year (12 quarter) forecast appears below. The first chart on the left shows the accuracy of the RAVI model since 1977 with the x-axis representing the model forecast and the y-axis representing the actual achieved ^SPXTR 3-year returns from the forecast date. The longer horizon yields a much higher correlation as is to be expected. An r-squared of 0.79 is respectable considering the short forecast horizon, and the correlation has actually improved to 0.86 over the last 10 years since 2005. The model is forecasting returns of 46.1% from end December 2014 through to end December 2017 which equates to 13.5% compound growth per annum.
The 5-year (20 quarter) forecast appears below. The first chart on the left shows the accuracy of the RAVI model since 1975 with the x-axis representing the model forecast and the y-axis representing the actual achieved ^SPXTR 5-year returns from the forecast date. The longer horizon yields a much higher correlation as is to be expected. An r-squared of 0.88 is pretty respectable considering the short forecast horizon, and the correlation has actually improved to an astonishing 0.96 over the last 10 years since 2005. The model is forecasting returns of 52.2% from end December 2014 through to end December 2019 which equates to 8.8% compound growth per annum.
BOTTOM LINE : If we were to ignore the ^SPXTR dividend component in the annual growth estimates and apply them directly to the SP-500 Index then this implies a growth range average of (6.6% + 13.5% + 8.8%)/3 = 9.63% for the SP-500. This means the SP-500 would need to rise from 2,059 as of 31 Dec 2014 to 2,257 as of 31 Dec 2015. This is slightly higher than the 2,246 target we penned in our 3Q2014 Forecast in December.
Bear-market and Recession Indicator
There is an interesting “by-product” of the RAVI as a bear-market and U.S recession indicator. Since the 1-year forecast “looks ahead” 4 quarters, we can average it by 4 quarters to attempt to obtain a “co-incident” view which warns of bear markets upon dipping below zero. We apply the same logic to the 8-quarter average of the 2-year forecast and the 12-quarter average of the 3-year forecast. Indeed, they all provide similar co-incident warnings of stock market corrections and 2-3 quarter warnings to recessions as displayed below. It is interesting to note that the rise of these “smoothed” indicators above zero also provide for good stock market re-entry signals:
BOTTOM LINE : “Whilst the smoothed forecasts that produce these signals have been weakening over the last few quarters and are no “pillars of strength” right now, they are not warning of any stock market corrections or recessions on the immediate horizon. In other words, the stock market is a lot less over-valued than various commentary on the Internet would lead you to think.”
It is intriguing that a valuation model can serve as stock market timing signal or recession indicator so this is certainly one new tool our clients will be monitoring closely going forward. It would be unwise to rely entirely upon this mechanism to attempt to gauge stock market risk and we recommend using it together with our existing Composite Market Health Index (CMHI) in the foreseeable future.
You should need no reminding that these are forecasts based on models utilising in-sample data and whilst our own research of the efficacy of the methodology in both robust in-sample and out-of-sample testing is highly encouraging, we should always bear in mind black-swans and geo-political events can always throw a spanner in the works and remind us that past performance is no guarantee for future returns. Also, shorter term forecasting is subject to more quarter-to-quarter variance than longer-term forecasting, despite perceived long-term correlation levels. This is especially the case when dealing with single-digit growth forecasts in above-average valuation scenarios such as we find ourselves in now when a 2-3% variance from expected outcomes actually makes a large difference. At the very least though, you can make estimates with some sound statistical reasoning behind them, since at the end of the day one always needs to be dealing with calculated risks in investing.
RecessionALERT subscribers can obtain quarterly updates of all the above charts in the CHARTS>MACRO>RAVI menu tab. The next quarterly update representing 1Q2015 is on 17 June 2015 and will be followed by a repeat of this analysis for subscribers. The 1,2,3 and 5-year forecasts also appear on the CHARTS>HOME>DASHBOARD tab, to the top-right of the right-hand panel as shown below, where forecasts can be viewed in real-time target (upside remaining), annualized or absolute terms (just click on the tab you desire.)
Clicking on the small tab with the “?” will display the RAVI Gauges User-Guide to assist you with interpretation of the various tabs.
Stresses are building up in the system
Despite a steady rise in the SP-500 Index, clear and persistent financial stresses are starting to build up in the system. We construct a composite of the St. Louis Fed Financial Stress Index, the Cleveland Financial Stress Index, the Kansas City Financial Stress Index and the Chicago Fed National Financial Conditions Index as shown below:
The average value of the composite, which begins in 1973, is designed to be zero. Thus, zero is viewed as representing normal financial market conditions. Values below zero suggest below-average financial market stress, while values above zero suggest above-average financial market stress. As a measure of financial stability, the composite exhibits two essential characteristics. Firstly, known periods of financial crisis correspond closely with peak periods of tightness in the composite, and secondly, the turning points of the composite coincide with well-known events in U.S. financial history. It is also useful to assess the current level of financial stress compared to the composites’ value during the past.
It is evident that the rise of the Financial Stress Composite above zero is a long-leading indicator for economic recession, but it also signifies a period of stock market volatility punctuated by frequent incidences of nasty stock market corrections. It is also clear that sustained periods when the Financial Stress Composite is below zero are mostly accompanied by low volatility U.S stock market “melt-ups”, with the recent 2012 to 2015 melt-up accompanied by historically low financial stress, being no exception.
It is probably premature to hit the panic button now. Also, a rise of the Financial Stress Composite above zero does not necessarily mean stock market gains are not possible, it just means that those gains are most likely going to come attached with considerable volatility, against a background of heightened risk for a non-trivial correction.
Of course, no stock market actions should be premised on readings of a single indicator alone (such as this one), and for this reason we should continue to examine readings on our Monthly Leading Economic Index, Long Leading Economic Index, Stock Market Health Composite, Recession Forecast Ensemble and our RecessionALERT Stock Market Valuation Index.
For the moment, the daily updated Stock Market Health Index available in your CHARTS menu is showing no cause for alarm, but most certainly looks like it is forming a top. The Diffusion is maxed-out with all six components in bull-market territory, but clearly technical price momentum, percentage of stocks in new long term rising trends, net advancing volume and weekly economic data have been fading for the last few months.
We are fortunate to have the tailwind of very benign seasonality at least until mid-April 2015, but it remains to be seen how all the remaining five stock market health components are going to be positioned to be able to endure the seasonality headwinds after this. If we find ourselves with five weak components and negative seasonality accompanied by an elevated Financial Stress Composite then serious consideration of some downside protection may be warranted. The timing of such protection measures could then be aided by the Fingerprints of a Market Top
The St. Louis Fed Financial Stress Index measures the degree of financial stress in the markets and is constructed from 18 weekly data series: seven interest rate series, six yield spreads and five other indicators. Each of these variables captures some aspect of financial stress.
The Cleveland Financial Stress Index is designed to track distress in the US financial system on a continuous basis giving the financial-system supervisors the ability to monitor stressful episodes as they are building in the economy. Such early detection is important because financial stress can quickly be amplified when stress is occurring in more than one market. The CFSI tracks stress in six types of markets: credit markets, equity markets, foreign exchange markets, funding markets (interbank markets), real estate markets, and securitization markets.
The Kansas City Financial Stress Index is a monthly measure of stress in the U.S. financial system based on 11 financial market variables.
The Chicago Fed National Financial Conditions Index measures over one hundred individual indicators for risk, liquidity and leverage in money markets, debt and equity markets and in the traditional and “shadow” banking systems.