Thursday, November 9th, 2017

The Only Game in Town

Last week we argued that the Chinese Technology sector was growing a bubble and that there are obvious parallels between its behaviour now and that of the US in the run-up to the dot-com bubble. This week we highlight the extent to which the Technology sector is becoming the only game in town for global investors. We don’t have space to discuss the lack of opportunities for country diversification within equities or carry trades within fixed income. Suffice it to say there aren’t many, and many of those which do exist are vulnerable to currency volatility. If CIOs wish to generate some alpha for their clients, they have to get their equity sector calls right at a regional and global level.

Until recently this wasn’t very difficult: a big underweight in Energy and Telecom everywhere and a big overweight in Technology did a lot of the work. Add in some regional overweights in Small/Mid-Caps in Europe and Financials in the US and the UK and this mix generated some consistent and healthy outperformance. The recommended active weight for all regions was at the top of its normal range during Q1 2017 and at a sector level it was common to find that an overweight in one region was offset by an underweight in another region (e.g. Healthcare in the US and Europe).

As of Q4 2017, two things have changed. The recommended active weight of the global equity portfolio has fallen to bottom of its normal range, and the recommended exposure to Tech has risen substantially. This is turn breaks down into two parts: an increase in exposure in regions where we were already overweight and an upgrade to overweight in regions where we were previously neutral. We are effectively at maximum overweight in the US, China and Japan and we are now at 50-60% overweight across Europe. Technology is ranked #1 in every region, apart from the Eurozone, where it is ranked #2.

At this point it is worth saying that our models reflect the cumulative decisions of investors everywhere. If they had done something else – as opposed to thinking or saying it – our models would look different. The models do not have a valuation term, but it is perfectly rational for an individual investor to believe that the Technology sector is overvalued and that it will become more overvalued. He will be happy to hold his position so long as he believes that his risk of loss is not increasing. At some point these rational decisions will become cumulatively irrational, and one way of identifying this moment is to observe the relationship between the excess risk and excess return of the sector versus the index.

On a global basis, Technology now accounts for 17% of the active weight of the portfolio. In the US and Japan, this figure is over 20% and in China it is 26%. The highest ever reading for any sector across all regions was 22% for Tech in 1999 and 2000. A reading above 15% is in the top quartile for any sector that has been ranked #1 going back over 20 years and the current reading for Tech is in the 90th percentile. We don’t want to bully our readers with statistics, but we need to emphasize that a lot of investors have already concentrated their sector bets on Technology.

We also think that the current concentration on Tech is likely to increase. The sector crossed into the top quartile in mid-October. Sometimes a sector will reach this level and retreat almost immediately. A typical run is between 6-10 weeks. The current run for Technology is 4 weeks, but in 1999/2000, the sector has a run which lasted 33 weeks. Small/Mid-Caps had a run of 32 weeks in 2004, and Staples had a run of 26 weeks in 2015. We cannot be sure, but given that its recommended weight is still increasing in Europe, there is no signal which suggests that level of concentration is likely to start falling in the near future.

If we don’t know how long it will last, we certainly cannot forecast how it will end. However, most times, when a sector has an extended run at the top, it is ranked in the bottom three on a global basis within a year of ceasing to be ranked #1. The only exception is Small/Mid-Caps in 2004, which went on to be #1 again for a period in 2005, but the recommended position was not as high a percentage of active weight, the duration of the run was much shorter, and it was ranked in the bottom three within six months.

Conclusions: (1) Everyone likes Tech at the moment. (2) It is the dominant sector in all regions apart from Europe, where it will probably get to this level soon. (3) The level of dominance in China is getting dangerous. (4) On a global basis, other sectors and Tech in 1999/2000 have enjoyed dominant episodes lasting up to 33 weeks. (5) Once this episode is over a period of sustained underperformance normally follows within a year. (6) Assuming that the current episode lasts as long as the previous Tech boom, the best we can hope for is that there is no significant deterioration until Q2 2018.

Leave a Reply

Wednesday, November 1st, 2017

The Days of Future Past

Last week, the FT ran a chart comparing the recent performance of the US Tech sector vs the US index to the period before and after the Tech bubble. It concluded that there were no similarities between the two periods: the relative outperformance was a lot smaller and there were solid earnings underpinning the valuation. The relevant quote is, “the current rally in tech, while remarkable, is a very different animal from the dotcom bubble.”.

We are never fans of the idea that it is “different this time”, but we agree on the narrow point regarding the US in the two different eras. And yet there are issues which are starting to concern us. The first is that since October the sector has been ranked #1 or #2 in all six of our regional equity sector models. It is very rare for any sector to be this dominant. Examples of extended runs include Energy from July to October 2004, Consumer Staples from October 2012 to February 2013, Financials from December 2005 to March 2006, and of course Technology from July 1999 to May 2000. We never know in advance how long a run is going to last, but anything more than four months is a long time, and the longer the run, the harder the subsequent fall. We don’t worry specifically about the Tech sector in the US, we worry about Technology everywhere.

Our second point is a subset of the first. The Chinese Technology sector relative to the Chinese index does look like the US Tech sector, just before it went exponential in Q3 1999. This is not yet as widely known as it should be because there is no single index which captures all of the China-centric Technology businesses in one place. Some of the majors, like Alibaba and Tencent have their primary listing in Hong Kong or the US and there are at least three onshore equity indices (CSI, Shenzhen and Shanghai) to which the sector could be compared. Our data is sourced from the FTSE All World, and all we can do is hope that it is as representative as any other data set which investors use.

Our comparison chart cannot be used to predict when the Chinese Tech sector is likely to go exponential or how high it would rise if it did. All we can say is that is has happened before in another country, when the global growth outlook was similar and when the sector was strong in every region. We also see parallels in the IPO market. The FT reports that there is a wave of Tech offerings which have been incubated by Baidu and Alibaba and others. This looks like the way in which Telecom companies used to float stakes in their mobile subsidiaries in the late 1990’s. It cannot be long before one group of analysts publishes research saying that a new IPO deserves a premium because it is backed by Alibaba, while another group argues that Alibaba’s business is only valued at a p/e of x, once the value of its quoted and unquoted investment portfolio is stripped out of the market cap.

In a growth sector like Technology, there will never be a stable consensus around how companies should be valued. That is the beauty of Harlyn’s approach. When the Chinese Tech sector stops producing an attractive return per unit of risk, it will start to underperform. We can observe this in real-time, and warn our clients accordingly. It hasn’t happened yet, but it will.

Leave a Reply

Thursday, October 26th, 2017

No Yellow Flags

We are marking the 30th anniversary of the 1987 crash by trawling through all the recent bear markets in the US and elsewhere to see if we can find any patterns which might help us to identify the onset of the next one. We found that a rise in excess volatility (equities volatility minus bond volatility) was a good indicator of an imminent bear market in the US in 2002 and 2007 and that it had a hit rate of 62% in non-US markets. The misses occurred when investors initially thought there would be little contagion in their domestic market from events in the rest of the world and then found out that they were wrong. This week we return to the US to see if the excess volatility test can be applied at the sector level.

The first lesson is that bear markets are much more complex affairs than they appear from just looking at index-level data. For the US as a whole there was no bear market in 1998, but there were six sectors which declined by 20% or more from a peak occurring sometime between April and December 1998. In addition, Healthcare had its own private bear market starting in April 1999. The same is true for 2011 and 2014/15 – no overall bear market but two sectors in each case which fell by 20% or more.

The second lesson is that even when there is a general bear market, individual sectors peak at very different times. In the 1999-2002 bear market, Financials peaked in April 1999, but Consumer Staples did not peak until May 2002. The peaks of the other sectors were spread evenly across the intervening time period. In 2007-08, Consumer Discretionary and Financials peaked in June 2007, but Energy and Materials did not peak until May 2008 and Staples not till September. Only with hindsight do we know that these developed into general bear markets, unlike 1998.

The third lesson is that there is no prescribed order in which sectors peak and then fall into bear territory. In the 2000 and 2007 Financials were #1 and #2, and Staples were #11 both times, but the overall correlation of the sector ranking is weak and it is non-existent for 1998 and 2011/14. Nor is there any link between the order of the peaks and the severity of the drawdown in the first year or in total. None of this should be a surprise, but it reinforces the lesson that every bear market is different.

The important point is that an increase in excess volatility works as an indicator of a sector-specific bear market, provided we look at those which peak early. This is when investors think they are dealing with a sector specific problem (or a series of them) as was the case in 1998 and 2011/14 but before the dynamics of contagion take over. By definition, this comes as a shock to the sectors which were not originally affected – just as it did to the international equity markets discussed above. This is the moment when excess volatility for individual sectors is no longer useful and we have to look at the index as a whole.

As far as the current situation is concerned, only Telecom has experienced a significant increase in excess volatility – up from 4% to 14% – since July. It is possible that this sector could fall into its own private bear market, but so far there is no sign of similar behaviour in, or contagion to, any other US sector, and that is good news.

Leave a Reply

Wednesday, October 18th, 2017

International Bears

Last week we asked how investors can determine whether a correction in the US equity market (-10% decline) is likely to turn into a full-scale bear market (-20% or more). We found that the behaviour of excess volatility (equity volatility minus bond volatility) had been an important indicator in advance of the 2000 and 2007 bear markets. It did not work in 1987, because there was no period when there was a correction, but not a bear market.

We cannot be precise on timing, but we found that there were material increases in excess volatility just before equities peaked and again as the correction turned into a full bear market. In each case the increase was at least 10 percentage points, and in the September and October 2008 it was almost 30 points. What mattered was the increase – the reassessment of the relative riskiness of equities and bonds – not the opening or the closing level.

At present this is a small sample and we are keen to extend the scope of this work. The problem is that there were not many bear markets in the US in the 1980s or 1990s. The LTCM crisis doesn’t conform to our definition and the one starting in November 1980 is best examined in the context the high inflation era of the 1970s, rather than the world of the Greenspan and Bernanke put. It’s only when we look at the US index in real, not nominal terms, that we understand just how serious the 1974, 1976 and 1980 bear markets were. However, it would be surprising if a measure of nominal volatility was a good indicator of a drawdown which had to be measured in inflation-adjusted terms. We will go back further into the low inflation era of the 1950’s and early 1960s, but we need to do some due diligence on the data before publishing our results.

In the meantime, we can look at non-US markets to see whether we pick up the same pattern. We look at three bear markets, 2000, 2007 and 2015. The last is the most interesting because it does not coincide with a bear market in the US, (-11% at the trough). It is important to understand that 2000 and 2007 bear markets in these countries do not necessarily start at the same time as the US. For instance, in Canada the GFC bear market did not start until June 2008, unlike Europe where it started in late 2007. In Japan both the 2000 and 2007 bear markets started several weeks before the US. In all cases we take our dates from the local not the US equity market. Our findings are as follows:

Japan is a text book example: in all three episodes there is an increase in excess volatility in the run up to or shortly after the onset of the first correction. Excess volatility then increases or stays at an elevated level as the correction turns into a bear market. Canada also works. In 2000 and 2015, there was a 10-point increase as the correction got underway, but in 2008 it only happened as the main bear market struck.

For the Eurozone, and the two main markets within it, the evidence is a bit more mixed. In 2000, France and the Eurozone saw a rise of just under 10 points but excess volatility in Germany hardly moved. Then, in late 2001 as the geo-political crisis deepened, excess volatility in Germany and the Eurozone soared by 15 points, but only 8 points in France. In 2007, there was no material increase anywhere in the Eurozone, followed by a surge of 30 points as the GFC struck, but this was a lagging not a leading indicator. By contrast, 2015 saw an 8-10 point increase ahead of the correction, with further 5-10 points as the bear market took hold.

For the other Anglo-Saxon markets, UK and Australia, the evidence is decidedly mixed. Neither saw any surge in excess volatility ahead of the 2000 correction, but both reacted after 9/11. In 2007, there was no significant increase until the GFC, but in 2015 both had an increase ahead of the correction and another at the onset of the bear market.

If we use a very crude system, scoring one point every time excess volatility acted as a lead indicator, Japan and Canada both score 3/3, UK and Australia both score 1/3, and for Eurozone, France and Germany the scores are 2/3 for 2000, 0/3 for 2007 and 3/3 for 2015. The total score is 13/21, or a hit-rate of 62%, which is good, but not spectacular.

Most of the misses can be explained using the same logic as for the US in 1987: investors did not think that what was happening elsewhere in the world would have an impact on their domestic market. Thus, the consensus in Europe, the UK and Australia did not think they would be affected by a US mortgage crisis until Lehmans went down.

This allows us to argue that a rise of about 10 points in excess volatility is a good indicator of an upcoming correction or a transformation to a full-scale bear market. It works best when the cause of the problem is primarily domestic and it can fail to move when investors do not understand the transmission mechanism from one region to another. However, whenever we see an increase of this magnitude, we are on notice that a significant and tradeable correction may be in the offing.

The fact that we haven’t seen anything like this so far in 2017 is one of the main reasons why we are overweight equities in all the main developed markets.

Leave a Reply


    Full Blog Archive:


  • Search Blogs by Category