Wednesday, February 21st, 2018

A Warning from Japan

Over the last two weeks, connoisseurs of financial journalism have feasted on some of their favourite stock articles ranging from “I told you so” to “It’s just a blip” and “Don’t do anything when markets are like this”. It is the last one which interests us. One of the key messages from our models since the sell-off began has been the almost total lack of differentiation between equity regions or equity sectors. This makes the few remaining cases of differentiation more important than usual. One of these is Japan.

Before the sell-off, Japan had been top of our regional equity allocation for 13-16 weeks, depending whether we looked at the euro or US dollar denominated model. This is by no means the longest run for a region in the #1 position, but it is longer than average, and we find that a major shift between risk-on and risk-off, however temporary, normally implies a change in regional leadership as and when the model moves back to risk-on.

In the real world, we note that the yen has strengthened against the dollar and broken up through a 5-year trend line. It could be on course to test 100, a level which would challenge the investment thesis of many international investors who have been long the local index and hedged on the currency. This may just lead to another bout of intervention by the authorities, providing some short-term relief, but it is hard to see it as a game changer versus other regions.

We are still underweight in the Eurozone, but we have started to reduce this and the process may accelerate if the Italian elections on March 4th produce a market-friendly result. We have recently upgraded our stance on Emerging Markets (mainly Asia) to reflect faster global growth, and we may have to increase the UK if Brexit fails to turn into the catastrophe that markets seem to expect. All of this has to be funded from somewhere.
Some it can come out of US equities, but realistically only from international investors reallocating within their non-domestic portfolios. US-based investors are too busy enjoying the windfall gains in US equities. Some of it could come from fixed income, but the surge in volatility has reminded every institutional investor of the importance of risk budgets. To the extent that Japan benefits from the haven trade, this will flow into fixed income not equities. However one looks at this, Japanese equities are more likely to be a source, rather than a destination, of funds over the next three months.

So far, we have not had confirmation of this move, but there is one important detail which is consistent with it. Our Japanese equity sector model has begun to reduce its underweight in defensives, in contrast to every other developed market. The contrast with the Eurozone is particularly clear. From January 2017 onwards, the defensive underweights were virtually identical. Since the start of February there has been a significant divergence. Yet both regions have a strong currency and central banks which say they are prepared to add stimulus if required. We only have three weeks’ data, but they suggest that investors are preparing to switch out of Japan and into the Eurozone. Unless something dramatic happens, we expect to downgrade Japanese Equities to neutral in the near future.

Leave a Reply

Thursday, February 15th, 2018

Shooting Bad Assets

It is not hard to tell that we are in the late part of the financial cycle. Investors have started to get out of speculative assets and bad investment strategies which no longer make sense. It’s a bit like a hostage scene from a Hollywood B-movie, where the terrorists take a couple of victims out of the room and shoot them noisily just to prove that they are serious. So far this year, the markets have shot Bitcoin and short-volatility ETFs. There will be other victims, the question is which?

Our list includes the usual suspects, such as the Tech sector in China, but others may be surprising. We start with the global “risk-free” asset itself: the US 7-10 year Treasury bond. Of course, this is not risk-free, because it is always exposed to the risk of rising yields, but this hasn’t mattered much for the last 30 years. Now it does. If we restate our usual formula to measure the risk-efficiency of Treasuries vs cash, they need to produce a return which is greater than the return on cash (i.e. interest rates) plus their excess volatility. There are many ways of showing this relationship – e.g. the Sharpe ratio – but we divide the difference between the required and the actual return by the annualised volatility.

For most of the last two years, this gap has been negative for US Treasuries. It is now 4.4 times the annualised volatility of returns, which means that they would have to stage a very large recovery just to get back to neutral. A low-volatility bear market is a frightening beast. Treasuries are losing money slowly and consistently. Even if there is a rally, it probably won’t be enough to take their returns back into risk-efficient territory. In theory, an increase in volatility would reduce the distance (measured in standard deviations) from the hurdle rate. The problem is that an increase in volatility is normally associated with a decline in the rate of return.

US High Yield and US Investment Grade are also in negative territory. The chart for High Yield Industrials was positive for most of the last two years, but it cracked badly in early January and has no obvious support. Sooner or later investors will have to discount the probability that the default rate will rise as we move past peak GDP growth. Investment Grade has a more negative chart, partly because volatility is lower. For the time being the issue is not default rates, but spreads over Treasuries, and the amount of leverage that some investors may have applied to their bond holdings. Leverage is an issue for all bond investors. If there is a structural upward shift in volatility – and we think there is – every investor has to reduce his VAR at the same time. Leverage increases the risk that this could become disorderly.

By contrast, the chart for US Equities is still in positive territory even after the correction, and will probably recover somewhat this week. It shows a big adjustment, from an exceptional high (almost 8 standard deviations above the hurdle rate). If the chart stabilises in a range of +1-+2 STD, this would be the equivalent of a substantial overweight in equities. The chart for EM Equities is actually better than the US at the moment, though it never reached the same highs. The chart for Eurozone Equities has fallen into negative territory, mainly thanks to the extra volatility caused by dollar/euro volatility. Despite this, it is still a lot better than any dollar-denominated fixed income category.

Leave a Reply

Wednesday, February 7th, 2018

Fear Volatility not Bond Yields

We are sometimes asked why we don’t use implied volatility, instead of realised volatility, as the basis of our probability models. Using our equity-bond model as an example, we argue that investors require an equity return which is equivalent to the bond return plus the excess volatility (equities minus bonds) in order to hold their benchmark weighting in equities. If the return on offer is substantially more than this, they will have an overweight position and vice versa. The formal answer to the question is that we want to measure the behavioural response of investors to their own experience, not a market-maker’s position. From now on, the short explanation will be: look what happened this week. The VIX spiked; people with short volatility positions got blown up; almost nobody cared.

None of this changes our view that volatility is the single most important variable in determining the behaviour of investors. We can now say with certainty that the Great Volatility Slide is Over (Synopsis – 10th January 2018) and that we are probably on a trend which will lead us back to median levels by the end of this year. In the note we argued that median volatility implies benchmark weightings in most major asset classes, all other things being equal.

Of course, they won’t be, so this note tries to put some numbers on how the change in volatility will interact with rising bond yields, and current estimates for earnings growth. All our usual caveats apply: these are illustrative numbers not forecasts. We include data for 2020, even though we have no idea what is going to happen then. We do, however, know how we would react to any given combination of inputs. Here are the main conclusions: –

• While bond yields are rising, especially in 2018 and 2019, the total return from bonds will be depressed, and probably negative. This significantly lowers the hurdle rate that equities need to beat in order to remain risk efficient.

• If bond yields rise to 3.5% by the end of 2018 and the volatility of both equities and bonds returns to its 22-year median, our model would still recommend a 50% overweight in equities, based on consensus earnings estimates. With volatility only going back to the first quartile, we would be 80% overweight, using the same bond yield.

• For 2019, using a 4.0% yield and consensus estimates, we are still 17% overweight equities in a median volatility regime. Volatility needs to move to the third quartile of its 22-year range for the model to generate a significant underweight in equities.

• 2020 is much more likely to be the danger year, even if we only stay in the median volatility regime. We make the assumption that bond yields stop rising, which means that the bond return is no longer depressed. In order to maintain a benchmark weight in equities we would need over 10% earnings growth and no PE compression, a tall order after two years of a “Trump boom”.

Leave a Reply

Thursday, January 25th, 2018

If You Have to Own Bonds…

All our tactical asset allocation models, no matter which currency they work in, are clear that they want as little exposure to fixed income as possible. If only the world were that simple. Every manager who runs a balanced mandate will have an irreducible minimum exposure to bonds, which he cannot go below. If you are one of these people and you are already at that limit, this note is addressed you. It tells you which style has worked best for each of the major asset pairs within US fixed income and hopefully you still have enough flexibility within the asset class to generate some outperformance.

Our probability-based approach has always assumed that excess returns (Asset A – Asset B) are serially-correlated in the short-run and that investors are risk-averse, i.e. they want to be rewarded for any extra risk greater than the volatility of the low-risk asset. However, these are just assumptions, so we should test them to see if they apply equally to all pairs of assets over the long run.

It is relatively easy to do this. Instead of our normal assumption, we can, for instance, assume that excess returns are mean-reverting and that investors are so confident of this behaviour that they will actively embrace extra risk if the price of one asset falls by a lot relative to another. This a doubling-down strategy. We can also assume that investors care whether they are in a momentum or mean-reverting regime, but are not too concerned about volatility, provided they get the first part of the question right. Finally, there could be investors who think that there is no relationship between current and future excess returns (effectively the efficient market hypothesis – EMH) and are only interested in minimizing (or maximising) volatility. In total there are nine possible combinations of style and risk-preference: anti-risk momentum, risk-indifferent momentum and pro-risk momentum and their equivalents for mean reversion and the EMH. The benchmark (constant 50/50) is the EMH portfolio where investors are indifferent to risk i.e. they make no attempt to minimise or maximise it.

Nine combinations give us nine possible ways in which we can trade each pair of assets, including one where the only trade is an automatic rebalancing to 50/50. All we have to do is compute the estimated returns, and crunch all the usual statistics for return on risk, maximum drawdown, and drawdown as a percentage of total return. We could do this for any pair of assets, but we in this note we focus on US$ fixed income. We look at six pairs of assets: emerging market sovereign bonds vs high yield, investment grade and US 7-10 year treasuries; high yield vs investment grade and US 7-10 year treasuries; investment grade vs US 7-10 year treasuries.

In the following pages, Figures 1-5 summarise the estimated returns for each of these pairs, for each style and risk-preference, grouped under five headings: total return, annualised volatility, return on risk, maximum drawdown and total return as percentage of maximum drawdown. The data cover twenty-two years from the beginning of 1995 to the end of 2017. We also show the raw data for the underlying assets. We highlight the best result in green and the worst result in yellow.

All this is so that we can get to the heart of the matter: which combination of style and risk-preference would have yielded the best overall result for each pair of assets over the last 22 years? In Figures 6-11, we assemble the data for each pair of assets, ranked in order of attractiveness (1=good). We then take the simple average of these ranks and produce a composite ranking. Conclusions as follows:

• In four of the six pairs, the best overall result is produced by one of the momentum styles. In the other two pairs, there is one winner each from mean reversion and the EMH.

• In five of the six pairs, the worst result is produced by one of the mean reversion styles, with the other coming from momentum.

• Within the momentum winners, the composite rankings favour an indifferent rather than an anti-risk preference. However, the difference between the underlying results is not significant, and is exaggerated by using a ranking system.

• For both of the non-momentum winners, the anti-risk preference produces the best result.

• The momentum winners are EM vs High Yield, EM vs Treasuries, High Yield vs Investment Grade and Investment Grade vs Treasuries.

• The mean reversion winner is EM vs Investment Grade, which is probably the least-arbitraged pair out of these six. The EMH winner is High Yield vs Treasuries.

Finally Figure 12 takes the average of the composite rankings for each pair of assets to produce an overall stylistic winner for the whole period. This is the style most likely to produce the best combination of total return, risk efficiency and low drawdown for most of the time. And the winner is…the momentum style with an anti-risk preference, which is why we make it the default, all-weather, all asset-class assumption.

In other words, in the absence of evidence to the contrary, fixed income investors should generally assume that excess returns are serially-correlated in the short run, and that investors are risk-averse. Trying to buy the dips or sell the tops is likely to lead to higher volatility and lower returns. Risk-budgeting is less important than understanding the basic stylistic relationship. We are well-aware that these conclusions contradict some of the most cherished beliefs of the arbitrage community, but this is what our long-run results prove.

Of course, there are times when the normal regime changes, but that will be the subject of another note.

Leave a Reply

    Full Blog Archive:

  • Search Blogs by Category