Thursday, 27 April 2017

Relationship Between XLF and Bond Yield Curve Steepness: A Study

 The Short Term View 

Since the Trump's presidential victory in Nov 2016, the immediate outcome of that event is the incredible spike in TNX (10 Year Treasury Note Yield), and the corresponding rise in financial stocks as captured by XLF (Financial Select Sector SPDR Fund). This is caused by the anticipation of Trump's pro-growth policies, which would be highly inflationary if the policies are successfully implemented.

Apart from such anticipation of financial deregulation (repeal of Dodd-Frank Act) and tax reduction, the only actual and material effect on XLF is caused the steep rise of TNX, which brought about the corresponding rise in XLF. As the media (almost all CNBC commentators) and others (on blogosphere)  suggest that the steep rise in TNX causes the yield curve to steepen, and that benefits financials' profitability. The financials need a wider spread between short-term and long-term bond yield to make money. So one can comfortably conclude: steeper yield curve leads to rising XLF prices.

I thought, this makes sense. And then, i also realise that reality is usually the opposite of the conventional wisdom, so i decide to dwell a little deeper into the matter.

Let's look at the 2 charts below to see how well this argument is being supported by data.

5 month chart of 10 yer and 2 year treasury note yield spread since 2016 election
5 month chart of 10 yer and 2 year treasury note yield spread since 2016 election

5 month chart of XLF ETF since 2016 election
5 month chart of XLF ETF since 2016 election

The yield spread chart, which is a standard measure for YCS (yield curve steepness), rises rapidly in the immediate aftermath of the election, tops in early December and declines steadily since. While XLF follows the steep rise of the YCS in the upwards direction, it doesn't really follow its downward move. In fact, while the steepness tops out in December 2016, XLF continues to rise until March 2017. The fall of XLF since March, one can argue is due to the general market correction.

Perhaps, one might also argue, the other 2 factors - tax cuts and Dodd-Frank repeal - keep XLF elevated relative to the YCS. After all, YCS isn't the only game in town for financials.

So these 2 charts above appears to suggest a good (and positive) correlation between the YCS and XLF, at least during Nov to Dec 2016. The correlation weakens after 12/2016.

I want to find more clues by looking at the longer time frame, to look at the relationship between XLF and yield curve steepness historically. Hopefully, these long term data shed some light.

 The Longer Term View 

The following 2 charts are simply the same 2 charts above with a longer time span from 1999 to 2017 (XLF data only goes back to 1999). We divide 6 distinct periods within the time span we study.

The numbered arrows (1 to 6) show the general trends in the corresponding 6 periods in both charts.

Yield curve steepness from 1998 to March 2017
Yield Curve Steepness (YCS) from 1998 to March 2017
Each arrow indicates a general trend direction for a given period

XLF from 1999 to March 2017
XLF from 1999 to March 2017
Each arrow indicates a general trend for a given period (that's corresponded to the YCS chart above)

The revelation of these 2 charts is suggesting something counter-intuitive and more complicated (as one should expect from life or financial market. There's never a simple correlation that works all the time. E.g. gold falls when DXY rises and vice versa. E.g. just look at a large drop in both DXY and gold price on the same day on 24/04/2017. Sometimes they do, sometimes they don't).

When we look at the falling trends of arrows 2 and 6 in YCS chart, they correspond to rising trends in XLF in the same period. It's the exact opposite to what i said above to the short period of Nov to Dec 2016.

Likewise, arrows 1 and 3 in YCS chart indicates rising trends while the same arrows 1 and 3 in XLF chart show the exact polar opposite trends, once again.

But the picture is more complicated with arrows in 4 and 5, where in both charts they seem to point to the same direction in both charts rather in the opposite.

Actually if we combine the two periods of 4 and 5 in both charts and consider them as one period (let's called this period 4A), then the trend in period 4A is negative in YCS chart, and positive in the XLF chart. Of course, i could just as well combine the periods 4, 5 and 6 into one period, and label this combined arrow 4B (the whole period 5 can be considered just a counter trend of the combined downtrend). In this case, 4B is in the general downtrend in YCS chart, and uptrend in XLF chart.

Problem solved.

Am i trying to change the data in order to fit my conclusion? Let's consider the following reasoning before you accuse me of data fiddling.

I think the reason why the periods 4 to 6 is somewhat troublesome compare to periods 1 to 3 is that the YCS trend is somewhat messy in periods 4 to 6 while the YCS trends in periods 1 to 3 are clean and pronounced. Because of the cleanness of the trends in periods 1 to 3, i tend to rely it more for the conclusion, which is a negative correlation between YCS and XLF.

More importantly, in periods 1 to 3, the arrows either start from the bottom of the chart (near 0%) to the top of the chart (just under 3%), or vice versa. In other words, the arrows always start from one end of the chart (top or bottom) to the opposite end (bottom or top) in YCS chart. The maximum spreads between 2 and 10 years bond yields seems to be kept under 3% since 1999.

You could say period 4B (which starts near the top of the chart) is still being played out to its final conclusion as it moved from the top to the current level and eventually come down to the bottom with 0% (flat yield curve) or negative (inverted yield curve).

This makes a lot of sense as we're heading into recession from economic growth in the next few years. This is perhaps, to me, provides an explanation for this negative correlations between YCS and XLF because economic expansion typically ends with a gradual flattening of yield curve to an inverted one, which typically corresponds with a peak in XLF as the stock market exhausts its bullish run. In this light, one is almost forced to conclude that decreasing YCS down to zero or negative should occur as XLF and the general bull market proceeds towards its eventual peak (just as it did in 2007 in the time span of our study).

This period 4B is considerably longer than its previous periods, but this current economic expansion is also one of the longest (could turn out to be the longest in US history, but not very long in world history. E.g. Australia's current economic expansion is 25.4 years old, and we don't know when it will end anytime soon. Netherlands' record is 103 years. Let's end another sacred mantra, "US economy should be in recession after 7 years". I'm not saying the next US recession won't occur for another 100 years. I'm just saying there's no fixed end days for economic recession. And 10 years isn't a long time for an economic cycle in world standard. My gut feeling - never rely on gut feel, mine or yours - tells me that this current US economic expansion would end in 2020 - 2021. Perhaps because i'm thinking that YCS is nowhere near zero. It's in fact, roughly half way between the top of 2.9% and the bottom of -0.5%. The median of these 2 points is 1.2%, and the current YCS is 1.04%, a little below the median).

Side note: I should simply have 4 periods to start with. I analyse the data while i wrote, and so i ended up writing down all my thinking process, instead of just the result. My apology (for i'm too lazy to do a complete rewrite). I write so that i can gain an understanding for myself, first and foremost (and as a bonus, share those analysis with the like-minded. Hence, the occasional messiness).


In short, falling YCS leads to rising XLF, there's a negative correlation between YCS and XLF,  for longer time span (measured in years). In the short term, such as the period between Nov to Dec 2016, the correlation between YCS and XLF is positive (measured in weeks). This short term relationship is probably transient phenomena.

If the YCS continues to fall from here and now, we should hold onto the financials because this is a positive signal for financials. We've already seeing this being played out in the last few months. As YCS continues to fall since January this year, we don't see the same percentage drop in XLF.

The rise in YCS in the immediate aftermath of Trump election is but a blip in the general downtrend in YCS in period 6 (using market jargon, it's just a relief rally in a bear market of YCS). And one is tempted to project that as YCS goes to zero, XLF would peak as it did in 2007.

In other words, if YCS (if there's such a ETF) is in bear market, then XLF is in bull market.

Am i wrong in drawing this conclusion because the time span i covered isn't long enough? I.e. only since 1999. Maybe. At the very least, we can debunk the seemingly sacred mantra that i heard in the mainstream media, "XLF goes up when YCS goes up" or "XLF falls when the yield curve flattens". That's just not evident in general. That mantras is dangerous for the long term investors as they would afraid to hold financials after constantly hearing and seeing a flattening yield curve being repeated in the media. At best, it confuses the viewers-investors. Maybe we should have a new mantra, "when yield curve is flattening, time to load up financials." This mantra doesn't hold true in the extreme. I.e. it holds true when it's flattening, not when it's flat as a pancake. Or a more intuitive mantra, "When TNX is rising, time to load up financials." And forget the whole business of YCS, which confuses us than helps us. Of course, YCS is more helpful for timing recession than TNX.

So what's that business about the financials need high YCS to be profitable? I don't know. The data since 1999 simply don't support that conclusion for extended period. Perhaps, the 10 year treasury yield (TNX, not YCS) that has a positive correlation with XLF. Let's have a look at that.

Here's a chart showing these 2 curves from 1999.

XLF vs TNX performance from 1999 to April 2017
XLF vs TNX performance from 1999 to April 2017

A couple of observations we could make.

1.  XLF outperforms TNX during period of rising TNX over the long term. We know that TNX has been in the bear market since early 1980s, and yet XLF is managed to outperform TNX with ever wider margins. The gaps between the 2 curves widen as TNX falls further towards zero. The financials may make less money with lower TNX, but they've managed to make more money as per TNX (by being more efficient, i imagine).

2.  This chart shows good positive correlation between the 2 curves (XLF and TNX), except for the relatively short period of 1 year or so (out of 18 years) between 12/2013 and 01/2015. The starting point of this brief negative correlation period coincides exactly with the starting point of period 6. I don't know the significance of this. Is it just a coincidence?

This positive correlation also nicely explains that why when YCS tops out in 12/2016 and yet XLF continues to move higher and tops out in 03/2017, this is because TNX also tops out in 03/2017. Actually we have something like a double tops in TNX between 12/2016 and 03/2017.

 In Summary 

TNX is indeed more important to financial profitability than YCS, which presents us with a counter-intuitive relationship.

 As i said before, YCS isn't the only game in town. There's also TNX. The more important game in town. Rising TNX signal inflation and therefore economic growth, hence financials' rising profitability because of higher loan growths. Therefore, loan growth trumps YCS. Falling YCS signals that economic growth is occurring (good for financials and stock market in general) and it's heading towards the monetary tightening that would result in an increasingly likelihood of an economic contraction.

Even when YCS are flattened or inverted, it could be year(s) before reaching the final peak of the stock market. In fact, the YCS can steepen (after inverted) before the top of the market, leading investors to think that things have returned to normal. Signals from yield curve isn't always straight forward (no pun intended). Bond market is made up of human players, so they could send the wrong signals, at least in the short period. Perhaps, the brief period of 2014 where the correlation between XLF and TNX being negative is just bond market getting it wrong?

No comments:

Post a Comment