Act 1: Open all hours
There is a bar in our neighbourhood that goes by the name Charley's Aunt. This is a source of some amusement to us, the name being known as answer to the question:
'Complete the phrase: Mrs Thatcher's heart may be in the right place, but her...'
The bar opens early and closes late - as the saying goes, it's always 5 o'clock somewhere.
The fact that it's always 5pm somewhere, and moreover, its always not a holiday somewhere, means that FX markets are not the place to work if you want a short working day and lots of days off.
Bond markets on the other hand... now we're talking.
One of the main task in pricing bonds is determining just exactly which days are holidays, so you can get your interest accruals for coupons correct. Doesn't that sound like the most fun thing ever?
While FX markets don't technically close, it can get farcical after 5pm New York, when traders have nothing better to do than go stop-hunting (which is one reason not to leave stop orders).
An extreme example of this was at the beginning of 2019, when lots of Japanese traders had their positions stopped out as soon as their trading year began.
Equity markets by contrast, very definitely have a close. With the rise of passive investing in recent years, the percentage of volume being transacted at the close has risen significantly. We mentioned last week that the price action leading into the close is rather distinct to the rest of the day. Let's take a look at what we can learn from it.
Act 2: A tale of four candles
Four candles, you know, 'andles for forks.
Passive funds typically transact at the closing price, because the index that they track is defined with reference to that price. Open end mutual funds, for example, need to buy or sell at the close to accommodate inflows or outflows. Market participants have until 3:45pm to submit orders for the close on NYSE. Shortly afterwards the exchange publishes the order imbalance. It is possible to get this data, but don't bother, because it is arbed away almost instantly by the high frequency folk.
'Oh, you have more buyers than sellers you say, in that case we'll buy now and sell to your buyers later.'
Not to worry though. We can observe the relevant information by examining the relationship between 4 specific candles: the first and last 15 minutes from today and yesterday.
For the first 15 minutes, we use the return from the prior close to the price at 9:45am. As mentioned last week, 9:45am is about the time at which overnight news has been somewhat efficiently priced by the market, so this is a return between two good prices.
We downloaded 15 minute data for a large number of stocks, and calculated the average correlation between the return at each time of day, and the return at the same time the prior day.
Generally this correlation is negative which we would attribute to mean reversion of idiosyncratic stock returns. However, the two highest positive values just happen to be the first and last of the day. That was either very lucky, or we expected this would happen.
Large investors need multiple days to change positions, and they need liquidity, so they trade in the liquid periods at the beginning and end of the day. The autocorrelation that this activity induces is arbitraged away to some extent, but it is still useful to know if we are trying to determine if the buyers and sellers we are expecting seem to be showing up.
What of these arbitrageurs?
Well using the same data, we calculated the average correlation of the overnight return to the return at each point in the day.
What appears to happen is that people try to trend the overnight move in the morning, but then in the afternoon the day-traders close out their positions and the correlation is negative. Once again, this pattern was something we expected to see.
The implication here is that if you've had a good run in the morning and are nearing your target, you might just want to take profit at lunch, as the stock will probably give some gains back in the afternoon. On the other hand, if your target is still some way away, then don't worry, as you may well see the same outperformance in the morning tomorrow.
Act 3: Lights out for the rilk
There are many instances where ETF rebalancing can lead to predictable action around the close. For example, when index constituents change, when bonds are downgraded, or when futures are rolled. Perhaps the most fun example is that of leveraged ETFs.
Like the mysterious rilk, some ETFs fly backwards. For example, there's a new one that is pleasingly named SARK. The most notorious examples however were the inverse VIX products. When VIX went up more than 100% in one day, these backwards products went extinct, and their 2x leveraged cousins essentially followed shortly after. However, apparently, they'll soon be back. Hopefully the new ones won't be as naff as the old ones.
VIX rebalancing flows are somewhat special, in the sense that VIX futures are pretty much only included in VIX specific products, so it is (relatively) easy to calculate rebalancing flows. There's a nice paper by Søren Bundgaard Brøgger which describes the mechanics of these leveraged fund flows in detail. The main point of the mechanics is that these rebalancing flows for leveraged funds are always buying high and selling low, which imposes an annual cost of around 3% on returns. This is a large reason why leveraged funds are not usually suitable as long term investments.
A second observation is that just like with the market imbalance data, the predictable nature of these flows serves to induce arbitrageurs to take the other side. When the flows are large they are presumably more obvious, so more offsetting flows are induced, reducing the relative market impact.
This sounds great, and the new leveraged VIX ETFs apparently will have a new settlement structure based on weighted averages that is supposed to mitigate volatility. However, it is worth remembering that when markets are under extreme duress, these arbitrageurs might well be taking a holiday (though not at Her Majesty's pleasure).
Market view: Barker is chasing tail
Last week, SPY did it's utmost to finish exactly flat, in line with our non-prediction. The result of a rangebound week appears to have been more option selling, which should serve to keep the market going round in a circle again.
Retail sentiment for the market remains strong, so if anything we have a slight bullish bias.