3rd CEPR-Imperial-Plato Market Innovator (MI3) Conference 2019
Keynote: Tick Size
Björn Hagströmer (Stockholm University)
Björn Hagströmer is Associate Professor in Finance. He joined Stockholm Business School after obtaining his PhD degree at Aston Business School (Birmingham, UK) in 2010. He is the course director for Financial Market Structure (FMS). Björn’s research interests include market microstructure and asset pricing. A selection of current working papers are shown below.
Main Takeaways and Quotes
Hagströmer opens his keynote by outlining the two phases of the presentation:
- Why tick size is important for financial markets
- A zoom in to some more in-depth research of his own (specifically ‘Bias in the Effective Bid-Ask Spread’, Feb 10, 2019)
Highlighting the conclusions that he draws, he explains how, “for an investor who doesn’t understand how tick size affects the trading environment, it could look like a bug, but it could act like a bug that bites and sucks the blood in terms of trading performance”.
The first comparison drawn is between two different trading environments: two identical securities that are both priced at around 25 dollars, but with different tick sizes due to different jurisdictions. One is binding and one is non-binding.
“Though the tick size is constant, in nominal terms, in a binding environment like the US (i.e. one penny), relative to the price it is very much varied – for a Ten dollar stock that will be ten basis points in relative terms, but for a one hundred dollar stock, that is one basis point; so a huge difference in trading environments.”
“in the EU, we have a step-wise tick size function, where the nominal tick size depends on price, so it fluctuates for the most traded stocks between one and two basis points, and then for the rest of the traded stocks it is wider, but it is still step-wise in the same way.”
Consequences of Binding Tikck Size
Hagströmer draws three main observations or consequences associated with binding tick size, along with key features of each consequence:
- Wider bid-ask spread:
- This incentivises the supply of liquidity
- It motivates SEC tick size pilot
- High frequency market makers broadly benefit from this feature.
- Higher depth at BBO:
- Greater revenue base leading to higher depth at best bid offer
- There is an unclear effect on the total depth
- There are larger trade sizes under binding ticks.
- There is speculation that fast traders will benefit:
- This is because it eliminates undercutting, and leads to competition on speed
- However, there is less price flickering (which HFTs benefit from), as seen in non-binding tick size markets.
There are other key areas and questions raised by tick size, including finding the optimal tick size, tick size and intermarket competition, the effect on asset pricing and overall liquidity measurement.
This final point is the crux of the discussion moving forward in the keynote speech. He identifies that there are “problems measuring [liquidity through the effective bid-ask spread, the tool mandated by the regulators] when the tick-size is binding.”
The effective Bid-Ask Spread
The next part of the discussion revolved around coloured toy cubes: Hagströmer used the toys to demonstrate the principle behind tick size’s effect on equity trades, using them as a visual aid to a trade taking into account the effect of binding and non-binding ticks.
He concludes that “traders prefer to pay this [pointing to one box] quarter of a cent spread than this three quarter of a cent spread. If that holds, there will be more trades here where the spread is overestimated than what we see here, where the spread is underestimated.”
This concept of the effective bid-ask spread is then demonstrated in terms of equations: the conceptual definition S = D (P – X), and empirical estimator Smid = D (P – Mid), where S = Spread, D = initiated trades (+ or – 1 for buyer or seller-initiated trades), P = transaction price, X = the fundamental value and Mid = the midpoint in the bid-ask spread.
“When the fundamental value is closer to the bid side, there will be more trades coming in at the bid side”
He then goes on to survey one week of S&P stock data from December 7-11th 2015; 55.7 million trade observations, to assess whether there is this bias present; a relatively small data set.
“I took the 55 million trades and categorised them by value deviation from the midpoint, comparing fundamental values of the midpoint, and you can see that most trades are at zero; around 25%.” The counterfactual is drawn at 50%.
“The result is a steep upward sloping probability curve […] This is strong evidence that there is a bias”
Methodology of Estimating fundamental Value
The next step, for Hagströmer, is to outline the methodology of estimating the fundamental value. He draws three main methodologies:
- The Midpoint:
- This can be used to measure liquidity, price discovery, returns and volatility,
- It’s relatively simple and the data underlying is widely available
- The data is discrete and it’s “not a martingale – it has been shown repeatedly that we can predict midpoint changes”
- The Weighted Midpoint:
- Not very attractive to the market maker as “the bid side spread is not very attractive – only a quarter of a cent, whereas on the ask side, you have a wide spread that is probably winder than if you had a smaller tick size”
- It effectively weights the best bid and offer prices by their inverse depth
- This midpoint is continuous and still not a martingale.
- The micro-price
- Due to time constraints, he didn’t explore this other fundamental value tool, but noted that he is happy to answer questions about it.
- It is based on best bid and offer prices and depth
- It is continuous and a martingale.
How big is the bias?
He then asks the question ‘how big is the bias?’ using the weighted midpoint effective spread of 1.61 basis points in the US, or a 1.37 micro-price effective spread (resulting in a 0.25 difference of 18%); “quite substantial, worth more than $200 mil.”
“All this has a lot to do with the binding tick size. As I said before, that will be more binding for the low price stocks […] In the category of stocks priced between five and 15 dollars, one gets almost 100% bias, but then it is falling, and is statistically significant up to somewhere around $115, and that is more than two thirds of the total trading volume.”
Does bias vary across investors
The data found that the bias is far more effective at trading against this fundamental value in high frequency traders, and effects large-cap stocks significantly more than small-cap stocks, though still effecting them.
Conclusion: Does this matter to investors?
“I promised to show you how this matters to investors […] if an investor overlooks valuation in the fundamental value that is not related to the code changes, so when the fundamental value moves back and fourth here, you always look at the midpoint, you overlook a lot of the liquidity valuation, and when you try to time your trades, you want to trade when liquidity is cheap. You can miss a lot of the action”
This is 27% action missed for the whole sample used, and 67% in the low-priced stocks (lower than $50).
If you are an institutional investor you may not monitor the midpoint spread, and so are more interested in pricing it, “but price, in fact, is often measured using mid points aswell”
“If you measure price impact from the point of trade until five minutes after, then you get 1.93 basis points on average using the midpoint, and if you use the micro-price impact you get 1.68 bps; so that’s 15%, and for the low price stocks its far more [37%]”.
In Conclusion, “this paper is showing that the midpoint effective spread is biased relative to the true effective spread in US equity markets, however it is not only limited to the US equity markets – I would expect this to apply to any security that trades with binding tick size”.
“There are various ways of overcoming that”.
Images: Muhammad Ashraf ©2019