Upload
clarence-lee
View
216
Download
0
Tags:
Embed Size (px)
Citation preview
Potential Pilot Problems
Charles M. JonesColumbia Business SchoolDecember 2014
1
The popular view about equity markets
2
Trading certainly looks different today…
20th century 21st century
Automation has driven out costs.Is it increasing liquidity and helping firms raise capital? 3
Two liquidity measures defined Effective bid-ask spreads
ESit = | Pit – Mit | Distance from prevailing midpoint Mit to trade price Pit
Actually a half-spread or one-way cost Defined for a single (child) transaction
Implementation shortfall More relevant for a parent order (e.g., buy 1mm shares
of IBM) For buys, ISit = – Mit
Distance (usually in bps) from decision-time price Mit to average trade price
Captures effect of driving prices up with sequences of buy orders
4
Source: spliced ITG research reports5
2005 2006 2007 2008 2009 2010 2011 2012 201320
40
60
80
0
10
20
30
40
50
60
US large-cap trading costs have trended down
IS Costs Commissions Average VIX
Co
sts
in b
ps
VIX
Source: Jun 2014 ITG Global Cost Re-view
What caused the improvements? There is a straightforward Econ 101 story
More competition within and across exchanges Scalable technology drives down costs
But we can’t be sure: correlation is not causality!
Many other things have changed over the past 20 years Various regulatory changes Perhaps less private information now
Can use market structure changes as instruments: Example: Hendershott, Jones and Menkveld (2010 JF)
But the gold standard for determining causal effects is randomized controlled trials
6
An example: 2007 repeal of short sale uptick rule Before 2005, NYSE short sales could only
happen: On an uptick (at a price higher than the last sale price) Or on a zero-plus tick (at the same price as the previous
transaction if the most recent price change was positive)
Regulation SHO: Adopted by the SEC in 2005. Initiated a pilot program suspending the NYSE’s uptick
rule and the Nasdaq’s analogous bid test.
All Russell 3000 stocks ranked by market value; every third stock assigned to the pilot.
Pilot continued into 2007.
SEC decided to repeal all price tests Announced June 13, 2007 Effective July 6, 2007
7
Empirical design for studying the 2007 repeal Takes advantage of virtually random assignment
Econometric approach: look before and after final repeal
Initial approach: treatment vs. control Treatment group (non-pilot stocks) experiences the
repeal Control group (pilot stocks) free of the uptick rule
throughout
Implemented via a differences-in-differences regression:Yit = β0 + β1Ti + β2At + β3TiAt + εit
whereYit is the outcome variable for stock i at time t,Ti = 1 if stock i is in the treatment group, Ti = 0
otherwiseAt = 1 if date t is after treatment (after repeal), else
At = 0
The interaction term β3 measures the average treatment effect.
8
More shorting since tick test repealed
Jan Feb Mar Apr May Jun Jul Aug0%
10%
20%
30%
40%
50%
60%
Shorting prevalence during 2007 in NYSE stocks
non-pilot (treatment) pilot (control)
Sh
ort
ing
as
a f
racti
on
of
trad
ing
vo
lum
e
Tick test repealed
9
Short-sale orders become more aggressiveShort order characteristics in NYSE stocks during 2007
0%
10%
20%
30%
40%
50%
Jan Feb Mar Apr May Jun Jul Aug
Frac
tion
of sh
ort s
ales
non-pilot marketable non-pilot passive pilot marketable pilot passive
Tick test repealed
Passive short-sale orders are those placed at or above the prevailing ask price. 10
The problem with this empirical design Doesn’t work if there are treatment spillover effects.
Spillovers mean control stocks are affected by the treatment too.
Controls aren’t actually controls.
Not clear what the difference-in-difference approach measures.
Seminal paper in econ: “Worms” (Miguel and Kremer, 2004) Study randomized deworming treatments on Kenyan village
children But children in the control group also benefit via less
transmission So can’t do simple treatment vs. control
These spillovers are called interference in the statistics literature.
11
What’s the problem with uptick repeal? Many short sale strategies are portfolio strategies
Example: index arbitrage. If the index is cheap: Buy futures or an index ETF Simultaneously short all of the underlying stocks
During the Reg SHO pilot, this strategy was hard to execute: Only about 1/3 of S&P500 stocks exempt from the uptick rule For all the rest, can’t short without complying with the uptick
rule Introduced substantial risk into this strategy.
After repeal, could short all stocks without this constraint Would expect more shorting of lists of stocks More shorting of pilot (control) stocks Voila! Treatment spillover.
Same is true for any list-based strategy (e.g., factors)12
Revisiting the evidence
Jan Feb Mar Apr May Jun Jul Aug0%
10%
20%
30%
40%
50%
60%
Shorting prevalence during 2007 in NYSE stocks
non-pilot (treatment) pilot (control)
Sh
ort
ing
as
a f
racti
on
of
trad
ing
vo
lum
e
Tick test repealed
13
More shorting of both treatment AND control stocks
Spillover effect could be quite large for effective bid-ask spreads
14
20070103200701042007010520070108200701092007011020070111200701122007011620070117200701182007011920070122200701232007012420070125200701262007012920070130200701312007020120070202200702052007020620070207200702082007020920070212200702132007021420070215200702162007022020070221200702222007022320070226200702272007022820070301200703022007030520070306200703072007030820070309200703122007031320070314200703152007031620070319200703202007032120070322200703232007032620070327200703282007032920070330200704022007040320070404200704052007040920070410200704112007041220070416200704172007041820070419200704202007042320070424200704252007042620070427200704302007050120070502200705032007050420070507200705082007050920070510200705112007051420070515200705162007051720070518200705212007052220070523200705242007052520070529200705302007053120070601200706042007060520070606200706072007060820070611200706122007061320070614200706152007061820070619200706202007062120070622200706252007062620070627200706282007062920070702200707032007070520070706200707092007071020070711200707122007071320070716200707172007071820070719200707202007072320070724200707252007072620070727200707302007073120070801200708022007080320070806200708072007080820070809200708102007081320070814200708152007081620070817200708202007082120070822200708232007082420070827200708282007082920070830200708310.04%
0.06%
0.08%
0.10%
0.12%
0.14%
0.16%
0.18%
non pilot mean res pilot mean res
Tick test repealed
Pilot (control) stocks see big changes: treatment spillover or unrelated?
This is not always a problem: no evidence of spillovers during 2008 shorting ban
Cross-sectional mean of short sales as a percentage of trading volume (RELSS) for stocks on the original Sep 2008 SEC ban list with matched non-banned stocks.
0%
5%
10%
15%
20%
25%
30%
Quartile 1 (Small-cap)
Banned stocks Non-banned match
Ban period
0%
5%
10%
15%
20%
25%
30%
35%
Quartile 2
Banned stocks Non-banned match
Ban period
0%
10%
20%
30%
40%
50%
60%
Quartile 3
Banned stocks Non-banned match
Ban period
0%5%
10%15%20%25%30%35%40%45%
Quartile 4 (Large-cap)
Banned stocks Non-banned match
Ban period
15
Tackling spillovers methodologically
Using notation from causal effects literature, Yi(zi,ψ) is the potential outcome for firm i given: its own treatment zi = {0, 1} ψ is the fraction of firms treated at random We only observe one of these outcomes; the other is the
unobserved counterfactual
Overall treatment effect (TE) moving from treatment strategy ψ to strategy ϕ:
TE(ψ, ϕ) = Σ E[Yi(1, ψ) – Yi(0, ϕ)] This can be rewritten as:
TE(ψ, ϕ) = Σ E[Yi(1, ψ) – Yi(0, ψ) + Yi(0, ψ) – Yi(0, ϕ)]
direct treatment effect indirect treatment effect
16
Tackling spillovers (cont’d.)
A treatment strategy ψ is often compared to no treatment (ϕ = 0). corresponds to the beginning of a regulatory pilot
program.
If the pilot is extended to all firms, treatment strategy changes from the original pilot fraction ϕ to ψ = 1.
In biostatistics, other fractions make sense: Vaccinating 75% vs. 50% of the population
Statistical inference is easier if you have many different groups with only within-group spillovers. Most stats papers discuss this case. Example: “Worms” studies randomized trials in many
villages.
17
But most regulatory pilots are one village One solution: identify off of differences-in-differences
regression with controls:Yit = β0 + β1Ti + β2At + β3TiAt + γXit + εit
whereYit is the outcome variable for stock i at time t,Ti = 1 if stock i is in the treatment group, Ti = 0
otherwiseAt = 1 if date t is after treatment (after repeal), else
At = 0Xit is a vector of control variables
The interaction term β3 measures the direct treatment effect.
β2 measures the indirect treatment effect (the average change in control firm outcome from moving to new treatment strategy).
Controls become quite important here.
18
Indirect effect non-trivial for uptick repeal
Jan Feb Mar Apr May Jun Jul Aug0%
10%
20%
30%
40%
50%
60%
Shorting prevalence during 2007 in NYSE stocks
non-pilot (treatment) pilot (control)
Sh
ort
ing
as
a f
racti
on
of
trad
ing
vo
lum
e
Tick test repealedDirect treatment effect: +5.8%Indirect treatment effect: +6.0%Total treatment effect: +11.8%
19
What’s HFT got to do with all this?
Pilot designers need to think about potential spillovers.
Currently in the U.S.: concern that current market structure is not ideal for small-cap firms.
20
2005 2006 2007 2008 2009 2010 2011 2012 201320
40
60
80
100
120
140
160
0
10
20
30
40
50
60
But small-cap trading costs remain high
IS Costs Commissions Average VIX
Co
sts
in b
ps
VIX
Source: Jun 2014 ITG Global Cost Re-view
21
SEC plans a new pilot program for smaller-caps To be a pilot stock, must satisfy all of the following:
Market cap of $5 billion or less Average daily volume (ADV) of 1 million shares or less Share price of $2 or more.
Pilot design: 1 control group and 3 test groups Approximately 300 securities in each of the four buckets
Test group 1: Quoted in nickels ($0.05), no other restrictions
Test group 2: Quoted & traded in nickels OR at the mid-point of the NBBO. Retail orders internalized only with price improvement of at
least $0.005. No price improvement required for trades off-exchange (e.g.,
dark pool).
Test group 3 same as group 2 plus: “Trade-at” requirement: off-exchange trades require
significant price or size improvement. Otherwise, must first execute against the full size of on-
exchange, protected quotations at the NBBO before executing off-exchange.
22
Overall conclusions
Equity market liquidity in large caps is clearly better than it was 10 years ago. Competition and cost reduction are probably the cause
Regulatory experiments have the potential to clearly identify causal effects. Would be great if Europe could start to do them Must think carefully about spillovers Must design the experiment carefully to maximize info
gained
My predictions and pleas: Due to the nature of information about small firms,
small cap liquidity will always be lousy regardless of market structure
Tick size and trade-at will have close to zero effect Trade-at should dramatically increase liquidity in large-
cap stocks; let’s try the pilot there! 23
For further reading
This talk incorporates elements from the following papers:
Ekkehart Boehmer, Charles M. Jones, and Xiaoyan Zhang (2013), “Shackling short sellers: the 2008 shorting ban,” Review of Financial Studies, 26:1363-1400.
Ekkehart Boehmer, Charles M. Jones, and Xiaoyan Zhang (2014), “Unshackling short sellers: the repeal of the uptick rule,” SSRN working paper.
Terrence Hendershott, Charles M. Jones, and Albert Menkveld (2010), “Does algorithmic trading improve liquidity?” Journal of Finance.
Terrence Hendershott, Charles M. Jones, and Albert Menkveld (2013), “Implementation shortfall and high-frequency price dynamics,” Chapter 9 of High Frequency Trading (edited by Maureen O’Hara, Marcos López de Prado and David Easley), Risk Books.
Charles M. Jones (2013), “What do we know about high-frequency trading?” SSRN working paper.
24