Volatility emerges as a key effect of the price discovery and order execution processes in financial markets. Microstructure aspects, like non-synchronous trading, price effects of volatility, and volume effects of volatility, can influence volatility though they may be ignored at longer horizons. Measures of order flow, like probability of informed trading (PIN), have been developed to help explain volatility and the transmission of private information in markets.
2. Introduction
Microstructure aspects can be safely ignored at longer horizons
They are however first symptoms of market asymmetries
First port of call for Private information ( as soon as the order is entered)
The research topic has hardly been explored in the Indian markets despite
the financial markets being deep and widely traded
Volatility in Prices seen as carrier of Information ( Filtration set) as information
is carried in Prices esp during Price Discovery and in Volumes traded
(Information about demand)
Global research in the areas of volatility and microstructure in financial
markets a salient area since the 80s
3. Microstructure as Volatility
The structure of markets ( Continuous Double Auction Markets in BSE and
NSE) lead to typical effects in the microstructure.
However once the latent effects of non synchronous trading are detected
and taken care of in the data
(spurious correlations) from Non Synchronous trading
(high frequency noise)
U shaped serial correlation in price volatility
Volatility emerges as a key effect of the Price discovery and Order
execution processes:
The price effect of volatility
The volume effect of volatility
4. The study of Price trends and Efficient
Markets
All forms of Market Efficiency rely on Price reflecting public and private
information in prices
As a construct, the Efficient Market Hypothesis continues to be a bulwark of
Financial Markets research
The existence of Price and Volume based trends and the existence of
trading around a denomination of Fundamental Value drive trading
algorithms and informed Human traders to the market for every trade in
Financial markets (Electronic exchanges, OTC markets and trading floors)
in Currencies, Equities, Derivatives, Fixed Income or Credit and complex
derivative combinations of these to satisfy the need to price and exchange
risks and with the underlying motivation of a (fair) profit
5. Model of Asynchronous trading
Lo and Mackinlay, created one of the first micro models for focusing on the
return generating process of N securities
In the absence of trading frictions or institutional rigidities , we can assign a
virtual return rit (CCR)reflecting company specific information and economy
wide effects for each security i
These virtual returns, intuitively differ from observed returns because of trading
frictions.
Lo and Mackinlay assume a non trading probability it in each period t,
assumed a s a IID sequence of coin tosses independent of the returns rit
rit
o , the Observed return of the security can be set to 0 if there is no transaction
as rit
o = ln (pt/pt-1), thus if we consider 5 periods consecutively in which the
security trades only in 1,2,5, the security reflects the same virtual return till period
2 and then observed return is 0 for 3 and 4 to get the 5th period return
accumulating that of 3 4 and 5
6. Lo and Mackinlay
One factor linear model for returns: ri = I + ifi + it
f is a 0 mean common factor and is idiosyncratic noise, both
independent of all leads and lags and f(t) Is independent of (it-k)
We use non trading probabilities to designate two new RVs
The first one is 1 while it does not trade while X is switched to 1 only when
the security trades
7. The duration of non trading k can then be expressed as a Product of all the
delta consecutively, and observed returns a re a function of X
This duration can be expressed as E(k) = pi/(1-pi), Var(k) = pi/(1-pi)^2
The variance of observed returns and other moments are thus:
8. With a security with non zero expected returns, variances are dilated and
negative serial correlation induced decays geometrically
The minimum autocorrelation as pi varies from 0 to 1 is achieved at
Min Corr(rit
o, rit+1
o)=-(CV/1+1.414*|CV|)^2
At Pi= 1/1+sqrt(2)|CV|
This spurious autocorrelation can be generalized and erased from a
portfolio of securities to reflect common news reflected in illiquid securities
with jumps
9. Bid-Ask Spread
Roll (1984)
If P*(t) be the fundamental value in a frictionless economy and s be the spread,
P(t) = P*(t) + I*s/2
Where I is IID in +1 for buyer initiated trades with p=0.5
and -1 for seller initiated trades with p = 0.5
Thus Change in Price = P = I(t) – I(t-1)*s/2
Here, the second moments of the process are
Var(P) = s2/2 , Cov(one period) = - s2/4, Cov k period k>1 = 0
Corr( one period) = -1/2
This intuition can be used to price the spread a s new information is produced
for the security
10. Ordered Probit Model
Lo and Mackinlay (1992) go on to present an ordered probit model which
allows volatility to be modeled in generally conditionally heteroscedastic
models and maps discrete time to continuous time assumed for price functions
This is solved using MLE and conditional volatility coefficients are squared to
ensure non negative volatility
Others:
Barrier Models by Cho and Freres and Marsh and Rosenfeld model transactions as
achievement of new discrete Price barriers, mapping the tau interval of trading
Rounding models (Gottleib and Kalay 1985)
Glosten and Harris (1988) – A permanent adverse selection component is present in
the data on bid ask spreads
43% adverse selection, 10% inventory holding costs and 47% transaction processing
11. Kyle Measures and Order Flow
Current literature on Microstructure still uses first principles.
Intuitively measures of Order flow (Bid Volume – Ask Volume) are the most
appealing
The latest measure by Easley and O Hara in PIN or VPIN measuring
Probability of Informed trading and Toxicity of the Order flow.
These measures initially helped explain a lot in High Frequency traded
markets in High Frequency Data
PIN measures are updated daily with methodological variants likely to
ensue in the future and also likely to be used by us in our work in this term
12. Volatility Spillovers and Cross Market
movements
Global integration of markets marked by a literature spanning the last decade
measuring commonality in movements between global indices
The literature shows obvious relations for dependent economies with the Dow
index .
Some literature also assumes a strengthening commonality of market moves a
sign of progress
Spillovers more interesting in Cross asset moves and hard to measure by
cointegration .
Currency markets to non traded financial asset markets most affected by equity
and derivatives markets trends
Without spillovers, yield curves can measure and predict foreing exchange
premiums (Ang and Chen, 2010) These predictions exhibit low skewness and low
correlation with carry returns
13. News Impact Curve and GARCH
models
Engel’s original study of volatility and its econometric measures guided by
and impactful is assessing the News impact of volatility of markets while
they assimilate information
A current study of news transmission and efficient price transmission in
markets may be handled by a microstructure approach
However microstructure approach effects of discrete markets already
available to discount from price and trading data
Arrival of news studied thru event studies a distinct class of analysis that
show the literature’s play with news arrival and assimilation at a larger level
14. Market Orders vs Limit Orders
A buy order is at a higher price than the mid rate, a sell order is at a lower
price than the mid rate
The early literature has assumed Market Orders signal informed trading as
the orders typically demand instant liquidity as if they have an expectation
of price (gains)
However this is changing as most mature markets in the trading zone
convert into Limit orders even if Market orders signal arrival of new
information.
Thus Limit orders are not only suppliers of liquidity but also arbitraging off the
availability of fundamental value and can be classified
15. The measures of illiquidity
It is generally known that if you take the effects of both price component of
liquidity and volume effects of liquidity you can mirror the intuition of less
traded stocks/assets also identifying significant Private information available
in the security vis a vis a general non interest in the stock/asset
Amihud and Mendelson and Hasbrouck and Gideon Saar set early
benchmark measures to determine the illiquidity premium
16. Asset pricing and the bid-ask spread
Amihud and Mendelsen 1986
The bid ask spread reflects the cost of execution and is found to be
negatively related to trading volume and the stock price continuity
First to reflect the increasing bid ask spread as the return premia observed
in the returns
Clientele effect implies here that longer horizon investors choose securities
with a higher spread
Slope of the return spread relationship decreases with the spread
Amihud (2002) time average absolute return with Dollar Volume of Trading
17. Assymetric information measures
Early measures quoted by Hasbrouck relate the Asymmetric component with
the plight of an uninformed dealer having to make market order trades
including the work of Kyle (1985) and Glosten and Milgrom(1984) as well as the
earliest work of Easley and O Hara (1987) and Glosten and Harris (1988)
The effect of inventory is available on specialized trading platforms aka the
NYSE with a contemporaneous and subsequent price impact
The Glosten and Harris measure ignores inventory
The quote revisions are treated as serially correlated with the quote revision
attributed to the assumption of trade event belonging to an informed trader,
assuming an Inventory state as white noise I=- t
q(t)-q(t-1)= v(t) – v(t-1) – Beta(I(t)-I(t-1))
Demand is then modeled as a moving average of the white noise terms
18. Liquidity Risk
Acharya and Pedersen(2005)
Liquidity is risky and is modeled for individual stocks with a common market
component
A Liquidity CAPM is drawn up using the ILLIQ premium C/P for C in transaction
costs and P the current price, return r = (D-C’+P)/P(t-1)
Here Dividend is modeled as D-C
Similar measures of illiquidity and return are aggregated for the market
component
The market return & stock return increases with the covariance of market and
stock liquidity,
Covariance between market liquidity and stock return negatively impacts the
idiosyncratic return (Pastor and Stambaugh, 2003)
Thirdly, investors accept a lower return on the liquid securities
19. Paper goes on to realise a differing role for the persistence of liquidity
Return also decreases with previous year’s volume and increases with spread
Bekaert (2003) implies illiquidity directly predicts returns in emerging markets
This may or may not be a temporary condition
The original volatility feedback effect is rarely documented in this literature .
As Expected returns increase the increase in required returns depresses the
current state of Price functions for the security
20. Michael Brennan
Brennan goes on in a series of studies linking volatilities in debt to a higher
peg than in emerging market and foreign equities
In Brennan Huh and Subrahmanyam (2013) authors build on the earlier
validation of the Amihud (2002) measure
They also build on their study of the symmetric Kyle measure in Brennan et
al (2010) where they find only seller initiated trades reflect a relationship (
trade at bid prices) (relation between return and changes in illiq)
Thus they decompose the Amihud measure for down markets and up
markets
Half Amihud measure is strongly priced in the cross section of returns on
down days
21. Liquidity and Asset Prices
Amihud, Mendelson and Pedersen(2005)
Measures in the impact of the volatility feedback effect with higher expected
returns depressing prices on reduction of liquidity
It can be assumed in general that hard to trade securities are trading at hard
positive discounts to their fair value
This series of literature thru 2013 reflects a desire to patch up Asset pricing
anomalies(Small Firm effect and others) using the liquidity risk priced in the
microstructure including Brunnermeier, Acharya , Pedersen, Subrahmanyam
and Brennan
The Paper merges different sources of (il)liquidity
Transaction costs ( exogenous),
Demand Pressure ( hitherto inventory risk, not in CDA markets)
Private Information
Order flow and search frictions
22. Price Formation
Joel Hasbrouck(2003) studies ETF and index futures trades
Volumes positively impacted on a large scale by presence of Index ETFs ,
Futures and exchange traded sector funds
Traded on Open Electronic Limit Order Books (available outside the ‘floor’)
A natural offshoot of trading for liquidity traders interested in diversification
Pennacchhi(1993) and Admati and Pfeiderrer(1988)
Brennan and Xia study Macroeconomic effects aas real interest rates on
different categories of stocks demarcated by Sharpe Ratio measuring the
impact on Pricing thru Discounted Cash flows
23. Probability of Informed Trading
Easley, Hvidkjaer and O Hara
Tools are also readily available in R for this measure
Initial validation paper with Engle and Wu in 2002(2008)
Bivariate Autoregressive intensity process
Arrival rates of informed and uninformed trades
Market Liquidity , Order Depth and Order Flow
Daily conditional rates construct forecasts of the PIN measure
PCA of PIN reflects a dominating impact of the systemic liquidity factor
Order Arrival used in a GARCH like process fitting to liquidty based returns
24. PIN
Measures components of order imbalance
Order flow information
Impact on market Depth
Liquidity
Price impact of transaction used as barrier to measure number of buy trades
required to reduce Price impact below barrier(HDFC Bank never exceeds 0.6%)
These PINS may be high before earnings announcement on expectations and
lower in post announcement trading as Private information converts into public
information
PINs work in a smaller time window in event studies (+/- 7 days)
Bad News with probability delta and Good news with probability 1-delta
(analogous to the Lo and Mackinlay measures)
25. PIN
Reflect the same concern with a delineation of only two types of
traders,uninformed = liquidity traders
Probability of news measured as alpha
Both traders modeled as Poisson processes with mu and epsilon arrival rate
parameters
Expected(Total Trades)
Order Imbalance
26. PIN results
High turnover stocks exhibiting continuing returns continue to show returns
unaffected by earnings events if they have high PINs ( liquidity)
They may have low PINS in this case they are subject to return reversals
High PIN stocks thus consistent with High Private information passing to
uninformed dealers keeping up returns
Uncertainty in Zhang (2006) also shown to work with PIN
27. Information uncertainty (Zhang 2006)
Blume, Ohara and Easley
Stock price continuations studied in market literature
A sequence of stock prices reflect information and a single price does not
reflect underlying information
Information is revealed slowly because of higher information uncertainty
In high Uncertaainty , as expected from the volatility feedback effect , the
required positive returns on good news increase but the negative impact of
bad news is minimized
Volumes can however not contribute to dissemination of information.
Can be shown that volume correlated the quality of information being supplied
to changes in expected prices/returns
One period model of trader demand ( needs explanation)
Zhang uses Firm age and Coverage as parameters in uncertainty model woth
analyst dispersion and SIGMA(volatility) as well as cash flow volatility
28. Trading Price jump clusters
Novotny , Petrov and Urga (2014)
Cluster of Price Jumps in Yen and Euro around tapering news and
Abenomics/ECB releases of liquidity
Assessment of trading oppoirtunities represented by price clusters
Show that Bid ask spreads eat into the returns
The market makers mark-up the short-term (overnight) implied volatility at the
foreign exchange markets as the uncertainty increases and
the impact of the announcements may increase the realized volatility
Verdelhan(2010) risk aversion rises around news announcements ( News Impact
relation with higher volatility)
Liquidity shocks timed with market news except for temporal inefficiencies
provided by central banks
29. News and Sentiment
Paul Tetlock study of the WSJ broadcast ‘Heard on the Street’ column
Negative sentiment as posited by author or Spreads
30. Stop Loss Rules
Kaminsky and Lo
Can substantially reduce volatility
Can substantially increase expected returns
31. Impossible Frontiers
Market portfolios do not all end up with positive weights
At least ten per cent combinations include short sales positions
Is that a microstructure component?
33. Using High – Lows and Variance
Corwin and Schultz (2012)
Bid-ask spreads can be estimated from Daily highs and lows
The Daily high is almost always a buy
The Daily low is almost always a sell
The difference is an indicator of the daily variance
Variance is proportional to the return interval(chosen)
Spread estimates can be developed from these high-low differences
Comparable with Roll(1984) spread estimators
Hasbrouck’s(2009) Gibbs estimator computer intensive
37. Earning from the Bid Ask spread
Menkveld 2013
HFT can trade assets cross markets and earn spreads as passive traders
Akin to liquidity making trades as quasi market maker
In Continuous Double Auction markets, traders need not be market makers
Yet Bid Ask spread can earn them a steady income without crossing the
spread and in volume trades that can be sustained by new information
IN US and Europe share of dominant stock exchanges down to 20%
38. Two scale realized Volatility
Recent research by Zhang, Mykland and Ait Sahalia(2002)
Using squared returns as volatility indicator
Non parametric estimation
Breaks down realized volatility and allows reconciliation to implied volatility
Y(it) = X(it) + e
Sum of squared returns then switches to
Use overlapping sub grids from t(k-1) with different time intervals(scales)
The presence of ht as endogenous to realized volatility
39. Ait Sahalia et al
To estimate the leverage effect in volatility and differentiate/integrate the
leverage and the volatility feedback effect
Related to high frequency estimation of volatility
40. Future Directions
Berger and Yang estimate Frontier Markets diversification benefits, showing
idiosyncratic variance more than ½ of the total variance in Variance
decomposition analysis
Marshall, Nguyen and Nuttawat(Advance) Frontier markets have high
transaction costs ( Illiquid spread)
Bandi et al _ Realised Volatility vs Implied Volatility Traders use different
Volatility estimates
Cornell and Green (!991) Spread to Price Performance of High yield Bonds
41. Signal or Noise
Banerjee and Green develop a model to recognize whether traders are
informed ( rather than just liquidity makers and takers)
They relate the phenomenon back to volatility clustering
Stein(2013)talks of traders inability to recognize other traders’ motive as
creating two externalities- One is the spread moving away from
fundamentals and the second is a fire sale effect
42. Private Information
M U and J shaped patterns in FX spot markets
Patterns in bid ask, trading volume and return volatility outside noise
Strong relationship between Cumulative order flow and exchange rates
(Easley) Unanticipated deviations from intraday trading volumes
Bilateral relationships between volatility and volume, volatility and bid ask
spreads
McGroarty et al study the introduction of the Euro and relate anticipated
and unanticipated order flow and volume and all three with bid ask spread
Best bid and ask orders per second with trade data
Log prices and log squares (volatility)
Total Volume from Bid and Ask trades added
43. Measures of Private Information
Order Flow
Volume
• Expected Volume not linked to market making , yet
a basic ‘inventory’ available in all traded assets
44. Relative liquidity as a future information
measure
Valenzuela et al(advance)
Quoted depth distribution in the Order book at different times
Indicator of consensus on traded price ( can be extended to other measures of
price /valuation stability)
Higher liq provision away from the book
Disagreement on current price
High volatility
(intuition) Exploration/spurious orders from a high volume investor/trader
PCA used to identify the important quotes in the book
Use Ait Sahalia, Mykland and Zhang TSRV
Goettler Parlour and Rajan, Frequency of quotes waiting at the traded price
45. Volatility thresholds
Kasch and Caporin 2013
Correlations increase during high volatility periods reducing diversification
effects
Dynamic behavior of cross correlations during switching of regimes
Volatility threshold using both Conditional volatility h^1/2 and covariance
matrix
The threshold Xt above which the variance becomes instrumental in cross
correlations,
Not like increased cross correlations are not contagion
46. Idiosyncratic Volatility
Stock specific/Name specific volatility
Can be an indicator of acquisition premium and higher market premiums in
Information poor economies
47. Volatility and Market Depth
Ahn Bae and Chan, Transitory volatility rises and then declines on increase in
market depth
Depth increases due to increase in Limit Orders
The TV affects mix between limit and market orders
These limit orders arise without market making yet provide liquidity
Earlier studies on superdot orders helped Hasbrouck identify migration of buyers
to limit orders and the phenomena of fleeting orders
Potential loss from trading with informed trader(“market orders”)
Foucault game theoretic model of a limit order market
Glosten(1994) Patient traders: Limit Orders, Urgent traders ( Market orders)
Limit Orders have to be wider spread to avoid market traders/informed traders
in high volatility markets
48. Order Book Slope and Price Volatility
Duong and Kalev (2008)
Negative relation between future volatility and variations in the liquidity
provision in the order book
Order book slope of the buy side more informative
Institutional orders more informative
Requirement of anonymous markets proved in ASX
Indian markets proved to be anonymous
Frontrunning minimized in anonymity yet information available in the order
book transforming private information to public information
49. Trade Autocorrelation
Chung and Li Using PIN measure
Price impact of trade
Positive correlation in direction of trade
Both measure positively related to trading frequency
Quote revisions earlier studies also because of liquidity, inventory and non
information reasons
PIN better measure than use of small firms and liquidity proxies
Kyle study initially on trade size and pace of information assimilation
Institutional trading may contribute more to clustering during high volatility
before a news break
50. Volatility of Liquidity
Volatility Feedback Puzzle (Bollerslev) resolved microstructure: Zhang and
Perreira 2010
Liquidity influences returns beyond trading costs
Liquidity modeled as stochastic price impact
Liquidity Premium = Additional return necessary (Adverse Price impact of
trading)
Not required for patient investors waiting on Limit Orders
Zhang and Perreira go on to use a CRRA risk aversion model
Granger causality from Price impact to trading activity
51. Volatility discovery across the LOB
Wang(2014)
Using trading in both stock and options markets
A new picking off risk in limit orders different from adverse selection because of
Private information
Informed traders use Limit orders established in the literature
Informative of future liquidity
Directions of future stock price movements
Future stock price volatility
Price aggressiveness of the Limit Order Book
Winners Curse
Aggressiveness of limit orders related to Options Price trading
Infer volatility from Options prices and change the aggressiveness of the Limit
order book
52. Emerging Limit Order Market
Jain and Jiang(2014)
Information content of the Limit Order Book on the Shanghai Stock
Exchange
LOB consistently predicts future price volatility from the slope of the order
book
Sell orders become more informative during extreme wide movement days
( higher no. of extreme moves on the Shanghai stock exchange)
Buy orders more informative during normal market and up move days
Order submission strategies
53. Sentiment
Berger and Turtle(2015)
Short term increases in sentiment precede positive returns
Sustained increases in sentiment precede negative returns
Especially true for market portfolios and Opaque portfolios with high
uncertainty and/or higher arbitrage for private information and greater
frictions
Sophisticated investors /arbs contribute to mispricing
Build up of short interest usually quieter than the behavioral bias in buying
Diminishing bubble growth rate
54. Indian trading/investing relations of
liquidity
Simultaneous price discovery in Banknifty options and futures markets
Large price moves across a bid spread and Price impact spanning Bid – Ask
pf 300 to 500, Price impact of 25*200 per lot in a single move
Actively traded with large volumes
Selling Puts in a monotonically rising trend
Risk induces trades on the other side, ensures high spreads and low volume
( in no. of lots) trading
Most of the float from Public sector banks
55. Continued Bayesian inference
Sequential deterministic modelling updated for new information
LOB as an information store
Information lost between the cup and the lip restored by Bayesian inference
Information to trade conversion based on Matching engine and order flow
Heavily traded securities can reveal the patchwork
First in First out at best prices
First few levels are broadcast, all levels are available readily ( except of some
exchanges)
Adds, Modifies and Cancels
Explicit vs Implied ( Hidden/iceberg)
Off book trades / Bulk trades mechanisms
56. Price Discovery and Liquidity
Impact of information arrival
Newswires on intra day movements
Riordan et al(2013)
Impact on liquidity and trading intensity
Liquidity decreases around negative messages ( May increase on lower
prices)
Negative messages induce stronger reactions
HFT data increased the study of such trading by HFT traders
Hoffman, fast and slow traders
57. Latency
Hasbrouck and Saar (2011)
Technology of trading
Easy execution
Flash crash potential from mistakes
Reduction in transaction costs but increase in spreads
58. Microstructure related increase in
concentration effects
Highly concentrated industries with low competition feature more
abnormal returns factoring product market cash flows in financial markets
Competitve industries see far lower risk premiums and abnormal returns
Industry concentration links to behavioral phenomena in the microstructure
59. Frontier markets and idiosyncratic risk
Berger and Yang(2013)
As mentioned earlier far away illiquid markets good idiosyncratic risk proxies
Frontier funds more illiquid extensions of emerging markets
1/3 of fund risk idiosyncratic for Frontier funds and ½ of regional funds,
commonality decreasing
The criticality of index proxies for Passive investors in Emerging Debt trading
Marshall, Nguyen et al fixate on transaction costs (Bid Ask spread) that
negate such diversification (Liquidity risk)
Chunky Price moves
60. Returns in Private Equity ‘ Markets’
Ang et al 2014
Model Partner cash flows
Unbalanced panel and Time varying PE premium ( Liquidity and chunky
transaction moves and related risk measures)
61. Limit Order Markets
Parlour and Seppi (2007)
LOMs : No Uniform clearing price
Each order filled at limit
Market orders changed to limit after partial fill (exchange characteristic/manual
in India)
Iceberg orders in the interest of market disclosure ( during sells)
In India cannot choose as exchange after large order arrives (market order) at
least technically
Dynamic trading strategies ‘ also’ depend on monitoring changing market
conditions ( frequency of monitoring) in news and transaction volume terms
Seen as fair in light of adverse selection frictions
Asset valuations now a social activity in Economics because most economic
assets have trading proxies
Coordination problems ruled over by ease of anonymity