FINANCIAL

MARKET
MICROSTUCTURE
TERM PAPER

110125 &
135368
CONTENTS

QUESTION ONE: .......................................................................................................................... 2

GLOSTEN & HARRIS, 1988..................................................................................................... 2

Introduction ............................................................................................................................. 2

Model ...................................................................................................................................... 2

HUANG & STOLL, 1997........................................................................................................... 4

Introduction ............................................................................................................................. 4

Model ...................................................................................................................................... 5

THE IMPACT ADVERSE SELECTION, ORDER PROCESSING & INVENTORY COSTS

ON PRICE?................................................................................................................................. 8

WEAKNESSES OF THE TWO-WAY DECOMPOSITION MODEL ...................................... 9

THREE-WAY DECOMPOSITION MODEL IMPROVEMENTS ............................................ 9

IMPLICATIONS AND CONCLUSIONS ................................................................................ 10

1
ON EMPIRICAL CONSIDERATIONS IN MARKET MICROSTRUCTURE

QUESTION ONE:
GLOSTEN & HARRIS, 1988
Introduction
This paper develops and implements a technique for estimating a model of the bid/ask spread. The
spread is decomposed into two components, one due to asymmetric information and the other
due to inventory costs, specialist monopoly power, and clearing costs. The authors explain the
adverse selection component, which accounts for information asymmetry, the inventory holding
component, related to market makers' need to hold securities, and the order processing component,
associated with executing and processing orders. The adverse-selection component is due to the
revision of market-maker expectations and has a permanent effect on all future prices, in the sense
that subsequent prices may go up or down but on average remain the same. The adverse selection
spread depends on order size and increases with quantity traded. The authors estimate the extent
to which spreads depend on order size. This order- size dependency is mainly due to asymmetric
information. The results of their estimation fail to reject the hypothesis that significant bid-ask
spreads are due to asymmetric information.

Model
The general two-component asymmetric information spread model is given by,

𝒎𝒕 = 𝒎𝒕−𝟏 + 𝒆𝒕 + 𝑸𝒕 𝒁𝒕 ……………………………………………………. (1a)

𝑷𝒕 = 𝒎𝒕 + 𝑸𝒕 𝑪𝒕 ……………………………………………………. (1b)
𝟏
𝑷𝟎𝒕 = 𝑹𝒐𝒖𝒏𝒅 (𝑷𝒕 , 𝟖) ..…………………………………………………… (1c)

𝒁𝒕 = 𝒛𝟎 + 𝒛𝟏 𝑽𝒕 ..…………………………………………………… (1d)
𝑪𝒕 = 𝒄𝟎 + 𝒄𝟏 𝑽𝒕 … .……………………………………………………. (1e)
𝒆𝒕 ~ 𝒊𝒊𝒅 𝑵𝒐𝒓𝒎𝒂𝒍 ...…………………………………………………… (1f)
{𝒇𝟏 (𝑻𝒕 ), 𝒇𝟐 (𝑻𝒕 )| 𝑻𝒕 }

2
Where, 𝒎𝒕 is the true price process, at time t, 𝑷𝒕 is the unrounded price process, 𝑷𝟎𝒕 is the
observed price process, 𝒁𝒕 the adverse selection spread component, 𝑪𝒕 the transitory spread
component reflecting order processing and inventory cost , zo, z1, co, and ct are constants and f1
and f2 are currently unspecified functions with f2 > 0.

Also, 𝑸𝒕 is the unobserved indicator for the bid/ask classification of 𝑷𝟎𝒕 (and is = +1 for an ask and
= - 1 for a bid), while 𝑽𝒕 is the observed number of shares traded in transaction t, and 𝒆𝒕 represents
any unobserved innovation or changes in true prices due to arrival of public information.

The true price innovations are of two types. The first, 𝒆𝒕 , due to the arrival of public information,
and the second, 𝑸𝒕 𝒁𝒕 , due to the revision in expectations conditional on the arrival of an order. If
𝒁𝒕 is positive, then a buy order will cause true prices to rise by 𝒁𝒕 while a sell order will cause
them to fall by -𝒁𝒕 . The adverse selection spread has a permanent effect on prices since it is due to
a change in expectations.

Expressing equations (1a) to (1e) in terms of the observed price change say 𝜟𝑷,

𝜟𝑷 = 𝑷𝟎𝒕 − 𝑷𝟎𝒕−𝟏
Defining the round off error, r, as
𝒓𝒕 = 𝑷𝟎𝒕 − 𝑷𝒕
Then,
𝒑𝟎𝒕 = 𝑷𝒕 + 𝒓𝒕

𝑷𝟎𝒕−𝟏 = 𝑷𝒕−𝟏 + 𝒓𝒕−𝟏

Thus,
𝜟𝑷 = 𝑷𝒕 − 𝑷𝒕−𝟏 + 𝒓𝒕 − 𝒓𝒕−𝟏

𝜟𝑷 = 𝒎𝒕 + 𝑸𝒕 𝒄𝒕 − 𝒎𝒕−𝟏 − 𝑸𝒕−𝟏 𝒄𝒕−𝟏 + 𝒓𝒕 − 𝒓𝒕−𝟏

3
 𝜟𝑷 = 𝒎𝒕−𝟏 + 𝒆𝒕 + 𝑸𝒕 𝒁𝒕 + 𝑸𝒕 𝒄𝒕 − 𝒎𝒕−𝟏 − 𝑸𝒕−𝟏 𝒄𝒕−𝟏 + 𝒓𝒕 − 𝒓𝒕−𝟏

𝜟𝑷 = 𝒆𝒕 + 𝑸𝒕 [𝒛𝟎 + 𝒛𝟏 𝑽𝒕 ] + 𝑸𝒕 [𝒄𝟎 + 𝒄𝟏 𝑽𝒕 ] − 𝑸𝒕−𝟏 [𝒄𝟎 + 𝒄𝟏 𝑽𝒕−𝟏 ] + 𝒓𝒕 − 𝒓𝒕−𝟏

Simplifying,
𝜟𝑷 = 𝒄𝟎 [𝑸𝒕 − 𝑸𝒕−𝟏 ] + 𝒄𝟏 [𝑸𝒕 𝑽𝒕 − 𝑸𝒕−𝟏 𝑽𝒕−𝟏 ] + 𝒛𝟎 𝑸𝒕 + 𝒛𝟏 𝑸𝒕 𝑽𝒕 + 𝒆𝒕 + 𝒓𝒕 − 𝒓𝒕−𝟏

Evaluating this expression for Qt-1 = 1 and Qt = - 1 gives the round-trip price change for a sale that
immediately follows a purchase of equal size. The absolute value of this quantity may be
interpreted as a measure of the effective spread. Its average value (assuming that et and rt both
have zero mean) is 2ct + zt. The quoted spread, the amount paid by a fully uninformed trader, is
2Ct + 2Zt.
This model was estimated using likelihood estimation. The results from the time-series analysis
are unable to reject the hypothesis that the adverse selection component is positive while the cross-
sectional analysis is unable to reject related predictions of the asymmetric information theory.

HUANG & STOLL, 1997

Introduction
In this paper, the authors provide a three-way decomposition of the spread into order processing,
inventory, and adverse information components. They show the underlying similarity of various
models and provide two approaches to a three-way decomposition of the spread. They show that
the existing trade indicator and covariance models fail to decompose the spread fully, for they
lump order processing and inventory costs into one category even though these components are
different. Inventory and adverse information components are difficult to distinguish because
quotes react to trades in the same manner under both. To separate these effects, the authors propose
and test two extensions of the basic trade indicator models. The first extension relies on the serial
correlation in trade flows. Quote adjustments for inventory reasons tend to be reversed over time,
while quote adjustments for adverse information are not. Trade prices also reverse (even if quotes
do not), which is a measure of the order processing component. This behavior of quotes and trade

4
prices after a trade is used to model the inventory and order processing costs. The second extension
relies on the cross-correlation in trade flows across stocks. This is because liquidity suppliers, such
as market makers, hold portfolios of stocks, and they adjust quotes in a stock in response to trades
in other stocks in order to hedge inventory. The reaction to trades in other stocks is used to infer
the inventory component as distinct from the adverse selection and order processing components.
The empirical results yield separate inventory and adverse information components that are
sensitive to clustering of transactions and to trade size.

Model
The unobservable fundamental value of the stock in the absence of transaction costs, Vt, is
modelled as follows:
𝑺
𝑽𝒕 = 𝑽𝒕−𝟏 + 𝜶 𝟐 𝑸𝒕−𝟏 + 𝜺𝒕 ,

(1)

where S is the constant spread, 𝜶 is the percentage of the half-spread attributable to adverse
selection, 𝑸𝒕 is the unobserved trade indicator for the bid/ask classification of the transaction price,
Pt and 𝜺𝒕 is the serially uncorrelated public information shock. Equation (1) decomposes the
change in Vt into two components. First, the change in Vt reflects the private information revealed
𝑺
by the last trade, 𝜶 𝟐 𝑸𝒕−𝟏 and second, the public information component captured by 𝜺𝒕 .

The quote midpoint, Mt, calculated from the bid-ask quotes that prevail just before a transaction,
is adjusted relative to the fundamental value on the basis of accumulated inventory. It is modeled
as follows:
𝒕−𝟏
𝒔
𝑴𝒕 = 𝑽𝒕 + 𝜷 ∑ 𝑸𝒊
𝟐
𝒕=𝟏

(2)
Where β is the proportion of the half-spread attributable to inventory holding costs, ∑𝒕−𝟏
𝒕=𝟏 𝑸𝒊 is the

cumulated inventory from the market open until time t - 1, and 𝑸𝟏 is the initial inventory for the
day. In the absence of any inventory holding costs, there would be a one-to-one mapping between
Vt and Mt,

5
The first difference of Equation (2) combined with Equation (1) implies that quotes are adjusted
to reflect the information revealed by the last trade and the inventory cost of the last trade:
𝒔
𝜟𝑴𝒕 = (𝜶 + 𝜷) 𝑸𝒕−𝟏 + 𝜺𝒕
𝟐
(3)
Where ∆ is the first difference operator.
The transaction price Pt is given as:
𝒔
𝑷𝒕 = 𝑴𝒕 + 𝑸𝒕 + 𝜼𝒕
𝟐
(4)
Where 𝜼𝒕 is the error term that captures the deviation of the observed half- spread, Pt - Mt, from the
𝒔
constant half-spread, and includes rounding errors associated with price discreteness.
𝟐
The basic regression model is a combination of equation 3 and 4 as shown below:
𝒔 𝒔
∆𝑷𝒕 = (𝑸𝒕 − 𝑸𝒕−𝟏 ) + (𝜶 + 𝜷) 𝑸𝒕−𝟏 + 𝜺𝒕 + 𝜟𝜼𝒕
𝟐 𝟐
(5)
The above model provides estimates of the traded spread, S, and the total adjustment of quotes to
𝒔
trades, (𝜶 + 𝜷) 𝟐 𝑸𝒕−𝟏 , where the adverse selection component is α and the inventory holding

component is β.
Equation (5) above is estimated using a Generalized Method of Moments, GMM and it provides
estimates of the traded spread, S. Although it is not possible to separate out the adverse selection
and inventory holding costs from this model, we can separate order processing costs from other
sources of the spread as, 1 – λ, where λ=(𝜶 + 𝜷).

Three-way decomposition of the spread

To distinguish the adverse selection component (α) and inventory component (β) of the traded
spread, Huang & Stoll first make use of the fact that in inventory models, changes in quotes affect
the subsequent arrival rate of trades. As trades reverse, quotes reverse. Consequently, under
inventory models negative serial correlation in quote changes (as well as in trades) is induced, and
this implication can be used to identify the inventory component. The negative serial correlation

6
in trades, Qt, and quote changes, ∆Mt is separate from, and in addition to, the negative serial
correlation in transaction price changes, ∆Pt resulting from the bid-ask bounce.
The model shown above is modified to reflect the serial correlation in trade flows. The conditional
expectation of the trade indicator at time t - 1, given Qt-2 is shown to be:
𝑬(𝑸𝒕−𝟏 |𝑸𝒕−𝟐 ) = (𝟏 − 𝟐𝝅)𝑸𝒕−𝟐
(6)
Where π is the probability that the trade at t is opposite in sign to the trade at t-1.
From equation (6), we can show that the change in the fundamental value will be given by:
𝒔 𝒔
∆𝑽𝒕 = 𝜶 𝑸𝒕−𝟏 − 𝜶 (𝟏 − 𝟐𝝅)𝑸𝒕−𝟐 + 𝜺𝒕
𝟐 𝟐
(7)
If π= 0.5, then equation (7) reduces to equation (1). Changes in the fundamental value, ∆𝑽𝒕 , are
serially un-correlated and unpredictable since they are as a result of trade innovations (the first two
terms) and unexpected public information releases (the last term).
Combining equation (2) and (7),
𝒔 𝒔
𝜟𝑴𝒕 = (𝜶 + 𝜷) 𝑸𝒕−𝟏 − 𝜶 (𝟏 − 𝟐𝝅)𝑸𝒕−𝟐 + 𝜺𝒕
𝟐 𝟐
(8)
Knowing that quote adjustments for inventory reasons depend on actual trades, not trade surprises
is what allows us to estimate separately the inventory and adverse information components.
Considering the expectation of ∆Mt conditional on the information available after Mt-1 is observed,
but before Qt-l and Mt are observed:
𝒔
𝑬(𝜟𝑴𝒕 |𝑴𝒕−𝟏 , 𝑸𝒕−𝟐 ) = 𝜷 (𝟏 − 𝟐𝜫)𝑸𝒕−𝟐
𝟐
(9)
Combining equation (9) and (4):
𝒔 𝒔 𝒔
𝜟𝑷𝒕 = 𝑸𝒕 + (𝜶 + 𝜷 − 𝟏) 𝑸𝒕−𝟏 − 𝜶 (𝟏 − 𝟐𝝅)𝑸𝒕−𝟐 + 𝒆𝒕
𝟐 𝟐 𝟐
(10)
Equation (10) is the analog of equation (5). By estimating equations (6) and (10) simultaneously,
we obtain an estimation of the traded spread S, the three components of the spread, α, β, and 1-α-
β, and the probability of a trade reversal π.

7
 THE IMPACT ADVERSE SELECTION, ORDER PROCESSING & INVENTORY
COSTS ON PRICE?

Two-Way Decomposition Model (Glosten & Harris, 1988):

• Adverse Selection:
Adverse selection refers to the information asymmetry between buyers and sellers and is seen to
widen bid-ask spreads, impacting transaction costs. Higher adverse selection costs are associated
with wider spreads, leading to a decrease in liquidity.
• Order Processing Costs:
Order processing costs involve the expenses incurred in executing trades, such as commissions
and fees. According to the model, higher order processing costs contribute to wider bid-ask spreads
and, consequently, lower liquidity. Increased costs of executing trades reduce market participants'
willingness to trade, negatively impacting liquidity.
• Inventory Holding Costs:
Inventory holding costs are associated with the capital tied up in maintaining a position. The model
suggests that higher inventory holding costs result in wider spreads as market makers adjust prices
to compensate for the risks associated with holding inventory. Increased inventory holding costs
can lead to a reduction in market liquidity.

Three-Way Decomposition Model (Huang & Stoll, 1997):

• Adverse Selection:
Similar to the Two-Way Decomposition model, adverse selection in the Three-Way model relates
to the impact of information asymmetry on bid-ask spreads. Adverse selection costs are considered
one of the components influencing liquidity and prices.
• Order Processing Costs:
Order processing costs, as in the Two-Way model, encompass the expenses associated with
executing trades. Higher order processing costs contribute to wider spreads and reduced liquidity.
• Inventory Holding Costs:

8
Like in the Two-Way model, inventory holding costs in the Three-Way model refer to the costs
incurred by market makers in holding positions. Increased inventory holding costs are associated
with wider spreads and reduced liquidity.

WEAKNESSES OF THE TWO-WAY DECOMPOSITION MODEL

➢ Simplification of Market Dynamics:

The Two-Way Decomposition model simplifies market dynamics by focusing primarily on adverse
selection, order processing costs, and inventory holding costs. This simplification may not fully
capture the complexity of real-world market conditions.

➢ Limited Consideration of Market Microstructure:

The model does not explicitly incorporate other aspects of market microstructure, such as market
depth, order book dynamics, or the impact of high-frequency trading. These factors can also play
a significant role in price discovery and liquidity.

➢ Assumption of Rational Behavior:

The model assumes rational behavior on the part of market participants. In reality, behavioral
factors and market sentiment can influence trading decisions, affecting price dynamics in ways not
explicitly considered in the Two-Way Decomposition.

THREE-WAY DECOMPOSITION MODEL IMPROVEMENTS

➢ Inclusion of Order Imbalance:

One significant improvement of the Three-Way Decomposition model is the inclusion of order
imbalance as a factor influencing liquidity. Order imbalance considers the relative proportion of
buy and sell orders, providing a more nuanced understanding of market dynamics.

➢ Enhanced Understanding of Price Impact:

9
By incorporating order imbalance, the Three-Way model offers a more comprehensive explanation
of how supply and demand imbalances contribute to bid - ask spread adjustments and,
consequently, price impact.

➢ Consideration of Dynamic Factors:

The Three-Way model considers dynamic factors, providing insights into how changes in order
flow and market conditions influence liquidity and price discovery over time.

IMPLICATIONS AND CONCLUSIONS

The Two-Way Decomposition Model despite its weaknesses, provides a straightforward

framework for understanding the impact of adverse selection, order processing costs, and
inventory holding costs on liquidity and prices. The Three-Way model offers a more nuanced
analysis by incorporating order imbalance, recognizing the importance of understanding the
relative proportions of buying and selling activity in the market. Both models emphasize the role
of bid-ask spreads as a key metric for measuring liquidity. Wider spreads are associated with higher
transaction costs and lower liquidity, while narrower spreads suggest a more liquid market.
Adverse selection and order processing costs contribute to transaction costs. Understanding and
quantifying these costs are crucial for market participants seeking to optimize trading strategies.

10
11