Hello, dear friend, you can consult us at any time if you have any questions, add WeChat: daixieit

Economics of Financial Markets

Market Microstructure, Final Exam

1.  Consider the Kyle (1985) model, but assume that instead of a single informed trader there are N > 1 informed traders, who all perfectly observe the final value of the security v but not the equilibrium price at the time that they determine their quantity demanded qi .

(a)  Suppose that market makers post the price schedule described by equation p(q) = µ + λq, where q is the net order flow      xi + u and µ = E[v]. Assuming that each informed trader uses the following order submission strategy:

xi  = Xi (v) = β(v 一 µ)  for i e f1, .., N{,

find the value of β for which a Nash equilibrium exists, determine how β is affected by N , and explain intuitively why.

(b)  Suppose now that investors follow the order submission strategy derived in step (a).

Show that in this case the market makers’ pricing strategy is given by equation p(q) = µ + λq, and find the value of λ that they optimally choose.

(c) What is the market depth in equilibrium, and how is it affected by an increase in the number of informed traders, N? What is the economic intuition for this result?

2.  Consider a modified version of the static model by Kyle (1985) where market makers conjecture that the informed trader, in equilibrium, may have the incentive to include a random component in his order, thus increasing the noise in the market.  Hence, their conjecture about the trading strategy of the insider is:

x = β(v 一 µ) + φ,

where φ is normally distributed with mean zero and variance σφ(ψ), (that is φ ~ N(0, σφ(ψ)) and it is such that Cov(φ, x) = Cov(φ, u) = Cov(φ, v) = 0).

(a) Assume that competitive market makers post the following price schedule: p(q) = µ + λq,

where q = x + u is the net order flow.  Find the competitive value of λ.  Is market depth higher or lower when compared to the baseline Kyle model? Explain intuitively why.

(b)  Solve the maximization problem of the informed trader, that is the value of x that

solves

max E [x(v 一 p)|v]

x

when the informed trader conjecture that the price schedule is p = µ + λq. What is the optimal value of φ, in equilibrium? Explain your answer.

3.  Consider the static model by Kyle (1985), where (i) market makers are risk neutral and perfectly competitive, (ii) the asset value is v ~ N (µ, σv(ψ)), (iii) the informed investor’s order is x = β(v 一 µ), and the noise traders’ order is u ~ N (0, σu(ψ)), independent of v; and (iv) market makers only observe the total net order q = x + u.  Suppose that the insider is risk-averse (with constant coefficient of absolute risk aversion b > 0) and that he liquidates any amount of the security that he buys at a liquidation value v + ∈, where ∈ ~ N (0, σe(ψ)), independent of u and v. At the time of trading the informed investor knows the realization of v but not that of ∈. This noise in his signal implies that in taking long or short positions based on his privileged information, he bears some risk.

(a) Write down the mean-variance objective function of the insider, and obtain his de-

mand function from the solution of the first order condition associated to the insider’s expected utility maximization problem. Identify the insider’s aggressiveness β RA .

(b) Execution risk is the risk that a transaction will execute at a price that is very far away from recent market prices. Denoting by βK  the insider’s trading aggressiveness in the baseline Kyle (1985) case, show that β RA   < β K   (you can assume that the equilibrium value of the market impact is positive: λ > 0). What is the intuition for this result? How does it relate to execution risk? What happens to market liquidity?

4. In the lecture, we have investigated the situation in which order-processing costs are assumed to be γ per share traded. Consider the following alternative assumptions:

(a) Assume that order-processing costs are k per transaction.   Compute the bid-ask

spread in this case and show that it is decreasing with the size of the transaction. Which features of the technology of trading would lead you to think that this is a realistic model of order-processing costs?

(b) Assume that order-processing costs are k per euro traded.  Show that the absolute

bid-ask spread is increasing in the security’s underlying value and the relative bid- ask spread is constant, in contrast with the expressions found in the lecture where order-processing costs are a constant γ per euro traded, irrespective of the share value.

 

Solutions

1. Kyle model with multiple informed traders.

(a) Informed trader i chooses a quantity xi  that maximizes his expected profit:

max E[xi (v  p)]

xi

where he realizes that his quantity demanded will affect the equilibrium price, since it is determined by the total demand q =     xj + u expressed by himself, his N 1 peers and the uninformed traders:

p = µ + λ [xi + (N 一 1) β (v 一 µ) + u] .

Observing that E[u] = 0 and substituting this expression for the expected market price into the maximand:

max xi [(v  µ)(1  λβ(N  1))  λxi]

xi

and taking the first-order condition with respect to xi  yields

1  λβ (N  1)

Equating coefficients with the informed traders’ assumed order submission strategy:

1  λβ (N  1)

yields the desired result.  As usual in Cournot Nash competition, the greater the number of competitors the less each one demands, though the sum of their demands is increasing in N .

(b) The competitive market makers set the price equal to the expected value of the

security, given the demand from both informed and noise traders:

p   =   µ + E ┌ (v  µ)|q = u + (v  µ)┐

N         ψ

( (N)(N)|(λ)ψ σv(ψ) + σu(ψ)

Equating the slope coefficient to the posited value λ:

λ =   ==  λ =   .

(c) This expression is decreasing in N > 1.  The more insiders compete with each other, the deeper the market, as λ decreases.

2. Kyle model with a noisy insider’s order.


(a) The conjectured trading by the informed trader is x = β(v 一 µ) + φ, so that the

net aggregate order flow is q = β(v 一 µ) + u + φ.  Hence market makers’ optimal

inference is

E[v|q] = µ + q = µ + q.

The zero-profit condition E[v|q] = p then implies

p = µ +         βσv(ψ)               q.

 

λ

Hence market depth is:

1       β ψ σv(ψ) + σu(ψ) + σφ(ψ)

=

λ              βσv(ψ)               .

Hence, we see that market depth is higher than in the baseline Kyle model. There- fore, by creating some noise, the informed trader can actually reduce the price impact of his order.

(b) As in the standard model by Kyle, the informed trader will choose x so as to

maxE [x(v 一 p)|v] = xE [(v 一 p)|v]

x

whose first-order condition is

v 一 µ = 2λx,

that is,

x = (v  µ) = β(v  µ)

Clearly, the optimal value of φ is zero, in equilibrium, since it can’t be optimal for the informed investor to add a noise component to his order, thus deviating from the profit maximizing order, even if this increases the liquidity in the market.

3. Kyle with a risk-averse insiders.

(a) The informed trader knows v but is uncertain about ∈ as well as about the (unrelated)

impact λu of the noise trades on the market price.  Assuming a linear price p = µ + λ(x + u), due to CARA and normality, his objective function is therefore given by:

E[x(v + ∈ 一 p)] 一 Var[x(v + ∈ 一 p)] = x(v 一 µ 一 λx) 一 x (σψ e(ψ) + λψ σu(ψ)).

Differentiating the above function with respect to x, equating the derivative to zero and solving for the insider’s strategy yields:

1

x =                              (v 一 µ).

↘                                                                               -

β RA

(b) The insider’s aggressiveness in the baseline Kyle model is given by βK  = 1/2λ, and

the result follows. Note that ó6hb execution risk and final valuation risk reduce the insider’s willingness to trade, given the market illiquidity equilibrium coefficient λ .


4. Alternative assumptions on order processing costs.

(a) If order-processing costs are k per transaction, they are k/q per share traded. Hence,

just replace γ = k/q in the expression for the spread, to obtain

St  = 2  + st(a) + st(b) ,

so that, in the absence of adverse selection (st(a)   = st(b)   = 0), the bid-ask spread is inversely proportional to shares traded.   Fixed costs per transaction can arise if the trading technology requires physical delivery of securities or paperwork for each trade. Electronic trading has reduced these fixed costs per transaction.

(b) If order-processing costs are k per euro traded, they amount to kp per share traded,

so that the transaction price is given by:

µt      

pt  = µt + kpt dt   == pt  =

where dt  e f一1, +1{. Hence, the bid and ask are given by:

µt                       µt    

 

and the bid-ask spread is

St  = 1  k ψ ,

so that in this case the bid-ask spread is increasing in the security’s underlying value. The relative bid-ask spread is instead constant:


= st  = µt


2k

1 k ψ .