Exchange Rate Economics: Theories and Evidence

354 Market microstructure approach
to pay a liquidity premium for the service of predictable immediacy. However,in
this model traders have inside,or asymmetric,information about,say,a potential
arbitrage opportunity,and can engage in favourable speculation against the market
maker. The adverse selection arises because in a market with competing market
makers,the one who interacts with the trader with inside information is the loser.
In a decentralised market with no consensus on price,setting a single price is a
dangerous strategy for a trader because he is vulnerable to the inside arbitrage
opportunity. By setting a bid–ask spread the market maker allows a tolerance for
error and it is easier to get spreads to overlap (which would represent a consensus)
than to get a scalar price to overlap. The spread therefore gives the market maker
some degree of protection from adverse selection arbitrage. Bagehot (1971) argued
that the losses incurred by the market maker from the better-informed traders
must be compensated by the less well-informed traders,and this idea has been
formalised in the asymmetric information models of,for example,Admati and
Pfleider (1988) and Subrahmanyam (1991) and we consider these models and their
predictions here.
In the model of Admati and Pfleider (1988),there are three types of agents:
informed traders,who only trade on terms advantageous to them; discretionary
liquidity traders who must trade over the trading day,but have some discretion
at what point in the day to trade; non-discretionary liquidity traders who must
trade at a given time during the day,irrespective of cost. In high volume periods
both informed and discretionary traders are attracted to trade. The informed are
attracted because they are better able to disguise their activity due to the behaviour
of the uninformed traders. The discretionary liquidity traders are also attracted
at this point because the increased activity amongst the informed traders implies
increased competition amongst them and the cost of trading is lowered for the
uninformed. Using this model they are able to explain some of the stylised facts of
the NYSE stock market: the high volume exhibited at the open and close of trade is
explained by the earlier kind of equilibrium mechanism while the concurrent high
variance at open and close follows on from the increased activity of the informed
traders who exploit previously private information.
Subrahmanyam (1991) builds on the model of Admati and Pfleider (1988) to
show that their key result,that increased activity by the informed trader lowers
the costs to the uninformed who pay the price of the informed,depends on the
assumption that the informed traders are risk neutral. If,in contrast,the informed
traders are risk averse then Subrahmanyam shows that increased activity by them
actually leads to an increase in the trading costs of liquidity traders.
A further way in which this has been modelled is in terms of inventory considerations.
For example,in the so-called random walk inventory model (starting with
Barnea and Logue 1975) the market maker has a desired inventory level (equal
to zero) and a constant spread is shifted up and down on a price scale to ensure
that the expected change in inventory is always zero; hence the level of inventory
follows a random walk. However,this has the unpleasant consequence that the
market maker inevitably becomes bankrupt. This follows because market makers
face finite capitalisation levels which,in turn,force upper and lower bounds on