Cold Case File? Inventory Risk and Information Sharing during the pre‐1997 NASDAQ

• DOI: http://doi.org/10.1111/eufm.12119
• Date: 01 September 2017
• Published date: 01 September 2017

Cold Case File? Inventory Risk

and Information Sharing during the

pre-1997 NASDAQ

Laurence Lescourret

ESSEC Business School, 3 avenue Bernard Hirsch, 95021 Cergy-Pontoise, France

E-mail: lescourret@essec.edu

Abstract

This paper shows that dealers in Over-The-Counter (OTC) ma rkets might

choose to share information about tr ansient price pressures. Using data from the

pre-1997 NASDAQ preopenin g, I find that the frequency and magn itude of

non-positive spreads (the i nformation-sharing vehic le) initiated by wholesaler s

(specialised market-make rs with a high exposure to in ventory risk) are strongly

related to opening price revers als and daily trading imbalan ces. This activity is

more likely to occur on days o f large liquidity shocks, a nd it is not observed for

other dealers. Overall, the ob ligation to absorb price press ure at a yet unknown

opening price might induce de alers to communicate the direc tion in which

the opening price should move. The findings contain lessons for the design of

today’sOTCmarkets.

Keywords: OTC markets, preopen, NASDAQ, information sharing, price reversals

JEL classification: G12,G14,D82

A NASDAQ trader, knowing he has a large volume of buy orders to fill [...] will enter

a bid that is the same as or higher than the offer to signal his need to get the price moving

and his orders filled. (‘Crossed and Locked Markets Create a Problem for NASDAQ’,

Wall Street Journal, 18 March 1999)

This paper greatly benefited from comments and suggestions of John Doukas (the Editor),

three anonymous referees, Thierry Foucault, Jos

e Miguel Gaspar, Sophie Moinas and

participants at ESSEC Business School, and CORE Econometric seminars. Laurence is

research fellow at CREST. Laurence gratefully acknowledges Imperial College for a

visiting position in 2011 and the French National Research Agency (ANR) for its financial

support (ANR-10-JCJC-1810-01). An earlier version of this paper circulated under the title

‘Why is there heterogeneity among dealers’behavior during the NASDAQ preopening

session?’The usual disclaimer applies.

European Financial Management, Vol. 23, No. 4, 2017, 761–806

doi: 10.1111/eufm.12119

© 2017 John Wiley & Sons, Ltd.

A global probe into foreign-exchange markets has unearthed evidence that traders

frequently colluded to try to manipulate currencies [...] A trader at one bank would

accumulate all of his institution’s‘buy’orders in a specific currency. Then he would share

this information with his competitors and exchange information about their overall

positions. (‘Forex Traders Said To Have Colluded In Effort To Profit’,Wall Street Journal,

19 December 2013)

1. Introduction

The practice of information sharing among dealers in Over-The-Counter (OTC) markets

has come to the forefront of regulatory and academic debate, after recent investigations

showed that OTC intermediaries share information about clients’orders.

1

Three

noticeable characteristics are common to all these cases of information sharing. First,

they occur in non-anonymous dealer markets in which a few big intermediaries’quoting

strategies (e.g., large banks) might influence other players. Second, information is shared

during a short period preceding the determination of a ‘benchmark’price used for

executing clients’orders. Third, dealers share information which concerns temporary

order imbalances unrelated to asset fundamentals. Such non-fundamental or ‘inventory’

information (Cao et al., 2006) might have common-value components related to the

market trading environment (for instance, a liquidity shock due to a distressed trader) and

specific private-value components related to participants (e.g., inventory adjustments,

speed of trading, etc.).

2

Inventory information is valuable during trading hours,

especially in dealer markets operated by risk-averse intermediaries in which inventory

concerns are more exacerbated, like most OTC markets such as the Foreign Exchange

(FX) market, or the NASDAQ before 1997 when it was a pure quote-driven market.

3

This paper examines the impact of non-fundamental information sharing among

dealers using the NASDAQ preopening before 1997. In an important contribution, Cao

et al. (2000) find that NASDAQ dealers used non-binding preopening quotes to signal

information related to fundamentals to ease price discovery at the open. My paper aims at

complementing their findings by focusing on non-fundamental information. I argue

that dealers find it desirable to signal information not necessarily about fundamentals, but

also about order flow imbalances or inventory shocks. I therefore ask the following

questions:

Do NASDAQ dealers use preopening quotes to share transient non-fundamental

information with the market?

Are opening trading pressures related to dealers’preopening quoting behaviour?

1

Examples include the rigging of the London Interbank Offered Rate (LIBOR), the allegation

of rigging at auctions of US Government debt, or the allegation of manipulation of Foreign

Exchange benchmarks to which the opening quote refers.

2

This information is also designated as ‘non-payoff relevant’(Barclay et al., 2006).

3

The order-handling rule implemented in 1997 enforced competition between dealers and

anonymous limit order traders by requiring the inclusion of limit orders posted on Electronic

Communication Networks (ECNs) in the calculation of the best market prices (National Best

Bid and Offer). Proliferation of ECNs followed significantly altering the way NASDAQ

operated, and changing the competitive behaviours of dealers.

© 2017 John Wiley & Sons, Ltd.

762 Laurence Lescourret

I use this setting for several reasons. First, the NASDAQ in 1996 operated as a pure

non-anonymous dealer market, in which transient order flow imbalances impacted risk-

averse intermediaries holding sub-optimal positions. Several studies show that inventory

control plays a central role in the price formation of dealer markets.

4

Second, at that time,

NASDAQ dealers had the obligation to execute incoming orders at the best price, either

because they posted this best price, or because they received orders from affiliated

brokers which they had guaranteed to execute (a practice referred to as preferencing

agreements).

5

The opening was especially risky due to overnight information

accumulation and the required guarantee to fill orders at the opening ‘print’price (the

first reported trade). Trading during preopening hours was in practice not feasible due to

prohibitive transaction costs and volatility. Quoted and effective bid-ask spreads were

more than four times larger in the preopen than during the trading day, mostly due to

adverse selection costs higher than at any other periods of the day (Barclay and

Hendershott, 2004).

6

Building a position minimising inventory exposure at the open

through trading could only be achieved at an high cost. Third, to our knowledge, there

exists no other dataset on quote-driven markets in which dealers’communicating

behaviour is clearly identified. Signaling preopening quotes arose partly because, during

our time period, NASDAQ dealers were unable to use telephone calls to solicit trades due

to governmental investigations of alleged collusion.

7

Using this unique dataset is thus

very relevant for analysing dealers’information-sharing behaviour.

The empirical approach consists of investigating the impact of dealers’non-binding

quotes (the information-sharing vehicle) on measures of trading pressure at the open

such as price reversals or trading imbalances. In particular, I focus on the behaviour of

wholesalers, a group of very large dealers specialised mainly in market-making during

the period in question. Wholesalers don’t provide sell-side research coverage from which

they could obtain fundamental information (e.g., Schultz, 2003; Madureira and

Underwood, 2008). They mainly handle retail order flow during the period of the

analysis (Griffinet al., 2003), which consists of mostly uninformed trades susceptible to

waves of investor sentiment (DeLong et al., 1990). Moreover, wholesalers do not

possess brokerage arms and therefore have to pay outside brokers to get access to order

4

See, among others, Hansch et al. (1998) for London Stock Exchange data; Lyons (1995) or

Bjønnes and Rime (2005) for foreign exchange markets, or Friewald and Nagler (2015) for

US corporate data. For NYSE specialist data, see, for example, Madhavan and Smidt (1993),

Comerton-Forde et al. (2010) or Hendershott and Menkveld (2014).

5

Preferencing and one of its various forms, internalisation, refers to practices that allow

brokers and institutional investors to have prearranged transactions with a preferenced dealer

(a quote-matcher), avoiding search costs for the best price. Launched by one of the NASDAQ

wholesalers –Madoff –this practice has been highly controversial but still widespread in the

US equity and option markets.

6

Using 112 trading days in 2000, Barclay and Hendershott (2004) find that the cost of the 50

most active NASDAQ stock during the preopen was 27.7 cents, compared to 7.8 cents during

the trading day. Due to a coarser tick size during my sample period, it probably implies that

transaction costs were even higher in 1996.

7

The Department of Justice and the SEC had started formal investigations against NASDAQ

dealers in 1995 following the avoidance of ‘odd-eight’quotes shown by Christie and Schultz

(1994) that suggested tacit collusion.

© 2017 John Wiley & Sons, Ltd.

Inventory Risk and Information Sharing during the pre-1997 NASDAQ 763