JFR Abstracts of the Forthcoming Articles: "Market Structure and Trading Volume
Anne-Marie Anderson and Edward A. Dyl"
Short Version :
Changes since 1992 have reduced the volume bias of NASDAQ but it still exists.
It has long been known that the reported volumes on the NASDAQ are overstated. This is due to double counting and inter-dealer dealings. As the describe it:
"The discrepancy between trading volumes reported in the two primary U.S. stock markets arises because Nasdaq is primarily a dealer market, whereas the NYSE is largely an auction market. In a dealer market...A dealer is therefore on one side of every transaction, which results in trading volume being overstated. When an investor sells 100 shares of firm X to a dealer, the dealer reports a 100-share transaction; when another investor buys the 100 shares of firm X rom the dealer, he reports another 100-share transaction. Only 100 shares of firm X have changed hands between the two investors, but trading volume of 200 shares has been reported."
The same 100 share trade on the NYSE would be reported as a 100 shares.
Moreover, "Trading volume can be further overstated due to inter-dealer trading." This occurs when a dealer trades with another dealer with no corresponding trade with investors. A common reason for this type of trade would be dealer inventory adjustments. Since NYSE specialists are involved in a smaller portion of trades (roughly 25% accoridng to Madhavan and Sofianos-1998) than NASDAQ dealers, this too adds to higher reported NASDAQ volume.
Thus, to compare volumes across markets, various algorithms have been used. The easiest is to merely divide NASDAQ volume by two. This idea received support by a 1997 paper by Atkins and Dyl that found when a firm moved from the NASDAQ to the NYSE, reported volume went down by approximately 50%.
However, that was then and this is now. For several important reasons, this overstating may no longer exist. For instance:
- ECN trading is making up a significant percentage of NASDAQ volume (44% in 2001). This is relevant since ECNS rarely double count their trades. However, if this increased trading is coming from dealers, then reported volumes may still go up.
- Rules changes in 1997 have led to increased use of public limit trades. These do not suffer from double counting. Hence we would see lower volumes.
- A 2001 rules change was designed to end double counting. Called "Riskless Principal Trade-Reporting Rules.....A riskless principal transaction is one where the market maker, after receiving an order to buy or sell a stock, purchases or sells the same security at the same price to fill the order. The Riskless Principal Trade-Reporting Rules require the dealer to report such a trade as one transaction."
In light of these changes Anderson and Dyl set out to see if reporting biases had changed. They examine the trading of 299 firms that changed from the NASDAQ to the NYSE in the 1997 to 2002 time period. They find that median volume does drop by about 37% which is less than the 50% number found by Atkins and Dyl (1997).
Thus, the authors conclude that NASDAQ volume numbers are still biased in comparison to NYSE reported volumes, but not by as much as they had been.
NOTE: While the paper is forthcoming, the JFR only lists abstracts. A previous version of this paper was presented at last year's FMA conference. A copy of that paper is available here.