Hostname: page-component-78c5997874-j824f Total loading time: 0 Render date: 2024-11-05T09:14:03.361Z Has data issue: false hasContentIssue false
  • English
  • Français

Dealer Inventory Behavior: An Empirical Investigation of NASDAQ Stocks

Published online by Cambridge University Press:  19 October 2009

Hans R. Stoll
Get access Rights & Permissions [Opens in a new window]

Extract

1. This paper presents and tests a model of dealer inventory response. The estimated inventory responsiveness coefficient is statistically significant and its magnitude is consistent with reasonable values of underlying variables which, it is hypothesized, determine the coefficient.

2. The sign of the inventory responsiveness coefficient indicates that dealers tend to be passive and acquire shares when prices fall and sell shares when prices rise. This type of behavior is sometimes termed “stabilizing.”

3. Dealer inventories tend to increase on days prior to price declines and tend to decrease on days prior to price increases; that is, inventory changes tend to be “destabilizing” with respect to future price changes. This implies that a fraction of the public trades on superior information and that dealers tend to lose money to such information traders.

4. There is a strong tendency for dealer inventory levels to return to normal, presumably zero. The implied typical inventory holding period is about 8 to 10 trading days.

5. Comparison of NASDAQ dealers and NYSE specialists shows that the pattern of inventory responsiveness is very much the same for the two. This suggests that both act in accordance with the underlying economic model and that differential regulation has little effect on typical inventory responsiveness.

6. This finding does not obviate the possibility that individual dealers or specialists behave in atypical or undersirable ways, and that the extent of such atypical behavior might depend on the degree of public regulation of dealer activities. An exhaustive comparative study of deviations from normal behavior was not possible. However, it was possible to compare the frequency of nonstabilizing transactions in which price change and inventory change on a given day are in the same, rather than opposite, direction. One could not conclude that NASDAQ dealers had more nonstabilizing activity than NYSE specialists.

Type
Research Article
Information
Journal of Financial and Quantitative Analysis , Volume 11 , Issue 3 , September 1976 , pp. 359 - 380
Copyright
Copyright © School of Business Administration, University of Washington 1976

Access options

Get access to the full version of this content by using one of the access options below. (Log in options will check for institutional or personal access. Content may require purchase if you do not have access.)

References

1 U.S. Securities and Exchange Commission, Institutional Investor Study Report of the SEC (IIS), Volume 4.Google Scholar

2 For a description and discussion of the tick test see Robbins, S., The Securities Market, pp. 196–199.Google Scholar

3 These costs are developed in H. R. Stoll, “The Supply of Dealer Services in Securities Markets,” mimeo.

4 If the dealer holds many stocks, Q is a weighted sum of dollar holdings of all stocks where the weights are the covariance between return on the stock under consideration and each stock. The model for a dealer with one stock is used throughout the paper and this is satisfactory so long as dealer wealth is interpreted to mean wealth devoted to making a market in the single stock under consideration, and this is the amount of wealth proportionate to the stock's share in the dealer's inventory. See Stoll, Ibid.

5 Information costs were first discussed in Bagehot, W. (pseud.), “The Only Game in Town,” Financial Analysts Journal (March/April 1971)CrossRefGoogle Scholar. For a theoretical model explaining why the dealer loses, see Jaffee, J. and Winkler, R., “Optimal Speculation Against an Efficient Market,” Journal of Finance, vol. 31 (March 1976).Google Scholar

6 There are (5 days) · (2052 stocks) = 10260 stock days in the sample. (One day of the original 6 is dropped in calculating returns.) Means and standard deviations for all variables are: ΔQi,t ($1000) = −.6332 (50.05) ΔQi,t−1 ($1000) = −.5845 (49.44)rit = 1.00631 (.0349);ri,t+1 = 1.0060 (.0348).

7 Using standard statistical tests which assume normality. The validity of the statistic depends as well on the independence of residuals. An indication that residuals are independent is given by examining the Durbin-Watson statistic. The regression results reported in the body of the paper are based on a data organization which reads in turn the five observations for each stock in the order in which the stocks appear (alphabetical). The D-W statistic cannot reject independence. The regressions in Table 1 are based on a data organization which ranks stocks on volume. Again, the D-W statistic cannot reject independence.

8 It might be argued that the dealer induces the price change between t and t+1 in order to dispose of the position acquired between t−1 and t. This is not correct. The price at the end of day t reflects the cost to the dealer of taking inventory position. If the dealer believes it necessary to change prices tomorrow in order to dispose of his inventory, that price change would in fact occur today. For example, if the dealer acquires 100 shares during the day, both bid and ask prices will be lower at the end of the day (given Pe) than at the beginning, and rt < 0. The fact of both bid and ask falling will provide incentives for the public to buy rather than to sell this particular stock during the next day. If there were no change in Pe, dealer sales out of inventory during the next day would cause him to raise his bid and ask with the result that rt+1 > 0. We, in fact, observe that rt+1 < 0 when ΔQt > 0.

9 Friend, I. and Blume, M., “The Demand for Risky Assets,” American Economic Review, vol. 65 (December 1975)Google Scholar.

10 In H. R. Stoll, “Revenues of Dealers on NASDAQ,” mimeo.

11 The accounting identity for the dealer's balance sheet is given by A = L + S + W (F.1)

where

A = assets (the securities he carries),

L = liabilities (i.e., bank loans),

S = subordinated loans (counted toward capital by SEC), and

W = dealer's equity.

The SEC's definition of net capital is

C − A(l−h) − L (F.2)

where

h − % “haircut” or deduction from market value of assets to reflect their marketability.

The net capital requirement is

where R is a ratio specified by the SEC or self-regulatory body.

Subordinated loans are here assumed to be some multiple of dealer equity: S = kW (F.4)

Using (F.1), (F.2), and (F.4) to substitute in (F.3) one gets

At the point of equality, i.e., maximum borrowing by the dealer

If R = 15, h = .13, and k = l, , and a dollar of dealer equity can support 5.8 dollars of stock in inventory.

12 From IIS, U.S. Securities and Exchange Commission, vol. 4, p. 1876.

13 Ibid., p. 1877.

14 Ibid., pp. 1873–1874.