عوامل مقطعی کنترل موجودی و اثرات ظریف ADRs ها

This paper addresses the effects of international cross-listings on the domestic market by investigating the properties of market maker inventories on the London Stock Exchange. We find that the unit-root hypothesis can only be rejected for a fraction of dealer inventories, and that mean reversion is generally slow. Inter-stock variations in mean reversion can be explained by liquidity- and risk-related variables, as well as the availability of ADRs on other exchanges. We find that ADRs do have an appreciable and significant impact on dealer behaviour and trading costs. The findings have implications for our understanding of the subtleties of linkages between markets.

There have been numerous studies concerned with linkages between markets where identical or closely correlated securities are traded, such as exchanges with cross-listed securities or markets for cash and derivative instruments. A very general version of the many questions that researchers are trying to address in this context is: Is there an appreciable impact of the existence of one market on the other, and what effects does the availability of another market have on trading behaviour, liquidity and trading costs? Ideally, questions like these would be examined using data on trading by investors and intermediaries on both exchanges. Unfortunately, researchers seldom get the chance to examine such detailed trading data. Therefore, empirical studies typically look at (intra-day) changes in prices, spreads and volumes, to examine volatility spillover and price discovery effects, for example. Studies on linkages between London and international markets include DeJong et al. (1995), Pagano and Röell (1991), and more recently Werner and Kleidon (1996) (hereafter WK96). Other researchers, notably Chan et al. (1996), have examined NYSE intra-day data on cross-listed securities. Unfortunately, we do not possess the ideal data either. However, we do have very detailed data on the trading behaviour of one very important subset of market agents on the London Stock Exchange, namely the market makers. In particular, we are able to trace their inventories over time. This is useful since microstructure theory makes certain predictions about the behaviour of dealer inventories, one of these being that they exhibit mean reversion. For example, Garman (1976) models a single dealer trying to follow a zero-inventory drift policy, while in Amihud and Mendelson's (1980) model the market maker has a preferred inventory level to which she tries to revert whenever order flow shocks cause her actual inventory to deviate. Other models with similar implications are Stoll (1978) and, for multiple dealer markets, Ho and Stoll (1983) and Biais (1993). Mean-reversion towards a preferred (or mean) level implies stationarity of the inventory series. Several studies have investigated this prediction empirically; papers include Hasbrouck and Sofianos (1993) and Madhavan and Smidt (1993) for NYSE specialists, and Snell and Tonks (1998) for the London Stock Exchange.1 These studies find that the unit-root hypothesis cannot be rejected for some of the dealer inventory series. Madhavan and Smidt (1993) as well as Snell and Tonks (1998) resort to intervention analysis to remedy some of their initial methodological problems. But all the authors agree that failure to reject the unit-root hypothesis is mainly due to the low power of unit-root tests, and economic factors which are not easily taken account of with current testing technology, such as time-varying preferred inventory level. Confirming the results of the above papers for market makers in large stocks on the London Stock Exchange, we show that the unit-root hypothesis cannot be rejected for some dealers, and give a number of reasons why the stationarity property might not be exhibited very clearly in the market from which our data are taken. As in the previous literature, these reasons range from methodological shortcomings to economic forces absent in the formulation of most inventory models. From the inventory control literature, we can derive a number of predictions of the cross-sectional determinants of the degree to which market makers control their share inventories. For example, one would expect the riskiness of the stock to play a vital role in determining the speed of mean reversion that a dealer will seek to achieve. If the dealer perceives the stock returns as not very volatile, she may pursue an inventory control policy with a lesser degree of mean reversion. Total risk will also be lowered by the diversification effect inherent in a market making portfolio, which would lead to slow observed mean reversion speeds in individual stocks. Similarly, turnover in the stock will be important, even though its effect on mean reversion coefficients is not quite clear ex ante. While it would be easier and thus possibly quicker to offload an unwanted inventory position in a stock with high turnover, this possibility also makes the market maker less concerned about the inventory held. Another determinant will be the number of market makers in the stock available for risk sharing. A large number of market makers could help to spread the inventory position via inter-dealer trading more quickly, while at the same time competition for the public order flow would be more fierce, implying slower mean reversion speeds. We examine the effects these factors have on mean reversion coefficients and frequencies with which the unit-root hypothesis is rejected. It is important to note some practices and institutional features of the London market, as discussed more fully by Hansch et al. (1999) and Menyah and Paudyal (2000), may also affect market maker inventory management. Preferencing arrangements between brokers and market makers, which imply that order flow does not necessarily go to the market maker with the most competitive contemporaneous quote, may limit market maker's ability to manage inventory in the desired direction. The practice of internalization, which means that order flow is not necessarily exposed to the whole market, may have a similar effect, by limiting other market makers' access to desirable order flow. However, given that recipients of preferenced or internalized order flow would presumably want to share risks through inter-dealer trading (as documented by Reiss and Werner, 1998), the detrimental effect of these practices on inventory control by other dealers is likely limited. Finally, the existence of a mandatory quote size, referred to as Normal Market Size (NMS), which is high relative to other markets such as Nasdaq, exposes market makers to substantial risk, and may influence their inventory control policy. As can be expected, NMS is highly correlated with turnover, and the latter is used in our empirical analysis. Interestingly, the analysis shows very clearly that incidents of non-rejection are much more frequent in ADR than non-ADR stocks. Mean reversion coefficients are significantly lower, and autocorrelations higher for ADR stocks. At first glance this fact could be due to a number of reasons such as the ones just described. For example, it is evident that more liquid stocks are more likely to have ADRs, and the number of non-rejections could also be influenced by non-ADR factors such as risk, and home market liquidity. Controlling for these other factors, however, it remains the case that market makers in ADR stocks behave significantly differently from market-makers in stocks which are not cross-listed on another exchange, typically in the US. This provides strong evidence that the availability of a closely related second market (with partly overlapping trading hours) has an appreciable and significant impact on market maker behaviour. This finding stands in contrast to the findings presented in WK96. These authors study a matched sample of British stocks which are cross-listed in ADR form on the NYSE. They find that New York intra-day patterns for volatility, volume, and spreads for British cross-listed stocks closely resemble those of the otherwise similar non-cross-listed stocks, and conclude that “despite [this] information spillover, London dealers do not respond by raising their spreads” and “New York patterns for British ADRs are almost completely unaffected by the fact that British stocks have been trading continuously in London for six hours prior to New York open.” These findings are consistent with the hypothesis that market makers do not use the US markets to manage their London books to a significant extent, and implies that NYSE trading has no real impact on the London market. In contrast to this, our evidence shows that market makers propensity to mean-revert their inventories towards their mean level is significantly lower for ADR cross-listed stocks, implying that the availability of ADRs does have an appreciable effect on market maker behaviour on the London Exchange as far as inventory management is concerned. We do not claim that market makers actually do trade significant amounts in the US markets, but that they benefit from the presence of that extra liquidity regardless. A possible explanation is that the ADR trading facility reduces the risk of holding a position in a stock after trading in London closes, and increases the overall liquidity in the stock. Both effects would lead to the market maker being less concerned about inventory control than she would be if ADR trading was not available. This in turn, should lower the inventory component of the bid–ask spread (e.g., see Stoll, 1989), and transaction costs. We thus investigate the relationship between mean-reversion speed and the size of the bid–ask spread, and find them to be negatively related. The remainder of the paper is organized as follows. Section 2 briefly explains the trading system used on the London Stock Exchange and describes the data used in our study, and summarizes the properties of dealer inventories. Section 3 studies the time-series properties of dealer inventories and implements the tests for unit-roots, pointing out some of the associated problems. Section 4 investigates the determinants of mean-reversion speeds in inventories, and shows that ADR availability plays a significant role. Section 5 summarizes the main findings of the paper and concludes with some directions for future research.

This paper seeks to find effects of international cross-listings on the domestic market by looking at the time series properties of share inventories of market makers on the London Stock Exchange. Firstly we find that in this multi-dealer environment, similar to results for single dealer markets, the unit-root hypothesis can only be rejected for a fraction of dealer inventories, and mean reversion, while typically faster than on the NYSE, is generally slow with implied inventory half-lives averaging around 2.5 days. Correspondingly, autocorrelations in inventories are found to be large at low lags and only slowly decaying. More importantly, inter-stock variations in mean reversion speeds are shown to be explained by liquidity and risk-related variables as well as the availability of ADR facilities other exchanges. Thus, both complementing and in contrast to WK96, we find that ADRs do have an appreciable and significant impact on the trading behaviour of dealers on the London Stock Exchange, and on investors in terms of trading costs. Finally, intra-stock variations can be linked to market maker size. The findings have implications for our understanding of the subtleties of linkages between the world's major securities markets.