Liquidity refers to the speed and cost of buying and selling (ie. liquidating) securities in the market. A liquid market is one in which a security can be bought or sold quickly and at a low transaction cost. There are at least 2 theories which predict the impact of the operation of competing execution venues such as dark pools on market liquidity. Interestingly, they provide conflicting predictions. On the one hand, there is the so-called “competition” hypothesis which implies that dark pools have a competitive effect on liquidity providers in lit markets. The hypothesis implies that dark pools provide competition for lit markets and put pressure on liquidity providers such as dealers to reduce their bid-ask spreads in order to attract order flow. In reducing bid-ask spreads they reduce the cost of trading and therefore increase liquidity. Competing with this view is the “fragmentation” hypothesis which implies that dark pools merely fragment liquidity in the lit market. The fragmentation view implies that because order flow is split across more than one trading venue, that the amount of trading in the lit market declines. This in turn implies that the amount of time that a limit order or quote is ‘alive’ in the lit market increases and the associated increase in the holding cost or risk faced by a limit order or quote provider increases. In order to obtain economic compensation for this increase in cost, the limit order or quote provider increases their bid ask spread. Hence the cost of trading increases, and the market is less liquid.
In US markets, it has historically been argued that specialists on NYSE and NASDAQ dealers have some form of market power which enables them to quote excessively wide spreads and extract economic rents from market order traders. In such a setting, the provision of competition through dark pools can entice specialists and dealers to tighten their bid ask spreads in order to compete with alternative trading venues for order flow. Therefore, while in US markets there may appear to be some economic rationale for the existence of dark pools, given the erosion in the market power of dealers and specialists through changes in market structure in the past two decades, it is questionable that they currently enjoy significant market power. Hence, there is a question mark over whether the “competition” hypothesis is valid. Of course, just about every other exchange in the world operates an open electronic limit order book, where no one single market participant is afforded any market power and all market participants compete against each other for market order flow. In such markets, the competition hypothesis is irrelevant, and the fragmentation view is likely to hold. Consequently, there would appear to be little rationale for entertaining dark pools.
the relationship between the percentage of trading in a stock executed off market and bid ask spreads is positive and highly statistically significant.
While research on the impact of dark pools on lit market liquidity is nascent, there has been some empirical research published. The most recent research published in the USA by Professor Dan Weaver out of Rutgers University uses a very straight-forward and sensible research design to test the impact of off market trading such as dark pools on liquidity. He uses regression analysis to estimate the relationship between the bid ask spread quoted for NYSE stocks and the portion of trading executed off-market, as well as a number of control variables for other factors that typically influence the bid-ask spread. Using data for October 2010 and a sample of 1,400 stocks traded on NYSE, Professor Weaver finds that the relationship between the percentage of trading in a stock executed off market and bid ask spreads is positive and highly statistically significant, and the parameter that measures the relationship between the two variables is 0.03. Hence his finding is that the greater the portion of trading done off market, the higher the bid ask spread. In fact, the parameter he estimates implies that if 40 percent of trading in NYSE stocks is executed off-market (as it is for a large number of stocks), that the cost of trading is 1.2 cents higher (0.4 x 0.03) in NYSE lit markets. The evidence documented by Professor Weaver supports the fragmentation hypothesis for US markets, and implies that trading in off-market venues such as dark pools impair the liquidity of lit markets. I have carried out an identical piece of research for the Australian Securities Exchange (ASX) using 483 stocks in the All Ordinaries Index and data sampled from October 2010. My analysis produces identical findings for the Australian market, except the coefficient between the proportion of trading in ASX stocks executed off market and percentage bid ask spreads is much higher. The reason that it is much higher is straight-forward. Trading activity in Australian stocks is much lower than trading activity in US markets, and therefore fragmenting trading activity in a smaller market has a larger impact on that market.
Forecasts of the impact of fragmenting markets can be obtained by estimating the sensitivity between trading activity and bid-ask spreads using regression analysis. Using data for the 200 largest stocks listed on the ASX for the 3 years ended 30 December 2011, I estimated the relationship between (the log of) bid ask spreads and (the log of) trading activity. This coefficient enables me to forecast what will happen to bid ask spreads if the portion of trading activity in stocks falls due to fragmentation. The average coefficient I estimate across the 200 largest stocks is -0.13, which implies that a 20 percent fall in trading activity on the ASX due to fragmentation will cause bid-ask spreads to increase by 2.6 percent (-0.13 x 0.20). Given that the average bid-ask spread across the top 200 stocks on the ASX is 0.35 percent, then the increase in the bid ask spread if 20 percent of lit market trading activity fragments is almost 1 basis point (0.026 x 0.35). This is approximately 3 times the size of the round-trip transaction fee charged by the ASX!
allowing smaller transactions which could easily be executed on the lit market merely fragments trading and impairs the liquidity of markets.
This evidence I present above clearly supports the fragmentation hypothesis and raises the question of whether there is any rationale in allowing dark pools to proliferate. Does this mean that there is no role at all for dark pools? I do not believe so. There is a role for dark pools in facilitiating transactions which simply cannot be executed on-market. Very large or “block” transactions can often destroy liquidity in the lit market for long periods of time, and it may be better for such transactions to be executed off-market in dark pools. However, allowing smaller transactions which could easily be executed on the lit market merely fragments trading and impairs the liquidity of markets. The regulatory implications of this logic are clear. They imply that the imposition of trade size thresholds above-which transactions can be executed off-market in dark pools and below which transactions must be executed in lit markets can provide an optimum market outcome which promotes overall liquidity and benefits the overall market.
Regulatory attempts to limit trading in dark pools to transactions which obtain price improvement over bid and ask prices in lit markets are unlikely to improve liquidity in the long run.
However, regulatory attempts to limit trading in dark pools to transactions which obtain price improvement over bid and ask prices in lit markets are unlikely to improve liquidity in the long run. If bid-ask spreads increase in lit markets because of fragmentation, then transactions which obtain price improvement relative to lit markets through trading off-market in dark pools may be no better off.