How long to leave an order in the dark?

The period of extreme volatility experienced by the global equity markets in Q4 2008 accentuated a dilemma habitually faced by buy-side traders that access liquidity via dark pools.
By None

The period of extreme volatility experienced by the global equity markets in Q4 2008 accentuated a dilemma habitually faced by buy-side traders that access liquidity via dark pools. Keep an order in too long and risk the market moving away, leaving the order unfilled; haul it out too quick and you risk missing out on finding the other side that could fill the trade in full with minimal market impact. If the subsequent clean-up trade required to complete the order, perhaps using a liquidity-seeking strategy across lit and dark pools, adds significant basis points to the cost, is the trader just wasting time by trying to trade in the dark?

Many traders cut the time they let orders rest but, regardless of market volatility, optimising passive strategies in the dark requires the ability to balance the expected reduction in market impact against the expected increase in short-term alpha loss and clean up costs. A key challenge is to counteract the impact of natural adverse selection, i.e. the tendency for passive orders to ?ll quickly when they might be expected to ?ll slowly and ?ll slowly when they should ?ll quickly.

This phenomenon can arise when placing limit orders in electronic limit order books or pegged orders in dark pools. It gets its name from the ‘natural’ presence of more buyers than sellers when prices are rising and more sellers than buyers when prices are falling. Unlike the practices that buy-side traders commonly worry about in dark pools, like gaming and pinging, natural adverse selection affects all passive executions.

In practice, adverse natural selection means that a passive buy order in a stock whose value is falling will fill more rapidly than a stock whose value is rising. This is because the presence of more buyers than sellers in a rising market makes it harder for the passive order to find its counterpart. Similarly, while one would expect a passive buy order for a falling stock to fill slowly because of the negative short-term alpha attached, such orders are completed faster because of the comparative ease with which a passive buyer can find a seller. Though a general trading phenomenon, natural adverse selection is of particular interest to dark pool traders.

In a recent study (Natural Adverse Selection: Theory and Evidence*), Goldman Sachs equity execution analysts David Jeria and George Sofianos assert that the “all-in cost of passive execution strategies will be higher in venues with high natural adverse selection than in venues with low adverse selection.” Dark pools could be considered high natural adverse selection venues, as well as multilateral trading facilities and electronic communication networks, not least because of the higher speed at which some of its more sophisticated participants react to changes in stock values.

To reduce natural adverse selection, some traders try to identify and exclude better-informed traders from their list of potential counterparties. This could also mean avoiding venues considered to have high proportions of toxic flow, i.e. dark pools. Jeria and Sofianos describe such attempts as futile, suggesting traders would be better off trying to “quantify the natural adverse selection trade-off (i.e. the reduced market impact against increases in short-term alpha loss, usually reflected in the high clean-up cost on the non-?lled shares) and optimise executions accordingly.”

An analysis of around 18,000 orders placed in Goldman Sachs’ dark pool, SIGMA X, by asset managers and hedge funds over three months found that orders were generally of relatively small ADV (3%), high short-term alpha as measured alpha-to-close, and that asset managers sometimes avoided trading with hedge funds despite similarly high levels of short-term alpha. The analysts reported a 47 basis-point (bp) differential between the alpha-to-close on filled and non-filled orders as a possible measure of natural adverse selection and characterised the pool as a high natural adverse selection venue for passive orders.

To determine whether a passive strategy could be justified in a high natural adverse selection venue, the analysts compared the all-in cost of the passive strategy including the clean-up cost against the estimated cost of an alternative aggressive strategy, basing their estimation of the cost of the alternative strategy on market impact and execution horizon alpha loss. The all-in cost of the passive strategy was the weighted average of actual shortfall on ?lled shares and clean-up cost on non-?lled shares, using ?ll rates as weights.

Acknowledging that clean up trade sometimes do not actually take place, Goldman Saches compared two best and worst case estimates, observing that “the wide range between the two clean-up estimates (24 to 50 bps) underscores the difficulty of estimating the clean-up cost”. Using their own dark pool, the Goldman Sachs analysts found that the cost of the worst case scenario for the passive dark pool strategy was the same as the cost of the aggressive strategy, while the overall cost of the best possible scenario, with limited clean up costs, was cheaper by about 14 basis points. The analysis also noted that different order types can also yield different results in terms of being more or less susceptible to the effects of natural adverse selection.

Though dark pools proliferate, already taking 7-8% of equity trading volume in the US, relatively little information exists to enable buy-side traders to evaluate execution quality effectively. This represents a serious performance measurement problem as buy-side traders increasingly conduct business in the dark. If only the shortfall on filled orders is reported, the full cost of passive strategies in dark pools is under-estimated, especially when fill rates are slower than expected due to natural adverse selection. Moreover, proliferation brings diversity. Given the individual attributes of different dark pools (e.g. varying levels of toxic flow, links to other pools etc.), buy-side traders may need to take nuanced approaches in each dark pool they access. After all, it’s the orders that are supposed to be left in the dark, not the traders.

Please click on the link to register your vote on The Trade’s poll on dark pools

* “A Guide to Liquidity” Institutional Investor, January 2009