file-o file-word file-excel file-powerpoint file-image file-archive file-audio file-movie file-code file-openoffice file-css menu googleplus facebook instagram twitter feed youtube vimeo2 lanyrd flickr picassa deviantart github wordpress blogger tumblr yahoo soundcloud skype linkedin lastfm delicious stumbleupon stackoverflow pinterest foursquare cross arrow-left arrow-down arrow-up arrow-right arrow-left2 arrow-down2 arrow-up2 arrow-right2 arrow-left3 arrow-down3 arrow-up3 arrow-right3 search

You are here

FIA PTG Whiteboard: Unintended Consequences of Frequent Batch Auctions as a Market Design

FIA PTG Whiteboard: Unintended Consequences of Frequent Batch Auctions as a Market Design

12 January 2015 6:30pm EST

The FIA Principal Traders Group (FIA PTG) Whiteboard is a space to consider complex questions facing our industry. As an advocate for data-driven decision-making, FIA PTG developed Whiteboard to sharpen the analytical tools being used to make policy decisions. 

This issue of FIA PTG's Whiteboard is being released in partnership with the FIA European Principal Traders Association (FIA EPTA). It examines frequent batch auctions and identifies areas where the application of additional data and analysis can help clarify the tradeoffs, benefits and costs associated with this idea. 

This Whiteboard is not meant to advocate for specific solutions, but rather to raise the standard of debate about key issues.  Click here to learn more about FIA PTG's position on frequent batch auctions.

Unintended Consequences of Frequent Batch Auctions as a Market Design

A recent paper by Budish, Cramton, and Shim ([BCS] The High-Frequency Trading Arms Race: Frequent Batch Auctions as a Market Design Response, 2013) proposes a potential alternative to continuous matching. According to their analysis, in the currently most common continuous limit order book (“continuous”) market matching design, there is a naturally occurring race to capture the “technical arbitrage”, or transact on stale quotes before they are adjusted to reflect new publically available information. BCS hypothesizes that liquidity providers usually lose this race and therefore have to account for this risk by posting wider bid-ask spreads with smaller quoting size than they otherwise would. As an alternative, BCS propose frequent batch auctions (FBA) – uniform-price sealed-bid double auctions conducted at frequent, but discrete time intervals. They argue that such a market design would not only decrease bid-ask spreads, but also eliminate the technology arms race, which in turn would free up capital deployed towards lowering latencies in liquidity providing operations.

This paper explains why eliminating the “technical arbitrage” via FBA would have limited benefits, and may introduce negative unintended consequences. There is a high probability that FBA would actually increase the cost of providing liquidity, such that the net effect would be wider bid-ask spreads and less liquidity than is provided in current continuous market designs.  Our goal in preparing this article is to help promote public discussion of this topic by identifying areas where the application of additional data and analysis can help clarify the tradeoffs of benefits and costs associated with the proposal.  In this article we have chosen to analyze certain aspects of the FBA proposal but, as with any complex market design change, there are likely additional aspects that merit further analysis.   

Scale of the “Technical Arbitrage”

BCS claim that FBA would eliminate the cost of liquidity provision associated with the “technical arbitrage”, and, as a result, bid-ask spreads would become narrower. While we concur that “technical arbitrage” would likely be reduced under FBA market design, BCS’s own data suggests that the magnitude of the technical arbitrage is limited and we believe that any benefit would likely be offset or outweighed by the added costs discussed in the next section.

BCS attempt to estimate the magnitude of the ES-SPY “technical arbitrage”; i.e. trading the CME E-mini S&P 500 futures versus the SPY ETF. Using CME and NYSE data from 2005-2011, they found an average of 800 arbitrage opportunities per day, with an average arbitrage opportunity having a quantity of 14 ES lots (7000 SPY shares) and a profitability of 0.09 in index points. Since 89% of the 800 arbitrages were initiated in the ES market, the total volume of SPY on NYSE potentially traded at a stale price would have been 4.98M (800*7000*0.89) shares per day.   Based on their analysis--this would equate to 10% of the average trading volume of SPY on NYSE during the same period.

On the other hand, 11% of the 800 arbitrages were initiated by price moves in SPY, which means the total volume of CME E-mini S&P 500 futures which traded at a stale price was 1281 lots per day, or 0.07% of the average daily volume.i  Therefore, the impact of “technical arbitrage” is insignificant for ES liquidity providers.

Assuming SPY liquidity providers participate in each trade and 10% of their trades cost them 0.09 index points on average as described by BCS, eliminating those arbitrages would equate to liquidity providers narrowing their bid-ask spreads only by 0.009 index points, or roughly 1/10 of the minimum increment in SPY ETF.ii

The Cost of Liquidity Provision

Two primary costs of liquidity provision are inventory risk and adverse selection risk. It is our view that under BCS’s proposed FBA market design both of these costs would actually increase, requiring liquidity providers to widen bid/ask spreads. We begin our analysis assuming a single trading venue, and then consider in the next section the implications of the proposed FBA when applied to multiple trading venues.

Inventory risk refers to the possibility that a change in price will cause the value of an inventory or accumulated position to decrease. Liquidity providers take price volatility and risk appetite into account when setting an inventory limit. As liquidity providers accumulate a position (i.e. inventory), they may hedge the associated market risk with related financial instruments in real time. Examples of related instruments for hedging purposes include different expiries of the same futures product, options hedged with underlying futures and different futures contracts along an interest rate curve.

Under the FBA market design, liquidity providers would be forced to delay hedging until the next batch auction occurs. This longer holding period, that is the time interval between auctions, even if this time is measured in milliseconds, would increase inventory risk as market participants could not immediately hedge new inventory. The frequent batch auction mechanism would also increase quantity risk by increasing the uncertainty of the quantity that liquidity providers would be able to hedge in each batch.  As risk and uncertainty increase, bid-ask spreads would be expected to widen and posted size, to decrease. Moreover, these risks and uncertainties could not be mitigated effectively by reducing the batch interval, because trades and price movements are often clustered. Hedging immediacy is needed most when positions accumulate quickly, and this occurs most often at the same moments when market changes and price volatility concentrate.iii

Adverse selection risk is closely related to information asymmetry in the context of liquidity providing. Investors who have a large quantity to execute possess information (i.e. their own large trading intention) not known to the marketplace and tend to drive the price of an asset in the direction in which they are trading. Therefore, liquidity providers face the risk of having their open orders hit by traders with superior information. Liquidity providers are exposed to adverse selection risk because they always accumulate inventory against the market’s supply and demand imbalance. Hence they tend to accumulate short positions when the market rallies and long positions when the market declines. Liquidity providers are compensated for this risk exposure by capturing the bid-ask spread. If adverse selection risk exposure increases, liquidity providers will widen the spreads they quote in order to account for this increased risk.

We believe that the FBA market design would increase the cost of adverse selection. To understand this, consider a simplified market model:

  • Investors (“customers”) only submit market ordersiv
  • Liquidity providers provide quotes around a theoretical price using limit orders
  • Theoretical price is completely driven by the net supply and demand imbalance from investors

In this simple model, liquidity providers earn half of the bid-ask spread in each instrument traded. This compensates them for the cost of adverse selection which can be measured by the change in theoretical price due to the net supply/demand. Under the FBA market design, liquidity providers’ total revenue would decrease because some investors’ orders offset each other in each auction. However, the cost associated with adverse selection stays the same because liquidity providers would still absorb the same imbalance of supply and demand as in the continuous model. To compensate for this, the break-even bid-ask spread for liquidity providers must be wider under the FBA market design.

In order to estimate the magnitude of this impact on the bid-ask spread, we use CME E-mini S&P 500 futures as an example. Over the period of 8/21/2014 – 8/27/2014, the average trading volume reduction for liquidity providers varies by the different batch intervals:v

Interval (in miliseconds)100050020010050
Volume reduction (%)2823191614

Assuming that in a continuous model, liquidity providers are on one side of each trade, this analysis shows that in an FBA model an average of 23% of trading volume would offset between “customers” if batch auctions are conducted every 500 milliseconds, for example. Therefore, the total trading volume of liquidity providers under an FBA market design would be reduced to 77% of current volume. To compensate for the fixed adverse selection cost, the breakeven bid-ask spread would have to increase by 30% (i.e. = 1/77% -1) under FBA with 500 millisecond auction

This rough estimate shows that the increase in the bid-ask spread due to adverse selection alone is on the same scale of, if not dominating, the reduction in the bid-ask spread resulting from eliminating “technical arbitrage”. This does not even include the higher inventory risk under FBA, which would require more model assumptions to quantify.

The Cost of Synchronized Implementation 

In this section we examine the issues of FBA when applied to multiple trading venues. It could be argued that a successful implementation of FBA would call for every trading venue to align with every other trading venue such that each auction period would end at precisely the same instant.  Without this level of precise coordination, idiosyncratic arbitrage opportunities may arise, reducing the likelihood that the perceived benefits of FBA would be realized. These new arbitrage opportunities could ironically lead to an increase in the technology arms race.

The cost of implementing such an effort across multiple exchanges in disparate geographic locations is likely to be prohibitively high, if not impossible. Such an implementation is complex in theory and unrealistic in practice. Further, synchronized FBA would add another potential source of coordination breakdown induced disorder.


Market structure considerations, including the mechanics of matching trades, are complex. A seemingly small change in market structure can result in significant negative and often unintended consequences and costs. While the proposed BCS FBA market design may improve liquidity by reducing some “technical arbitrage”, it may, on a net basis, decrease the quality of overall liquidity (i.e. wider bid-ask spreads and decreased posted size) due to the costs associated with inventory risk and adverse selection. Prior to implementing a change to market structure such as FBA, it is necessary to conduct a comprehensive and quantitative analysis to best understand the consequences. 


i. We only consider the front-month contract when calculating the average daily volume. (return)

ii. Assume that, as BCS describes, 10% of NYSE SPY volume is executed as part of an arbitrage and each of those has a profitability of 0.09 index points.  This profitability translates to an “arbitrage” cost of 0.09 index points on 10% of all NYSE SPY volume for liquidity providers, or, in other words, an expected cost of 0.009 index points per trade.  It is assumed that liquidity providers account for that cost by widening bid-ask spreads.  Thus, if that arbitrage was eliminated, liquidity providers may be able to tighten bid-ask spreads by the same amount.  (return)

iii. In the classic Gaussian environment, cutting the batch interval by half can reduce the variance risk by half since this scales linearly in time. It is well-known, however, that volatility clusters and the distribution of intra-day price changes is not Gaussian. Consider the extreme scenario when most of the trades and price movements of a day occur in a few bursts. In such a scenario, inventory risk cannot be reduced until the batch interval is sufficiently small. However, as the interval gets smaller, the benefit of reducing “technical arbitrage” diminishes.  (return)

iv. In practice, “customers” may submit a variety of order types including market and limit orders.  (return)

v. We use the tick by tick data from the CME market data feed where each trade is labeled as buy-initiated or sell-initiated. According to the label, we can calculate how many buy and sell quantities would offset in each interval.  (return)

vi. It is assumed that liquidity providers need to maintain an equal level of profit before and after any market design change.  In this simplified model we define liquidity provider profit as (Revenue – Adverse Selection Cost – Inventory Risk – Other Costs) where Revenue can be defined as the size of the bid/ask spread multiplied by the liquidity provider’s volume.  If volume for liquidity providers is reduced by 30% and all other aspects of the profit calculation remain constant it will be necessary for liquidity providers to increase their bid/ask spread size by 30% to maintain an equal level of profit after this proposed change.  (return)



FIA Principal Traders Group

FIA PTG is an association of more than 20 firms that trade their own capital on exchanges in futures, options and equities markets worldwide. FIA PTG members engage in manual, automated, and hybrid methods of trading, and they are active in a wide variety of asset classes, including equities, fixed income, foreign exchange and commodities. FIA PTG member firms serve as a critical source of liquidity, allowing those who use the markets, including individual investors, to manage their risks and invest effectively. FIA PTG advocates for open access to markets, transparency, and data-driven policy. For more information, contact Heather Vaughan at + 1 202-466-5460.

FIA European Principal Traders Association

FIA EPTA is an association of European principal traders formed in June 2011 under the auspices of the Futures Industry Association (FIA). FIA EPTA members consist of 25 principal trading firms that provide significant amounts of liquidity to European regulated markets and multilateral trading facilities (MTFs). FIA EPTA members deal on own-account using proprietary capital and do not have clients or act as deposit takers in any form. FIA EPTA members are authorised by various EU national competent authorities and/or supervised by the regulated markets on which they trade, and are further subject to Union and national conduct of business regulation. For more information, contact Johannah Ladd at +31 20 369 0138. 





Back to FIA