High Frequency Trading in Brazil: Mirage or Miracle?

Christian Zimmer, Head of Quantitative Trading and Research, and Hellinton Hatsuo Takada, Quantitative Trader, of Itaú Asset Management reveal the truth about high frequency trading in Brazil.
Conference panels, discussions and articles on High Frequency Trading (HFT) generally start with its definition. The term HFT is like ‘Cleopatra’ – sexy and mysterious and everyone is keen to know more about it. But the term HFT speaks for itself, so is it wasting time to go over it again?
Probably, because the term ‘high’ only has meaning relative to an external point of reference, just like cold, hot, sweet or other adjectives. This subjectivity is all the more interesting, as it is extremely difficult to measure an investor’s  brief holding period in most financial markets and, therefore, determine if it really is ‘high’. Unlike in the US, where the exchanges do not register the origin of the trade, Brazilian regulation allows BM&FBOVESPA to identify the final client on every trade. Consequently, it is much easier to measure the holding period of an investor for each asset. Also, this rule is the means by which the exchange determines whether an investor’s trade is classified as a ‘day trade’ and is thus eligible for reduced fees.
Naturally, BM&FBOVESPA does not classify a trader opening a position in the morning and closing it at the end of the day as a high frequency trader. There should be far more trading than this to qualify as HFT.  But how much more? It depends on the exchange’s criteria and reference point for ‘high’.
Figures for HFT published by BM&FBOVESPA in their April 2011report show 3.9% of the BM&F segment is high frequency and 5.9% of the BOVESPA segment. Consequently, the reduced fees are presented to the Brazilian trading community as less of an issue, as they say there is evidence of HFT taking hold. But HFT volume is not really increasing and is still far off the US figures which are often cited at around 60-70%. After carefully observing BM&FBOVESPA market prices, it is easy to conclude that it would take some time (possibly hours) to have a change in the prices sufficiently large enough to pay the transaction costs.Remember that HFT strategies are very sensitive to transaction costs.
Our suggestion is to step away from making subjective references to ‘high frequency’. Instead, one should look at the underlying trading strategies. The incentives an exchange should create to attract flow must be adjusted to the strategies that are really needed. Each strategy deserves a different set of policies and this will help the diversification of the traders’ strategies.
A trader using a market maker strategy can live with exchange fees as long as the bid-ask spread is sufficiently high. If the spread narrows, the costs become crucial and the exchange must lower the fees in order to keep this client in the market. On the other hand, a directional trader has different issues; if the fees are high, a trader must wait longer for a relevant price move so that they can capitalize on their position. Contrary to the market maker, the directional trader loves to see narrow bid-ask spreads. There would be no need to lower fees when the spread is close. The same is true for the statistical arbitrage traders.
When looking at the third party analyses of HFT in the international markets, we often see that the most common strategy is the market maker approach. This fact is strongly influenced by market fragmentation, which we do not have in Brazil. Fragmentation creates new inter-market trades, which could qualify as arbitrage trades, but not necessarily as market maker trades. Fragmentation also makes exchanges and other venues compete for the customers that provide liquidity and, as a result, give incentives to market makers. As mentioned above, Brazil does not have a fragmented market and BM&FBOVESPA does not see it necessary to ask for more liquidity. At least not as long as international capital flows are strong and increasing. Liquidity is needed in second tier shares and below.
It remains to be seen whether the inventive BM&FBOVESPA program to exempt the officially designated market makers from exchange fees will be enough to stimulate other participants to trade. At least theoretically, this provides an entry/ exit point for statistical arbitrage traders. However, as long as the allowed spreads can be as large as 1%, the strategy might not be necessarily profitable. At this moment it is worth noting that most of the Brazilian statistical arbitrage trades are long-short trades in stocks focusing on preferred-common stock relationships (in Brazil they are known as PNON, with PN standing for preferred stocks and ON for common ones).
It is also interesting to look at statistical arbitrage trades that are latency dependent, i.e. true arbitrage trades. Are these the ‘true’ high frequency traders? If there are only a few trading opportunities per day, it does not seem as if BM&FBOVESPA could classify them as high frequency. Latency sensitive traders typically use what the exchange refers to as the DMA3 (clients directly sending orders through a connection to the exchange) or DMA4 (co-location) categories. Trades through these categories can easily be measured. Unfortunately, the ability to measure the latency sensitive flow is lost because the DMA3 category is also used for any direct sponsored customer trades, so all that remains is to  measure the flow from the co-location model.
If we use the DMA4 numbers as the reference point for HFT, then we reach a HFT participation figure of 2.8% in the BM&F segment and about 2% in the BOVESPA segment (as at April 2011). The BM&FBOVESPA DMA4 measurements are significantly lower than their HFT percentages. This suggests they accounted additional strategies into this pool, such as market making strategies. Theoretically market makers could have contributed to this figure, but because of a very narrow spread in the high volume stocks and high fees, it is reasonable to assume that the market making strategy does not contribute too much to the HFT volume.
One might argue that there are still the directional trades. Yet, as this strategy needs a certain price move before it can make money and the number of trades per day is limited. On the other hand, the number of traders that might be using this strategy is not limited, as the models are nearly all different. There are only about ten Brazilian players able to successfully run intra-day directional trades. Perhaps we should conclude that the international players have better models or a better understanding of the market?
Recently, BM&FBOVESPA announced a new pricing model for high-frequency traders, which uses the Average Daily Trading Value (ADTV) to calculate fees in its equity market. Fees range from 0.019% for R$20 million ADTV up to 0.01% for firms trading over R$500 million ADTV. Ironically, almost no firms were able to qualify as ‘high frequency’ players within the exchange’s cost reduction program.
*     *     *
Open For Discussion: Click here to see what readers had to say about HFT in Brazil and hear the author’s response.