Petra Wikstrom, global head of Execution and Alpha solutions at BNP Paribas, talks to Profit & Loss about why FX TCA benefits from “a pragmatic engineering approach”.
Profit & Loss: When it comes to producing meaningful TCA, what are the big data challenges facing market participants?
Petra Wikstrom: Over the last five years we’ve seen a constant uptick in the electronification of FX, but the number of venues offering FX liquidity has increased far beyond that, which means that similar volumes are now offered across more venues.
This naturally means that when we use electronic venues for the assessment of pricing information that we need to tap into more of these venues. Then we have to interpret that data to seek a representative view of the market.
FX TCA requires a pragmatic engineering approach for drawing the right conclusions. We don’t have access to the full data pool of liquidity or prices across the market and so we have to make educated estimates from what’s available. For example, for the ECNs this means looking at a combination of the primary and also secondary venues.
From this we can estimate benchmarks, but then we have to assess which is the right type of benchmarks depending on the individual investor’s underlying rationale of the execution, and how to appropriately interpret the results.
So on the post-trade side, the biggest challenge is that we don’t have a full universe of data for benchmark calculation. But creating meaningful post-trade TCA is not an impossible task by any means.
P&L: And on the pre-trade side?
PW: On the pre-trade side, we want to be able to make an educated estimate of the performance for our future execution as a way of understanding market impact for different sized orders and execution strategies given the current liquidity at hand.
So we have to measure those liquidity conditions and the pool of liquidity that is utilised to try and get an as representative as possible view of the overall market. Now in the liquid G10 currencies, where between 80% and 90% of the market is traded electronically, there is enough data to give us a fairly good representation of the market.
The same approach is more limited in the forwards, swaps and NDF market, even though the amount of electronic NDF trading has increased in the past couple of years. But if we can get some information and data around these products we see investors are usually very interested in it, especially as combined with profound trading experience.
P&L: There seems to be some skepticism from market participants regarding bank provided post-trade TCA. This seems to centre around concerns that banks are really justifying the price they gave rather than providing a thorough assessment of that execution. What do you make of this?
PW: When it comes to post-trade TCA and post-trade evaluation of performance, I think that it’s absolutely key for any provider to be highly transparent around what universe of data is included in the report and how the benchmark is calculated. And as the market becomes increasingly transparent there are ways to essentially get a second opinion to confirm the benchmark calculation by another provider.
The advantage that sell side TCA can provide is the intellectual capital and quantitative expertise to help improve the client’s execution. So the sell side has the intellectual capital across numerous business centres – whether it’s electronic market making, automated execution, voice trading, etc, – that it can leverage to help investors gain insight about their execution process.
What this does is helps investors find areas where there are outliers or patterns that show room for improvement in their execution, which in turn creates a positive feedback loop to the pre-trade solution.
P&L: Using algos is often pushed forward as the solution for improving execution now. Why is this?
PW: There are a couple of things going on simultaneously here. Although it depends on the investor, there is generally a push for more automation of process and audit trail of execution where there can be.
Then, depending on the liquidity of the market, where we have seen significant structural changes over the past few years, some clients might choose to split an order up over time to limit market impact. But then another reason for using algos could be that a particular execution strategy will allow a firm to get easier access to a larger pool of liquidity, but then naturally adapt to market conditions as the execution is live.
Overall, I think that more firms are using algos right now as a means to limit market impact and save costs, specifically for large order size. But it’s important to remember that when we talk about large orders that everything is relative. For example, what is “large” in EUR/USD would be different if you were trading the Polish zloty or the Korean won.
The maturity and existence of these more automated strategies is really in the liquid G10 space and some deliverable EM, but that doesn’t mean that firms can’t still work an order over a period of time in a more manual fashion to try and limit market impact for other pairs.
P&L: When it comes to using algos, isn’t there a problem of being able to judge the performance of different algos against each other?
PW: There are a few points to note here. If a firm executes a large order in a certain fashion, then it’s hard to directly compare that to another strategy executed at a different point in time because the liquidity conditions could have been very different. That’s one aspect to think about, there could, for example, have been an economic event happening, which could have impacted liquidity conditions, and price volatility, and made the executions different.
Another aspect to think about is that in order to accurately judge the performance of any execution strategy, firms need a large sample of executions. Someone who has done five trades with a strategy could say that it works excellently, it could be a drift phenomenon or it could actually be excellent, but without enough statistics, it’s hard to rigorously evaluate.
Yet another thing to consider is whether the algos being compared are being used for the same objective, because if they aren’t then it could, in addition to the relative performance, be down to the trade-off between limiting market impact and taking price risk. Overall, to address the relative performance of strategies, one needs to have the same objective and use for the same-sized orders and over the same liquidity conditions.
So a passive investor might choose to work an order over a long period of time to limit market impact, and in doing so, be willing to take on some price risk. In contrast, someone else might choose to do a more aggressive execution and offload the risk more quickly – even though this will probably have a higher cost – because they don’t want to take on this price risk. So it’s important to remember that it’s not necessarily a one size fits all solution.