This article was written by Mike Googe, BTCA Product Manager at Bloomberg.
Trading cost analysis (TCA) is now an essential activity for fixed income traders seeking quantitative execution quality insights. This is driving a virtuous circle where the demand for higher quality analytics is pressing change in market practices and micro structure. The next step is to introduce greater context to allow relative performance to be considered. With competition for business increasing, firms are being challenged to monitor and most importantly demonstrate they achieve the best trading performance. But how do you know what is ‘best’?
Relative measurement
When you make a trade, there will be trading costs, but what’s the best way to measure those costs? The basic principle in TCA is to compare your achieved execution price to some observed benchmark, such as overall order slippage from arrival time to completion, or capturing the bid/ask spread impact using a far touch benchmark. These measures provide valuable insight but do so in absolute terms. They don’t tell the whole story, and it isn’t as simple as saying positive performance is ‘good’ and negative performance is ‘poor’. Whatever the performance you can always ask ‘Could I have done better’?
One approach to gaining perspective is to use relative benchmarks. These generally fall into two categories. Predictive pre-trade estimates which model data to predict a likely trading cost, and peer benchmarks which represent the aggregation of performance for a community.
Peer benchmarks
Peer benchmarks allow traders to compare their absolute performance against the aggregate performance sourced from a community of contributors. But what are peer benchmarks? To clear up a misconception, peer is not about trying to cluster similar firms together. Doing so can dilute the sample data in to a set of peer groups. Each firm is unique and trades and operates in a distinct way. Peer is therefore a peer of order characteristics which are clustered together to make meaningful comparisons.
Creating useful fixed income peer data has unique challenges due to the liquidity and seasoning characteristics of bonds so care is needed to select the most appropriate benchmarks and groupings to use. When creating peer data for fixed income, it’s critical to make an accurate comparison. For example, grouping by buys or sells produces lots of data, but doesn’t yield anything of value. However, creating a peer group at the level of an individual bond issue might not provide enough data,and also raises the risk of information leakage. The key characteristic to deliver meaningful insight is achieved by balancing groupings such as rating, maturity and order difficulty to allow clustering orders with similar characteristics for a comparison. Now you have the basis for a meaningful absolute to relative comparison, and are getting closer to answering ‘is this a good trade?’
Another key feature of a useful peer data set is ranking which shows where your absolute performance sits in comparison to other contributors in the selected grouping. Just because you outperformed the peer by 3 bp in our example doesn’t tell you if you are in the top 5 percent of contributors or just above the mean. Ranking therefore shows where in the peer group you sit. Peer data quality is critical for confident usage. Enforcing a minimum number of orders, curated for possible outliers, and contributors, whilst excluding your own contributions before a benchmark can be calculated will deliver reliable results.
Finally, peer utility doesn’t stop there. It can also inform best execution monitoring. Suggesting performance thresholds based on the tails of the peer data can give a solid frame of reference to establish a defendable best execution policy.
Putting costs in context
After measuring your performance within the peer group, your next step is to analyze your costs. To make meaningful comparisons, you need context, and this can be achieved through thoughtful grouping of results and using the spread to benchmark to gain perspective.
As with the initial peer analysis, the group you compare your costs with is important. The basic characteristics of an order are the issue, the size and quantity, and the side of the trade (buy or sell). However, there are many other characteristics you can compare trades on, such as maturity and credit rating, and these can also be overlaid with broader market variables such as market volatility, momentum or bid/ask spread to deliver higher-resolution analysis. For example, comparing performance where maturity is long- versus short-dated, or during high- versus low-volatility market conditions, might give you a more accurate picture of trade quality. It might not make sense to set the same threshold for trading performance when market volatility is high as opposed to low.
Proper grouping is the first step in building context but this can be significantly reinforced by adding other relative perspective such as the spread to benchmark and comparison to an index.
Spread is ubiquitous tool for pricing bonds but can also help inform your TCA. For example, if an order arrives at 10am we can observe the arrival price and use this to calculate cost against the final execution price. Let’s say this resulted in a -35 bp slippage. In reviewing your post trade TCA you see execution time was 10:15am and in that 15 minutes the price moved against you. When you break the analysis down you see that performance at time of execution showed only a -3 bp slippage. Execution relative to the market at the time of execution was decent but still an overall order cost of -35 bp raises questions. Here is where you can utilise the spread to benchmark for more context. Calculating the spread at order arrival time we see that it was 25 bp over its default benchmark. If we perform the same operation at execution time and see that the spread remained constant then you can represent this as real slippage in absolute terms but critically you can now explain the spread target was maintained over the life of the order.
The spread to benchmark can perhaps be thought of as an index of one security. The next step up is to draw that comparison with an index to which the bond issue belongs such as the Bloomberg Barclays BRAIS. For example by comparing the move in price for your target issue to its related index over a given time range before the order arrived or after the order was completed you can infer whether the price of the target issue under or over performed the index. This informs if the issues you are trading are moving adversely or favourably in relative terms not just absolute. This extra layer of refinement can inform on how best to manage/react to the behaviour of order originators (e.g. portfolio managers or customers) and then demonstrate whether it is issue specific or market wide.
So was that a good trade ?
Fixed income analytics are evolving rapidly as participants seek to unlock the insights that allow them to improve their performance and demonstrate value. By using emerging relative measurements such as peer benchmarks or making use of more refined context data such as the spread to benchmark or index comparisons, participants now have a clearer frame of reference to understand and explain their post trade analysis.
So the question you need to ask yourself is: Have I made the best possible trade yet? Even with the best available data and the most sophisticated analyses, it’s a difficult question to answer. However, by comparing your trade to a well-constructed peer group, and by putting your costs in context, you can get much closer to “yes.”