Blue Sky Group & MJ Hudson: A deep dive into transition manager results

4 nov 2021

 
Little had been published regarding transition track records, until Blue Sky Group and MJ Hudson published a paper revealing the track record of transition costs.

Track Record Data

The paper used data captured from transition managers representing around 2,200 individual transition events, executed between 2014 and 2019, worth collectively nearly $ 900 billion, including both equity and fixed income transitions (broadly excluding transitions funded from or delivering to cash). The data include:

• Pre-estimate of transition implementation shortfall
• Pre-estimate range of potential outcomes (that is one standard deviation estimate of risk)
• The actual, realised implementation shortfall result

This data was segmented by:

• Client type
• Client location (region)
• Size of event (legacy asset value)
• Legacy and target asset classes

Understandably, transition managers are protective of their performance data, so whilst most agreed to share, some did not.

Transition Benchmarking

To help understand the terminology used in this paper, we first explain a few technical transition terms. Evaluating a transition starts with an effective transition benchmark, which is universally recognised as ‘implementation shortfall’. This benchmark compares the difference in the value of the transition portfolio (legacy assets sold and target assets purchased) during the transition to the target portfolio (the portfolio to be built). This includes all direct costs (commission, taxes, et cetera) and indirect costs (market impact and bid-ask spread) of execution plus any change in value related to underlying market movement (opportunity cost). Whilst the cost of trading is always a negative impact, opportunity cost is the risk of transition and can have either a negative or a positive impact on the transition result. Opportunity cost is the primary reason for transition results to differ from estimates.

 

 

Transition managers are typically assessed by their realised implementation shortfall versus their estimate. These estimates are generally represented on a normal distribution curve (Figure 1), with a mean cost at the centre representing estimated explicit and implicit costs and a risk range (±) representing opportunity cost. Collectively, they represent a low and a high estimate, typically based on one standard deviation (σ) of outcomes. These estimates are then compared to the actual result at the end of a transition. Like ordinary asset managers, transition managers also achieve a ‘relative performance’ which we will call the transition result in this paper. There is no concept of outperformance per se – any deviation, cheaper or more expensive than the estimated outcome, likely represents greater risk exposure than forecasted. However, a transition result which delivers a profit rather than a loss is a more favourable experience.

Transition Performance Results

Firstly, we analyse whether actual transition results (versus estimates) remain within their expected risk boundaries (one standard deviation) as provided by transition managers in their pre-trade analysis. Assuming a normal distribution, one would expect approximately 68% of the events to be within one standard deviation.

Second, we investigate both the number and magnitude of events which were less or more expensive than estimated. Assuming a normal distribution, one would expect an equal number of events to deliver cheaper or more expensive results.

Data in Figure 2 shows a count of each transition’s z-score (difference between estimate and result divided by the risk estimate) rounded to the nearest 0.1, along with a theoretical normal distribution of z-scores for the same number of events. For example, the point at the top of the graph, represents 148 events which had a z-score of zero (no difference between estimate and outcome) compared to expecting 88 if the results were normally distributed.

 

 

Figure 2 clearly reveals two important observations:

• Firstly, a much greater proportion of transitions (80%) fall within an ‘acceptable’ outcome (inside 1σ) than might be ‘normally’ expected (68%).
• Secondly, a higher proportion of transitions (2%) fall at the ‘extremes’ of outcomes (outside 3σ) than might be ‘normally’ expected (0.3%).

This shows that on average transitions are likely to deliver a result within expected risk estimates more than would be normally expected¹. However, when they go ‘wrong’, there is a much greater likelihood of them being much worse than expected. The same applies to favourable outcomes. We will further elaborate on this below.

The analysis also considered the dispersion of results. Whilst this revealed an even balance of cheaper (1,106) and more expensive (1,091) results, similar to the distribution seen in figure 2, the results were not exactly symmetrical, particularly at the extremes.

In particular, 42 events were at least 100 bps more expensive than estimated, as opposed to 20 that were at least -100 bps cheaper. The overall average performance of all events was 3 bps more expensive than estimated.

So why do transition managers appear to perform better than normal expectations in some circumstances and worse in others?

• Firstly, transition managers’ expertise, experience and skill in managing costs and risks would improve results beyond normally distributed outcomes, where estimates are accurate.
• Secondly, if estimated ranges of outcomes were routinely inflated to increase the probability of showing a better-than-expected outcome to the client, then this would artificially increase the proportion of results falling within one standard deviation of risk.
• Lastly, rather than any deliberate manufacturing of estimates, transition managers may simply struggle to routinely estimate transition risks accurately.

^{1)  Assuming expectation are aligned with a normal distribution of outcomes.}

 

SUMMARY The average cost of all transitions in the survey was 36 bps with an estimated (1 SD) range of +/- 38 bps. Of the transition results 80% was within the transition manager’s estimated range of outcomes, which implies they outperformed normal distribution of 68% (1 SD). This either represents transition managers expertise routinely delivering better results than would be expected, or that the range of estimates is inflated to deliver better than expected outcomes.

Bijlagen

• FI-6 - Blue Sky Group (SW).pdf