The idea behind a universe or peer group analysis is to compare a manager against its peers through the analysis of various statistical values at different percentiles or their ranking within that peer group.
The most commonly observed percentiles are the 5th, 25th, 50th, 75th and 95th. One of the most common universe analyses is to separate the universe into four groups or quartiles, where the 1st quartile represents the top 25% of managers in a universe, the 2nd quartile represents the next 25% of managers, and so forth. When graphing the quartiles, the top and bottom 5% of all managers are removed. Removing the top and bottom 5% prevents outliers from distorting the graph.
Below you can see how the graph can get distorted. The two graphs below show a peer group analysis, highlighting the three year time period with the first graph including outliers, so the 0th-100th percentiles and the second graph excluding outliers by removing the 0th-5th and 95th-100th percentiles, showing only the 5th-95th percentiles.
In comparing the two, we can see the outliers tend to distort the graph by stretching it. The one with outliers has returns from 6.32% to 71.19%, giving it a range of 64.96%. Removing outliers gives the Universe returns from 17.6% to 29.8%, dropping it to a range of 12.2%. This change provides a better range relative to a majority of the returns within the peer group over this three year time period.
Now, as mentioned, the most commonly observed percentiles are those when split into quarters, however functionality within StyleADVISOR allows one to break out the percentiles into any combination, i.e. deciles. For more information here, please contact our product support team.
We can also illustrate a peer group analysis by rank. By converting universes to ranks, the white space in the chart is eliminated. It is usually easier to see where the manager plots relative to its peers when universes are stretched out to show rankings. Shown below is the Sharpe Ratio rank.
Rolling Period Analysis
One bias in any universe comparison is a phenomenon known as end point bias. This occurs when the beginning or ending points can be chosen arbitrarily and may affect the outcome of an analysis. This can be mitigated by allowing StyleADVISOR to show all of the potential periods using a trailing period chart rather than discrete periods. Depending upon the type of analysis, end point bias can also occur when the latest period receives more weight in the analysis making long term analyses biased towards the present. Showing rolling periods rather than single periods helps eliminate this.
We can see from above ending September of 2011, Manager A has a return of -13.06%. Moving to the graph in the bottom left, we can see the trailing one year return of Manager A in the bottom quartile or the 75th to 95th percentile. This would imply that compared to his peers over the past year the manager has not performed very well. However, if we were to show a 12 month rolling window over the past year, we actually see in the bottom right graph, Manager A is in the top quartile or the 5th to 25th percentile 9 out of the last 12 months. This would imply the manager has actually performed well over the past year compared to its peers. This shows an example of end point bias, where ending September 2011 Manager A has a bad monthly return, causing a bad one year return against its peers. But, if we show a rolling period analysis, we can eliminate that end point bias and show that the manager actually performed very well against its peers over the past year.
If you have any questions on running a peer group analysis in StyleADVISOR, please feel free to call our product support team at (800) 789-5323 and hit “2” or email email@example.com