If You're Samurai You're Good

This post is a detailed review of benchmarks. If that's useful, read on.

A benchmark is vital for evaluating a portfolio manager’s performance against. We covered this first with tracking error/misfit risk. Specifically a benchmark can be used as:

1. A reference point

2. A guidance for fund managers or plan sponsors

3. To ID risk exposures

4. For performance measurement and attribution

5. For manager appraisal and selection

6. Marketing

7. Compliance

A valid benchmark has 7 components that go by the mnemonic SAMURAI. These are often tested in a multiple choice question. Furthermore, and this is really important, all benchmark questions should be compared against the tenants of the SAMURAI properties. Because as long as you are SAMURAI, you’re good.

-Specified in advance : Benchmark is known to all at start of evaluation period
-Appropriate : The benchmark should accurately reflect the manager’s performance style
-Measurable : You must be able to measure the results
-Unambiguous : A good benchmark’s components should be known
-Reflective of manager’s current investment expertise
-Accountable : Manager should agree that the benchmark is an appropriate measure
-Investable : You should be able to replicate and invest in a benchmark

Just as there are 7 properties of a valid benchmark there are 7 common types of benchmarks. This is captured in this table (it's an image so if you dont see it make sure you download the image in the email):

To construct a custom security-based benchmark:

1. ID the manager’s process, asset selection (cash), and weights

2. Those weights are now the benchmark weights

3. Select securities consistent w/ process

4. Assess & rebalance to replicate manager as needed or on a schedule

A few notes on manager universe benchmarks (because it has its own LOS). It fails all benchmark criteria except for being measurable. It is not investable (cannot ID the median portfolio in advance), It is ambiguous, You cannot verify appropriateness (because again, it is ambiguous), Have to rely on compiler representation of data, Survivor bias is present (underperforming managers fall off, biased upwards).

Testing Benchmark Quality

There are a number of factors to consider when evaluating a benchmark’s quality or when identifying the appropriate portfolio. On exam day, however, make sure you know these two facts:

  • Pick the portfolio whose beta is closest to 1
  • Pick the index with the lowest tracking error

Other signs of a good benchmark (If you know 2 or 3 of these you should be OK)*:

1. Minimal systematic bias relative to the benchmark

2. A manager’s security selections, (A, from M = P + S + A), should be uncorrelated with style (S)

3. Manager’s systematic risk should be similar to the benchmark, i.e. similar risk characteristics

4. There should be a strong correlation between (P-M) and (B-M)

5. A higher coverage ratio is better…because it is a closer replication of the benchmark

6. Low turnover…if the portfolio is passively managed…too much turnover makes it uninvestable

7. Low tracking error...it should be less than tracking error of Portfolio and the Market

8. It should have positive positions or exposures…if the manager is expected to be long-only

And if you know the reasons behind #2 and #4 you really understand this section