Micro Attribution Overview
George Santayana once said, "Those who cannot remember the past are condemned to repeat it." Since most investment managers don't always beat their benchmark, it probably makes sense for them to learn from past decisions. This is what attribution is all about.
Looking at a list of portfolio securities may not be very helpful. What do you think might be a better approach to analyze this?
Yes!
Not exactly.
Organizing securities into groupings that share common elements is a better way.
Another way to examine the active return is to use a factor model. The one-factor market model predicts what the portfolio should return based on a broad market index.
$$\displaystyle R_p = a_p + b_1 F_1 + \epsilon_p$$
But expected return can also be calculated based on a benchmark model. The difference in the results is the portfolio's differential expected returns of the portfolio.
Why do you think the incremental return doesn't always explain the difference between the actual return and the benchmark?
Exactly!
The difference between the actual return and the incremental expected return is the result of the manager's investment skill.
No.
While the factor model can be expanded to include more factors, it doesn't guarantee better explanatory power. In other words, adding additional factors may not reduce the difference between what is predicted and what actually happens.
The difference between the actual return and the incremental expected return is the result of the manager's investment skill.
In summary:
[[summary]]
Attribution for an investment manager is narrower in scope than macro attribution and focuses on the particular investment strategy. __Micro attribution__ examines portfolio returns relative to its benchmark to decompose the active return.
Suppose the portfolio only has US common stocks. At its most basic level, the portfolio return is the weighted sum of the returns of all the securities.
Why won't this provide too much insight on a well-diversified portfolio?
No.
Investment managers use all kinds of data, so that's not the issue. But a list of security-level attribution doesn't provide much insight because the impact of any one security is relatively small in the context of a well-diversified portfolio.
Exactly!
A list of security-level attribution doesn't provide much insight because the impact of any one security is relatively small in the context of a well-diversified portfolio.
Why do you think it's very common for investment managers to group securities by economic sector when conducting a micro attribution analysis?
Exactly!
Examining securities in a way that's consistent with the investment process makes the conclusions of the analysis more insightful. For instance, if there is a weighting decision based on a particular investment theme (i.e., big companies will be investing heavy in technology) it will likely affect securities exposed to those same factors.
There are also other ways to look at the data besides organizing it into particular groupings.
Not quite.
While you'll probably find this information in marketing materials, that's a poor basis of reasoning for an attribution framework.
Think more about the investment process.
Not really.
Although most benchmark securities can be put into economic sectors, that isn't always the case. When that happens, firms can classify those securities and disclose their methodology.
Think more about the investment process.
Securities that share common elements tend to be correlated because they are exposed to many of the same factors (industry employment figures, interest rates, market capitalization, etc.).
How might such groupings be used as part of the investment process?
Exactly!
No.
Analyzing specific securities would take the same amount of effort. But grouping the securities by relevant factor exposures would make it easier for a manager to take bets away from the benchmark.
Organize securities into groupings that share common elements
Sort the securities by the largest return differences compared to the benchmark
The models don't account for the manager's skill
There will always be a difference due to model specification error
It's unlikely that any individual securities have a significant impact
Investment managers don't typically use attribution data that is too granular
Sector weights are commonly used in marketing material
Sector exposures are examined as part of the portfolio construction process
It's necessary for comparison since benchmarks provide their sector exposures
It would make the analysis of individual stocks easier
It could help a manager take specific bets away from the benchmark
Continue
Continue
Continue
