Performance analysis

[ajsa]

Hi David,

Performance analysis is to test the significance of the attributed returns. but i could not find an example. Is it basically to test beta(i) is significant with t-stat of (beta(i) / (standard error(i))? Does each factor have an “info ratio”? I am asking because Schweser says “the active systematic returns and residual retunrs (from common and specific factors) can be tested for statistical signicance using t-stat for info ratios”

Thanks.

 

[David Harper CFA FRM]

Hi ajsa,

If you have Grinold, the example is Table 17-4 where he breaks down total active returns into elements (factors); e.g., active beta.
Each of those has an IR (e.g., 0.23 for active beta surprise). Then he multiplies that by SQRT(5) for a 5-year time period (1988 to 1992): t stat ~ IR * SQRT(T)

David

 

[ajsa]

Hi David,

I see.. now i put the pieces together... But looking at the table, I wonder how IR for each factor is calculated? (what 'benchmark' should be used for each factor?)

Thanks!

 

[David Harper CFA FRM]

ajsa,

I think (but I am not sure) that for each factor (e.g., active beta surprise) you have a time series
e.g., if you look at my XLS http://www.bionicturtle.com/premium/spreadsheet/8.a.3_grinold_perf_ch_17/
e.g., column (I) is a 10-month series of active beta surprise
...it requires a benchmark, mine is hypothetical, but the text uses S&P500;...
i *think* the return is just the average and the risk is just the stddeviation of the series, so the IR is not our alpha-IR, but just an IR for the series of the factor.
then i think return/risk = IR (differs only due to rounding)

David

 

[ajsa]

Hi David,

In the table for the first row (Active beta surprise), return/risk=0.02/0.16=0.125 <> 0.23 (IR). so it seems not that way..

Thanks.

 

[David Harper CFA FRM]

...that's a good point, rounding doesn't explain that first one...hmmm...totals hold up ...not sure...David

 

[ajsa]

OK, thanks anyway David!