time-period bias

[tia5882]

can someone please explain it to me? i am not graspong the concept correctly
thanks!

 

[ditchdigger2CFA]

Rather than Jan 1st to December 31st performance history, a shady company will say, “We held a portfolio return of 59.3% from October 12th 1998 to April 2nd, 2005.”
Data mining is present to make the time period look great, however this is not easily compared or contrasted to other firms.

 

[daj224]

also, time period bias in arith vs geometric returns. in arithm rtns, timing of money inflows can overestimnate performance. this is in the quant section

 

[kuromusha]

ditchdigger2CFA wrote:Rather than Jan 1st to December 31st performance history, a shady company will say, “We held a portfolio return of 59.3% from October 12th 1998 to April 2nd, 2005.” Data mining is present to make the time period look great, however this is not easily compared or contrasted to other firms.

My understanding of time-series bias is that it should be short term so that the bias only exist in the short period and not in the long run. 1998 to 2005 is too long for time period bias to exist. What you’re saying is simply a quoting method, not a bias.

 

[cdealbuquerque]

One would prefer to use a full business cycle to be able to make meaningful comparisons.

 

[Flashback]

Using sample serie which is too short or not seasonally adjusted or even too long sample time series can cause biased estimators. The example is beta coefficient calculation.