Investments, tenth edition

  Why Not Replace the Entire TB Cake with the BL Icing? 

 This question is raised by the need to use the BL technique if the overall portfolio is to 

include forecasts from comparative economic and international finance analyses. It is 

indeed possible to use the BL model for the entire process of constructing the efficient 

portfolio. The reason is that the alpha compiled for the TB model can be replaced with BL 

views. To take a simple example, suppose only one security makes up the active portfolio. 

With the TB model, we have macro forecasts,  E ( R  

 M 

 ) and  s  

 M 

 , as well as alpha, beta, and 

residual variance for the active portfolio. This input list also can be represented in the 

 following form, along the lines of the BL framework:

   R

5 3E(R

M

), E(R

A

)

5 b

A

E(R

M

)

4

P

5 ¢0, 1 1

a

A

b

A

E(R

M

)

≤

PR

T

5 Q 1 e 5 a

A

1 e

Q

E

5 0

D

5 a

A

s

2

(

e) 5 Var(forecasting error) in Equation 27.6

 

s

2

(D)

5 s

2

(

e) 1 s

2

(e)

 

 

(27.15)

   

 where   e  is the residual in the SCL regression of Equation 27.5. Calculation of the condi-

tional expectations from Equation 27.15 as in Equation 27.13 will bring us to the same 

adjusted alpha as in Equation 27.7 of the TB model. 

 In this light, the BL model can be viewed as a generalization of the TB model. The BL 

model allows you to adjust expected return from views about alpha values as in the TB 

model, but it also allows you to express views about  relative  performance that cannot be 

incorporated in the TB model. 

 However, this conclusion might produce a false impression that is consequential to 

investment management. To understand the point, we first discuss the degree of confi-

dence, which is essential to fully represent a view in the BL model.  Spreadsheet 27.2  and 

 Figure 27.5  illustrate that the optimal portfolio weights and performance are highly sensi-

tive to the degree of confidence in the BL views. Thus, the validity of the model rests in 

large part on the way the confidence about views is arrived at. 

 When a BL view is structured to replace a direct alpha estimate in a TB framework, 

we must use the variance of the forecasting error taken from Equation 27.7 and applied to 

Equation 27.15. This is how “confidence” is quantified in the BL model. Whereas in the 

TB framework one can measure forecast accuracy by computing the correlation between 

analysts’ alpha forecasts and subsequent realizations, such a procedure is not as easily 

applied to BL views about relative performance. Managers’ views may be expressed about 

different quantities in different time periods, and, therefore, we will not have long forecast 

histories on a particular variable with which to assess accuracy. To our knowledge, no 

promotion of how to quantify “confidence” appears in academic or industry publications 

about the BL model. 

  

11

 The BL model can also be used to introduce views about relative performance of various U.S and foreign 

corporations. 

bod61671_ch27_951-976.indd   969

bod61671_ch27_951-976.indd   969

7/31/13   7:24 PM

7/31/13   7:24 PM

Final PDF to printer

---

970 

P A R T   V I I

  Applied Portfolio Management

 This raises the issue of adjusting forecasts in the TB model. We are not aware of actual 

results where analysts’ track records are systematically compiled and used to adjust alpha 

forecasts, although we cannot assert that such effort is nowhere expended. However, indi-

rect evidence suggests that alphas are usually not adjusted, leading to the common “com-

plaint” that the TB model is not applied in the field because it results in “wild” portfolio 

weights. Yet, as we saw in Section 27.3, those wild portfolio weights are a consequence of 

failing to adjust alpha values to reflect forecast precision. Any realistic  R -square that can 

be obtained even by excellent forecasters will result in moderate portfolio weights. Even 

when “wild” weights do occasionally materialize, they can be “tamed” by a straightfor-

ward restriction on the variance of the tracking error. 

 It is therefore useful to keep the two models separate and distinct; the TB model for the 

management of security analysis with proper adjustment of forecasts and the BL model for 

asset allocation where views about relative performance are useful  despite  the fact that the 

degree of confidence must in practice be inaccurately estimated.    

  

12

 Alex Kane, Alan Marcus, and Robert R. Trippi, “The Valuation of Security Analysis,”  Journal of Portfolio 

Management  25 (Spring 1999). 

    27.5 

The Value of Active Management 

  We showed in Chapter 24 that the value of successful market timing is enormous. Even 

a forecaster with far-from-perfect predictive power would contribute significant value. 

 Nevertheless, active portfolio management based on security analysis has even greater 

potential. Even if each individual security analyst has only modest forecasting power, the 

power of a  portfolio  of analysts is potentially unbounded.