A System of Nonlinear General Least Square

[Elysia Guo]

Hi all!

I am trying to work with a system of nonlinear GLS, with two equations. Previous papers have used the Mishkin Test. After running the following program, I got the error r(198), and I couldn't fix it.

Specifically, I want to study the market pricing of earnings component with respect to their implications for one year ahead earnings.

The system of equations I am estimating is
1) EBXI(t+1) =a0+a1*CFO(t)+a2*NAC(t)+a4*ABNAC(t)+e
2) ABNORMALRETURN(t+1)=b0+b1*[EBXI(t+1)-{a0}-{a1*}*CFO(t)+{a2*}*NAC(t)+{a4*}*ABNAC(t)]+v


mishkin ABNORMALRETURN EBXI, predvars(CFO NAC ABNAC)
matrix betaF=e(b_fcst)
matrix score predF=betaF
matrix betaR=e(b_retn)
matrix score predR=betaR

Stata was able to run forecast equation, return equation and estimation equation, but it stops at market efficiency tests (as follows).

Forecast equation -
------------------------------------------------------------------------
Variable Coef. Std.Err. z-stat P<|Z|
------------------------------------------------------------------------
Constant 0.0000 0.0000 . .
CFO, Winsori 0.0068 0.0155 0.4354 0.6632
-1.604e+04 31048.8811 -0.5167 0.6054
e[ISIN,t] -1.0819 2.5532 -0.4238 0.6717
------------------------------------------------------------------------

Returns equation -
------------------------------------------------------------------------
Variable Coef. Std.Err. z-stat P<|Z|
------------------------------------------------------------------------
-0.0000 0.0000 -0.7114 0.4769
Constant 0.0000 0.0000 . .
CFO, Winsori -0.1732 2.3346 -0.0742 0.9409
-6.960e+04 4.646e+06 -0.0150 0.9880
e[ISIN,t] -246.0851 514.3370 -0.4785 0.6323
------------------------------------------------------------------------

Estimation summary
------------------------------------------------------------------------
Equation DF RMSE R2 Chi2 P>Chi2
------------------------------------------------------------------------
Forecast 3 623.9315 0.0001 0.7 0.87143
Returns 3 0.6961 0.0001 0.9 0.92200
Observations 11328
------------------------------------------------------------------------

Market efficiency tests
------------------------------------------------------------------------
Test Chi2 DF P>Chi2
------------------------------------------------------------------------
ISIN,t invalid name
r(198);

where ISIN is the identifier for companies, and t is not a specified variable.

I am not very familiar with stata, could any one help me out here please? Is the mishkin file correct? Or is it my syntax? Any input would be greatly appreciated.

Cheers,
Ellie

[William Lisowski]

Welcome to Statalist!

I'm not sure through what path you obtained the user-written mishkin.ado - Stata's search mishkin returns no results - so there will likely be fewer users here familiar with it than with other user-written commands distributed through the usual channels. And it is complex enough code that it is difficult to understand at a simple reading. Your syntax seems correct; it appears that the problem lies in the code itself.

A quick search turns up the author's personal web site at http://sites.google.com/site/judsoncaskey/ - with the version of mishkin.ado identical to yours in the Data section, and a request at the top of that page that problems be reported to him. I found his email address among the information given at http://www.anderson.ucla.edu/faculty...faculty/caskey I suggest you should send email to the author.

[Elysia Guo]

Hi William,

Thank you for the swift reply! I tried to contact the author but still waiting for the reply. Is there other ways to estimate a system of nonlinear equations?

I have tried the nlsur function, and it gives me an error as well. My code is below:

local missingvar "EARN1,ABRETURN2, ABNAC_CF,CF_ACC_L, CFO"

nlsur(EARN1={A0}+{A1}*CFO+{A2}*CF_ACC_L+{A3}*ABNAC _CF)(ABRETURN2={B0}+{B1}*(EARN1-{A0}-{A1*}*CFO-{A2*}*CF_ACC_L-{A3*}*ABNAC_CF)),variables(`missingvar') ifgnls

and the error I am getting is

Calculating NLS estimates...
could not evaluate equation 1
starting values invalid or some RHS variables have missing values
r(480);

I have tried removing missing variables from EARN1 and ABRETURN2 but still getting this message. Any thoughts?

Thanks again!
Ellie

[William Lisowski]

I am not by any means experienced with nonlinear systems, but one thing comes to my notice. You wrote

nlsur(EARN1={A0}+{A1}*CFO+{A2}*CF_ACC_L+{A3}*ABNAC _CF)(ABRETURN2={B0}+{B1}*(EARN1-{A0}-{A1*}*CFO-{A2*}*CF_ACC_L-{A3*}*ABNAC_CF)),variables(`missingvar') ifgnls

Did you perhaps mean to write

nlsur(EARN1={A0}+{A1}*CFO+{A2}*CF_ACC_L+{A3}*ABNAC _CF)(ABRETURN2={B0}+{B1}*(EARN1-{A0}-{A1}*CFO-{A2}*CF_ACC_L-{A3}*ABNAC_CF)),variables(`missingvar') ifgnls

I don't see any justification for including the asterisk within the braces in what I read of help nlsur although I might have missed it.

Beyond that, the error message suggests that you have missing values for one or more of your right-hand-side variables (CFO, etc.) in some observations.