[Limdep Nlogit List] compensating variation

[Philomela at gmx.net]

Dear all, 
i try to estimate compensating variation in NLOGIT for a MNL model:

matrix; b1=[0.914027/
2.082979/
1.821922/
1.221991/
-1.122365/
0.342412/
-0.746888]$

matrix;varb1=[0.0169,0.00854,0.00314,0.00210,-0.010,0.004246,-0.00871775,0.0966889/
0.00854,0.0812,0.0426,0.00290,-0.0215227,0.0265288,-0.0174158,0.00462964/
0.00314,0.04268,0.0614,0.002316,-0.0199,0.02375,-0.008286,-0.027/
0.0024,0.0029,0.0028,0.00075,-0.002,0.00363,0.000174344,0.0144727/
-0.01,-0.021,-0.0199203,-0.00272772,0.0263839,-0.0161,0.015078,0.0133787/
0.00,0.0265,0.0237526,0.00363972,-0.0161393,0.0546738,0.0187702,0.0193871/
-0.0087,-0.0174158,-0.00828,0.000174,0.015,0.0187702,0.0551388,0.0246523]
$
wald; start=b1; var=varb1; labels=a1,a2,a3,a4,a5,a6,a7,a8;pts=10000;
    fn1=( log   (exp (a1*7+a2*1+a4*18+a5*5.5)- exp (a1*4+a2*1+a4*15+a5*5.5)) *1/a5)$

I think there must be in error in this formula since i get negative CV and with excel (as it should be for an improvement) a positive one. Is their a mistake in the formula or any other explanation?

Thank you, 
Philio
-- 
Neu: GMX FreeDSL Komplettanschluss mit DSL 6.000 Flatrate + Telefonanschluss für nur 17,95 Euro/mtl.!* http://dslspecial.gmx.de/freedsl-surfflat/?ac=OM.AD.PD003K11308T4569a