COMPARISON OF REGRESSION ESTIMATES:
ORIGINAL FIVE AND "NEW 3" MEMBERS OF THE
EUROPEAN COMMUNITY (Pooled GLS/ARMA)
Data Basis: EU8 ( Original Five versus New Three )
Source: Eichenberg and Dalton, 1993, p. 527.
Dependent Variable: See foot of Table.
____________________________________________________________
3. 4.
Original Five New Three
1973-1988 1973-1988
N=155 N=93
____________________________________________________________
COEFFICIENT T-RATIO COEFFICIENT T-RATIO
____________________________________________________________
R2 .674 .523
CONSTANT 29.064 2.591 -55.735 -3.631
INFLATION -1.870 -3.026 -1.883 -2.996
UNEMPLOYMENT -.100 -.366 -.068 -.153
GDPINDEX 8.322 .891 13.039 .966
INTRA-EC EXPORT % .322 3.106 1.150 10.336
EAST-WEST CONFLICT -.005 -1.058 -.013 -1.741
UK REFERENDUM 32.593 3.873
DANISH SEA 24.577 3.707
IRISH SEA 11.636 1.223
EC ELECT '79 5.374 2.651 1.580 .439
NETHERLANDS 14.945 6.301
ITALY 22.358 10.030
FRANCE 3.650 1.484
____________________________________________________________
DIAGNOSTIC INFORMATION BY CROSS SECTION
____________________________________________________________
Residual | Residual
Mean RVR | Mean RVR
|
BELGIUM .174 1.276 |
GERMANY -.145 1.138 |
FRANCE -.192 1.400 |
ITALY .004 .621 |
NETHERLANDS .004 .566 |
DENMARK | .033 .338
IRELAND | .031 .887
UK | -1.537 1.775
____________________________________________________________
Explanatory Notes
RVR is a "residual variance ratio", the ratio of actual residual variance for a cross section to the expected variance. Expected variance is the pooled error variance
divided by the number of cross sections.
OLS Diagnostics: The OLS versions of Equations 3 and 4 produced the following diagnostic information. Starred values of the first-order autocorrelation function (ACF1)
were used as the value of Rho in the GLS/ARMA versions of the equations.
EQUATION 3.| EQUATION 4.
|
RESIDUAL| RESIDUAL
CROSS SECTION ACF1 VARIANCE| ACF1 VARIANCE
____________________________________|_______________________
|
BELGIUM .17 9.088 |
DENMARK | .03 9.756
GERMANY .05 7.926 |
FRANCE *.50 9.174 |
IRELAND | .23 20.667
ITALY .09 5.389 |
NETHERLANDS .28 4.435 |
UK | *.57 38.088
____________________________________|_______________________
this is a regression analysis of the "good thing" Eurobarometer question: "Generally speaking, do you think that (our country's) membership in the Community is a good thing, a bad thing, or neither good nor bad?"
The specific operationalization in the equation is "Net Support", that is, the percentage who respond "good thing" minus the percentage who respond "bad thing".
For design related questions or comments, please contact Carlos E. Granier.
For content related questions or comments, please contact Prof. Eichenberg.
© 1998, Richard C. Eichenberg. For classroom use at Tufts University only. Citation or
reproduction of this material is prohibited without the written consent of the author.