Power in the European Union: an evolutionary computing approach

Figures & data

Table 1. Expert estimates of inter-institutional relative power

Institution	Procedure	Range
Council	COD CNS	62.50–100 71.43–100
Commission	COD CNS	22.22–100 10–100
European Parliament	COD CNS	44.44–100 0–75

Note: The practitioners survey asked 18 informants to rank each institution’s relative power on a scale of zero to one hundred. Data from Thomson (2002). CNS = consultation, COD = codecision.

Table 2. Expert estimates of member state relative power

Member State	Range
Austria	0–80
Belgium	25–60
Denmark	10–60
Finland	20–55
France	45–100
Germany	45–100
Greece	10–45
Ireland	10–80
Italy	30–90
Luxembourg	0–95
Netherlands	35–65
Portugal	20–50
Spain	40–100
Sweden	30–55
UK	80–100

Note: Data from Bailer (2004). The practitioners survey asked fifteen informants to rank each Member State’s power on a scale of zero to one hundred.

Figure 1. Evolution with one-point crossover.

Table 3. Prediction errors using formal and estimated power scores

	COM	EP	AUS	BEL	DK	FIN	FR	GER	GRE	IRE	IT	LUX	NL	POR	SP	SW	UK	Avg. Error
CNS/QMV (55)
SSI	100	0	14.20	16.89	11.46	11.46	34.36	34.36	16.89	11.46	34.36	7.37	16.89	16.89	28.40	14.20	34.36	23.3
Experts range	100	61.42	36.77	11.49	4.60	25.28	20.69	45.97	20.69	4.60	13.79	0	16.09	21.4	45.97	25.28	36.77	21.9
Optimum	100	70.39	33.18	0	0	35.06	0	29.61	57.03	0	0	0	0	0	52.91	65.89	6.14	20.5
CNS/UNAN (39)
SSI	0	0	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	17.3
Experts range	40	0	0	25	60	20	100	100	45	80	90	0	35	20	40	55	100	15.3
Optimum	0	0	0	0.06	100	0	54.47	0.001	0	0	35.78	0	0	0	0	7.51	18.19	10.8
COD/QMV (56)
SSI	0	100	10	12.14	7.81	7.81	26.10	26.10	12.14	7.81	26.10	4.55	12.14	12.14	21.34	10	26.10	27.3
Experts range	50	100	0	14.20	34.09	31.25	25.57	32.21	5.68	45.45	17.05	0	36.93	11.36	22.73	31.25	56.82	22.9
Optimum	9.78	29.35	0	0	0	82.61	0	0	4.35	94.57	0	40.22	10.87	0	15.22	100	0	19.4
COD/UNAN (12)
SSI	0	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	100	19.2
Experts range	26.31	100	0	7.47	17.94	16.44	13.45	13.45	13.45	23.92	8.97	28.40	10.46	14.95	11.96	8.97	29.90	12.0
Optimum	0.19	40.09	0	0.01	0	0	18.08	0	28.81	100	0.03	0.05	0	2.06	14.02	0.01	3.88	6.1

Note: Cell entries are power scores entered into the compromise model for each actor. For the SSI models, scores are ‘variant two’ Shapley Shubik Index values, normalised so that the strongest actor scores 100. For the other models, scores are from an evolutionary algorithm constrained so that each actor’s power can be any value from 0 to 100 (optimum) or any value within the range assessed by experts. QMV = qualified majority voting, UNAN = unanimous voting, CNS = consultation, COD = codecision.

Figure 2. Estimated power and frequency of missing positions, by procedure.

Each panel plots power estimates from the experts range models in Table 3 againt the number of missing positions for each actor in the DEU data.

Figure 3. Estimated power and extremity (average distance), by produre.

Each panels plots power estimates from the experts range models in Table 3 against each actor’s average absolute distance from the mean positions in the DEU data.

Figure 4. Estimated power and extremity (min or maxpositions), byprocedure.

Each panels plots power estimates from the experts range models in Table 3 against the percentage of times each actor took a position at the minimum or maximum of the polocy alternatives in the DEU data.

Figure 5. Estimated power and council presidency, byprocedure.

Each panel plots power estimates from the experts range models in Table 3 against the number of issues in the DEU data decided under each Member State’s Council Presidency.

Thomson, R. (2002) “Estimation of the Relative Capabilities of the Commission, the Council and the European Parliament Using Key Informants,” 18 February, mimeo.

 Google Scholar

Bailer, S. 2004. “Bargaining Success in the European Union.” European Union Politics 5 (1): 99–124. doi:https://doi.org/10.1177/1465116504040447.

 Web of Science ®Google Scholar