Multilingual federalism in times of crisis

Figures & data

Table 1. Dimensions and characteristics of the conjoint analysis.

Dimension	Characteristic
Responsible level	1. Decision made at the [federal/national]* level 2. Decision made at the [federal/national]* level, but each [State/province/community and region/canton/region]** can be more restrictive 3. [States/Provinces/Communities and regions/Cantons/Regions]** coordinate with each other and decide together 4. Each [province/community and region/canton/region]** decides for itself 5. Each [municipality/county or city]*** decides for itself
Role of scientists	1. No influence 2. Some influence 3. Strong influence
Role of citizens	1. No influence 2. Some influence (consultative citizens’ forums)
Role of stakeholders (employers, trade unions, etc.)	1. No influence 2. Some influence 3. Strong influence
Transparency	1. None 2. Some (decision is explained) 3. Full (decision is explained and all data publicly available)

Note: The conjoint survey was conducted in Dutch, English, French, German, and Italian; see Figure A1 for examples. *France: ‘national’; **Belgium: ‘Communities and Regions’; Switzerland ‘Cantons’; France: ‘Regions’; USA and Australia: ‘States’; ***USA: ‘county or city’.

Table 2. Sample size(s) and distribution of languages by country.

Country	Mother tongue of survey respondents
	French	English	Dutch	German	Total
Canada	610	610	0	0	1220
Belgium	760	0	752	0	1512
Switzerland	760	0	0	760	1520
France	1118	0	0	0	1118
Australia	0	1093	0	0	1093
USA	0	1136	0	0	1136
Total	3248	2839	752	760	7599

Figure 1. French speakers and governance attitudes across countries.

Figure 2. English speakers and governance attitudes across countries.

Figure 3. Differences between French speakers by country. Reading example: Belgian francophones are more favourably inclined towards centralization than the French.

Figure 4. Differences between English and French speakers by residence in Canada. Note: ‘anglo-RoC’ = English speakers living in rest of Canada; ‘anglo-quebecois’ = English speakers living in Quebec; ‘franco-quebecois’ = French speakers living in Quebec; ‘franco-RoC’ = French speakers living in rest of Canada.

Figure 5. Differences between French and non-French speakers by country.

Table 3. ANOVA test results for between-group differences.

	Francophones (F, B, CAN, CH)	Anglophones (AUS, US, CAN)	Belgium (French vs. Dutch)	Canada (French vs. English)	Switzerland (French vs. German)
Overall	6.620***	1.982**	1.594^+	4.634***	1.640^+
Level	13.14***	1.401	0.585	3.635**	0.126
Citizens	0.015	2.671*	0.064	1.162	1.584
Experts	1.299	2.008^+	1.261	9.778***	1.975
Stakeholders	0.302	0.899	0.159	0.005	0.102
Transparency	0.142	0.251	2.937*	1.767	3.111*

Note: Shown are F-statistics comparing models with and models without group interactions (country of residence or mother tongue). Assessed is therefore whether these interaction coefficients are distinguishable from zero. Levels of significance: *** = 0.001, ** = 0.01, * = 0.05, ⁺=0.1. Shaded cells = in line with our hypotheses.

Table A1. Sample composition, in % of group and category.

Group	Age					Gender	Education			Place of residence				Current personal economic situation				Religion				Ideology (0–10)
	18–24	25–34	35–44	45–54	55+	Female	Primary	Secondary	Tertiary	Countryside	Town	Suburbs	Big city	Very difficult	Difficult	Coping	Comfortable	None	Catholic	Protestant	Other	Left (0-3)	Centre (4-6)	Right (7-10)
F-French	17	24	25	17	16	66	4	43	53	17	24	16	43	11	26	40	24	45	28	3	24	24	44	32
CAN-French	12	12	20	16	40	58	6	45	50	16	28	27	29	9	20	43	27	29	61	3	6	17	64	19
CAN-English	8	14	17	13	39	63	3	43	54	7	26	33	34	8	14	41	37	34	25	23	17	27	50	23
B-French	24	20	15	16	26	61	5	45	49	31	31	14	24	12	26	45	17	46	32	4	18	24	47	29
B-Dutch	17	19	18	18	27	41	5	47	48	29	36	18	18	7	21	46	25	42	43	2	13	20	40	40
CH-French	16	20	19	21	24	51	4	54	42	31	31	15	23	13	27	42	17	37	32	15	16	24	47	28
CH-German	13	20	22	18	36	44	7	62	31	37	23	17	23	8	15	38	39	42	25	22	10	25	52	23
AUS-English	16	17	17	13	37	59	3	59	39	9	26	47	19	9	16	46	29	47	21	14	18	20	55	26
US-English	13	15	17	15	40	56	12	36	52	10	33	36	21	6	13	36	45	27	21	26	26	31	34	35

Table A2. Personal COVID-19 unaffectedness by type and group, in % of respondents per group.

	I contracted the virus at some point and suffered greatly.	I contracted the virus at some point but had only mild or no symptoms.	I or somebody close to me had to quarantine at some point.	Somebody close to me contracted the virus at some point but had only mild or no symptoms.	I have suffered financially because of the measures taken by governments to contain the spread of COVID-19	Somebody close to me contracted the virus at some point and suffered greatly.	I have lost my job because of the measures taken by governments to contain the spread of COVID-19.	I have not been personally affected by the pandemic in a negative way.
F-French	6%	8%	15%	21%	25%	15%	8%	49%
CAN-French	2%	3%	15%	8%	20%	8%	11%	63%
CAN-English	2%	3%	22%	10%	25%	7%	10%	57%
B-French	5%	10%	26%	27%	21%	16%	8%	48%
B-Dutch	5%	9%	32%	27%	25%	17%	7%	49%
CH-French	3%	9%	24%	28%	25%	20%	9%	51%
CH-German	3%	6%	34%	24%	23%	12%	7%	54%
AUS-English	3%	3%	24%	7%	24%	5%	9%	60%
US-English	5%	10%	45%	40%	21%	27%	11%	36%

Note: While more than one response could be chosen by respondents, each had to select at least one.

Table A3. Power analyses.

Sample	N individuals	Final N	Power
Francophones	3248	25 984	0.72
Anglophones	2839	22 712	0.67
Canada	1220	9760	0.35
Belgium	1512	12 096	0.41
Switzerland	1520	12 160	0.42

Note: Final N = number of individuals times number of profiles from which to choose (2) times number of tasks (4) minus missings, if applicable. Power analyses as per Schuessler and Freitag (2020), performed with the R-package ‘cjoint’ (version 2.1, for interactive conjoints, conditional AMCE = 0.05).

Figure A1. Screenshot from the conjoint experiment.

Figure A2. Conditional marginal means for all nine language-country groups.

Figure A3. Conditional marginal means for all nine language-country groups, facetted by group.

Figure A4. Conditional marginal means for all nine language-country groups, matching respondents only.

Figure A5. Conditional marginal means for all nine language-country groups, passed attention check only.

Figure A6. Minimum required effective sample sizes. Note: shown are the required effective sample sizes ( =  N individuals * N scenarios * N iterations) for a conjoint experiment with a maximum of five levels ( = the responsible government, in our case) to detect effect sizes AMCE > 0.02 and < 0.1 by the different levels of statistical power for our sub-samples (cf. Table A3) at significance level 0.05. Graph made using the cjpowR package developed by Schuessler and Freitag (2020).

Schuessler, J., and M. Freitag. 2020. Power Analysis for Conjoint Experiments. Working Paper. 1 September 2023. https://osf.io/preprints/socarxiv/9yuhp/.

 Google Scholar