Exploration of the opinions about USA, China and Russia


Attaching package: 'reshape2'

The following object is masked from 'package:tidyr':

    smiths

Linking to GEOS 3.11.0, GDAL 3.5.3, PROJ 9.1.0; sf_use_s2() is TRUE

We have presented in a previous post the database that we have elaborated on the basis of the Global Attitude Survey in order to follow the evolution of world opinion about the three major powers that are USA, China and Russia.

Before to develop more sophisticated analysis on polarisation, we will try here to propose simple explorations of data in order to check the quality and shortcomings of the data.

We remind briefly the structure of the database of 617345 observation and 18 variables.

	wav	sta	wgt	sex	age	opi_rel	opi_USA	opi_CHN	opi_RUS	opi_UN	opi_EU	opi_IRN	opi_DEU	opi_FRA	opi_JPN	opi_IND	opi_TWN	opi_BRA
617342	2023	ZAF	1.0502763	Male	24	Somewhat important	Somewhat favorable	Very favorable	Somewhat favorable	Somewhat favorable	Somewhat favorable	NA	NA	NA	NA	Somewhat favorable	Somewhat favorable	NA
617343	2023	ZAF	0.3254254	Male	39	Very important	Very unfavorable	Very unfavorable	Very unfavorable	Very unfavorable	Very unfavorable	NA	NA	NA	NA	Very unfavorable	DK	NA
617344	2023	ZAF	2.4937630	Male	28	Very important	DK	Very unfavorable	Very unfavorable	DK	Very unfavorable	NA	NA	NA	NA	Somewhat favorable	DK	NA
617345	2023	ZAF	0.7725435	Female	24	Very important	Somewhat favorable	Somewhat favorable	Somewhat favorable	Somewhat favorable	Somewhat favorable	NA	NA	NA	NA	Somewhat favorable	Somewhat favorable	NA

I. Evolution of the opinion of UK citizens about USA (2002-2023)

Let’s take a example the evolution of the evolution of the opinion of inhabitants of one country about another. We will use the case of the opinion of UK citizens on USA as it is one case where the survey has been realized every year from 2002 to 2023. In the majority of case, we don’t have complete time series.

Data selection

      wav            wgt              sex             age       
 Min.   :2002   Min.   :0.05043   Female: 9707   Min.   :18.00  
 1st Qu.:2009   1st Qu.:0.60548   Male  :10161   1st Qu.:36.00  
 Median :2014   Median :0.94839                  Median :52.00  
 Mean   :2014   Mean   :1.00422                  Mean   :50.97  
 3rd Qu.:2019   3rd Qu.:1.11110                  3rd Qu.:65.00  
 Max.   :2023   Max.   :5.70402                  Max.   :97.00  
                   opi      
 Very favorable      :2455  
 Somewhat favorable  :9061  
 Somewhat unfavorable:4735  
 Very unfavorable    :2092  
 DK                  :1429  
 Refused             :  96

wav	wgt	sex	age	opi
2002	0.893	Male	39	Somewhat favorable
2002	1.039	Male	23	Somewhat unfavorable
2002	0.893	Male	35	Somewhat unfavorable
2002	0.956	Female	59	Very favorable
2002	1.094	Female	22	Somewhat favorable
2002	0.955	Male	41	Somewhat favorable

Opinions by years

Keeping all modalities of answer, we can create a weighted table of answers with as many lines as wawes of wurvey and as many columns as possibilities of choices.

Opinions of UK on USA (raw count)
wav	Very favorable	Somewhat favorable	Somewhat unfavorable	Very unfavorable	DK	Refused
2002	134	239	62	21	41	3
2003	90	258	68	59	24	0
2004	75	214	122	49	39	0
2005	90	314	202	82	55	1
2006	98	403	181	117	94	0
2007	87	410	288	124	66	6
2008	56	335	187	88	65	7
2009	95	418	105	49	71	3
2010	105	384	135	44	74	2
2011	116	483	217	62	109	6
2012	105	505	244	74	84	4
2013	100	484	215	75	111	8
2014	130	530	192	77	68	2
2015	152	474	167	63	91	6
2016	200	678	294	95	161	10
2017	135	371	289	126	95	9
2018	114	364	262	156	68	10
2019	164	410	269	116	34	1
2020	89	314	326	208	30	2
2021	113	517	211	104	39	5
2022	184	643	315	91	29	2
2023	130	458	281	99	26	6

We can easily transform the table in percentage for a better visualisation of the probability of opinion of each type :

Opinions of UK on USA (%)
	Very favorable	Somewhat favorable	Somewhat unfavorable	Very unfavorable	DK	Refused
2002	26.7	47.8	12.4	4.3	8.1	0.7
2003	18.1	51.7	13.6	11.9	4.7	0.0
2004	15.0	42.8	24.4	9.9	7.9	0.0
2005	12.1	42.2	27.2	11.0	7.4	0.1
2006	11.0	45.1	20.3	13.1	10.5	0.0
2007	8.8	41.8	29.4	12.7	6.7	0.6
2008	7.6	45.3	25.3	12.0	8.9	0.9
2009	12.8	56.4	14.2	6.6	9.6	0.4
2010	14.2	51.6	18.1	6.0	9.9	0.3
2011	11.7	48.7	21.8	6.2	11.0	0.6
2012	10.3	49.7	24.0	7.3	8.3	0.4
2013	10.0	48.7	21.7	7.5	11.2	0.8
2014	13.0	53.0	19.2	7.7	6.8	0.2
2015	16.0	49.7	17.5	6.6	9.5	0.7
2016	13.9	47.2	20.5	6.6	11.2	0.7
2017	13.2	36.2	28.2	12.3	9.3	0.9
2018	11.7	37.4	26.9	16.0	7.0	1.0
2019	16.5	41.2	27.1	11.6	3.4	0.1
2020	9.2	32.4	33.6	21.5	3.1	0.2
2021	11.4	52.3	21.3	10.5	3.9	0.5
2022	14.6	50.9	24.9	7.2	2.3	0.2
2023	13.0	45.8	28.1	9.9	2.6	0.6

We can now produce different time series according to our hypothesis on the phenomena under investigation.

Who has an opinion ?

We can firstly investigate the % of people that refused to answer or did not made a choice.

As we can see the evolution of this parameter is characterized by a slow increase of people without opinion between 2002 and 2016, between 8 and 10% of the sample. But after the election of D. Trump in 2017, the proportion of people without opinion declined dramatically to less than 5 % and the election of J. Biden did not modify the situation.

Polarisation of opinion

This first results are in line with another analysis related to the choice of extreme situation of the Likert scale. The figure below indicates the proportion of people that has chosen the answers “very favorable” or “very unfavorable”.

We notice here a high polarisation in 2002 (probably in relation with September 11 and the “war to terrorism” of G.W. Bush). But we reach the lowest level of polarisation in 2010-2014 during the presidency of B.Obama. The polarisation increase slowly during the second presidency of Obama and dramatically with the election of D. Trump. J. Biden election is characterized by a reduction of polarisation.

Balance of opinion.

We use here the classical index of attraction-repulsion \(I\) to evaluate the balance of positive and negative opinion.

\(I = \frac{Favorable-Unfavorable}{Favorable+Unfavorable}\)

Finally we can produce the classical diagram of the sum of favorable opinion that reveals expected correlation with the different presidents of the US. But this figure is only one component of the analysis and the two previous one about opinion and polarisation are certainly as much important.

Trajectory with CA

We can use a correspondance analysis in order to visualize the general trajectory of opinions (excluding answers of people that refused or did not answer). The advantage of this statisticial method is to provide a good picture of the period of change or stability of opinion and to identify the most important turning points.

[1] "data.frame"

The Axis 1 (74% of inertia) is clearly associated to the opposition between favorable opinion (left) and unfavorable opinion (right). The Axis 2 is related to the degree of polarisation that can be higher (top) or lower (down).
The most extreme situations are observed in 2002 (very favorable opinion) and 2020 (very unfavorable opinion). They are both related to international events that has dramatically influenced the opinion of UK citizens on the US : sympathy in the case of the folow-up of septembr 11. attacks ; antipathy in the case of Trump présidence and covid crisis. These extreme points are generally not observed during a long period and we observe very strong distance between these years and the one located immediately before and after.
Looking at the most important distances between two years, we notice very important jumps in 2002-2003 (begining of the second war in Iraq), 2008-2009 (election of Obama), 2016-2017 (election of Trump) and 2020-2021 (election of Biden).

This results suggest that it is not obvious to evaluate the opinion of a country about USA if we don’t have several points of observation through time and, ideally, a complete view of the trajectory.

II. Benchmarking the opinions of different countries about USA in 2007

In this second exploration, we select several countries but only one wave of survey. We choose the year 2007 because it is the one for which we have the maximum of countries (47) and also because it is located just before the crise of subprimes that started in july 2007.

Data selection

      sta             wgt              sex             age       
 CHN    : 3141   Min.   : 0.0100   Female:23028   Min.   :18.00  
 IND    : 2040   1st Qu.: 0.8655   Male  :21965   1st Qu.:27.00  
 USA    : 1999   Median : 1.0000                  Median :38.00  
 PAK    : 1975   Mean   : 1.0003                  Mean   :40.47  
 UGA    : 1116   3rd Qu.: 1.0033                  3rd Qu.:51.00  
 NGA    : 1092   Max.   :18.7495                  Max.   :97.00  
 (Other):33630                                                   
                   opi       
 Very favorable      : 6917  
 Somewhat favorable  :14836  
 Somewhat unfavorable:11096  
 Very unfavorable    : 9208  
 DK                  : 2664  
 Refused             :  272

sta	wgt	sex	age	opi
ARG	0.8136704	Female	57	DK
ARG	0.8136704	Male	34	Very unfavorable
ARG	0.8136704	Male	34	DK
ARG	0.8136704	Female	67	Very unfavorable
ARG	0.8136704	Female	80	Somewhat unfavorable
ARG	0.8136704	Female	29	Somewhat favorable

Opinions by country

Keeping all modalities of answer, we can create a weighted table of answers with as many lines as states of survey and as many columns as possibilities of choices.

Opinions by states on USA in 2017(raw count)
sta	Very favorable	Somewhat favorable	Somewhat unfavorable	Very unfavorable	DK	Refused
ARG	26	107	248	332	76	12
BGD	170	359	148	258	48	4
BGR	61	186	122	80	39	4
BOL	65	280	273	155	51	10
BRA	38	396	384	128	50	3
CAN	114	426	274	140	24	4
CHL	110	327	196	88	66	14
CHN	78	1012	1477	318	244	12
CIV	363	265	59	21	0	0
CZE	45	358	357	94	28	14
DEU	23	281	466	195	29	7
EGY	68	141	319	457	13	2
ESP	10	159	158	142	29	2
ETH	293	252	100	57	6	2
FRA	47	345	446	164	1	1
GBR	87	410	288	124	66	6
GHA	318	244	52	48	44	1
IDN	36	251	415	245	45	1
IND	407	804	373	210	209	37
ISR	263	437	132	43	10	2
ITA	32	234	138	51	38	8
JOR	83	122	255	521	11	8
JPN	59	404	254	21	23	1
KEN	429	434	77	33	24	0
KOR	21	397	234	33	32	1
KWT	68	158	97	133	36	8
LBN	162	313	239	278	7	2
MAR	43	106	163	396	266	26
MEX	82	381	217	125	17	7
MLI	311	248	62	66	8	6
MYS	26	161	210	272	28	1
NGA	486	272	103	194	37	0
PAK	83	223	282	1076	313	4
PER	99	390	162	92	55	1
POL	61	248	124	28	40	3
PSE	30	63	121	551	12	0
RUS	78	326	314	160	104	10
SEN	181	299	131	71	13	5
SVK	31	339	330	155	38	6
SWE	88	367	368	116	53	3
TUR	17	68	82	729	74	1
TZA	143	186	107	172	86	8
UGA	326	390	93	121	183	2
UKR	48	222	97	98	31	4
USA	941	662	236	118	39	8
VEN	97	356	147	178	20	5
ZAF	215	404	148	152	79	2

We can easily transform the table in percentage for a better visualisation of the probability of opinion of each type :

Opinions by states on USA in 2007 (%)
	Very favorable	Somewhat favorable	Somewhat unfavorable	Very unfavorable	DK	Refused
ARG	3.3	13.3	31.0	41.5	9.5	1.5
BGD	17.2	36.4	15.0	26.1	4.9	0.4
BGR	12.3	37.9	24.9	16.2	8.0	0.7
BOL	7.8	33.6	32.7	18.6	6.1	1.2
BRA	3.8	39.6	38.4	12.8	5.0	0.3
CAN	11.6	43.4	27.9	14.3	2.4	0.4
CHL	13.7	40.9	24.5	11.0	8.2	1.8
CHN	2.5	32.2	47.0	10.1	7.8	0.4
CIV	51.3	37.5	8.3	2.9	0.0	0.0
CZE	5.0	40.0	39.8	10.5	3.1	1.6
DEU	2.3	28.1	46.6	19.5	2.9	0.7
EGY	6.8	14.1	31.9	45.7	1.3	0.2
ESP	2.0	31.8	31.6	28.4	5.8	0.4
ETH	41.3	35.5	14.2	8.0	0.8	0.3
FRA	4.7	34.4	44.4	16.3	0.1	0.1
GBR	8.8	41.8	29.4	12.7	6.7	0.6
GHA	45.0	34.5	7.4	6.8	6.2	0.1
IDN	3.7	25.3	41.8	24.6	4.5	0.1
IND	20.0	39.4	18.3	10.3	10.2	1.8
ISR	29.7	49.3	14.9	4.8	1.1	0.2
ITA	6.4	46.6	27.5	10.2	7.6	1.7
JOR	8.3	12.2	25.5	52.1	1.1	0.8
JPN	7.7	53.0	33.3	2.8	3.0	0.1
KEN	43.0	43.5	7.7	3.3	2.4	0.0
KOR	2.9	55.3	32.6	4.6	4.5	0.1
KWT	13.6	31.6	19.4	26.6	7.2	1.6
LBN	16.2	31.3	23.9	27.8	0.7	0.2
MAR	4.3	10.6	16.3	39.6	26.6	2.6
MEX	9.9	46.0	26.2	15.1	2.0	0.8
MLI	44.4	35.4	8.8	9.4	1.1	0.9
MYS	3.7	23.1	30.1	39.1	4.0	0.1
NGA	44.5	24.9	9.4	17.8	3.4	0.0
PAK	4.2	11.3	14.2	54.3	15.8	0.2
PER	12.4	48.8	20.3	11.5	6.9	0.2
POL	12.0	49.2	24.6	5.6	7.9	0.7
PSE	3.9	8.1	15.6	70.9	1.5	0.1
RUS	7.8	32.9	31.7	16.1	10.5	1.0
SEN	25.9	42.7	18.7	10.1	1.9	0.7
SVK	3.4	37.7	36.7	17.2	4.2	0.7
SWE	8.8	36.9	37.0	11.7	5.3	0.3
TUR	1.8	7.0	8.4	75.1	7.6	0.1
TZA	20.4	26.5	15.2	24.5	12.3	1.1
UGA	29.3	35.0	8.3	10.9	16.4	0.2
UKR	9.6	44.3	19.4	19.6	6.2	0.9
USA	47.0	33.0	11.8	5.9	1.9	0.4
VEN	12.1	44.3	18.3	22.2	2.5	0.6
ZAF	21.5	40.4	14.8	15.2	7.9	0.2

Who has an opinion ?

We can firstly investigate the % of people that refused to answer or did not made a choice.

As we can see on the figure, there is a great variability in the proportion of people that has formulated an opinion in the different countries where the survey has been realized. The countries where the maximum of people has refused to answer are Morocco (29%), Uganda (17%), Pakistan (16%) and Tanzania (13%). The reverse situation is observed in Côte d’Ivoire, France, and Lebanon where less than 1% of people did not formulated an opinion on USA.

Polarisation

We can then investigate the degree of polarisation, defined as the proportion of people that used the extreme values of the likert scale.

The most important levels of polarisation are observed in countries from Middle East characterized by very negative opinion of USA (Turkey, Palestiniant territories, Jordan, Pakistan, …) but also in countries from subsaharan Africa characterised by very positive opinion of USA (Nigeria, Côte d’ivoire, Mali, …). The countries with low level of polarisation are rather located in Asia (Korea, China, Japan, …), Europe (Czech republic, Italia, Poland, …) and Latin America (Brazil, Peru,…).

Balance of favorable and unfavorable opinions

We use here the classical index of attraction-repulsion \(I\) to evaluate the balance of positive and negative opinion.

\(I = \frac{Favorable-Unfavorable}{Favorable+Unfavorable}\)

The index of attraction-repulsion is theoretically bounded between -1 (all opinions are unfavorable) and + 1 (all opinions are favorable) with an equilibrium around 0 (equal number of favorable and unfavorable opinions). In present case, we observe a very strong diversity of positions from -0.81 in Turkey to +0.77 in Ivory Coast. The countries are to some extent clustered in regions with very negative opinions in Middle east and very positive in subsaharan Africa. But they are exceptions and we should not conclude too quickly because it is possible that people has systematically provide positive or negative answers to foreign countries proposed to their evaluation. As we will see later, the absolute opinion on USA is less interesting than the comparison between USA and other countries of reference like Chinaor Russia.

Mapping with CA

We can use a correspondance analysis in order to visualize the positions of countries about USA. We will obtain by this method a map of the world opinion where distance between positions reveals differences in opinions.

The figure reveals clearly a “Guttman effect” which means that axis 1 and 2 should not be interprated separately but jointly. The countries are indeed distributed on a parabol that follow the order of opinions from the most unfavorable (topleft) to the moderate (bottom center) and the most favorable (topright).
But the countries are not regularly distributed along the parabol and we can easily identify clusters of countries sharing the same opinion on USA in 2007 : very unfavorable in arabic and muslim countries but also in Argentina ; very favorable in subsaharan Africa, Israel and USA ; moderate and equilibrated in other countries.

But this result is only a snapshot of the situation in 2007 and what we have to perform in next step is an analysis of the dynamic of positions of countries through time.