Saturday, July 23, 2016

Analyzing Multi-Question Sets in Pew’s Survey Data to Estimate the Percentages of Muslims Who Hold Moderate or Extremist Views (Part 1 of 2)


                                                                       Summary

This article documents an attempt to find out, through analyses of responses to sets of survey questions, what percentages of Muslim adults are “moderate” or extremist. Pew data were analyzed for numerous multi-question sets. Results, weighted by countries’ adult Muslim population sizes, were as follows for four of the main analyses:
66% across twenty-one countries favor at least one of death for apostasy, stoning of adulterers, or honor killings. Only 24% oppose all three of those forms of murder.
46% across seven countries viewed favorably at least one of three terrorist groups. Only 23% viewed unfavorably all three terrorist groups. Among the subset of respondents who gave definite answers to all three questions, the majority (62%) viewed favorably at least one of the three terrorist groups.
80% across thirty-six countries endorsed at least one of death for apostasy, stoning of adulterers, whippings and cutting off of hands, or making sharia (Islamic law) the official law of the country. Only 16% opposed all four extremist items.
95% across twenty-one countries endorsed at least one of a seven-item set that included death for apostasy, stoning of adulterers, whippings and cutting off of hands, honor killings, making sharia (Islamic law) the official law of the country, that a wife “must always obey” her husband, or that a wife should not have the right to divorce. Also, 84% endorsed at least two, 70% endorsed at least three, and 56% endorsed at least four of the seven items. Only 3% opposed all seven extremist items.
The results for the larger sets of questions indicate that a large majority of Muslim adults from a diverse sampling of countries holds hardline fundamentalist/extremist views, and that only a small minority could possibly be moderate.

Note: This is Part 1 of a two-part article. For Part 2 click here.


Introduction

During the past fifteen years many people in the West have discussed the question of what percentage of Muslims are “moderate.”[1]  The question can be summarized as follows: What percentage of Muslims support (a) an important set of human rights and freedoms, such as the freedom to criticize their religion publicly; freedom to openly leave their religion; etc., and equality of people regardless of gender, religion, race, sexual orientation, language, etc.; and oppose (b) hardline sharia, horrendous practices (such as female genital mutilation, honor killings, marriage of adult males to child brides, etc.), terrorism, and terrorist groups? The flip-side of that question asks What percentage of Muslims oppose (a) and support (b)?  In other words, what percentage are hardline fundamentalists/extremists?

Unfortunately, Pew Research and others have not designed their surveys with the primary aim of answering this major, two-sided question. For example, questions concerning harsh punishments for “blasphemy,” homosexuality, premarital sex, and so on, have not been asked in a standard questionnaire in a sufficiently large number of countries, either by Pew or others, to my knowledge. That is also the case for questions about other punishments, including imprisonment, for such transgressions of Islamic law. That said, Pew’s research on Muslims’ opinions, summarized in their 2013 report [2], does include some questions concerning fundamentalist or extremist views, and covers the largest number of countries and the largest overall population. I will show in this article how it is possible to make better use of Pew’s data to help answer the major question.

Pew makes the data sets for many of their surveys freely-available for download and analysis by members of the general public. With access to the microdata, one can do analyses other than those reported by the original researchers. The results of the analyses reported in this article can be verified by readers who have adequate knowledge and skill.[3] Previously, I described how to get started with a simple analysis of a Pew data set using the free statistical software PSPP. For another article, using PSPP and a common spreadsheet, I estimated the population-weighted percentages of Muslim adults who, for each question, favored or opposed harsh punishment, sharia, honor killing, and restrictions of women’s rights and freedoms. In the present article, I again use PSPP to explore some Pew data sets [4] on Muslims’ opinions on those issues, but this time I analyze responses within sets of questions.

To begin to estimate the percentages of Muslims who are moderate or extremist, it is important to gather together in the same analysis responses to multiple questions. Analyzing responses to a set of questions can reveal, for example, the percentages of Muslims who favor at least one of death for apostasy or stoning for adultery. A respondent who favored even only one of those forms of murder would not be “moderate” in the sense described briefly above. I would classify such a respondent as a likely supporter of at least one murderous practice, and thus an extremist. (Call that the “one-is-enough” criterion). Those who opposed both of those forms of murder could possibly be moderates. We could not make that conclusion definitively because many other important questions weren’t asked by the researchers. But we could then expand the analysis to include additional relevant questions, increasing our confidence in estimating the percentage of possible moderates. The same logic can be used in classifying respondents' views toward terrorist or militant fundamentalist groups such as al-Qaeda, Hamas, Hezbollah, and the Taliban. Holding a favorable view of one or more of them would indicate support for murderous extremism. A moderate would view all of them unfavorably. To get a sense of the extent of the extremism, I will also examine how many such items are favored (or opposed), and by what percentages of respondents, per country and overall. Due to the limited number of relevant questions in each dataset, the resulting figures will to some extent underestimate the percentage of extremists and overestimate the percentage of possible moderates.

Pew’s survey questions vary widely in validity, in my opinion. I’ve included in these analyses relevant questions that I believe to be, if not problem-free, at least of adequate validity and clarity. The subject matter of these questions deals directly with real-world actions, laws, policies, or active groups that are hardline fundamentalist or extremist. As I will discuss in a forthcoming article, I excluded Pew’s “suicide bombing and other forms of violence” question due to its lack of clarity. Subsequent articles will present critical evaluations and quantitative analyses of other questions in Pew’s surveys of Muslims’ views as well as those from other research organizations.

All results reported in this article are for the general samples of adult Muslims surveyed by Pew, except for analyses of subsets in 5.2, 5.3, and 5.5. Using Pew’s age structure estimates of the Muslim population in each country, I estimated the country population sizes for Muslims 18 years of age or older (see Appendix A of the previous article). I’ve used those estimates in calculating the population-weighted percentages for each response set reported below. Following Pew’s “Instructions for Downloading Datasets,” I’ve applied their weight variable in obtaining the response choice percentages for individual countries. Readers who wish to replicate analyses from the present article may find it helpful to see the previous article’s instructions for PSPP in Analysis 5 (“Honor” killings) and Appendix B. The discussion of the estimated margins of error for response choice percentages in the previous article is applicable to the combined analyses in the present article.


Analyses

1.0 Summary of the Main Analyses

Hardline fundamentalism/extremism versus “moderation” in populations of Muslim adults as tested by analyses of multi-question sets.

Note: Some question wording not shown in this article is quoted in the previous article.

Table 1.0

Percentages of Muslims Who Favor at Least One Vs. Oppose All Extremist/ Hardline Fundamentalist Items Per Question Set
Pew
Survey
Global Attitudes 2010

Percentage
The World’s Muslims 2013, Tolerance and Tension 2010

Favor at
Oppose
Rmndr.*
Question Set

least one
all
Death for apostasy, stoning adulterers
58
37
6
Death for apostasy, stoning adulterers, honor killings
66
24
11
Death for apostasy, stoning adulterers, whippings and cutting off hands
63
32
5
Death for apostasy, stoning adulterers, whippings/cutting, honor killings
69
21
9
Hamas, Hezbollah, al-Qaeda
46
23
31
Hamas, Hezb., al-Q., death for apostasy, stoning adulterers, whip./cut.
78
12
9
Death for apost., ston. adultr., whip/cut, terror groups, restrict women
84
10
7
Death for apostasy, stoning adulterers, whippings/cutting, sharia
80
16
4
Death for apostasy, stoning adulterers, whip./cut., honor killings, sharia
83
11
6
Death for apostasy, stoning adulterers, whip./cut., sharia, religious leaders should only be men
91
7
2
Death for apostasy, stoning adulterers, whip./cut., honor killings, sharia, wife must always obey husband, wife has no right to divorce
95
3
2
Notes. Percentages are weighted by countries’ Muslim adult population sizes. Displayed numbers are rounded.
Datasets: Pew Global Attitudes 2010, Pew Africa Survey 2009, Pew The World’s Muslims 2012.
Wording of questions and response options is simplified in this table. See text and original Pew sources for exact wording.
*Remainder is the percentage of Muslim adults who neither favored at least one nor opposed all items in the question set.
Pew’s terrorist group questions elicited some very high percentages of don’t know responses. See analyses 5.1, 5.2, 5.3.



1. Favor death for apostasy, or stoning for adultery, or both? Oppose both?

1.1. Survey Report: Pew Global Attitudes, 2010, “Muslim Publics Divided on Hamas and Hezbollah,” page 35. [5]

Pew’s response code: 1 = favor, 2 = oppose, 8 = don’t know, 9 = refused.

Instructions for doing the combined analysis with PSPP: Select Transform, then Compute Variable. Write the variable label and numeric expression (see below). Each numeric expression makes a binary variable, scoring 1 (one) for cases that fit the criteria and 0 (zero) for cases that do not. Once you’ve created the new variables, check a small random sample of their cases to make sure their scores are correct, then run a Crosstabs analysis on each to obtain the percentages. The remainder consists of the cases that don’t fit the criteria of the new variables. Readers can fairly easily verify the percentages for individual countries listed in the tables below by following this procedure.

Expressions:

Favor at least one: (Q108d = 1) | (Q108c = 1)

Oppose both: (Q108d = 2) & (Q108c = 2)


Table 1.1

Combined analysis of question set: Q108d death penalty for apostasy, Q108c stoning of adulterers.

Percentage
Population
Favor at
Oppose
both
Rmdr.
Favor at
least one
Oppose
both
Remainder
Sum
least one
Egypt
92
5
2
44143034
2551620
1173745
47868400
Jordan
91
6
3
3060198
197098
89904
3347200
Pakistan
86
8
6
83489461
7717434
6333105
97540000
Nigeria
64
34
2
22176146
11615229
832625
34624000
Indonesia
47
49
4
67272650
69191680
5302070
141766400
Lebanon
24
74
1
442182
1346784
27034
1816000
Turkey
17
76
7
8119219
37183186
3427595
48730000





Population Sum
228702891
129803031
17186079
375692000
Population-Weighted Percentage
60.88
34.55
4.57
100.00
Note: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.


Comment: A population-weighted 61% of Muslim adults across seven countries favor either death for apostasy or stoning adulterers, or both, while 35% oppose both.


1.2. Survey Reports: Pew, “The World’s Muslims,” 2013, p. 219 (Q92b apostasy) and p. 221 (Q92d adultery), full report pdf; Pew Forum, “Tolerance and Tension: Islam and Christianity in Sub-Saharan Africa,” 2010, full report pdf.[6] Note that in the sub-Saharan full report and in its data file, the questions are numbered Q95c (apostasy) and Q95e (adultery).

Expressions:

Favor at least one: (Q92b = 1) | (Q92d = 1)

Oppose both: (Q92b = 2) & (Q92d = 2)


Table 1.2

Combined analysis of question set: Q92b death penalty for apostasy, Q92d stoning of adulterers.

Percentage
Population
Favor at
Oppose
both
Rmdr.
Favor at
least one
Oppose
both
Remainder
Sum
least one
Egypt
96
3
1
46106625
1275859
485916
47868400
Jordan
92
6
2
3066845
216939
63416
3347200
Pakistan
91
3
6
88581211
3346948
5611842
97540000
Afghanistan
91
7
2
13238391
1031726
311484
14581600
Palestin. T.
83
12
5
1654714
234584
100302
1989600
Djibouti
73
20
7
361477
97157
35366
494000
Niger
73
21
6
4929947
1384972
426681
6741600
Malaysia
68
17
15
7560401
1856102
1652697
11069200
Iraq
66
22
12
10364508
3420345
1913546
15698400
Bangladesh
65
28
7
53917596
23485849
6058556
83462000
Senegal
63
35
2
3732635
2031568
122197
5886400
Mali
62
29
9
4181635
1942536
571429
6695600
Guinea Biss.
62
32
6
213563
110893
19544
344000
DR Congo
54
36
11
231392
155037
45972
432400
Tanzania
48
50
2
3833103
4025885
160612
8019600
Nigeria
45
50
5
15682415
17244451
1697134
34624000
Indonesia
45
49
6
63944464
68854139
8967797
141766400
Ghana
44
51
5
898173
1052169
102457
2052800
Uganda
44
53
4
744922
899227
63850
1708000
Chad
43
56
1
1267286
1671680
21434
2960400
Kenya
42
57
1
749760
1019040
26400
1795200
Mozambique
41
55
4
847896
1131540
73764
2053200
Cameroon
41
56
2
692901
946736
41162
1680800
Liberia
41
52
7
97204
125591
17204
240000
Ethiopia
36
62
2
4942967
8543400
335634
13822000
Tunisia
30
67
3
2241164
4981983
193252
7416400
Kyrgyzstan
29
64
7
850599
1891619
190983
2933200
Tajikistan
27
61
12
1007065
2267448
466687
3741200
Lebanon
23
75
2
411025
1360557
44417
1816000
Russia
20
70
10
2168004
7558809
1105587
10832400
Azerbaijan
16
77
7
1059479
5027123
470197
6556800
Turkey
13
84
3
6267223
40807015
1655762
48730000
Kosovo
10
70
19
114768
780318
212114
1107200
Bosnia-Hrz.
9
87
5
119459
1180543
62797
1362800
Albania
7
83
10
135280
1525484
186036
1846800
Kazakhstan
6
90
4
459135
6788848
310017
7558000





Population Sum
346675233
220274121
33824246
600773600
Population-Weighted Percentage
57.70
36.67
5.63
100.00
Note: 36-Country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Sources: Pew Research, Africa Survey 2009 and The World’s Muslims 2012 Datasets.

Comment: A population-weighted 58% of Muslim adults across 36 countries favor death for apostasy or stoning of adulterers, or both, while 37% oppose both. About 347 million Muslim adults across the 36 countries—with about 89 million in Pakistan, 64 million in Indonesia, and 54 million in Bangladesh—favor at least one of the two extreme punishments.



2. Favor at least one of death for apostasy, or stoning for adultery, or “honor” killing? Oppose all three?

Survey Report: Pew, “The World’s Muslims,” (2013). The “honor” killing Q53 and Q54 tables are on p. 190 of the complete report pdf.

Note: I presented results of an analysis of Q53/Q54 (“honor” killing of a male or female) in the previous article. Consistent with that previous analysis, I do not score “rarely justified” responses as moderate or as (sufficiently) extremist in the present analysis.

Pew’s response code for Q53/Q54: 1 = often justified, 2 = sometimes justified, 3 = rarely justified, 4 = never justified, 8 = don’t know, 9 = refused

My classification for Q53/Q54: moderate option = 4, clear extremist option <= 2.

Expressions:

Favor at least one:  (Q92b = 1) | (Q92d = 1) | ((Q53 <= 2) | (Q54 <= 2))

Oppose all three: (Q92b = 2) & (Q92d = 2) & ((Q53 = 4) & (Q54 = 4))

Table 2

Combined analysis of three-item set: Q92b death penalty for apostasy, Q92d stoning of adulterers, Q53/Q54 “honor” killings.

Percentage


Population

Favor at
Oppose all
Remainder

Least One
Three
Egypt
98
0
1
47868400
Afghanistan*
95
2
3
14581600
Jordan
93
3
4
3347200
Pakistan
92
2
5
97540000
Palestinian T.
91
4
5
1989600
Iraq*
85
6
10
15698400
Niger
78
10
12
6741600
Bangladesh
74
13
13
83462000
Malaysia
73
11
16
11069200
Indonesia
49
39
12
141766400
Lebanon
46
40
14
1816000
Tajikistan
42
33
24
3741200
Kyrgyzstan
41
38
21
2933200
Tunisia
41
46
13
7416400
Russia
32
44
24
10832400
Kosovo
27
43
29
1107200
Turkey
27
57
16
48730000
Albania
22
56
22
1846800
Azerbaijan
21
64
15
6556800
Bosnia-Herz.
15
69
16
1362800
Kazakhstan
10
73
17
7558000

Population
340645530
122778303
54541366
517965200
Pop.-Wtd. %
65.77
23.70
10.53
100.00

Notes. 21-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Research, The World’s Muslims 2012 Dataset.
*In Afghanistan and Iraq, an alternative wording for Q53/Q54 was used.

Comments: Among Muslim adults across the 21 countries, about 66% favor at least one of the three forms of murder, while only 24% oppose all three. If Q54 is included in the analysis instead of Q53/Q54, then 65% favor at least one, 25% oppose all three, and a remainder of about 10% had other response combinations.



3. Favor at least one of death for apostasy, or stoning for adultery, or whipping/ cutting off of hands for crimes like theft? Oppose all three?

3.1. Survey Report: Pew Global Attitudes, 2010, “Muslim Publics Divided on Hamas and Hezbollah.”

Favor at least one: (Q108d = 1) | (Q108c = 1) | (Q108b =1)

Oppose all three: (Q108d = 2) & (Q108c = 2) & (Q108b = 2)


Table 3.1

Combined analysis of three-question set: Q108d death penalty for apostasy, Q108c stoning of adulterers, Q108b whippings and cutting off of hands.

Percentage
Population
Favor at
Oppose all
Remainder
Least One
Three
Egypt
95
3
2
47868400
Jordan
95
4
2
3347200
Pakistan
89
6
5
97540000
Nigeria
72
27
2
34624000
Indonesia
53
44
3
141766400
Lebanon
26
73
1
1816000
Turkey
20
74
6
48730000

Population
245795331
115954179
13942490
375692000
Pop.-Wtd. %
65.42
30.86
3.71
100.00
Notes: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.


3.2. Survey Reports: Pew, The World’s Muslims, 2013 (Q92b apostasy, Q92d adultery, Q92c whipping/cutting); Pew, Sub-Saharan, 2010, (Q95c apostasy, Q95e adultery, Q95d whipping/cutting)

Favor at least one: (Q92b = 1) | (Q92d = 1) | (Q92c = 1)

Oppose all three: (Q92b = 2) & (Q92d = 2) & (Q92c = 2)


Table 3.2


Combined analysis of three-question set: Q92b death penalty for apostasy, Q92d stoning of adulterers, Q92c whippings and cutting off of hands.


Percentage
Population
Favor at
Oppose all
Remainder
Least One
Three
Egypt
97
2
1
47868400
Jordan
94
4
2
3347200
Pakistan
92
3
5
97540000
Afghanistan
92
6
2
14581600
Niger
86
11
4
6741600
Palestinian T.
84
11
5
1989600
Djibouti
80
14
6
494000
Iraq
74
17
9
15698400
Malaysia
72
14
14
11069200
Bangladesh
71
22
7
83462000
Mali
71
23
6
6695600
Senegal
70
28
2
5886400
Guinea Bissau
69
26
5
344000
DR Congo
62
28
10
432400
Cameroon
57
41
2
1680800
Tanzania
56
43
2
8019600
Liberia
54
40
6
240000
Nigeria
54
43
3
34624000
Chad
54
46
1
2960400
Ghana
53
43
4
2052800
Uganda
52
44
3
1708000
Indonesia
50
43
6
141766400
Mozambique
50
47
4
2053200
Kenya
45
53
2
1795200
Kyrgyzstan
44
52
4
2933200
Ethiopia
39
58
3
13822000
Tajikistan
33
56
11
3741200
Tunisia
33
65
3
7416400
Russia
31
60
9
10832400
Lebanon
24
73
3
1816000
Azerbaijan
19
75
7
6556800
Turkey
18
79
4
48730000
Kazakhstan
15
80
5
7558000
Bosnia-Herzeg.
15
80
5
1362800
Kosovo
12
69
19
1107200
Albania
12
77
11
1846800

Population
375990689
194211558
30571353
600773600
Pop.-Wtd. %
62.58
32.33
5.09
100.00
Note: 36-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Sources: Pew Research, Africa Survey 2009 and The World’s Muslims 2012 Datasets.



4. Favor at least one of death for apostasy, or stoning for adultery, or whipping/ cutting off of hands for crimes like theft, or honor killing? Oppose all four?

Survey Report: Pew, The World’s Muslims, 2013

Favor at least one: (Q92b = 1) | (Q92d = 1) | (Q92c = 1) | ((Q53 <=2) | (Q54 <= 2))

Oppose all four: (Q92b = 2) & (Q92d = 2) & (Q92c = 2) & ((Q53 = 4) & (Q54 = 4))


Table 4

Combined analysis of four-item set: Q92b death penalty for apostasy, Q92d stoning of adulterers, Q92c whippings and cutting off of hands, Q53/Q54 “honor” killings.

Percentage
Population
Favor at
Oppose all
Remainder
Least One
Four
Egypt
99
0
1
47868400
Afghanistan*
96
2
3
14581600
Jordan
95
3
3
3347200
Pakistan
94
2
5
97540000
Palestinian T.
92
3
4
1989600
Iraq*
88
5
7
15698400
Niger
88
5
6
6741600
Bangladesh
78
11
11
83462000
Malaysia
76
9
15
11069200
Indonesia
54
35
11
141766400
Kyrgyzstan
53
31
16
2933200
Tajikistan
48
31
21
3741200
Lebanon
46
40
14
1816000
Tunisia
42
44
13
7416400
Russia
41
38
21
10832400
Turkey
30
55
15
48730000
Kosovo
28
43
29
1107200
Albania
23
55
22
1846800
Azerbaijan
23
63
14
6556800
Bosnia-Herz.
21
64
15
1362800
Kazakhstan
19
65
15
7558000

Population
358027551
110956431
48981218
517965200
Pop.-Wtd. %
69.12
21.42
9.46
100.00
Notes. 21-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Research, The World’s Muslims 2012 Dataset.
*In Afghanistan and Iraq, an alternative wording for Q53/Q54 was used.



5. View favorably at least one of the following terrorist/militant groups: Hamas, or Hezbollah, or Al-Qaeda, (or the Taliban)? View unfavorably all three?

 5.1 Survey Report: Pew Global Attitudes, 2010, “Muslim Publics Divided on Hamas and Hezbollah.”

Q7. “Please tell me if you have a very favorable, somewhat favorable, somewhat unfavorable or very unfavorable opinion of (INSERT)?
…n.   Hamas               o.   Hezbollah             p.   al Qaeda…”

Pew’s response code: very favorable = 1, somewhat favorable = 2, somewhat unfavorable = 3, very unfavorable = 4, don’t know = 8, refused = 9.

Expressions:

View favorably at least one: (Q7n <= 2) | (Q7o <= 2) | (Q7p <= 2)

View unfavorably all three: ((Q7n = 3) | (Q7n = 4)) & ((Q7o = 3) | (Q7o = 4)) & ((Q7p = 3) | (Q7p = 4))

Note: For Pakistan only, the al-Qaeda question was asked in Q35b. To obtain the results for Pakistan, use Q35b instead of Q7p in the above expressions.


Table 5.1

Combined analysis of three-question set: Q7n Hamas, Q7o Hezbollah, Q7p al-Qaeda.

Percentage
Population
View Favorably
View Unfavorably
Remainder
At Least One
All Three
Jordan
85
11
3
3347200
Egypt
68
24
8
47868400
Nigeria
63
15
22
34624000
Indonesia
54
21
25
141766400
Lebanon
53
42
4
1816000
Pakistan*
33
8
59
97540000
Turkey
10
65
25
48730000





Population
171207405
87318253
117166342
375692000
Pop.-Wtd. %
45.57
23.24
31.19
100.00
Notes: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.
*In Pakistan the question about al-Qaeda was Q35b.

Comments:

A plurality of 46% viewed favorably at least one, and a minority of 23% viewed unfavorably all three of the terrorist groups.

The very large percentages of “don’t know” responses to these questions in Nigeria, Indonesia, Turkey, and (especially) Pakistan contribute to the high figures in the remainder column and complicate the interpretation of the results. Therefore, the next analysis (5.2) will include only the subset of respondents who gave definite answers (very/somewhat favorable or very/somewhat unfavorable) to all three questions. Analysis 5.2 will examine the distribution of extremism scores, as described below.


Calculating extremism scores. Instructions for PSPP: For each of the relevant question variables in the set, use the Compute Variable function to make a new variable that scores 1 point for an extremist response, and 0 (zero) for all other responses. For example, to make the variable for the al-Qaeda question Q7p, you would write the numeric expression Q7p <= 2 to score a one or a two “favorable” response. (You could label the “Target Variable” in this example FavQ7p). The numeric expression produces a binary variable consisting of 1s and 0s, where 1 indicates that a respondent favored the terrorist group and 0 indicates the respondent did not view it favorably. Note that “did not view it favorably” includes all responses other than the “favorable” 1 and 2, including the “unfavorable” responses 3 or 4, or “don’t know,” or “refused” (though for the definite answer subset in 5.2, below, the dk/refs are excluded anyway). Once you’ve made a binary (1 versus 0) extremism variable for each item, then sum the scores of those binary variables using the Compute Variable function, making a summary extremism variable that has a score for each respondent. Use Crosstabs to obtain the distribution of extremism scores for each country by entering country as rows and summary extremism variable as columns.

Extremism Score (Scale 0 – 3): (FavQ7n) + (FavQ7o) + (FavQ7p)

Reminder: For Pakistan only, the al-Qaeda question was asked in Q35b, not Q7p.

5.2 Analysis of the definite answer subset: The subset includes only respondents who gave a definite opinion (very favorable, somewhat favorable, somewhat unfavorable, very unfavorable) for all three questions. Create a filter variable which you can then use to select only respondents with definite answers to all three questions for the analysis. In a previous article, I provided instructions for how to make and apply a filter variable. The expression for the definite answer subset here can be as follows: (Q7n <= 4) & (Q7o <= 4) & (Q7p <= 4)


Table 5.2        
 

Analysis of subset who had definite answers to all three questions.
Combined analysis of three-question set: Q7n Hamas, Q7o Hezbollah, Q7p Al-Qaeda.
Extremism Score Scale: 0 – 3.

Percentage
Subset
Number of Terrorist Groups Viewed Favorably
Zero
One
Two
Three
At Least 1
Population
%
Jordan
13
33
39
15
87
3001415
90
Nigeria
22
12
18
48
78
23570033
68
Egypt
27
45
22
6
73
42714127
89
Pakistan
31
13
34
23
69
25660861
26
Indonesia
32
15
30
23
68
94125769
66
Lebanon
44
4
50
1
56
1719337
95
Turkey
88
6
2
4
12
36091207
74

Population
87318253
43172976
53092651
43298869
139564496
226882749

P.-W. %
38.49
19.03
23.40
19.08



P.W.C %
100
62
42




Notes: Seven-country subset population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.
*In Pakistan, the question about al-Qaeda was Q35b.


Comments:

1. The definite-answer subset proportion of the total population is .6039 (i.e., 226882749 / 375692000). Multiplying .6039 by .3849 (the proportion of the subset who viewed none of the three terrorist groups favorably) = .2324, or 23.24%, confirming the result shown in Table 5.1, that is, the percentage who viewed all three unfavorably.

2. About 62% (61.51%) of the subset viewed favorably at least one of the three terrorist groups.

3. A subset population-weighted 38.49% did not indicate a favorable view any of the three terrorist groups. Because this subset has no (zero) don’t know/refused responses, and 61.51% viewed favorably at least one terrorist group, we can deduce that 38.49% viewed all three terrorist groups unfavorably (also see Comment 1).

4. In Pakistan, about 69% of those with definite answers for all three questions favored at least one of the terrorist groups. However, Pakistan’s definite-answer subset is only about 26% of its general sample, so the 69% should be regarded with caution.

5. Indonesia has a 41% share of the subset population.


One can also calculate the mean extremism scores for each country. In PSPP, use the Means option from Compare Means (located under the Analyze menu), using country as the independent variable and the summary extremism score variable (see “Calculating extremism scores” above, just before 5.2) as the dependent variable, to obtain the country mean extremism scores. The output can then be exported to a spreadsheet for population-weighting. For the definite answer subset in 5.2, the (subset) population-weighted mean extremism score was 1.23 (range: .22 for Turkey to 1.91 for Nigeria), which indicates that on average the definite answer subset viewed favorably a little more than one of the three terrorist groups. Of the seven countries, only Turkey had a mean of less than 1.

5.3 The subset of those who responded “don’t know/refused” to at least one of the three terrorist group questions constitutes .3961 of the total population (1 - .6039). It includes all the remaining data that was not captured in the definite answer subset. Structurally, its maximum number of definite responses is two, though many of its respondents had one or zero definite responses. Within this subset, a (subset) population-weighted 21.3% viewed at least one terrorist group as favorable, and 7.2% viewed two as unfavorable. The remaining 71.5% of the subset either responded with don’t know/refused to all three questions, or with one unfavorable and two don’t know/refused responses. With the majority of the subset dominated by don’t know/refused responses, it would not be sensible to calculate a mean extremism score. Pakistan has a 48% share of this subset population. For Pakistan, the subset made up 73.7% of its general sample. In the subset in Pakistan, 19.5% viewed favorably at least one terrorist group while 3.8% viewed two as unfavorable.

5.4 Survey Report: Pew Global Attitudes, 2013, “Muslim Publics Share Concerns About Extremist Groups.” [7]

Q9. “Please tell me if you have a very favorable, somewhat favorable, somewhat unfavorable or very unfavorable opinion of (INSERT)?
…j. Hamas                  k. Hezbollah               l. al Qaeda…               …q. The Taliban…” 

Pew’s response code: very favorable = 1, somewhat favorable = 2, somewhat unfavorable = 3, very unfavorable = 4, don’t know = 8, refused = 9.

View favorably at least one: (Q9j <= 2) | (Q9k <= 2) | (Q9l <= 2) | (Q9q <= 2)

View unfavorably all four: ((Q9j = 3) | (Q9j = 4)) & ((Q9k = 3) | (Q9k = 4)) & ((Q9l = 3) | (Q9l = 4)) & ((Q9q = 3) | (Q9q = 4))

Note: For Pakistan only, questions for al-Qaeda and the Taliban were Q41b and Q41d, respectively. To obtain the results for Pakistan, use Q41b instead of Q9l, and Q41d instead of Q9q, in the above expressions.

Table 5.4

Combined analysis of four-question set: Q9j Hamas, Q9k Hezbollah, Q9l al-Qaeda, Q9q the Taliban.

Percentage
Population
View Favorably
View Unfavorably
Remainder
At Least One
All Four
Egypt
77
20
3
51280120
Lebanon
65
33
2
1899040
Palestinian T.
64
26
10
2251320
Jordan
62
34
5
3740560
Tunisia
54
19
27
7706920
Malaysia
48
17
35
12110320
Indonesia
45
22
33
148928360
Nigeria
34
18
47
38928040
Pakistan*
30
8
63
105930760
Senegal
22
32
46
6546880
Turkey
15
58
27
51275800

Population
175932900
95273757
159391463
430598120
Pop.-Wtd. %
40.86
22.13
37.02
100.00
Notes: 11-country population (2013 est.) is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2013 Dataset.
*In Pakistan, questions about Al-Qaeda and the Taliban were Q41b and Q41d, respectively.


Comments about Table 5.4:

1. The 2013 Muslim (age 18+) population estimates for the countries were calculated by making a linear interpolation between Pew’s 2010 and 2020 estimates, but otherwise were calculated the same way as the 2010 Muslim (age 18+) population estimates.

2. A population-weighted 41% of Muslim adults viewed favorably at least one of the four terrorist/extremist groups, while 22% viewed unfavorably all four of them.

3. Except for Egypt, Lebanon, the Palestinian Territories, and Jordan, the results for the other seven countries are obscured by very high levels of “don’t know” responses, resulting in the high percentages in the Remainder column. These very high levels of don’t know responses also significantly obscure the population-weighted results overall because the populations of these other seven countries together make up 86% of the eleven-country population.

5.5 Analysis of the definite-answer subset of these 2013 data showed that a (subset) population-weighted 56.7% viewed favorably at least one of the four terrorist/extremist groups, while 43.3% viewed all four unfavorably. The definite-answer subset constituted 51.04% of the general sample.



6. Favor at least one of death for apostasy, stoning for adultery, whipping/cutting off hands for crimes like theft, Hamas, Hezbollah, or Al-Qaeda. Oppose/ view unfavorably all of these?

Survey Report: Pew Global Attitudes 2010, “Muslim Publics Divided on Hamas and Hezbollah.”

Favor at least one: (Q108d = 1) | (Q108c = 1) | (Q108b = 1) | (Q7n <= 2) | (Q7o <= 2) | (Q7p <= 2)

Oppose/ view unfavorably all six: (Q108d = 2) & (Q108c = 2) & (Q108b = 2) & ((Q7n = 3) | (Q7n = 4)) & ((Q7o = 3) | (Q7o = 4)) & ((Q7p = 3) | (Q7p = 4))

Reminder: For Pakistan only, the question about Al-Qaeda was Q35b, not Q7p.

Table 6.1

Combined analysis of six-question set: Q108d death penalty for apostasy, Q108c stoning of adulterers, Q108b whippings/cutting off of hands, Q7n Hamas, Q7o Hezbollah, Q7p Al-Qaeda

Percentage
Population

Favor at
Oppose/Unfav.
Remainder

least one
all six
Jordan
98
1
1
3347200
Egypt
97
3
1
47868400
Pakistan*
94
1
5
97540000
Nigeria
87
6
8
34624000
Indonesia
77
11
12
141766400
Lebanon
65
31
4
1816000
Turkey
27
51
22
48730000

Population
294271761
45805268
35614971
375692000
Pop.-Wtd. %
78.33
12.19
9.48
100.00
Notes: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.
*In Pakistan, the question about al-Qaeda was Q35b.


6.2 Extremism Score (0 – 6): (FavQ108d) + (FavQ108c) + (FavQ108b) + (FavQ7n) + (FavQ7o) + (FavQ7p)


Table 6.2

Combined analysis of 6-question set: Q108d death penalty for apostasy, Q108c stoning of adulterers, Q108b whippings and cutting off of hands, Q7n Hamas, Q7o Hezbollah, Q7p Al-Qaeda.
Extremism Score Scale: 0 – 6.

Mean
Percentage of Respondents

Extrem.
Number of Extremist Items Favored

Score
Zero
One
Two
Three
Four
Five
Six
Jordan
3.62
2
3
14
28
25
21
7
Egypt
3.41
3
5
11
30
34
14
4
Nigeria
3.16
13
8
13
24
12
15
15
Pakistan*
2.95
6
8
11
50
12
8
5
Indonesia
2.15
23
14
22
20
10
7
4
Lebanon
1.45
35
6
41
14
3
0
0
Turkey
0.52
73
12
9
5
1
1
0
Pop.-W. Mean
2.47

Pop.-Wtd. %

21.67
10.38
15.34
27.56
12.60
7.88
4.57
P.-W. Cml. %

100
78
68
53
25
12

Notes. Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.
*In Pakistan, the question about al-Qaeda was Q35b.

Comments about Table 6.2:

1. Note that 21.67% favored zero items. That 21.67% consists of 12.19% who opposed all six items, plus 9.48% remainder (compare with Table 6.1).

2. At the bottom row, the population-weighted cumulative percentage of those who favored at least four of the six extremist items was 25, while 53% favored at least three, and 68% favored at least two.

3. The high percentages of “don’t know” responses to the terrorist group questions in Nigeria, Pakistan, Indonesia, and Turkey complicate the interpretation of these results.



7. Favor at least one of sharia as law of the land, death for apostasy, stoning for adultery, whipping/cutting off hands for crimes like theft? Oppose all four of these?

Survey Report: Pew, The World’s Muslims, 2013.

Favor at least one: (Q79a = 1) | (Q92b = 1) | (Q92d = 1) | (Q92c = 1)

Oppose all of them: (Q79a = 2) & (Q92b = 2) & (Q92d = 2) & (Q92c = 2)

Extremism Score: (FavQ79a) + (FavQ92b) + (FavQ92d) + (FavQ92c)

Pew Sub-Saharan 2010 (Q95a sharia question, Q95c apostasy, Q95e adultery Q95d whipping/cutting).
Note: Q79aRUS was the alternative version of Q79a used in Russia.


Table 7

Combined analysis of four-question set: Q79a sharia, Q92b death penalty for apostasy, Q92d stoning of adulterers, Q92c whippings and cutting off of hands.  Extremism Score Scale: 0 – 4.

Mean
Percentage of Respondents

Extrem.
Number of Extremist Items Favored
Oppose
Rmdr.

Score
One
Two
Three
Four
>= 1
All Four
Afghanistan
3.44
8
7
18
67
100
0
0
Pakistan
3.30
5
9
24
59
97
0
3
Egypt
3.12
7
16
34
43
100
0
0
Palestin. T.
3.04
11
11
19
53
95
3
2
Djibouti
2.81
13
11
13
52
89
8
3
Jordan
2.73
12
26
31
29
98
0
1
Niger
2.68
13
16
41
25
95
3
2
Malaysia
2.59
21
13
19
39
91
2
7
Iraq
2.44
26
20
26
25
97
1
2
Bangladesh
2.29
27
21
24
22
94
3
3
Mali
2.19
11
17
21
28
77
18
5
Senegal
2.08
14
15
23
24
76
22
2
Dem.R.Congo
2.05
22
17
16
26
80
11
9
Guinea Biss.
1.90
13
26
18
17
75
22
3
Nigeria
1.82
30
14
17
18
79
20
1
Ghana
1.77
17
13
12
25
66
30
4
Indonesia
1.68
35
16
20
10
81
14
6
Uganda
1.64
31
21
13
13
78
20
2
Kenya
1.62
29
13
10
19
71
26
3
Cameroon
1.57
22
18
20
9
70
28
2
Mozambique
1.57
32
22
11
12
77
21
2
Chad
1.55
24
14
10
18
66
33
1
Tanzania
1.55
16
18
16
14
64
34
2
Liberia
1.53
23
17
14
14
68
24
8
Ethiopia
1.37
42
12
11
10
74
25
1
Tunisia
1.30
29
7
10
14
60
36
3
Kyrgyzstan
1.07
25
17
11
3
57
35
8
Russia*
0.90
32
11
8
3
53
28
18
Lebanon
0.85
17
4
5
11
37
57
6
Tajikistan
0.81
22
12
6
5
44
39
18
Kosovo
0.43
16
4
4
2
26
40
34
Turkey
0.41
15
5
4
1
25
65
10
Bosnia-Herz.
0.38
16
5
2
1
24
68
8
Azerbaijan
0.31
17
5
1
0
23
63
14
Kazakhstan
0.30
15
4
2
0
21
67
11
Albania
0.29
9
6
2
0
18
52
30
Pop.-W. Mean
2.08

Pop.-Wtd. %

21.59
13.54
19.66
25.04
79.84
15.76
4.40
P.-W. Cml. %

80
58
45




Notes. 36-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Sources: Pew Research, Africa Survey 2009 and The World’s Muslims 2012 Datasets.
*In Russia, the alternative version of Q79a referred to making sharia the law in the “Muslim Republics of Russia.”

Comments about Table 7:

1. The percentage who favored zero (0) items of the set is the percentage of the remainder (4.40) plus the percentage who opposed all four (15.76), which = 20.16%.

2. As the bottom row shows, a population-weighted cumulative 45% favored at least three of the four extremist items, while 58% favored at least two.


(Part 2 of this article continues in the next post with Analyses 8 - 12, Discussion, and Notes and References)

2 comments:

  1. Where can one see how the Ahmadis answered, to connect their identification as Ahmadis with the positions on death for apostasy, etc.?

    ReplyDelete
  2. Hi Reason on Faith,

    The denomination identification information is in Pew's sub-Saharan African data set. It involves some work, but if you download the Pew data file and open it using PSPP (a freely-available stats package), you can then go in to the file and do a crosstabs analysis of q37rec (which gives the denomination -- Sunni, Shia, Ahmadiyya, etc.) as rows and q95c (apostasy), q95d (whippings/cutting), and q95e (stoning adulterers) as columns. For most other questions in that data set you will have to filter the analysis by major religion, but not for the q95 series because it was only asked of Muslims.

    Pew's sub-Saharan data (2009):
    http://www.pewforum.org/datasets/tolerance-and-tension-islam-and-christianity-in-sub-saharan-africa/

    Pew's published report for that (2009) data:
    http://www.pewforum.org/2010/04/15/executive-summary-islam-and-christianity-in-sub-saharan-africa/
    (see Complete Report pdf)

    How to open a Pew data file and do a crosstabs analysis using PSPP, described here (skip down to the end of the article):

    http://empethop.blogspot.ca/2015/02/a-look-at-pews-2013-worlds-muslims.html

    How to make a filter variable (see Appendix B in the linked article):

    http://empethop.blogspot.ca/2016/07/estimated-numbers-and-percentages-of_3.html


    Probably the main reason Pew didn't report on the denomination subsets in their sub-Saharan data is that the Muslim sample sizes weren't big enough for that type of analysis. 298 people (raw count) across several countries is not bad though for getting a very rough idea of where Ahmadis stand on these issues. The Ahmadis explicitly categorized as such "volunteered" that identification. There may be additional Ahmadis in some of the other categories listed for q37rec.

    I should add that there are other survey data sets available online where the specific denomination information is given, so even if the researchers haven't presented those particular results in their reports, you can still go in and find them in the data file.

    ReplyDelete