Sunday, July 03, 2016

Estimated Numbers and Percentages of Muslim Adults Who Favor or Oppose Harsh Punishments, Sharia, and Restrictions on the Rights and Freedom of Women


Pew Research Center’s surveys of Muslims’ opinions, published in 2010 and 2013, reported the percentages of those in each country who favored or opposed sharia or harsh elements of sharia.[1, 2, 3] However, Pew did not estimate the numbers and mean percentages of Muslims over all surveyed countries who favored or opposed those items. In the ongoing debates about the extent of hardline fundamentalist beliefs held by Muslims, it is critically important to have scientifically-based estimates of these numbers and percentages. Consequently, some laypersons, journalists, and activists have attempted to make their own estimates, based on Pew’s findings. In the present article, I will use the Pew survey data sets [4] as well as Pew’s estimates of the age structure of Muslim populations [5] in the countries surveyed, to estimate the appropriate numbers and population-weighted percentages.

I will examine Pew’s data for questions concerning the death penalty for leaving Islam, stoning of adulterers, whippings and cutting off of hands as punishments, making sharia the official law of the nation, and “honor” killings. I will also examine Pew’s data for two questions that measured attitudes towards women’s rights and freedom, specifically whether a wife “must always obey” her husband and whether a wife should have the right to divorce her husband. Results for additional questions relevant to extremism, including support for specific terrorist groups, will be examined in subsequent articles. Complete and exact wording for the questions and response options quoted below is taken from the Questionnaire file in the relevant Dataset folder.

All results reported in the analyses below are for the general samples of adult Muslims surveyed by Pew. Using Pew’s age structure estimates for the Muslim population in each country for 2010, I’ve estimated the country population sizes for Muslims 18 years of age or older [shown in Appendix A]. I’ve then used those estimates to calculate the population-weighted percentages for the response choices. Following Pew’s “Instructions for Downloading Datasets,” I’ve applied their weight variable in the analyses within individual countries. Using the freely-available statistical software PSPP, plus a common spreadsheet, readers can verify the results obtained here by analyzing the relevant data files made available to the public by Pew. In Appendix B, and in Analysis 5 (“Honor killings”), I describe some procedures in PSPP for analyzing these data files.

Results for Thailand, shown in The World’s Muslims (2013) report, are not included in the present analyses because they are not nationally representative; Pew surveyed only that country’s five southernmost provinces. Data for Algeria is in Pew’s data file but was not included here due to relatively low (75%) national representativeness (see the Codebook pdf, page 3, in The World’s Muslims 2012 Dataset folder).

For practical reasons, I did not attempt to formally estimate the margins of error for the combined samples analyzed here, but I discuss them briefly and make rough estimates, with caveats, in Appendix C. Given the large sizes of the combined samples (6688, 27385, 28894, 34744, and 35199), and other considerations, I don’t regard margin of error as a major concern in this article.



Analyses

1.0 Summary of Analyses 1.1 to 4 

Support for and opposition to harsh sharia punishments and making sharia the official law of the nation.

Table 1.0

Muslims’ Support for and Opposition to Harsh Punishments and Sharia as National Law
Survey
Item
Percentage
Number of
Combined
Report
Favor
Oppose
Dk/Ref
countries
sample n
Pew
death penalty for apostasy
48
47
5
7
6688
Global
stoning of adulterers
56
40
4
7
6688
2010
whippings and cutting off hands
53
43
4
7
6688







Pew
death penalty for apostasy
39
54
6
36
34744
World's
stoning of adulterers
51
43
6
36
34744
Muslims
whippings and cutting off hands
50
46
5
36
34744
2013
sharia as law of the nation
69
24
7
36
35166
Notes. Percentages are weighted by countries’ Muslim adult population sizes. Samples are of Muslim adults.
Datasets: Pew Global Attitudes 2010, Pew Africa Survey 2009, Pew The World’s Muslims 2012.


1. Death Penalty for Apostasy

1.1. Survey Report: Pew Global Attitudes, 2010b, “Muslim Publics Divided on Hamas and Hezbollah,” (Q108d, page 35). Combined 7-country Muslim sample size = 6688.

Q108. “Do you favor or oppose making the following the law in (survey country)? … d. Death penalty for people who leave the Muslim religion” 

Table 1.1


Q108. Do you favor or oppose making the following the law in (survey country)?
… d. Death penalty for people who leave the Muslim religion


Favor
Oppose
Don't Know
Refused
Total
Population
180651065
175147300
17645489
2248145
375692000
Percentage
48.08
46.62
4.70
0.60
100.00
Notes: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.

Comments:
1.     As shown in Table 1.1, the seven-country population-weighted percentage of Muslims 18 years of age or older who favored death for apostasy is 48.08. The comparable figure for the seven-country all-ages Muslim population is 49.79% and for the seven-country age 15-and-older Muslim population is 48.40%.

2.     An alternative though slightly less accurate method for estimating the 2010 age ≥ 18 Muslim populations uses (a) the U.N.’s published estimates for the general (not religion-specific) age ≥ 18 population for each country, multiplied by (b) Pew’s published estimates of the proportions of Muslims (not age-specific) in each country. Using these estimates, the above seven-country population-weighted percentage of Muslim adults who favor death for apostasy is 48.14.


1.2. Survey Report: Pew “The World’s Muslims,” 2013 (Q92b, page 219), also includes results from Pew “Sub-Saharan” 2010a (Q95c, page 291). Combined 36-country Muslim sample size = 34744.

Countries Included in /Excluded from the Analysis: 36 included, Thailand excluded.

Q92. “Do you favor or oppose the following? … b. the death penalty for people who leave the Muslim religion”

Table 1.2


Q92. Do you favor or oppose the following: … b. the death penalty for people who leave the Muslim religion?


Favor
Oppose
Don't Know
Refused
Total
Population
236811546
325531116
33706500
4724438
600773600
Percentage
39.42
54.19
5.61
0.79
100.00
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.
                   

Comments:

1.     As shown in Table 1.2, the 36-country population-weighted percentage of Muslims ≥ 18 years of age who favored the death penalty for apostasy is 39.42. The comparable figure for the 36-country all-ages Muslim population is 40.57%, and for the 36-country age ≥ 15 years Muslim population is 39.64%. I suspect that 39.42% is a slight underestimate of the level of expressed support for Q92b for Muslim adults in the survey period (late 2008 to early 2012), but not far off. Using the U.N. × Pew method of estimation (described above in Comment 2 under Table 1.1), the 36-country population-weighted percentage of Muslim adults who favor death for apostasy is 39.11.

2.     Multiplying the “favor” proportion obtained from the 36-country population by the estimated 2010 world population of Muslim adults (age ≥ 18, see Appendix A) gives us 0.394177684 × 954481153 = 376,235,170.



2. Stoning of Adulterers

2.1. Survey Report: Pew Global Attitudes, 2010b, “Muslim Publics Divided…” (Q108c, page 35). Combined 7-country Muslim sample size = 6688.

Q108   “Do you favor or oppose making the following the law in (survey country)? … c. Stoning people who commit adultery”

Table 2.1


Q108. Do you favor or oppose making the following the law in (survey country)?
… c. Stoning people who commit adultery


Favor
Oppose
Don't Know
Refused
Total
Population
209493782
149675855
14288223
2234140
375692000
Percentage
55.76
39.84
3.80
0.59
100.00
Notes: Seven-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Global Attitudes, 2010 Dataset.


2.2. Survey Report: Pew “The World’s Muslims,” 2013 (Q92d, page 221), includes some results from Pew “Sub-Saharan Africa” 2010a (Q95e, page 293). Combined 36-country Muslim sample size = 34744.

Countries Included in or Excluded from the Analysis: 36 included, Thailand excluded.

Q92. “Do you favor or oppose the following? … d. stoning people who commit adultery”

Table 2.2


Q92. Do you favor or oppose the following: …d. stoning people who commit adultery?


Favor
Oppose
Don't Know
Refused
Total
Population
306014063
258513394
29205305
7040838
600773600
Percentage
50.94
43.03
4.86
1.17
100.00
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.
                   


3. Punishments Like Whippings and Cutting Off of Hands for Crimes Like Theft and Robbery

Survey Report: Pew Global Attitudes, 2010b, “Muslim Publics Divided on Hamas and Hezbollah,” (Q108b, page 35). Combined 7-country Muslim sample size = 6688.

Q108. “Do you favor or oppose making the following the law in (survey country)? … b. Punishments like whippings and cutting off of hands for crimes like theft and robbery”

Table 3.1.


Q108. Do you favor or oppose making the following the law in (survey country)?
… b. Punishments like whippings and cutting off of hands for crimes like theft and robbery


Favor
Oppose
Don't Know
Refused
Total
Population
198765675
162968130
12305379
1652816
375692000
Percentage
52.91
43.38
3.28
0.44
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 (Q92c, page 220) and Pew “Sub-Saharan Africa,” 2010a (Q95d, page 292). Combined 36-country Muslim sample size = 34744.

Countries Included in or Excluded from the Analysis: 36 included, Thailand excluded. 

Q92. “Do you favor or oppose the following? … c. punishments like whippings and cutting off of hands for crimes like theft and robbery”

Table 3.2


Q92. Do you favor or oppose the following: …c. punishments like whippings and cutting off
of hands for crimes like theft and robbery?


Favor
Oppose
Don't Know
Refused
Total
Population
298187526
273862422
24064991
4658660
600773600
Percentage
49.63
45.58
4.01
0.78
100.00
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.



4. Making Sharia (Islamic Law) the Official Law of the Land in the Country

Q79a (page 201) in Pew’s (2013) The World’s Muslims, full report; and Q95a (page 289) in Pew’s (2010a) “Sub-Saharan Africa” full report. Combined 36-country Muslim sample size = 35166.

Countries included or excluded: See Comments below Table 4.

Q79 “Do you favor or oppose a. making the sharia, or Islamic law, the official law of the land in our country?”

Table 4


Q79. Do you favor or oppose a. making the sharia, or Islamic law, the official law of the land in our country?


Favor
Oppose
Don't Know
Refused
Total
Population
420564114
146095099
33217272
11235515
611112000
Percentage
68.82
23.91
5.44
1.84
100.00
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.

Comment about Table 4: The total population here differs from that of the Q92 series because Morocco (standard Q79a) is included and Russia is excluded. 

Excluded countries with non-standard wording were Iran, Russia, and Thailand. In Russia and Thailand the wording was modified to refer to the prospect of making sharia the law within specific regions within the country instead of the country as whole, e.g., in Russia, (Q79aRUS) “…the Muslim Republics of Russia,” and in Thailand, (Q79aTHA) “…the provinces where the Muslim population forms a majority.” In Iran, the sharia question was as follows: (Q80) “Do you favor or oppose the implementation of the sharia, or Islamic law in our country?” 

Results for the excluded countries (% favor, oppose, dk/ref): Iran 83, 15, 2; Russia 42, 37, 21; Thailand 77, 11, 12.



5. “Honor” Killing of Either a Male or Female

This analysis combines male and female honor killing responses into one response set.

Survey Report: Pew, “The World’s Muslims,” 2013, full report (see Q53 and Q54, page 190). Combined 23-country Muslim sample size = 28894.

Countries Included in or Excluded from the Analysis: 23 included, Thailand excluded. Niger is included here, as it is in the data set with the other countries surveyed between 2011 and 2012, though Pew (2013) did not report it on page 190.

Q54 “Some people think that if a woman engages in premarital sex or adultery, it is justified for family members to end her life in order to protect the family’s honor. Others believe that this practice is not justified, no matter the circumstances. Do you personally feel that this practice is often justified to defend the family honor, sometimes justified, rarely justified, or never justified?”

Q53 asks about killing a male family member. It has the same wording as Q54, except that it uses “man” and “his.” In the alternative version (Q53AIU and Q54AIU) fielded in Afghanistan, Iraq, and Uzbekistan, the phrase “brings dishonor to his/her family” was used instead of the standard “engages in premarital sex or adultery.”

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

My classification: moderate option = 4, clear extremist option <= 2.

Comment: The response that honor killings are “rarely justified” is, in my opinion, not moderate. The “rarely justified” response indicates extremism, but not as strongly compared to the “often” or “sometimes justified” options. The criteria for categorizing a response as “extremist” should be fairly stringent in these analyses, so I decided not to lump “rarely” with “often” or “sometimes.” 

Setting up for a combined analysis of two (or more) questions requires some familiarity with software that can be used to compute new variables out of existing variables. In Appendix B, I describe briefly how to create a filter variable in PSPP. Those instructions are also applicable to computing variables for the combined analysis of Q53 and Q54. Here are some brief 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. Expressions for the alternative versions Q53AIU and Q54AIU follow the same pattern as for Q53 and Q54. To obtain the percentages, enter the newly computed variables instead of the original variables into the Crosstab analysis.

Numeric Expressions:

Often or sometimes justified for one or the other or both: (Q53 <= 2) | (Q54 <= 2)

Rarely for both: (Q53 = 3) & (Q54 = 3)

Never and rarely, or vice versa: ((Q53 = 4) & (Q54 = 3)) | ((Q53 = 3) & (Q54 = 4))

Never for both: (Q53 = 4) & (Q54 = 4)

Table 5


“Honor” Killing: Combined Analysis of Responses to Questions 53 (man) and 54 (woman).

“Some people think that if a man/woman engages in premarital sex or adultery, it is justified for family members to end his/her life in order to protect the family’s honor. Others believe that this practice is not justified, no matter the circumstances. Do you personally feel that this practice is often justified to defend the family honor, sometimes justified, rarely justified, or never justified?”


Percentage
Population

Often/
Both
One Rarely,
Both
Remainder

Sometimes
Rarely
One Never
Never
Afghanistan*
66
6
6
19
3
14581600
Iraq*
62
6
6
19
7
15698400
Egypt
48
10
21
14
7
47868400
Jordan
43
2
23
31
2
3347200
Pakistan
42
3
2
42
10
97540000
Palestin. T.
41
7
4
40
8
1989600
Lebanon
41
6
5
43
4
1816000
Bangladesh
40
16
6
31
7
83462000
Niger
36
17
1
37
9
6741600
Uzbekistan*
32
7
10
41
10
17083200
Tajikistan
31
10
10
41
8
3741200
Tunisia
29
6
6
54
5
7416400
Kyrgyzstan
23
11
9
49
9
2933200
Russia
22
6
5
54
12
10832400
Turkey
21
3
5
63
8
48730000
Malaysia
21
8
3
55
12
11069200
Kosovo
20
13
3
57
6
1107200
Albania
18
8
3
64
6
1846800
Morocco
16
3
5
57
18
21170800
Indonesia
9
8
3
79
1
141766400
Bosnia-Hrz.
8
9
4
76
2
1362800
Azerbaijan
8
2
5
79
6
6556800
Kazakhstan
6
4
7
78
5
7558000

Population
164575214
43746320
31666985
279492613
36738069
556219200
Pop.-Wtd. %
29.59
7.86
5.69
50.25
6.60
100.00
Notes. 23-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, Iraq, and Uzbekistan, the phrase “brings dishonor to his/her family” was used instead of “engages in premarital sex or adultery.”



6. Whether a Wife Must Always Obey Her Husband

Survey Report: Pew, The World’s Muslims, 2013, full report, Q78, page 200. Combined 23-country Muslim sample size = 28894.

Countries Included/ Excluded: 23 included, Niger included, Thailand excluded.

Q78 “Now I am going to read you a statement. Please tell me if you completely agree with it, mostly agree with it, mostly disagree with it or completely disagree with it: A wife must always obey her husband.”

Table 6

Q78. Please tell me if you completely agree, mostly agree, mostly disagree or
completely disagree: A wife must always obey her husband.

Completely
Mostly
Mostly
Completely
Don’t
Refused
Total

agree
agree
disagree
disagree
know
Population
305176143
172199148
44689038
21515340
9521618
3117913
556219200
Percentage
54.87
30.96
8.03
3.87
1.71
0.56
100.00
Notes. 23-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Research, The World’s Muslims 2012 Dataset.

Comment on Table 6: About 86% agree at least mostly, and about 12% disagree at least mostly, that a wife “must always obey” her husband.


7. Whether a Wife Should Have the Right to Divorce Her Husband

Survey Report: Pew, The World’s Muslims, 2013, full report, Q77, page 199. Combined 22-country Muslim sample size = 27385.

Countries Included or Excluded: 22 included, Niger included, Afghanistan excluded (Q77 was not asked there), Thailand excluded.

Q77 “I will read you two statements, please tell me which comes closer to your view, even if neither is exactly right.
1—A wife should have the right to divorce her husband
OR
2—A wife should not have the right to divorce her husband”

Table 7

Q77. “I will read you two statements, please tell me which comes closer to your view, even if neither is exactly right.”

A wife should have the right to divorce her husband
A wife should not have the right to divorce her husband
Neither/ depends (vol.)
Don’t know
Refused
Total
Population
236244564
243407513
41359315
15968599
4657609
541637600
Percentage
43.62
44.94
7.64
2.95
0.86
100.00
Notes. 22-country population is of Muslims age 18 and older. Displayed numbers are rounded.
Data Source: Pew Research, The World’s Muslims 2012 Dataset.

Comment: A plurality (45%) says that a wife should not have the right to divorce. An additional 8% volunteered answers that Pew categorized as “neither/depends.”





Discussion

Since the terrorist attacks of September 11, 2001, debates have carried on among the general publics in the West about the prevalence of extremist or hardline fundamentalist views in the world’s Muslim population. Pew’s (2013) survey evidence happens to be relevant to these debates because Pew asked multiple questions about harsh fundamentalist views and sampled a broad range of countries and vast numbers of Muslim adults. The analyses reported here on Pew’s data indicate that majorities, pluralities, and large minorities of Muslim adults, in vast numbers across many countries, favor extremely harsh fundamentalist practices and agree with severe restrictions on the rights and freedoms of women. A substantial majority (69%) of Muslim adults favors making sharia the law of their country.

How representative of Muslim adults generally are these results? Pew’s 36-country sample that was asked the three harsh punishment questions was very diverse and represents roughly 60% of the world’s Muslim adults. It is probably reasonable to expect that the percentages in favor of such harsh punishments would be close to but slightly higher for the remaining 40% who were not surveyed or not asked those questions. Pew (2013, pp. 147-149) was not able to survey some countries, or did not ask some questions in some surveyed countries, due to cultural hypersensitivities, safety and security concerns, active conflict, and/or government restrictions. Overall, those factors suggest that the sample Pew did obtain was biased toward more “moderate” views. According to Pew (2013), “In some countries, pretest results indicated the need to suppress certain questions to avoid offending respondents and/or risking the security of the interviewers. In other countries, interviewers considered some questions too sensitive to pretest.” (p. 147).  Pew (2013) did not ask the three harsh sharia punishment questions in Iran, Morocco, or Uzbekistan; did not survey India, the GCC (Saudi Arabia, UAE, et al.), Sudan, Yemen, China, Syria, Burkina Faso, Somalia, Guinea, Ivory Coast, or Libya. The 75% nationally representative data for Algeria, not included in Pew’s (2013) report, indicated that clear majorities there favored each of the three harsh punishments. India’s Muslims were not surveyed because Pew determined that their questions would be deemed too sensitive to ask there. 




Disclaimer

Pew Research is not responsible for my analyses or my interpretation of their data. From the Pew instructions for downloading data sets: “All manuscripts, articles, books, and other papers and publications using our data should reference the Pew Research Center’s Religion & Public Life Project as the source of the data and should acknowledge that the Pew Research Center bears no responsibility for the interpretations presented or conclusions reached based on analysis of the data.”



References and Notes


[1] Tolerance and Tension: Islam and Christianity in Sub-Saharan Africa, Pew Forum on Religion & Public Life (April, 2010), full report pdf. http://www.pewforum.org/files/2010/04/sub-saharan-africa-full-report.pdf

[2] Muslim Publics Divided on Hamas and Hezbollah, Pew Research Center Global Attitudes Project (December 2, 2010) http://www.pewglobal.org/files/2010/12/Pew-Global-Attitudes-Muslim-Report-FINAL-December-2-2010.pdf

[3] The World’s Muslims: Religion, Politics and Society, Pew Research Center (April 30, 2013), full report pdf http://www.pewforum.org/files/2013/04/worlds-muslims-religion-politics-society-full-report.pdf

[4] The corresponding data sets drawn from here for the analyses include the Pew Forum on Religion & Public Life’s Sub-Saharan Africa Survey, Pew Global Attitudes Spring 2010 Survey Dataset, and Pew Religion’s The World's Muslims Dataset 2012. These data set packages contain multiple files in addition to the data files proper, such as a Codebook and detailed Questionnaire.

[5] Pew’s religion-specific age structure estimates for 2010 can be obtained through the following interactive website: http://globalreligiousfutures.org/countries. A detailed report for these estimates is in The Global Religious Landscape: A Report on the Size and Distribution of Major Religious Groups as of 2010, The Pew Forum on Religion and Public Life (December, 2012) http://www.pewforum.org/files/2014/01/global-religion-full.pdf




Appendix A


Estimating the Age 18-and-Over Muslim Populations

In its opinion surveys examined here, Pew surveyed only Muslim adults, i.e., those 18 years of age or older. However, many authors who try to estimate the numbers and population-weighted percentages of Muslims who hold extremist views use in their calculations the entire (all-ages) populations of Muslims in each country surveyed. That is questionable for at least two reasons. First, we don’t know how closely the responses of younger teens and children would match those of the adults at the time of the survey. Also, obviously, it would not make sense to generalize these results to infants and toddlers. Second, there are big differences between countries in age structure (or age distribution). For example, the 0-14 age group makes up about 46% of the Muslim population in Afghanistan but only about 26% in Turkey. To ignore these differences and use the overall all-ages Muslim populations for each country to weight the percentages of Muslims who hold a particular view would result in estimates that are less accurate than those that use the age ≥ 18 Muslim population sizes to weight the percentages.

From what I've seen, the previous estimates of the percentages of Muslims who hold particular views also make use of the rounded percentages displayed in Pew’s published tables and figures. Estimates based on rounded figures will be somewhat inaccurate. To improve the accuracy of the estimates, percentages (or proportions) displayed to several decimal places can be obtained easily from the data sets that Pew has published (see Appendix B, below). Original figures accurate to several decimal places are also useful in maintaining a high degree of accuracy in intermediate calculations and in checking for errors.

Another type of estimate of the percentage of Muslims favoring harsh sharia punishments uses only the subset of respondents who favor both sharia as the official law of the nation (as phrased in Q79a, in Pew, 2013, The World’s Muslims) and the specific harsh sharia punishment, such as the death penalty for apostasy. Authors who make this type of estimate perhaps assume that they can ignore the respondents who opposed, or who gave no definite response to, the prospect of making sharia the law of country. That assumption is incorrect, because there is in fact a significant minority of respondents who either opposed sharia (as presented in Q79a) or indicated that they didn’t know, but who favored specific harsh sharia punishments. In addition, some authors assume that only respondents who favored sharia as the national law (Q79a) were asked about the specific harsh punishments for apostasy, theft, and adultery. That assumption is also incorrect, as I confirmed with the primary researcher and—as readers can confirm—by checking the data file. In countries where the sharia question and the three questions about the harsh sharia punishments were asked, all Muslim respondents were asked all four of these questions. Hence, those who make these errors underestimate the total percentages of Muslims who favor harsh punishments.

Pew [5] published religion-specific age structure estimates for many countries, including those surveyed. However, those estimates are in groupings for ages 0-14, 15-29, etc.; they did not make an age ≥ 18 group. That limitation will be dealt with, with sufficient accuracy for the present purposes, by using Pew’s published age group numbers to estimate the numbers of Muslims 18 years of age and older. I made a linear estimation that uses the simple assumption that there are equal increments of difference between each single-year age group’s population and the one adjacent to it. With that assumption, one can use Pew’s published population size estimates for the age 0-14 and age 15-29 groups, and the age ≥ 15 group, to estimate the size of the age ≥ 18 group. There will be a very slight loss of accuracy using a linear estimation, but that loss should be small enough that it does not have a significant impact on the results reported here. The linear method I describe below is quick and easy to use in a spreadsheet.

Step 1: Find d, the mean increment of population difference from a single-year age group to the next. The formula for d = [(x / 15) – (y / 15)] / 15, where x is the population size of the age 15-29 group, y is the population size of the age 0-14 group, 15 is the number of years in each age group, and 15 is also the number of differences from one year to the next (adjacent) year between ages 22 and 7. Ages 22 and 7 are the median years of their respective age groups.

Step 2: Find the population sizes of the age 17, 16, and 15 (single-year) groups, respectively. The numbers of single-year increments from 22 for each of those groups are 5, 6, and 7 respectively. Hence the population size for the age 17 group = (x / 15) + (d × 5), for the age 16 group = (x / 15) + (d × 6), and for the age 15 group = (x / 15) + (d × 7).

Step 3: Obtain the sum of the age 17, age 16, and age 15 populations.

Step 4: Subtract the sum of the age 17, 16, and 15 populations from the age ≥ 15 population. The result is the linear estimate of the age ≥ 18 population.


Table A1


Estimated* Muslim Population Sizes Based on Pew’s 2010 Age Structure Numbers

Age Group
≤ 14 yrs.
17 yrs.*
16 yrs.*
15 yrs.*
15+16+17*
≥ 15 yrs.
≥ 18 yrs.*
Afghanistan
14,550,000
706,444
732,800
759,156
2,198,400
16,780,000
14,581,600
Albania
600,000
44,889
44,400
43,911
133,200
1,980,000
1,846,800
Azerbaijan
1,880,000
161,333
157,733
154,133
473,200
7,030,000
6,556,800
Bangladesh
42,750,000
2,723,333
2,736,000
2,748,667
8,208,000
91,670,000
83,462,000
Bosnia-Hrz.
270,000
22,889
22,400
21,911
67,200
1,430,000
1,362,800
Cameroon
1,660,000
80,000
83,067
86,133
249,200
1,930,000
1,680,800
Chad
2,820,000
138,222
143,200
148,178
429,600
3,390,000
2,960,400
DR Congo
470,000
21,556
22,533
23,511
67,600
500,000
432,400
Djibouti
310,000
18,444
18,667
18,889
56,000
550,000
494,000
Egypt
24,500,000
1,530,222
1,540,533
1,550,844
4,621,600
52,490,000
47,868,400
Ethiopia
12,890,000
637,111
659,333
681,556
1,978,000
15,800,000
13,822,000
Ghana
1,560,000
80,000
82,400
84,800
247,200
2,300,000
2,052,800
Guinea-Biss.
290,000
14,889
15,333
15,778
46,000
390,000
344,000
Indonesia
56,060,000
3,767,556
3,764,533
3,761,511
11,293,600
153,060,000
141,766,400
Iran
16,900,000
1,488,000
1,451,867
1,415,733
4,355,600
56,680,000
52,324,400
Iraq
13,540,000
681,778
703,867
725,956
2,111,600
17,810,000
15,698,400
Jordan
2,260,000
132,444
134,267
136,089
402,800
3,750,000
3,347,200
Kazakhstan
3,110,000
207,333
207,333
207,333
622,000
8,180,000
7,558,000
Kenya
1,850,000
91,778
94,933
98,089
284,800
2,080,000
1,795,200
Kosovo
470,000
30,889
30,933
30,978
92,800
1,200,000
1,107,200
Kyrgyzstan
1,460,000
99,111
98,933
98,756
296,800
3,230,000
2,933,200
Lebanon
640,000
44,889
44,667
44,444
134,000
1,950,000
1,816,000
Liberia
210,000
9,556
10,000
10,444
30,000
270,000
240,000
Malaysia
5,970,000
348,667
353,600
358,533
1,060,800
12,130,000
11,069,200
Mali
6,790,000
329,111
341,467
353,822
1,024,400
7,720,000
6,695,600
Morocco
8,940,000
607,556
606,400
605,244
1,819,200
22,990,000
21,170,800
Mozambique
1,870,000
88,667
92,267
95,867
276,800
2,330,000
2,053,200
Niger
7,480,000
332,889
349,467
366,044
1,048,400
7,790,000
6,741,600
Nigeria
37,300,000
1,711,111
1,788,667
1,866,222
5,366,000
39,990,000
34,624,000
Pakistan
59,120,000
3,543,556
3,583,333
3,623,111
10,750,000
108,290,000
97,540,000
Palestin. T.
1,670,000
87,778
90,133
92,489
270,400
2,260,000
1,989,600
Russia
2,780,000
226,667
222,533
218,400
667,600
11,500,000
10,832,400
Senegal
5,260,000
269,778
277,867
285,956
833,600
6,720,000
5,886,400
Tajikistan
2,460,000
148,000
149,600
151,200
448,800
4,190,000
3,741,200
Tanzania
6,680,000
350,667
360,133
369,600
1,080,400
9,100,000
8,019,600
Tunisia
2,450,000
186,889
184,533
182,178
553,600
7,970,000
7,416,400
Turkey
18,790,000
1,268,222
1,266,667
1,265,111
3,800,000
52,530,000
48,730,000
Uganda
1,860,000
90,667
94,000
97,333
282,000
1,990,000
1,708,000
Uzbekistan
7,830,000
548,222
545,600
542,978
1,636,800
18,720,000
17,083,200
Sum
378,300,000
22,871,111
23,106,000
23,340,889
69,318,000
760,670,000
691,352,000
Notes. The 39 countries above are from Pew’s The World’s Muslims (2013) report. Displayed numbers are rounded.
Data Source: Pew Global Religious Futures Project, 2010 Age Structure Estimates for Muslims.
*Estimates for age groups 15, 16, 17, and ≥ 18 years were made by Empethop based on Pew’s numbers.


The justification for weighting response percentages by the population sizes of the countries should be clear from a perusal of Table A1. The adult Muslim populations of some countries are many times larger than those of others. (Indonesia's adult Muslim population here is about 590 times the size of Liberia's). To estimate the percentages of Muslim adults holding particular views, it is important to take these country population sizes into account. However, if our interest were, for example, to compare the different countries according to their percentages of Muslim adults holding particular views, it could make sense to give countries equal weight.

Those who delve into Pew’s population estimates may notice that there are, for 15 of the 39 countries, discrepancies of ± 10000 between Pew’s published estimates of the total Muslim population and the totals obtained by summing the Muslim population age groups in Pew’s age structure estimates. (For the other 24 countries there were no discrepancies between the totals). The discrepancies sum to -30000 across 39 countries, a number that is too small to have a significant impact on the results reported here for overall population-weighted percentages of response types. (Note: Pew updates their population estimates from time to time, so the discrepancies that I report here may be gone or may have changed by the time you read this).

The Pew opinion surveys reported above were carried out starting in December of 2008 for the sub-Saharan African study and ending in February of 2012 for The World’s Muslims study. I chose 2010 as the year for the population estimates because Pew published estimates for that year, and it is the same year as the Pew Global (2010) survey, or mostly within one year of the start of the Sub-Saharan and the end of the World’s Muslims survey period. In addition, doing my own separate estimates for each year in the range would be more time-consuming and, I believe, would not gain much more accuracy. For the 36 countries included in the analyses of the three harsh punishment items, the estimated age ≥ 18 Muslim population is 600773600. The general, all-ages Muslim population for those 36 countries is 1,006,890,000, obtained from Pew’s numbers. Hence, the (2010) age ≥ 18 Muslim population is about 59.67% of total Muslim population of those 36 countries.  Pew’s estimate for the 2010 population of Muslims of all ages worldwide is 1,599,700,000 (see here), and 59.67% of that is roughly 954,481,000. That’s assuming that the Muslim population age structure of those 36 countries combined can be generalized to that of the remaining countries that were not surveyed.




Appendix B



Using PSPP to Work on the Pew Data Files

If you choose to attempt to replicate the entire set of results reported in the analyses above, I should caution that that would probably be a very time-consuming project for most people. I would recommend for those who have some (but not a great deal) of experience working with data that a more manageable project would be to focus on verifying the results for one question that interests you most. For those who have little to no experience working with data, I would recommend that you not attempt to replicate these results unless you have some guidance from someone who is experienced, or until you learn the knowledge and skills needed.

In a previous article I gave some instructions for how to easily open up a Pew data file and do a simple analysis of the response option percentages for particular questions for each country. Readers who have adequate experience in using spreadsheets and an introductory knowledge of statistics and research methods can apply the instructions to reproduce and verify the results shown in the present article using the freely-available statistical software PSPP. It will be necessary to use a spreadsheet such as Excel, or the freely-available Calc, or others, with which to work on the PSPP output. You can also verify the results reported in this article using software other than PSPP, but obviously some of the instructional details will differ between systems. I recommend PSPP for these simple analyses because it is free, easy to use, and can handle the large Pew data files.

In calculating population percentages and numbers, it is important to have original figures that are accurate to several decimal places. These are available only from Pew’s data files. PSPP is set by default to display results of analyses (i.e., output numbers) to two decimal places. If you want to reproduce the results reported above, maintaining a good level of accuracy through the intermediate calculations, I suggest that you set PSPP to display the output percentages to at least nine (9) decimal places. The brief syntax for this modification can be cut and pasted from the PSPP site into a syntax file. Change the 6 in their example SET FORMAT F22.6. to a 9 when you enter it into a PSPP syntax editor (see Figure 1 below). Save it as a PSPP syntax file.

Figure 1





When you have the data file open, open this syntax file, and select Run and All before doing the analysis. That means PSPP will run the command to display output to the specified number of decimal places for all subsequent analyses, as long as that syntax file is open and is not modified further. 

Calculating the population percentages and numbers.

Carry out a simple Crosstabs analysis, with country as rows and the question as columns, to obtain a table of the response choice percentages for each country (see instructions in a previous article). Make sure the appropriate weight variable is turned on for the analysis. Once you’ve obtained the response choice percentages precise to several decimal places, the next step is to export this PSPP output to a spreadsheet (e.g., Excel, Calc, etc.) where the percentages can be converted to proportions. To export the output, go to the Output window and click on File, then click on Export. Name the file and save it in .csv format and to a folder for your outputs for this project. You can then go to that file and open it with your preferred spreadsheet program. [When you open it in the spreadsheet, then format the percentages as proportions, showing, for example, numbers carried to 11 decimal places if your output percentages were displayed to 9 decimal places. Save the numbers in the proportion format (e.g., to 11 decimal places) before closing the file]. Note that to reproduce the results reported here, you will need to make several such output files. Enter the estimated adult Muslim populations of the relevant countries (e.g., see Appendix A). The population-weighted percentages of responses are obtained by the following steps, for each question:

1. Multiply the proportion for each response (e.g., favor, oppose, don’t know, refused) for each country by the estimated adult Muslim population for each country. These products are the estimated numbers of Muslim adults in each country for each response.

2. For each response, sum the products in the previous step. (E.g., if there are 36 countries, obtain the sum of the 36 products).

3. Divide each of the sums obtained in the previous step by the sum of the adult Muslim populations of all the relevant countries. The quotients are the proportions of adult Muslims over all the relevant countries for each type of response.

A Note on Analyzing the Sub-Saharan African (2009) Data

Pew’s 2012 The World’s Muslims dataset has only data for Muslims, but their (2009) sub-Saharan African data set includes data for multiple religious groups. To select only the data for Muslims, it is necessary to filter non-Muslim cases out of the analysis. Respondent’s religion is recorded in the “RELIGrec” variable just after Q31. Muslim is coded as 2 in the RELIGrec column. Use the Compute function, located in the Transform menu, to create a filter variable, as shown in this example where I’ve named the filter variable “Muslim” (see Figure 2, below).

Figure 2





Then press OK. This procedure creates a binary variable where all Muslim cases are coded as one (1) and all non-Muslim cases as zero (0). You can then use “Select Cases” (located under the Data menu) to put on your Muslim filter variable for the analysis. Select “Use filter variable,” scroll down to and then click on the “Muslim” variable, then press OK to put on that filter. If you are planning to do numerous analyses of the Muslim data, for convenience you may wish to make a separate data file with only the Muslim data. To do that, save a new copy of the file (e.g., name it to indicate it is only Muslim data), and in the new copy follow the above steps but, in the “Select Cases” dialog box, choose the “Deleted” option for unselected cases, then save.



Appendix C


Margins of Error and Other Sources of Variability

The combined samples are much larger than those reported for the individual countries shown on page 150 in Pew, 2013; page 66 in Pew's 2010 sub-Saharan report; and page 21 in Pew Global, 2010. Therefore one can expect the margins of error for the combined samples to be much smaller. The details of how Pew calculated the margins of error for each individual country’s sample are not published, but they are evidently complex, taking into account factors such as their weighting and stratified random sampling schemes (e.g., see pages 147-148 in Pew, 2013). One can still get a rough idea of the size of the margins of error by using online calculators (e.g., here and here). Using these calculators that assume simple random sampling, the Pew Global (2010) 7-country combined (raw, unweighted) sample of 6688 would have a margin of error of about ± 1.2%, or roughly plus or minus one percent, assuming a response choice proportion of .5 (50%), with 95% confidence. (Response choice proportion is the proportion of respondents who chose a particular answer, e.g., the proportion who said they favored stoning of adulterers). Margin of error decreases as response choice proportion deviates from .5 toward 1 or 0. These calculators also indicate that the Pew (2013) 36-country combined (raw, unweighted) sample of 34744 examined here would have a margin of error of about ± 0.53%, or roughly plus or minus half of one percent, again assuming a response choice proportion of .5, with 95% confidence. While these simple margin of error estimations are not strictly appropriate for Pew’s data, they are based on large samples and so are probably not far off of what the appropriate estimates would be.

The statistical margins of error should also be viewed in broader perspective as one kind of variability among multiple potential sources of variability. For example, the level of support or opposition to some items may have changed over the years since the surveys were conducted (2008-2012). Slight changes in the wording of questions can sometimes produce significant changes in response choices. The overall level of support or opposition to some items could also change if Pew were to include additional countries in the survey, such as Saudi Arabia, Libya, Somalia, Sudan, Yemen, Syria, and so on. (That said, given that Pew sampled numerous countries that span a wide spectrum of opinion, and the majority of the world’s Muslim adults, it is likely that the averages that I report here are not very far off of what they would be had Pew sampled from all the world’s Muslim adults).

No comments:

Post a Comment