Peer Effects of the Same and Different Religions on Faithfulness: A Comparison between Indonesia and India

Abstract

By employing questionnaire surveys to empirically examine peer effects on religious faithfulness, this study mainly compares Muslims in Indonesia and India as examples. This study uses religious restrictions on foods as the main component of the questionnaire. A total of two variables were selected to examine peer effects: (1) the percentage of respondents’ close friends who follow a different religion and (2) the percentage of people in the respondents’ city who follow the same faith. Ordinary least squares/generalized least squares regression was conducted, and six models were estimated. The results reveal that Indian/Indonesian respondents are more affected by those who follow the same/different religions, respectively, suggesting that relatively smaller groups have larger peer effects on religious faithfulness. Although further investigations are required, these symmetric results may be attributed to the fact that tensions among people from different religions are high/low, and that the percentage of people who follow a different faith in the respondents’ city is high/low in India and Indonesia, respectively.

1. Introduction

It is often the case that a variety of religions exist in a country, and people follow different faiths (Taylor 2002; Bellah and Hammond 2013). There are multiple main religions in some countries (Type A countries), while a certain religion dominates in other countries (Type B countries). Type B countries include Indonesia and India, where 86.7% and 80.5% of people follow Islam and Hinduism, respectively (Kementerian Dalam Negeri 2020; Ministry of Home Affairs 2020). Religious faithfulness or religiosity can be affected by the same and other religious groups, but there seems to be a lack of accumulated scientific knowledge. The present study seeks to fill this knowledge gap. To examine the peer effects on religious faithfulness, Type B countries may be more relevant. People’s attitudes toward religious taboos, such as that which is considered haram for Muslims, are selected as the variables of spiritual faithfulness because such taboos, especially those for foods, are the most understandable, and people face it daily.

If there are some main religions in a country (i.e., Type A countries), people may be more careful with their religious restrictions. Malaysia, where Islam is the official religion, can be regarded as a typical Type A country, and Malay people are expected to follow Islam (Hooker 2004; Shamsul 2001), but other religions also account for a large portion of the population. The share of Islam is 61.3%, while other religions such as Buddhism, Christianity, and Hinduism account for 19.8%, 9.2%, and 6.3%, respectively (Department of Statistics Malaysia 2011). Because nearly 40% of the population are non-Muslim, halal certification plays a vital role in distinguishing halal products for Malaysian Muslims (Abdul et al. 2009; Talib et al. 2017; Nasirun et al. 2019). The Malaysian halal certificate is one of the most respected and influential Islamic credentials, and Malaysia is becoming a leading country in the global halal market (Mahiah Said et al. 2014; Shirin Asa 2017; Kawata et al. 2018). Both Muslims and non-Muslims recognize the Malaysian halal logo; for example, based on a recent study, for the question “have you ever seen the Malaysian halal logo?” 561 (80%) and 139 (20%) of Malaysian respondents selected “definitely yes” and “probably yes,” while no respondents selected “maybe yes,” “maybe no,” or “definitely no” (Kawata and Salman 2020). Malaysian Muslims may be careful in selecting products in their daily lives to better follow religious beliefs (Muhamad et al. 2016; Haque et al. 2018). To do so, they may maximize the use of halal certification.

The situation may be different if a particular religion is dominant (Type B countries). For example, while Indonesian Muslims are faithful to Islamic teachings (Hasan 2009), they are not concerned about halal certification when purchasing products because most products sold in Indonesia are halal (Viverita et al. 2017; Anwar et al. 2018; Nusran et al. 2018). This situation may be similar to that in India. It can be inferred that people who live in countries where a particular religion dominates (Type B countries) are more affected by people of both the same and other religions while being less concerned about certification. Thus, people in these countries are more appropriate as subjects for this study. Thus, in this study, we selected Type B countries to examine peer effects. Indonesia and India were selected because Indonesia is the largest Muslim-majority country (Roslan Mohd Nor and Malim 2014) and India is the largest Hindu-dominant country. There are two methods for examining peer effects. One involves comparing the religious majority of each country (that is, Muslims in Indonesia and Hindus in India), and the other involves comparing those from the same faith. This study employed the latter comparison. Since people from different religions may have different traditions, religious habits, and ways of thinking, the former method of comparison would be more complicated.

People may be affected by people of the same and other religions in their daily lives. Using data obtained through a questionnaire survey, this study selected the following two variables to examine peer effects on people’s religious faithfulness: (1) the percentage of close friends whose religion is different from that of the respondent, and (2) the rate of people in the respondents’ city who believe in the same faith. Both close friends and people living in the same city are expected to significantly affect religious faithfulness (Yasid et al. 2016). It is also essential to select a criterion that appropriately represents religious faithfulness. Since the main targets of this study are Muslims, while replies from followers of other religions can be included, evident and common criteria such as halal for Muslims are preferable (Othman et al. 2016). This study creates such criteria based on respondents’ self-evaluation of their product (mainly foods) selection (Sun et al. 2012). If people follow either Islam or Hinduism, they are expected to avoid consuming pork and beef in their daily lives; these commonly consumed meats are easily obtainable in Indonesia and India (Meyer-Rochow 2009; Erwanto et al. 2014). Thus, a promising criterion of religious faithfulness can be created by asking the respondents the degree to which they adhere to religious rules regarding taboo subjects/aspects in their daily lives.

This study empirically presents the symmetric results of peer effects on religious faithfulness regarding the above two variables for Indonesian and Indian Muslim residents.

2. Materials and Methods

2.1. Questionnaire Survey

The questionnaire was circulated in both Indonesia and India. The Indonesian survey was conducted in the middle of September 2020 via the Internet for citizens in and around Jakarta (Jabodetabek). The Indian surveys were conducted between September and December using snowball sampling in Bengaluru, Chennai, Cochin, Hyderabad, and Vijayawada. The religious ratios of Jakarta and the five study sites in India are tabulated in

Table 1. The religious ratios in the study sites (unit: %).

	Indonesia	India
	Jakarta	Bengaluru	Chennai	Cochin	Hyderabad	Vijayawada
Muslim	83.7	13.0	9.5	17.6	43.5	9.1
Christian	12.5	5.3	7.7	38.1	2.2	3.6
Hindu	0.2	80.3	80.7	43.8	51.9	85.2
Buddhist	3.6	0.1	0.1	0.0	0.0	0.0
Others	0.0	1.4	2.0	0.5	2.4	2.1

Sources: Kementerian Agama (2022), Census Organization of India (2022).

. As this study mainly compares Muslims in both countries, five study areas of different Indian states with varied Muslim ratios were selected (Table 1).

The following questions were asked: place of residence (Q1), gender (Q2), age (Q3), marital status (Q4), religion (Q5), educational background (Q6), occupation (Q7), if respondents had family members (including friends living together) who believed in a different religion (Q8), the percentage of respondents’ close friends who believe in a different religion (Q9), and the percentage of people who believe in the same religion in the respondents’ city (Q10). Moreover, 11 questions regarding the religious restrictions on foods (Q11–Q21) were also investigated using a 5-point Likert-type scale. The 11 questions are presented in

Table 2. Questions on the religious restrictions on foods.

Awareness	✧ Q11: You can differentiate taboo and non-taboo foods ✧ Q12: You avoid buying taboo foods ✧ Q13: You eat only non-taboo foods
Religious Belief	✧ Q14: You avoid taboo foods due to your religious beliefs ✧ Q15: You select non-taboo foods because your religion teaches you to do so
Health Reasons	✧ Q16: You believe non-taboo foods are good for your health ✧ Q17: You believe the consumption of taboo foods would harm your health
Logo Certification	✧ Q18: You always check if a food logo for your religion is put on the food ✧ Q19: If you do not find a logo, you will check the ingredients
Exposure	✧ Q20: You always update your knowledge about taboo and non-taboo foods ✧ Q21: You often talk about taboo foods with family members, friends, and/or colleagues

.

2.2. Analytical Methods

The primary purpose of this study was to examine the peer effects on religious faithfulness. The ordinary least squares (OLS) and generalized least squares (GLS) regressions were applied for this purpose. The index of faithfulness was constructed using the average values of Q11–Q21 and was used as the explained variable $y_i$ (

Table 3. Variables.

Variable	Explanation/Sign Condition
Explained variable	Respondents’ religious faithfulness (average value of Q11–Q21)
Explanatory variables
GEN	Gender (Q2) =1 if men =0 if women
AGE	Age (Q3) =1 if 19 or less =2 if 20–29 =3 if 30–39 =4 if 40–49 =5 if 50–59 =6 if 60–69 =7 if 70 or more
MAR	Marital status (Q4) =1 if single =0 if married
CHR	Religion (Q5) =1 if Christian =0 if Hindu or Muslim
HID	Religion (Q5) =1 if Hindu =0 if Christian or Muslim
EDU	Educational background (Q6) =1 if high school or less =2 if college =3 if bachelor =4 if master =5 if doctor Sign condition (+): Respondents’ religious faithfulness is expected to increase with a higher “educational background”.
DIF	Q9 =1 if 10% =2 if 30% =3 if 50% =4 if 70% =5 if 90% Sign condition (–): Respondents’ religious faithfulness is expected to decrease as “the percentage of respondents’ close friends who believe in different religion” increases.
SAME	Q10 =1 if 15% =2 if 30% =3 if 45% =4 if 60% =5 if 75% =6 if 90% Sign condition (+): Respondents’ religious faithfulness is expected to increase as “the percentage of people who believe in the same religion in respondents’ city” increases.
D1	=1 if a respondent is from Hyderabad, India; =0 otherwise
D2	=1 if a respondent is from Bengaluru, India; =0 otherwise
D3	=1 if a respondent is from Cochin, India; =0 otherwise
D4	=1 if a respondent is from Chennai, India; =0 otherwise
D5	=1 if a respondent is from Vijayawada, India; =0 otherwise Note that D5 was removed in India_A and India_M (see Table 4) to avoid perfect multicollinearity.

). The candidate explanatory variables were Q2–Q6, Q9, and Q10. The estimations were performed in R (R Core Team 2019). The regression procedure was as follows: OLS was applied, and the Breusch–Pagan (BP) test was conducted to check for heteroscedasticity. If the H₀: BP test statistic = 0 was rejected, GLS was applied; otherwise, the OLS result was used. After selecting OLS/GLS, the variance inflation factor (VIF) of all variables was calculated to check for multicollinearity. VIF < 10 was typically used as the rule of thumb, and this study followed this criterion.

There are six models as tabulated in

Table 4. Models.

	Muslim + Non-Muslim (A)	Muslim (M)
All data (A)	A_A model	A_M model
Indonesian data	Indonesia_A model	Indonesia_M model
Indian data	India_A model	India_M model

.

Specifically, models were constructed as follows:

A_A and A_M models

$$\begin{array}{ll}{y}_i={\beta }_i^0& +{\beta }_i^{D1}D1+{\beta }_i^{D2}D2+{\beta }_i^{D3}D3+{\beta }_i^{D4}D4+{\beta }_i^{D5}D5+{\beta }_i^{GEN}GEN\\ & +{\beta }_i^{AGE}AGE+{\beta }_i^{MAR}MAR+{\beta }_i^{CHR}CHR+{\beta }_i^{HID}HID+{\beta }_i^{EDU}EDU\\ & +{\beta }_i^{DIF}DIF+{\beta }_i^{SAME}SAME+{\epsilon }_i\end{array}.$$

(1)

Indonesia_A and Indonesia_M models:

$$\begin{gathered} y_i={\beta }_i^0+{\beta }_i^{GEN}GEN+{\beta }_i^{AGE}AGE+{\beta }_i^{MAR}MAR+{\beta }_i^{CHR}CHR+{\beta }_i^{HID}HID+ \\ {\beta }_i^{EDU}EDU+{\beta }_i^{DIF}DIF+{\beta }_i^{SAME}SAME+{\epsilon }_i. \end{gathered}$$

(2)

India_A and India_M models:

$$\begin{gathered} y_i={\beta }_i^0+{\beta }_i^{D1}D1+{\beta }_i^{D2}D2+{\beta }_i^{D3}D3+{\beta }_i^{D4}D4+{\beta }_i^{GEN}GEN+{\beta }_i^{AGE}AGE \\ +{\beta }_i^{MAR}MAR+{\beta }_i^{CHR}CHR+{\beta }_i^{HID}HID+{\beta }_i^{EDU}EDU+{\beta }_i^{DIF}DIF+{\beta }_i^{SAME}SAME+{\epsilon }_i \end{gathered}$$

(3)

Note that the dummy variable D5 was removed in the India_A and India_M models to avoid the occurrence of perfect multicollinearity.

In the above models, $\beta$s are parameters, $\epsilon$ is the error term, and $i\in I$ represents the respondents. The questionnaire also asked whether respondents had family members (including friends living together) who believed in a different religion (Q8). Because those who lived with family members whose faith was different were expected to be exceptional, the result of Q8 was examined separately and was not used as the candidate of the explanatory variable of OLS/GLS regressions.

3. Results

3.1. Descriptive Statistics for Q1 to Q7

The descriptive statistics for Q1 to Q7 are presented in

Table 5. Descriptive statistics for Q1 to Q7.

	Total	Indonesia	India
Place of residence (Q1)
Indonesia	131	131
India	450
Hyderabad			90
Bengaluru			90
Cochin			90
Chennai			90
Vijayawada			90
Gender (Q2)
Men	302	61	241
Women	279	70	209
Age (Q3)
19 or less	69	8	61
20–29	88	46	42
30–39	185	25	160
40–49	180	33	147
50–59	53	18	35
60–69	6	1	5
70 or more	0	0	0
Marital status (Q4)
Single (including widow, divorcee)	137	60	77
Married	444	71	373
Religion (Q5)
Muslim	389	102	287
Christian	92	26	66
Hindu	98	1	97
N/A	2	2	0
Education background (Q6)
High school or less	29	1	28
College	80	30	50
Bachelor	216	26	190
Master	232	71	161
Doctor	23	2	21
N/A	1	1	0
Occupation (Q7)
Private sector	206	49	157
Government sector	84	35	49
Self-employed	127	5	122
Housewife	66	0	66
Students	91	35	56
No job/Retired	0	0	0
N/A	7	7	0

. The Indonesian sample size was 131, while the Indian sample size was 90 for each region, resulting in 450 respondents. In total, 581 respondents were included. The percentage of Muslims among the respondents in the five Indian study areas was 68.9% (Bengaluru), 45.6% (Chennai), 85.6% (Cochin), 66.7% (Hyderabad), and 52.2% (Vijayawada); thus, replies from Muslims accounted for the majority of the sample in all the study sites.

3.2. Explanatory Variable Concerning Peer Effects (Q9 and Q10)

The most critical candidates for the explanatory variables were the replies to Q9 and Q10. The results of these questions are presented in Figure 1 and Figure 2. Q9 asked about the percentage of respondents’ close friends who believed in a different religion. As illustrated in Figure 1, the Indian respondents tended to have more close friends who believed in different religions compared to Indonesian respondents. Q10 asked about the percentage of people who believed in the same faith in the respondents’ city. As illustrated in Figure 2, the Indonesian respondents tended to live in cities where the percentage of citizens with the same religion was higher.

3.3. Explained Variable (Q11–Q21)

The index of faithfulness for each respondent $i$ was constructed using the average values of Q11–Q21 and was used as the explained variable $y_i$ for all the models. The scores of Q11–Q21 for the Indonesian respondents (n = 115–117) and Indian respondents (n = 281) are presented in Figure 3. Some respondents did not answer this question because they did not think they were under religious restrictions, such as that which is considered haram for Muslims. Such samples were not included in the estimation because the value of $y_i$ was not available. If they provided answers to most of the questions in Q11–Q21 but not for some, these data were used in the analyses. As illustrated in Figure 1, the Indonesian respondents provided higher scores for most questions. The exceptions were exposure (Q20 and Q21). The most apparent difference between the Indonesian and Indian respondents was observed for Q12 ("You avoid buying taboo foods"). The average score was 4.27 for all the respondents, 4.41 for the Indonesian respondents, and 4.21 for the Indian respondents, respectively.

3.4. Examination of Peer Effects

The estimation results of the six models are reported in

Table 6. Estimation results.

A_A Model					A_M Model
	Estimate	t-value	p-value		Estimate	t-value	p-value
Intercept (Indonesia)	4.43	23.85	0.00	***	4.56	23.70	0.00	***
Intercept (Vijayawada)
D1 (Hyderabad)	−0.32	−2.90	0.00	***	−0.38	−3.42	0.00	***
D2 (Bangalore)	−0.38	−3.53	0.00	***	−0.46	−4.22	0.00	***
D3 (Cochin)	−0.33	−3.81	0.00	***	−0.39	−4.30	0.00	***
D4 (Chennai)	−0.10	−0.90	0.37		−0.17	−1.53	0.13
D5 (Vijayawada)	−0.05	−0.44	0.66		−0.12	−1.08	0.28
GEN (Q2)	0.05	1.31	0.19		0.05	1.41	0.16
AGE (Q3)	0.02	0.66	0.51		0.02	0.87	0.39
MAR (Q4)	0.09	1.18	0.24		0.11	1.53	0.13
CHR (Q5)	−0.85	−4.94	0.00	***
HID (Q5)	−0.28	−1.59	0.11
EDU (Q6)	0.08	2.60	0.01	**	0.07	2.54	0.01	**
DIF (Q9)	−0.10	−4.14	0.00	***	−0.10	−4.39	0.00	***
SAME (Q10)	−0.01	−0.45	0.65		−0.04	−1.54	0.12
n	395				369
Breusch–Pagan test	126.92				42.69
(p-value)	0.00				0.00
	GLS				GLS
AIC	457.81				366.38
RSE	23.90				1.14
R2
Adj.R2
F-statistic
(p-value)
Indonesia_A Model					Indonesia_M Model
	Estimate	t-value	p-value		Estimate	t-value	p-value
Intercept (Indonesia)	4.06	10.65	0.00	***	3.96	10.66	0.00	***
Intercept (Vijayawada)
D1 (Hyderabad)
D2 (Bangalore)
D3 (Cochin)
D4 (Chennai)
D5 (Vijayawada)
GEN (Q2)	0.01	0.13	0.90		0.01	0.09	0.93
AGE (Q3)	−0.05	−0.88	0.38		−0.02	−0.37	0.71
MAR (Q4)	−0.04	−0.26	0.80		0.06	0.46	0.64
CHR (Q5)	−0.68	−2.65	0.01	**
HID (Q5)
EDU (Q6)	0.19	2.76	0.01	**	0.22	3.37	0.00	***
DIF (Q9)	−0.13	−2.68	0.01	**	−0.13	−2.99	0.00	***
SAME (Q10)	0.05	1.17	0.25		0.02	0.55	0.58
n	114				99
Breusch–Pagan test	36.66				4.11
(p-value)	0.00				0.66
	GLS				OLS
AIC	195.13				−133.30
RSE	41.81				0.42
R2					0.32
Adj.R2					0.28
F-statistic					7.44
(p-value)					0.00
India_A Model					India_M Model
	Estimate	t-value	p-value		Estimate	t-value	p-value
Intercept (Indonesia)
Intercept (Vijayawada)	4.51	34.80	0.00	***	4.52	34.86	0.00	***
D1 (Hyderabad)	−0.13	−2.65	0.01	**	−0.12	−2.56	0.01	**
D2 (Bangalore)	−0.42	−6.55	0.00	***	−0.42	−6.42	0.00	***
D3 (Cochin)	−0.21	−4.22	0.00	***	−0.21	−4.20	0.00	***
D4 (Chennai)	−0.06	−1.52	0.13		−0.07	−1.57	0.12
D5 (Vijayawada)
GEN (Q2)	0.06	1.95	0.05	*	0.07	2.07	0.04	**
AGE (Q3)	0.06	2.92	0.00	***	0.06	2.93	0.00	***
MAR (Q4)	0.08	1.04	0.30		0.08	1.03	0.30
CHR (Q5)	−0.46	−1.15	0.25
HID (Q5)	−0.32	−1.72	0.09	*
EDU (Q6)	−0.04	−1.66	0.10	*	−0.04	−1.69	0.09	*
DIF (Q9)	−0.02	−0.79	0.43		−0.02	−0.93	0.35
SAME (Q10)	−0.08	−3.08	0.00	***	−0.09	−3.16	0.00	***
n	281				269
Breusch–Pagan test	57.11				57.27
(p-value)	0.00				0.00
	GLS				GLS
AIC	178.76				158.32
RSE	84,267.15				133,352.70
R2
Adj.R2
F-statistic
(p-value)

***, **, and * indicate statistical significance at the 1%, 5%, and 10% levels, respectively.

. Those who did not answer Q1–Q21, except for Q7 and Q8, were removed. The sample sizes were 395 for the A_A model and less for the other models. Based on the results of the BP test, OLS was applied for one case (Indonesia_M model), whereas GLS was applied for the other five cases. The value of VIF was less than 6 for all cases, suggesting that multicollinearity was not a concern.

The parameter value of the dummy variables (D1 to D5) was negative for the A_A and A_M models, implying that Indonesian respondents had a higher score ($y_i$). This result was consistent with the result that the average score of the Indonesian respondents (4.41) was higher than that of the Indian respondents (4.21). The variables GEN (Q2) and AGE (Q3) were not statistically significant at the 10% level, except for the India_A and India_M models. The MAR (Q4) variable was not statistically significant at the 10% level for all six models. The parameter values of CHR (Q5) and HID (Q5) in the A_A, Indonesia_A, and India_A models were negative, implying that Christians and Hindus had lower scores than Muslims. The variable EDU (Q6) parameter values were positive for Indonesian cases (Indonesia_A and Indonesia_M models), meaning that those with higher education had a higher score, and the sign condition was satisfied. However, the variable EDU (Q6) parameter values were negative for Indian cases (India_A and India_M models). The variable DIF (Q9) parameter values were negative for all six models, implying that the score decreases as “the percentage of respondents’ close friends who believe in different religion” increased. However, the variable DIF (Q9) was not statistically significant at the 10% level for the Indian cases (India_A and India_M models). The sign condition was satisfied for the rest of the models (A_A, A_M, Indonesia_A, and Indonesia_M models). The parameter values of the variable SAME (Q10) were positive for the Indonesia_A and Indonesia_M models but were not statistically significant. The parameter values for the variable SAME (Q10) were statistically significantly negative for Indian cases (India_A and India_M models) at the 1% level. The sign condition was satisfied for Indonesian cases (Indonesia_A and Indonesia_A models) but not for Indian cases (India_A and India_M models).

3.5. Living Together with Family Members of a Different Religion (Q8)

As expected, the percentage of respondents who lived with family members (or friends) whose religion was different was small (

Table 7. Living together with family members of a different religion (Q8).

Respondents	Live Together with	Indonesia	India
Muslim	Christian	2	2
	Hindu	1	13
Christian	Buddhist	2	-
	Hindu	1	1
	Muslim	8	6
Hindu	Christian	-	2
	Muslim	-	2
Total		14	26
Number of respondents		11	26

Some respondents live with two or more members of different religions (e.g., Muslim respondents live together with Christians and Hindus).

). There were 11 (8.4%) and 26 (5.8%) respondents in Indonesia and India, respectively. Because some of the Indonesian respondents live with two or more members whose religion is different, the total number in Table 7 is 14. Although the sample size is small, it can be inferred that religious minorities tend to live with members of different religions. For example, in Indonesia, Christians are a minority, but the number of cases was 11, while the number of Muslims, the majority, was 3. In India, Muslims and Christians are religious minorities, but the number of cases was 22, while there were 4 instances of Hindus. Because the percentage of Muslim respondents was the highest in Indonesia and India, it is more accurate to use percentages for this investigation. The proportion of people living with members of a different religion was 3.0% (3/102) and 4.1% (4/97) for the religious majority in Indonesian Muslims and Indian Hindus, and 41.0% (11/27) and 6.2% (22/353) for religious minorities in Indonesia and India, respectively.

4. Discussion

4.1. Symmetric Results for Indonesian and Indian Models

Many existing studies consider religiosity or religious faithfulness with other factors such as self-identity, knowledge, education, and certification logos in the analyses of purchase behavior (Rahman et al. 2020), business ethics (Kum-Lung and Teck-Chai 2010), job satisfaction (Wening and Choerudin 2015; Amaliah et al. 2015), and economic achievement (Yusof et al. 2018). As such, religious faithfulness is an essential component in the relevant analyses, but the number of studies that examine peer effects on religiosity is limited. Some existing studies discuss peer effects on adolescents’ religiosity in Indonesia (French et al. 2011) and other countries such as the United States (Schwartz et al. 2005; King and Roesner 2009). Others have revealed that both parents and friends influence adolescents’ attendance of religious services (Regnerus et al. 2004). To the best of our knowledge, this study is the first to incorporate peer effects in examining religious faithfulness while considering different countries (Indonesia and India) and religions (e.g., Islam and Hinduism).

Peer effects are essential components by which to explain religious faithfulness based on both existing studies (e.g., Yasid et al. 2016) and the results of the current study. In what follows, some relevant studies are briefly reviewed before discussing the main results of this study. Lon and Widyawati (2019) provide an interesting case from Flores, Indonesia, which investigates the relationship between Catholics and Muslim families. In the communal dining of Catholic and Muslim families, Catholic families offer a pork-free menu to Muslim families. In contrast, Muslim families usually provide a menu with pork for Catholic families. This tradition implies that mutual understanding among different religions is deep, and there are no serious tense relationships. Because the six religions (i.e., Islam, Christianity, Hinduism, and three more) are officially acknowledged, there is freedom of faith in Indonesia (Bräuchler 2014). Note that there are few exceptions, such as conflicts among Christians and Muslims in the Moluccas (Qurtuby 2019). On the other hand, it has been highlighted that there is tension between Hindus, the religious majority, and people of other religions such as Muslims in India (Sen and Wagner 2005; Tausch et al. 2009; Mitra and Ray 2014).

The DIF (Q9) and SAME (Q10) variables are symmetric for the Indonesian and Indian cases (

Table 8. Summary of the main results.

	Indonesia_A and Indonesia_M Models	India_A and India_M Models
DIF (Q9)	Significantly negative	Not significant
	Interpretation: The respondents’ religious faithfulness (i.e., explained variable $y_i$) takes a lower value if the percentage of respondents’ close friends who believe in a different religion (Q9) increases in Indonesia
	Impacts of the religious minority The percentage of people who believe in a different religion in the respondents’ city is low (Figure 2), and the Indonesian respondents are affected only by DIF (Q9)
SAME (Q10)	Not significant	Significantly negative
		Interpretation: The respondents’ religious faithfulness (i.e., explained variable $y_i$) takes a lower value if the percentage of people who believe in the same religion in the respondents’ city (Q10) increases in India
		Impacts of the religious minority: The percentage of respondents’ close friends who believe in the same religion is low (Figure 1), and the Indian respondents are affected only by SAME (Q10)

). DIF (Q9) is significantly negative for the Indonesian cases (Indonesia_A and Indonesia_M models), while SAME (Q10) is significantly negative for the Indian cases (India_A and India_M models) (Table 6). On the other hand, DIF (Q9) is not a statistically significant variable for the Indian cases, and SAME (Q10) is not a statistically significant variable for the Indonesian cases. These results imply that the respondents’ religious faithfulness (i.e., explained variable $y_i$) takes a lower value if the percentage of respondents’ close friends who believe in a different religion (Q9) increases in Indonesia. On the other hand, respondents’ religious faithfulness (i.e., explained variable $y_i$) takes a lower value if the percentage of people who believe in the same religion in the respondents’ city (Q10) increases in India.

The surrounding situations for the Indonesian and Indian respondents were also symmetric (Table 8). As illustrated in Figure 1 and Figure 2, the percentage of respondents’ close friends who believe in a different religion (Q9) is low and high for Indonesia and India, respectively. On the other hand, the percentage of people who believe in the same faith in the respondents’ city (Q10) is high and low for Indonesia and India, respectively. Based on the above results, it may be inferred that relatively smaller groups have larger peer effects on respondents’ religious faithfulness. In the Indonesian case, the percentage of people who believe in different religions in the respondents’ city is low, and the Indonesian respondents are affected by DIF (Q9). Similarly, in the Indian case, the percentage of respondents’ close friends who believe in the same religion is low, and the Indian respondents are affected by SAME (Q10).

These results seem to be consistent with the aforementioned studies. There are tensions among people of different religions in India. Thus, it is natural that people would be more self-restrictive when surrounded by people of other religions and vice versa, resulting in statistically significant negative results for SAME (Q10) in the Indian models. There are no such tensions among people of different religions in Jabodetabek. Thus, it is natural that people would be affected by those who believe in different religions and vice versa, resulting in statistically significant and negative results for DIF (Q9) in the Indonesian models. However, limited studies have been conducted on this topic, and future studies are required to confirm the possible influence of religious tensions.

4.2. Influence of Family Members Whose Religion Is Different (Q8)

The number of Indonesian and Indian respondents who indicated that they lived with family members whose religion was different from theirs was 11 and 26, respectively. The percentage of people living with members of a different religion was 3.0% and 4.1% for the religious majority in Indonesian Muslims and Indian Hindus, and 41.0% and 6.2% for religious minorities in Indonesia and India, respectively. There seems to be a tendency for religious minorities to live with members of different religions.

4.3. Validity of Self-Evaluation Results

The results of Q11 to Q21 were averaged, and the average score was used as the explained variables in the models. The average scores of all the respondents, the Indonesian respondents, and the Indian respondents were 4.27, 4.41, and 4.21, respectively. The Indonesian value was 0.2 points higher than that of India, which is statistically significant at the 1% level (two-sided t-test, p = 0.0027, d.f. = 140). The results seem to be valid because religion permeates the daily lives of Indonesian people (French et al. 2008), resulting in high self-evaluation.

5. Conclusions

This study investigated peer effects on religious faithfulness in Indonesia and India using regression analyses with (1) the percentage of respondents’ close friends who follow a different religion and (2) the percentage of people in the respondents’ city who follow the same faith as the main variables. Since religious restrictions on foods are an easily understandable issue for subjects, this study employed 11 questions on food restrictions (Q11–21), and the average scores of these 11 questions were used as the explained variables. The score of the Indonesian sample was higher for most questions, especially Q12, while the score of the Indian sample for Q20 and Q21 was higher than that of the Indonesian sample. Since both Q20 and Q21 are related to exposure, a higher score for these questions among the Indian sample indicates that Muslims in India need to gather information as they constitute the religious minority; additionally, Indonesian Muslims concentrate on religious faithfulness, resulting in a high score for Q12.

In the regression analyses, two variables were used in the examinations: (1) the percentage of respondents’ close friends whose religion is different (DIF (Q9)), and (2) the percentage of people in the city in which the respondent lives who believe in the same faith (SAME (Q10)). The results indicate that the Indonesian respondents are affected more by those who believe in the same religion, while the Indian respondents are affected more by those who believe in a different religion, suggesting that relatively smaller groups have larger peer effects on religious faithfulness. These symmetric results may be attributed to the fact that tensions among people from different religions are high and low in India and Indonesia, respectively, and that the percentage of people who believe in a different faith in the respondents’ city is high and low in India and Indonesia, respectively.

There seems to be a tendency for religious minorities to accept family members whose religion is different. Based on the discussions in Section 4.1 and Section 4.2, it can be concluded that relatively smaller groups such as religious minorities are more important when considering peer effects on religious faithfulness. Although policies in Type B countries may stress the benefits of the religious majority, this study revealed the important roles of religious minorities, and policy makers need to consider the influence of such groups.

Finally, this study has the following limitations. First, only one city from Indonesia was selected. Because Indonesia is varied in both religions and ethnicities, it may be better to gather data from the whole country. However, our samples are still valid because (1) the samples were collected from citizens in Jabodetabek, which is a substantially heterogeneous society (Ariane 2020), and (2) the primary purpose of this study was to check the peer effects of different situations, which may not necessarily require an entire Indonesian sample. Moreover, Indonesia is composed of more than 13,000 islands with varied languages; thus, it is realistic to gather samples from a large region such as Jabodetabek that attracts immigrants from other parts of the country. Second, while Hindus constitute the religious majority in India, most of the Indian sample in this study were Muslims. As pointed out in the Introduction section, different faiths could also have been compared, but people from different faiths may follow different traditions, habits, and ways of thinking, which makes comparisons difficult. Comparisons of different faiths should be addressed in future studies. Third, this study examined only Indonesia and India. These countries were selected because religion is an essential part of the respective populations’ lives, and the religions of Indonesia, especially Islam, are strongly tied to citizens’ daily lives (Snibbe and Markus 2002; Cohen et al. 2005; French et al. 2011); in addition, we selected India because there are tensions among people of different religions in India.