Determinants of early marriage in Iran: a multivariate survivorship analysis using census microdata

ABSTRACT

Recent reductions in female age at marriage in Iran correspond with the rise in the proportion of girls marrying at young ages. This paper aims to determine to what extent the probability of surviving in the never-married state by age 18 is affected by different time-invariant demographic, socio-economic, and cultural factors. A multivariate survivorship analysis was applied to the 2% micro-data from two successive censuses. This method is useful where longitudinal family surveys are not available. The findings indicate that (1) gender and ethnicity remain two important determinants of early marriage after adjusting for other covariates and (2) having literate parents substantially increases the relative probability of survival, with the literacy of the father displaying a larger influence regardless of gender and place of residence. These indicate the relevance of gender system, ethnicity, and socioeconomic standing in designing policies and interventions aimed at preventing early marriage in Iran.

KEYWORDS:

• Early marriage
• survival analysis
• gender
• education
• ethnicity
• Iran

Introduction

Worldwide, one in every five young women (aged 20 to 24) marry by age 18 and Asia is home to more than 60% of child brides, comprising more than 400 million women (UNICEF, 2018). The practice widely varies across Asian countries; ranging from 4% in Maldives, Qatar, and Oman to 59% in Bangladesh (Torabi, 2023). According to the 2010 Multiple Indicator Cluster Survey (MICS), 17% of young women in Iran marry before they reach their 18^th birthday (UNICEF global databases, 2020). The country has been characterized by early and universal marriage (Momeni, 1972) within a segregated gender system (Tashakkori & Thompson, 1988). As the country became more urbanized and more population gained access to education, marriage was postponed to higher ages although remained universal (see Torabi et al., 2013). Legal interventions have not reinforced the trend towards later marriage. In compliance with Sharia laws, the female legal age at marriage was dropped to 9 (from 15) after the Islamic Revolution of 1979 but raised to 13 in 2003, leaving marriage at lower ages to the approval of the girl’s guardian and a judge or a medical doctor (Tremayne, 2006).

Changes in the timing of marriage in Iran are shown in Figure 1. First consider the female singulate mean age at marriage, which increased between the mid-1980s and the mid-2000s (from 19.7 to 23.3 years), before a reversal in the following decade particularly in rural areas. The male singulate mean age at marriage rose by 4 years (23.2 to 27.4) during the same period, widening the age difference between wives and husbands. Rural-urban differences in the timing of marriage remained relatively constant for men. For women, however, a notable gap emerged since 2011, resulting in a difference of 1.4 years between the age at marriage in rural and urban areas in 2016. The recent fall in age at marriage in rural areas, which constitute 26% of the population of Iran (Statistical Centre of Iran, 2016), has corresponded with reductions in the proportion of never-married young women, including those aged 15 to 19. As illustrated in Figure 2, 78.1% of rural women aged 15–19 were never married in 2006 compared to 71.6% and 67.3% of their counterparts in 2011 and 2016, respectively.

Figure 1. Singulate mean age at marriage (SMAM) by gender and place of residence, Iran, 1986–2016.Source: Calculated from census published results (Statistical Centre of Iran, 1986, 1996, 2006, 2011, and 2016)

Figure 2. Percentage of never married women in rural areas by age group, Iran, 2006–2016.Source: Calculated from census published results (Statistical Centre of Iran, 2006, 2011, and 2016)

The reduction in female mean age of marriage in Iran is a new phenomenon and has not attracted much research interest at the national level. A large body of research, however, exists regarding continued marriage postponement. Using a systematic review method, Torabi and Shams-Ghahfarokhi (2022) found various determinants of marriage in Iran, including economic (e.g. unemployment, low income, and difficulties in providing housing), social (e.g. education), place of residence (e.g. living in urban areas), demographic (e.g. marriage squeeze), lifestyle (e.g. luxurious marriage, high expectation from marriage, and materialistic attitude), psychological (e.g. fear from divorce), family background (e.g. low socioeconomic status), and ethnicity. However, the recent reduction in the female age of marriage in rural areas of Iran and the determinants of early marriage need careful attention.

Early or child marriage – defined as marriage by age 18 – has various adverse consequences for women and for society and its elimination is on the global development agenda (United Nations General Assembly, 2015). According to UNICEF (2014), these consequences include social isolation, limited education and employment opportunities, early pregnancy, lower access to medical care during pregnancy, and vulnerability to sexually transmitted infections, including HIV. Examining macro-level data from 35 Asian countries, Torabi (2023) shows that early marriage is associated with poor health outcomes among women as well as children. Considering women’s reproductive behaviour and health outcomes, a higher prevalence of early marriage was found to be associated with higher fertility, lower mean age of childbearing, higher maternal mortality ratio as well as female youth deaths, higher unmet need for contraception and lower prenatal care and births attended by skilled staff. A higher prevalence of early marriage was also associated with higher neonatal, infant, and child mortality as well as a higher prevalence of anaemia, wasting, stunting, and underweight among children under age 5.

The recent trend towards a higher prevalence of early marriage in Iran has happened in a context where the majority of women have access to education even in rural areas. According to the results of the Iranian 2016 census (Statistical Centre of Iran, 2016), 84.2% of women aged 6 and above are literate (88.0% in urban and 72.8% in rural areas). These figures are substantially higher when young women are considered; 97.1% of women aged 6 to 24 are literate (97.9% in urban and 95.1% in rural areas). The country adopted a pro-natalist population policy in the early 2010s as a response to persistent postponement of marriage and low levels of fertility (see Torabi & Abbasi-Shavazi, 2015). As noted earlier, the prevalence of early marriage started to rise in 2006. Thus, its commencement cannot be attributed to the shift in population policy. However, this policy could have contributed to its persistence. The fact that the prevalence of early marriage in Iran is notably higher than in some Asian countries with similar levels of female education, urbanization, and net migration (Torabi, 2023) indicates the role of micro-level factors in determining early marriage in Iran. This multidimensionality of determinants of early marriage raises the question as to what extent this phenomenon is related to the characteristics of young individuals and their families. This question is best answered by longitudinal data at the individual and household levels, which unfortunately are not available. This paper applies a multivariate survivorship analysis proposed by Hill (1999) to the 2% micro-data from the 2011 and 2016 Iranian censuses to determine to what extent the probability of surviving in the never-married state by age 18 is affected by different time-invariant demographic, socio-economic and cultural factors. The next section draws on the literature to identify determinants of early marriage. Then, the data and method are introduced before the findings are presented, discussed, and concluded.

Determinants of early marriage

The literature identifies various determinants of early marriage, which are generally interwoven and beyond the control of children, and include gender, educational attainment, family’s economic status, religion, social norms and traditions, and place of residence (Johansson, 2015; Wodon et al., 2017).

Gender is an important determinant of the timing of marriage, in general, and early marriage, in particular. Early marriage affects girls disproportionately, although boys also marry as children (UNICEF, 2014). Several factors may explain such gender differences. For instance, child brides tend to live in rural areas and attain little or no education. The place of residence and education have been considered as two crucial determinants of early marriage (Kumchulesi et al., 2011). People living in urban areas experience later marriage than their rural counterparts partly due to more individualistic and material lifestyles and weaker social control. Furthermore, delayed marriage is not socially acceptable, infrastructures are less adequate, work opportunities are more limited, gender values are more rigid, and education is less prevalent in rural areas (UNICEF & UNFPA, 2018). A recent rise in the proportion of married teenage girls in rural areas of Iran has already been shown (See Figure 2). As noted by Lowe et al. (2019), however, determinants of early marriage may vary across different societies. Chowdhury et al. (2020) show that the probability of early marriage is higher in urban than the rural areas in Bangladesh.

Education is strongly associated with early marriage. Low levels of education are both a cause and a consequence of early marriage (Santhya et al., 2010). Most studies in developing countries emphasize the importance of expanding girls’ education in preventing early marriage and have shown that parents’ and children’s education strongly affect its occurrence (Bates et al., 2007; Hasanah et al., 2021; Hervish & Feldman-Jacobs, 2011; Ikamari, 2005; Jain et al., 2011; Rumble et al., 2018; UNICEF, 2005; Yüksel-Kaptanoğlu & Ergöçmen, 2014). Studies report a negative association between early marriage and educational attainment in different Asian countries (see, e.g. Pintu, 2019; Roest, 2016; Rumble et al., 2018) because higher education increases knowledge (Notoatmodjo, 2012), opens new windows of social and economic opportunities (Caldwell et al., 1983), increases the young person’s autonomy over marriage decisions (Jejeebhoy, 1995) and enhances logical thinking (Hasanah et al., 2021). Higher education has been found to reduce the risk of the marriage of girls in Iran too (Torabi et al., 2013).

Family background plays an important role in early marriage, especially in societies where parents’ approval is still important in marriage formation. Bagi (2022) shows that although the role of parents in the marriage decisions of their children has diminished in Iran, 27% of marriages in the 2010s were still decided by parents. Social values and norms regarding gender roles and the maintenance of girls’ safety and chastity have also been identified as drivers of early marriage (Parsons et al., 2015; Roest, 2016). Bicchieri et al. (2014) mentioned 12 factors related to early marriage, including family pressure. In societies where chastity is still important, the risk of losing virginity outside marriage increases as girls grow older and become more exposed to love affairs or sexual harassment. Social conformity may be another reason for family pressure to marry off their young girls (Chowdhury et al., 2020; Lowe et al., 2019).

Poverty has been recognized as another important driver of early marriage (Nasrullah et al., 2014; Rumble et al., 2018). The practice is 2.4 higher in the least compared to the most affluent households in Indonesia (BPS, 2020) and the median age at first marriage is 4.3 years lower for women in the poorest than those in the wealthiest quintiles in India (UNICEF, 2014). Versluys (2019) shows that although education and poverty influence early marriage, deep-rooted gender norms have a larger impact on the likelihood of girls being forced into early marriage. Child marriage is deeply rooted in gender inequality, especially in patriarchal societies (Kandiyoti, 1988).

Early marriage is also influenced by cultural factors, including those associated with ethnicity (Jain et al., 2011; UNICEF & UNFPA, 2018; Wijayati, 2017). These differences may partly be responsible for inter- and intra-country variations in the practice of early marriage (Safavi & Minaei, 2015). Laws prohibit early marriage in most countries, but their very existence indicates the impact of other factors. Yüksel-Kaptanoğlu and Ergöçmen (2014) believe social norms, local customs, and traditions that justify early marriage may be more powerful than national laws. For instance, the concept of ‘honor’ inclines families to favour early marriage for girls in Turkey before they experience sexual relations and bring dishonour to their families. This concept seems to be powerful in all societies where sexual relations outside of marriage are prohibited and considered scandalous.

Various studies have confirmed ethnic differences in marriage and its patterns (Arnaldo, 2004; Aryal, 2007; Landale & Tolnay, 1993; Lesthaeghe et al., 1989; Mpilambo et al., 2017; Raley et al., 2015). Hasanah et al. (2021) consider ethnicity as an important determinant of early marriage because customary and cultural factors still play an important role in the process of spouse selection. Lowe et al. (2019) show the role of ethnicity in causing anxiety and fear of premarital sex, leading to a high rate of early marriage in Gambia. Ethnic differences in early marriage are also evident in Turkey, where Kurds display the highest rate of early marriage due to their strong cultural traditions (Yüksel-Kaptanoğlu & Ergöçmen, 2014). The timing of marriage of women in Iran is influenced by ethnicity after adjusting for several demographic and socioeconomic characteristics at individual and contextual levels, indicating the important role of cultural factors associated with ethnic groups (Torabi & Baschieri, 2010; Torabi et al., 2013). Accordingly, we expect early marriage in Iran to be associated with ethnicity too.

The next section introduces the data and method applied to ascertain to what extent early marriage in Iran is determined by these demographic, socio-economic, and cultural factors.

Data and method of analysis

To identify factors affecting the survivorship of boys and girls aged less than 18 in the never-married state, a pooled two-sample data set was assembled from the 2% census micro-data for cohorts aged 10–12 at the 2011 census and 15–17 five years later, at the 2016 census. The reason for including this age range is that the question about marital status has been asked from the population aged 10 + .The data set includes 126,788 never-married individuals (67069 individuals or 52.9% and 59,719 individuals or 47.1% from the 2011 and 2016 censuses, respectively). According to Hill (1999) who proposed the multivariate model of individual risk, the cohort is assumed to be closed to migration and all forms of pseudo-migration caused by misclassifications. Thus, all individuals who migrated to Iran or between rural and urban areas during the interval between the two censuses (6300 persons) have been excluded from the analysis. In addition, in households where more than one eligible person existed, only one person was selected randomly, excluding 7610 individuals from the data set. The final data set includes 112,574 never-married individuals (58247 individuals or 51.7% and 54,300 individuals or 48.2% from the 2011 and 2016 censuses, respectively). Both samples are stratified simple random samples of the enumerated population of Iran. The sampling weights have been included in the data set and applied to the analysis.

Before turning to the multivariate analysis, the intersample and the group-specific survival probabilities are estimated. The intersample survival probability for the closed cohort is estimated by

(1) $\hat{P}=\frac{g.n_2}{\mathit{f}.n_1}$(1)

where $n_1$ and $n_2$ represent the size of cohorts at times$t_1$ (or 2011) and $t_2$ (or 2016), respectively. Let $\mathit{f}$ and $\mathit{g}$ represent the sample correction factors that inflate the sample to full population sizes.

To compare the survival probabilities of subgroups within a closed cohort, the group-specific probability is estimated as follows:

(2) $\hat{P}{}^x=\frac{g{n}_2^x}{\mathit{f}{n}_1^x}$(2)

and

(3) $\hat{P}{}^{\bar{x}}=\frac{g{n}_2^{\bar{x}}}{\mathit{f}{n}_1^{\bar{x}}}$(3)

where $\mathit{x}$ and $\bar{x}$ indicate the presence and absence of a time-invariant characteristic. These variables include gender, urban or rural place of residence, ethnicity indicated by the predominant ethnic group in the province of residence as estimated by Torabi and Baschieri (2010), and the literacy of parents as an indicator of the socio-economic standing of the family. The educational attainment of parents has not been included because the preliminary results showed that it is not a time-invariant characteristic. The variables of the age of parents and their age difference have also been included in the analysis but measured at an interval level. Older parents and those with a wider age difference are expected to share more positive attitudes regarding early marriage. Thus, their children are considered to be more likely to marry at an early age. The ratio of two survival probabilities (Equations 2 and 3) yields $\hat{R}$, the relative probability of survival associated with the presence of $\mathit{x}$ or the relative risk:

(4) $\hat{R}=\frac{{\hat{P}}^{\mathit{x}}}{{\hat{P}}^{\bar{x}}}=\frac{g{n}_2^x/\mathit{f}{n}_1^x}{\mathit{g}{n}_2^{\bar{x}}/\mathit{f}{n}_1^{\bar{x}}}=\frac{n_2^x{n}_1^{\bar{x}}}{n_2^{\bar{x}}{n}_1^x}$(4)

Thus, if the sampling fraction is the same for subgroups $\mathit{x}$ and $\bar{x}$ at the time t, the estimation of relative survival probabilities for a closed cohort is possible even when sample fractions are unknown.

Using the model in Equation (1)(1) $\hat{P}=\frac{g.n_2}{\mathit{f}.n_1}$(1) , the effects of a vector of covariates on survival during the intersample period are explored. ${\pi }_i$ represent the probability of surviving from $t_1$ to $t_2$ for individual $\mathit{i}$ ($\mathit{i}=1,2,\dots ,\mathit{N})$. This probability depends on the abovementioned time-invariant covariates $(X_i$) through a log-probability model:

(5) $\mathit{log}{\pi }_i=\mathit{\alpha }+\mathit{\beta }{ }^{\mathrm{\prime }}{X}_i$(5)

where $\mathit{\alpha }$ represents the log of baseline probability and ${\beta }_k$ represents the log of the relative risk associated with the covariate $\mathit{k}$, adjusted for other covariates.

Using the pooled data, the parameters of the log probability model in Equation (1)(1) $\hat{P}=\frac{g.n_2}{\mathit{f}.n_1}$(1) can be estimated by fitting a logit regression model to the dummy variable $y_i$ equal to 1 if the observation $\mathit{i}$ is in $t_2$ sample (or 2016 sample) and 0 otherwise:

(6) $\mathit{log}\frac{{\pi }_i}{1-{\pi }_i}={\alpha }^{\ast }+\mathit{\beta }{ }^{\mathrm{\prime }}{X}_i$(6)

where ${\alpha }^{\ast }=\mathit{\alpha }+\log \left(\frac{f}{g}\right)$. This model allows us to estimate relative survival differences but not the absolute survival probabilities because the estimation of $\mathit{\alpha }$ requires the sampling fractions. In this paper, $\mathit{\alpha }$ cannot be estimated because the sampling fractions for individuals are not clear.

Findings

This section starts with describing the variables included in the analysis (Table 1). Males constitute just over half of the sample. Seven out of every ten individuals included in the analysis live in urban areas. The majority of parents are literate but literacy is more widespread among fathers than mothers. Fathers are on average 5.7 years older than mothers. The highest share of the sample belongs to the regions predominated by Persians (more than one-third), followed by those predominated by Persians and Turks, Turks, and a mixture of ethnic groups (13–18%). Other groups constitute less than 5% of the sample.

Table 1. Descriptive statistics of the variables included in the analysis.

Variable		% or Mean (SD)	n	Variable		% or Mean (SD)	n
Gender	Male	52.2	58782	Ethnicity	Persian	36.2	40743
	Female	47.8	53765		Persian and Turk	18.0	20181
Place of residence	Rural	28.6	80348		Gilak	3.2	3554
	Urban	71.4	32199		Mazandarani	3.9	4408
Father’s literacy	Illiterate	13.6	14019		Turk	13.6	15336
	Literate	86.4	89132		Kurd	4.2	4761
Mother’s literacy	Illiterate	19.5	21165		Lur	3.2	3650
	Literate	80.5	87412		Baluch	4.5	5071
Father’s age (year)		44.7 (7.6)	112547		Mixed	13.2	14843
Mother’s age (year)		39.9 (6.9)	112547
Parent’s age difference (year)		5.7 (4.7)	112547

SD stands for standard deviation; Among different variables, only the sum of the father’s and mother’s literacy is not equal to 112,547 due to missing values.

Table 2 presents the survival probabilities of individuals with different gender, place of residence, ethnicity, and parents’ literacy as well as the relative risks associated with the presence of a particular characteristic (i.e. being female, living in urban areas, living in a non-Persian- predominated region, having a literate father and having a literate mother). Girls are less likely to stay never-married ($\hat{P}=0.879)$ than boys ($\hat{P}=0.983$), whose chances of getting married at an early age are close to zero. Calculating the corresponding figures by place of residence, the survival probability of urban boys is one but that of rural boys is lower ($\hat{P}=0.891)$. For girls, the survival probability of those living in urban areas is equal to 0.956 which is much higher than that of those living in rural areas ($\hat{P}=0.705)$. The value of $\hat{R}$ indicates that girls are 11% less likely to remain never married than boys between 2011 and 2016.

Table 2. Intersample group-specific survival probabilities and relative risks, Iran, 2011–2016.

Variable		n	Survival probability ($\hat{\mathit{P}})$	Relative risk ($\hat{\mathit{R}})$	t-test/F-Anova
Gender	Male (reference)	58782	0.983		9.414 **
	Female	53765	0.879	0.894
Place of residence	Rural (reference)	80348	0.799		16.241 **
	Urban	32199	0.991	1.239
Ethnicity	Persian (reference)	40743	0.932		F= 5.795 **
	Persian and Turk	20181	0.990	1.062
	Gilak	3554	0.980	1.051
	Mazandarani	4408	0.993	1.065
	Turk	15336	0.892	0.957
	Kurd	4761	0.895	0.960
	Lur	3650	0.929	0.996
	Baluch	5071	0.901	0.966
	Mixed	14843	0.891	0.956
Father’s literacy	Illiterate (reference)	14019	0.811		12.095 **
	Literate	89132	0.949	1.169
Mother’s literacy	Illiterate (reference)	21165	0.807		8.656 **
	Literate	87412	0.971	1.202

Relative risks have been calculated as the ratio of the survival probability of each category of the variable of interest to the survival probability of the reference category. **: significant at 0.001.

Considering the place of residence, residents of rural areas appear more likely to marry at an early age, as we have already noticed. The highest survival probability belongs to the young people living in regions predominated by the Mazandarani ethnic group, followed by the region predominated by the Persian and Turk and Gilak groups, where $\hat{P}$ is around 0.99. The lowest figures, on the other hand, belong to those living in the Kurd- and Turk-predominated regions as well as the region occupied by a mixture of ethnic groups, where $\hat{P}$ is less than 0.89. Finally, those whose parents are literate are less likely to marry at an early age. The value of $\hat{R}$ indicates that young people whose father is literate are 17% more likely to remain never-married and the corresponding figure for the variable of mother’s literacy is 20%. It is important to note that applying appropriate tests indicates statistically significant differences in the survival probability of all variables.

The results of multivariate survivorship models are presented in Table 3. Model 1 includes a full model for both males and females. First, consider the variable of gender. After adjusting for other variables, girls are 11% less likely than boys to remain never married. Place of residence does not influence early marriage, after accounting for other variables but living in all regions predominated by non-Persians makes survival more likely. The largest gap belongs to the Baluch group, which is 22% more likely than Persians to remain never married. Thus, lower survival probabilities of the Baluch as well as Turk, Kurd, and Lur groups, which were observed earlier, appear to be related to their differences with Persians in the variables included in the analysis. It is interesting to see that controlling for these differences results in comparable relative risks among Mazandarani and Lur groups, whose timing of marriage considerably differs (see Torabi & Baschieri, 2010).

Table 3. Coefficients and risk ratios indicating the relative probability of survival in the never-married state, Iran, 2011–2016.

		Males and females (1)	Female in urban areas (2)	Females in rural areas (3)	Males in rural areas (4)
Variable		$exp\left(\mathit{\beta }\right)$	$exp\left(\mathit{\beta }\right)$	$exp\left(\mathit{\beta }\right)$	$exp\left(\mathit{\beta }\right)$
Gender	Male (reference)
	Female	0.887**	-
Place of residence	Rural (reference)
	Urban	0.998NS	-
Ethnicity	Persian (reference)
	Persian and Turk	1.016**	0.996NS	1.454**	1.123**
	Gilak	1.025**	0.947**	1.222**	1.143**
	Mazandarani	1.100**	1.134**	1.265**	1.075**
	Turk	1.078**	1.064**	1.128**	1.156**
	Kurd	1.041**	1.053**	1.126**	1.008NS
	Lur	1.095**	1.027*	1.312**	1.293**
	Baluch	1.220**	0.971**	1.514**	1.426**
	Mixed	1.028**	0.994NS	1.204**	1.055**
Father’s literacy	Illiterate (reference)
	Literate	1.816**	1.756**	2.153**	1.760**
Mother’s literacy	Illiterate (reference)
	Literate	1.374**	1.382**	1.382**	1.401**
Father’s age		1.096**	1.099**	1.091**	1.079**
Mother’s age		1.066**	1.069**	1.065**	1.066**
Parent’s age difference		0.941**	0.954**	0.934**	0.943**
Constant		0.001**	0.001**	0.001**	0.002**

NS: Not significant; *: p value<0.05; ** p value<0.01.

Having literate parents increases the relative probability of survival; young people with literate fathers and mothers are 81% and 37% more likely than those with illiterate fathers and mothers to remain never married, respectively. Having older parents also increases the relative probability of survival. This finding was not anticipated because older parents are expected to share more positive values regarding early marriage. Because the income and assets of the household have not been controlled, this finding may reflect the economic advantage of older parents over younger ones, which prevents them from marrying off their children (and particularly their girls) due to financial hardship. However, a larger age difference between parents appears to reduce the relative probability of survival. This finding is expected from the parent’s own marriage experience, which can be considered less modern and more in line with the practice of early marriage.

Models 2 and 3 include females who live in urban and rural areas, respectively. Similar to what was observed for the whole sample, the literacy of parents increases the relative probability of survival but the magnitude of the influence of the father’s literacy is larger for rural girls. Again, having older parents increases the relative probability of survival but a larger age difference between parents reduces it. The magnitude of the influence is comparable for urban and rural girls.

The major difference between rural and urban areas is the direction and magnitude of the coefficients of ethnicity. The direction of the coefficients remains unchanged in rural areas, with living in all non-Persian regions increasing the relative probability of survival after adjusting for other variables. Baluch girls display the largest gap with Persian girls. In urban areas, however, adjusting for other variables only makes the relative probability of survival in the Mazandarani-, Turk-, Kur- and Lur-predominated regions higher than the Persian-predominated region. Interestingly, both the Gilak and Baluch girls with the farthest timings of marriage (Torabi & Baschieri, 2010) would display similar relative probabilities of survival if they had similar characteristics.

Model 4 only includes males residing in rural areas because as previously noted, the probability of male early marriage is zero in urban areas. Similar to what we have observed for rural girls, living in all regions predominated by non-Persians makes survival more likely but here, the largest gap belongs to the Baluch group, which is 43% more likely than Persians to remain never married and the Lur group comes next. Similar to the previous analyses, the literacy of parents increases the relative probability of survival and the effect of the literacy of fathers is greater than the literacy of mothers. As we have already observed in previous analyses, having older parents increases the relative probability of survival but a larger age difference between parents reduces it.

Summary and conclusion

The recent reduction in female age at marriage in Iran is partly related to the rise in the proportion of girls marrying at young ages. This paper applied a multivariate survivorship analysis proposed by Hill (1999) to the 2% micro-data from the 2011 and 2016 Iranian censuses to determine to what extent the probability of surviving in the never-married state by age 18 is affected by different time-invariant demographic, socio-economic and cultural factors.

The results indicate that girls experience early marriage in both rural and urban areas, although the experience is more common in rural areas. For boys, the experience is non-existence in urban areas. Gender remains an important determinant of surviving in the never-married state after adjusting for other covariates. The fact that girls would be more likely to experience early marriage even if they had similar characteristics to boys, suggests the existence of a gender system in Iran that defines the values and norms associated with family patterns and is reflected in other behaviours such as the division of roles between spouses in Iran (see Torabi, 2020; Torabi, 2021).

The place of residence also remains an important determinant of surviving in the never-married state after adjusting for other covariates, with rural girls being more likely to experience early marriage. This is consistent with previous research (UNICEF, 2014) and may result from more inadequate infrastructures and limited educational and paid work opportunities in these areas (UNICEF & UNFPA, 2018). Literate parents are not expected to be as motivated as illiterate ones to marry off their children at an early age because they are more likely to enjoy a higher economic standing and invest in the education and skills of their children. Having literate parents was found to increase the relative probability of survival, with the literacy of the father displaying a larger influence regardless of either the gender or the place of residence. This argument is reinforced by the finding that the older the parents are, the less likely their children will be to experience early marriage. Considering that older parents are expected to share more positive attitudes about early marriage, this result shows that their better economic standing probably outweighs their potentially encouraging attitude. Previous research has found prosperous households are less likely to marry off their children at an early age (BPS, 2020; Nasrullah et al., 2014). Economic empowerment of less affluent and vulnerable households can prevent early marriage in Iran too.

The age difference between parents was another indicator of the tendency of parents towards early marriage because parents with larger age differences can be expected to be more in favour of this practice and perhaps have a similar experience. The results appeared in the expected direction; the larger this difference is, the more likely the children will be to marry at an early age. Ethnic groups can also convey specific values and norms regarding gender roles and the timing of marriage. Consistent with previous research on the timing of marriage in Iran (Torabi & Baschieri, 2010; Torabi et al., 2013) ethnicity remained a determinant of early marriage after adjusting for demographic and socio-economic covariates. Thus, the success of interventions aimed at preventing early marriage in multi-ethnic countries such as Iran depends on taking the cultural values, norms, and customs of specific ethnic groups into consideration.

This paper found early marriage to be influenced by various demographic, socio-economic, and cultural factors. Future studies can benefit from more detailed information about the determinants of early marriage at the individual (e.g. educational attainment and paid work opportunities) and household level (e.g. economic standing and number of siblings) and its social, economic, and health consequences to provide a deeper and more comprehensive understanding about early marriage in Iran.

Disclosure statement

The authors do not have any relevant financial or non-financial competing interests.

Correction Statement

This article has been corrected with minor changes. These changes do not impact the academic content of the article.