The effect of prenatal exposure to Ramadan on human capital: evidence from Turkey

Abstract

This study aims to analyze the effect of prenatal exposure to certain nutritional disruptions in utero on human capital formation. In order to resolve the endogeneity driven by maternal dietary deficiencies, we use the Islamic holy month of Ramadan, during which observers abstain from eating and drinking entirely from sunrise to sunset, as a natural experiment to assess the effects of fetal malnourishment on educational and labor market outcomes. We compare schooling and employment measures of individuals who are prenatally exposed to Ramadan with those who are not using the 2016 round of the Turkish Family Structure Survey dataset. Our results indicate that men who are exposed to Ramadan in utero in the first or third trimester complete 0.53 fewer years of schooling and are less likely to obtain middle school, high school, and college degrees. These effects are stronger when Ramadan coincides with summer.

1 Introduction

Disparities in children’s well-being and cognitive and non-cognitive competencies arise in the initial periods of life and significantly modify adulthood outcomes. Current literature points out the fact that dietary deficiencies in early developmental phases, especially in utero, have the most significant detrimental impact on health as well as schooling and employment outcomes (Almond and Currie 2011a, b; Almond et al. 2018; Comfort 2016; Greve et al. 2017; Schultz-Nielsen et al. 2016; Abiona and Ajefu 2023; Neelsen and Stratmann 2011; Chen et al. 2020; Bryce et al. 2008; de Rooij et al. 2010). Approximately 200 million children below the age of 5 will not be able to realize their whole capacity in the long run due to nutritional deficiencies (Grantham-McGregor et al. 2007). These findings and the immensity of the affected population emphasize the need to better understand the relationship between early prenatal exposure to shocks and later life outcomes and accordingly draw attention to the necessity of early investments in the prenatal period to improve human capital formation (Heckman 2007; Doyle et al. 2009; Almond and Currie 2011a).

Identifying the causal effect of exposure to in-utero shocks, however, is not straightforward. Experiencing prenatal nutritional deficiencies is not random; on the contrary, it carries unobservables such as maternal education, health, and income. In order to eliminate biased estimates and form a causal relationship between prenatal undernutrition and later life outcomes, the general practice in the literature is to utilize extreme natural experiments such as famines,¹ natural disasters, or pandemics. The results show that early intrauterine exposure to these severe events poses a greater danger of obesity, coronary heart disease, metabolic and cognitive impairments, and unfavorable labor market outcomes (Almond and Currie 2011b; Schulz 2010; Scholte et al. 2015; Roseboom et al. 2006). These events capture causality since they are orthogonal to several unobservable factors affecting an individual’s health during gestation. Nevertheless, these are unusual circumstances that mainly occurred in the early twentieth century.² Therefore, the obtained findings are not suitable to apply to milder, daily events that affect millions of children, for instance, malnutrition due to prenatal dietary deficiencies to which hundreds of millions of children, particularly from developing nations, have fallen victim (Victora et al. 2008). In addition, since extreme unfortunate events are not under the control of human behavior, it highly limits the extent of policy interventions to curb the intensity of negative outcomes (Schultz-Nielsen et al. 2016).

In this study, we adopt a life-course approach and present new causal evidence on the detrimental effects of intrauterine exposure to nutritional deficiencies on individuals’ educational attainment and labor market outcomes. An ongoing and less extreme form of nutritional shock or, from a broader perspective, an array of lifestyle changes that pregnant women face is observing Ramadan, the holiest month of Islam, during which each healthy adult Muslim should fast during daylight hours for around 30 days. We address the endogeneity problem of prenatal malnourishment by treating Ramadan observance as a natural experiment for a milder nutritional shock.³ Utilizing the 2016 wave of the “Turkish Family Structure Survey” prepared by the Turkish Statistical Institute (TurkStat), we compare the educational and labor market outcomes of individuals who are prenatally exposed to Ramadan with those of their contemporaries. We use the intention-to-treat (ITT) approach to assess the impact of in-utero Ramadan exposure on human capital formation. In addition, since we are testing multiple hypotheses simultaneously, in order to prevent Type I errors, we employ the Romano-Wolf correction method as a robustness check and report the adjusted p-values (Clarke et al. 2020).

Every year, 60 to 90% of Muslim pregnant women practice fasting, 29 million Muslim babies get exposed to Ramadan prenatally, and currently, 1.2 billion Muslim population experienced it to some extent in utero (Majid et al. 2019; Karimi and Basu 2018; Almond et al. 2018). However, despite the immensity of the affected population, the interest in the literature on Ramadan fasting is still scarce, though growing. Further, a large part of the literature concentrates on differences in the health of exposed and non-exposed individuals either during childhood or adulthood and reports worsened health among older people, higher rates of disabilities in adults, and lower birthweight and decreased growth in children (van Ewijk 2011; Almond and Mazumder 2011; Pradella and van Ewijk 2018; Karimi and Basu 2018; Karimi et al. 2021). On the other hand, the effect of intrauterine Ramadan exposure on education and labor market outcomes has not received enough attention. With the aim of filling this gap, our study provides new insight into the findings of the current Ramadan literature in various aspects.

First, to our knowledge, this is the first study to utilize the length of fasting for measuring the effects of intrauterine exposure to Ramadan on academic and labor market outcomes.⁴ Ramadan follows the lunar calendar instead of the widely used Gregorian, shifting 10 to 11 days backward each year. Thus, unlike the equatorial countries (e.g., Indonesia, Uganda), where daylight hours and temperatures are almost constant throughout the year, the changes in the timing of Ramadan mean that the duration of fasting in Turkey would range between 11 to 17.5 hours depending on the date as opposed to 12 hours on the equator. Metzger et al. (1982) observe certain biomarkers⁵ of 21 pregnant and 27 non-pregnant women in the morning after a 12-hour nightlong fast and at noon—6 hours after the first test (they assess the 18-hour fast results). The authors report that the initial glucose levels of pregnant women are already lower than those of non-pregnant women and that while the biomarkers of non-pregnant women remained stable from morning to noon, pregnant women experienced accelerated starvation; in other words, their blood measures substantially deviated when fasting was prolonged for 6 more hours. Therefore, since even skipping breakfast after overnight fasting causes such dramatic changes in expectant mothers, simply identifying someone as exposed might fail to recognize the importance of the length of nutritional disruptions during pregnancy. On the other hand, employing varying durations of daylight hours for an extended period allows us to capture the potentially detrimental effects of intrauterine dietary disturbances more precisely and improves the applicability of our findings. We address the differences in the length of fasting by using a relative measure of the moderated values of daylight hours for the first time and test the effect of exposure duration on human capital formation.

Second, the gap in the literature does not only arise from the number of studies analyzing the relationship between prenatal conditions and human capital being very limited but also that the existing narrow research mainly targets Muslim immigrants in developed countries and reports adverse outcomes for the exposed groups, e.g., differences in academic test scores among 7-year-old Muslim immigrant children in the UK (Almond et al. 2015), differences in the academic performances of ninth-grade Muslim immigrant students in Denmark (Greve et al. 2017), and differences in labor market outcomes of Muslim males in Denmark (Schultz-Nielsen et al. 2016). Nevertheless, migrants do not represent a random sample and might differ at core levels from the residents in their countries of origin (Borjas 1987). Thus, the obtained results might not be generalizable to other countries. One exception in the current Ramadan literature on human capital formation in a developing country context is Indonesia. Majid (2015) and Majid et al. (2019) consider the impact of intrauterine Ramadan exposure on children’s test scores and adult labor market outcomes. Their findings present lower academic performances and working hours among the exposed individuals. However, as an equatorial country, fasting duration and temperature do not vary throughout the year in Indonesia. Therefore, acknowledging the abovementioned critical risk of accelerated starvation among pregnant women, we believe it is crucial to examine if the picture drawn by the limited number of studies applies to a developing Muslim country like Turkey, with 99% Muslim citizens (CIA 2022).⁶

Our findings suggest a high sensitivity of educational and labor market outcomes of individuals to prenatal conditions, and it is particularly gender-specific. Concerning the academic outcomes, similar to the previous literature, we find that intrauterine Ramadan exposure has more pronounced adverse effects when it overlaps with the first or third trimester. Men who are exposed to Ramadan in utero in the first or third trimester complete 0.53 fewer years of schooling and are less likely to obtain middle school, high school, and college degrees, while no evidence of effects is present among females. This finding might be attributable to the differences between boys and girls in terms of their level of vulnerability to intrauterine shocks, as earlier literature states that boys are more easily and severely affected than girls by negative prenatal conditions (Eriksson et al. 2010; Di Renzo et al. 2007; Wang et al. 2023).

Contrary to our estimates, by studying the impact of in-utero Ramadan exposure on the academic performances of ninth-grade Muslim immigrant students in Denmark, Greve et al. (2017) report negative effects of exposure only for females. These opposite findings might result from a potential selection bias among migrants, since they might not be a perfect representative of the population, and/or the seasonality caused by using the 1985–1995 birth cohorts. As Ramadan’s adherence to the lunar calendar leads to an annual shift of 10 to 11 days backward, this timing change introduces seasonality problems into the analysis. It takes around 33 years for Ramadan to occur on the same date once again. This suggests that one needs to have at least 33 years of variation in Ramadan exposure to eliminate the biases in estimates caused by seasonal dietary, sunlight, and/or temperature differences. Since we have a broad sample of individuals between the ages of 33 and 97, following a similar approach to both van Ewijk (2011) and Almond and Mazumder (2011), we can separate the effect of Ramadan from seasonality and provide healthier results.⁷ On the grounds that the season of birth significantly affects educational outcomes (Verachtert et al. 2010), it is imperative not to confuse the effect of Ramadan with that of the seasons of birth.

Finally, our study contributes to the multiple lines of research within the realms of health and developmental studies. Proposing “the womb may be more important than the home,” Barker (1990, p. 1111) includes fetal life into the equation for the limitations of human capabilities and supports the “fetal origins hypothesis,” which suggests that what happens in utero has permanent effects on the offspring. During pregnancy, while the fetus goes through “critical” developmental phases, unfavorable in-utero conditions occurring at any of these stages might generate major adverse effects on the health of the offspring and might alter several later life outcomes (Godfrey and Barker 2001; Lumey et al. 2011; Brown et al. 2000). Being highly reviewed in epidemiological literature, recently, there is a growing interest in the fetal origins hypothesis in the economic literature.

The rest of the paper is organized as follows: Section 2 summarizes background information on Ramadan observance, Section 3 describes the data, Section 4 defines the methodology, Section 5 presents the results and conducts robustness analysis, Section 6 discusses the findings and potential policy implications, and Section 7 concludes.

2 Background on Ramadan fasting during pregnancy

During Ramadan, the holy month of Islam, healthy adult Muslims must fast from dawn to dusk. Starting from sunrise, adult Muslims should abstain from certain activities, including eating, drinking, smoking, and sexual intimacy, until sunset; after that, they can break fasting together with their families and enjoy traditional foods and sweets. Though fasting is mandatory for healthy adult Muslims, people traveling, elderly or sick, and women on their periods are excused. In addition, pregnant or breastfeeding women can also be freed from the immediate obligatoriness of fasting if fasting is harmful to them or their babies. However, they have to make up for these days as soon as the health of theirs or their offspring allows before the next Ramadan (Rashed 1992), or they have to make compensation for not fasting in terms of money or feed a disadvantaged person for each missed Ramadan day (Quran, Surah Al-Baqarah 2:184–185).

Since the exemption for pregnant women is not clearly stated, many Muslim women continue to fast during pregnancy despite its potentially harmful consequences (Seckin et al. 2014; Safari et al. 2019). The very vague expression regarding the exemption of expecting mothers from fasting makes some Muslims think that it is imperative to fast during pregnancy (van Ewijk 2011). Also, some pregnant women keep their fast on the grounds that they do not think of pregnancy as an illness to free them from the obligation; they think fasting is not harmful to a healthy woman or the fetus (Lou and Hammoud 2016; Mubeen et al. 2012). Another motivating factor for Muslim pregnant women is that they find fasting spiritually relieving, while not fasting is spiritually disturbing for them (Robinson and Raisler 2005). This emphasizes the crucial role of religious judgments on health-associated routines (Padela et al. 2018).

In addition, some women are concerned about the adverse consequences of observing the missed days right after delivery, i.e., during lactation, since they believe fasting might decrease their breastmilk (Ertem et al. 2001). Further, the fact that Ramadan is traditionally observed with family and friends, although exempt, since they must compensate for the missed days later on, pregnant women might find it challenging to make up for those days alone afterward (Seiermann et al. 2021). Moreover, they might want to avoid paying compensation. Thus, if they find fasting not so hard during pregnancy, they might opt to observe it during Ramadan.

On the other hand, the most important reasons for expectant mothers to refrain from fasting are their fear of harming the fetus and/or themselves, feelings of weakness, lack of nutrition, not being able to carry out daily activities, dizziness, nausea, and hunger and/or thirst (Mubeen et al. 2012; Joosoph et al. 2004; Awwad et al. 2012).

A recent qualitative medical study conducted in the western region of Turkey interviewed 17 pregnant women in detail to analyze their thoughts and practices regarding Ramadan observance (Uludağ and Göral Türkcü 2022). In this sample of 17 women, eight continued fasting during pregnancy. The ages of the respondents range from 19 to 30 years old. The education or socioeconomic status of those women does not seem to help explain their preferences for fasting decisions since the shares of high school and university graduate pregnant women are close to each other. They report similar motives listed above for the expectant mothers’ decision to fast: believing that they are obliged to fast, fearing the necessity of compensating for the missed days alone afterward, feeling spiritual relief, considering it healthy and safe, and abiding by social pressure. The study discloses that women feel discomfort, feelings of embarrassment, and disrespect if they eat anything while the rest of the family is observing Ramadan. This social aspect is more prominent among women living in an extended family structure with parents, parents-in-law, or grandparents. Considering the present-day approaches of pregnant women to fasting, the results suggest a strong resistance of religious beliefs both to time and medical knowledge.

Although in some studies, there is a very slight decrease with the increased level of maternal education, the rate of pregnant women observing Ramadan is calculated between 60 and 90% among Muslims around the world.⁸ Though there are quite a few high-quality medical studies conducted in Turkey during Ramadan to monitor the differences in fetal development between fasting and non-fasting pregnant women, there is no data to suggest a percentage of fasting among expectant mothers. However, the High Council of Religious Affairs, which is the responsible government body for expressing opinions, answering religious questions, and informing the public about the practices of Islam in Turkey, approaches the subject in the same vague manner by suggesting pregnant women not fast if they fear of harm to themselves or their babies and fast if both themselves and their babies are not at risk.⁹ Therefore, there is no particular reason to assume that the percentage of fasting during pregnancy will be dissimilar to other countries.¹⁰

3 Data

We use data from the TurkStat questionnaire of the “Turkish Family Structure (TFS) Survey” in 2016, which includes 17,239 households with 35,475 individuals aged 15 and over. The survey provides information on the individuals’ value judgments about “the lifestyles of the family and individuals” according to variables such as gender, age, education status, employment status, family structure, income group of the household, and other characteristics in individual and household level provided in 12 regions. People who did not report their exact birthdates are excluded from the analysis, which reduced the sample to 31,829 observations with people between the ages of 15 and 102.¹¹

Our primary interest is to analyze the effect of prenatal exposure to Ramadan on the educational attainment and labor market outcomes of people in Turkey. We use six separate measures for schooling: illiteracy, not having any formal education, years of schooling, middle school, high school, and college completion. We also evaluate the following labor market outcomes: employment, earnings, having a regular income, and precarious employment.

Independent from their ages, we eliminate the individuals who continue their education. In addition, we control for the change in the education policy that would have resulted in significant biases in our estimates. In 1996, the Ministry of National Education of Turkey (MoNE) implemented a new law on compulsory schooling, which became into action in 1997 and obligated people who were in the fourth grade or lower in 1996 to complete 8 years of schooling. The group of people who are affected most by the change in the law was born after January 1987. However, since there was no strictly implemented law on the school starting age in Turkey at that time, there were also people who started primary school 1 or 2 years later than their peers, or there were those who had to repeat grades before the change in the law. Therefore, in order to eliminate any effect of the change in the compulsory schooling law on the level of education that the people in our sample obtained, we limit our analysis to those older than 32 years of age at the time of the interview. In the remaining sample with complete birthdate information, there is some obvious heaping in specific dates of birth, among which January 1^st consists of around 20.4%.¹² Since the frequency of January 1^st birthdate is excessively high, we exclude these respondents from the sample entirely for the health of our analysis.¹³

Table 1 below presents the descriptive statistics of the variables used in our analysis for the pooled sample and then for females and males separately. Note that for schooling outcomes, we focus on those older than 32 years old. However, since labor market trends change substantially after retirement, we restrict our sample to individuals in the working-age group, between 33 and 65 years old. The age cut-off being the sole difference, the demographics in both groups are quite similar other than the mean age, which is 50 in the first group and 46 in the second.¹⁴ The share of females and males is equal. Around 40% of our sample grew up in rural areas; other than a small group growing up abroad, the rest lived in urban or suburban areas till mid-adolescence.

Table 1 Descriptive statistics of the study sample

On average, 8% of individuals in our sample are illiterate with a substantial difference between women and men, 14 and 2%, respectively. The gender discrepancy remains in the uneducated group, constituting 12% of the whole group. Overall, the mean years of schooling is calculated as 7.6 years, with men obtaining 2 more years of education than women. Favoring the males, gender disparities persist regarding middle school, high school, and college completion rates, with mean percentage values of 43, 33, and 16, respectively, in the pooled sample. For the labor market outcomes, employment and income statistics vary between women and men, while having a regular income or working at a precarious job is similar. Observing the differences, we also run separate regressions for each group.

4 Methodology

4.1 Identifying exposure

For determining prenatal exposure to Ramadan, the method used by van Ewijk (2011) is followed. The average duration of a human pregnancy is 266 days from conception; therefore, by calculating backward from the date of birth, we check whether the pregnancy overlapped with the corresponding Ramadan between 1913 and 1984.

Due to the nature of the timing of Ramadan, which shifts backward 10 or 11 days each year depending on the moon cycle, the base group of comparison is the ones conceived during a relatively brief period, which is approximately 60 days every Islamic year. Nevertheless, some pregnancies are longer than 266 days, so those who stayed longer in utero would be misclassified as not exposed while they instead were. In order to prevent the noise caused by this misclassification, we create a separate category for that group by setting a 3-week margin since pregnancies extending 3 weeks past their expected term are uncommon (Kieler et al. 1995). This category covers everyone whose mother’s pregnancy lasted longer, and it also separates those conceived right after Ramadan since there might be differences between them and the others.

We generate two types of exposure variables:

4.1.1 Overlap between Ramadan and pregnancy

For each person in our remaining sample, we create five binary variables to capture the overlap between the gestational periods and Ramadan. The first measure is a dummy variable for being certainly in utero, while the other three measures determine the trimester-specific exposure. The first trimester denotes the days of 1–89 in utero (266–177 days before birth), the second trimester represents the days of 90–178 in utero (176–88 days before birth), and the third trimester describes the days of 179–266 in utero (87–0 days before birth). Lastly, the fifth category is for those with a 3-week margin to reduce the noise caused by potential post-term births and post-Ramadan conception. It is worth pointing out that premature deliveries cannot be miscategorized as not exposed but can only be miscategorized as exposed since we assume the entire length of pregnancy is 266 days (Fig. 1). Limited research has been conducted concerning the impact of maternal fasting during Ramadan on the length of gestation. By utilizing data on pregnancy duration, Almond and Mazumder (2011) report a minor reduction in gestation length by roughly a day due to maternal fasting observance, while Karimi (2018) finds no particular evidence regarding the influence of Ramadan exposure on gestation duration.

Fig. 1

Assigning intrauterine Ramadan exposure of individuals born in 1976–1977

4.1.2 Duration of fasting

The duration of fasting changes depending on the hours of daylight, and in Turkey, it varies between 11 and 17.5 hours, depending on the season. As pregnant women suffer from accelerated starvation even after skipping breakfast, measuring prenatal exposure only based on the overlap between Ramadan and trimesters might not fully observe the effects of fasting at differing lengths. Therefore, we create an exposure variable based on the duration of fasting.

In order to account for the changes in the duration of fasting, previous studies use daylight hours rather than the actual duration of fasting while assessing the effect of exposure on health outcomes. However, fasting begins with the morning prayer (fajr) and, depending on the date, fajr might start up to 2 hours before sunrise.¹⁵ Nevertheless, the enhancement in our methodology does not stem from using the actual duration of fasting instead of daylight hours. Earlier literature addresses the changes in the duration of nutritional disruptions with an hours-of-exposure measure by summing up the daylight hours of all days overlapping with Ramadan during pregnancy (Almond and Mazumder 2011; Karimi and Basu 2018; Karimi et al. 2021). While employing the cumulative hours of exposure for each trimester is practical for interpretation, this approach could potentially yield inaccurate conclusions since it considers the exposure of person A, who encounters Ramadan during the first trimester for a span of 20 days, with an average fasting period of 9 hours (resulting in a cumulative exposure of 180 hours) as equivalent to that of person B who also undergoes Ramadan in the first trimester but for 10 days, during which the fasting period averages 18 hours (equating to 180 hours of exposure). This implies that in-utero experience of fasting for 9 hours across 20 days holds the same implications as fasting for 18 hours across 10 days in terms of potential adverse effects.¹⁶ As accelerated starvation occurs a lot quicker in pregnant women, totaling the hours of fasting would be misleading in estimating the effect of the duration of nutritional disruptions.

The timing of fasting is calculated based on the astronomical position of the Sun in the sky, thus it changes between longitudes. The solar elevation angle used for determining the astronomical twilight ranges between -15° and -20° across countries (Herdiwijaya 2020). Turkey has adopted the -18° angle for the calculation of dawn.¹⁷ Considering its geographical coordinates, there is around a 70-minute difference at the beginning and end time of fasting (which depends on the longitude) between the easternmost city, Igdir, and the westernmost city, Canakkale. In contrast, the duration of fasting (which depends on the latitude) is 25 minutes shorter or 55 minutes longer in the northernmost city, Sinop, than in the southernmost city, Hatay, depending on the season. In addition, the duration of fasting, conditional on the time of the year, increases or decreases in the same city by 30 minutes on average from the first day of Ramadan to the last day. However, our dataset does not provide individuals’ place of birth information. Therefore, also taking the population density into account, we choose Ankara as the weighted center of Turkey to base our exposure variable for the length of fasting (Fig. 2). Since the duration either increases or decreases for half an hour from the beginning to the end of Ramadan, we use the length of fasting on the 15^th day of the relevant Ramadan days as the average duration of fasting. Then, instead of summing up the total hours of fasting, we create an exposed hours measure by dividing the length of fasting on the 15^th day by the maximum hours of fasting a day in that year and assign that calculated percentage on trimesters overlapping with Ramadan. Having a value of 1 as the relative measure of exposure means that the individual got exposed to Ramadan in June when the duration of fasting is at its maximum, 17.5 hours.

Fig. 2

Provinces and region map of Turkey

4.2 Empirical framework

The epidemiological literature analyzes the differences in perinatal outcomes of fasting and non-fasting pregnant women by treating the preference over fasting as external. However, this approach fails to take self-selection bias into account since women who choose to fast might be different at base levels than those who choose not to. Therefore, instead of building upon the individual’s decision to fast, we follow the ITT methodology for identifying intrauterine exposure, which is the commonly used method in this flourishing literature where we check if Ramadan coincides with pregnancy (van Ewijk 2011; Almond and Mazumder 2011; Almond et al. 2015; Greve et al. 2017; Schultz-Nielsen et al. 2016). ITT approach eliminates the endogeneity caused by treating the fasting preferences of pregnant women as exogenous, thus turning Ramadan exposure into a natural experiment for mild nutritional disruptions during pregnancy. Consequently, those whose prenatal period did not overlap with Ramadan constitute the control group.

We should point out that our data does not have information on the fasting behavior of pregnant women (see Section 2 for Ramadan observance during pregnancy). Therefore, the ITT approach allows us to capture the impact of Ramadan overlapping with pregnancy rather than the effect of confirmed fasting. We assume that there is no selective timing of pregnancy present; in other words, Ramadan and pregnancy randomly coincide with each other (i.e., exogenous), yet the decision to fast is endogenous.¹⁸ Then, it becomes possible to measure the differences between the educational and labor market outcomes of individuals who experienced Ramadan in utero and those who did not. The fact that not all women whose pregnancy overlapped with Ramadan practiced fasting, but some of them certainly did to various extents, suggests that our ITT results are underestimated since the real impact is equally distributed among the sample consisting of both the fasting and non-fasting mothers, thus diluted. Therefore, if there is perfect compliance with observing Ramadan, ITT estimates would be equal to the average treatment effect (ATE). If women whose own health or infants’ health are at a higher risk in case of fasting choose not to fast, then the underestimation of ATE would be more pronounced in the findings. On the other hand, if women who are in great health abstain from fasting, then the underestimation would be slight. In addition, as the percentage of pregnant women who choose to fast changes between trimesters, the level of underestimation of ATE varies accordingly as well. We address the issue of seasonality with data covering more than 33 years of birth cohorts and an additional variable for month-fixed effects.¹⁹

For our analysis, we use the following four regressions for each outcome measure, ${y}_{icm}$:

$${y}_{icm}={a}_{\mathit{0}}+ \beta {{hours}\_{exposed}}_{cm}+{\mathit{a}}_{\mathit{1}}{{age}}_{{icm}}+ {a}_{\mathit{2}}{{age}}_{icm}^{\mathit{2}}+ {a}_{3}{{age}}_{icm}^{\mathit{3}}+{ a}_{4}{{age}}_{icm}^{\mathit{4}}+ {\omega }_{c}+{\gamma }_{m}+{\theta X}_{icm}+\delta {{sex}}_{icm}+{\varepsilon }_{icm}$$

(1)

$${y}_{icm}={a}_{\mathit{0}}+ {\sum }_{\mathit{\tau =1}}^\mathit{3}{\beta }_{\tau }{hour}{s\_{exposed}\_{trimester}}_{cm\tau }+{a}_\mathit{1}{{age}}_{icm}+ {a}_\mathit{2}{{age}}_{{icm}}^\mathit{2}+ {a}_\mathit{3}{{age}}_{icm}^\mathit{3}+{ a}_\mathit{4}{{age}}_{icm}^\mathit{4}+ {\omega }_{c}+ {\gamma }_{m}+{\theta X}_{icm}+\delta {{sex}}_{icm}+{\varepsilon }_{icm}$$

(2)

$${y}_{icm}={a}_\mathit{0}+ \beta {{exposed}}_\mathit{cm}+{a}_\mathit{1}{{age}}_\mathit{icm}+ \mathit{a}_\textit{2}{{age}}_{icm}^\mathit{2}+ \mathit{a}_\mathit{3}{{age}}_{icm}^\mathit{3}+{ a}_\mathit{4}{{age}}_{icm}^\mathit{4} {+\, {\omega }_{c}+ \gamma }_{m}+{\theta X}_{icm}+\delta {{sex}}_{icm}+{\varepsilon }_{icm}$$

(3)

$${y}_{icm}={a}_\mathit{0}+ {\sum }_\mathit{\tau =1}^\mathit{3}{\beta }_{\tau }{{trimester}}_{cm\tau }+{a}_\mathit{1}{{age}}_{icm}+ {a}_\mathit{2}{\mathrm{age}}_{icm}^\mathit{2}+ {a}_\mathit{3}{{age}}_{icm}^\mathit{3}+{ a}_\mathit{4}{{age}}_{icm}^\mathit{4}+ {\omega }_{c}+ {\gamma }_{m}+{\theta X}_{icm}+\delta {{sex}}_{icm}+{\varepsilon }_{icm}$$

(4)

where $\it {y}_{icm}$ represents the outcomes of interest for the individual i, born in cohort c, and month m. We check for illiteracy, having no formal education, years of schooling, middle school, high school, and college completion for educational measures; employment, log of income, having a regular income, and precarious employment as the labor market outcomes. ${hours}\_{expose}{d}_{cm }$ and ${hours}\_{expose}{{d}\_{trimester}}_{cm\tau }$ capture the duration of fasting as a fraction of the maximum length of fasting a day where the first measure assesses the overall exposure during pregnancy; the latter identifies trimester-specific variations in fasting duration. Similarly, ${{exposed}}_{cm}$ is the dummy indicator, which takes the value of 1 if the individual was in utero during Ramadan, and ${{trimester}}_{cm\tau }$ is a binary variable for Ramadan overlapping or starting in the corresponding gestational period. Those who are certainly not exposed to Ramadan constitute our control group for the analyses. ${age}_{icm}$ is the age of the respondent, ${\omega }_{c}$ and ${\gamma }_{m}$ are the birth cohort and month fixed effects, respectively; and ${X}_{icm}$ is the individual level variables such as mother tongue, place the individual grew up till the age of 15, and the current region that the respondent resides. We also control for gender in the analysis with the variable ${sex}_{icm}$. Lastly, ${\varepsilon }_{icm}$ shows the weighted error terms.

5 Results

We first analyze the effects of intrauterine Ramadan exposure on the educational outcomes of individuals by utilizing the length of fasting as a fraction of the maximum duration of Ramadan fast a day. Observing the differential effects by gender, we check for the impacts separately for females and males. Next, we explore the labor market outcomes of individuals. Thereafter, we check whether our findings are robust to alternative exposure measures by using exposure dummies, which is the commonly addressed approach in the growing Ramadan literature. In addition, since we run multiple hypotheses simultaneously, in order to prevent the risk of erroneously rejecting the null hypotheses, we employ the Romano-Wolf method for multiple hypothesis testing to ensure robust and consistent estimations.

Table A2 in the Appendix presents the summary statistics of outcome measures for the exposed and not exposed individuals.

5.1 Main results

5.1.1 Effects of prenatal exposure to Ramadan on educational outcomes

We utilize the Eqs. (1) and (2) to analyze the effect of prenatal exposure to Ramadan on educational outcomes. Excluding those who reported their birthdays as January 1^st, for all the remaining respondents who already completed their studies and are above the age of 32, the main results are mostly weak for the pooled sample (Table A5). In order to assess the possible gender differences in terms of vulnerability to intrauterine shocks, we control for the interaction terms between the dummy variables of exposure and gender. Figures 3 and 4 present the point estimates of the interaction variables and the corresponding 95% confidence intervals, where complete results are reported in Panel A and Panel B of Table A6, respectively. With the additional male interaction terms, Fig. 3 demonstrates the overall effect of intrauterine exposure to Ramadan. At the same time, Fig. 4 compares the effects of Ramadan exposure in distinct stages of pregnancy with those who are not exposed at all.

Fig. 3

The overall effect of prenatal exposure to Ramadan with male interactions

Fig. 4

The trimester-specific effects of prenatal exposure to Ramadan with male interactions

The obtained coefficients for the pooled sample suggest a differential male vulnerability to in-utero shocks, while the overall exposure estimates become insignificant following the introduction of interaction terms (estimation results without interactions are presented in Table A7). We find a significantly negative effect of being exposed to Ramadan in utero in the first or third trimester on years of schooling and completing at least middle school for males. This result is consistent with the findings of Majid (2015), who also reports negative effects of exposure in the first and third trimesters on math scores.

One explanation behind this finding could be that maternal fasting in the third trimester alters cognition by failing to provide a favorable growth environment. To better understand the effect of Ramadan fasting on fetal health, Sakar et al. (2015) conducted a case-control study in Turkey with 166 pregnant women who were in their second or third trimester during the month of Ramadan in 2013. In that year, Ramadan occurred in the summer season when the duration of fasting ranged from 16 to 17 hours. Both the fasting and the non-fasting samples had 83 pregnant women. Doppler ultrasound was used to assess the fetal biometric measurements both at the beginning and the end of Ramadan. Their findings show that the amniotic fluid,²⁰ which is vital for the fetus to go through healthy developmental phases in the womb, built up substantially in a month among pregnant women who did not observe Ramadan compared to fasting pregnant women.

Though by the second trimester all key brain structures have developed (Sur and Rubenstein 2005), the third trimester is characterized by rapid brain advancement and enhancement of connections occurring across various domains (Kostovic and Vasung 2009). When fetal blood flow is disturbed, the brain gets threatened by developmental impairment during the third trimester (McQuillen et al. 2010). Mirghani et al. (2005) conduct a cross-sectional study with 124 fasting and non-fasting pregnant women who are otherwise healthy. The authors collect blood samples and trace fetal heart rate. They find significantly lower numbers of large accelerations in fetal heart rates among fasting women. Therefore, another explanation for the negative impact of maternal nutritional disruptions in the third trimester on schooling outcomes could be the changes in the fetal heart rate caused by maternal fasting behavior.

As both our estimates and the medical literature suggest that the severity of the effect of malnourishment during pregnancy on women and men differs, we test the outcomes for females and males separately. Table 2 below presents the effects of intrauterine nutritional disruptions on schooling measures by gender while capturing the differences in the length of nutritional disruptions. Our results suggest no damaging impact of prenatal exposure to Ramadan on the educational outcomes among females. However, we find substantially detrimental effects of experiencing fast in utero on schooling measures for males. Our findings in Panel A show that men who experience Ramadan in the summer season obtain over half a year less education compared to non-exposed males. Our Panel B results indicate that intrauterine exposure to Ramadan in the first or third trimester during summer results in 0.55 and 0.62 fewer years of education among men, respectively. For middle school completion, both first- and third-trimester shocks pose a significant risk for males. Exposure in the third trimester presents a larger impact than the first, yet this minor difference is statistically insignificant (p-value > 0.73). Coming across a period of fetal malnourishment in the third trimester implies that a man would be less likely to complete high school. For finishing a graduate degree, both first- and third-trimester conditions seem critical for men.

Table 2 Effects of hours of prenatal Ramadan exposure on educational outcomes by gender

Prior to drawing attention to the gender differences, we should note that the limited number of studies focusing on in-utero programming within the context of Ramadan also report pronounced effects of the shocks experienced in the first or third trimester on academic outcomes (Almond et al. 2015; Majid 2015; Greve et al. 2017). Similarly, the key role of the initial and final trimester conditions on human capital formation is evident within various frameworks of the fetal origins hypothesis literature. For instance, middle school children in Sweden who were prenatally exposed to the Chernobyl disaster within the first 5 months are observed to have lower math scores and reduced likelihood of obtaining high school education (Almond et al. 2009). Likewise, Tian et al. (2022) note a negative impact of enduring the earthquake in the third trimester on high school completion rates in China. Rather than solely focusing on the negative, as an example of a positive shock, Field et al. (2009) analyze the success rate of prenatal iodine supplementation on schooling outcomes in Tanzania. The authors also report that supplementation during the first trimester of pregnancy results in an increase in schooling of around 0.5 years, and girls are found to benefit more.

Figure 5 plots the point estimates of trimester-specific duration of exposure and the 95% confidence intervals for women and men separately (Panel B results shown in Table 2).²¹ The figure makes the male disadvantage regarding the persistent effects of intrauterine dietary disruptions on schooling more evident, while female academic outcomes are not significantly disturbed. This might be a result of the already low level of education among women in Turkey (see Table 1), which might not be hindered further by the in-utero exposure to Ramadan. However, the explanation for the gender discrepancies regarding the outcomes of intrauterine shocks might not be straightforward. The epidemiological literature provides evidence that male fetuses are more vulnerable to unfavorable in-utero conditions. Similarly, Karimi et al. (2021) and Karimi and Basu (2018) examine the effect of prenatal exposure to fasting on height in Iran and also report negative results only among the exposed male children.

Fig. 5

Point estimates of trimester-specific duration of exposure and the 95% confidence intervals for women and men

In the case of in-utero shocks, boys are more defenseless against external prenatal traumas than girls (Hansen et al. 1999). Even if they survive, adverse events during pregnancy are more likely to damage the male fetuses than females (Di Renzo et al. 2007).²² One of the explanations behind this phenomenon is the fetal size differences between genders. The fetuses optimize their development under the constraints of the finite nutritional resources that their mothers provide. At each stage of pregnancy, boys progress quicker and are taller than girls (Eriksson et al. 2010). Though this might seem like an evolutionary advantage, on the contrary, it puts males at a greater risk of fetal malnourishment compared to females since their increased placental size requires more dietary intake (Pedersen 1980).

In addition, the male brain is larger with a bigger hypothalamus, which is a potential result of elevated levels of testosterone secretion in the third trimester (Kraemer 2000; Earls 1987). Previously, the hypothalamus was only credited as a component of the limbic system; however, recent studies present that it is an integral part of cognitive functions (Burdakov and Peleg-Raibstein 2020). Therefore, suffering from nutritional shortages in the third trimester might hinder the healthy cognitive development of infants, especially males, which might explain the pronounced gender discrepancies in our findings.

5.1.2 Effects of prenatal exposure to Ramadan on labor market outcomes

The effects of in-utero Ramadan exposure on labor market outcomes are assessed by the following measures: (1) labor market participation, (2) log earnings, (3) having a regular income, and (4) precarious employment. We treat the variable of having a regular income as a financial security and working on a daily basis as an indicator of precarious employment. For the labor market outcomes in the pooled sample, while we find no significant results for workforce participation and the logarithm of income, our results marginally suggest an unexpectedly positive effect on having regular income and job security among those who experienced maternal fasting in the third trimester (Table 3).²³

Table 3 Effects of hours of prenatal exposure to Ramadan on labor market outcomes

We then turn our attention to the gender differences in labor market outcomes given the estimates of the male interaction variables (Table A9) and the significant reduction in educational attainment due to intrauterine Ramadan exposure among men. Employing Eqs. (1) and (2), Table 4 presents the effects of exposure for females and males separately by accounting for variations in fasting duration. While no substantial effects are observed for females, the results indicate positive effects for males, where men who are prenatally exposed to Ramadan in the second or third trimester are more likely to have a regular income and secure employment.

Table 4 Effects of hours of prenatal exposure to Ramadan on labor market outcomes by gender

A possible explanation we can consider for these unforeseen outcomes is the presence of governmental and societal support mechanisms specifically aimed at assisting disabled individuals.²⁴ Our dataset lacks information related to disabilities. Similarly, Schultz-Nielsen et al. (2016) also encounter a limitation of not having a direct measure of disability in the Danish dataset of male workers. To compensate for this, they utilize “early retirement” as a proxy for disability and observe that Muslim immigrants who were exposed to Ramadan during the seventh month of gestation exhibit a higher likelihood of retiring early. In addition, Almond and Mazumder (2011) report increased rates of disability among Iraqi and Ugandan men who experienced Ramadan in utero in the ninth month. However, additional data and analyses are required to confidently elaborate on these unexpected findings.

5.2 Robustness analyses

5.2.1 Alternative exposure measures

We run several robustness analyses to support our primary findings. Consistent with the common practice in the current Ramadan literature, we employ exposure dummies to assess the effect of in-utero exposure to fasting on human capital formation. Observing the gender discrepancies in terms of vulnerability to prenatal nutritional deficiencies (Table A6), using Eqs. (3) and (4), Table 5 presents estimates for the effect of prenatal exposure to Ramadan on educational outcomes for females and males separately.

Table 5 Effects of prenatal exposure to Ramadan on educational outcomes by gender

Notably, though the magnitudes differ, we reach similar conclusions using both approaches: while no significant effects are observed for females, the results indicate substantially negative effects for males. We should note that the current estimates are smaller compared to our baseline findings, where we address the potential accelerated starvation by employing differing durations of fasting. The exposure variable in Panel A points out that men who are exposed to Ramadan complete around 0.48 fewer years of schooling compared to those who are not exposed (while the initial approach suggests over half a year less education as a result of exposure in summer). They are also less likely to obtain middle school, high school, and college education. When we check for the impact of intrauterine exposure at different stages of pregnancy, our results in Panel B indicate that men experiencing Ramadan in the first or third trimester, on average, complete 0.53 fewer years of schooling (whereas employing hours of fasting measures denotes a decrease of 0.59 years of education on average). For middle school completion, experiencing the nutritional shock in the first or third trimester has similar degrees of negative impact (with a 1 percentage point difference from the previous approach), while for high school completion, though the level is similar, only the exposure in the third trimester is found to cause significant limitations for men. For college completion, our findings imply that, though slight, experiencing Ramadan in the first or third trimester has damaging impacts on males.

Subsequently, we employ the exposure dummies for the labor market estimates. Due to the apparent gender discrepancies regarding the vulnerability to intrauterine shocks (Table A9), Table 6 presents the effect of intrauterine nutritional shocks on employment measures for women and men separately.²⁵ Our findings are notably robust to alternative measures of exposure where undergoing Ramadan during the second or third trimester yields unexpected positive effects on both having a regular income and improvement in job security among males.

Table 6 Effects of prenatal exposure to Ramadan on labor market outcomes by gender

5.2.2 Restricted sample

The initial sample we use includes individuals conceived or born during Ramadan. Therefore, we limit our sample to those who experienced a complete Ramadan in utero. We find no significant effect of a full month of Ramadan exposure on the educational outcomes for the pooled sample or females. Since illiteracy and lacking formal education are less common among men, we focus our attention on educational attainment. Table 7 below presents estimates for males, which are similar to our baseline findings. Both the first and third trimesters remain significant in the restricted sample. The negative effects of exposure in the first trimester are slightly larger than our initial findings if the individual experiences a full month of Ramadan in utero. Finding no effect on high school completion rates, our results show that men who experience a complete month of Ramadan in summer obtain 0.57 fewer years of education. In addition, they are less likely to complete middle school and college education.

Table 7 Effects of full-month Ramadan exposure in utero on educational outcomes among males

Similarly, we observe no notable impact of an entire month of Ramadan exposure on labor market results for either the combined group or for females. However, for men, consistent with our baseline findings, experiencing a complete Ramadan in summer overlapping with the second or third trimester yields unexpected positive results regarding the consistency of their earnings (Table 8).

Table 8 Effects of full-month Ramadan exposure in utero on labor market outcomes among males

5.2.3 Multiple hypothesis testing correction

Due to the simultaneous testing of multiple hypotheses, the p-values obtained from Eqs. (1)–(4) could potentially be underestimated (Romano and Wolf 2005a, b). In order to mitigate the risk of erroneous rejection of null hypotheses, we utilize the Romano-Wolf correction technique to simultaneously test multiple hypotheses regarding the educational and labor market outcomes of men (Clarke et al. 2020).²⁶ Table 9 reports the model, resample, and Romano-Wolf step-down adjusted p-values computed by employing both continuous and dichotomous measures of exposure to maternal fasting.

Table 9 Multiple hypothesis testing (Romano-Wolf) for educational outcomes among males

The Romano-Wolf step-down adjusted p-values demonstrate statistical significance in both models. Years of schooling and middle school completion rates remain significantly lower among men when gestation overlaps with Ramadan. Our findings indicate that men who are prenatally exposed to Ramadan during summer complete 0.53 fewer years of schooling, while exposure to maternal fasting implies a reduction of 0.48 years in education. Both approaches demonstrate the damaging impacts of intrauterine exposure to Ramadan on middle school completion. Though the trimester-specific duration of fasting measures become insignificant, using the alternative measure of exposure, we find robust evidence for the detrimental effects of intrauterine Ramadan exposure in the first or third trimester on educational outcomes among males. Our findings ensure the detrimental effects of maternal dietary disruptions on academic achievements of males. On the contrary, the Romano-Wolf adjusted p-values are insignificant for the labor market outcomes (Table A11). Overall, we find that the results from different sample restrictions and methods are consistent with our baseline findings.

6 Discussion

The significant indication of our findings is that the differences in prenatal well-being might be one of the very explanations behind the varying levels of human capital among individuals. Although Ramadan observance is practically a form of daytime intermittent fasting, it is still potent enough to negatively affect academic and labor market outcomes. Therefore, for creating more qualified human capital, simple investments during pregnancy might be more effective than the conventional and late policy implications in the areas of education. We should note that our findings apply not only to the Muslim population but also to developed or other developing countries. Skipping meals or morning sickness is common during pregnancy, especially at the beginning of gestation, since pregnancy often goes unrecognized. However, nutritional deficiencies in the first trimester substantially disturb healthy cognitive development.

In addition, a study comparing the perinatal outcomes of newborns to mothers who were exposed to Ramadan in utero and who were not finds lower birthweights and narrower placental sizes among the infants of exposed mothers (Alwasel et al. 2013). The fact that the effect of mild fetal nutritional disruptions extends beyond generations proves that simple investments during pregnancy would be more effective than the conventional and late policy implications for creating more qualified human capital. Thus, simple interventions such as using media support to highlight the detrimental effects of maternal fasting, dieting, or skipping breakfast on the cognitive capacities of their babies might mitigate the potential losses among children. In addition, providing nutritional support to disadvantaged pregnant women becomes important, especially in developing countries where education is imperative for escaping the poverty cycle. Given the intergenerational transmission of education and socioeconomic status, the positive effects of such implications would be highly multiplied across generations.

In as much as we imply that our findings are induced by fasting, dietary disruptions are not the only changes occurring during Ramadan (Lessan and Ali 2019). Changes in caloric intake, an increase in the consumption of fatty and sweet foods, disturbed sleep cycles, and dehydration could all affect the development of the offspring and shape the very assets that s/he will be born with, hence forming her/his future.

7 Conclusion

This paper provides new evidence on the long-term effects of maternal nutritional disruptions on the human capital accumulation of the next generation. We utilize the Islamic holy month of Ramadan as a natural experiment to assess the effects of fasting during pregnancy on educational and labor market outcomes, which we treat as proxy variables for the cognitive abilities of individuals. Differing from the previous studies, we assess the effect of intrauterine Ramadan exposure on these measures by employing variations in the duration of fasting. Our findings present detrimental consequences of unfavorable in-utero conditions on human capital formation, and these effects are only robust among males, with no significant effect present among females. Our analysis indicates that men who are exposed to Ramadan in the first or third trimester obtain 0.53 fewer years of schooling on average and that this impact is stronger in summer. They are also less likely to attend middle school, high school, and college. These findings are robust both to alternative measures of exposure and the multiple hypothesis testing correction methods. Therefore, we present evidence that prenatal exposure to Ramadan in the first or third trimester onwards has a persistent negative effect on males.

These highly pronounced gender discrepancies regarding the vulnerability to intrauterine shocks could be explained by fetal programming. It is an organic phenomenon suggesting that the changes in the prenatal conditions induce alterations in the nutrition supply to the baby, which results in a process of long-lasting transformations (Barker and Thornburg 2013). Though this might be beneficial in the womb, it dramatically fails when the outside environment is different than in utero. Male fetuses are bigger than females; this size difference requires higher levels of placental nutrition, and in return, it makes boys more vulnerable to fetal malnourishment compared to girls. Since pregnant women enter the accelerated starvation phase a lot quicker than non-pregnant women, even mild maternal dietary shortages might put the male fetuses at greater risk and limit cognitive advancement. Our results lend support to the idea that even a temporary and less severe form of dietary deficiency, such as Ramadan, might hinder the cognitive development of individuals and shape the partially fixed assets that they will be born with.

Thinking within the frame of the economic discipline, Heckman (2007) proposes different phases of child development as separate inputs that produce the “capacity” of the child, where I₁ is the prenatal investment while I₂ is postnatal. According to his model, supported by medical literature, since early childhood cognitive development has its unique periods, I₁ and I₂ cannot perfectly replace each other. Therefore, the capacity cannot surpass the level set in the prenatal period. This very framework stresses the importance of fetal programming and prenatal investments and signals that such early interventions might be more efficient and economical both for the individual and the society as a whole (Doyle et al. 2009; Almond and Currie 2011a, b).