1
The Fasting and the Spurious:
The Effect of in Utero Ramadan Exposure on Socioeconomic Outcomes
Tristan Al-Shehab
University of Amsterdam
Supervised by Pauline Rossi
June 15, 2017
Abstract
Many individuals assume that pregnant Muslim women do not fast during the holy month of Ramadan. As one of the five pillars of Islam, it is in fact expected of them to do so, unless there is a serious concern that it could do harm to themselves or their unborn foetus. In fact, the majority of Muslim women do indeed fast during Ramadan. This impact could be large considering over 80% of the 1.8 billion Muslims worldwide are likely to have been in utero during Ramadan. This paper uses the occurrence of Ramadan as a natural experiment to measure the impact of in utero Ramadan exposure on socioeconomic outcomes, namely educational attainment and earnings. Younger, male Muslims exposed in utero to Ramadan obtain up to 0.4 years less education than Muslims not exposed to Ramadan. Muslims in the top quintile of earnings experience up to a 10% decrease in income when exposed to Ramadan in utero. This paper agrees with previous literature which conclude that the effect of in utero exposure for the Muslim population as a whole can be mixed when considering both health and socioeconomic outcomes, but is an area that demands more attention.
2
1. Introduction
Expectant mothers are often advised to use caution with their consumption behaviour during pregnancy, whether that be smoking (Meyer, 1977), drinking (Mills, 1984) or eating (Lechtig, 1975). This is not only because it may cause adverse health effects to the mother, but also the foetus when it is born, which could cause long lasting damage. The month of Ramadan for Muslims represents a situation where the mother’s behaviour could indeed cause long lasting damage to their unborn children if they decide to fast. The majority of pregnant Muslims still fast in Ramadan, meaning that most children of Muslim parents have been exposed to some degree of Ramadan fasting in utero. Therefore, while the health risks of fasting during Ramadan might be considered, is it at all possible that this behaviour has long term socioeconomic effects on their children? This question has scarcely been researched in previous literature, but could provide insights into how early childhood development affects an individual’s educational attainment and earnings potential. The returns to education have been well established (Patrinos, 2004), hence adjusting fasting behaviour could represent a cost effective policy to improve socioeconomic outcomes, in comparison to traditional educational interventions (Heckman, 2000).
In this paper, I use data from the fifth wave of the Indonesian Family Life Survey in 2014 to study the impact of in utero Ramadan exposure on educational attainment and income, seeing if there exists a channel of reduced cognitive ability. The effect on income in fact has not been studied previously, hence adds a new dimension to what has been written about Ramadan fasting to date. This paper builds upon the findings of van Ewijk (2012) which conclude from an earlier wave of the same survey that in utero Ramadan exposure can lead to a significant decrease in body mass index (BMI). I discuss the reasons why my research contradicts these findings, despite employing a similar empirical strategy. I decide to also use data from Indonesia for two important empirical reasons. Firstly, Indonesia has the largest Muslim population in the world, allowing for a representative view of outcomes. Secondly, being close to the equator, daylight hours remain fairly constant throughout the year in Indonesia, and therefore fasting hours are also constant, regardless of the month that Ramadan falls on the Gregorian calendar.
My identification strategy involves calculating an individual’s exposure to Ramadan in utero by estimating their date of conception from their date of birth, and seeing if Ramadan in that year overlapped with their gestation period. This helps remove any seasonality effects and allows the inclusion of many cohorts, unlike most medical studies. This intention to treat (ITT) approach results in conservative estimates of any true effect of in utero Ramadan exposure and I explain why this is the case. The key assumption with this strategy is that parents do not plan births around Ramadan in order to improve either their health or socioeconomic outcomes. I present evidence showing that this scenario is unlikely to be a legitimate concern to bias my results.
I address the confusion that both Muslims and non-Muslims have when considering what groups of individuals are exempt from fasting, and whether pregnant mothers are well informed about the potential consequences of fasting. This paper also explains the economic reasons for engaging in potentially harmful fasting behaviour, using the club good model of Iannaccone (1992).
I find that younger, male Muslims are most negatively affected by in utero Ramadan exposure, in terms of the years of education they receive. Furthermore, earnings for those in the top 20% of the population are lower than Muslims not exposed to Ramadan in utero. Individuals in this bracket of earnings are working in skilled occupations that require education. There is some evidence to suggest that the channel for these findings is through a reduction in cognitive ability, however this is not robust to the type of cognitive ability test used. I do not obtain similar results to any outcome when considering the non-Muslim population.
3
The remainder of this paper is organised as follows. Section 2 outlines background on the behaviour of fasting, as well as studies that explore health and socioeconomic outcomes. Section 3 describes the data set used, how outcome variables were constructed, and the empirical strategy. Section 4 presents findings on health, education, income, and cognitive ability. Section 5 discusses the results, empirical issues, and future research.
2. Background and Previous Literature
2.1 Ramadan and Islam
As one of the five pillars of Islam, taking part in Ramadan is central to Islamic beliefs. Ramadan is supposed to bring people closer to God, acting as a reminder of those less fortunate. By removing worldly pleasures, it allows Muslims to focus on their inner self. Ramadan prohibits the consumption of food and drink during daylight hours for approximately 30 days each year according to the Islamic calendar, where Ramadan is the ninth month. In addition to this, smoking, caffeine and sexual intercourse are generally prohibited from dawn until dusk. Muslims generally start fasting in Ramadan when they reach puberty1.
Exemptions for fasting can be provided for individuals due to: travel, menstruation, severe illness, pregnancy, and breastfeeding. However, there are caveats with this statement. Firstly, those who miss the fasting due to these reasons are encouraged to make up the fasting at a later date. Secondly, it seems in general that there is a misunderstanding whether pregnant women are actually exempt, because if all pregnant Muslim women skipped fasting, there would be no motivation for this research. Pregnant Muslim women are only advised to skip fasting if there are health complications during pregnancy or that they have a reason to believe that fasting will cause harm to themselves and their baby. Because fasting may not cause serious health damage to a pregnant women or the foetus, Muslim women still generally join in with fasting to connect with their faith and the celebratory breaking of the fast each day known as Iftar.
In addition, many women in fact do not believe that fasting while pregnant is harmful. Joosoph et al. (2004) examined pregnant Muslim women in Singapore to find that 79% of subjects believed fasting was not harmful to either themselves or their foetus during Ramadan. Overall, 89% fasted for at least one day, with 74% fasting for at least 20 days. I calculated a weighted average of 22 days of fasting from their study, which is equivalent to 73% of the length of Ramadan. There also seems to be large encouragement from spouses and family to continue fasting, with only 7% of spouses and 5% of families against fasting during pregnancy. The authors also note that the convenience and camaraderie of fasting combined with the difficulties of fasting outside of Ramadan encourage women to fast e.g. food shops are often closed during daylight hours in Ramadan.
Other estimates of the number of pregnant women who fast in Ramadan are of a similar magnitude to the above, including 86% in Walsall, United Kingdom (Gatrad, 1994), around 70% (depending on the trimester) in Hamadan Province, Iran (Arab, 2001), and 78% in Malaysia (Salleh, 1989).
Economic literature exploring the effect of Ramadan fasting has generally ignored explanations of why this behaviour takes place in the first place from an economic perspective. Religion is not widely explored in economic literature, but religion can actually form a key part in determining and individual’s behaviour and hence the allocation of scarce resources within a society. Iannaccone (1992) presents religion as an industry that has low barriers to entry, is highly competitive, and has no intellectual property rights. A
1 An overview of guidelines while fasting in Ramadan can be found here, from the Islamic Center of Raleigh:
https://raleighmasjid.org/imam/fiqh-ramadan.html. Guidelines can vary between types of Islam and location, therefore I have presented a representative view of what other similar websites state.
4
member’s utility/religious satisfaction is dependent upon their own inputs, and that of others. This means that as a club good (non-rivalrous but excludable), members with low participation are tempted to free ride off members with higher levels of participation. This free rider problem can be avoided by increasing costs to screen out members with low participation, increasing utility for the whole group (religion). Such a cost can be any religious practice that may seem to be unproductive, including fasting in Ramadan. An alternative to screening would be to have membership fees that then subsidise participation, however this could only work if the group could observe an individual’s participation. Therefore members must either participate fully, or not at all, with those remaining finding their utility increased, especially in a society with higher participation rates in the religion. In the model described, individually maximizing Nash behaviour is not Pareto optimal and Nash equilibrium utility will always be less than the Pareto-optimal utility. Although fasting in Ramadan is difficult to enforce as it is largely private behaviour, feelings such as guilt, habit and self-enforcement mechanisms can help encourage high participation even when there are potential health costs to a pregnant woman and her baby.
2.2 Education and Socioeconomic Outcomes
Almond and Mazumder (2011) use census data from Uganda to primarily look at health outcomes, but also socioeconomic ones. They use a strategy similar to the OLS estimation I shall present, calculating one’s in utero Ramadan exposure by seeing if Ramadan overlapped between their date of birth and estimated date of conception. They however do not have precise birth dates available and only the month of birth, making it a less accurate estimation. They find that men exposed to Ramadan in the first two months of gestation are around 2% less likely to own a home, which is not particularly large given that the rate of home ownership in Uganda is 73%. They do not find Ramadan exposure to have significant effects on illiteracy, years of schooling or the probability of being employed, however certain individual gestation periods can be found to be significant. This is not particularly surprising though, considering they measure 9 different gestation months, making it statistically likely that one is significant at the 10% level. Overall they conclude that the socioeconomic effects are small and far less consistent than their results for health.
Almond et al. (2015) found that test scores for 7 year olds in maths, reading, and writing are between 0.05 and 0.08 standard deviations lower for Pakistani and Bangladeshi students in England who were exposed to Ramadan in utero. The authors use a difference in difference strategy, with the first difference being whether a child was exposed to Ramadan in utero or not, and the second difference is between students of a Pakistani and Bangladeshi origin compared with Caribbean students. With a sample of students from 1998 to 2007, the results are strongest in the first three months of pregnancy for all subjects. While it is questionable that Caribbean students serve as suitable control group, their results are robust to comparing against white British students. This strategy serves as viable alternative when using a smaller number of cohorts.
Azizi et al. (2004) studied the long-term impact of fasting on human capital outcomes using a historical cohort study on 191 children between 4 and 13 years old. Of these children, 98 had mothers who fasted during Ramadan (treatment group) and 93 had mothers who did not fast (control group). They find no significant effect of maternal fasting behavior, during the third trimester of pregnancy, on the intelligence quotients (IQs) of school-age children.
5
There exists medical theory to suggest that fasting during pregnancy is harmful for the mother and her foetus. Fasting creates a shortage of nutrition that hinders the growth of the foetus, namely the lack of glucose (known as hypoglycaemia or low blood sugar). The mother’s body can partially compensate for this by increasing the rate of fat metabolism, where fat is broken up into fatty acids and ketone bodies. If the number of ketone bodies is too high (ketoacidosis) then blood pH levels can decrease, which then destroys proteins in the body. This can then lead to tissue damage, organ failure and even death. Pregnant women are therefore at an increased risk of reaching hypoglycaemia and ketoacidosis because of the increase energy requirements of the placenta and foetus. Metger et al. (1982) developed a phrase known as ‘accelerated starvation’, which describes the increased speed of reaching bodily states seen in starvation for pregnant women. Najimudeen (2016) summarises the main medical findings of fasting while pregnant. Najimudeen concludes that there exists studies that show adverse effects when fasting during pregnancy and also studies that report no significant impact. While many of the studies are small or contain methodological flaws, it appears that the pregnant mother’s body adapts to the different nutritional status, hence cannot be called starving or malnutrition. The paper highlights the lack of evaluation of long term outcomes for the child.
Van Ewijk (2012) uses a sample of 18 years old and above from the third wave of the IFLS in 2000 to find a significant reduction in body mass index (BMI) when exposed to Ramadan in utero. The strategy employed is analogous to that of Almond and Mazumder (2011), where exposure to Ramadan in utero is calculated by seeing if Ramadan overlapped with an individual’s gestation period, estimating their conception date 266 days before their date of birth. Using an OLS estimation, the results are significant for being conceived in Ramadan, and exposure to Ramadan in the second and third trimester. There is however a questionable use of household fixed effects in the regressions, since with a sample of 18 years old and above means that the household members are far more likely to be spouses and hence not genetically related family members, assuming children are likely to leave their home once adults. Household fixed effects would therefore incorrectly estimate variation within households since height is largely determined by genetics (McEvoy, 2009).
Almond and Mazumder (2011), using Michigan Natality files, find that babies in utero in Ramadan were around 18 grams lighter, with larger effects in the first and second trimester. They use a sample of Arabs in their data, rather than explicitly Muslims. They also use census data from Iraq and Uganda with only data on birth months, rather than day. In Uganda, the sample is 20-80 year olds, and find a 22% increase in the disability rate when exposed to Ramadan in utero. The Iraq data uses a sample of 20-39 year olds, finding a 23% increase in the rate of disability.
3. Data and Methodology
3.1. Indonesian Family Life Survey (IFLS)
The Indonesian Family Life Survey is a longitudinal socioeconomic and health survey, based on a sample of households representing 83% of the population in Indonesia. It consists of 5 waves conducted every 7 years, the most recent being IFLS5 in 2014, which I use for my analysis. This data set not only has the benefit of being the most recent at the time of writing, but also the most rich, with over 16,000 households and 50,000 individuals interviewed. The IFLS5 is also the first wave that has moved from a paper questionnaire to a computer assisted personal interview system (CAPI), which reduces errors made by interviewers, mainly with respect to skipping questions. This wave also introduced ‘showcards’ which help respondents understand questions in several key languages in addition to Bahasa Indonesia.
6
Indonesia is useful for my analysis for two primary reasons. Firstly, it is a majority Muslim country with the largest population of Muslims in the world, with over 87% of the population (approximately 203 million people) identifying as Muslim (Pew Forum on Religion & Public Life, 2009). Secondly, being near the equator it means that daylight hours are fairly constant over each year and hence each Ramadan, at around 13.5 hours of daylight per day. Since fasting in Ramadan depends on daylight hours, each cohort will have a very similar exposure to Ramadan and hence effect from fasting.
Years of Schooling- I have constructed the variable years of schooling using data from the highest
level of schooling completed and the highest grade of that school completed for an individual. Indonesian schooling includes Elementary School (age 6-11), Middle School (age 12-14), High School (age 15-18), and University. Net enrolment rates for elementary schooling in Indonesia was 90% in 2015, secondary schooling was 76%, and tertiary at 24% (UNESCO Institute for Statistics). There are Islamic equivalents for Elementary, Middle and High school, and vocational equivalents to Middle and High schools which I have included in the sample. 16% of schools follow the Islamic education system, which is overseen by the Ministry of Religious affairs, with the remaining 86% of mainstream schools overseen by the Ministry of National Education (The World Bank, 2014). I have excluded Adult Education institutions and Disabled schools since it is not possible to verify the length of study and at what age they started studying. I have also excluded any degree beyond Masters Level, known as Sarjana 3, which is the equivalent to a doctorate degree. This is due to the small number of observations (13 out of 32,000), and the fact that a doctorate takes a minimum of 3 years to complete with no maximum, which could lead to measurement error. Only individuals above the age of 15 answered questions pertaining to their highest level of schooling completed.
Income and Consumption- While an exact measure of income is not measured in the IFLS, it is
possible to calculate household and hence per capita expenditure. I use consumption and expenditure aggregates from the RAND Corporation’s website to adapt such measure from the IFLS4 in 2007 to the IFLS52. The types of household expenditures used to make the aggregates are: Food, Non-Food (frequently and non-frequently purchased), Education, and Housing. The aggregates do not contain deflators and hence are in nominal terms, but this should not be a great concern considering the fact I am using cross-sectional data and am not comparing expenditures over time3.
Cognitive Ability- An adaptive number sequence test was given to participants aged 15 and above,
where 6 questions were answered. The first 3 questions were compulsory for all participants, and the next 3 were given on the basis of how many of the first 3 they answered correctly. A score which allows for comparability between questions is included in the data set. The questions involve participants selecting the correct number that aligns with previous given numbers to create a sequence. The test was developed by Dr John McArdle, which involved extensive pre-tests in Indonesia and Mexico4.
Other Cognitive Assessments- There are a number of other cognitive tests used in the IFLS to
compliment the one above. Firstly, participants were given 8 cognitive questions which were an abridged version of the Raven’s test, and 5 questions on simple arithmetic, totalling 13 questions. The Raven’s test is used to estimate abstract reasoning and non-verbal form of fluid intelligence. All questions consist of identifying the missing piece of a visual geometric design. The benefit of the Raven’s test in the context of a developing country survey is its independence of language, reading, and writing skills (Raven, 2003). The tests are administered for those aged 15 and above, with those above 60 years old not receiving arithmetic
2 These aggregates can be downloaded here: http://www.rand.org/labor/FLS/IFLS/download.html 3
The expenditure aggregates do not use spatial deflators to obtain a real value between provinces. Further details and notes on the construction of these variables can be found in the document ‘Note on the Construction of the IFLS Consumption Expenditure Aggregates’, Witoelar (2009).
4
7
questions to avoid refusal of participation. Secondly, there is an immediate word recall test which involved participants repeating as many of 10 nouns as possible that were stated randomly by the interview. Finally, there is a delayed word recall given 4 to 5 minutes after the first, and contained the same words as before. Word recall tests are typically used as a proxy for short term memory in psychology and mental health (Paivio, 1973).
3.2 Ramadan Data
To construct the Ramadan exposure variables, I first converted historic Ramadan dates from the Islamic calendar to the Gregorian calendar, using a website hosted by the Institute of Oriental studies at the University of Zurich5. To calculate an individual’s exposure to Ramadan in utero, I checked if the month of Ramadan overlapped between their exact date of birth and 268 days before they were born, which is assumed to be the standard gestation length of 9 months (Jukic et al., 2013). I then further separate this identification into the following 5 categories which are mutually exclusive: Conceived during Ramadan, Ramadan starts in the first trimester of pregnancy (days 1-89 of gestation), Ramadan starts in the second trimester of pregnancy (days 90- 178 of gestation), Ramadan starts in the third trimester of pregnancy (days 179- 268 of gestation), and born during Ramadan. I excluded data from individuals who were conceived less than 21 days after the end of a Ramadan (calculated as birth date minus 268 days) because, if they were born post term (a birth that has extended beyond two weeks past the normal gestation period), their classification as to whether they were exposed to Ramadan in utero might be incorrect. This bandwidth of 21 days is therefore larger than necessary as it removes almost all children that were conceived in the festivities following Ramadan, who may have different characteristics to other children.
3.3 Specification
The basic specification I use for all analyses is an OLS model which controls for sex, age, age squared and month of birth, using robust standard errors. Age is measured in days from the date 1 January 2016 to provide a more precise estimate than years and remove any errors in participants stating their age incorrectly. Month of birth is important to control for since it is plausible that parents plan conception and birth around the academic calendar and other public holidays.
The model is as follows:
𝒀𝒊= 𝜶 + 𝜷𝑹𝒂𝒎𝒂𝒅𝒂𝒏𝑬𝒙𝒑𝒐𝒔𝒖𝒓𝒆𝒊+ 𝜸𝑺𝒆𝒙𝒊+ 𝜽𝟏𝑨𝒈𝒆𝒊+ 𝜽𝟐𝑨𝒈𝒆𝑺𝒒𝒖𝒂𝒓𝒆𝒅𝒊+ 𝝁𝑴𝒐𝒏𝒕𝒉𝒐𝒇𝑩𝒊𝒓𝒕𝒉𝒊+ 𝜺𝒊
Where 𝒀𝒊 is the outcome of an individual and 𝑹𝒂𝒎𝒂𝒅𝒂𝒏𝑬𝒙𝒑𝒐𝒔𝒖𝒓𝒆𝒊 is either the exposure to
Ramadan at any point during pregnancy, or one of the stages of pregnancy listed in section 3.2. Controls include: 𝑨𝒈𝒆𝒊 , 𝑨𝒈𝒆𝑺𝒒𝒖𝒂𝒓𝒆𝒅𝒊 , and 𝑴𝒐𝒏𝒕𝒉𝒐𝒇𝑩𝒊𝒓𝒕𝒉𝒊 for an individual, 𝒊.
The treatment group to identify the intention to treat (ITT) consists of Muslims who were exposed to Ramadan in utero, and the control group consists of Muslims who were not exposed to Ramadan in utero and hence born in a relative small gap of around 60 days each year. I find this control group to be more appropriate than using non-Muslims exposed to Ramadan due to the possibility that their socioeconomic status and ethnicity may differ to Muslims.
5 The calendar converter can be found here: http://www.oriold.uzh.ch/static/hegira.html (Institute of Oriental Studies at Zurich University, 2006)
8
An ITT analysis includes all randomised participants, regardless of the treatment they actually ended up receiving (misclassification in this case). It is considered to be more appropriate if outcome data is available for all randomised subjects. The difficulty of deviation from treatment (exposure to Ramadan) limits attrition bias (Gupta, 2011).
The estimated ITT from the above regression likely underestimates the average treatment effect (ATE) for a number of reasons. First, it is unknown which mothers have fasted during Ramadan and how long for, hence these estimates assume that all Muslim mothers fasted for the entire duration of Ramadan. Second, I do not know the religion of an individual’s mother, only their own religion. Therefore any changes in religion between generations have the potential to bias estimates, however the net effect of people joining minus leaving Islam may be minimal. Third, due to the heterogeneity of the length of pregnancy between women, there will be some misclassification, especially for prematurely born children who may incorrectly be classified as exposed. Misclassification can also happen between trimesters for shorter term pregnancies. Fourth, there is survival bias stemming from the fact that exposure to Ramadan could increase mortality rates, especially in utero and hence attrition will bias estimates towards zero. Therefore I limit my sample to below 60 year olds in case Ramadan exposure increases the likelihood of illness and mortality towards the end of life. Lastly, all individuals who have a missing date of birth have to be excluded from the analysis, which may also underestimate the true effect. Indeed, those with a missing date of birth have less years of schooling and lower per capita income than those with date of births available, possibly due to being a foundling or adopted.
Since Ramadan occurs around 10-11 days on the Gregorian calendar earlier every year, it takes around 33 years before Ramadan occurs on the same Gregorian date again. This shifting of Ramadan dates over the years effectively creates an exogenous sources of randomisation, where seasonal effects can be separated.
This strategy is valid if one assumes that individuals do not plan their births around Ramadan to avoid their child being exposed in utero. I check this by looking at the proportion of individuals exposed for each month of birth. If these proportions are similar for each month, then it can be assumed that individuals do not plan births around Ramadan, regardless of the fact that some months of birth may be more popular i.e. planned around school start dates. The proportion of individuals exposed to Ramadan with a sample of 15-60 year olds is 87.7%. The month with the lowest exposure rate is February, with 82.9%, and the highest is October, with 92.9% of individuals exposed to Ramadan in utero. There are no sudden jumps between months however, with a gradual increase between February and October. Therefore while birth planning may occur, it does not appear to be a behaviour that drastically interferes with Ramadan exposure rates, and may partially be explained by the frequency Ramadan occurs in a given month for the sample selected.
A further method of checking to see if birth planning is a concern is to see whether there are spikes in months of births 9 months after the end of a Ramadan. I check this by looking at the months of birth for a given year and cross-referencing this to the end of Ramadan in the preceding year. After looking at birth months for every decade from 1950-2000, I conclude that this is not a legitimate concern, and it appears the main spike in births from 1980 onwards is in July, which coincides with the start of the academic year. Finally, as mentioned in section 3.2, the exclusion of individuals who were conceived less than 21 days after the end of a Ramadan goes some way to remove data of children whose parents planned their birth around Ramadan.
9
4.1 Weight and Height
I begin my results by investigating if there is indeed a health channel that could influence results on socioeconomic outcomes. Weight and height are the primary outcome variables in van Ewijk (2012), thus using a similar empirical strategy and a later wave of the IFLS, the results should be somewhat comparable.
Tables 1 and 2 show that there is generally no significant impact of Ramadan exposure on the weight and height of Muslims, with the notable exception of females, where there appears to be a positive effect. These are however rather small, with an increase of around half a kilogram and a third of centimetre increase when exposed to Ramadan in utero. Results for males tend to be negative for both weight and height, but are not significant when looking at all stages of pregnancy.
A possible reason for the difference in my results presented is the absence of fixed effects, which may not be a suitable type of regression. This is due to the fact that height and weight differences cannot be explained within a household if members are not related to each other, which is likely if the sample contains over 18 year olds that often leave home.
Table 1: Effect of in utero Ramadan exposure on weight of Muslims
Muslims
(1) (2) (3) (4) (5) (6)
Gestation Period All Muslims Males Females
15-24 years old 25-39 years old 40-60 years old In Utero during Ramadan 0.169 -0.175 0.508* 0.584 -0.087 -0.014
(0.222) (0.332) (0.298) (0.403) (0.346) (0.421) Conceived in Ramadan 0.115 -0.380 0.571 0.287 0.256 -0.190 (0.313) (0.467) (0.421) (0.630) (0.508) (0.545) 1st Trimester -0.086 -0.675* 0.466 0.034 0.043 -0.446 (0.251) (0.372) (0.338) (0.488) (0.393) (0.460) 2nd Trimester 0.396 -0.134 0.887*** 0.890* 0.229 0.029 (0.253) (0.376) (0.340) (0.461) (0.408) (0.485) 3rd Trimester 0.050 -0.196 0.309 0.347 -0.499 0.584 (0.277) (0.417) (0.370) (0.531) (0.430) (0.529) Born in Ramadan -0.145 -0.345 0.077 0.22 -0.901* 0.946 (0.325) (0.494) (0.430) (0.585) (0.499) (0.633) N 24095 11186 12909 5814 10249 8032 Mean 57.4 59.3 55.7 52.7 58.6 59.2
Dependent Variable: Weight in kilograms
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation,
10
controlling for: age in days, age squared, and month of birth, with robust standard errors. Columns (1), (4), (5), and (6) also control for sex. Column (1) uses a sample of all Muslims aged 15-60. Standard errors in parentheses.
* p<0.1, ** p<0.05, *** p<0.01
Table 2: Effect of in utero Ramadan exposure on height of Muslims
Muslims
(1) (2) (3) (4) (5) (6)
Gestation Period All Muslims Males Females
15-24 years old 25-39 years old 40-60 years old In Utero during Ramadan 0.112 -0.139 0.338** 0.196 -0.176 0.261
(0.114) (0.172) (0.150) (0.233) (0.168) (0.208) Conceived in Ramadan -0.155 0.537** 0.192 0.025 -0.311 0.027 (0.160) (0.246) (0.209) (0.353) (0.249) (0.276) 1st Trimester 0.040 -0.363* 0.395** 0.418 -0.152 0.085 (0.127) (0.193) (0.167) (0.285) (0.188) (0.226) 2nd Trimester 0.202 -0.095 0.477*** 0.304 -0.021 0.455* (0.129) (0.195) (0.171) (0.265) (0.200) (0.239) 3rd Trimester 0.056 -0.058 0.167 0.239 -0.258 0.344 (0.144) (0.222) (0.186) (0.298) (0.217) (0.261) Born in Ramadan -0.005 -0.180 0.165 -0.091 -0.302 0.714** (0.168) (0.267) (0.212) (0.321) (0.255) (0.311) N 24138 11205 12933 5827 10263 8048 Mean 156.8 163.0 151.4 157.6 157.2 155.6
Dependent Variable: Height in centimetres
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: age in days, age squared, and month of birth, with robust standard errors. Columns (1), (4), (5), and (6) also control for sex. Column (1) uses a sample of all Muslims aged 15-60. Standard errors in parentheses.
* p<0.1, ** p<0.05, *** p<0.01
11
Table 3 presents the effect of Ramadan exposure in Utero on the number of years of schooling an individual completes, separating the effects into 5 further stages of pregnancy. The full sample consists of 15-60 year old Muslims. The lower bound of 15 years old for this sample is justified by the fact that individuals younger than 15 did not answer questions on their highest level of education completed. The upper bound of 60 years old is justified by the fact there may be survival bias for higher ages, so it is helpful to limit this potential bias by lowering the age of the sample.
For the full sample I find that there is no significant effect, although there is a suggestion of a negative impact with all signs being negative. The only subsamples with significant effects regardless of stage of pregnancy is for those that were 15-24 years old at the time of interview. Negative effects are also found in certain stages for males and those aged 25-39. The size of these effects are quite large, ranging from 0.1 to 0.4 years less education, depending on the stage of pregnancy that the individual was exposed. This compares to overall sample average of 8.8 years of education. There does not appear to be any trend as to which stage of pregnancy has the most significant or largest effect, except it is noted that there are no significant effects when conceived in Ramadan. The fact that males are affected by exposure to Ramadan whereas females are not, could partially be explained by the fact that males received 8.9 years of education compared to 8.7 for females in the sample. There is an anomaly in finding that 40-60 year olds born in Ramadan receive 0.75 more years of education- a very large magnitude considering the fact that other stages of pregnancy have no effect. The lack of effect for this age group compared to under 40 year olds is supported by the column for those who haven’t completed High School. Those under 40 have on average 9.6 years of education compared with just 7.4 for those over 40. This sharp rise in enrolment in Indonesia seems to have meant the younger and more educated generation stand to lose the most in terms of education from exposure to Ramadan in utero.
Table 4 shows the same analysis, however with a non-Muslim sample. No results are found to be significant with the exception of a large and positive effect of in utero Ramadan exposure for 15-24 year olds in the first trimester. The results in general appear to be positive, however cannot be distinguished from zero. Non-Muslims on average receive 0.7 years more education than Muslims in the sample used.
12
Table 3: Effect of in utero Ramadan exposure on years of education of Muslims
Muslims
(1) (2) (3) (4) (5) (6)
Gestation Period
All
Muslims Males Females
15-24 years old 25-39 years old 40-60 years old In Utero during Ramadan -0.015 -0.193 0.134 -0.318*** -0.088 0.045
(0.083 (0.124) (0.110) (0.116) (0.125) (0.169) Conceived in Ramadan -0.011 -0.173 0.121 -0.143 0.135 0.019 (0.120) (0.182) (0.158) (0.191) (0.183) (0.226) 1st Trimester -0.121 -0.326** 0.064 -0.201 0.046 -0.215 (0.094) (0.141) (0.125) (0.138) (0.144) (0.184) 2nd Trimester -0.030 -0.234* 0.156 -0.440*** -0.123 0.174 (0.094) (0.142) (0.125) (0.136) (0.149) (0.194) 3rd Trimester -0.093 -0.217 0.009 -0.178 -0.276* 0.092 (0.103) (0.156) (0.137) (0.169) (0.156) (0.208) Born in Ramadan -0.054 -0.137 0.017 -0.240* -0.375** 0.750*** (0.116) (0.175) (0.153) (0.145) (0.178) (0.253) N 18949 8631 10318 4150 8061 6738 Mean 8.8 8.9 8.7 9.9 9.4 7.4
Dependent Variable: Years of education
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: age in days, age squared, and month of birth, with robust standard errors. Columns (1), (4), (5), and (6) also control for sex. Column (1) uses a sample of all Muslims aged 15-60. Standard errors in parentheses.
13
Table 4: Effect of in utero Ramadan exposure on years of education of non-Muslims
Non-Muslims
(1) (2) (3) (4) (5) (6)
Gestation Period
All
Non-Muslims Males Females
15-24 years old 25-39 years old 40-60 years old
In Utero during Ramadan 0.170 0.163 0.241 0.290 0.336 -0.380
(0.263) (0.332) (0.427) (0.450) (0.396) (0.480) Conceived in Ramadan 0.360 0.096 0.908 0.815 0.921 -0.167 (0.394) (0.503) (0.634) (0.723) (0.594) (0.745) 1st Trimester 0.041 0.238 -0.243 1.107* 0.185 -0.630 (0.297) (0.374) (0.483) (0.580) (0.436) (0.540) 2nd Trimester 0.147 -0.086 0.548 0.340 0.345 -0.399 (0.299) (0.381) (0.478) (0.504) (0.469) (0.552) 3rd Trimester 0.117 0.274 -0.118 -0.476 0.553 -0.584 (0.322) (0.408) (0.517) (0.596) (0.488) (0.577) Born in Ramadan 0.201 0.245 0.265 0.291 -0.115 1.160 (0.372) (0.466) (0.614) (0.545) (0.566) (0.729) N 2297 1369 928 447 1047 803 Mean 9.5 9.3 9.7 9.8 9.9 8.6
Dependent Variable: Years of education
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: age in days, age squared, and month of birth, with robust standard errors. Columns (1), (4), (5), and (6) also control for sex. Column (1) uses a sample of all non-Muslims aged 15-60. Standard errors in parentheses.
* p<0.1, ** p<0.05, *** p<0.01
4.3 Income
Table 5 shows the impact of in utero Ramadan exposure on log per capita income of Muslims, using a sample different to the analyses above. The sample uses only individuals who responded as being the head of a household and were 25-60 years old. This is because household consumption and income was calculated at the household level, hence the income earners needed to be identified to avoid picking up observations that did not contribute to household income. The lower bound of 25 years old in the sample is justified by assuming that individuals are capable of working a full time job at this age, rather than studying or training. Furthermore, self-reported wages often fail to identify those who don’t earn a wage i.e. entrepreneurs, so
14
while this is not a perfect way of identifying earners in a household, it can serve as a good alternative to self-reported wages, which can be inaccurate. The sample is split into quintiles based on the whole survey sample, hence the difference in the number of observations per quintile. All regressions control for years of education, in addition to the controls listed in the initial specification.
There is no effect of in utero Ramadan exposure when considering the first four quintiles, with the exception of an increase of 1.9% in income when conceived in Ramadan for the 3rd quintile. There is however a significant effect for those in the 5th quintile, equivalent to a decrease of 6.4% in income when in utero during Ramadan. There are also significant effects for the first trimester and being born in Ramadan, resulting in decreases of 7.8% and 10.4% in income respectively.
Table 5: Effect of in utero Ramadan exposure on log per capita income of Muslims
Muslims (1) (2) (3) (4) (5) (6) Gestation Period All Muslims 1st Quintile 2nd Quintile 3rd Quintile 4th Quintile 5th Quintile
In Utero during Ramadan -0.022 0.001 0.000 0.005 -0.005 -0.064*
(0.025) (0.024) (0.009) (0.007) (0.01) (0.035) Conceived in Ramadan 0.029 -0.018 -0.006 0.019* -0.001 -0.035 (0.034) (0.037) (0.013) (0.011) (0.014) (0.048) 1st Trimester -0.020 -0.006 0.003 0.006 -0.011 -0.078** (0.027) (0.026) (0.01) (0.008) (0.011) (0.038) 2nd Trimester -0.045 -0.002 -0.007 0.001 -0.001 -0.058 (0.028) (0.026) (0.010) (0.008) (0.011) (0.039) 3rd Trimester -0.017 0.016 0.008 0.005 -0.003 -0.049 (0.030) (0.028) (0.011) (0.009) (0.013) (0.042) Born in Ramadan -0.029 0.022 -0.004 -0.002 -0.006 -0.104** (0.035) (0.032) (0.013) (0.012) (0.015) (0.046) N 6407 1207 1288 1328 1240 1344
Dependent Variable: Log per capita income
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: years of education, age in days, age squared, and month of birth, with robust standard errors. Column (1) uses a sample of all Muslims aged 25-60 who are heads of households. Standard errors in parentheses. Quintiles constructed for the whole survey population, with the 1st quintile the poorest and the 5th quintile the richest.
15
In order to gain a better understanding of those in the fifth quintile and how they may be affected by in utero Ramadan exposure, I examine the types of occupations that these individuals have. There are larger numbers in occupations such as professional, managerial, clerical and service work in this quintile compared to lower quintiles. These jobs tend to require more skill and education than agricultural or production workers, where the fifth quintile have lower numbers in these occupations. This compliments the earlier finding that younger and more educated individuals are affected the most by in utero Ramadan exposure.
Table 6: Distribution of occupations per quintile of income for Muslims
Muslims Quintile
Occupation 1 2 3 4 5 Total
Professional/Technical 1.5% 2.3% 4.5% 6.7% 9.9% 5.1%
(20) (34) (69) (102) (158) (383)
Administrative and Managerial 0.2% 0.3% 0.2% 0.6% 1.5% 0.6%
(2) (5) (3) (9) (24) (43)
Clerical and Related Workers 1.2% 2.6% 4.4% 6.0% 11.6% 5.3%
(16) (38) (68) (92) (184) (398) Sales Workers 11.4% 14.2% 15.2% 21.6% 21.5% 17.0% (152) (209) (233) (330) (341) (1265) Service Workers 8.1% 10.9% 13.3% 15.1% 17.2% 13.1% (108) (160) (205) (230) (274) (977) Agricultural Workers 44.6% 35.4% 27.6% 19.8% 12.8% 27.4% (596) (520) (425) (302) (203) (2046)
Production and Labourers 33.0% 34.3% 34.8% 30.2% 25.5% 31.5%
(441) (505) (535) (461) (406) (2348)
Total 100% 100% 100% 100% 100% 100%
(1335) (1471) (1538) (1526) (1590) (7460)
Notes: Quintiles constructed for the whole survey population, with the 1st quintile the poorest and the 5th quintile the richest. Those in the table are all Muslims aged 25-60 who are heads of households. Percentages are per column. Absolute numbers are in parentheses.
Table 7 uses the same analysis as table 5 but with a non-Muslim population. The results do not tend to show any association between in utero Ramadan exposure and log per capita income, with the exception of the second quintile. The reduction of 4.1% in income for the second quintile is considerably smaller than the 6.4% found for the fifth quintile of the Muslim sample.
17
Table 7: Effect of in utero Ramadan exposure on log per capita income for non-Muslims
Non-Muslims (1) (2) (3) (4) (5) (6) Gestation Period All Non-Muslims 1st Quintile 2nd Quintile 3rd Quintile 4th Quintile 5th Quintile In Utero during Ramadan -0.085 -0.041 -0.041* -0.004 0.012 -0.048
(0.061) (0.091) (0.023) (0.022) (0.022) (0.09) Conceived in Ramadan 0.005 -0.114 -0.054 0.006 -0.032 0.085 (0.096) (0.102) (0.036) (0.035) (0.032) (0.131) 1st Trimester -0.109 -0.042 -0.039 -0.024 0.033 -0.149 (0.068) (0.099) (0.026) (0.025) (0.028) (0.098) 2nd Trimester -0.110 -0.008 -0.043 0.013 0.025 -0.018 (0.073) (0.098) (0.028) (0.025) (0.029) (0.100) 3rd Trimester -0.051 -0.008 -0.045 0.015 -0.008 0.027 (0.079) (0.104) (0.029) (0.026) (0.029) (0.111) Born in Ramadan -0.087 -0.118 -0.021 -0.034 -0.001 -0.093 (0.092) (0.134) (0.036) (0.031) (0.035) (0.106) N 805 132 155 153 168 197
Dependent Variable: Log per capita income
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: years of education, age in days, age squared, and month of birth, with robust standard errors. Column (1) uses a sample of all non-Muslims aged 25-60 who are heads of households. Standard errors in parentheses. Quintiles constructed for the whole survey population, with the 1st quintile the poorest and the 5th quintile the richest.
* p<0.1, ** p<0.05, *** p<0.01
Table 8 is comparable to table 6, with a non-Muslim sample. The findings are similar, and focussing specifically at the second quintile, there is a large majority of individuals working in agriculture and production.
18
Table 8: Distribution of occupations per quintile of income for non-Muslims
Non-Muslims Quintile
Occupation 1 2 3 4 5 Total
Professional/Technical 2.5% 1.7% 2.2% 3.0% 13.6% 5.2%
(4) (3) (4) (6) (33) (50)
Administrative and Managerial 0.0% 0.6% 0.0% 0.5% 2.9% 0.9%
(0) (1) (0) (1) (7) (9)
Clerical and Related Workers 1.9% 1.7% 5.9% 12.1% 9.1% 6.6%
(3) (3) (11) (24) (22) (63) Sales Workers 7.6% 11.9% 16.7% 17.2% 22.3% 15.8% (12) (21) (31) (34) (54) (152) Service Workers 7.6% 13.1% 14.0% 16.7% 14.1% 13.3% (12) (23) (26) (33) (34) (128) Agricultural Workers 51.9% 33.0% 27.4% 20.2% 13.6% 27.5% (82) (58) (51) (40) (33) (264)
Production and Labourers 28.5% 38.1% 33.9% 30.3% 24.4% 30.6%
(45) (67) (63) (60) (59) (294)
Total 100% 100% 100% 100% 100% 100%
(158) (176) (186) (198) (242) (960)
Notes: Quintiles constructed for the whole survey population, with the 1st quintile the poorest and the 5th quintile the richest. Those in the table are all non-Muslims aged 25-60 who are heads of households. Percentages are per column, absolute numbers are in parentheses.
4.4 Cognitive Ability
Cognitive ability results are explored as a potential mechanism that is driving the negative impact of in utero Ramadan exposure on years of education and income. While health in terms of height and weight were not seen to be significantly affected, it is plausible that fasting in Ramadan could have an effect on a child’s brain development. This effect on brain development could then prevent a child from attaining more years of education and in turn lower their earnings potential.
In table 9, I estimate the effect of in utero Ramadan exposure on Muslims’ adaptive number series test score. The results are in line with my findings on height, weight and education, where males and those
19
under 40 are negatively affected by in utero Ramadan exposure. The effect here however is rather small, with every negative result being less than 1% of the average.
Results of the other cognitive ability measures, including the Raven’s test and word recall were not significantly affected by in utero Ramadan exposure. Therefore, the channel of cognitive ability to explain earlier results does not appear to be robust.
Table 9: Effect of in utero Ramadan exposure on adaptive number series test score for Muslims
Muslims
(1) (2) (3) (4) (5) (6)
Gestation Period
All
Muslims Males Females
15-24 years old 25-39 years old 40-60 years old In Utero during Ramadan -0.789 -2.849* 1.026 -1.386 -1.999 2.064
(1.220) (1.636) (1.785) (2.257) (1.716) (2.548) Conceived in Ramadan 0.618 -0.765 1.934 -1.319 1.924 1.888 (1.777) (2.495) (2.512) (3.556) (2.573) (3.466) 1st Trimester -1.396 -2.322 -0.340 -1.411 -1.006 -0.712 (1.393) (1.885) (2.024) (2.841) (1.970) (2.801) 2nd Trimester -1.086 4.115** 1.753 -2.807 -3.254 2.184 (1.390) (1.882) (2.022) (2.537) (2.066) (2.972) 3rd Trimester -0.94 -3.960* 1.592 -1.689 -4.474** 5.195 (1.528) (2.100) (2.200) (2.903) (2.184) (3.166) Born in Ramadan 0.374 -2.845 3.031 -1.302 -2.743 8.336** (1.703) (2.385) (2.413) (2.896) (2.476) (3.701) N 24258 11262 12996 5881 10306 8071 Mean 513.9 519.8 508.8 530.7 521.2 492.3
Dependent Variable: Adaptive number series test score
Notes: Two separate regressions are run for each column- one for the overall exposure to Ramadan (in utero) and one that separates by the stage of pregnancy. All regressions use ordinary least squares estimation, controlling for: age in days, age squared, and month of birth, with robust standard errors. Columns (1), (4), (5), and (6) also control for sex. Column (1) uses a sample of all Muslims aged 15-60. Standard errors in parentheses.
20
5. Discussion
This paper has shown that Muslims that are male and under the age of 40 can be negatively affected by in utero Ramadan exposure in terms of their educational attainment, receiving up to 0.4 years less education than Muslims not exposed to Ramadan in utero, depending on the stage of pregnancy when Ramadan fell. Furthermore, individuals in the top quintile of earnings earn around 7% less income when exposed to Ramadan in utero, depending on the stage of pregnancy. Subsamples with negative effects tend to be more educated, indicating that exposure to Ramadan tends not to have an impact on socioeconomic outcomes when individuals were not able to reach higher levels of education so easily.
The mechanism that is driving these results however remains unclear. Exploring the physical health channel, there seems to be little effect of in utero Ramadan exposure on height and weight, contradictory to the previous literature that uses a similar empirical strategy to this paper. Males are however the most negatively affected, in line with the results on education and income. When looking at the mental health channel, males and young Muslims are the most affected, however the effect is small and not particularly robust when considering several measures of cognitive ability. This suggests that exposure to Ramadan in utero may in some way slow brain development, which would result in less academic ability and hence lower earnings (even when controlling for years of education). Further analysis could also investigate the channel of access to healthcare. It is worth noting that all estimates likely underestimate the average treatment effect, primarily as not all Muslim women fast during Ramadan.
One interesting result is that there are significant negative results found for several outcomes for those born in Ramadan (for specific subsamples), including height, years of education and income. A possible explanation is that those born in Ramadan may be born in hospitals with considerably less resources if staff such as doctors and nurses are working less due to fasting. This may be further exacerbated during the celebration of Eid al-Fitr, where staff may be allowed holidays6. Therefore, those with serious health conditions may not be treated as well or as quickly as those not born in Ramadan, leading to long term problems. This however is not supported by the fact that non-Muslims are not significantly affected by being born in Ramadan. Although the sample size is considerably smaller than that of Muslims, the results for Muslims would only be explained by the assumption that Muslims are treated by Muslims at non-Muslim hospitals, which is unlikely considering the huge majority of non-Muslims in Indonesia.
The heterogeneity of results between stages of pregnancy, subsamples and outcomes (in the case of cognitive ability) is likely to be attributed to the estimation technique, which can result in an underestimation of the ATE. The estimation is lacking information on whether mothers fasted, for how long, and the length of gestation. Therefore while this may not be a large issue on a sample in the tens of thousands, the relatively small effects of in utero Ramadan exposure may not be shown to be significant with a smaller sample or subsample with only hundreds of observations. I have shown however that birth planning is not a large concern with this estimation technique, and the use of up to 45 cohorts in my sample helps remove any seasonal effects.
Exploring the reasons of why mothers would fast in Ramadan in the first place highlight two important points. Firstly, mothers and Muslims as a whole generally do not believe that fasting while pregnant is harmful to either themselves or their foetus. Secondly, participation in this behaviour can be explained in economic terms using the club good model, where costly behaviour increases participation rates in a religion to remove the free rider problem. These two points are interlinked, as educating pregnant Muslims about the potential health risks of fasting should raise cost of engaging in fasting from their perspective, whereby the utility of not fasting is higher than that of fasting. This is likely to be more effective
6
21
in countries with lower numbers of Muslims, since the environment is likely to support those not fasting more i.e. more help from friends/colleagues and shops do not close.
A potential implication from my research is that fasting in Ramadan needs to be considered when designing future medical and economic studies. The strategy I have employed to measure the long term effects in this paper is a good estimation with limited information, however is far from perfect and can be fraught with significant bias, especially with a smaller sample size. In an ideal scenario, studies will monitor whether a mother has fasted in Ramadan and for how long, tracking the health and economic outcomes of their child over time. This however would be difficult to include in the IFLS since individuals are only surveyed every 7 years, whereas Ramadan is an annual event. Even with an annual survey, it is difficult to enforce measurement of fasting accurately since people may forget to record this information or not record it accurately. Assuming these measurements were to be included in a study immediately and accurately, it would take at least 20 years to find anything meaningful on socioeconomic outcomes or adult health.
The policy implications are difficult to pinpoint, since the cause of such behaviour is through religion, and regulation on this is likely to infringe on human rights and difficult to enforce. Furthermore, the huge variance in magnitude and significance of the literature surrounding long term effects of fasting in pregnancy mean that medical professionals are not able to give definite advice to mothers who are considering fasting. This is also contingent on the fact that pregnant mothers will visit a doctor, which may not be the case in a developing country context. Education therefore may be best targeted towards religious communities and ensuring that pregnant women are adequately supported during Ramadan to understand that there are potential long term consequences of fasting, particularly in developing countries. Since Indonesia is considered to interpret Islam in a less austere way than the Middle East, Indonesians may be more open to this type of education than other nationalities.
There are other effects of Ramadan that I feel are worth exploring for future research on education and income. Rather than the channel of exposure to Ramadan fasting in utero, there exists a channel of reduced educational ability and earnings through the physiological effects of fasting when studying or working. Educational attainment while fasting may be hampered, especially when Muslim schools can often close cafeterias, making it difficult to not fast in Ramadan. Furthermore, in an increasingly international labour and educational market, Muslims could be significantly disadvantaged compared to non-Muslims, particularly in non-Muslim countries7. Campante and Yanagizawa-Drott (2015) show that Ramadan fasting has a negative impact on output growth of Muslim countries, indicating that microeconomic studies on education and individual earnings are likely to find negative results. The paper also shows that subjective well-being increases amongst Muslims while observing Ramadan, consistent with the club good model of Iannaccone (1992). This provides further suggestion that advice to reduce fasting behaviour, for medical or economic reasons, is unlikely to be politically popular.
7 Schools in England have expressed their concerns of children fasting during exam season:
http://www.telegraph.co.uk/education/2016/06/03/schools-say-muslim-students-should-break-ramadan-fast-to-avoid-b/
22
References
A.M. Jukic, D. B. (2013). Length of human pregnancy and contributors to its natural variation. Human
Reproduction, 28(10): 2848–2855.
Aaron Lechtig, J.-P. H. (1975). Effect of Food Supplementation During Pregnancy on Birthweight. American
Academy of Pediatrics, 56(4):508-20.
Azizi F, S. H. (2004). Intellectual development of children born of mothers who fasted in Ramadan during pregnancy. International Journal of Vitamin and Nutrition Research, 74: 374-80.
Brian P. McEvoy, P. M. (2009). Genetics of human height. Economics & Human Biology, 7(3):294-306. Butt, S. (2017, July 3). Absence from duty on Eid: Eight doctors, 14 nurses of DHQ hospital suspended.
Retrieved from Dawn: https://www.dawn.com/news/1342870
Douglas Almond, B. M. (2015). In Utero Ramadan Exposure and Children's Academic Performance. The
Economic Journal, 125 (December), 1501–1533.
Espinoza, J. (2016, June 3). Schools say Muslim students 'should break Ramadan fast' to avoid bad grades. Retrieved from The Telegraph: http://www.telegraph.co.uk/education/2016/06/03/schools-say-muslim-students-should-break-ramadan-fast-to-avoid-b/
Gatrad, A. R. (1994). Attitudes and beliefs of Muslim mothers towards pregnancy and infancy. Archives of
Disease in Childhood, 71: 170-174.
Gupta, S. K. (2011). Intention-to-treat concept: A review. Perspectives in Clinical Research, 2(3): 109–112. Heckman, J. J. (2000). Policies to foster human capital. Research in Economics, 3-56.
Iannaccone, L. R. (1992). Sacrifice and Stigma: Reducing Free-riding in Cults, Communes, and Other Collectives. Journal of Political Economy, Vol. 100, No. 2, pp. 271-291.
Institute of Oriental Studies at Zurich University. (2006, May 26). Conversion of Islamic and Christian dates. Retrieved from http://www.oriold.uzh.ch/static/hegira.html
Islamic Association of Raleigh. (n.d.). Rules of the Fasting of Ramadan. Retrieved from https://raleighmasjid.org/imam/fiqh-ramadan.html
J Joosoph, J. A. (2004). A survey of fasting during pregnancy. Singapore Med J, Vol 45(12) : 583-586. M.B. Meyer, J. T. (1977). Maternal smoking, pregnancy complications, and perinatal mortality. American
Journal of Obstetrics and Gynecology, 494-502.
Mazumder, D. A. (2011). Health Capital and the Prenatal Environment: The Effect of Ramadan Observance During Pregnancy. American Economic Journal: Applied Economics, 3: 56–85.
Metzger BE, R. V. (1982). "Accelerated Starvation" and the Skipped Breakfast in Late Normal Pregnancy. The
Lancet, 13;1(8272):588-92.
Mills JL, G. B. (1984). Maternal alcohol consumption and birth weight. How much drinking during pregnancy is safe? Journal of the American Medical Association, 252:1875-1879.
Najimudeen, M. (2016). Pregnancy and Ramadhan Fasting. British Journal of Medicine & Medical Research, 16(12): 1-14.
23
Nasrollahi, M. A. (2001). Interrelation of Ramadan Fasting and Birth Weight. Medical Journal of Islamic
Academy of Sciences, 14:3, 91-95.
Paivio, A. (1973). Picture superiority in free recall: Imagery or dual coding? Cognitive Psychology, 5, 176-206. Patrinos, G. P. (2004). Returns to investment in education: a further update. Education Economics, 111-134. Pew Forum on Religion & Public Life. (2009). Mapping the Global Muslim Population: A Report on the Size
and Distribution of the World's Muslim Population. Washington D.C.
RAND Corporation. (n.d.). IFLS Data and Documentation. Retrieved from http://www.rand.org/labor/FLS/IFLS/download.html
Raven, J. R. (2003). Rave Progressive Matrices. In R. S. McCallum, Handbook of Nonverbal Assessment (p. 390). Springer.
Reyn J. G. van Ewijk, R. C. (2013). Associations of Prenatal Exposure to Ramadan with Small Stature and Thinness. American Journal of Epidemiology, 177(8):729–736.
Salleh, H. (1989). Ramadan fasting among pregnant women in Muar district, Malaysia and its association to health outcomes. Malaysian Journal of Reproductive Health, 7(1):69-83.
The World Bank. (2014, September 1). World Bank and Education in Indonesia. Retrieved from
http://www.worldbank.org/en/country/indonesia/brief/world-bank-and-education-in-indonesia United Nations Educational, S. a. (n.d.). Retrieved from UNESCO Institute for Statistics:
http://uis.unesco.org/country/ID
Witoelar, F. (2009). Note on the Construction of the IFLS Consumption Expenditure Aggregates. The World Bank.
Yanagizawa-Drott, F. C. (2015). Does Religion Affect Economic Growth and Happiness? Evidence from Ramadan. The Quarterly Journal of Economics, 130 (2): 615-658.
