Dose birthweight matter to quality of life? A comparison between Japan, the U.S., and India

Yamane, Chisako; Tsutsui, Yoshirō Article Dose birthweight matter to quality of life? A comparison between Japan, the U.S., and India Health Economics Review Provided in Cooperation with: Springer Nature Suggested Citation: Yamane, Chisako; Tsutsui, Yoshirō (2022) : Dose birthweight matter to quality of life? A comparison between Japan, the U.S., and India, Health Economics Review, ISSN 2191-1991, Springer, Heidelberg, Vol. 12, Iss. 1, pp. 1-25, https://doi.org/10.1186/s13561-022-00393-9 This Version is available at: https://hdl.handle.net/10419/285285 Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated licence. https://creativecommons.org/licenses/by/4.0/ Yamaneand Tsutsui Health Economics Review (2022) 12:48 https://doi.org/10.1186/s13561-022-00393-9 RESEARCH Dose birthweight matter toquality oflife? Acomparison betweenJapan, theU.S., andIndia Chisako Yamane1* and Yoshiro Tsutsui2 Abstract Background: Birthweight is a widely accepted indicator of infant health and has significant and lasting associations. Several studies have found that low and high birthweight have significant negative associations with adult health. A new study in the field of social sciences has established that birthweight has significant negative associations with not only adult health but also social attributes, such as income and occupation; however, no studies have evaluated the associations between birthweight and quality-of-life (QOL) attributes such as happiness. Methods: In this study, we use data from Japan, the U.S., and India, collected in 2011, in which the respondents were asked about their own birthweights to examine the long-term associations between low and high birthweight and eight outcome variables related to the QOL: adolescent academic performance, height, education, marital status, body mass index, income, health, and happiness. We regressed each of the eight outcome variables on low and high birthweight and the interaction terms of the old age and the birthweight dummies for each country. We estimated both the reduced and the recursive-structural forms. While the former estimates the total, that is, the sum of direct and indirect associations between birthweight and each outcome, the latter reports the direct association between birthweight and each outcome. Results: In Japan, while low birthweight is negatively associated with all outcomes, the associations of high birthweight were limited. In the U.S., low birthweight was not associated with any outcomes, but high birthweight had significantly negative associations with health and happiness. In contrast, in India, high birthweight was significantly and positively associated with income, health, and happiness, while low birthweight was associated with several outcomes negatively, similar to Japan. These associations were stronger in youth than in old age. Conclusion: Our study demonstrated that the associations of birthweight with QOL are widely diversified across countries: low birthweight, rather than high birthweight, is a problem in Japan and India. However, the opposite is true for the U.S., indicating that policymakers in developed countries must pay closer attention to the problems caused by high birthweight, whereas those in developing countries are better to focus on low birthweight. Keywords: Low birthweight, Hight birthweight, Long-term associations of birthweight, Japan, the U.S., India © The Author(s) 2022. Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article’s Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article’s Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/. The Creative Commons Public Domain Dedication waiver (http:// creat iveco mmons. org/ publi cdoma in/ zero/1. 0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Introduction This study investigates whether birthweight matters to quality of life (QOL). Studies of birthweight are classified into those in the domains of medicine and social science, each of which is classified into those investigating the causes and consequences of birthweight. In the field of social sciences, researchers are interested in whether Open Access *Correspondence: [email protected] 1 Faculty of Economics, Hiroshima University of Economics, 5-37-1, Gion, Asaminami, Hiroshima-shi, Hiroshima 731-0192, Japan Full list of author information is available at the end of the article Page 2 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 inequality—which is measured by social attributes, such as income, education, and occupation—is passed from generation to generation, and therefore they focused on low birthweight, which might be a link in the transmission of poverty [1]. For example, with regard to the causes, it is known that parents’ education level, income level, and employment types are critical factors [2], while considering the consequences, low birthweight was found to have a negative association with grades of mathematics, as well as with health and employment in adulthood [3]. The present study falls in the social science domain, and investigates the associations of birthweight with QOL in adulthood. Compared with previous studies, this study has a merit of investigating various outcomes including not only academic performance, height, and education in adolescence, but also marital status, body mass index (BMI), income, health, and happiness in adulthood. Among these, the associations with happiness have not been reported in literature. Furthermore, this study is unique as it estimates not only reducedform for each outcome separately, as done in most previous studies, but also the recursive and structural forms considering the simultaneity of the outcomes. While the reduced form estimates the total association, i.e., the sum of direct and indirect associations, between birthweight and each outcome, the latter reports the direct association between birthweight and each outcome.1 The second contribution of this study is that it analyzes data from three distinct countries—Japan, the U.S., and India. Though there have been many studies investigating QOL’s link with birthweight in the U.S., in Japan, there have been no substantial studies except [4, 5]. Furthermore, in India, there have been no studies, to our knowledge, which investigated the associations between birthweight and the long-term socio-economic consequences in adulthood [refer to Additionalfile1: Supplementary material A for a detailed review]. As shown in Table1, the distributions of birthweight in our data are quite different among these countries.2 India is one of the countries where the percentage of low-birthweight births is the highest in the world.3 In contrast, the share of the high-birthweight births was 15% in the U.S., which was much higher than that of Japan (1.3%) and India (0%). These facts suggest that QOL’s link with low birthweight may differ across countries, and that not only low birthweight, but also high birthweight might be related to QOL. The latter suggestion is consistent with the knowledge in obstetrics that high birthweight is associated with various dysfunctions, such as an increase in frequency of labor dystocia and consequently an increase in Caesarean section rates [8, 9]. High birthweight is caused by maternal obesity and gestational diabetes, which exposes the fetus to elevated levels of fuels such as glucose and fatty acids throughout gestation [10, 11]. Considering that overweight is a major health-related risk factor Table 1 Distribution of birthweights SBW represents standard birthweight Birthweight (kg) Japan U.S. India Number of observations Rate Number of observations Rate Number of observations Rate (%) (%) (%) LBW < 2.5 302 6.23 LBW < 2.5 309 6.17 LBW < 2.5 128 12.3 SBW 2.5–2.999 1364 28.1 SBW 2.5–2.999 1371 27.4 SBW 2.5–2.999 266 25.7 3.0–3.499 1629 33.6 3.0–3.999 1474 29.4 3.0–3.499 209 20.2 3.5–3.999 307 6.33 HBW 4.0–4.499 632 12.6 Q_HBW 3.5–3.999 17 1.64 HBW 4.0–4.499 58 1.2 4.0–4.499 0 0 4.5 or more 3 0.06 V_HBW 4.5 or more 145 2.89 4.5 or more 0 0 Do not know 1187 24.5 Do not know 1079 21.5 Do not know 417 40.2 Total 4850 100 5010 100 1037 100 1 In addition, estimation by recursive and structural forms report the dependency between outcomes. 2 For example, according to our survey data, the rate of low birthweight was 12% in India, while it was 6% in Japan and the U.S. 3 According to [6], the percentage of low-birthweight infants in India was one of the highest globally at 30% similar to Yemen and Sudan. It states that “India alone accounts for 40% of low birthweight births in the developing world and more than half of those in Asia. There are more than 1 million infants born with low birthweight in China and nearly 8 million in India.” Using a representative sample of 60 low- and middle-income countries, [7] investigated the ratio of people whose BMI is lower than 16, and reported that India was the highest at 6.2% in the world; Bangladesh was the secondhighest. Though the ratio of Bangladesh has been declining by 0.52% per annum, the decline in India has been only around 0.11% per annum. Page 3 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 in adulthood,4 high birthweight has a risk of reducing QOL. In this context, the third contribution of this study is to investigate how high birthweight, in addition to low birthweight, influences QOL in adulthood, which has been, to our knowledge, seldom investigated, except for the associations with health. The rest of this study is organized as follows: In the methods section, expositions of outline of survey, definitions of variables, hypotheses of this study, and estimation methods are given. In the results section, the estimation results of the reduced form, those of recursive and structural forms, and those by full information maximum likelihood (FIML) method are presented. In the discussion section, we discuss the innovation of our results and link them to those from previous studies. In the conclusion section, we argue for the uniqueness of this study, mention its limitations, and review possible future work. Methods Outline ofthesurvey This study uses the data of Japan, the U.S., and India for 2011 from the Household Panel Survey on Consumer Preferences and Satisfaction (JHPS-CPS) conducted by Osaka University.5 Respondents were not asked whether they were singletons or twins in the survey. In Japan, the survey was executed by Central Research Services Inc., a company with extensive experience in academic research and government contracts. Double stratified random sampling from the entire population was used to create a representative sample with respect to sex, age, and living locations. Enumerators visited the homes of the selected respondents to administer the questionnaires. The completed questionnaires were collected after several days. Meanwhile, in the U.S., the survey was conducted by a large survey company, TNS Custom Research, which mailed an English translation of the same questionnaire used in Japan to residents, randomly selected from its pool of registered respondents using the census divisions of the U.S., except for Alaska and Hawaii. In India, Nikkei Research Inc. implemented the survey in six major cities (Delhi, Mumbai, Bangalore, Chennai, Kolkata, and Hyderabad). Each city is separated into four areas; in each area, 15 points were chosen from which five respondents were selected considering their sex, age, and social class (Socio-Economic-Classification; SEC),6 and were interviewed because many people with no education were involved. Accordingly, respondents from the poorest to the richest strata are included in the survey. The questions asked in the three countries are nearly the same. Most of the questions in Japanese and English were checked by a bilingual researcher to ensure consistency. In India, the questionnaires in five local languages, Bengali, Hindi, Kannada, Tamil, and Telugu, were used at the interviews in addition to the questionnaire in English. The number of surveys distributed and collected (response rate), and the observations used for analysis were as follows: Japan, 5316, 4934 (92.8%), 4850; U.S., 7046, 5313 (75.4%), 5075; India, 1280, 1037 (81.0%), 1037.7 Variable definitions8 Birthweight dummies Respondents were asked how much they weighed when they were born, allowing them to choose from one of the following: “less than 2.5 kg,” “2.5–2.999 kg,” “3.0– 3.499 kg,” “3.5–3.999 kg,” “4.0–4.499 kg,” “4.5 kg or more,” and “don’t know”.9 The distribution of these responses is shown in Table1. Compared to 6.2% of the respondents in Japan and the U.S., 12.3% answered “less than 2.5 kg” in India.10 We defined low-birthweight dummy (LBW), which takes the value of 1 for those who chose “less than 2.5 kg” and 0, otherwise. High-birthweight dummy (HBW) in Japan was defined as 1 if birthweight is equal to or more than 4 kg, and 0 otherwise because only three 4 Schellong etal., [12] conducted a meta-analysis of 66 studies across 26 countries, which investigated the relationship between birthweight and subsequent overweight, and found that there exists a linear relationship between them—high (low) birthweight results in higher (lower) risk of overweight. 5 The questionnaires and the aggregated results are available on the website of the Institute of Social and Economic Research, Osaka University. (https:// www. iser. osaka-u. ac. jp/ survey_ data/ top_ jp. html, viewed on December 18, 2020). 6 With regard to social class, the respondents were selected evenly across five strata as defined by the education level (illiterate to post-graduate) and occupation (unskilled labor to intermediate and senior civil officials/executives) of householders. 7 A lower number of observations were used in the analysis, primarily because many people did not answer the questions about their income and parent’s education. 8 It is possible and sometimes effective to use the continuous variable of birthweight instead of the low- and high-birthweight dummies. However, because this study focuses on the association between low- and high-birth- weight and QOL, and six options were provided for birthweight, we used birthweight dummies rather than a continuous birth weight variable. 9 In the U.S., pounds (lb) are used instead of kg, and the six options are as follows: “less than 5.5 lb.,” “5.5 to 6.9 lb.,” “7.0 lb. to 8.4 lb.,” “8.5 lb. to 9.9 lb.,” “more than 10 lb.,” and “don’t know.” Hereafter, for simplicity, 2.5 kg represents 5.5 lb.; 4 kg, 8.5 lb.; and 4.5 kg, 10 lb. for the U.S. data. 10 According to the most recent available data from OECD, the percentage of low birthweight is 28.0% in India, 9.4% in Japan, and 8.3% in the U.S. (https:// stats. oecd. org/ Index. aspx? Query Id= 81145). The main reason why the percentage of low birthweight in our survey is lower than these values is that many respondents answered “don’t know.” Assuming that respondents who answered “don’t know” has the same distribution over the birthweight of the respondents who remembered birthweight, the percentages of low birthweight become 8.3% in Japan, 7.9% in the U.S., and 20.6% in India, which is closer to the official data. Page 4 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 respondents answered “4.5 kg or more” in Japan. In the U.S., 632 respondents between 4 kg and 4.5 kg were defined as HBW, and 145 respondents who weighed 4.5 kg or more are defined as very high birthweight (V_HBW). In India, since there are no respondents who weighed 4 kg or more, we defined the 17 respondents who weighed 3.5 kg or more as having a “quasi-high birthweight” (Q_HBW).11 Furthermore, we defined the “do not know dummy” (DONTKNOW), which takes 1 for those who chose “don’t know” and 0 otherwise, for all the estimations in the three countries. Outcome variables12 Academic performance, height, education, and health: Academic performance (ACADEMIC) is based on a self-assessment of “grade ranks of all subjects” at the age of 15 years on a five-point scale. HEIGHT (in meters) is the height at the time of the survey. EDUCATION was defined as the highest level of education completed.13 Marital status MARRIAGE is a binary dummy variable, which takes 1 for “currently married” and 0 otherwise. BMI BMI was defined as weight (kg) / [height (m)]2 at the time of the survey. Personal income For Japan and the U.S., the respondents were asked to select their personal income (before taxes and including bonuses) in 2010 from a list of 10 options ranging from 1 = “None” to 10= “JPY 14 million (USD 140 thousand) or more.” We estimated the value (unit = JPY 1 million; USD 10 thousand) of each classification by applying lognormal distribution to the frequency distribution of these responses. For India, respondents were asked directly for their monthly personal income (before taxes and including bonuses; unit = INR 10 thousand) in 2010, from which annual personal income was calculated by multiplying with 12. We defined this as “personal income” (INCOME). Though personal income of housewives/househusbands, students, and the retired is often extremely low, it is not appropriate to evaluate that their QOL is low. Therefore, this study excluded them from the sample for the estimation of personal income.14 HEALTH was defined as the responses to the question “How would you describe your current health status?” based on a five-point Likert scale. Happiness Respondents were asked the question “Overall, how happy would you say you are currently?” and were requested to choose on a scale of 0 to 10, with 0 being “very unhappy” and 10 being “very happy”. We defined the variable HAPPINESS with the responses. Control variables The control variables include parents’ education (F_EDU- CATION: father’s education, M_EDUCATION: mother’s education), parents’ age at birth (F_AGE_BIRTH: father’s age at birth, M_AGE_BIRTH: mother’s age at birth), standard of living at 15 years old (S_LIVING), no siblings at age 15 (ONLYCHILD), and mother’s employment status at age 15 (M_FULLTIME: full-time work dummy, M_PARTTIME: part-time work dummy, and M_HOUSEWIFE: housewife dummy).15 Other control variables include male dummy (MALE), age (AGE), age squared (AGESQ), depth of religious faith (RELIGION), and prefectures and state at the age of 15 years in Japan and the U.S., respectively, and current residence in the six cities in India (region dummies).16 The definitions of these variables are summarized in Table2. Also, Table3 presents the descriptive statistics of the variables used in the estimation. In Fig.1, we present the mean and 95% confidence interval (95%CI) of each outcome variables by birthweight categories by country. Looking at the means, Fig.1A (Japan) reveals that LBW is lower for all the categories except MARRIAGE and BMI, while HBW is higher for HEIGHT, EDUCATION, BMI, HEALTH, and HAPPINESS. Figure1B (U.S.) reveals that LBW is lower for ACADEMIC, HEIGHT, EDUCATION, INCOME, HEALTH, and HAPPINESS, while HBW is higher for HEIGHT, BMI, INCOME, and lower for MARRIAGE and HAPPINESS. Figure1C (India) reveals that LBW is lower for HEIGHT, EDUCATION, INCOME, HEALTH and HAPPINESS, while HBW is higher for ACADEMIC, HEIGHT, EDUCATION, BMI, INCOME, HEALT, and HAPPINESS. In sum, the tendency of lower outcomes for LBW is recognized in three countries, while HBW is associated with lower outcome in the U.S. but higher 14 To check the robustness, we also estimated the income equation deleting the respondents with zero income. The results were qualitatively same. 15 In India, however, the data of the parents’ age at birth were not available. 16 Since there is no data on the region of residence at the age of 15 years for India, we used the city of current residence. 11 According to [13], “ideal birth weight” in terms of being less likely to develop diabetes in adulthood is lower by 1 kg in India compared with the U.S. Therefore, it would be worthwhile to examine the association of “quasi-high birth weight.” 12 Validity of the outcome variables, especially marriage and BMI, is discussed in Additionalfile2: Supplemental material B. 13 In India, the options include “illiterate” and “literate, but not in school.” Page 5 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 outcomes in India and Japan. Since 95% CI is large from several cases, however, regression analysis is required to get more reliable results. Hypotheses As mentioned in the Introduction, studies in developed countries have found that low birthweight has a negative association with health and education [also refer to Additional file 1: Supplementary material A]. Therefore, we speculate that LBW has an unfavorable association with QOL and examine: Hypothesis 1: LBW has a negative association with the outcomes concerning QOL, specifically, academic performance, height, highest educational attainment, personal income, health, and happiness. While several studies that have reported associations between LBW and health in young adulthood, relatively few have reported associations of LBW with outcomes in adulthood and beyond. Therefore, we posit the following hypothesis. Hypothesis 2: The association between LBW and outcomes concerning QOL is stronger in youth, and weaker in old age. Although there have been only a few studies, some indicated that HBW causes overweight later in life, resulting Table 2 Definition of the variables Name of the variables Definition of the variables Academic performance (ACADEMIC) Self-assessment of “Average of all Subjects” at age 15 on a five-point Likert scale, with 1 being the lowest and 5 being the highest performance. Height (HEIGHT) Height in meters at the time of the survey. Education (EDUCATION) Highest level of education completed. For Japan, from 1 = “Grade School” to 11 = “Doctoral Degree.” For the U.S., from 1 = “Grade School” to 9 = “Doctoral Degree.” For India, from 1 = “Illiterate” to 8 = “Graduate/ Post-Graduate-Professional.” Marital status (MARRIAGE) 1 = married, 0 = otherwise BMI (BMI) Weight in kilograms divided by the square of height in meters (kg/m2). Personal Income (INCOME) Annual personal income (before taxes and including bonuses). JPY 1 million for Japan, USD 10 thousand for the U.S. and INR 10 thousand for India. Housewives/househusbands, students, and retired individuals are excluded. For Japan and the U.S., respondents chose from 1 = “No income” to 10 = to “14 million JPY (140 thousand USD) or more.” The lognormal distribution was applied to the frequency distribution to estimate the class values. For India, monthly income (10 thousand rupees) multiplied by 12. Health (HEALTH) Self-assessment of health status at the time of response on a five-point Likert scale: from 1 = “not good” to 5 = “good.” Happiness (HAPPINESS) Self-assessment of happiness at the time of response, from 0 = “very unhappy” to 10 = “ver y happy.” Low birthweight (LBW) 1 = weighing less than 2.5 kg, 0 = otherwise High birthweight (HBW) For Japan, 1 = weighing more than 4 kg, 0 = otherwise. For the U.S., 1 = weighing 4–4.499 kg, 0 = otherwise. Quasi-high birthweight (Q_HBW) For India, 1 = weighing more than 3.5 kg, 0 = otherwise. Very high birthweight (V_HBW) For the U.S., 1 = weighing more than 4.5 kg, 0 = otherwise. Do not know (DONTKNOW) 1 = do not know the birthweight, 0 = otherwise Age (AGE) Respondent’s age Age-squared (AGESQ) Squared term of the respondent’s age Old age dummy (OLD) 1 = the respondent’s age is higher than or equal to 50 years old, 0 = otherwise. Gender (MALE) 1 = male, 0 = female Father’s education (F_EDUCATION) Education level of the respondent’s father Mother’s education (M_EDUCATION) Education level of the respondent’s mother Father’s age at birth (F_AGE_BIRTH) Father’s age when the respondent was born. Mother’s age at birth (M_AGE_BIRTH) Mother’s age when the respondent was born. Mothers’ employment status dummy Mother’s employment status when the respondent was 15 years old. Full-time worker (M_FULLTIME), part-time worker (M_PARTTIME), housewife (M_HOUSEWIFE) Standard of living at age 15 (S_LIVING) Self-assessment of “Standard of living” at the age of 15 years, on a 11-point Likert scale, from 0 = “lowest” to 10 = “highest.” Only child dummy (ONLYCHILD) 1 = no siblings at age 15, 0 = otherwise Religious beliefs (RELIGION) Self-assessment of degree of religious beliefs at the time of response, with 1 = “doesn’t hold true at all” to 5 = “particularly true” to the statement “I am deeply religious.” Page 6 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 in various health problems [12, 14, 15]. Moreover, as pointed out in the Introduction, a high BMI prevails in the U.S. (a developed country), whereas underweight prevails in India, which leads us to the following: Hypothesis 3: HBW has an association with outcomes concerning QOL. Specifically, it has a negative association in the U.S. and a positive association in India. Method ofestimation To test Hypotheses 1 through 3, we regressed each of the eight outcome variables on LBW and HBW as well as the interaction terms of the old age dummy and the birthweight dummies for each country.17 We estimated both reduced form and recursive-structural form. Reduced form We estimated the reduced form based on the literature in this field, regressing each outcome variable over exogenous variables including birthweight dummies. All exogenous variables were used in all the equations without venturing into the issue of identification. The estimates of the reduced form represent the total association of each regressor with each outcome variable as the outcomes of QOL. The formula of the reduced form is given by Eq. (1). Here, OUTCOMEik is the kth outcome of the i th individual, BWim is the mth birthweight dummy variable (LBW and HBW, etc.), Xi is the set of control (1) OUTCOME ik =𝛼k+𝛽k,m ∑ m BW im +γ k,m ∑ m BW imOLD i + 𝛿 k OLD i+ 𝜀 k X i+ e ik ,k = 1, … ,8 Table 3 Descriptive statistics of the variables Obs. is number of observations. Calculations are based on the sample of valid responses to the birth weight question. Birthweight categories are presented in Table1 Japan U.S. India Obs. mean median IQR Obs. mean median IQR Obs. mean median IQR ACADEMIC 4717 3.39 3.0 1.0 2772 3.69 4.0 2.0 926 3.06 3.0 2.0 EDUCATION 4795 4.27 3.0 4.0 3680 5.02 5.0 2.0 1037 4.57 5.0 1.0 HEIGHT (m) 4797 1.62 1.62 0.14 3283 1.70 1.70 0.15 1037 1.59 1.6 0.1 WEIGHT (kg) 4716 59.9 59.0 15.0 3586 82.2 79.4 24.9 1036 58.5 58 15 BMI 4714 22.6 22.3 3.98 3203 28.2 26.8 7.70 1036 23.1 22.8 5.4 INCOME 2989 3.54 2.86 3.40 2774 4.16 2.88 3.41 484 11.8 9.6 8.4 HEALTH 4802 3.35 3.0 1.0 3678 3.37 3.0 1.0 1037 3.46 3.0 1.0 HAPPINESS 4738 6.39 7.0 3.0 3563 7.29 8.0 3.0 1037 7.39 8.0 1.0 AGE 4850 52.25 53.0 21.0 3703 52.59 53.0 23.0 1037 45.72 45.0 23.0 F_EDUCATION 4589 2.64 2.0 2.0 3624 3.58 3.0 3.0 1037 3.86 4.0 3.0 M_EDUCATION 4586 2.26 2.0 2.0 3632 3.56 3.0 1.0 1037 3.24 3.0 3.0 F_AGE_BIRTH 4234 31.3 31.0 6.0 3703 30.1 29.0 9.0 – – – – M_AGE_BIRTH 4274 27.7 27.0 6.0 3703 27.0 26.0 9.0 – – – – S_LIVING 4800 4.78 5.0 2.0 2777 4.47 5.0 3.0 1037 5.80 6.0 2.0 RELIGION 4841 1.66 1.0 1.0 3610 2.95 3.0 2.0 1037 3.95 4.0 2.0 Freq. ratio (%) Cum. Freq. ratio (%) Cum. Freq. ratio (%) Cum. Gender MALE 2591 46.6 46.6 2257 45 55.4 469 45.2 45.2 FEMALE 2259 53.4 100 2798 55.4 100 568 54.8 100 OLD age dummy OLD 2033 41.9 41.9 3047 60.0 60.0 407 39.3 39.3 NOT OLD 2817 58.1 100 2028 40.0 100 630 60.8 100 Martial status MARRIAGE 3881 80.2 80.2 3010 60.0 60.0 837 80.7 80.7 UNMARRIAGE 957 19.8 100 2006 40.0 100 200 19.3 100 Mother’s employment status M_FULLTIME 1861 39.3 39.3 1957 46.9 46.9 64 6.2 6.17 M_PARTTIME 1171 24.7 64.1 873 20.9 67.9 31 3.0 9.16 M_HOUSEWIFE 1550 32.8 96.8 1177 28.2 96.1 930 89.7 98.8 Otherwise 151 3.19 100 162 3.89 100 12 1.2 100 Siblings ONLYCHILD 256 5.45 5.45 283 7.46 7.46 60 5.79 5.79 Otherwise 4442 94.55 100 3512 92.54 100 977 94 100 17 Note that this study does not use twin data. Page 7 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 variables, and eik is the disturbance term. If the mth birthweight dummy lowers the kth outcome, then βk, m will be negative. We also expect that the association of birthweight dummy is stronger for younger people and will gradually disappear as they get older (Hypothesis 2). To measure this change, we created an old age dummy (OLDi), which takes 1 if the respondent’s age is higher than or equal to 50 years old, and 0 otherwise. We made its interaction terms with birthweight dummies. Therefore, the coefficient of birthweight dummies, βk, m, represents its association for young respondents, while the coefficient of the interaction terms, γk, m, represents the difference in the associations between the old and the young. For BMI, INCOME, HE ALTH, and HAPPINESS, γk, m is expected to take the opposite sign of βk, m, because the influence on these outcomes will be smaller for the older respondent. On the other hand, because ACADEMIC, HEIGHT, EDUCATION, and MARRIAGE are outcomes corresponding to a younger age for most respondents, we expect that they are independent of their age when the survey was conducted, so that γk, m is expected to be zero. As control variables, we used sex and age as basic attributes, parents’ education and parents’ age at birth (as confounding factors), standard of living at the age of 15 years, presence or absence of siblings at the age of 15 years, mother’s employment status at the age of 15 years, region of residence at the age of 15 years, and the depth of religious faith, which we regarded as exogenous to the outcome variables. We estimate Eq. (1) with ordinary least squares (OLS) for HEIGHT and BMI, with logit for MARRIAGE and with ordered logit for ACADEMIC, EDUCATION, INCOME, HE ALTH, and HAPPINESS.18 2.8 3 3.2 3.4 3.6 ACADEMIC < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT a abc 1.55 1.6 1.65 1.7 1.75 HEIGHT < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT b 3.5 4 4.5 5 5.5 EDUCATION < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT c .7 .75 .8 .85 .9 MARRIAGE < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT d 22 23 24 25 BMI < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT e 2.5 3 3.5 4 4.5 INCOME < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT f 3.2 3.4 3.6 3.8 4 HEALTH < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT g 6.2 6.4 6.6 6.8 7 7.2 HAPPINESS < 2.5 2.5-2.999 3.0-3.499 3.5-3.999 4.0≤ BWEIGHT 95%CI mean h 3.4 3.5 3.6 3.7 3.8 3.9 ACADEMIC < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 1.66 1.68 1.7 1.72 1.74 1.76 HEIGHT < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 4.4 4.6 4.8 5 5.2 EDUCATION < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT .5 .55 .6 .65 MARRIAGE < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 27 28 29 30 31 BMI < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 3 3.5 4 4.5 5 5.5 INCOME < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 3 3.1 3.2 3.3 3.4 3.5 HEALTH < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 6.6 6.8 7 7.2 7.4 7.6 HAPPINESS < 2.5 2.5-2.999 3.0-3.999 4.0–4.499 4.5≤ BWEIGHT 95%CImean 2.8 33.2 3.4 3.6 3.8 ACADEMIC < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 1.56 1.58 1.6 1.62 1.64 1.66 HEIGHT < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 44.5 5 5.5 6 EDUCATION < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT .7 .8 .9 1 1.1 MARRIAGE < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 22 24 26 28 BMI < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 10 15 20 25 INCOME < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 3 3.5 4 4.5 5 HEALTH < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 7.5 8 8.5 9 9.5 HAPPINESS < 2.5 2.5-2.999 3.0-3.499 3.5≤ BWEIGHT 95%CImean Fig. 1 A Japan: Each mean outcome variable by birthweight categories with 95% CIs. Notes: This figure implies associations between birthweight and each outcome variable in the case of Japan. LBW is lower for all the categories except MARRIAGE and BMI, while HBW is higher for HEIGHT, EDUCATION, BMI, HEALTH, and HAPPINESS in Japan. B USA: Each mean outcome variable by birthweight categories with 95% CIs. Notes: This figure implies associations between birthweight and each outcome variable in the case of the U.S. LBW is lower for ACADEMIC, HEIGHT, EDUCATION, INCOME, HEALTH and HAPPINESS, while HBW is higher for HEIGHT, BMI, INCOME and lower for MARRIAGE, and HAPPINESS in the U.S. C India: Each mean outcome variable by birthweight categories with 95% CIs. Notes: This figure implies associations between birthweight and each outcome variable in the case of India. LBW is lower for HEIGHT, EDUCATION, INCOME, HEALTH and HAPPINESS, while HBW is higher for ACADEMIC, HEIGHT, EDUCATION, BMI, INCOME, HEALTH, and HAPPINESS 18 In the estimation by ordered logit, odds ratio, which is the exponential of the coefficient, can unambiguously test if the coefficient differs from zero with statistical significance. Page 8 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Recursive andstructural form Since the eight outcome variables are endogenous, the problem of identification arises. Therefore, in this study, in addition to the reduced form, we estimate the recursive and structural form.19 Figure2 depicts the timeline, or possible paths, from birth until the survey, through which birthweight has a connection with each outcome concerning QOL. As Fig.2 shows, birthweight is connected with the adult outcomes directly and indirectly. We assume the recursive system of the outcomes to be occurring in the order of ACADEMIC, HEIGHT, EDUCATION, and MARRIAGE.20 Therefore, the specification for ACADEMIC is the same as in Eq. (1). As for HEIGHT, the specification becomes the one which added ACADEMIC to Eq. (1). For EDUCATION, HEIGHT is added to this equation, and for MARRIAGE, EDUCATION is further added to this equation. Each equation is estimated separately by OLS. Note that the time-ordering simply implies that predetermined variables are not affected by later events. However, the ordering is not precise, so there should exist some who do not follow it: For example, some people graduate from university after they marry. Therefore, this timeline includes advantages and drawbacks. We should consider the tradeoff between assuming slightly unsound time-ordering and solving the difficult endogeneity problem among eight instead of four outcome variables. This paper chose the former approach in consideration of the fact that the burden of the endogeneity problem is enormous. Among the assumed orderings, many people may question if higher academic performance will lead to higher height. However, we do not argue this causation, or any causation, in Fig.2. We just assume that height, which is determined around 18 years of age for many people, does not affect academic performance in 15-year-olds. Many people may also question if height affects performance in later life. Though we again do not assume this causation, there have been studies that report that height is positively associated with wage level Fig. 2 Path diagram: Associations of birthweight on long-term outcomes. Note: This figure shows the timeline path from birth until the survey, meaning that birthweight is associated with each outcome related to quality-of-life (QOL) 19 Since appropriate instrument variables are often difficult to find, we regard the results of the recursive- and structural-form with instrumental variable method for a robustness check of the results of the reduced form. 20 As per academic performance, the survey question asked scores in school at the age of 15 years. Height is determined by around the age of 20 years, and final education is, for most people, between 15 and 22 years. Many people marry after completing school, and according to the results of the questionnaire surveys, divorce and remarriage are relatively rare (especially in Japan). Page 15 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Table 6 (continued) ACADEMIC HEIGHT EDUCATION MARRIAGE BMI INCOME HEALTH HAPPINESS OLS Ordered logit OLS OLS Ordered logit OLS Logit OLS OLS Ordered logit OLS Ordered logit OLS Ordered logit Coef. Odds Ratio Coef. Coef. Odds Ratio Coef. Odds Ratio Coef. Coef. Odds Ratio Coef. Odds Ratio Coef. Odds Ratio (0.0316) (0.052) (0.00246) (0.0472) (0.054) (0.0111) (0.091) (0.122) (0.465) (0.083) (0.0251) (0.086) (0.0357) (0.081) Cons 4.608*** 1.581*** 3.329*** −0.951*** 16.32*** −9.509* 2.990*** 5.862*** (0.551) (0.0311) (0.743) (0.201) (1.891) (5.585) (0.431) (0.748) Obs. 926 926 1037 1037 1037 1037 1037 1036 484 484 1037 1037 1037 1037 R20.175 0.182 0.247 0.185 0.093 0.271 0.294 0.392 The coefficients are estimated by OLS and odds rations by orderd logit. The definition and unit of each variable are presented in Table2. Regression coefficients are presented in the upper rows. The region dummies representing the prefecture where respondents lived at the age of 15 years are included in the estimation, but are not shown here to save space. Robust standard errors are in parentheses. Because the estimated coefficients are very small, the coefficients and standard errors for AGE are multiplied by 100 and AGESQ by 1000. In the case of linear regression, if the units of explanatory variables are changed, such as when dividing by 100, the estimated coefficients and standard errors are multiplied by 100 while the t-values remain the same. *** p < 0.01, ** p < 0.05, * p < 0.1 Page 16 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 in Additional file 3: Supplemental material C; Table C-1). Age and gender exhibited the largest associations with many outcomes: the association with LBW was merely one tenth to one fourth. Though low birthweight has a smaller association with education than that of father’s education and living standards in childhood, it has a comparable association with personal income and health. The U.S. In Table5, we present the estimates of Eq. (1) for the U.S. LBW was associated with no outcomes. The magnitudes of the coefficients of LBW for various outcomes were less than half of the corresponding estimates in Japan. In addition, the interaction term of LBW and old age dummy was not significant. In contrast, HBW had a significantly positive association with HEIGHT and BMI, and a significantly negative association with MARRIAGE. V_HBW had a significant and positive association with HEIGHT and BMI, while it had a significantly negative association with HE ALTH and HAPPINESS. Further, as for the outcomes significantly associated with HBW and/ or V_HBW, the interaction terms with old age dummy took the opposite signs, indicating that these associations are mitigated for older adults. The above results are summarized as follows: (3) in the U.S., though LBW was not associated with any outcomes significantly, HBW had a negative association with HE ALTH and HAPPINESS, and (4) the significant associations of HBW for those under 50 years old were reduced for older adults. Though these results were not consistent with Hypothesis 1, they support Hypotheses 2 and 3.23 These conclusions are confirmed with odds ratio estimated with an ordered logit model though there were two differences in the significance level. According to the standardized regression, the magnitude of the coefficient of V_HBW in the equation for HE ALTH and HAPPINESS was around − 0.06, which is tantamount to one sixth of that on age. It was comparable with that of S_LIVING for HE ALTH and HAPPINESS, and F_EDUCATION and M_AGE_BIRTH for HE ALTH, revealing that the association was fairly large. (Estimates are presented in Additional file3: Supplemental material C; Table C-2). India The estimates for India are presented in Table6, which are summarized as follows: (5) LBW is negatively associated with HEIGHT, EDUCATION, INCOME, and HE ALTH, and positively associated with MARRIAGE. (6) As for these outcomes, the coefficients of interaction dummies with the old age dummy took the opposite signs to those on LBW except for HE ALTH, indicating that the associations with these outcomes were mitigated for older adults. However, as for HE ALTH, the negative association of LBW was augmented for those over 50 years old. (7) Q_HBW was significantly positive for BMI, HE ALTH, and HAPPINESS.24 These conclusions are confirmed with the odds ratio estimated with an ordered logit model except for the negative association between LBW and INCOME, though there were three differences in the significance level. The standardized regression analysis revealed that the magnitude of the coefficient of LBW for EDUCATION, INCOME, and HE ALTH were about half to one-third of that of F_EDUCATION for these outcomes. The coefficient of Q_HBW for HE ALTH was about two-thirds of that on F_EDUCATION and half of that on M_HOUSE- WIFE. (Estimates are presented in Additional file3: Supplemental material C; Table C-3). Comparing the results of the reduced form estimations across the three countries, while LBW was associated with various outcomes negatively in Japan and India, it was not associated with any outcome in the U.S. However, in the U.S., HBW was negatively and significantly associated with MARRIAGE, and V_HBW was negatively and significantly associated with HE ALTH and HAPPINESS. In contrast, in India, Q_HBW was positively and significantly associated with HE ALTH and HAPPINESS. These different associations of HBW in the U.S. and India support Hypothesis 3. Estimation results ofrecursive andstructural form We estimated the recursive system consisting of consisting of ACADEMIC, HEIGHT, EDUCATION, and 23 We briefly mention the association with controls. MALE was positive for HEIGHT and INCOME, negative for ACADEMIC, and insignificant for EDUCATION. Age demonstrates an inverted U-shape with respect to EDUCATION, MARRIAGE, BMI, and INCOME, while it is U-shaped for HE ALTH and HAPPINESS. F_EDUCATION was related to many outcomes as in Japan. In the U.S., M_EDUCATION also had similar associations with a couple of outcomes. S_LIVING had a positive association with ACADEMIC, HEIGHT, EDUCATION, HE ALTH, and HAPPINESS, and a negative association with BMI as in Japan. M_HOUSEWIFE was positively related to INCOME. However, the effects of F_AGE_BIRTH, M_AGE_BIRTH, and ONLYCHILD were limited as in Japan, except for a positive association of mother’s age at birth with EDUCATION and HE ALTH. 24 As for the associations of the controls, male dummy was significantly positive for ACADEMIC, HEIGHT, EDUCATION, INCOME, whereas it was insignificant for MARRIAGE, HE ALTH, and HAPPINESS. MARRIAGE, BMI, and INCOME showed an inverted U-shaped association with age. Whereas F_EDUCATION was significantly and positively associated with ACADEMIC, EDUCATION, INCOME, and HE ALTH, M_EDUCATION was only positively and significantly associated with HAPPINESS. M_HOUSEWIFE was associated with HEALTH and HAPPINESS positively. S_LIIVNG was positively and significantly associated with all the outcomes except for BMI, INCOME, and HAPPINESS. ONLYCHILD has a positive association only on happiness. Page 17 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 MARRIAGE using OLS, while we estimated the structural form using BMI, INCOME, HE ALTH, and HAPPINESS, considering that these four variables are endogenous. Specifically, we adopted the averages of the endogenous outcomes across the samples of the same gender, age group, and residence as the instruments.25 Though we included all the control variables in all the estimations, they are not shown in the tables to save space. Japan The estimation results for Japan are presented in Table7.26 LBW had a significantly negative association with HEIGHT and EDUCATION, and a significantly positive association with MARRIAGE and BMI. Though all coefficients had the same signs as those of the reduced form, their absolute values were smaller, except for HEIGHT and BMI. These results suggest that the indirect associations have the same signs as direct associations, except HEIGHT and BMI. Meanwhile, INCOME, HE ALTH, and HAPPINESS, whose coefficients were significantly negative in the reduced form estimations, became insignificant in the structural estimations, indicating that LBW was associated with these outcomes through indirect paths, but not directly. HBW was significantly positive for HEIGHT and BMI, but insignificant for other outcomes. In addition, the magnitudes of the significant coefficients were similar to the estimates of the reduced form, suggesting that the indirect associations of HBW were marginal. In sum, these results are consistent with those of the reduced form. The results concerning Hypothesis 2 (i.e., interaction terms of old age and birthweight dummies) by the recursive-structural form are consistent with those of the reduced form.27 The U.S. We present the estimation results for the U.S. in Table8. Estimates of LBW were not significant for any outcomes, as were in the reduced form. Coefficients of HBW were significantly positive for HEIGHT and significantly negative for MARRIAGE, as in the reduced form, whereas it was insignificant for BMI. In addition, the absolute values of the estimates became smaller than those of the reduced form representing the total associations, suggesting that indirect associations have the same signs as the direct associations. V_HBW had a significantly positive association with HEIGHT and BMI, and a significantly negative association with HE ALTH.28 India The results for India are presented in Table9. The magnitude of the absolute values of the coefficients of LBW were similar to those in the reduced form, albeit slightly smaller. Consequently, the association of LBW with HE ALTH became insignificant. The associations of Q_ HBW became insignificant for BMI and HAPPINESS, and significant only for HEALTH.29,30 Results byfull information maximum likelihood (FIML) method In addition to the recursive-structural form, the whole system of the eight outcomes associated with QOL was estimated with FIML. The estimates did not differ largely from those by recursive-structural estimations, both of which represent the direct association. They confirmed the conclusions deduced in the reduced form estimation. (Estimates of FIML are presented in Additionalfile 4: Supplemental material D; Tables D-1 to D-3.) The SEM command of STATA reports the estimates of direct, indirect, and total associations based on the FIML estimation. (The estimates of direct, indirect, and total associations based on FIML are presented in Additionalfile5: Supplemental material E; Tables E-1 to E-3.) Therefore, we can compare the estimates of the total 27 As per the dependences between outcomes, ACADEMIC was significantly and positively associated with all outcomes except BMI. EDUCATION and HEIGHT were associated with INCOME significantly, while MARRIAGE was associated with HAPPINESS and HE ALTH significantly. Regarding the outcomes estimated in the structural form, both HEALTH and INCOME each showed a mutual positive association with HAPPINESS. Further, HE ALTH was associated with INCOME negatively. These results, except for the last, are consistent with our intuitions. 28 As for the associations between outcomes in the recursive system, ACADEMIC was associated with HEIGHT, EDUCATION, and INCOME; EDUCATION was associated with INCOME and HEALTH; and MARRIAGE was associated with BMI, INCOME, and HEALTH. Regarding the outcomes of the structural system, HAPPINESS had positive and mutual associations with HEALTH and INCOME, as was in Japan. In addition, INCOME and HEALTH were associated with negatively and mutually. 29 Note that the number of observations of BMI, HE ALTH, and HAPPINESS equations in India became half of that of the reduced form estimations because INCOME should be included as the regressor for the structural estimation. The decline in significance of BMI and HAPPINESS may be due to smaller observations. 30 Regarding associations between outcomes of the recursive system, ACADEMIC was associated with HEIGHT, EDUCATION, and HAPPINESS significantly and positively. HEIGHT was associated with BMI significantly and negatively; EDUCATION was associated with INCOME significantly and positively; MARRIAGE was associated with BMI and INCOME significantly and positively. Regarding the four outcomes of the structural form, HEALTH affected INCOME and HAPPINESS significantly and positively. In addition, HAPPINESS and INCOME were mutually associated significantly and negatively. 25 Because we adopted the same number of instruments as the number of endogenous variables, the system is just identified. Regarding weak identification, we showed the results of F-test of the excluded instruments and Kleibergen-Paap rk Wald statistic in Tables5-1 to 5-3. These statistics were large for all the estimations, thereby rejecting the null of the weak identification. 26 As the equation for ACADEMIC is the same as the reduced form, it is not reported here. Page 18 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Table 7 The estimates of the recursive and structural forms: Japan The coefficients are estimated by OLS. The definition and unit of each variable are presented in Table2. Regression coefficients are presented in the upper rows. All control variables are included in the estimation, but their estimates are not shown in the table to save space. The recursive system is estimated with OLS, while structural system is estimated with 2SLS using instrumental variables: the average value of each variable (BMI, income, health, and happiness) with the same three attributes of gender, age, and place of present residence. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1 Recursive Structural HEIGHT EDUCATION MARRIAGE BMI INCOME HEALTH HAPPINESS LBW −0.0282*** − 0.324** 0.0685** 0.689* −0.401* − 0.0672 − 0.196 (0.00563) (0.149) (0.0333) (0.389) (0.243) (0.109) (0.201) LBW×OLD 0.000661 0.176 −0.114** − 0.922* 0.978* −0.0428 0.506* (0.00738) (0.206) (0.0487) (0.534) (0.515) (0.158) (0.281) HBW 0.0390*** −0.142 0.0217 1.146** −0.178 0.0354 0.084 (0.0104) (0.308) (0.0642) (0.511) (0.470) (0.141) (0.281) HBW×OLD −0.0219 0.652 0.0458 −2.286*** −0.984 −0.348 0.496 (0.0236) (0.630) (0.132) (0.803) (0.836) (0.303) (0.463) DONTKNOW −0.00418 −0.166 − 0.0932** − 0.526* 0.158 − 0.0802 −0.228 (0.00484) (0.154) (0.0395) (0.282) (0.268) (0.0999) (0.165) DONTKNOW×OLD −0.00369 0.0607 0.103** 0.28 −0.104 −0.0015 0.256 (0.00542) (0.174) (0.0429) (0.329) (0.320) (0.112) (0.188) OLD −0.00617* 0.281*** −0.0953*** − 0.0372 0.271 − 0.0724 − 0.139 (0.00333) (0.0958) (0.0223) (0.227) (0.214) (0.0661) (0.120) ACADEMIC 0.00294*** 0.643*** 0.0168** −0.0432 0.313*** 0.0455* 0.0927** ######### (0.0271) (0.00677) (0.0828) (0.0725) (0.0235) (0.0416) HEIGHT 0.717 0.148 −1.78 2.260** −0.221 0.631 (0.493) (0.118) (1.182) (1.055) (0.346) (0.600) EDUCATION −0.00133 0.0178 0.267*** 0.00878 0.0135 (0.00386) (0.0451) (0.0409) (0.0140) (0.0245) MARRIAGE 0.124 0.234 0.146* 0.768*** (0.287) (0.235) (0.0812) (0.117) BMI 0.107 −0.0001 0.000208 (0.102) (0.0370) (0.0619) INCOME 0.0612 −0.0372 0.126** (0.0953) (0.0286) (0.0497) HEALTH −0.0583 −0.761** 0.704*** (0.488) (0.368) (0.223) HAPPINESS −0.14 0.359** 0.185*** (0.213) (0.169) (0.0623) Cons 1.496*** −2.222** −1.358*** 24.18*** −13.21*** 3.431*** 3.263 (0.0173) (0.897) (0.213) (3.347) (3.447) (1.160) (2.188) Obs. 3789 3771 3765 2587 2587 2587 2587 R20.641 0.376 0.167 0.121 0.278 0.163 0.229 Weak identification test F(3, 2515) 18.569 17.691 21.232 17.261 Kleibergen-Paap Wald rk F statistic Underidentification test χ2(3) 46.338*** 44.916*** 59.690*** 44.982*** Kleibergen-Paap rk LM statistic F-value of the first stage BMI – 20.56*** 19.97*** 20.57*** Income 41.30*** – 41.44*** 40.77*** Health 18.95*** 18.94*** – 18.94*** HAPPINESS 23.26*** 22.60*** 23.16*** – Page 19 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Table 8 The estimates of the recursive- and structural-forms: U.S. The coefficients are estimated by OLS. The definition and unit of each variable are presented in Table2. Regression coefficients are presented in the upper rows. All control variables are included in the estimation, but their estimates are not shown in the table to save space. The recursive system is estimated with OLS, while structural system is estimated with 2SLS using instrumental variables: the average value of each variable (BMI, income, health, and happiness) with the same three attributes of gender, age, and place of present residence. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1 Recursive Structural HEIGHT EDUCATION MARRIAGE BMI INCOME HEALTH HAPPINESS LBW −0.0114 −0.0785 0.0018 −0.838 0.197 −0.0416 0.315 (0.0117) (0.197) (0.0820) (1.258) (0.654) (0.166) (0.420) LBW×OLD 0.0157 0.109 0.0373 1.041 −0.714 − 0.0277 − 0.431 (0.0157) (0.276) (0.100) (1.635) (0.919) (0.226) (0.591) HBW 0.0186*** −0.0367 − 0.0941** 1.252* − 0.0962 0.0218 − 0.0371 (0.00672) (0.140) (0.0419) (0.759) (0.451) (0.0882) (0.223) HBW×OLD −0.00088 −0.0315 0.0317 0.318 −1.184* −0.133 0.205 (0.0100) (0.225) (0.0642) (1.205) (0.664) (0.153) (0.368) V_HBW 0.0674*** −0.315 0.0767 3.148* − 0.951 −0.301 − 0.119 (0.0190) (0.318) (0.0798) (1.739) (0.621) (0.193) (0.514) V_HBW×OLD −0.0501* 0.721 −0.227* −1.14 0.793 0.328 0.311 (0.0303) (0.441) (0.125) (2.486) (1.103) (0.286) (0.675) DONTKNOW −0.00775 0.187 −0.0383 − 0.375 0.682 − 0.105 − 0.239 (0.00781) (0.154) (0.0474) (0.849) (0.523) (0.106) (0.252) DONTKNOW×OLD 0.00474 −0.0604 −0.0204 0.751 −1.342** 0.099 0.347 (0.00935) (0.198) (0.0602) (1.159) (0.670) (0.145) (0.340) OLD −0.0022 0.0923 −0.116*** − 0.805 0.0535 0.011 0.0177 (0.00634) (0.134) (0.0401) (0.639) (0.433) (0.0913) (0.232) ACADEMIC −0.00382** 0.611*** −0.00065 0.155 0.197* 0.043 −0.0251 (0.00167) (0.0362) (0.0112) (0.200) (0.112) (0.0272) (0.0667) HEIGHT 0.646 0.138 −7.085** 1.928 0.158 0.398 (0.518) (0.152) (3.466) (1.875) (0.417) (1.077) EDUCATION 0.01 −0.206 0.790*** 0.0803* − 0.0323 (0.00686) (0.306) (0.0913) (0.0411) (0.106) MARRIAGE −0.832* 0.737*** 0.137** 0.177 (0.500) (0.278) (0.0681) (0.185) BMI 0.169* 0.00154 −0.0588 (0.0968) (0.0217) (0.0481) INCOME 0.0438 −0.0285 0.118 (0.355) (0.0495) (0.112) HEALTH −0.918 −1.045* 0.745** (1.165) (0.599) (0.328) HAPPINESS −0.524 0.684** 0.185*** (0.548) (0.319) (0.0686) Cons 1.636*** −2.044** −1.131*** 40.37*** −18.12*** 1.902 6.612** (0.0249) (0.971) (0.294) (8.052) (5.771) (1.452) (2.991) Obs. 1885 1876 1860 1361 1361 1361 1361 R20.517 0.303 0.179 0.144 0.045 0.189 0.128 Weak identification test F(3, 1298) 5.789 9.262 5.307 6.746 Kleibergen-Paap Wald rk F statistic Underidentification test χ2(3) 17.652*** 26.120*** 16.429*** 20.115*** Kleibergen-Paap rk LM statistic F-value of the first stage BMI – 14.32*** 14.32*** 13.72*** Income 9.99*** – 10.66*** 10.20*** Health 18.89*** 19.41*** – 19.50*** Happiness 13.39*** 13.67*** 13.75*** – Page 20 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Table 9 The estimates of the recursive- and structural-forms: India The coefficients are estimated by OLS. The definition and unit of each variable are presented in Table2. Regression coefficients are presented in the upper rows. All control variables are included in the estimation, but their estimates are not shown in the table to save space. The recursive system is estimated with OLS, while structural system is estimated with 2SLS using instrumental variables: the average value of each variable (BMI, income, health, and happiness) with the same three attributes of gender, age, and place of present residence. Robust standard errors are in parentheses. *** p < 0.01, ** p < 0.05, * p < 0.1 Recursive Structural HEIGHT EDUCATION MARRIAGE BMI INCOME HEALTH HAPPINESS LBW −0.0135 −0.377*** 0.0869** −0.626 −2.392** − 0.162 0.0237 (0.00988) (0.126) (0.043) (0.653) (1.210) (0.172) (0.254) LBW×OLD 0.00849 0.25 −0.161** 3.103** −1.604 − 0.424 0.165 (0.0161) (0.222) (0.0787) (1.237) (2.369) (0.291) (0.418) Q_HBW 0.0133 −0.227 0.0574 3.129 −0.851 0.769*** −0.308 (0.0222) (0.324) (0.0790) (1.955) (3.194) (0.204) (0.381) Q_HBW×OLD −0.0265 0.421 0.0489 −0.946 6.119* −0.576 0.859* (0.0356) (0.480) (0.0832) (2.213) (3.669) (0.400) (0.518) DONTKNOW −0.0119* 0.041 −0.0372 − 0.502 −2.437** 0.077 − 0.393** (0.00711) (0.119) (0.0379) (0.560) (1.200) (0.113) (0.192) DONTKNOW×OLD 0.0121 0.568*** −0.00128 1.382* 3.634* −0.637*** 0.36 (0.0100) (0.181) (0.0536) (0.811) (1.965) (0.157) (0.343) OLD 0.00354 −0.277 −0.132** 0.0772 0.214 0.294* −0.394 (0.0110) (0.179) (0.0539) (0.761) (1.991) (0.158) (0.274) ACADEMIC 0.00488** 0.310*** −0.00912 − 0.408** − 0.0977 − 0.0213 0.137** (0.00228) (0.0392) (0.0126) (0.204) (0.468) (0.0442) (0.0685) HEIGHT 0.671 −0.213 −22.79*** −7.172 − 0.491 1.571 (0.562) (0.179) (2.574) (8.651) (0.937) (1.687) EDUCATION 0.000795 0.274 2.102*** 0.0206 0.113 (0.0104) (0.189) (0.405) (0.0454) (0.0715) MARRIAGE 1.108** 3.129*** 0.0817 0.0995 (0.505) (1.107) (0.122) (0.210) BMI −0.116 0.0161 0.00237 (0.300) (0.0362) (0.0683) INCOME −0.0473 0.0173 −0.0511* (0.0607) (0.0140) (0.0266) HEALTH −0.0672 2.395* 0.836*** (0.669) (1.239) (0.202) HAPPINESS 0.11 −1.126 0.192*** (0.320) (0.729) (0.0705) Cons 1.584*** 2.514** −0.436 53.64*** −1.99 3.016 0.311 (0.0364) (1.043) (0.362) (5.915) (20.22) (2.103) (3.935) Obs. 926 926 926 450 450 450 450 R20.179 0.245 0.208 0.295 0.325 0.36 0.349 Weak identification test F(3, 420) 9.217 15.374 9.964 12.613 Kleibergen-Paap Wald rk F statistic Underidentification test χ2(3) 17.790*** 32.265*** 21.420*** 28.743*** Kleibergen-Paap rk LM statistic F-value of the first stage BMI – 16.80*** 16.74*** 16.22*** Income 12.17*** – 12.16*** 11.90*** Health 28.79*** 29.06*** – 27.15*** Happiness 41.40*** 39.83*** 41.49*** – Page 21 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 associations by FIML with the coefficients estimates in the reduced form. These results were similar, indicating consistency of the estimates by the reduced form and structural form (Additional file5: Supplemental material E; Tables E-4 to E-6). Discussions The estimation results across the three countries were quite different: while LBW was negatively associated with QOL in Japan and India, it did not matter in the U.S. In contrast, HBW was a negatively associated with QOL in the U.S. but it was a positively associated with QOL in India. We now explore whether these results are consistent with previous studies [refer to Additional file1: Supplemental material A: Literature survey]. As confirmed in our study, in Japan, [4] reported a significantly negative association of low birthweight on academic performance, education, and personal income while using OLS [5]. found that low birthweight was adversely associated with academic performance, which is confirmed in our results. In addition, our finding that the adverse association among younger respondents was mitigated in older respondents is consistent with their finding. However, their conclusion that low birthweight was not associated with education and primary job status contradicts our results.31 In sum, using a representative sample of Japan, our study found significant associations with more life outcomes. Furthermore, the analysis of high birthweight has not been done before in Japan. Since various contradicting results have been reported on the associations of birthweight in the U.S., it is not easy to compare them with ours. For example, though [22] found a positive relationship between birthweight and the variables educational attainment, height, and wage rates, [23] reported only a minor association with education, but no association with health [24]. reported that birthweight showed a significant positive impact on math and reading ability in childhood. Meanwhile, we found that LBW was not significant for any outcome, whereas HBW had a significant association with several outcomes, including having a negative association with HE ALTH and HAPPINESS. In addition, the comparison became more difficult because most of the previous studies in the U.S. investigated the linear relationship between birthweight and outcomes concerning QOL, while we examined the associations of birthweight dummies. The linear specification of birthweight precludes the possibility that both low and high birthweights are simultaneously worse than the standard birthweight, which possibly produces contradicting results from ours. To ease the comparison, we estimated Eq. (1) substituting birthweight (BWEIGHT) for birthweight dummies such as LBW, HBW, and V_HBW. (Estimates are presented in Additional file6: Supplemental material F; Table F-1). The estimates revealed that BWEIGHT had a significantly positive relationship with HEIGHT and BMI, whereas it was significantly and negatively correlated with HE ALTH and HAPPINESS. These results are consistent with previous studies of the U.S. At the same time, these results correspond to the results of the estimates on V_HBW for these four outcomes, indicating that the insignificant estimates of LBW in the U.S. (Table5) does not necessarily contradict the results of previous studies, which showed that birthweight is significantly associated with some outcomes concerning QOL. In addition, [22] reported that the upper quartile (highest 25%) of birthweight negatively affected wage rate, whereas the bottom quartile (lowest 25%) of birthweight had a positive association, which is consistent with our result of negative associations of V_HBW with various outcomes.32 There have been no papers in India, which to our knowledge, investigated the associations between birthweight and QOL in the field of social science. This study found that LBW has a significant and negative association with ACADEMIC, HEIGHT, EDUCATION, and HE ALTH, while it has significantly positive association with MARRIAGE. Meanwhile, Q_HBW has a significant and positive association with BMI, INCOME, and HE ALTH: that is, higher birthweight has an opposite association compared to that in the U.S. In sum, the associations of birthweight are widely diversified across the countries: in Japan, while low birthweight was negatively associated with many adult outcomes, the associations of high birthweight were limited. In India, while low birthweight was associated with many outcomes negatively as in Japan, quasi-high birthweight was associated with income, health, and happiness positively. In the U.S., while low birthweight did not have any significant associations, very high birthweight was negatively associated with health and happiness. These are visualized using the estimates by standardized regression (Fig.3A and B). 31 Though [5] identified the association between birthweight and various life outcomes, their data, based on the Japanese Study of Age and Retirement (JSTAR), is biased as it is limited to a sample of older adults aged 54 and above, and is geographically limited to 10 cities. Furthermore, the sample size was about 270 for their estimation. In contrast, the data of Japan in this study is a representative sample of 3800 people with respect to sex, age, and region. The differences in their results might come from the differences in the data used. 32 In addition, among the medical studies in the U.S., [15] compared those who were born with a birthweight between 2.0–2.5 kg and over 5.0 kg, reporting that the association with the appearance of orthopedically impaired was larger for those over 5.0 kg. Page 22 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Probably, the diverse results across the three countries come from, in part, the differences in BMI in adulthood. It is known that obesity due to affluent society is a problem in the U.S., while malnutrition due to poverty is a serious problem in India.33 In our data for the U.S., BMI is negatively correlated with HEIGHT, EDUCATION, MARRIAGE, INCOME, HE ALTH, and HAPPINESS at the 1% level, whereas in India, it is positively correlated with MARRIAGE, HE ALTH, and HAPPINESS at the 1% level (results not shown). Meanwhile, we found that HBW had a significantly positive association with BMI in the three countries (Tables4, 5 and 6). These results suggest the possibility that high birthweight led to high BMI in adulthood, which in turn connected negatively to the outcomes in the U.S. and positively to those in India. Furthermore, the result that HBW was associated with the outcomes in adulthood, such as MARRIAGE, BMI, HE ALTH, INCOME, and HAPPINESS, but was not associated with ACADEMIC and EDUCATION in adolescence, is opposite to the results of LBW, suggesting that the mechanism of the associations of low and high birthweight might be different. -.1 -.05 0 .05 .1 ACADEMIC JAPAN ab USA INDIA LBW -.15 -.1 -.05 0 .05 HEIGHT JAPAN USA INDIA LBW -.15 -.1 -.05 0 .05 EDUCATION JAPAN USA INDIA LBW -.1-.05 0 .05 .1 .15 MARRIAGE JAPAN USA INDIA LBW -.15-.1-.05 0 .05 .1 BMI JAPAN USA INDIA LBW -.2-.15-.1-.05 0 .05 INCOME JAPAN USA INDIA LBW -.15-.1-.05 0.05 .1 HEALTH JAPAN USA INDIA LBW -.1 -.05 0 .05 .1 HAPPINESS JAPAN USA INDIA LBW -.05 0 .05 .1 .15 ACADEMIC JAPAN USA INDIA -.05 0 .05 .1 .15 HEIGHT JAPAN USA INDIA -.1 -.05 0.05 EDUCATION JAPAN USA INDIA -.04-.02 0 .02.04.06 MARRIAGE JAPAN USA INDIA 0 .05 .1 .15 BMI JAPAN USA INDIA -.1 -.05 0 .05 .1 INCOME JAPAN USA INDIA -.1-.05 0 .05 .1 .15 HEALTH JAPAN USA INDIA -.15-.1-.05 0 .05 .1 HAPPINESS JAPAN USA INDIA HBW V_HBW Q_HBW Fig. 3 A Mean and 95% CI of the coefficient on LBW estimated by standardized regression. Notes: The graph is based on the estimates reported in tables C-1 to C-3 in Additional file 3: Supplemental material. In this figure, we present the estimates by standardized regression because it allows the comparison of the magnitudes of estimated coefficients among different regressions. B Mean and 95% CI of the coefficients on HBW, V_HBW, Q_HBW for Japan, the U.S., and India, respectively, estimated by standardized regression. Notes: The graph is based on the estimates reported in tables C-1 to C-3 in Additional file 3: Supplemental material. In this figure, we present the estimates by standardized regression because it allows the comparison of the magnitudes of estimated coefficients among different regressions 33 According to [25], Southern Asia, including India, has very high stunting prevalence and the highest wasting prevalence. Page 23 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Conclusions This study analyzed the long-term association between birthweight and QOL using the large-scale survey data conducted in Japan, the U.S., and India in 2011. This study is unique in the following ways: First, it investigated the eight outcomes concerning QOL, adolescent academic performance, height, education, marriage, BMI, income, health, and happiness; second, it analyzed the data for three countries, Japan, the U.S., and India, which are widely diversified with respect to birthweight. Third, it investigated whether the associations tend to be mitigated when older; Fourth, it investigated not only the association of low birthweight, but also the association of high birthweight; Fifth, it investigated using not only the estimation of reduced form, but also the recursivestructural forms, accounting for the transmission of the associations between outcomes. The estimates of the reduced form, which represent the total association of birthweight dummies with outcomes, revealed that low birthweight had significant and negative associations with most of the outcomes in Japan, while high birthweight did not have an association with any outcomes. In contrast, in the U.S., whereas low birthweight did not have an association with any outcomes significantly, very high birthweight was significantly and negatively associated with some outcomes including happiness. In India, low birthweight was associated with most of the outcomes significantly and negatively, whereas quasi-high birthweight was associated with some outcomes including happiness significantly and positively. In addition, these associations were stronger for younger respondents than older ones. Further, the results of the recursive-structural form were consistent with those of the reduced form. These findings have important implications. First, the finding that the association between low and high birthweight and later QOL are different between countries suggests that the association may partly depend on how society helps citizens facing difficulties from their birthweight. In particular, this study found that, whereas high birthweight has a negative association with HE ALTH and HAPPINESS for younger people in the U.S., no such association was found for older people (Table5), and that in India, whereas there was a positive association between high birthweight and INCOME for older people, no such association was found for younger people. These complex findings might have a root in diversified social conditions among countries. In this study, we not only investigated the associations assuming them to be independent of each other, but also assuming that they constitute a system in which they heavily depend on each other. Such an approach can elucidate how the association between low- and high-birth- weight and one outcome transfers to another outcome. Because the transmission can be prevented by appropriate treatment, such a finding might elucidate weaknesses in society to mitigate the association between low- and high-birthweight and later QOL. We hope that such a study will be made in the future. This study has limitations. First, the birthweight data were self-reported. Therefore, the data may suffer from memory bias. How might this affect the findings in this study? Since older people probably recalled their birthweights less accurately, the use of recalled birthweights could partly explain the weaker associations between birthweight and outcome variables in the older age group. To be immune from this problem, however, we should collect longitudinal cohort data, which has not been done in Japan.34 Second, the lack of information regarding gestational period such as gestational age at birth is another serious limitation of this study, though this study controlled for confounders such as parents’ education and their ages at birth, living standards in childhood, mother’s working status in childhood, and the presence of brothers and sisters. To confirm the causality from birthweight to life outcomes, the information on gestational period is important. Analysis using twin fixed-effects model is an alternative method. Nonetheless, as twin fixed-effects model estimates the effect of the difference in birthweight that occurred in gestation period, the mechanism of how the difference emerges may not be identical between twins and singletons [refer to Additional file1: Supplementary material A]. Therefore, whether results using twins have external validity to singletons is questionable, which calls for research using singletons, as in this study. Third, to focus on the associations between low and high birthweight and QOL, we set the entire standard birthweight as the baseline. However, people’s fate may depend on their birthweight even when they are born within the standard birthweight, which constitutes an interesting future study. 34 Similar to our study, previous studies in Japan also used self-reported birthweight data. Page 24 of 25 Yamaneand Tsutsui Health Economics Review (2022) 12:48 Abbreviations JHPS-CPS: Japan Household Panel Survey on Consumer Preferences and Satisfaction; BMI: Body Mass Index; JPY: Japanese Yen; USD: US Dollar; OLS: Ordinary Least Squares; 2SLS: Two-Stage Least Squares; FIML: Full-Information Maximum Likelihood; SEM: Structural Equation Modeling; OECD: Organization for Economic Co-operation and Development; COE: Centers of Excellence. Supplementary Information The online version contains supplementary material available at https:// doi. org/ 10. 1186/ s13561- 022- 00393-9. Additional file1: Supplemental material A. Survey of the literature. Additional file2: Supplemental material B. The validity of the outcome variables. Additional file3: Supplemental material C. The estimates of the standardized regression. Additional file4: Supplemental material D. Estimates using full-infor- mation maximum likelihood (FIML) method. Additional file5: Supplemental material E. Direct, indirect, and total associations. Additional file6: Supplemental material F. The estimation results using birthweight instead of birthweight dummies: U.S. Acknowledgements This study used microdata from the Preference Parameters Study of Osaka University’s twenty-first Century COE Program, “Japan Household Panel Survey on Consumer Preferences and Satisfaction.” We are grateful to all contributors, especially Yoshiro Tsutsui, Fumio Ohtake, and Shinsuke Ikeda. We would like to thank Professor Koichi Maekawa, Professor Daiji Kawaguchi, Professor Tomo Nishimura, and Associate Professor Sayaka Nishimura for their guidance and comments. We would also like to express my appreciation for the financial support from Japan Society for the Promotion of Science (JSPS KAKENHI Grant Number 21 K01534). Authors’ contributions CY conducted literature searches, reviewed articles, analyzed data, and drafted the manuscript. YT contributed to conceptualizing and designing the study, synthesizing the findings, and preparing the manuscript. Both CY and YT read and approved the final manuscript. Funding This study was supported by JSPS KAKENHI (Grant Number 21 K01534). Availability of data and materials The datasets used and/or analyzed in this study are available on the website of the Institute of Social and Economic Research, Osaka (https:// www. iser. osaka-u. ac. jp/ survey_ data/ top_ jp. html, viewed on December 18, 2020), and is obtained with written permission. The conclusions of this article and the associated files are included within the article as supplementary information files. Declarations Ethics approval and consent to participate Not applicable. Consent for publication Not applicable. Competing interests The author declares that they have no competing interests. Author details 1 Faculty of Economics, Hiroshima University of Economics, 5-37-1, Gion, Asaminami, Hiroshima-shi, Hiroshima 731-0192, Japan. 2 Faculty of Social Relations, Kyoto Bunkyo University, Senzoku-80 Makishimacho, Uji, Kyoto 611-0041, Japan. Received: 15 October 2021 Accepted: 2 September 2022 References 1. Pearson H. The Life Project: The Extraordinary Story of Our Ordinary Lives: Penguin Books; 2007. 2. Currie J. Healthy, Wealthy, and Wise: Socioeconomic Status, Poor Health in Childhood, and Human Capital Development. J Econ Lit. 2009;47:87–122 Available from: https:// pubs. aeaweb. org/ doi/ 10. 1257/ jel. 47.1. 87. 3. Currie J, Hyson R. Is the impact of health shocks cushioned by socioeconomic status? The case of low birthweight. Am Econ Rev. 1999;89:245–50. 4. Nakamuro M, Uzuki Y, Inui T. The effects of birth weight: Does fetal origin really matter for long-run outcomes? Econ Lett. 2013;121:53–8. 5. Matsushima M, Shimizutani S, Yamada H. Life course consequences of low birth weight: Evidence from Japan. J Jpn Int Econ. 2018;50:37–47. 6. UNICEF and WHO. Low Birthweight: Country, regional and global estimates: Unicef; 2004. 7. Razak F, Corsi DJ, Slutsky AS, Kurpad A, Berkman L, Laupacis A, et al. Prevalence of body mass index lower than 16 among women in low- and middle-income countries. JAMA. 2015;314:2164–71. 8. Campbell S. Fetal macrosomia: a problem in need of a policy. Ultrasound Obstetr Gynecol. 2014;43:3–10 Available from: https:// onlin elibr ary. wiley. com/ doi/ 10. 1002/ uog. 13268. 9. Ishikawa H, Kanayama N, Matsubara S. Management of Fetal Macrosomia. Notice Points to “Guideline for Obstetrical Practice in Japan 2008”. Acta Obstet Gynaecol Jpn. 2008;60:424–31. 10. Rich-Edwards JW, Colditz GA, Stampfer MJ, Willett WC, Gillman MW, Hennekens CH, et al. Birthweight and the risk for type 2 diabetes mellitus in adult women. Ann Intern Med. 1999;130:278–84. 11. Mccance DR, Pettitt DJ, Hanson RL, Jacobsson LTH, Knowler WC, Bennett PH. Birth weight and non-insulin dependent diabetes: Thrifty genotype, thrifty phenotype, or surviving small baby genotype? BMJ. 1994;308:942–5. 12. Schellong K, Schulz S, Harder T, Plagemann A. Birth Weight and Long-Term Overweight Risk: Systematic Review and a Meta-Analysis Including 643,902 Persons from 66 Studies and 26 Countries Globally. PLoS One. 2012;7:e47776. Available from. https:// doi. org/ 10. 1371/ journ al. pone. 00477 76. 13. Gluckman PD, Hanson MA. Living with the past: Evolution, development, and patterns of disease. Science (1979). 2004;305(5691):1733–6. 14. Flensborg-Madsen T, Mortensen EL. Birth Weight and Intelligence in Young Adulthood and Midlife. Pediatrics. 2017;139:e20163161 Available from: http:// pedia trics. aappu blica tions. org/ lookup/ doi/ 10. 1542/ peds. 2016- 3161. 15. Avchen RN. Birth Weight and School-age Disabilities: A Population-based Study. Am J Epidemiol. 2001;154:895–901 Available from: https:// acade mic. oup. com/ aje/ artic lelookup/ doi/ 10. 1093/ aje/ 154. 10. 895. 16. Case A, Paxson C. Stature and status: Height, ability, and labor market outcomes. J Polit Econ. 2008;116:499–532. 17. Gao W, Smyth R. Health human capital, height and wages in China. J Dev Stud. 2010;46:466–84. 18. Yamamura E, Smyth R, Zhang Y. Decomposing the effect of height on income in China: The role of market and political channels. Econ Hum Biol. 2015;19:62–74. 19. Angrist JD, Pischke J. Mostly Harmless Econometrics: An Empiricist’s Companion: Princeton University Press; 2009. 20. Greene HW. Econometric Analysis: Global Edition. 8th ed: Pearson Education Limited; 2020. 21. Abe Y, Kikuchi T, Nagasaki K, Hiura M, Tanaka Y, Ogawa Y, et al. Lower birth weight associated with current overweight status is related with the metabolic syndrome in obese Japanese children. Hypertens Res. 2007;30:627–34. 22. Behrman JR, Rosenzweig MR. Returns to birthweight. Rev Econ Stat. 2004;86(2):586–601. 23. Royer H. Separated at girth: US twin estimates of the effects of birth weight. Am Econ J Appl Econ. 2009;1:49–85.