100269 THE IMPACT OF IMMIGRATION ON CHILD HEALTH: EXPERIMENTAL EVIDENCE FROM A MIGRATION LOTTERY PROGRAM STEVEN STILLMAN, JOHN GIBSON and DAVID MCKENZIE∗ This paper uses a unique survey designed by the authors to compare migrant children who enter New Zealand through a random ballot with children in the home country of Tonga whose families were unsuccessful participants in the same ballots. We find that migration increases height and reduces stunting of infants and toddlers, but also increases BMI and obesity among 3- to 5-yr-olds. These impacts are quite large even though the average migrant household has been in New Zealand for less than 1 yr. Additional results suggest that these impacts occur because of dietary change rather than direct income effects. (JEL J61, I12, F22) I. INTRODUCTION will generally experience large gains in income and increased access to health care and clean Childhood obesity is a major public health water, this migration also potentially introduces problem both globally and in the United States unhealthy lifestyle patterns, such as increases in (Institute of Medicine 2004; Troiano et al. fat and refined sugar-rich diets and decreases 1995). At the same time, extensive immigration in regular physical activity (Clemens, Montene- to the United States, Canada, Europe, Australia, gro, and Pritchett 2008; McKenzie, Gibson, and and New Zealand (NZ) has led to large increases Stillman 2009; Popkin and Udry 1998). Thus, in the number of foreign-born children in these migration may potentially have negative impacts countries, with many, if not most, of these on health, particularly of still-growing children children being born in developing or transition who are most affected by environmental and countries. Although economic migrants moving dietary changes.1 from a developing to a developed country Child health is of intrinsic interest, both as a current measure of well-being and a *We thank the Government of the Kingdom of Tonga for permission to conduct the survey there, the New Zealand source of future human capital. Moreover, given Department of Labour Immigration Service for providing the strong economic argument for increas- the sampling frame, attendees at the An International Per- ing international migration, it is important for spective on Immigration and Immigration Policy Conference in Canberra, Australia for helpful comments, Halahingano economists to also examine other impacts that Rohorua and her assistants for excellent work conducting migration can have on well-being and whether the survey, and most especially the survey respondents. these impacts lower the net benefit of migrating Financial support from the World Bank, Stanford Univer- sity, the Waikato Management School, and Marsden Fund grant UOW0503 is gratefully acknowledged. The study was approved by the multiregion ethics committee of the New 1. For example, see http://vivirlatino.com/2006/03/02/ Zealand Ministry of Health. The views expressed here are immigration-to-the-us-harmful-to-your-health.php (accessed those of the authors alone and do not necessarily reflect the March 4, 2007). opinions of the World Bank, the New Zealand Department of Labour, or the Government of Tonga. ABBREVIATIONS Stillman: Senior Fellow, Motu Economic and Public Policy ATT: Average Treatment Effect on the Treated Research, Level 1, 97 Cuba Street, PO Box 24390, BMI: Body Mass Index Wellington, New Zealand. Phone 64-4-939-4250, Fax 64-4-939-4251, E-mail stillman@motu.org.nz CDC: Centers for Disease Control Gibson: Professor of Economics, Department of Economics, DoL: Department of Labour University of Waikato, Private Bag 3105, Hamilton, New ITT: Intention to Treat Effect Zealand. Phone 64-7-856-2889, Fax 64-7-838-4331, LATE: Local Average Treatment Effect E-mail jkgibson@waikato.ac.nz OLS: Ordinary Least Squares McKenzie: Senior Economist, Development Research Group, PAC: Pacific Access Category The World Bank, 1818 H Street NW, Washington DC PINZMS: Pacific Island–New Zealand Migration 20433. Phone 202-458-9332, Fax 202-522-3518, Survey E-mail dmckenzie@worldbank.org 62 Economic Inquiry doi:10.1111/j.1465-7295.2009.00284.x (ISSN 0095-2583) Online Early publication March 11, 2010 Vol. 50, No. 1, January 2012, 62–81 © 2010 Western Economic Association International STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 63 for individuals and for society as a whole. How- not drawn in the ballot. This survey instrument ever, identifying the causal impact of migration collected information on both parental-reported on child health requires comparing the current health and measured anthropometrics, as well health of migrant children to what their health as additional data on household income, diets, would have been had they stayed in their home and access to health care facilities. Thus, we are country. This counterfactual is typically unob- able to examine whether migration has a causal served, and thus the current literature settles for impact on child health or whether migration either comparing the health of immigrant chil- just changes parents’ reference points for what dren to that of native-born groups in the des- “good health” means, and examine the pathways tination country (e.g., Bell and Parnell 1996; through which changes in child health occur. Frisbie, Cho, and Hummer 2001; Gordon et al. In the short-term, migration is found to 2003; Institute of Medicine 1998; Kirchengast increase height and reduce stunting of 0- to 2- and Schober 2006) or comparing the health of yr-olds, and increase weight, body mass index immigrant children in the destination country (BMI) and obesity of 3- to 5-yr-olds, and have to that of similar nonimmigrant children in the no impact on anthropometrics but lead to bet- source country (e.g., Smith et al. 2003). Both of ter parental-reported health for 6- to 18-yr-olds. these approaches assume that there is no selec- The scale of these impacts is quite large even tivity into migration and thus the health of non- though the average migrant household has been migrant children can be used as an appropriate in NZ for less than 1 yr. Additional results sug- counterfactual for what the health of migrant gest these health effects operate through dietary children would have been in the absence of change rather than as direct income effects. It is migration.2 These approaches are not very con- well known that the first 3 yr of life are when vincing because migrant families are likely to height is most susceptible to nutritional changes, differ from nonmigrant families along a host and it is exactly for this age-group that we see of unobserved dimensions, some of which are migration affecting height. For older children, a likely to be correlated with both child health richer, higher calorie diet has a limited impact and migration. on height, but instead increases body mass. This paper overcomes this problem by exam- Tongan migrants to NZ are not atypical of ining the impact of migration on children’s the average developing country migrants else- health in the context of a unique survey of where in the world. For example, the average participants in a migrant lottery program. The adult Tongan migrant in our sample has 11.7 yr Pacific Access Category (PAC) under NZ’s of education, compared to 11.0 yr for the aver- immigration policy allows an annual quota of Tongans to migrate to NZ. The other options age 18- to 45-yr-old new arrival in the United available for Tongans to migrate are fairly lim- States. However, unlike many developing coun- ited, unless they have close family members tries, there are already high levels of childhood abroad. Many more applications are received obesity in Tonga (Fukuyama et al. 2005). Thus, than the quota allows, so a ballot is used by the from the standpoint of the worldwide problem NZ Department of Labour (DoL) to randomly with childhood obesity, it is discouraging to find select from among the registrations. The same that migration leads to increased obesity even survey instrument, designed by the authors, was among an already overweight population. This applied in both Tonga and NZ and allows exper- suggests that the increased global movement of imental estimates of the impact of migration people will serve to strengthen the worldwide on child health to be obtained by comparing convergence toward a mostly overweight popu- the health of immigrant children whose parents lation. However, this is not meant to imply that were successful applicants in the ballot to the migration to NZ has necessarily been bad for health of those children whose parents applied to the health of the Tongan children in our sam- migrate under the quota, but whose names were ple. Previous studies have shown that stunting has a negative association with cognitive devel- opment and adult labor market outcomes (Case 2. A much smaller literature looks at children who remain in their home countries while a parent migrates (e.g., and Paxson 2008; Colombo, de Andraca, and Hildebrandt and McKenzie 2005; Kanaiaupuni and Donato Lopez 1988; Jamison 1986). Thus, the increased 1999). These studies can at best determine the impact of height and reduced stunting of Tongan children having a migrant parent on the health of children, but do not provide information on the health impacts of the child in NZ may have a larger positive effect on their themselves migrating. lifetime well-being than any negative effects 64 ECONOMIC INQUIRY from increases in weight and obesity caused by As noted earlier, the majority of the immi- migration. grant health literature compares immigrants to The next section briefly discusses a sim- native-born in the destination country. In the ple theoretical model of why migration might United States, much attention has been given affect child health, and summarizes the find- to the “healthy immigrant paradox,” which has ings of existing literature on migration and found Hispanic immigrants to be of better health child health. Section III provides background than U.S. natives of similar socioeconomic sta- and context on Tongan health and migration, tus (Institute of Medicine 1998). However, there and describes the survey and measures of child is some evidence that this is in part due to health used in this study. Section IV calculates selectivity, with healthier individuals migrating the treatment effect of migration on child health. (Rubalcava et al. 2008) and in many other con- Section V then explores the mechanisms under- texts immigrant children have been found to lying the measured impacts on child health, and be in poorer health than natives. For example, Section VI concludes. Kirchengast and Schober (2006) report higher rates of obesity among Turkish and Yugoslav immigrant children in Austria than Austrian II. HOW MIGHT MIGRATION AFFECT CHILD HEALTH? children; and Meulmeister, Berg, and Wedel (1990) find higher rates of micronutrient defi- A. Theoretical Model ciencies and malnutrition among Turkish and The literature has identified many potential Moroccan immigrant children than Dutch chil- channels through which immigration may affect dren in the Netherlands. the health of children. The Grossman (1972) The studies most closely related to ours health production function provides a theoretical in terms of geographic focus have compared framework which we use to summarize these anthropometric outcomes for Pacific Island chil- various effects. The health of child i at a dren in NZ to those for other children in NZ. particular point in time can be written as: Pacific Island children are taller and heavier for their age than both international reference (1) Hi = h(Mi , Ti , Ki , Bi , εi ) standards and Caucasian children in NZ. For where Mi represents medical and nutritional example, the prevalence of obesity in 3–7-yr- inputs, Ti encompasses the time inputs of the old Pacific Island children ranges from 42% parent and the time use of the child, Ki is to 49%, depending on the criteria used, ver- parental health knowledge, Bi represents biolog- sus only 7%–13% for comparable Caucasian ical endowments such as genetic factors, and εi children (Gordon et al. 2003). The mean height represents random health shocks. Migration may and weight of Pacific Island children tracks affect child health through changes in Mi —such the 95th percentile of international reference as changing diets and changes in access to health charts until about age 10–11, with height then care; through changes in Ti —such as less time falling back toward the reference median while breastfeeding (Carballo, Divino, and Zeric 1998) weight remains high (Salesa, Bell, and Swinburn and changes in the level of physical activity 1997). Both genetic and dietary differences may of children (Unger et al. 2004); and changes account for some of these differences across eth- in Ki , if parents gain more health knowledge nic groups, with Pacific Island children having when abroad (Hildebrandt and McKenzie 2005). significantly higher fat intakes than non-Pacific However, the main challenge to identifying the Island children (Bell and Parnell 1996). How- impact of migration is that the migration deci- ever, none of these studies distinguish between sion of a household might be correlated with immigrant Pacific Island children and those born variables unobserved by the researcher, such as in NZ and thus have little to say about the impact either a child’s genetic health status, Bi , or with of migration. random health shocks, εi . As discussed earlier, immigrants differ from natives in many observable and unobservable dimensions, making it difficult to ascribe any of B. Related Literature these differences to the impact of migration per Although there is a large literature on the se. A number of other studies explore the impact health of immigrant children, this identifica- of acculturation by comparing the health of tion challenge makes it difficult to ascribe most immigrants who have been abroad for differing of the findings to the effects of immigration. amounts of time (see Institute of Medicine 1998, STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 65 for a review). But, there are several problems (WHO 2005) reports that there is no chronic which prevent this strategy from giving us the undernutrition and no important micronutrient full impact of immigration on health. First, a deficiencies in Tonga. However, earlier work number of health effects may occur very soon suggests that malnutrition may occur during after migrating (or even during the migration infancy and early childhood due to delays in the journey in some cases) and thus comparing the introduction of supplementary food or lack of health of a child who has been abroad 1 yr to nutritionally valuable weaning foods and diets one who has been abroad 5 yr will clearly miss too low in protein among children under 2 yr the health impacts which occur during the first of age (Bloom 1986; Lambert 1982). Among year. Second, both because the effect of migra- adolescents and adults, noncommunicable dis- tion on health is likely to vary with age at arrival, eases are the most important health problem. and because the unobservable characteristics of The adult obesity rate was 60% in 2004 (WHO migrants are likely to vary over time, it is not 2005), whereas a recent study of 5- to 19-yr- possible to identify the impact of years in the olds also found high rates of childhood obesity, destination country on health (e.g., it is not pos- especially among girls (Fukuyama et al. 2005). sible to separately identify age, cohort, and year effects).3 Third, individuals in either the origin B. Migration Context and the PAC or the destination country may have experienced health shocks (say a drought) during the inter- Emigration levels are high, with 30,000 Ton- vening period which should be accounted for gans living abroad, the vast majority in NZ, when measuring the impact of immigration. Australia, and the United States. However, dur- Overall, the scarcity of surveys which contain ing the 1990s, the opportunities for emigration information on both migrants in the destination became more limited, as NZ followed Australia country and nonmigrants in particular source in introducing a points system for migration, countries, and the challenge of separating the with points awarded for education, skills, and impact of migration from migrant selectivity, business capital. Few Tongans qualified to emi- limits the ability of the existing literature to grate under these systems, and so most Tongan identify the health impacts of migration for emigration was through family reunification cat- children. In the next section, we discuss how egories, as the spouse, parent, or child of an the unique data used in this paper helps resolve existing migrant. For example, in 2004/2005, both these problems. only 58 Tongans gained residence to NZ through the business/skilled categories, compared to 549 through family categories. Australia admitted III. CONTEXT AND SURVEY DATA 284 Tongans during the 2004/2005 financial A. Background and Health Context year, whereas the United States admitted 324 Tongans in 2004, of which 290 were under fam- The Kingdom of Tonga is an archipelago ily categories.5 of islands in the South Pacific, about two- In early 2002, another channel was opened thirds of the way from Hawaii to NZ. The up for immigration to NZ through the creation population is just more than 100,000, with a of the PAC, which allows for a quota of 250 gross domestic product (GDP) per capita of Tongans to emigrate to NZ each year regardless approximately U.S.$2,200 in PPP terms. One- of their skill level or socioeconomics status.6 third of the labor force is in agriculture and Specifically, any Tongan citizens aged between fishing, with the majority of workers in the 18 and 45, who meet certain English, health, and manufacturing and services sectors, which are dominated by the public sector and tourism. 5. Source: Australian Government Department of Immi- Tonga’s infant mortality rate is 20 deaths per gration and Multicultural Affairs, U.S. Department of Home- 1,000 live births, comparable to Ukraine, Brazil, land Security Office of Immigration Statistics, and New and Paraguay, and much higher than the 5.3 per Zealand Department of Labour. 1,000 in NZ.4 Data on malnutrition and stunt- 6. The Pacific Access Category also provides quotas for 75 citizens from Kiribati, 75 citizens from Tuvalu, and, ing is scarce. The World Health Organization prior to the December 2006 coup, 250 citizens from Fiji to migrate to New Zealand. There have been some changes 3. In addition, selective return migration can cause the in the conditions for migration under the Pacific Access characteristics of migrants who have been in the country Category since the period we examine in this paper (see longer to differ from those who have been in the country Gibson and McKenzie 2007 for details)—here we describe for shorter periods. the conditions that applied for the potential migrants studied 4. Source: World Bank Central Database, data for 2005. in this paper. 66 ECONOMIC INQUIRY character requirements,7 can register to migrate The first group consists of a random sample to NZ.8 Many more applications are received of 101 of the 302 Tongan immigrant households than the quota allows, so a ballot is used by the in NZ, who had a member who was a successful NZ Department of Labour (DoL) to randomly participant in the 2002–2005 PAC ballots.10 select from among the registrations. During the Administrative data show that none of the ballot 2002–2005 period we study, the odds of having winners had returned to live in Tonga at the one’s name drawn were approximately one in time of the survey, nor had any of them after a ten. Individuals whose names are not selected further 2 yr. There are 171 children aged ≤18 in can apply again the next year. these households. The second group consists of a Once their ballot is selected, applicants must sample of households of successful participants provide a valid job offer in NZ within 6 mo from the same random ballots who were still in in order to have their application to migrate Tonga at the time of surveying. We sampled 26 approved. This offer can be for essentially any of the 65 households in this group, focusing our full-time job, and most of the migrants began sampling on households located in villages from work in typical entry level occupations, such as which the migrants in our first survey group packing groceries in supermarkets and working had emigrated. Most of this group consists of in construction. After a job offer is filed along individuals whose applications were still being with their residence application, it typically processed at the time of surveying. There are takes 3–9 mo for an applicant to receive a 56 children aged ≤18 in these households. In residence decision. Once receiving approval, forming all of our experimental estimates, we they are then given up to 1 yr to move. The weight the sample so that it reflects the actual median migrant in our sample moved within 1 ratio of migrants to successful ballots still in mo of receiving their residence approval. At the Tonga at the time of the survey. time of our survey, the median migrant child The third survey group consists of house- had spent 6 mo in NZ (mean of 7.6 mo). Thus, holds of unsuccessful participants in these same this paper examines the short-term impact of ballots. The full list of unsuccessful ballots from migration on child health. these years was provided to us by the NZ DoL, but the details for this group were less infor- mative than those for the successful ballots, as C. Pacific Island–NZ Migration Survey only a post office box address was supplied and there were no telephone numbers. We used two The data used in this paper are from the strategies to derive a sample of 119 households first wave of the Pacific Island–NZ Migration with a member with an unsuccessful ballot from Survey (PINZMS), a comprehensive household this list, with this sample size again dictated by survey designed to measure multiple aspects of our available budget. First, we used information the migration process and take advantage of on the villages where migrants had come from the natural experiment provided by the PAC.9 to draw a sample of unsuccessful ballots from The survey design and enumeration, which was the same villages (implicitly using the village overseen by the authors in 2005–2006, covered of residence as a stratifying variable). Second, random samples of four groups of households, we used the Tongan telephone directory to find surveying in both NZ and Tonga. contact details for people on the list. To over- come concerns that this would bias the sample 7. Data supplied by the New Zealand Department of to the main island of Tongatapu, where people Labour for residence decisions made between November are more likely to have telephones, we delib- 2002 and October 2004 reveals that out of 98 applica- erately included in the sample households from tions only 1 was rejected for failure to meet the English requirement and only 3 others were rejected for failing other requirements of the policy. See McKenzie, Gibson, and Still- 10. A large group of the 302 immigrant households man (2009) for more details on this policy. were unavailable for us to survey because they had been 8. The person who registers is a Principal Applicant. reserved for selection into the sample of the Longitudinal If they are successful, their immediate family (spouse Immigrant Survey, conducted by Statistics New Zealand. and children under age 24) can also apply to migrate as In McKenzie, Gibson, and Stillman (2009), we describe in Secondary Applicants. The quota of 250 applies to the total detail the tracking of the sample in New Zealand, showing of Primary and Secondary Applicants and corresponds to a contact rate of more than 70%. The main reasons for about 90 migrant households each year. noncontact were incomplete name and address details, which 9. Further details about this survey and related papers should be independent of child health and therefore not a produced from these data can be found at http://www. source of sample selectivity bias. There was only one refusal pacificmigration.ac.nz. to take part in the survey in New Zealand and none in Tonga. STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 67 the Outer Islands of Vava’u and ‘Eua’. There are been shown to be indicative of health status 281 children aged 18 and under in these house- and correlated with economic prosperity. The holds. remaining seven measures of child health are The final survey group consists of households derived from height and weight data. These living in the same villages as the PAC applicants measurements were directly collected by trained but from which no eligible individuals applied interviewers during the in-person surveys, and for the quota in any of our sample years (e.g., are adjusted for whether the child is measured 2002–2005). We randomly selected 90 nonap- lying down or standing, whether they are wear- plicant households with at least one member ing shoes, and the type of clothing being worn.11 aged 18–45. There are 271 children aged ≤18 in We examine three continuous measures of child these households. These households will be used anthropometry: height, weight and BMI, each to look at the process of health selection into standardized by age in months and gender.12 migration, and for examining the cross-sectional These measures are each expressed as z-scores correlates of child health in Tonga. which show how many standard deviations each The fact that a random ballot was used to child is away from the age- and gender-specific select among applications gives us a group median height, weight, or BMI in a reference of migrants and a comparison group who are population of well-nourished children.13 similar to the migrants in both observable and Our final four measures are threshold mea- unobservable dimensions, but remain in Tonga sures derived from the standardized height and only because they were not successful in the BMI z-scores and based on U.S. Centers for Dis- ballot. This allows experimental estimates of ease Control (CDC) recommendations: stunting the impact of migration on child health to be is defined as having standardized height below obtained by comparing the health of children the 5th percentile of the reference population whose parents were successful applicants in the and indicates chronic undernutrition and poor ballot to the health of those children whose health, underweight as having standardized BMI parents applied to migrate under the quota, but below the 5th percentile, overweight as having whose names were not drawn. standardized BMI between the 85th and 95th percentiles, and obese as having standardized BMI above the 95th percentile of the reference D. Measuring Child Health population (Kuczmarski, Ogden, and Grummer- Our analysis focuses on nine interrelated Strawn 2000).14 measures of child health. The first two are parent-reported measures of each child’s health 11. Height was measured to the nearest 0.1 cm using a portable stadiometer (Schorr Height Measuring Board, status in the current year and their health status Olney, Maryland) and weight was measured to the nearest compared to 1 yr ago on 5-point scales. Self- 0.1 kg on a digital scale (Model UC-321; A&D Medical, reported health status has the virtue of being Milpitas, California). 12. BMI refers to the body mass index which is mea- quick to collect, making it a common ques- sured as weight in kilograms divided by height in meters tion on multipurpose surveys, such as the New squared. This has been shown by nutritionists to best mea- Immigrant Survey in the United States (Jasso sure energy intakes net of energy output. 13. We use the 1990 reference standards for the United et al. 2004), despite evidence of systematic dif- Kingdom, as derived in Cole, Freeman, and Preece (1998), ferences in responses by socioeconomic status for each of these measures as they are available for children (Sindelar and Thomas 1991). These questions of all ages. We find similar results using nonstandardized provide an indication of the level of and changes measures of height, weight, and BMI, but focus on the standardized results for comparability with the literature. in overall health status; however, there are rea- 14. There is considerable debate about the validity of sons to worry that parental responses to these using universal BMI cutoff points for comparing obesity questions may change with migration, regard- prevalence across ethnic groups. Rush, Plank, and Davies (2003) show that for the same BMI, the percent body-fat for less of whether health actually changes. For Pacific Island children is lower than that for NZ children of example, when reporting whether or not their European origin. Rush et al. (2004) report similar findings child is in good health, migrant parents may for young adults, for example, they find that the average body-fat for a young adult Pacific Islander with a BMI of compare their children to a reference group of 33 is the same as that for a young adult of European origin NZ children, rather than to the health standards with a BMI of 30. However, because we are comparing BMI of children in Tonga. for Tongan children in New Zealand to Tongan children Physical indicators of nutrition are not sub- in Tonga, as opposed to comparing immigrant children to natives, as is common in much of the literature, this ject to respondent-specific reporting error and debate about using ethnic-specific BMI cutoffs should not are of direct interest themselves as they have be a concern. 68 ECONOMIC INQUIRY Child height (or stature) is generally known lack of a strong income gradient in child health to be a sensitive indicator to the quality of eco- in Tonga, which we show later in the paper. nomic and social environments (Steckel 1995), Nevertheless, even in the absence of migra- whereas child weight and, more typically, BMI tion selectivity in terms of child health, the have been demonstrated to be good measures for results from a nonexperimental study still will be identifying short-run effects on health (Strauss biased either if the migration decision of adults and Thomas 1998). A number of studies have depends on their underlying desires for invest- shown that the relationship between socioeco- ing in their children’s futures, including making nomic status and child health varies with the future investments in child health, or if house- age of the child (Case, Lubotsky, and Paxson holds experience shocks (such as a drought) 2002; Sahn and Alderman 1997). Thus, we strat- which drive both their migration decision and ify our analysis of the impact of migration on directly affect future child health. The PAC bal- child health into four age-groups across which lot allows us to produce an unbiased experimen- impacts are likely to differ: 0–2, 3–5, 6–12, tal estimate of the causal impact of migration and 13- to 18-yr-olds. on child health, regardless of potential unob- Environmental factors are especially impor- servables that are correlated with a household’s tant determinants of child height in early child- desire to emigrate. hood. Therefore, the World Health Organization recommends focusing analysis of height mea- IV. THE EFFECT OF MIGRATION ON CHILD HEALTH sures to 0- to 5-yr-olds (WHO 1986). The stature of infants and children is particularly vulnerable This section focuses on estimating the impact to nutritional stresses and, in our example, these of migration to NZ on the health of Tongan children changed environments during this vul- children. We rely on the fact that the PAC nerable stage in life (all 0- to 2-yr-olds in our ballot, by randomly denying eager migrants sample were born in Tonga, because they had the right to move to NZ, creates a control to be included in the ballot application to be group of children that should have the same included in our sample, and thus were mainly outcomes as what the migrant children would brought to NZ as infants). Thus, we further split have had if they had not moved. Evidence that the 0–5 age-group. Teenagers are often dropped the control group of nonmigrants is statistically when examining child health, because the onset identical to the successful ballots in terms of ex- of puberty is thought to be weakly related to ante characteristics is reported in Table 2. We underlying health status, thus making it diffi- cannot reject equality of means for any variable cult to measure the true relationship between among all children (0- to 18-yr-olds), which is other covariates and health status. Instead of consistent with the random selection of ballots dropping teenagers, we examine their outcomes among applicants to the PAC.15 separately. A. Sample Means and Intent-to-Treat Effects E. Migration Selection and Child Health Table 3 presents the proportion of parents The PAC randomizes among the group of reporting their children are in very good health, households interested in migrating to NZ under as opposed to good or average health; the pro- the PAC. It is thus of interest to examine portion of parents reporting their children are in whether children in households which apply to much better health now compared to 1 yr ago, migrate under the PAC have different health than as opposed to somewhat better now, about the children in households which do not apply to same now, or somewhat worse health now; the migrate. Table 1 compares the characteristics of mean z-score for each anthropometric measure; children and their parents in the unsuccessful and the proportion of children that are stunted, ballot households to those for the nonapplicant underweight, overweight, and obese among chil- children. We see positive selection into the PAC dren in each of the four age-groups whose par- applicant pool in terms of parental education ents were either successful or unsuccessful in the and household income. However, there is no PAC ballot (and standard errors for each which significant difference in any of our nine child health measures between children in nonappli- 15. McKenzie, Gibson, and Stillman (2009) provide further evidence that the PINZMS captures a random sample cant households and children in households with of both successful and unsuccessful PAC ballots and that unsuccessful ballots. This is consistent with the winning the ballot is properly randomized. STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 69 TABLE 1 Selection of Families into the Pacific Access Category Ballot (Comparison of Characteristics of Children ≤18 in Nonapplicant and Unsuccessful Ballots) Sample Means in Tonga Unsuccessful t -test of Equality of Ballots Nonapplicants Means p Value Proportion children 0–2 yr old 0.15 0.22 .05 Proportion children 3–5 yr old 0.20 0.25 .10 Proportion children 6–12 yr old 0.41 0.35 .11 Proportion children 13–18 yr old 0.23 0.17 .13 Age in months 104.2 89.1 .03 Proportion female 0.46 0.45 .77 Proportion living with both parents 0.93 0.93 .98 Number of children in household 4.8 4.3 .38 Father’s age 38.9 38.9 1.00 Father’s years of education 11.6 10.8 .01 Father’s height 170 163 .28 Mother’s age 37.0 36.9 .94 Mother’s years of education 11.3 10.6 .03 Mother’s height 164 165 .65 Total real household cash income 17,553 9,348 .00 Total real household own production 10,427 7,399 .06 Very good parent-rated health 0.51 0.55 .57 Much better health since last year 0.34 0.38 .54 Standardized height for age −0.25 −0.19 .76 Standardized weight for age 1.05 0.93 .55 Standardized BMI for age 1.50 1.35 .47 Stunted—height for age ≤ 5th percentile 0.17 0.17 .95 Underweight—BMI for age ≤ 5th percentile 0.04 0.05 .58 Overweight—BMI for age 85th–95th percentile 0.16 0.16 .89 Obese—BMI for age ≥ 95th percentile 0.44 0.42 .77 Total sample size 281 271 — Note: Test statistics account for clustering at the household level and survey stratification. account for clustering at the household level and However, in concordance with the high lev- survey stratification and weighting). els of obesity in Tonga as a whole, children Consider first children in households where are on average heavier than the reference pop- the parent had been unsuccessful in the PAC ulation, with 39% of 0- to 2-yr-olds, 48% of ballot. These children remain in Tonga, and their 6- to 12-yr-olds, and 64% of 13- to 18-yr- health indicates what health conditions would be olds classified as obese. For the children ≥6 yr, like in the absence of migration. Infants and tod- mean weight for age and BMI for age are over dlers (aged 0–2) are generally short in stature one standard deviation higher than the reference compared to the reference population, with 36% population. The exception is 3- to 5-yr-olds, defined as stunted. Mean standardized height is which are only slightly heavier than the refer- closer to the reference population for older chil- ence population. One explanation of these dif- dren but, in each age-group, a larger proportion ferent patterns among 0- to 2-yr-olds compared than expected are stunted (12%, 13%, and 17%, to 3- to 5-yr-olds may be that Tongan children respectively for 3–5, 6–12, and 13- to 18-yr- have growth (height) spurts at slightly older olds versus 5% in the reference population by ages than British children under 5 who form definition). This is consistent with the findings in the reference population. However, because our early studies such as Lambert (1982) and Bloom analysis only uses this reference group for stan- (1986) that suggested malnutrition could be an dardization purposes, this only affects interpre- issue in the early years. tation of the levels of obesity and stunting, and 70 ECONOMIC INQUIRY TABLE 2 Test for Randomization (Comparison of Ex-ante Characteristics of Children ≤18 in Successful and Unsuccessful Ballots) Sample Means Applicants t -test of Equality of Successful Ballots Unsuccessful Ballots Means p Value Proportion children 0–2 yr old 0.12 0.15 .21 Proportion children 3–5 yr old 0.21 0.20 .72 Proportion children 6–12 yr old 0.43 0.41 .68 Proportion children 13–18 yr old 0.24 0.23 .84 Age in months 107.5 104.2 .63 Proportion female 0.47 0.46 .83 Proportion living with both parents 0.98 0.93 .12 Number of children in household 4.3 4.8 .27 Father’s age 39.6 38.9 .47 Father’s years of education 11.7 11.6 .86 Father’s height 162 170 .24 Mother’s age 37.9 37.0 .34 Mother’s years of education 11.6 11.3 .47 Mother’s height 159 164 .30 Proportion in New Zealand 0.80 — — Months in New Zealand 7.6 — — Total sample size 247 281 — Note: Test statistics account for clustering at the household level and survey stratification and weighting. not of the changes in these variables driven by B. The Impact of Migration on Child Health migration. In a perfect randomized experiment, the Simple comparison of means between the impact of the treatment (here, migration) on successful and unsuccessful ballots identify each outcome can be obtained through a sim- whether there are significant intention-to-treat ple comparison of means or proportions in the effects, that is, whether getting a successful bal- control group (unsuccessful ballots) with the lot leads to changes in child health outcomes.16 treatment group (successful ballots), as done in For 0- to 2-yr-olds, we see that winning the bal- the previous subsection. However, as discussed lot causes significantly greater height and less in Heckman et al. (2000), this simple experi- stunting, with no changes in weight or parental mental estimator of the treatment effect on the perceptions of health. Only 5% of 0- to 2-yr- treated is biased either if control group members old children in households with a winning ballot substitute for the treatment with a similar pro- are stunted, compared to 36% of 0- to 2-yr-old gram or if treatment group members drop out of children in households with unsuccessful ballots. the experiment. In our application, substitution For 3- to 5-yr-olds in contrast, we see winning bias will occur if PAC applicants who are not the ballot results in no significant changes in drawn in the ballot migrate through alternative height, but increases in weight, leading to higher means and dropout bias will occur if PAC appli- BMI and a much higher proportion obese. There cants whose name are drawn in the ballot fail to are no significant changes in either height or migrate to NZ. weight for older children, but parents of both We do not believe that substitution bias is of 6- to 12-yr-olds and 13- to 18-yr-olds are more serious concern in our study, as individuals with likely to say their children are in very good the ability to migrate via other arrangements will health in winning ballot households. likely have done so previously given the low odds of winning the PAC ballot.17 Furthermore, 17. We did not come across any incidences where 16. These t tests account for clustering at the household remaining family members told us that the unsuccessful level and survey stratification and weighting. applicant had migrated overseas during our fieldwork. TABLE 3 Summary Statistics—Sample Means Very Good Much Better Standardized Standardized Stunted Height Underweight Overweight Obese BMI for Parent-rated Health Since Height for Weight for Standardized for Age ≤ 5th BMI for Age BMI for Age ≥ 95th Health Last Year Age Age BMI for Age Percentile ≤ 5th Percentile Age Percentile Children 0–2 yr old Unsuccessful ballots 0.70 0.27 −0.91 0.09 1.35 0.36 0.06 0.16 0.39 Successful ballots 0.70 0.44 0.63 0.43 0.58 0.05 0.21 0.05 0.34 Raw intent to treat 0.00 0.17 1.54 0.34 −0.77 −0.32 0.14 −0.11 −0.05 t test of ITT = 0 (p value) .97 .28 .00 .63 .23 .00 .15 .18 .72 Subsample size 65 47 51 53 49 51 49 49 49 Children 3–5 yr old Unsuccessful ballots 0.66 0.36 0.04 0.47 0.52 0.12 0.08 0.13 0.13 Successful ballots 0.69 0.40 0.09 1.32 1.50 0.19 0.02 0.22 0.42 Raw intent to treat 0.03 0.05 0.05 0.86 0.97 0.07 −0.06 0.10 0.29 t test of ITT = 0 (p value) .76 .66 .73 .02 .01 .36 .20 .20 .00 Subsample size 106 106 96 98 90 96 90 90 90 Children 6–12 yr old Unsuccessful ballots 0.47 0.34 0.09 1.39 1.76 0.13 0.02 0.16 0.48 Successful ballots 0.70 0.43 0.04 1.40 1.64 0.12 0.00 0.17 0.42 Raw intent to treat 0.23 0.09 −0.05 0.01 −0.11 −0.01 −0.02 0.00 −0.06 t test of ITT = 0 (p value) .00 .30 .62 .97 .67 .91 .16 .97 .49 Subsample size 220 220 204 210 208 204 208 208 208 Children 13–18 yr old Unsuccessful ballots 0.35 0.35 0.43 1.46 1.87 0.17 0.03 0.16 0.64 Successful ballots 0.69 0.34 0.36 1.66 2.07 0.12 0.02 0.26 0.67 STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH Raw intent to treat 0.34 −0.02 −0.07 0.20 0.20 −0.05 −0.02 0.09 0.03 t test of ITT = 0 (p value) .00 .87 .79 .48 .50 .45 .67 .30 .81 Subsample size 123 123 108 112 111 108 111 111 111 Total sample size 514 496 459 473 458 459 458 458 458 Note: Test statistics account for clustering at the household level and survey stratification and weighting. ITT, Intention to treat effect. 71 72 ECONOMIC INQUIRY as discussed earlier, the other options available cases, whether an individual has migrated to NZ for Tongans to migrate are fairly limited, unless is instrumented by whether their household was they have close family members abroad. How- successful in the PAC ballot. All standard errors ever, as shown in Table 2, dropout bias is a more use the appropriate survey weights to account relevant concern; only 80% of ballot winners for the sampling rates for each group and are (weighted by the number of their children) had clustered at the household level. migrated to NZ at the time of our survey. Many For 0- to 2-yr-olds, we find that migration of the other ballot winning households were still causes a significant increase in height and reduc- in the process of moving, whereas the others tion in stunting. Immigrant children of this age either decided not to move, or were unable to are 1.8 to 1.9 standard deviations taller as a move due to the lack of a valid job offer in NZ result of migration, and 36–42 percentage points for the household principal applicant. less likely to be stunted.20 This greater height is Instrumental variables provide an approach associated with lower BMI for age, but despite for estimating average treatment effects with large magnitudes, the effect on BMI is not sig- experimental data. In our application, the PAC nificant, although there is a greater tendency to ballot outcome can be used as an excluded be underweight for age and a reduced likeli- instrument because randomization ensures that hood of being overweight for age. For 3- to success in the ballot is uncorrelated with unob- 5-yr-olds, we find strong and significant evi- served individual attributes which might also dence that migration increases weight. Migration affect child health, and that success in the ballot leads to a significant 0.9 to 1.0 standard devi- is strongly correlated with migration.18 This esti- ation increase in weight for age, a 0.9 to 1.2 mate is called the local average treatment effect standard deviation increase in BMI for age, a (LATE) and can be interpreted as the effect of 10–18 percentage point increase in the likeli- treatment on individuals whose treatment status hood of being overweight (only significant when is changed by the instrument. Angrist (2004) including control variables), and a 32–36 per- demonstrates that in situations where no indi- centage point increase in the likelihood of being viduals who are assigned to the control group obese. For neither 0- to 2-yr-olds nor 3- to 5- receive the treatment (e.g., there is no substi- yr-olds is there any significant difference in the tution), the LATE is the same as the average likelihood that a parent views the child’s health treatment effect on the treated (ATT). as very good, or being better than last year as a Table 4 presents three sets of results using result of migration, although the point estimates the ATT estimator for each outcome and age- for better health than last year show a positive, group. The first row presents linear instrumental but insignificant, effect of approximately 20 per- variables estimates with no control variables, centage points for 0- to 2-yr-olds. and the second row presents linear instrumental For older children, migration is found to variables estimates with controls added for each have no significant impact on anthropometric child’s gender, age in months, age in months measures. Moreover, most of the point estimates squared, birth order position, and their parent’s are relatively small in size; however, in contrast age and height. Including controls for these pre- to younger children, parents are significantly determined variables should increase the effi- more likely to view their 6- to 18-yr-olds as ciency of our estimates. In almost all cases, the being in very good health after migration. For point estimates are very similar when adding 6- to 12-yr-olds, parents are 28–29 percentage these controls, which is consistent with random- points more likely to view them as having ization balancing these covariates. Finally, the very good health, whereas for 13- to 18-yr- third row presents marginal effects from bivari- olds parents the corresponding figure is 33–41 ate probit models for each discrete outcome, percentage points. with no control variables added.19 In all three Overall, the results appear consistent with children receiving more food intake with 18. Validity of the instrument also requires that the bal- lot outcome does not directly affect child health conditional on migration status. It seems unlikely to us that winning the adding a balanced covariate to a nonlinear model such as a ballot and not being able to migrate would impact the health probit can change the point estimates (Raab et al. 2000). status of children in the household. 20. Although the size of these impacts are quite large, 19. Bivariate probit results using controls were generally previous research has suggested that, if the circumstances similar in magnitude and significance, but the bivariate of undernourished children change at a young enough age, probit had trouble converging in a few cases when the almost a complete reversal of stunting is possible (Golden controls were added. Furthermore, unlike in a linear model, 1994). TABLE 4 IV Estimates of Experimental Impact of Migration on Child Health Very Good Much Better Standardized Standardized Stunted Height Underweight Overweight BMI Obese BMI for Parent-rated Health Since Height for Weight for Standardized for Age ≤ 5th BMI for Age for Age 85th–95th Age ≥ 95th Health Last Year Age Age BMI for Age Percentile ≤ 5th Percentile Percentile Percentile Children 0–2 yr old Linear IV: No control variables −0.007 0.230 1.795∗∗∗ 0.438 −1.180 −0.369∗∗∗ 0.220 −0.171 −0.077 (0.169) (0.203) (0.579) (0.868) (1.046) (0.101) (0.174) (0.126) (0.216) Linear IV: Control variables 0.011 0.194 1.855∗∗ 0.370 −1.092 −0.424∗∗∗ 0.334∗ −0.042 −0.186 (0.181) (0.219) (0.787) (0.762) (1.060) (0.147) (0.191) (0.156) (0.243) Bivariate probit: No controls −0.008 0.220 −0.364∗∗∗ 0.278∗∗ −0.161∗∗ −0.073 (0.200) (0.212) (0.078) (0.113) (0.068) (0.199) Subsample size 65 47 51 53 49 51 49 49 49 Children 3–5 yr old Linear IV: No control variables 0.039 0.059 −0.161 1.012∗∗ 1.195∗∗∗ 0.083 −0.072 0.119 0.362∗∗∗ (0.126) (0.134) (0.462) (0.440) (0.459) (0.090) (0.057) (0.090) (0.126) Linear IV: Control variables 0.069 0.032 −0.052 0.901∗∗ 0.878∗ 0.077 −0.019 0.183∗∗ 0.317∗∗ (0.117) (0.138) (0.453) (0.456) (0.465) (0.089) (0.046) (0.082) (0.135) Bivariate probit: No controls 0.040 0.058 0.071 −0.059 0.097 0.357∗∗∗ (0.134) (0.134) (0.079) (0.046) (0.076) (0.100) Subsample size 106 106 96 98 90 96 90 90 90 Children 6–12 yr old Linear IV: No control variables 0.285∗∗∗ 0.111 0.169 0.013 −0.138 −0.009 −0.023 0.003 −0.078 (0.090) (0.105) (0.342) (0.317) (0.323) (0.076) (0.016) (0.071) (0.113) Linear IV: Control variables 0.275∗∗∗ 0.113 0.666∗ 0.402 0.041 −0.046 −0.037 0.003 −0.057 (0.098) (0.101) (0.388) (0.354) (0.351) (0.083) (0.026) (0.077) (0.120) Bivariate probit: No controls 0.290∗∗∗ 0.109 −0.009 −0.018 0.003 −0.078 (0.092) (0.105) (0.081) (0.013) (0.067) (0.114) Subsample size 220 220 204 210 208 204 208 208 208 Children 13–18 yr old Linear IV: No control variables 0.410∗∗∗ −0.022 0.081 0.241 0.242 −0.064 −0.018 0.111 0.032 (0.111) (0.136) (0.299) (0.333) (0.351) (0.082) (0.043) (0.107) (0.134) STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH Linear IV: Control variables 0.330∗∗∗ −0.106 0.186 0.340 0.249 −0.090 −0.021 0.089 0.055 (0.089) (0.123) (0.285) (0.323) (0.325) (0.089) (0.044) (0.120) (0.130) Bivariate probit: No controls 0.386∗∗∗ −0.019 −0.073 −0.033 0.120 0.033 (0.096) (0.122) (0.087) (0.033) (0.118) (0.139) Subsample size 123 123 108 112 111 108 111 111 111 Total sample size 514 496 459 473 458 459 458 458 458 Notes: Standard errors account for clustering at household level and use survey weights. Control variables are child’s gender, age in months, age in months squared, birth order, 73 parent’s age, and parent’s height. Ballot success is used to instrument migration to New Zealand in each regression. ∗ Significant at 10%; ∗∗ Significant at 5%; ∗∗∗ Significant at 1%. 74 ECONOMIC INQUIRY migration, and with this greater food intake hav- to an average annual total household income of ing differential effects depending on the age of 19,840 NZ dollars among unsuccessful lottery children.21 As noted earlier, there is some evi- applicants in Tonga.22 A number of studies find dence that a late transition to solid food and a strong relationship between household income inadequate nutritional content of weaning foods and child health (Case, Lubotsky, and Pax- has resulted in malnutrition during early child- son 2002), thus we first examine whether these hood in the Pacific. International evidence has income increases are likely to be related to the shown nutritional supplementation to only have estimated impacts of migration on child health. an impact on stunting and height under the age In Table 5, we present results from estimating of 3 (Branca and Ferrari 2002; Schroeder et al. the relationship between child health and child 1995). Beyond this age, additional nutrition is and parent characteristics,23 log total house- unlikely to have much impact on height. How- hold cash income,24 log total household imputed ever, excess energy intake through an increase value of own-production,25 and log distance in calories can of course still lead to weight from the nearest doctor26 among all children in increases, as has happened here with the 3- to all households in Tonga (e.g., a combined sam- 5-yr-olds. Interestingly, the large increase in the ple of unsuccessful ballot applicants, successful propensity of being underweight for 0- to 2-yr- ballot applicants still in Tonga, nonapplicants olds is entirely driven by the large increase in and previous household members of successful average height, because it is not accompanied by migrants now in NZ that are still in Tonga). We any change in the average weight of these chil- estimate ordinary least squares (OLS) models dren. In the next section, we examine in more for each of the continuous outcomes and probit detail the evidence for greater resource intake. models for the discrete outcomes and present marginal effects and their associated standard V. HOW MIGHT MIGRATION BE AFFECTING CHILD errors which account for clustering at the house- HEALTH? hold level. The results are only associations, not causal relationships. Nonetheless, if income In this section, we attempt to understand has a strong causal impact on child health, we some of the channels through which these would expect to see a significant association in effects may operate. Returning to Equation (1), these regressions. However, we see there is only we see that health outcomes may change as a a weak relationship between income and most result of changes in material inputs, time inputs, measures of child health.27 The exceptions are and health knowledge. Our data only allow us height for age, where children are significantly to examine the impact of changes in material taller in households with higher cash incomes inputs, although we will discuss how changes in and income from own-production, and stunting, the other two types of factors could relate to our where children in households with higher cash results. incomes are less likely to be stunted. However, Increases in income alter the ability of a the magnitude of these effects are quite small household to purchase food and medical inputs that affect child health production. As shown in McKenzie, Gibson, and Stillman (2009), migra- 22. Total household income includes labor earnings, tion from Tonga to NZ results in large increases agricultural income, pension and investment income, the receipt of social benefits, and the imputed value of own- in earned income among principal applicants. produced foods that are consumed by the household. Re-estimating the main treatment effect model 23. We include all of the covariates from the treatment from that paper to examine the impact on total effects regressions as well as controls for the total number of children in the household, whether the child lives with household income among migrant households both of their parents, and each parent’s years of education. with children, we find that migration increases 24. We also estimate the same models controlling for a annual total household income by approximately quadratic in income. The models using log income best fit 14,990 NZ dollars for these households relative the data and results are qualitatively the same in each case. 25. The value of own-production is imputed using self- reported valuations of own produce consumed in the week 21. In unreported results, we also examined whether before the survey. We control for this separately because impacts are related to the amount of time the children lived own-production is likely to be directly related to child in New Zealand. We find migration has significant impacts anthropometrics due to the different foods consumed by on the same outcomes and that the magnitude of these households with crops versus those without own production. impacts grow linearly with time spent in New Zealand (e.g., 26. This is calculated using GPS data on the location of the average monthly impact equals the total impact reported each household and of each medical center. in Table 4 divided by the mean number of months living in 27. This is the case even if we do not control for parent New Zealand for children in each age-group). characteristics in the regression model. TABLE 5 Correlates of Health Status in Tonga (Probit Marginal Effects for Outcomes (1)–(2), (6)–(9), OLS for Remainder) Very Good Much Better Standardized Standardized Stunted Height Underweight Overweight BMI Obese BMI for Parent-rated Health Since Height for Weight for Standardized for Age ≤ 5th BMI for Age for Age 85th–95th Age ≥ 95th Health Last Year Age Age BMI for Age Percentile ≤ 5th Percentile Percentile Percentile Log total household cash income −0.0200∗∗ −0.011 0.0595∗ −0.046 −0.052 −0.0147∗∗ 0.004 0.013 −0.002 (0.010) (0.017) (0.034) (0.045) (0.050) (0.007) (0.003) (0.010) (0.019) Log total household own-production 0.012 −0.001 0.111∗∗∗ 0.004 −0.055 −0.009 0.001 −0.011 −0.019 (0.017) (0.016) (0.042) (0.053) (0.048) (0.008) (0.003) (0.007) (0.019) Log distance from nearest doctor 0.009 0.023 −0.014 −0.032 −0.057 0.007 0.008 0.021 −0.020 (0.024) (0.024) (0.100) (0.092) (0.095) (0.019) (0.006) (0.016) (0.027) Female dummy −0.034 −0.028 0.096 0.420∗∗∗ 0.399∗∗∗ −0.015 −0.0301∗∗∗ 0.012 0.115∗∗∗ (0.044) (0.040) (0.147) (0.113) (0.133) (0.033) (0.012) (0.029) (0.039) Age in months/12 0.000 −0.001 0.0144∗∗ 0.0180∗∗∗ 0.0140∗∗ −0.00357∗∗∗ −0.00108∗∗∗ 0.001 0.001 (0.001) (0.002) (0.006) (0.006) (0.006) (0.001) (0.000) (0.001) (0.002) Age squared/144 0.000 0.000 −0.00815∗∗∗ −0.00642∗∗∗ −0.00391∗ 0.00166∗∗∗ 0.000428∗∗∗ 0.000 0.000 (0.001) (0.001) (0.002) (0.002) (0.002) (0.000) (0.000) (0.001) (0.001) Birth order position −0.025 0.004 0.071 0.038 0.006 −0.009 −0.007 −0.024 0.031 (0.023) (0.020) (0.076) (0.079) (0.075) (0.014) (0.007) (0.017) (0.025) Number of children in household 0.005 0.002 −0.059 0.007 0.008 0.007 −0.002 −0.005 −0.014 (0.014) (0.016) (0.053) (0.059) (0.054) (0.010) (0.004) (0.011) (0.017) Lives with both parents 0.110 0.066 −0.006 −0.422 −0.015 0.027 −0.006 0.039 0.024 (0.080) (0.075) (0.353) (0.376) (0.375) (0.057) (0.025) (0.039) (0.106) Father’s age −0.003 −0.008 −0.011 −0.004 0.003 0.002 0.000 −0.005 0.009 (0.005) (0.006) (0.021) (0.020) (0.019) (0.003) (0.001) (0.004) (0.006) Mother’s age 0.006 0.0109∗ 0.021 0.029 −0.003 −0.003 0.000 0.003 −0.001 (0.006) (0.006) (0.027) (0.020) (0.022) (0.004) (0.002) (0.004) (0.007) Father’s years of education 0.010 0.008 −0.012 0.056 0.058 0.010 0.0112∗∗∗ −0.014 0.020 (0.017) (0.016) (0.060) (0.053) (0.060) (0.014) (0.004) (0.011) (0.019) Mother’s years of education 0.013 −0.003 0.108∗ 0.124∗∗ 0.042 −0.0222∗ −0.00947∗∗ 0.009 0.006 (0.014) (0.015) (0.057) (0.049) (0.062) (0.012) (0.005) (0.011) (0.021) Father’s height 0.000 −0.001 0.005 0.000 −0.002 0.000 0.000 −0.001 0.000 STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH (0.001) (0.001) (0.003) (0.002) (0.002) (0.001) (0.000) (0.000) (0.001) Mother’s height −0.001 0.002 0.009 0.007 0.004 −0.001 0.001 0.001 0.002 (0.002) (0.002) (0.006) (0.004) (0.005) (0.001) (0.001) (0.001) (0.002) Observations 609 568 540 561 542 540 542 542 542 R-squared 0.09 0.03 0.10 0.17 0.08 0.07 0.17 0.04 0.06 Notes: Robust standard errors in parentheses, clustered at household level. Regressions also control for survey year. ∗ Significant at 10%; ∗∗ Significant at 5%; ∗∗∗ Significant at 1%. 75 76 ECONOMIC INQUIRY with a 100% increase in cash income (income foods are consumed, five of which are com- from home production) associated with a 0.06 posites. These foods are: rice, roots, fruits, and (0.11) standard deviation increase in height for nonroot vegetables, fish, fats, meats, milk, and age and a 1.5 percentage point decrease in the sweets.29 We estimate linear instrumental vari- likelihood of being stunted. However, children ables models for each using whether the house- whose parents earn more cash income are sig- hold was successful in the PAC ballot to instru- nificantly less likely to be reported in very good ment for whether the individual has migrated to health. NZ. These models are estimated with one obser- We then ask whether there is any relationship vation per-child to allow all covariates from the between the health of Tongan children who have second specification of the child health regres- recently immigrated to NZ and the change in sion models to be included in these regressions income that their families experienced as a result as well (results are presented both with and of migrating. Table 6 shows that the magnitude without covariates), and thus the results can be of the change in earnings experienced from interpreted at the impact of migration on the moving to NZ and the number of months they diet of the average child in the sample.30 We have lived in NZ have almost no relationship also estimate a third specification that includes with the health of immigrant children.28 The additional controls for the number of male and only significant associations are that a 100 NZD female adults in the household and the number increase in weekly earnings is associated with of other children. As discussed below, migration a 2.3% point reduction in the likelihood of leads to significant changes in household com- parents reporting their children as having much position which could have a mechanical impact better health than 1 yr ago and 0.9 percentage on the number of daily meals consumed by a point increase in the likelihood of children being household. Although we cannot jointly identify stunted. the impact of both migration and changes in Taken together, these results suggest that, household composition on diet, if the estimates even with the large income gains experienced by of the direct impact of migration are unchanged migrant households, changes in income explain in this specification, we can rule out that changes little of the estimated impact of migration on in household composition are responsible for our child health. Dietary change is another path- findings. way through which migration is likely to affect These results indicate that migration leads child health. Not only is the availability of foods to a significant increase in the consumption hugely different between Tonga and NZ, the of meats, fats, and milk. These changes in relative prices of foods available in both coun- diet are large—consumption of meats and fats tries also differ immensely. Existing literature both almost double while consumption of milk also suggests that major dietary changes occur quadruples—and are robust to including con- for Pacific Islanders following migration to NZ trols for household composition. Although we (Harding et al. 1986). Thus, we next examine cannot directly relate changes in diet to changes whether changes in diet are likely to be related in child health because we do not know which to the estimated impacts of migration on child household members are consuming which food, health. these results suggest that dietary change is Table 7 presents results from estimating the directly related to changes in child health. ATT of migration on diet. Specifically, we col- lect information from all households on whether 29. Roots include taro (swamp taro), taro taruas (chi- any of 30 different foods were eaten by any nese taro), kumara (sweet potato), taamu/kape, yams, cas- member of the family during the day prior to sava/manioc, and potato. Fruits and nonroot vegetables the interview. For 27 of these foods, we also include other vegetables, coconut (fresh and dry), banana, asked during how many meals were these foods mango, pawpaw, and other fruits. Fish includes tinned fish and fresh fish. Fats include corned beef, mutton, and coconut eaten. The list of foods is identical in Tonga and (fresh and dry). Meats include corned beef, mutton, fresh NZ making a direct comparison of diet composi- beef, chicken, pork, and other meat (e.g., sausage). Sweets tion possible. To focus our analysis, we examine are one of the foods where the number of meals is not recorded, thus this is a discrete outcome. the cumulative number of meals in which eight 30. Again, standard errors are presented which account for clustering at the household level and all regressions use 28. Only one Tongan immigrant child is underweight, the appropriate survey weights to account for the sampling thus this outcome is dropped from this analysis. Again, the rates for each group. We also include day of the week results are robust to not controlling for parent characteristics fixed effects in the regressions with covariates to account besides household income. for temporal patterns in food consumption. TABLE 6 Correlates of Health Status in New Zealand (Probit Marginal Effects for Outcomes (1)–(2), (6)–(8), OLS for Remainder) Very Good Much Better Standardized Standardized Stunted Height Overweight BMI Obese BMI for Parent-rated Health Since Height for Weight for Standardized for Age ≤ 5th for Age 85th–95th Age ≥ 95th Health Last Year Age Age BMI for Age Percentile Percentile Percentile Change in total household −0.002 −0.0229∗ −0.025 −0.014 −0.004 0.009∗∗∗ −0.009 0.000 earnings (00s NZD) (0.008) (0.013) (0.023) (0.036) (0.036) (0.004) (0.004) (0.013) Months in New Zealand/12 0.114 0.046 0.009 0.174 0.179 0.017 −0.024 0.065 (0.080) (0.125) (0.307) (0.243) (0.196) (0.039) (0.039) (0.085) Female dummy 0.009 0.060 0.028 0.405∗∗ −0.032 −0.002 −0.047 0.002 (0.052) (0.084) (0.285) (0.153) (0.217) (0.035) (0.051) (0.092) Age in months/12 −0.030 0.013 −0.040 0.081 0.028 0.010 0.017 −0.051 (0.026) (0.044) (0.127) (0.148) (0.129) (0.020) (0.022) (0.046) Age squared/144 −0.037 −0.075 0.400 0.276 0.456 −0.097 −0.057 0.479∗∗ (0.118) (0.180) (0.587) (0.668) (0.575) (0.103) (0.109) (0.235) Birth order position 0.053 −0.027 −0.285 −0.168 0.019 0.026 −0.036 0.024 (0.033) (0.060) (0.188) (0.141) (0.147) (0.024) (0.028) (0.052) Number of children in household −0.017 0.037 0.004 0.104 0.077 0.001 0.028 −0.011 (0.026) (0.049) (0.151) (0.148) (0.104) (0.014) (0.021) (0.043) Lives with both parents Perfect Perfect 2.102 −0.389 −2.373∗∗∗ −0.438∗ Perfect Perfect Predictor Predictor (1.437) (0.682) (0.521) (0.369) Predictor Predictor Father’s age −0.0240∗ −0.016 −0.020 −0.036 0.054 −0.001 0.000 0.024 (0.014) (0.022) (0.046) (0.052) (0.045) (0.006) (0.008) (0.019) Mother’s age 0.0527∗∗∗ 0.018 0.016 −0.036 −0.110∗∗ 0.000 0.004 −0.033 (0.016) (0.024) (0.046) (0.047) (0.048) (0.007) (0.009) (0.020) Father’s years of education −0.020 −0.014 −0.007 0.063 0.001 0.004 −0.0224∗∗∗ 0.008 (0.024) (0.023) (0.043) (0.048) (0.037) (0.007) (0.007) (0.016) Mother’s years of education −0.010 0.033 −0.031 0.114 0.101 0.012 −0.0347∗∗ 0.046 (0.021) (0.042) (0.069) (0.080) (0.062) (0.011) (0.016) (0.029) Father’s height 0.000 0.000 0.000 0.004 0.001 0.001∗∗ 0.000 0.000 (0.001) (0.002) (0.003) (0.003) (0.002) (0.001) (0.001) (0.001) Mother’s height 0.00248∗ 0.0181∗ 0.004 0.002 0.002 0.000 0.009∗∗∗ −0.002 STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH (0.001) (0.011) (0.005) (0.004) (0.003) (0.001) (0.002) (0.001) Observations 184 180 166 169 161 166 158 158 R-squared 0.26 0.31 0.11 0.17 0.26 0.22 0.14 0.16 Notes: Robust standard errors in parentheses, clustered at household level. One individual is underweight. One individual is dropped from each discrete model where lives with both parents are a perfect predictor. NZD, New Zealand Dollar. ∗ Significant at 10%; ∗∗ Significant at 5%; ∗∗∗ Significant at 1%. 77 78 TABLE 7 Linear IV Estimates of Experimental Impact on Diet Composition No. of Meals No. of Meals No. of Meals No. of Meals No. of Meals No. of Meals No. of Meals Anyone Ate Rice Roots Fruits/Vegs Fish Fats Meats Milk Sweets Mean unsuccessful ballots 0.224 1.733 2.477 0.580 0.705 1.053 0.448 0.146 Relative price (Pa’anga/NZD) 1.966 0.504 0.769 0.567 0.654 1.262 1.657 NA No control variables −0.097 0.221 1.013∗∗ −0.264∗∗ 0.640∗∗∗ 0.911∗∗∗ 1.121∗∗∗ 0.034 (0.095) (0.191) (0.414) (0.113) (0.185) (0.161) (0.140) (0.089) Main control variables −0.084 0.319 0.424 −0.183 0.649∗∗∗ 0.960∗∗∗ 1.210∗∗∗ 0.047 (0.098) (0.207) (0.433) (0.121) (0.172) (0.170) (0.133) (0.083) Added controls for household size −0.050 0.404∗ 0.098 −0.173 0.611∗∗∗ 0.972∗∗∗ 1.212∗∗∗ 0.009 (0.100) (0.215) (0.450) (0.124) (0.170) (0.170) (0.133) (0.093) Total sample size 528 528 528 528 528 528 528 528 ECONOMIC INQUIRY Notes: Standard errors account for clustering at the household level and all regressions use survey weights. Models with main control variables include controls for the child’s gender, age in months, age in months squared, birth order position, their parents’ age and height, and day of the week fixed effects. The final specifications include additional controls for the number of male and female adults in the household and the number of children. Ballot success is used to instrument for being in NZ in each regression. The market exchange rate is 1.372 Pa’anga per NZD. Roots include taro (swamp taro), taro taruas (chinese taro), kumara (sweet potato), taamu/kape, yams, cassava/manioc, and potato. Fruits and vegetables include other vegetables, coconut (fresh and dry), banana, mango, pawpaw, and other fruits. Fish includes tinned fish and fresh fish. Fats include corned beef, mutton, and coconut (fresh and dry). Meats include corned beef, mutton, fresh beef, chicken, pork, and other meat (e.g., sausage). NZD, New Zealand Dollar. ∗ Significant at 10%; ∗∗ Significant at 5%; ∗∗∗ Significant at 1%. STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 79 Increased consumption of meats and milk would infants in NZ, whereas reductions in physical lead to increased protein and other micronutri- activity might play an important role in explain- ent intake, which have been shown to increase ing the increased BMI of pre-teens. It is also the stature of infants and toddlers (Branca and possible that maternal health knowledge about Ferrari 2002). However, increased consumption nutrition during early childhood may improve of these goods along with fats would lead to an in NZ. increase in overall calorie and fat intakes, which is directly related to weight gain. A number of factors could contribute to VI. CONCLUSIONS changing diets. As mentioned earlier, relative This paper overcomes the selection prob- food prices are quite different in NZ versus lems affecting previous studies of the impact Tonga and most migrant households have expe- of migration on child health by examining a rienced large increases in income. Table 7 also migrant lottery program. The PAC under NZ’s displays the relative Tongan to NZ market price immigration policy allows an annual quota of for each food item. The estimated changes in Tongans to migrate to NZ in addition to those diet are somewhat consistent with relative prices approved through other migration categories, also being a factor—for example, meats and such as skilled migrants and family streams. milk are relatively cheaper in NZ than in Tonga Many more applications are received than the compared to other foods (in particular, roots quota allows, so a ballot is used to randomly and fish). However, we find low cash income select from among the registrations. A unique elasticities for most foods in Tonga.31 Perhaps, survey designed by the authors allows exper- more importantly, the marketing of foods and imental estimates of the impact of migration the availability of different foods is likely to on child health to be obtained by comparing be vastly different between these countries. Fur- the health of immigrant children whose parents thermore, many Tongan households grow or were successful applicants in the ballot to the raise some of their own food, whereas none of health of those children whose parents applied the Tongan migrant households in our survey to migrate under the quota, but whose names do so. were not drawn in the ballot. Overall, these results suggest that dietary Migration is found to affect child health in a change is an important channel through which manner consistent with increased food intake. migration impacts child health and that changes Infants and toddlers suffer less stunting after in income, both the direct effect of these changes migration, whereas 3- to 5-yr-olds gain weight. and their impact on diet, are of limited impor- Older children show no change in anthropomet- tance. Differences in relative prices may explain ric measures, but have better parental reported some of this dietary change, but it seems likely health. Dietary change appears to be an impor- that other important mechanisms are also driv- tant channel through which migration impacts ing this. Another potentially important channel child health, whereas changes in income, both is changes in household structure. For example, the direct effect of these changes and their ATT estimates indicate that the share of adult impact on diet, are of limited importance. Dif- women in migrant households declines by 19 ferences in relative prices may explain some percentage points following migration. We sus- of this dietary change, but it seems likely that pect that having fewer female extended family other important mechanisms, such as changes in members around to help prepare meals could household structure, are also driving this. be a large contributor to a shift toward less It is important to note that there are a number healthy diets. It is also important to note that of other channels through which migration may there are a number of other channels through affect child health that our data do not allow us which migration may affect child health that our to examine. For example, changes in antenatal data do not allow us to examine. For example, practices, such as breastfeeding, might explain changes in antenatal practices, such as breast- the increased stature of infants in NZ, whereas feeding, might explain the increased stature of reductions in physical activity might play an important role in explaining the increased BMI 31. Households with higher cash incomes are not con- of pre-teens. Further research is needed to exam- suming significantly different amounts of fruits, vegetables, ine these effects, as well as to determine inter- milks, or meat. In contrast, consumption patterns do vary with the level of own food production, which does not take ventions which can help lower the rate of obesity place among Tongan households in New Zealand. among older children in immigrant households. 80 ECONOMIC INQUIRY It also must be emphasized that these results Zealand Pacific Children is a Public Health Concern.” reflect the short-run impacts of migration on Journal of Nutrition, 133(11), 2003, 3456–60. Grossman, M. “On the Concept of Health Capital and the child health and therefore it will be important Demand for Health.” Journal of Political Economy, to continue monitoring these children to see 80(20), 1972, 223–55. whether the increases in BMI and obesity per- Harding, W., C. Russell, F. Davidson, and I. Prior. “Dietary Surveys from the Tokelau Island Migrant Study.” sist for the cohort of 3- to 5-yr-olds and whether Ecology of Food and Nutrition, 19(1), 1986, 83–97. similar changes occur for the younger cohort of Heckman, J., N. Hohmann, J. Smith, and M. Khoo. “Sub- 0- to 2-yr-olds once they are too old for addi- stitution and Dropout Bias in Social Experiments: A tional food intakes to impact their height. Study of an Influential Social Experiment.” Quarterly Journal of Economics, 115(2), 2000, 651–94. Hildebrandt, N., and D. McKenzie. “The Effects of Migra- tion on Child Health in Mexico.” Economia, 6(1), REFERENCES 2005, 257–89. Institute of Medicine. “Childhood Obesity in the United Angrist, J. “Treatment Effect Heterogeneity in Theory and States: Facts and Figures.” 2004. Accessed February 8, Practice.” Economic Journal, 502, 2004, C52–C83. 2008. http://www.iom.edu/Object.File/Master/22/606/ Bell, C., and W. Parnell. “Nutrient Intakes of Tongan and FINALfactsandfigures2.pdf. Tokelauan Children Living in New Zealand.” New . From Generation to Generation: The Health and Zealand Medical Journal, 109(1034), 1996, 435–438. Well-Being of Children in Immigrant Families. Wash- Bloom, A. “A Review of Health and Nutrition Issues in the ington DC: National Academy Press, 1998. Pacific.” Asia-Pacific Population Journal, 1(4), 1986, Jamison, D. “Child Malnutrition and School Performance in 17–48. China.” Journal of Development Economics, 20, 1986, Branca, F., and M. Ferrari. “Impact of Micronutrient Defi- 299–309. ciencies on Growth: The Stunting Syndrome.” Annals Jasso, G., D. Massey, M. Rosenzweig, and J. Smith. “Immi- of Nutrition and Metabolism, 46(1), 2002, 8–17. grant Health-selectivity and Acculturation,” in Criti- Carballo, M., J. Divino, and D. Zeric. “Migration and Health cal Perspectives on Racial and Ethnic Differences in in the European Union.” Tropical Medicine & Interna- Health in Late Life, edited by N. Anderson, R. Bulatao, tional Health, 3(12), 1998, 936–44. and B. Cohen. Washington, DC: National Academy Case, A., and C. Paxson. “Stature and Status: Height, Ability Press, 2004, 227–66. and Labor Market Outcomes.” Journal of Political Kanaiaupuni, S., and K. M. Donato. “Migradollars and Economy, 116(3), 2008, 499–532. Mortality: The Effects of Migration on Infant Survival Case, A., D. Lubotsky, and C. Paxson. “Economic Status in Mexico.” Demography, 36(3), 1999, 339–53. and Health in Childhood: The Origins of the Gradient.” Kirchengast, S., and E. Schober. “To be an Immigrant: A American Economic Review, 92(5), 2002, 1308–34. Clemens, M., C. Montenegro, and L. Pritchett. “The Place Risk Factor for Developing Overweight and Obe- Premium: Wage Differences for Identical Workers sity During Childhood and Adolescence?” Journal of across the U.S. Border.” Working Paper No. 148, Biosocial Science, 38(5), 2006, 695–705. Center for Global Development, 2008. Kuczmarski, R., C. Ogden, and L. Grummer-Strawn. “CDC Cole, T. J., J. V. Freeman, and M. A. Preece. “British 1990 Growth Charts: United States.” National Center for Growth Reference Centiles for Weight, Height, Body Health Statistics Advance Data from Vital and Health Mass Index and Head Circumference Fitted by Max- Statistics 314, 2000. imum Penalized Likelihood.” Statistics in Medicine, Lambert, J. “The Effect of Urbanization and Western 17(4), 1998, 407–29. Foods on Infant and Maternal Nutrition in the Colombo, M., I. de Andraca, and I. Lopez. “Mental Devel- South Pacific.” Food and Nutrition Bulletin, 4(3), opment and Stunting.” in Linear Growth Retardation 1982. Accessed December 2009. http://www.unu.edu/ in Less Developed Countries, edited by J. Waterlow. unupress/food/8F043e/8F043E04.htm New York: Raven Press, 1988, 201–14. McKenzie, D., J. Gibson, and S. Stillman. “How Important Frisbie, W., Y. Cho, and R. Hummer. “Immigration and Is Selection? Experimental Vs Non-experimental Mea- the Health of Asian and Pacific Islander Adults in sures of the Income Gains from Migration.” Journal the United States.” American Journal of Epidemiology, of the European Economic Association, 2009. 153, 2001, 372–80. Meulmeister, J., H. Berg, and M. Wedel. “Vitamin D Sta- Fukuyama, S., T. Inaoka, Y. Matsumura, T. Yamauchi, tus, Parathyroid Hormone and Sunlight in Turkish K. Natsuhara, R. Kimura, and R. Ohtsuka. “Anthro- Moroccan and Caucasian Children in the Netherlands.” pometry of 5–19-year-old Tongan Children with Spe- European Journal of Clinical Nutrition, 44(4), 1990, cial Interest in the High Prevalence of Obesity among 461–70. Adolescent Girls.” Annals of Human Biology, 32(6), Popkin, B. M., and J. R. Udry. “Adolescent Obesity 2005, 714–23. Increases Significantly in Second and Third Genera- Gibson, J., and D. McKenzie. “The Impact of an Ex- tion US Immigrants: The National Longitudinal Study Ante Job Offer Requirement on Labor Migration: The of Adolescent Health.” Journal of Nutrition, 128, New Zealand-Tongan Experience,” in International 1998, 701–06. Migration, Economic Development and Policy, edited Raab, G. M., S. Day, and J. Sales. “How to Select Covari- by C. Ozden, and M. Schiff. Washington DC: World ates to Include in the Analysis of a Clinical Trial.” Bank, 2007, 215–33. Controlled Clinical Trials, 21, 2000, 330–42. Golden, M. H. “Is Complete Catch-up Possible for Stunted Rubalcava, L., G. M. Teruel, D. Thomas, and N. Goldman. Malnourished Children?” European Journal of Clinical “The Healthy Migrant Effect: New Findings from the Nutrition, 48(Suppl. 1), 1994, S58–S70. Mexican Family Life Survey.” American Journal of Gordon, F., E. Ferguson, V. Toafa, T.-E. Henry, A. Gould- Public Health, 98(1), 2008, 78–84. ing, A. Grant, and B. Guthrie. “High Levels of Child- Rush, E., L. Plank, and P. Davies. “Body Composition and hood Obesity Observed Among 3- to 7-year-old New Physical Activity in New Zealand Maori, Pacific and STILLMAN, GIBSON & MCKENZIE: IMMIGRATION AND CHILD HEALTH 81 European Children Aged 5–14 Years.” British Journal of Children in Maya Immigrant Families in the U.S.” of Nutrition, 90(6), 2003, 1133–39. Economics and Human Biology, 1(1), 2003, 145–60. Rush, E., L. Plank, V. Chandu, M. Laulu, D. Simmons, Steckel, R. “Stature and the Standard of Living.” Journal of B. Swinburn, and C. Yajnik. “Body Size, Body Com- Economic Literature, 33(4), 1995, 1903–40. position, and Fat Distribution: A Comparison of Young Strauss, J., and D. Thomas. “Health, Nutrition, and Eco- Men of European, Pacific Island, and Asian Indian Eth- nomic Development.” Journal of Economic Literature, nicities.” New Zealand Medical Journal, 117(1207), 36(2), 1998, 766–817. 2004, 1–9. Troiano, R. P., K. M. Flegal, R. J. Kuzmarski, S. M. Camp- Sahn, D., and H. Alderman. “On the Determinants of Nutri- bell, and C. L. Johnson. “Overweight Prevalence and tion in Mozambique: The Importance of Age-specific Trends for Children and Adolescents.” Archive of Pedi- Effects.” World Development, 25(4), 1997, 577–88. atric Adolescent Medicine, 149, 1995, 1085–91. Salesa, J., C. Bell, and B. Swinburn. “Body Size of New Unger, J., K. Reynolds, S. Shakib, D. Sprujit-Metz, P. Sun, Zealand Pacific Islands Children and Teenagers.” New and C. A. Johnson. “Acculturation, Physical Activity, Zealand Medical Journal, 110(1046), 1997, 227–9. and Fast-Food Consumption among Asian-American Schroeder, D. G., R. Martorell, J. Rivera, M. Ruel, and and Hispanic Adolescents.” Journal of Community J. Habicht. “Age Differences in the Impact of Nutri- Health, 29, 2004, 467–81. tional Supplementation on Growth.” Journal of Nutri- WHO Working Group. “Use and Interpretation of Anthro- tion, 125, 1995, 1051–59. pometric Indicators of Nutritional Status.” Bulletin of Sindelar, J., and D. Thomas. “Measurement of Child Health: the World Health Organization, 64(6), 1986, 929–41. Maternal Response Bias.” Yale University Economic WHO, Regional Office for the Western Pacific. “Tonga: Growth Center, Discussion Paper 633, 1991. Health Situation.” 2005. Accessed February 28, Smith, P., B. Bogin, M. Sarela-Silva, and J. Loucky. “Eco- 2007. http://www.wpro.who.int/countries/ton/health_ nomic and Anthropological Assessments of the Health situation.htm.