Food Policy 38 (2013) 70–81

Contents lists available at SciVerse ScienceDirect

Food Policy journal homepage: www.elsevier.com/locate/foodpol

Food consumption patterns and malnourished Indian children: Is there a link? Pushkar Maitra a, Anu Rammohan b,⇑, Ranjan Ray a, Marie-Claire Robitaille b a b

Department of Economics, Monash University, Clayton Campus, VIC 3800, Australia Discipline of Economics, University of Western Australia, Perth 6006, Australia

a r t i c l e

i n f o

Article history: Received 27 January 2011 Received in revised form 24 May 2012 Accepted 7 October 2012 Available online 13 November 2012 Keywords: India Weight-for-height Height-for-age Calorie intake Expenditure patterns

a b s t r a c t Despite its economic success, India has made little progress towards meeting its Millennium Development Goal targets of reducing undernourishment, particularly among children. In this paper, we use nationally representative datasets, the National Family Health Surveys (NFHS II and NFHS III) and the National Sample Survey (55th and the 61st rounds) to analyse the link, if any, between child nutritional outcomes and calorie intakes. Our analysis finds evidence of an improvement in the heightfor-age z-scores, but a worsening in weight-for-height z-scores for children aged 0–3 over the period 1998/1999–2005/2006. There is also evidence of a sharp decline in per adult equivalent calorie intake from the principal food items over roughly this same period. Moreover, this decline was observed across all the expenditure quintiles. Our analysis is therefore able to identify a co-movement of declining nutritional intake for both adults and children and a lack of progress in improving nutritional outcomes of children. Ó 2012 Elsevier Ltd. All rights reserved.

Introduction Despite India’s impressive economic growth in recent years, child nutrition outcomes continue to be poor. The World Bank notes that ‘‘South Asia . . . still has the highest rates and the largest numbers of undernourished children in the world’’, and adds that ‘‘the high economic growth experienced by South Asian countries has not made an impact on the nutritional status of South Asian children’’ (WorldBank, 2011). UNDP (2007–2008) estimates show that 47% (51%) of all Indian children aged below 5 years were classified as being under-weight for age (under-height for age) between 1996 and 2005. Several recent studies have focused on malnutrition among Indian children during this period of rapid economic growth. See for example Lokshin et al. (2005), Gragnolati et al. (2005), Tarozzi and Mahajan (2007), and Pathak and Singh (2011). There is also strong evidence of a decline in per capita calorie consumption in India over the last 20 years (see Ray and Lancaster, 2005; Ray, 2007; Deaton and Dreze, 2009), which has resulted in an increase in the rate of undernourishment. The trends observed in those two strands of the literature suggest that perhaps there is a nexus between the poor nutritional outcomes of young children on the one hand, and the lower calorie consumption of households. Recent evidence from other developing countries like Indonesia

⇑ Corresponding author. E-mail addresses: [email protected] (P. Maitra), anu.rammohan@u wa.edu.au (A. Rammohan), [email protected] (R. Ray), marie-claire.robita [email protected] (M.-C. Robitaille). 0306-9192/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved. http://dx.doi.org/10.1016/j.foodpol.2012.10.004

(Block and Webb, 2009) and China (Osberg et al., 2009) provide evidence of a link between spending patterns, especially food expenditure and child malnutrition. However in the Indian context there has been virtually no attempt to draw a link between the two. In this paper we examine the link, if any, between the declining nutritional intake of Indian households and the lack of progress on child nutritional outcomes. We use information from the National Family and Health Survey (NFHS) and the National Sample Survey (NSS) datasets in a comprehensive analysis that focuses on both child anthropometric measures and household consumption expenditure patterns. To the best of our knowledge, Deaton and Dreze (2009) is the only other study which uses both these datasets to analyse the links between nutrition, calorie intakes and food expenditure. However, their focus is not on any specific group, but rather on the big picture of generally poor nutritional outcomes in India. Our analysis on the other hand focuses on very young children: we examine the nutritional status of children aged 0–3 years, as well as the household expenditure and consumption patterns of households with children in this age group. Malnutrition at an early age has long term implications (McGregor-Grantham, 1995; Martorell and Habicht, 1986; Martorell, 1985). Poor nutrition has implications for a child’s development, since a lack of adequate calories and nutrients to sustain normal growth puts children at a greater risk of being vulnerable to diseases and has adverse effects on their physical, cognitive and mental development (Barker, 1994). It is argued that eliminating malnutrition could cut child mortality by over 50%, and reduce the burden of disease by about 20% (see Murray and Lopez, 1997; Tomkins and Watson, 1989;

P. Maitra et al. / Food Policy 38 (2013) 70–81

Pelletier, 1998). Poor nutrition at childhood also impacts negatively on children’s future productivity (Strauss and Thomas, 1995). We use data from the nationally representative National Family and Health Survey (NFHS) and the National Sample Survey (NSS). The NFHS conducted in 1998/1999 (NFHSII) and 2005/2006 (NFHS III) contain detailed information on child nutrition, which allows us to examine the trends in child nutritional outcomes during a period of rapid economic growth in India. However, in studying child nutrition, it is also important to consider calorie intake and food expenditures by households with young children. Unfortunately, the NFHS data sets do not contain any information on household expenditure patterns or calorie intakes. In order to address this shortcoming, we use data from the 55th (1999/2000) and 61st (2004/2005) rounds of the NSS (NSS 55 and NSS 61 respectively) to examine calorie intakes at the household level. The data used in this paper, therefore, come from parallel and independent surveys on child health and household expenditures that were conducted by different statistical agencies. However, they are both nationally representative. The NFHS II and the NSS 55, on the one hand, and NFHS III and the NSS 61 on the other, were carried out over (almost) contemporaneous periods. This allows us to reconcile the results from NFHS with that from the NSS datasets, although we cannot infer any causality. Our results show that while the height-for-age z-scores have improved for rural children less than 3 years of age between the years 1998 and 2006, the weight-for-height z-scores have worsened for these same children. Estimates from the NSS datasets show that over the same period, there has been a switch away from food to non-food consumption in households with children aged 0–3.1 We also find evidence of a sharp and consistent decline in calorie intake in these households over the study period, suggesting a possible explanation for the worsening of weight-for-height outcomes of children aged 0–3. The rest of the paper is organised as follows. In section ‘Data and descriptive statistics’, we present the data and descriptive statistics for the variables used in the analysis. In section ‘Estimation method and results’, we present our estimation methods and results from our empirical analysis. The final conclusions and policy implications are presented in section ‘Conclusion’.

Data and descriptive statistics NFHS data The NFHS data sets are nationally representative and provide a 3-year retrospective collection of statistical records on maternal and child health practice and outcomes, along with demographic and economic information on the mothers, their children and other selected family members. Our estimating sample contains information on 27,411 children aged 0–3 years (15,104 children from NFHS II and 12,307 from NFHS III). The analysis is based on questions from the women’s questionnaire, which in NFHS II was administered to every-married women aged 15–49 years and to all women aged 15–49 years in NFHS III.2 The sample is restricted to children residing in rural areas of the major states of India.3 Our 1 Our findings are consistent with those of Patnaik (2007), who finds that hunger and deprivation is increasing, especially in rural areas, and that people are purchasing fewer calories. 2 Having children out of wedlock is an extremely rare event in India and, indeed, all children in our sample are born to a married woman. It is therefore unlikely that this difference in sampling procedure affects our results. 3 The states included in our analysis are: Andhra Pradesh, Assam, Bihar, Gujarat, Haryana, Karnataka, Kerala, Madhya Pradesh, Maharashtra, Orissa, Rajasthan, Tamil Nadu, West Bengal, Himachal Pradesh, Punjab and Uttar Pradesh.

71

key variables of interest are the two anthropometric measures of child nutrition: height-for-age and weight-for-height z-scores. In an influential article, Waterlow et al. (1997) established that height-for-age and weight-for-height are good indicators of a child’s nutritional status. A child’s height-for-age is an indicator of his/her long-run nutritional status, reflecting the child’s past nutritional experience, while a child’s weight-for-height is regarded as an indicator of short-run or current nutritional status. We use the z-score method, with the reference population being the commonly used US National Center for Health Statistics (NCHSs) standard.4 The z-scores have an important advantage over simple measures of height and weight in that they are less sensitive to changes at the extremes of the distribution of these variables. They also facilitate comparisons across measures that use different units of measurement. Finally, the use of z-scores makes it possible to pool children of different ages and gender. A negative z-score indicates that the child’s nutritional status is worse than the nutritional status of the average child in the reference population. Descriptive statistics for key variables used in the analysis are presented in Table 1. Over the period 1998/1999–2005/2006, although the average height-for-age z-score of rural Indian children aged 0–3 years has improved from 1.94 to 1.59, it continues to remain below the reference median. The weight-for-height z-scores, however, have worsened during this period from 0.86 standard deviations in NFHS II to 1.08 standard deviations in NFHS III. Specifically, the proportion of children wasted (defined as weight-for-height z-score being <2) has increased from 17% to 19% over the period under consideration, while the proportion of children stunted (defined as height-for-age z-score being <2) has decreased from 49% to 40% over the same period. In Fig. 1 we present the non-parametric locally weighted regressions of the height-for-age and weight-for-height z-scores on the age of the child. As discussed above, both the height-for-age and the weight-for-height z-scores of Indian children are below those of the reference population. The curves in every case decline rapidly until 18 months of age and then stabilise around 2 for the height-for-age z-scores, and around 1 for the weight-for-height z-scores. Beyond 18 months, the relationship between a child’s age and z-score is fairly non-monotonic, irrespective of the anthropometric measure considered. However, the extent of wasting is much less than the extent of stunting in the sample, and the extent of wasting also decreases for older children. Next, we examine changes in the distribution of z-scores. In Fig. 2, we present the kernel density estimates of the height-forage and weight-for-height z-scores for children aged 0–3 years. Using the Kolmogorov–Smirnov test, we always reject the null hypothesis that the distributions are the same over the two survey rounds. The mass of the distribution for the height-for-age z-scores for the NFHS III dataset lies to the right of the NFHS II dataset, indicating substantial improvement in height-for-age z-scores (regarded as a measure of long-term nutrition in children). With respect to the weight-for-height z-scores, the mass of the distribu-

4 The z-score of child i is the difference between the observed value for child i and the median value of the reference population, all divided by the standard deviation of the reference population. Our results remain unchanged if we use the mean value of the reference population as opposed to the median. In this paper we consider two alternative measures: the height-for-age and the weight-for-height z-scores, which reflects the long-term and short-term nutritional status of the children. A third measure is the weight-for-age z-score. However 95% of the variance in the weightfor-age z-score is explained by the variance in the height-for-age z-score and the variance in the weight-for-height z-score (Keller, 1983; cited in WHO, 1986), and so in our analysis we restrict ourselves to the height-for-age and the weight-for-height zscores. The z-scores are used to describe how far a measurement is from the median (average), conditioning on the age/gender of the child. See http://www.who.int/ childgrowth/training/module_c_interpreting_indicators.pdf.

72

P. Maitra et al. / Food Policy 38 (2013) 70–81

Table 1 Sample means for key variables used in the analysis (NFHS II and NFHS III).

Table 1 (continued) NFHSII (1998)

Sample size Height-for-age Weight-for-height Height-for-age <3 Height-for-age from 3 to 2 Height-for-age from 2 to 1 Height-for-age from 1 Height-for-age <3 Weight-for-height from 3 to 2 Weight-for-height from 2 to 1 Weight-for-height from 1 to Male Age of child: 0–6 Age of child: 7–12 Age of child: 13–18 Age of child: 19–24 Age of child: 25–29 Age of child: 30–36 Number of sisters Number of brothers Birth weight low Birth order = 1 Birth order = 2 Birth order = 3 Birth order = 4 Age of mother at birth: 19 Age of mother at birth: 20–24 Age of mother at birth: 25–29 Age of mother at birth: 30–34 Age of mother at birth: 35 or higher Mothers education no education Mothers education primary schooling Mothers education secondary or higher Mother is wife of the household head Fathers education no education Fathers education primary schooling Fathers education secondary or higher Know ORS Use ORS Wealth quintile: poorest (Q1) Wealth quintile: poor (Q2) Wealth quintile: middle (Q3) Wealth quintile: rich (Q4) Wealth quintile: richest (Q5) Household has radio Household has television Household has access to pipe water Hindu Other caste Andhra Pradesh Assam Bihar Gujarat Haryana Himachal Pradesh Karnataka Kerala Madhya Pradesh Maharashtra Orissa Rajasthan Tamil Nadu Uttar Pradesh West Bengal Punjab Child was breastfeed for 6–24 month Child was given: water Child was given: milk Child was given: green vegetable Child was given: fruit Onset of breastfeeding: 1 h or less Onset of breastfeeding: 1 h to 1 day Onset of breastfeeding: more than 1 day At least 1 household member smokes Mothers BMI: <16.5 Mothers BMI: 16.5–18.5

NFHSII (1998)

NFHS III (2005)

15118 1.939 0.863 0.259 0.229 0.239 0.273 0.032 0.134 0.318 0.516 0.524 0.210 0.158 0.196 0.133 0.186 0.117 0.751 0.641 0.253 0.272 0.252 0.184 0.292 0.229 0.399 0.238 0.094 0.040 0.591 0.157 0.252 0.492 0.312 0.180 0.508 0.609 0.062 0.258 0.248 0.232 0.184 0.078 0.339 0.232 0.094 0.841 0.665 0.045 0.034 0.126 0.041 0.044 0.039 0.047 0.029 0.109 0.043 0.068 0.123 0.043 0.129 0.045 0.036 0.181 0.747 0.419 0.254 0.163 0.150 0.229 0.611 0.539 0.101 0.312

12307 1.585 1.084 0.179 0.223 0.263 0.335 0.035 0.157 0.370 0.438 0.521 0.171 0.169 0.173 0.153 0.175 0.160 0.739 0.610 0.224 0.293 0.274 0.172 0.261 0.197 0.423 0.243 0.093 0.043 0.506 0.149 0.345 0.486 0.290 0.159 0.550 0.712 0.055 0.294 0.249 0.210 0.164 0.083 0.255 0.306 0.104 0.810 0.635 0.030 0.047 0.112 0.041 0.043 0.032 0.047 0.030 0.121 0.039 0.055 0.062 0.034 0.211 0.059 0.037 0.202 0.828 0.407 0.232 0.221 0.218 0.268 0.381 0.215 0.104 0.311

Mothers BMI: 18.5–25 Mothers BMI: 25–30 Mothers BMI: >30 Mothers anaemia level: Mothers anaemia level: Mothers anaemia level: Mothers anaemia level: Diarrhoea

mild moderate severe missing

0.560 0.023 0.003 0.157 0.165 0.017 0.045 0.198

NFHS III (2005) 0.542 0.036 0.006 0.414 0.186 0.017 0.012 0.127

tion for the NFHS III sample lies to the left of the NFHS II sample, indicating a worsening of the overall distribution. In Fig. 3 we present the difference in cumulative distribution functions for the two survey rounds. For a given value z of z-scores, letting F denote the cumulative distribution function while the subscript denotes the survey round, we compute the differences in distributions as FIII(z)  FII(z). Improvements are reflected as negative numbers. Fig. 3 shows that while there has been an improvement in the height-for-age z-scores, this is not the case with the weight-for-height z-scores, supporting the results presented in Table 1 and Fig 2. Turning to the descriptive statistics for the other key variables (Table 1), between 1998/1999 and 2005/2006, there have been large improvements in the educational attainment of both parents in our sample, particularly that of mothers: the proportion of mothers with no schooling fell from 59% to 51% between the two periods. The proportion of fathers with no schooling decreased from 31% to 29%. There is also an improvement in the proportion of parents (both mothers and fathers) who have attained secondary schooling and above. In 1998/1999, 25% (51%) of the mothers (fathers) had education levels of at least secondary schooling. By 2005/2006, this figure has increased to 35% for mothers and 55% for fathers. We observe that the proportion of children residing in households from the lowest and highest wealth quintiles has increased (from 0.26 to 0.29 for Q1 and from 0.78 to 0.84 for Q5), while there has been a decrease in the proportion of children residing in households belonging to the intermediate wealth quintiles. Other factors such as maternal health status may also influence child nutritional outcomes. In particular, maternal iron deficiency (anaemia) increases the risk of pre-term labour, low birth weight, infant mortality, and predicts the likelihood of iron deficiency in infants after 4 months of age (Brabin et al., 2001). Maternal anaemia is classified as a severe public health problem in India by the World Health Organisation (WHO, 2008). In our sample, 18% of the mothers were in the moderately or severely anaemic category in 1998/ 1999, and this number increased to 20% in 2005/2006. Similarly, low maternal body-mass index (BMI) can influence child nutritional outcomes. Typically an individual with a BMI under 18 is regarded as being underweight and possibly malnourished. Around 41% of the mothers in our sample have a BMI below 18.5. The proportion of mothers with BMI below 16.5 and BMI in the range of 16.5– 18.5, has remained stable over the two survey years. Turning to child feeding practices, several results are noteworthy. First, we observe that the proportion of children that were breast-fed between 6 and 24 months has increased from 18% to 20% between the two survey years. However, we see a decline in the proportion of children who were given milk and green vegetables in the 7 days prior to the survey. NSS data Next, using data from the NSS 55 and NSS 61, we describe the household expenditure patterns for a variety of food items, including both cereal and non-cereal items such as milk, milk products,

73

P. Maitra et al. / Food Policy 38 (2013) 70–81

HAZ: Rural

-1 -2

Locally weighted regression; bw = 0.5

-3

-3

-2

-1

Locally weighted regression; bw = 0.5

0

0

WHZ: Rural

0

6

12

18

24

30

36

0

6

12

18

24

30

36

Age

Age NFHS II

NFHS III

NFHS II

NFHS III

Fig. 1. Lowess plots of HAZ and WHZ by year.

HAZ: Rural .4 .3 .2

Kernel Density Estimates

0

.1

.3 .2 .1 0

Kernel Density Estimates

.4

WHZ: Rural

-4

-2

0

2

4

6

NFHS II

-5

0

5

z-score

z-score NFHS III

NFHS II

NFHS III

Notes: Kolmogrov-Smirnov test p-values: HAZ: Combined K-S 0.0534; p-value = 0.0000 WHZ: Combined K-S 0.0739; p-value = 0.0000 Fig. 2. Kernel density estimates of HAZ and WHZ by year. Notes: Kolmogrov–Smirnov test p-values: HAZ: combined K–S 0.0534; p-value = 0.0000. WHZ: combined K–S 0.0739; p-value = 0.0000.

protein and pulses. Both the NSS 55 and the NSS 61 contain information on the physical quantities of consumption of different food items. The calorie intake from food items was obtained by using item specific calorie conversion figures provided by the FAO (2011), and from the calorie conversion tables that are available in Gopalan et al. (1999). The latter also provided the conversion factors of fat and protein that were used to calculate the intake of these nutrients from the principal food items and milk products. Finally to compute the adult equivalent calorie intake, we have used the age and gender specific calorie requirements of the Indian population as prescribed by the Indian Council of Medical Research

(ICMR, 2011). These requirements are presented in Table 10 (see also ICMR, 2002). Consider, for example, a household consisting of two males aged 61and 35 years, two females aged 55 and 30 years, one girl aged 9, and two boys aged 5 and 2 years. Then the adult equivalent household size (in terms of nutritional requirement) is 1950/2800 (male aged 61) + 2200/2800 (female aged 55) + 1 (male aged 35) + 2200/2800 (female aged 30) + 21 00/2800 (girl aged 9) + 1500/2800 (boy aged 5) + 1200/2800 (boy aged 2) = 4.982. In the calculation of the POU rates, the equivalences implicit in these calorie requirements were automatically taken into account,

74

P. Maitra et al. / Food Policy 38 (2013) 70–81

HAZ: Rural

.05 -.1

-.1

-.05

0

Difference by Year

0 -.05

Difference by Year

.05

.1

.1

WHZ: Rural

-5

-4

-3

-2

-1

0

1

2

3

4

5

-5

z-score

-4

-3

-2

-1

0

1

2

3

4

5

z-score

Fig. 3. Difference in cdf of HAZ and WHZ. Improvements are denoted by negative numbers.

since we calculated required calories as being equal to calorie requirements for a person in the age-gender group multiplied by the number of household members in that age-gender group summed over all possible age-gender groups within the household. This way a household is considered to be undernourished if its calorie intake is less than the minimum calorie requirement. In our analysis we restrict our analysis to rural households with at least one child aged 0–3 years. Our final sample then consists of 17,223 households in NSS 55 and 16,732 households in NSS 61. Table 2 presents the descriptive statistics for the key variables in the two survey years. There does not appear to be a great deal of movement across a large set of socio-economic variables. The two main exceptions are land holding and the proportion of workers employed in agriculture-related activities, both of which have declined sharply over the two survey years. The fall in agriculture as a source of income over this period (see Bardhan, 2005), land fragmentation among siblings, and increased distress land sales are possibly some of the reasons for the decline in the size of land holding.5 In addition while there is a slight decline in the years of schooling of the household head, there is a large increase in both per capita and per adult equivalent expenditure over this period. This appears to suggest households in rural India have indeed become richer. The poverty rate has remained stable at around 20% over the period, though there is a slight increase in the percentage of under nourished households (defined as households for whom the actual calorie intake is less than the required calorie intake). In Table 3, we compare the average (in per-adult equivalent terms) consumption of four major food items and in calorie intake of households with at least one child aged 0–3 years for the two survey years. These figures portray a rather depressing picture. There has been a decline in per-adult equivalent consumption of

5 Note that the definition of land holding has changed between the 55th and the 61st rounds of the NSS. For the 55th Round, land holding refers to total land owned, whereas for the 61st round we are subtracting the land leased out from total land possessed, which may explain the drop in land holding across two rounds. However, even when we re-compute the size of land holdings using a consistent definition, we still observe a large drop in the mean land holding size. However, since this issue is not central to our analysis, and therefore we do not dwell on this any further.

all the major food items, and consequently a 90 kcal a day decline in calorie input (again in per-adult equivalent terms). These comparisons could potentially provide important insights on the observed nutritional outcomes of children aged 0–3 years. To summarise, we find evidence of a large decrease in calorie intake over the period 1999/2000–2004/2005 for households with young children (0–3 years), which is consistent with the decline in nutritional outcomes observed in the NFHS data. This is an issue of considerable concern since this was a period of rapid economic growth in India, and could be indicative of increasing inequality and a worsening of living conditions in rural India. Estimation methods and results Methods Using NFHS data, we estimated OLS models for weight-forheight and height-for-age z-scores for a pooled sample of children aged 0–3 years in the NFHS II and NFHS III data sets (Panel A). We also estimated a probit model of the likelihood of a child being stunted (height-for-age z-score <2), and being wasted (weightfor-height z-score <2). Given the focus of our paper, the primary variable of interest is the NFHS III dummy, which is presented in Table 4. The complete set of results is presented in Table A1 in the Supplementary appendix. All of the regressions control for an extensive set of individual, parental, household and community characteristics. They include the child’s age in months, the number of male and female siblings, the child’s birth-weight and birth-order, the age of the mother at birth, the highest educational attainment of the mother and the father, the mother’s knowledge of and experience in using ORS, the wealth quintile of the household, whether the household has a television and a radio, whether the main source of drinking water is piped water, the religion and caste of the household, whether the mother is underweight, overweight or obese and the mother’s anaemia status. To control for seasonal changes in food availability, we also include the month of measurement. Finally, a set of state dummy variables are included to control for any state specific

P. Maitra et al. / Food Policy 38 (2013) 70–81 Table 2 Descriptive statistics for key variables used in the analysis (NSS 55th and NSS 61st rounds).

Sample size Hinduism Islam Christianity Sikhism Jainism Buddhism Household type: non-agricultural Household type: agriculture labour Household type: other labour Household type: self employed in agriculture Scheduled tribe (ST) Scheduled caste (SC) Other backward classes (OBCs) Years of schooling household head Male household head Age of household head Marital status Household size Household size per adult equivalent Land holding (in hectare) Per capita expenditure (in Rs.) Expenditure per adult equivalent (in Rs.) Poverty rate POU rates Andhra Pradesh Assam Bihar Gujarat Haryana Karnataka Kerala Madhya Pradesh Maharashtra Orissa Rajasthan Tamil Nadu West Bengal Uttar Pradesh Himachal Pradesh Punjab

NSS 55

NSS 61

17,223 0.825 0.127 0.016 0.023 0.001 0.004 0.163 0.279 0.080 0.392 0.108 0.199 0.389 3.777 0.951 43.340 0.92 6.952 5.08 4.092 449.019 618.981 0.31 0.89 0.060 0.052 0.136 0.038 0.021 0.036 0.030 0.094 0.059 0.046 0.067 0.043 0.065 0.203 0.020 0.031

16,732 0.820 0.128 0.016 0.027 0.001 0.004 0.254 0.153 0.122 0.359 0.109 0.199 0.426 3.208 0.947 44.136 0.93 6.854 5.03 1.562 522.377 716.841 0.30 0.91 0.056 0.039 0.134 0.035 0.027 0.032 0.037 0.101 0.066 0.054 0.068 0.039 0.065 0.190 0.028 0.031

Notes: POU rates are the fraction of people who are undernourished. Calculation of POU rates is based on the criterion of ‘actual calories < required calories’.

Table 3 Differences in food consumption across the two NSS rounds (households with a child aged 0–3 years). Per adult equivalent consumption of

NSS 55

NSS 61

Rice (kg/month) Wheat (kg/month) Other cereals (kg/month) Pulses (kg/month) Milk (kg/month) Milk products (kg/month) Egg, fish, meat (kg/month) Vegetables (kg/month) Fresh fruits (kg/month) Calorie (kcal/day)

8.139 6.313 2.009 1.091 7.776 0.909 0.987 7.255 1.846 2092.6

7.699 5.657 1.911 0.929 7.194 0.646 0.899 6.634 1.630 2002.5

policies that can have an effect on child nutritional outcomes (the reference category is that the child resides in Uttar Pradesh, the most populous state in India). The NFHS based estimations of child anthropometric measures are followed by the NSS based OLS estimations of key results for the expenditure shares of four major categories (food, education, medical institutional and medical non-institutional) in Table 6 (Panel A), per adult equivalent consumption of cereals (Table 6 Panel B), per adult equivalent consumption of milk, milk products, meat, vegetables, and fruit. In Table 7 we report the results on nutritional

75

intake (the per adult equivalent calorie, protein and fat consumption in kcal/month) from the principal food items and for milk products. Principal food items consists of rice, wheat pulses, vegetables and mil. Again the regressions control for a full set of household and community characteristics. Note that all analysis using the NSS data is conducted at the household level, with the sample restricted to households that have at least one child aged 0–3. The full set of results are presented in Tables A5–A8 in the supplementary appendix. Below we discuss the main findings of our analysis, focusing first on the NFHS datasets and then on the NSS datasets. Anthropometric measures The regression results presented in Table 4 (Panel A) show that even after controlling for a large number of observables that can potentially affect child nutritional outcomes, overall height-forage z-scores are higher in 2005/2006 (NFHS III is positive and statistically significant), whereas the weight-for-age z-scores have significantly decreased in 2005/2006 (NFHS III dummy is negative and statistically significant). More specifically, the average heightfor-age z-score is 0.38 standard deviations higher in 2005/2006, while the average weight-for-height z-score is 0.21 standard deviations lower. Given that the average height-for-age z-score is 1.94 in 1998/1999, this amounts to an increase of 20% in average height-for-age z-scores over the relevant period, which cannot be accounted for by the other control variables used in the model. On the other hand, given that the average weight-for-height zscore was 1.59 in 1998/1999, there has been a 20% worsening of the average weight-for-height z-score over the relevant period. The marginal effects from the probit regressions for stunting and wasting (Panel B) show that children are 11-percentage points less likely to be stunted in 2005/2006 but are 2-percentage points more likely to be wasted in 2005/2006, and the effect is statistically significant in both cases. The results then corroborate the descriptive statistics presented earlier. How do some of the other observables affect the height-for-age and weight-for-height z-scores? The full set of results presented in Table A1 in the appendix show that boys have higher height-forage z-scores, but there is no statistically significant difference between boys and girls in terms of weight-for-height z-scores. In other words, while the long-term nutritional status of boys is better than that of girls, there are no gender differences in short-term nutritional outcomes. Not surprisingly, low birth-weight is negatively associated with both height-for-age and weight-for-height z-scores in the OLS estimates. Similarly the probit estimates also show that low birth weight significantly increases the probability of a child being stunted or wasted. While a child’s birth order has no statistically significant influence on their weight-for-height, we note that relative to a first-born child, a child who is later born has significantly lower height-for-age z-score. This is possibly indicative of sibling competition for scarce resources leading to poor long-term health or maternal depletion. Several of the maternal characteristics are influential in a child’s anthropometric outcomes. Specifically, relative to having a mother with no education, having a mother educated at primary, secondary and above levels of schooling, significantly improves heightfor-age z-scores. A mother’s secondary education similarly has a positive and significant influence on a child’s weight-for-height z-score. Not surprisingly, household wealth significantly improves both weight-for-height and height-for-age z-scores. Relative to a child born in the richest wealth quintile, children from the other wealth quintiles have poorer anthropometric outcomes. The nutritional outcomes of young children are likely to be closely linked to maternal health. We included among our explanatory variables the mother’s BMI category, and anaemia status to account for the influ-

76

P. Maitra et al. / Food Policy 38 (2013) 70–81

Table 4 Regressions for z-scores, stunting and wasting. Panel A (OLS)

Height-for-age z-score

NFHS III (year = 2005–2006)

0.378

***

Weight-for-height z-score 0.206*** (0.017)

(0.022)

Panel B (probit)

Stunting

Wasting

NFHS III (year = 2005–2006)

0.107*** (0.008)

0.017*** (0.006)

Observations

27,425

27,507

Notes: Stunting is defined as having a height-for-age z-score <2; Wasting is defined as having a weight-for-height z-score <2. Figures in parenthesis are standard errors clustered at the mother level. Regressions control for a full set of individual, parental and household characteristics. Full set of results are presented in Table A1 in the Appendix.  p < 0.05.  p < 0.1. *** p < 0.01.

Table 5 Robustness of results to alternative estimation methods.

Panel A: Ordered probit estimatesa NFHS III (year = 2005/2006) Coefficient estimate Marginal effects

Severely stunted/wasted Moderately stunted/wasted Mildly stunted/wasted Normal

Height-for-age z-score

Weight-for-height z-score

0.278*** 0.074*** 0.036*** 0.018*** 0.092***

0.141*** 0.009*** 0.026*** 0.021*** 0.056***

(0.017) (0.005) (0.002) (0.001) (0.006)

(0.017) (0.001) (0.003) (0.003) (0.007)

Panel B: SUR estimatesb NFHS III (year = 2005/2006)

0.378*** (0.023)

0.208*** (0.018)

Panel C: OLS regressions excluding the six northern states NFHS III (year = 2005/2006)

0.292*** (0.031)

0.184*** (0.025)

0.376*** 0.386*** 0.368*** 0.344*** 0.468***

0.159*** 0.206*** 0.218*** 0.243*** 0.331***

Panel D: OLS regressions by wealth quintile NFHS III (year = 2005/2006)

Poorest (Q1) Poor (Q2) Middle (Q3) Rich (Q4) Richest (Q5)

(0.044) (0.046) (0.048) (0.051) (0.071)

(0.033) (0.035) (0.038) (0.042) (0.060)

Notes: Figures in parenthesis are standard errors clustered at the mother level. Regressions control for a full set of individual, parental and household characteristics.  p < 0.05.  p < 0.1. *** p < 0.01. a Full set of results are presented in Supplementary Appendix, Tables A2 and A3. b Full set of results are presented in Supplementary Appendix, Table A4.

Table 6 Pooled OLS regressions: expenditure and inputs. Panel A: Expenditure shares (four major categories)a Food NSS 61st round (Year = 2005–2006) 0.018*** (0.001)

Education 0.010*** (0.001)

Medical-institutional 0.008*** (0.001)

Medical-non-institutional 0.004*** (0.001)

Panel B: Per adult equivalent consumption of cerealsb Rice (kg/month) NSS 61st round (Year = 2005–2006) 0.513*** (0.052)

Wheat (kg/month) 0.990*** (0.053)

Other cereals (kg/month) 0.104*** (0.040)

Pulses (kg/month) 0.244*** (0.033)

Panel C: Per adult equivalent consumption of milk, milk products, meat, vegetables and fruitsc Milk (kg/ Milk products (kg/ Egg, fish and meat (kg/month) month) month) NSS 61st round (year = 2005– 1.345* (0.806) 0.311** (0.132) 0.221*** (0.058) 2006)

Vegetables (kg/ month) 0.953*** (0.082)

Fresh fruits (kg/ month) 0.270*** (0.063)

Notes: Figures in parenthesis are robust standard errors. Regressions control for a full set of individual, parental and household characteristics. a Full set of results are presented in the Supplementary Appendix Table A5. b Full set of results are presented in the Supplementary Appendix Table A6. c Full set of results are presented in the Supplementary Appendix Table A7. *** p < 0.01. ** p < 0.05. * p < 0.1.

ence of maternal health on child nutrition. Our analysis shows that relative to having a mother with BMI in the normal range, a child

whose mother is severely underweight (BMI < 16.5) or underweight (BMI 2 [16.5, 18.5]), has lower weight-for-height and

77

P. Maitra et al. / Food Policy 38 (2013) 70–81 Table 7 SUR estimates for per adult equivalent daily intake of calorie, protein and fat. Calorie Panel A: Per adult equivalent daily intake of calorie, protein and fat from principal food items NSS 61st round (year = 2005–2006) 193.5*** (8.923)

Protein

Fat

0.291*** (0.024)

0.116*** (0.036)

0.133** (0.055)

0.143** (0.059)

a

b

Panel B: Per adult equivalent daily intake of calorie, protein and fat from milk products NSS 61st round (year = 2005–2006) 2.882 (2.029) Notes: Figures in parenthesis are robust standard errors. Regressions control for a full set of individual, parental and household characteristics. Principal food items consist of rice, wheat, pulses, vegetables and milk.  p < 0.1. a Full set of results are presented in the Supplementary Appendix Table A8. b Full set of results are presented in the Supplementary Appendix Table A9. *** p < 0.01. ** p < 0.05.

height-for-age z-scores. Similarly, having a mother with moderate or severe anaemia rather than being in the normal range, is negatively associated with child’s weight-for-height and height-for-age z-scores. This relationship between maternal and child health becomes more important when we discuss some of the channels through which a decline in nutritional intake within households can affect the health/nutritional outcomes of children. Robustness to alternative estimation techniques The results are robust to alternative estimation methods. First, rather than using a binary classification (stunted or not and wasted or not), following Kassouf and Senauer (1996) we categorise the nutritional status of children into four categories, namely: (1) Severely wasted/stunted: z-score is less than 3; (2) Moderately wasted/stunted: z-score lies in the interval (3, 2); (3) Mildly wasted/stunted: z-score lies in the interval (2, 1); (4) Normal: z-score >1. Given that there is a natural ordering of these categories, the appropriate model to use in this case is the ordered probit model. The marginal effects associated with the NFHS III dummy variable from the corresponding ordered probit regressions are presented in Table 5, Panel A. They show that relative to 1998/1999, in 2005/2006 children were significantly more likely to be in the mildly stunted category (1.8 percentage points) and in the normal category (9.2 percentage points) and significantly less likely to be in the severely and moderately stunted categories. On the other hand, with regards to wasting, relative to 1998/1999, in 2005/2006 children are significantly more likely to belong to the severely (0.8 percentage points), moderately (2.6 percentage points) or mildly (2 percentage points) wasted categories; and significantly less (5.5 percentage points) likely to belong to the normal weight category. Second, we present estimates from a SUR system of regressions for the height-for-age and weight-for-height z-scores. This allows the errors in the two equations to be correlated, since they relate to the same child. As an additional (identifying) variable in the weight-for-height equation, we include a dummy variable taking the value of one if the child has had diarrhoea in the 2 weeks prior to the survey. Diarrhoea has been shown to have an impact on short-term nutrition but is not expected to influence a child’s long-term nutrition, except in chronic cases.6 The SUR estimates presented in Table 5, Panel B are almost identical to the corresponding OLS estimates of height-for-age and weight-for-height z-scores presented in Table 4.

literature which argues that the liberalisation process has not been uniform across the country (see for example, Basu and Maertens, 2007; Siggel, 2010). According to the WorldBank (2006), the rural areas of some Indian states (such as Bihar and Orissa) possess levels of poverty and food insecurity comparable to the poorest nations in sub-Saharan Africa, whilst others (such as Punjab and Kerala) are similar to middle-income nations. Menon et al. (2008) find that although India’s overall rank on the Global Hunger Index (GHI) was 66 (a higher figure indicates a poorer hunger index), there is substantial heterogeneity across the major states in India. For example, the GHI ranges from 34 for Punjab (placing it between Nicaragua and Ghana) to 82 for Madhya Pradesh (placing it between Chad and Ethiopia). Around 10 of the 17 individual Indian states ranked above the Indian average.

Geographical differences India is a heterogeneous country in terms of attitudes, food habits and in the overall standards of living. There is a fairly large 6 To obtain the standard errors adjusted for clustering at the mother level, we bootstrap the standard errors (with 1000 replications). The full set of results are presented in the supplementary Appendix (Tables A2–A4). Due to space considerations we do not discuss these results.

Fig. 4. Estimated effect for the NFHS III dummy for the different states.

78

P. Maitra et al. / Food Policy 38 (2013) 70–81

These findings are also consistent with our analysis, and the OLS estimates presented in Table A1 show a fair amount of geographical variation in child nutritional outcomes among the Indian states. To explore this further, we compute OLS regressions for height-for-age z-score and weight-for-height z-score separately for each of the states included in our sample. The estimated coefficients (and the 95% confidence interval) of the NFHS III dummy, are presented in Fig. 4. The largest improvements in both the height-for-age z-scores and the weight-for-height z-scores have occurred in the state of Haryana. Indeed there has been a significant improvement in the long-term health status of children in all the northern states (Madhya Pradesh, Rajasthan, Uttar Pradesh, Punjab, Haryana and Himachal Pradesh); considerably less so in the other states. To examine if the results, particularly the improvements in long-term nutrition (height-for-age z-scores), is driven by the performance in the six northern states, we re-estimated the OLS regressions for heightfor-age and weight-for-height, excluding children who reside in these six northern states. The coefficient estimates from these regressions are presented in Table 5, Panel C. Notice that while still statistically significant, the coefficient estimate of the NFHS III dummy for the height-for-age z-score regression is lower (declining from 0.38 to 0.29). Excluding these northern states does not have any influence on the magnitude and statistical significance of the NFHS III dummy in the weight-for-height z-score regression. These results indicate that while most of the progress in terms of children’s long-term nutrition has occurred in the northern states, the worsening of short term nutritional status does not follow any clear geographical pattern. Differences by wealth quintiles It is also worth examining whether the changes in wasting and stunting are occurring in specific socio-economic strata. We therefore re-estimate the OLS regressions for height-for-age and weightfor-height z-scores and the probit regressions for stunting and wasting, separately for the different wealth quintiles: poorest (Q1), poor (Q2), middle (Q3), rich (Q4) and richest (Q5). In Table 5 Panel D we present the coefficient estimates for the NFHS III dummy for the five quintiles. The OLS regressions for heightfor-age z-scores suggest (as expected) that the long term nutritional status has increased for children across all socio-economic strata and the effect is the highest for those for the richest household (Q5). On the other hand, the weight-for-height z-scores have declined for children across all socio-economic strata but surpris-

ingly the effect is the strongest for those in the richest households (Q5). Expenditures and nutritional intakes We now move to our analysis of household expenditure patterns using the NSS data sets. For consistency with the NFHS analysis, the sample is restricted to households with at least one child in the 0–3 age category. Table 6 presents the estimated coefficients of the NSS 61 dummy from a set of pooled regressions of the expenditure shares on 4 major items (Panel A); per adult equivalent consumption of cereals (Panel B); and per adult equivalent consumption of non-cereals such as milk, milk products, meat, vegetables and fruits (Panel C). Table 7 presents the corresponding estimated coefficients from a SUR regression of calorie, protein and fat intake from the principal food items and milk products. Recall that the household size deflators that were used to convert all the consumption figures into ‘‘per adult equivalent’’ figures were nutritionally determined by the age and gender specific calorie requirements specified by ICMR (2011), see Table 10. A negative sign of the coefficient estimate of NSS 61 indicates a decline between the period 1999/2000 and 2005/2006, after controlling for a full set of household and community characteristics. The results presented in Table 6 Panel A show that over the period 1999/2000 and 2005/2006, the share of food as a proportion of household budget has been declining and the share of education has been increasing. This result is consistent with Engel’s law: as households have become more affluent, they appear to be directing their spending away from food (and to a lesser extent, from medical expenses) towards education. Panels B and C show that this decline in expenditure on food has been accompanied by large and significant decline in the consumption of the principal food items. The fact that the consumption of milk and milk products has declined significantly (see Panel C) is important. Recall that these are households with young children (aged 0–3), and milk and milk products are crucial for the mental and physical development of these children. Milk (and milk products) is a nutrient rich food comprising of nine essential nutrients, including calcium, vitamins A, B12 and D, protein, phosphorous, riboflavin, potassium and niacin, which are crucial for the growth of the child. A fall in consumption of milk and milk products in these households is therefore likely to have a strong direct and negative impact on the health of these children.

Table 8 SUR estimates for per adult equivalent daily intake of calorie, protein and fat by wealth/expenditure quintiles. Protein

Fat

Panel A: Per adult equivalent daily intake of calorie, protein and fat from principal food items Q1: 0–20% 46.26** (23.13) Q2: 20–40% 141.38*** (16.40) Q3: 40–60% 182.43*** (15.61) Q4: 60–80% 215.48*** (16.57) Q5: 80–100% 290.77*** (28.30)

Calorie

0.09*** (0.03) 0.181*** (0.02) 0.23*** (0.02) 0.39*** (0.13) 0.42*** (0.04)

0.01** (0.01) 0.03*** (0.006) 0.04*** (0.01) 0.27 (0.22) 0.13*** (0.02)

Panel B: Per adult equivalent daily intake of calorie, protein and fat from milk products Q1: 0–20% 4.17 (4.74) Q2: 20–40% 2.76 (3.65) Q3: 40–60% 0.09 (3.39) Q4: 60–80% 0.34 (3.63) Q5: 80–100% 7.288* (3.86)

0.001 (0.01) 0.01 (0.01) 0.02* (0.01) 0.38 (0.34) 0.127*** (0.04)

0.001 (0.01) 0.01 (0.01) 0.02* (0.01) 0.40 (0.36) 0.14*** (0.040)

Notes: Coefficient estimate of the NSS 61st round (year = 2005–2006) dummy presented. Figures in parenthesis are robust standard errors. Regressions control for a full set of individual, parental and household characteristics. Full set of results are available on request. Principal food items consist of rice, wheat, pulses, vegetables and milk. *** p < 0.01. ** p < 0.05. * p < 0.1.

79

P. Maitra et al. / Food Policy 38 (2013) 70–81

In addition to this direct effect operating through a lower consumption of the principal food items, there is a possible indirect effect, which works through the worsening of the nutritional status of mothers with young children due to their lower consumption of all the principal food items. Recall that using NFHS data, we have already observed that there is a close link between maternal health and child nutrition (see Table A1). The results presented in Table 7 confirm that, ceteris paribus, over the period 1999/2000 to 2004/2005, per adult equivalent calorie intake fell by 4500 kcals per month (translating to 150 kcals a day on an average). This is a significant decline over a relatively short span of 5 years, especially if we note that we have considered here only a subset of the food items, so the actual decline is likely to have been larger. This decline in calorie intake is accompanied by a sharp and significant decline in the per adult equivalent daily intake of protein and fat as well. This result holds true for the principal food items (Panel A), and for milk products (Panel B). The estimates presented in Table 7 are SUR estimates, which allows for substitution between the nutrients via changing composition of the food items and correlation between the error terms. In our estimation we allow for the possibility of non-diagonal variance–covariance matrix. The full set of results (available in the Supplementary Appendix Table A8) show a negative association between household size and intake of calorie, protein and fat from the principal food items. The variable per capita expenditure is however positively signed and statistically significant for intakes from calorie, protein and fat. We have also computed the corresponding OLS estimates, but due to to space considerations they are not presented here as they are almost identical to the SUR estimated presented in Table 7. Summing up, there is a strong decline in the weight-to-height z-score of children aged 0–3 and the results presented in Table 4 (for the NFHS data) and Tables 6 and 7 (for the NSS data) are indicative of a strong co-movement in declining calorie intakes in households with children and poorer nutritional outcomes of children. Why there is no effect on long term health/nutritional status over the period under consideration remains an open question. Note that our results should be interpreted with caution: given the lack of relevant data we cannot infer causality from our results.

Panel A: Per adult equivalent intake of calorie, protein and fat from principal food items

Panel B: Weight-to-Height z-score. Differences by expenditure quintiles Is the decline in nutritional intake restricted to any particular socio-economic group or is it across the board? To examine this question we re-estimate separately by wealth quintiles the peradult equivalent calorie, protein and fat intake from the principal items and for milk products and present the coefficient estimates from the NSS 61 dummy in Table 8. Between the two rounds, there was a decline in the intake of all three nutrients, namely, calorie, protein and fat obtained from the principal food items taken together as a group. The decline took place across all the expenditure classes, though, in proportionate terms, more for the richest households (those in the top 20% of the expenditure distribution): the coefficient estimate of the NSS 61 dummy is greater in magnitude for the richest households (in Q5): the only exception being input of fat from the principal food items, which is the highest for households in Q4. If we however focus only on milk products, the calorie intake was fairly steady, but that of protein and fat declined over this period. Fig. 5 combines the results from the change in weight-forheight z-scores by socioeconomic status (wealth quintile) and change in the nutritional intake (in per-adult equivalent terms) of calorie, fat and protein from the principal food items across the different surveys. Once again there is evidence of systematic co-movement in nutritional intake and short run health outcomes. In general inputs and outputs are the worst for the richest

Fig. 5. Co-movement of nutritional intake and outcomes by expenditure/wealth quantiles.

households. Unfortunately at this stage we are unable to speculate why the situation has worsened more for the richest households. Geographical differences Table 9 presents the state specific changes in per adult equivalent intake of calorie, protein and fat from the consumption of the principal food items (Panel A) and milk products (Panel B). Almost everywhere (and with only minor exceptions), there is a decline in calorie intake from the consumption of the principal food items and this decline is statistically significant in most cases. There is

80

P. Maitra et al. / Food Policy 38 (2013) 70–81

Table 9 State wise differences in per adult equivalent daily intake of calorie, protein and fat from principal food items and milk products between 55th and 61st round in households with children aged 0–3 years. p-Value of diff.

Fat

p-Value of diff

Panel A: Per adult equivalent daily intake of calorie, protein and fat from principal food items Andhra Pradesh 114.80 0.000 0.09 Assam 95.90 0.002 0.12 Bihar 124.59 0.000 0.29 Gujarat 20.71 0.424 0.07 Haryana 369.21 0.001 0.65 Himachal Pradesh 130.77 0.000 0.17 Karnataka 80.80 0.004 0.07 Kerala 186.43 0.000 0.26 Madhya Pradesh 112.36 0.000 0.11 Maharashtra 48.01 0.265 0.04 Orissa 297.39 0.000 0.38 Punjab 55.43 0.061 0.03 Rajasthan 49.00 0.099 0.08 Tamil Nadu 183.58 0.000 0.25 Uttar Pradesh 120.25 0.006 0.08 West Bengal 172.74 0.000 0.32

Calorie

p-Value of diff.

Protein

0.002 0.000 0.059 0.049 0.003 0.000 0.018 0.000 0.000 0.333 0.000 0.197 0.000 0.000 0.144 0.000

0.01 0.05 0.24 0.03 0.21 0.04 0.02 0.05 0.00 0.01 0.09 0.01 0.03 0.03 0.08 0.14

0.326 0.000 0.360 0.366 0.007 0.006 0.297 0.004 0.898 0.698 0.003 0.469 0.002 0.022 0.059 0.059

Panel B: Per adult equivalent daily intake of calorie, protein and fat from milk products Andhra Pradesh 18.11 0.127 0.01 Assam 24.06 0.073 0.06 Bihar 1.67 0.819 0.33 Gujarat 14.77 0.112 0.04 Haryana 0.17 0.984 0.24 Himachal Pradesh 2.64 0.893 0.13 Karnataka 19.60 0.072 0.04 Kerala 66.35 0.001 0.02 Madhya Pradesh 2.21 0.746 0.02 Maharashtra 21.19 0.114 0.01 Orissa 5.26 0.740 0.00 Punjab 23.72 0.083 0.17 Rajasthan 8.78 0.169 0.07 Tamil Nadu 15.35 0.369 0.01 Uttar Pradesh 6.95 0.206 0.00 West Bengal 0.24 0.983 0.04

0.348 0.365 0.021 0.051 0.332 0.137 0.000 0.112 0.313 0.566 0.124 0.029 0.058 0.003 0.337 0.193

0.01 0.06 0.35 0.05 0.26 0.14 0.04 0.02 0.02 0.02 0.00 0.18 0.07 0.01 0.00 0.04

0.348 0.365 0.021 0.051 0.332 0.137 0.000 0.112 0.313 0.566 0.124 0.029 0.058 0.003 0.337 0.193

Notes: Principal food items consist of rice, wheat, pulses, vegetables and milk.

Table 10 Per capita calorie requirement per day by gender and age group. Gender

Male Female

Per capita calorie requirement per day (kcal) for the age group (in years) <3

3–6

6–9

9–12

12–15

15–18

18–60

>60

1200 1200

1500 1500

1800 1800

2100 2100

2500 2200

3000 2200

2800 2200

1950 1800

Reference: ICMR (2002).

however a fair amount of geographic variation: for example we see that the decline is not large in magnitude and is indeed not statistically significant in the western states of Gujarat and Maharashtra; on the other hand the southern states of Andhra Pradesh, Karnataka, Kerala and Tamil Nadu appear to have performed very poorly over the period. This decline in calorie intake is matched with a decline in intake of fat and protein from the principal food items. The decline is not as strong in magnitude or statistical significance when we restrict ourselves to nutritional intake from milk products. Conclusion It is now widely acknowledged that Indian children are among some of the most malnourished in the world and almost all the available statistics show that nutrition intakes are declining. What is really worrying is that this has occurred against a backdrop of high economic growth rates in the Indian economy. The paradox is heightened by the fact that, in recent years the weight-for-height and height-for-age z-scores have moved in oppo-

site directions. Over the period 1998/1999 to 2005/2006, coinciding with a period of what has been termed the second-generation economic reforms, height-for-age z-scores have improved. However, this gain in long-term nutrition has been matched with a significant decline in weight-for-height z-scores, indicating a decline in shortrun nutrition. This might not necessarily be a problem as the shortterm losses can be easily reversed. Unfortunately, the available datasets cannot tell us whether this decline in weight-for-height is indeed a short-run phenomenon, or whether it is the harbinger of what is to come. After all, persistent adverse short-run effects will ultimately accumulate and have long-term effects. Two other caveats are worth mentioning. First, since our analysis relies on expenditure data, we acknowledge that it is likely that there is an underestimation of consumption, particularly in households that do not rely on the market for food consumption. However, lack of data availability prevents us from conducting any further quantity-based analysis. Second, while we are able to identify a contemporaneous co-movement of declining child nutrition and declining nutritional intake, with the data at hand, we cannot identify any form of causality.

P. Maitra et al. / Food Policy 38 (2013) 70–81

We have identified a co-movement of declining calorie, protein and fat intake for both adults and children and the lack of progress in improving nutritional outcomes of children. Previous studies have simply reported the ‘‘puzzle’’, but not explained it. In studying the child nutritional issues in a comprehensive manner involving two different data sets covering very similar periods, we have tried to nudge the literature forward. We cannot however demonstrate that this link is a causal one. However, the key result of our analysis that there has been a decline in the consumption of the major food items is matched by a decline in overall calorie intake in households with children aged below 3 years is of considerable significance. It points to the role of policy initiatives that can stem the decline in nutritional intake, and hence lead to advances in child health. Such initiatives need to target households with young or, perhaps, to nip the problem in the bud, those with expectant mothers. A possible approach would be to introduce conditional cash transfer programs like that in Mexico (Progressa) or Colombia (Familias en Accion) aimed specifically at improving the health and nutritional outcomes of young children. The Vietnamese experience of impressive performance in improving child nutrition outcomes and nutritional intake against a background of economic performance not dissimilar to India’s (see Mishra and Ray, 2009), also holds several policy lessons for India. The programs documented in Hop (2003) in the Vietnam context as having successes in achieving higher nutrition and reducing malnutrition are worth emulating in India. This is clearly a topic of considerable policy and academic importance and merits further research. Acknowledgements Pushkar Maitra, Anu Rammohan and Ranjan Ray acknowledge funding from Australian Research Council Discovery Grants. The authors are grateful to Ankita Mishra for her painstaking research assistance. They have benefitted from comments and suggestions made by anonymous referees of this journal, Ashok Kotwal, Lisa Magnani, Kunal Sengupta and participants at the Australian Development Economics Workshop (ADEW). The usual caveat applies. Appendix A. Supplementary material Supplementary data associated with this article can be found, in the online version, at http://dx.doi.org/10.1016/j.foodpol.2012.10. 004. References Bardhan, P., 2005. Economic Reforms, Poverty and Inequality in China and India. University of California, Berkeley, Mimeo. Barker, D., 1994. Mothers, Babies and Disease in Later Life. BMJ Publishing, London. Basu, K., Maertens, A., 2007. The pattern and causes of economic growth in India. Oxford Review of Economic Policy 23 (2), 143–167. Block, S., Webb, P., 2009. Up in smoke: tobacco use, expenditure on food, and child malnutrition in developing countries. Economic Development and Cultural Change 58 (1), 1–24. Brabin, B.J., Hakimi, M., Pelletier, D., 2001. An analysis of anemia and pregnancyrelated maternal mortality. Journal of Nutrition 131, 604S–614S. Deaton, A., Dreze, J., 2009. Food and nutrition in India: facts and interpretations. Economic and Political Weekly XLIV (7), 42–65. FAO, 2011. .

81

Gopalan, C., Sastri, B.V.R., Balasubramanian, S.C., 1999. Nutritive Value of Indian Foods. National Institute of Nutrition, ICMR, Hyderabad. Gragnolati, M., Shekar, M., Das Gupta, M., Bredenkamp, C., Lee, Y.-K., 2005. India’s undernourished children: a call for reform and action. Health, Nutrition and Population (HNP) Discussion Paper. The World Bank. Hop, L.T., 2003. Programs to improve production and consumption of animal source foods and malnutrition in Vietnam. Journal of Nutrition 133, 4006S–4009S. ICMR, 2002. Nutrient Requirements and Recommended Dietary Allowances for Indians. A Report of the Expert Group of the Indian Council of Medical Research. ICMR, New Delhi. ICMR, 2011. . Kassouf, A.L., Senauer, B., 1996. Direct and indirect effects of parental education on malnutrition among children in Brazil: a full income approach. Economic Development and Cultural Change 44 (4), 817–838. Keller, W., 1983. Choice of indicators of nutritional status. In: Schurch, B. (Ed.), Evaluation of Nutrition Education in Third World Communities. Hans Huber Publishers, Bern. Lokshin, M., DasGupta, M., Gragnolati, M., Ivaschenko, O., 2005. Improving child nutrition? The integrated child development services in India. Development and Change 36 (4), 613–640. Martorell, R., 1985. Child Growth Retardation: A Discussion of its Causes and its Relationship to Health. Nutritional Adaptation in Man(ed). John Libbey, Londres. Martorell, R., Habicht, J.-P., 1986. Growth in early childhood in developing countries. In: Falkner, F., Tanner, J.M. (Eds.), Human Growth: A Comprehensive Treatise, vol. 3. Plenum Press, New York. McGregor-Grantham, S., 1995. A review of studies of the effect of severe malnutrition on mental development. Journal of Nutrition 125 (8), 2233S– 2238S. Menon, P., Deolalikar, A., Bhaskar, A., 2008. India State Hunger Index: Comparisons of Hunger across States. International Food Policy Research Institute, Washington, DC. Mishra, V., Ray, R., 2009. Dietary diversity, food security and undernourishment: the Vietnamese evidence. Asian Economic Journal 23 (2), 225–247. Murray, C., Lopez, A., 1997. Global mortality, disability and the contribution of risk factors: global burden of disease study. Lancet 349, 9063. Osberg, L., Shao, J., Xu, K., 2009. The growth of poor children in China 1991–2000: why food subsidies may matter. Health Economics 18 (Suppl. 1), S89–S108. Pathak, P., Singh, A., 2011. Trends in malnutrition among children in India: growing inequalities across different economic groups. Social Science and Medicine 73 (4), 576–585. Patnaik, U., 2007. Neoliberalism and rural poverty in India. Economic and Political Weekly 42 (30), 3132–3150. Pelletier, D.L., 1998. The potentiating effects of malnutrition on child mortality: epidemiologic evidence and policy implications. Nutrition Review 52 (12), 409– 415. Ray, R., 2007. Changes in food consumption and the implications for food security and undernourishment: India in the 1990s. Development and Change 38 (2), 321–343. Ray, R., Lancaster, G., 2005. On setting the poverty line based on estimated nutrient prices: condition of socially disadvantaged groups during the reform period. Economic and Political Weekly 40, 46–56. Siggel, E., 2010. Poverty alleviation and economic reforms in India. Progress in Development Studies 10, 247. Strauss, J., Thomas, D., 1995. Human resources: empirical modelling of household and family decisions. In: Srinivasan, T., Behrman, J. (Eds.), Handbook of Development Economics, vol. 3A. North Holland Press, Amsterdam, pp. 1883– 2023. Tarozzi, A., Mahajan, A., 2007. Child nutrition in India in the nineties. Economic Development and Cultural Change 55 (3), 441–486. Tomkins, A.M., Watson, F., 1989. Malnutrition and Infection: A Review. Nutrition Policy Discussion Paper, # 5. ACC/SCN, Geneva, ACC/SCN State of the art series. UNDP, 2007–2008. Human Development Report: Fighting Climate Change: Human Solidarity in a Divided World. Oxford University Press, Oxford and New York. Waterlow, J.C., Buzina, R., Keller, W., Lane, J., Nichaman, M., Tanner, J., 1997. The presentation and use of height and weight data for comparing the nutritional status of groups of children under the age of 10 years. Bulletin of the World Health Organisation 55 (489–498). WHO, 1986. Use and interpretation of anthropometric indicators of nutritional status. Bulletin of the World Health Organisation 64, 929–941. WHO, 2008. Worldwide Prevalence of Anaemia, 1993 to 2005. WHO, Geneva. WorldBank, 2006. India Development Policy Review. World Bank, Washington. WorldBank, 2011. An Urgent Call for Action: Undernourished Children of South Asia. .