IFPRI Discussion Paper 01423

March 2015

Agriculture, Nutrition, and the Green Revolution in Bangladesh

Derek D. Headey

John Hoddinott

Poverty, Health, and Nutrition Division

INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE

The International Food Policy Research Institute (IFPRI), established in 1975, provides evidence-based policy solutions to sustainably end hunger and malnutrition and reduce poverty. The Institute conducts research, communicates results, optimizes partnerships, and builds capacity to ensure sustainable food production, promote healthy food systems, improve markets and trade, transform agriculture, build resilience, and strengthen institutions and governance. Gender is considered in all of the Institute’s work. IFPRI collaborates with partners around the world, including development implementers, public institutions, the private sector, and farmers’ organizations, to ensure that local, national, regional, and global food policies are based on evidence. IFPRI is a member of the CGIAR Consortium.

AUTHORS Derek D. Headey (d.headey@cgiar.org) is a senior research fellow in the Poverty, Health, and Nutrition Division of the International Food Policy Research Institute (IFPRI), Washington, DC. John Hoddinott (j.hoddinott@cgiar.org) is deputy division director of the Poverty, Health, and Nutrition Division of IFPRI, Washington, DC and H.E. Babcock Professor of Food and Nutrition Economics and Policy at the Charles H. Dyson School of Economics and Management at Cornell University, Ithaca, NY.

Notices 1. IFPRI Discussion Papers contain preliminary material and research results and are circulated in order to stimulate discussion and critical comment. They have not been subject to a formal external review via IFPRI’s Publications Review Committee. Any opinions stated herein are those of the author(s) and are not necessarily representative of or endorsed by the International Food Policy Research Institute. 2. The boundaries and names shown and the designations used on the map(s) herein do not imply official endorsement or acceptance by the International Food Policy Research Institute (IFPRI) or its partners and contributors.

Copyright 2015 International Food Policy Research Institute. All rights reserved. Sections of this material may be reproduced for personal and not-for-profit use without the express written permission of but with acknowledgment to IFPRI. To reproduce the material contained herein for profit or commercial use requires express written permission. To obtain permission, contact the Communications Division at ifpri-copyright@cgiar.org.

iii

Contents

Abstract v

Acknowledgments vi

1. Introduction 1

2. Agricultural Production and Nutrition in Bangladesh 3

3. Conceptual Linkages between Rice Productivity Growth and Nutritional Outcomes in Bangladesh 6

4. Data and Methods 8

5. Results 11

6. Conclusions 16

References 17

iv

Tables

2.1 Sources of growth in Bangladeshi agriculture, 1997–2011 (in percentages) 5

4.1 Descriptive statistics 9

5.1 Fixed-effects estimates of the impact of rice yields on chronic and acute child nutrition indicators 11

5.2 Adding control variables to the WHZ and mild wasting fixed-effects models 12

5.3 Fixed-effects estimates of the impact of yield growth on child-feeding indicators (in percentages) 13

5.4 Fixed-effects estimates of the associations between child feeding indicators and child nutrition outcomes 14

5.5 Assessing the robustness of the impact of yields on child weight: Fixed-effects and Blundell-Bond generalized methods of moments estimates 15

5.6 Assessing the robustness of effects of yields on child-feeding indicators: Fixed-effects and Blundell-Bond generalized methods of moments estimates (in percentages) 15

Figures

1.1 Trends in child undernutrition and rice yields, 1997–2011 2

2.1 A late bloomer catching up: Rice yields in Bangladesh relative to Indonesia 4

2.2 Trends in irrigation, modern variety uptake, and rice yields 4

v

ABSTRACT

Although agriculture is widely thought to be an important sector for influencing nutrition outcomes, there is a remarkable dearth of rigorous evidence on the role of large scale agricultural programs, such as Asia’s Green Revolution. This paper therefore analyzes agriculture and nutrition linkages in Bangladesh, a country that achieved rapid growth in rice productivity at a relatively late stage in Asia’s Green Revolution, as well as unheralded progress against undernutrition. To do so the authors create a synthetic panel that aggregates nutritional data from five rounds of the Demographic Health Surveys (1997 to 2011) with district-level estimates of rice yields. Using various panel estimators, they find rice yields significantly explain weight gain in young children but not linear growth. The authors further show that rice yields have large and positive effects on the timely introduction of complementary foods for young children but not on dietary diversity indicators and that this complementary feeding indicator is positively associated with child weight gain but not with linear growth. The reverse holds for dietary diversity indicators, which influence linear growth but not weight gain. The results therefore suggest that Bangladesh’s Green Revolution has made a large contribution to improvements in weight gain but has had little effect on postnatal linear growth. Achieving the latter would appear to require greater efforts to diversify diets out of cereals and into more micronutrient-rich foods.

Keywords: undernutrition; Green Revolution; agricultural productivity; rice; Bangladesh

vi

ACKNOWLEDGMENTS

We would especially like to thank Wahid Quabili and Aminul Islam Khandaker for invaluable research assistance and Akhter Ahmed and Purnima Menon for comments and advice. This work has partly been supported by the Department for International Development (United Kingdom) through its funding of the Transform Nutrition Consortium and the Leveraging Agriculture for Nutrition in South Asia Consortium as well as the Bill & Melinda Gates Foundation through its funding of the Advancing Research on Nutrition and Agriculture project.

1

1. INTRODUCTION

Despite a surge of recent interest in identifying the impact of agricultural interventions on maternal and child nutrition, the existing scientific literature has brought little light to bear on the core question of how large-scale agricultural programs and policies influence nutrition outcomes (Ruel and Alderman 2013). Much of the literature has been confined to cross-sectional studies for which even indirect policy attribution is very difficult (Bhagowalia, Headey, and Kadiyala 2012; Dillon, McGee, and Oseni 2014; Hoddinott, Headey, and Dereje 2014) or to more experimental studies of small-scale livestock or homestead gardening interventions (Berti, Krasevec, and FitzGerald 2004; Leroy and Frongillo 2007; Masset et al. 2012).

Yet advocates of agriculture-led development typically have in mind much larger-scale agricultural programs and policies in the spirit of Asia’s “Green Revolutions” (Bezemer and Headey 2008; Diao et al. 2008; Diao, Hazell, and Thurlow 2010; Hazell 2009; Mellor 1976; Pinstrup-Andersen 2013). These Green Revolutions were led by the research and development of improved rice, wheat, and maize varieties, which—along with associated policies to promote the expansion of irrigation, fertilizers, and other inputs—have contributed to rapid growth in Asian food production during the past 40 years. Rice yields, for example, have increased by around 150 percent in Bangladesh, northern India, Indonesia, and Pakistan since the 1960s, while wheat yields in these countries increased by some 250 percent (Food and Agriculture Organization of the United Nations [FAO] 2014).

Despite its fundamental contribution to poverty reduction, surprisingly little is known about the impact of Asia’s Green Revolution on nutrition, and much of what has been written is speculative at best (see Hazell 2009 for a review). Optimists have focused on the contributions of Green Revolution investments to calorie consumption and national food security (Pinstrup-Andersen and Jaramillo 1991), while pessimists point to the adverse micronutrient consequences of reduced biodiversity in monocropping systems, particularly lower consumption of pulses, coarse grains, and fish (Bouis 2000), and to the harmful health and nutritional impacts of excessive use of fertilizers and pesticides (Brainerd and Menon 2014). Yet the majority of this evidence has not examined the impacts of agricultural growth on changes in nutrition outcomes. This knowledge gap exists because the Green Revolutions of the 1960s, 1970s, and 1980s largely preceded the kinds of large, multitopic surveys that are typically a prerequisite for identifying the welfare impacts of large-scale interventions (Elbers and Gunning 2013; Barrett and Carter 2010).

This paper seeks to fill this knowledge gap by studying the nutritional impacts of rice productivity growth in Bangladesh. Bangladesh is an ideal case study for several reasons.

First, for political reasons, Bangladesh was a relatively late adopter of Green Revolution technologies (Evenson and Gollin 2003), meaning that much of its productivity growth occurred during more recent periods of improved statistical surveillance. From 1997 to 2011 (the period of our analysis) yield growth for rice averaged 3.6 percent per year (Figure 1.1) on the back of increased adoption of improved varieties and the rapid expansion of the irrigated dry season rice crop. Second, productivity growth in Bangladesh coincided with substantial improvements in preschooler nutritional status. In 1996/1997, fully half of preschooler children in Bangladesh were stunted (height for age z (HAZ) score < –2) or mildly wasted (weight for height (WHZ) z score < –1). These very high rates of undernutrition were among the highest in the world at that time, although from 1996/1997 to 2011 all three indicators fell dramatically (Figure 1.1). Third, Bangladesh has a relatively rich array of nutritional and agricultural data, the dearth of which has undoubtedly been a constraint in this literature.

2

Figure 1.1 Trends in child undernutrition and rice yields, 1997–2011

Source: FAO (2014). Note: mt/ha = metric ton per hectare. Stunting refers to HAZ scores of –2 or less, for children 0 to 24 months of age. Mild

wasting refers to WHZ scores of –1 or less, for children 0 to 24 months of age measured from October to December of the calendar year in question, to control for seasonality of wasting and the different survey timings of the various Demographic Health Surveys rounds. Wasting values for 2004 and 2007 are imputed with a linear trend.

In this paper, we exploit these rich data sources to construct a synthetic panel dataset comprising nutritional indicators from five rounds of the Bangladesh Demographic Health Surveys (DHS) (during 1996/1997 to 2011) and district-level data on rice yields from the Bangladesh Bureau of Statistics. With these data, we examine the impact of rice productivity growth on preschool nutritional status, controlling for district-level fixed effects and time period effects as well as other time-varying factors that might also explain nutritional improvements in Bangladesh, such as the rapid expansions in women’s education, sanitation, health and family planning services, women’s empowerment, and nongovernmental organization (NGO) services. We assess the robustness of our time and location fixed-effects models through the use of panel generalized methods of moments (GMM) estimation techniques that address endogeneity concerns regarding the estimated impacts of rice yields. Hence we are able to purge our regression models of factors that potentially confound the inferences drawn from the cross-sectional econometric literature referenced above.

The remainder of this paper is structured as follows. Section 2 provides an overview of the nutritional and agricultural situation in Bangladesh. Section 3 describes a conceptual model used to motivate our empirics. Section 4 describes the construction of the synthetic dataset and the empirical models we use to analyze that data. Section 5 presents our results, and Section 6 summarizes our findings.

Stunting

Mild wasting

Rice yields

0.0

0.5

1.0

1.5

2.0

2.5

3.0

3.5

4.0

4.5

0%

10%

20%

30%

40%

50%

60%

1997 2000 2004 2007 2011

Rice

yie

lds (

mt/

ha)

Und

ernu

triti

on p

reva

lenc

e (%

)

3

2. AGRICULTURAL PRODUCTION AND NUTRITION IN BANGLADESH

Bangladesh is characterized by a uniquely intensive agricultural production system that largely takes place on very small family farms engaged in multiple cropping seasons. Most of the poor are landless farm or nonfarm laborers, or smallholders (Balagtas et al. 2014; Hossain 2004). Traditionally, rice production used relatively low levels of irrigation and other modern inputs. Production was a highly seasonal affair, with the vast majority of production taking place in the monsoonal aman season, with production in the dry season—particularly the boro season—constrained by lack of water. But like other Asian countries, Bangladesh was to benefit substantially from the research and development of high-yielding varieties (HYVs) and the associated adoption of irrigation and other modern inputs: in short, the so-called Green Revolution package (Ahmed, Haggblade, and Chowdhury 2000; Hossain, Bose, and Mustafi 2006; Naher 1997).

Bangladesh’s Green Revolution got off to a sluggish start (Figure 2.1). In 1967 the Bangladesh Academy of Rural Development imported the IR8 variety from the International Rice Research Institute in the Philippines and introduced it in the dry season, while IR20 was introduced in 1970 for the wet season. The spread of these varieties was slow in the 1970s, delayed by the war of independence and the rebuilding process as well as the vulnerability of new varieties to pests and disease (Hossain, Bose, and Mustafi 2006). By the mid-1980s only around 27 percent of the rice area was planted with modern varieties, and yield growth was averaging around 2.2 percent per year (FAO 2014). The 1990s saw more dramatic changes, however. First, there was an acceleration of HYVs adoption (Figure 2.2) driven via improved linkages between agricultural extension and research (including increased tolerance of disease and local farming cycles) and better collaboration between the public and private sectors (including influential domestic non-government organizations (NGOs) such as BRAC). By the late 1990s around two-thirds of the rice area was planted with HYVs. Second, the government progressively liberalized agricultural inputs, particularly the imports of small-scale irrigation equipment, including diesel pumps and shallow tubewells (Nazneen et al. 2007). As a result, the irrigated area doubled from 1990 to 2010, and the share of the once minor boro crop in total production increased from around 15 percent in the 1970s to 58 percent in 2010 (Figure 2.2).1 In 1997, rice accounted for almost two-thirds of total production value (Table 2.1). From 1997 to 2011 rice production grew by 80.7 percent, or 5.8 percent per year, and accounted for 61.4 percent of total production growth. In contrast, growth of other cereals was immaterial. While growth in livestock and “other food products” was relatively rapid, it started from a low base; Hossain, Bose, and Mustafi (2006) argue that the development and adoption of fast-maturing rice varieties allowed farmers to plant more non-rice crops, suggesting that some of the growth in other food products might have resulted from rice productivity gains.

1 Note, however, that fertilizer application per cropped area changed only modestly during the same period.

4

Figure 2.1A late bloomer catching up: Rice yields in Bangladesh relative to Indonesia

Source: FAO (2014). Note: mt = metric ton; DHS = Demographic Health Surveys.

Figure 2.2 Trends in irrigation, modern variety uptake, and rice yields

Source: Data are from the Bangladesh Bureau of Statistics (various years). Note: mt = metric ton; MT/Ha = metric tons per hectare; MV = modern variety.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

1961

1963

1965

1967

1969

1971

1973

1975

1977

1979

1981

1983

1985

1987

1989

1991

1993

1995

1997

1999

2001

2003

2005

2007

2009

2011

Rice

yie

lds (

mt p

er h

ecta

re)

BangladeshIndonesia

0

0.2

0.4

0.6

0.8

1

1.2

1.4

0%

10%

20%

30%

40%

50%

60%

70%

80%

90%

1991

1992

1993

1994

1995

1996

1997

1998

1999

2000

2001

2002

2003

2004

2005

2006

2007

2008

2009

2010

Rice

yie

lds (

mt/

acre

)

Area

und

er ir

rigat

ion

& M

Vs (%

)

MV area (%) Irrigated area Yield (MT/Ha)

_ _ _ DHS rounds

5

Table 2.1 Sources of growth in Bangladeshi agriculture, 1997–2011 (in percentages)

Variable

Initial share of production value,

1997

Growth rate of production 1997–2011

Share of total agricultural

growth Food

Rice 63.5 5.8 61.4

Other cereals 3.1 6.1 3.2

Livestock 15.3 4.0 10.2

Other food products 14.3 9.7 23.3

Nonfood 3.7 3.2 2.0

All agriculture 100.0 5.8 Source: FAO (2014).

This combination of rapid growth in yields combined with a more even spread of production across the wet and dry seasons is plausibly linked to substantial improvements in food security. FAO (2014) estimates suggest that from 1990 to 2010, the percentage of the population that was calorie deprived fell by half, from 35 percent to 17 percent. An additional benefit of the rapid emergence of the boro crop was reduced intra-annual rice price variation (Dorosh and Shahabuddin 2002), which may have reduced seasonal hunger, especially for landless laborers, who are the poorest group of net food consumers.

However, there is thus far no evidence of any impact on rice productivity growth on nutritional change, and there are numerous other potential explanations of Bangladesh’s rapid reductions in maternal and child undernutrition. Bangladesh is arguably better known for its microfinance revolution (Khandker 2005), focus on women’s empowerment and secondary school education for girls in particular (World Bank 2005), significant improvements in community-led health and family planning initiatives (Chowdhury et al. 2013; World Bank 2005), pioneering Community-led Total Sanitation campaigns (Kar 2003), sustained growth in labor-intensive manufacturing (Zhang et al. 2013), and massive surge in overseas remittances during the past ten years (Balagtas et al. 2014; Orr et al. 2009). The various rounds of the DHS also show fairly substantial improvements in several important infant and young child feeding indicators, particularly immediate postnatal breast-feeding and timely introduction of complementary foods (Hanif 2013), suggesting that nutritional programs in Bangladesh might have been more effective than is commonly thought (World Bank 2005).

Consistent with the hypothesis that nutritional change was a multisectoral process, Headey et al. (2015) conduct statistical decompositions of changes in child HAZ scores and stunting during the same five DHS rounds (1996/1997 to 2011) and find that asset accumulation and gains in parental education were the two largest drivers of change, followed by significant improvements in health, sanitation, and family planning outcomes. However, their model—which notably excludes any agricultural indicators—accounts for only slightly more than half of the actual nutritional change in rural areas.2

2 Their preferred model explains around 70 percent of the improvement in urban areas.

6

3. CONCEPTUAL LINKAGES BETWEEN RICE PRODUCTIVITY GROWTH AND NUTRITIONAL OUTCOMES IN BANGLADESH

The agricultural household model developed by Singh, Squire, and Strauss (1986) and extended by Behrman and Deolalikar (1988) of nutrient intake and anthropometric outcomes provides a basis for linking rice productivity growth to nutrition.3 These models conceptualize the household decisionmaking process as one wherein parents are concerned about the nutritional status of preschool children and their nutrient intakes. Households command various resources—their labor, time, and agricultural assets—and make utility-maximizing production and consumption decisions in the context of standard budget constraints. The resulting models cast nutrition outcomes and nutrient intakes as a function of farm income, off-farm wage earnings, food and nonfood prices, time resources devoted to child care, community characteristics, and exogenous consumption preferences (tastes).

These models posit a number of important linkages between exogenous technological shocks—such as the availability of modern varieties and irrigation technologies—and individual nutrition outcomes. A first-order effect of a technology shock will be on farm income, with farm households making adoption, planting, input, and marketing decisions based on the relative profitability of the technology. As stressed by Singh, Squire, and Strauss (1986), under the strong assumptions of complete markets farmers do not consider household nutrition requirements in these production decisions. However, various market failures might motivate farmers to make production decisions for nonincome reasons. For example, farmers might specialize in a staple food like rice because it is less perishable than fruits or vegetables, generally less susceptible to flooding, and hence less vulnerable to marketing failures (Shahabuddin and Dorosh 2002). In the absence of other effective means of securing household food security, it is possible that farmers allocate more resources to rice production than would be optimal under pure profit incentives in the context of complete well-functioning markets.

For the large number of nonfarm households in rural Bangladesh, however, it is the second-order effects of the technology shock that matter most. The availability of HYVs in Bangladesh and the liberalization of input markets resulted in pervasive multiple cropping. This, in turn, translated into large increases in the demand for farm labor, which is expected to benefit landless households through higher wages, increased working hours, or a combination of both (Ravallion 1990). Similarly, the income gains accruing to rice farmers increase their demand for local nonfarm goods and services (such as house construction), which also serves to increase wages. Another second-order effect pertains to reductions in real rice prices. Since the bulk of the poor are landless net food consumers, lower real rice prices are expected to substantially improve real incomes for the poorest segments of society.

However, a key question facing the agriculture-nutrition literature is whether these household welfare gains also translate into nutritional improvements in young children. A well-established nutritional literature has shown that child undernutrition largely manifests itself in the first 1,000 days of life (in utero and the first 24 months of childhood), after which there is substantial stabilization (Victora et al. 2009). Child nutrition in this period is affected by maternal nutrition insofar as the latter affects birth size and weight (both of which are low in Bangladesh), but thereafter the growth-faltering process is largely influenced by inappropriate infant and young child care practices and the disease environment.

Among the former, poor feeding practices are widely recognized as a critical constraint in South Asia (Senarath et al. 2012; Kabir et al. 2012; Menon 2012). While the introduction of complementary foods (solid or semisolid) should be almost universal by six to eight months of age, only 20 percent of such children were given complementary foods in the 1996/1997 round of the DHS, though there was rapid improvement by 2004, with almost 70 percent of children six to eight months of age being given solid foods. In extremely food-insecure settings it is possible that parents may delay the introduction of complementary foods and instead rely on prolonged breast-feeding (we hereafter term this the rationing hypothesis). Since rice is easily the most common complementary food for infants and young children in

3 For a recent discussion, with new extensions, see LaFave and Thomas (2012).

7

Bangladesh—with more than 60 percent of children in the six to eight month age group consuming rice in 2007, compared to much smaller proportions consuming micronutrient-rich foods—it is possible that rice productivity growth contributed to observed improvements in feeding practices.

While rice productivity growth should substantially improve basic household food security—and perhaps also food availability for young children—it is less obvious that these gains have translated into significant dietary diversity. Theoretically, the real income effects of lower rice prices will lead to dietary diversification (according to Bennett’s law), although reductions in the price of rice relative to other calories could lead to substitution (in both production and consumption) out of more nutritious foods (pulses, fish) in favor of rice production and consumption (Bouis 2000). Using multiple rounds of nutrition and price data collected in high-frequency surveillance data, Torlesse, Kiess, and Bloem (2003) find that reductions in rice prices increase the share of household food expenditure devoted to nonrice foods (that is, dietary diversity), which suggests that income effects dominate. However, at the aggregate level the FAO Food Balance Sheets show very modest dietary diversification in Bangladesh during the past two decades (FAO 2014), while the various rounds of DHS used herein also show very limited dietary diversification among young children, as do other surveys.4

Productivity growth in agriculture might also have some negative impacts on nutrition. New agricultural technologies can lead to more demands on women’s time use (Headey, Chiu, and Kadiyala 2012), though rice production in Bangladesh is primarily a male occupation. More generally, discrimination against women and girls is widely thought to be a significant constraint on nutrition in Asia (Jayachandran and Pande 2013; Smith et al. 2003; Molini, Nubé, and van den Boom 2010), implying that girls might benefit less from productivity growth than do boys. However, as we noted above, Bangladesh is internationally renowned for the promotion of greater empowerment of women, so in this particular instance it is possible that the benefits of agricultural productivity growth were actually enhanced by the simultaneous improvement in women’s status. Finally, the overuse and inappropriate use of fertilizers and pesticides is a significant concern in South Asian agriculture. A study on fertilizer use in India used a natural experiment to gauge the nutritional impact of excessive fertilizer use and found significant negative effects on child growth outcomes (Brainerd and Menon 2014).5 There may well be similar concerns in Bangladesh, where a World Bank study found that in 2005 around half of farmers were overusing pesticides, while more than 87 percent admitted to using little or no protective measures while applying pesticides (Dasgupta et al. 2005). On the other hand, since Bangladeshi women tend to work less in rice production and more in the homestead, it is not obviously the case that the effects are equally severe in Bangladesh. Thus, while we still have relatively strong expectations of observing a significant and positive effect of rice productivity growth on maternal and child nutrition, there are also mechanisms that might weaken these effects or lead to impacts on some nutrition indicators but not on others.

4 The Demographic Health Surveys (DHS), unfortunately, measure different food groups during different periods, and for

some specific groups, like dairy, they phrase the questions inconsistently. While we have made every effort to ensure consistency, it is possible that the different phrasing of the questions across rounds creates some systematic inconsistencies. In our regression models these inconsistencies are addressed through time period (round) effects.

5 One study conjectures that rice-based diets may constrain nutritional improvement. In a very detailed intrahousehold study in very poor villages in northern Bangladesh, Ahmed (1993) found that preschooler children often struggled to eat the rice allocated to them by their parents. He conjectured that this stemmed partly from poor health environments but also from the starchiness of rice and the tendency toward eating large, twice-daily meals that overfill small stomachs rather than snacking more frequently on more diverse food preparations.

8

4. DATA AND METHODS

Empirical research into agriculture-nutrition linkages has been greatly constrained by lack of data, with nutritional surveys rarely collecting agricultural data and long-running economic surveys rarely collecting anthropometric data. For these reasons we create a synthetic panel by merging five rounds of data on rural households and children from the nationally representative Bangladesh DHS collected in the years 1996/1997, 2000, 2004, 2007, and 2011 with district-level agricultural data (Bangladesh Bureau of Statistics 2014). We focus on rural areas only. While rice productivity growth might benefit urban areas through lower rice prices, high levels of market integration across urban areas would presumably result in very little spatial variation in rice prices. Moreover, the urban sample in DHS is relatively small, leaving some districts with very few urban observations.

Each observation in our rural panel is a year-district mean. There are 23 rural districts and five observations per district, but because we are missing data for several districts in several rounds, our sample has 109 observations. When constructing this sample, we were concerned that we had a sufficient number of children in each district. In our main sample the median number of children per district-year is 352, though there are a handful of districts where many fewer children were measured (the minimum being 17 children), mostly in the Chittagong Hill districts. We test the robustness of our results to the exclusion of such districts.

Our outcome variables reflect two dimensions of undernutrition in children: acute undernutrition (weight for height) and chronic undernutrition (height for age). We also apply cutoffs to create indicators of undernutrition prevalence at the district level. For weight-for-height z scores we use a threshold of –1 standard deviations to identify mild wasting prevalence and –2 to identify moderate wasting. For height-for-age z scores we use –2 standard deviations for moderate stunting and –3 for severe stunting. We focus on weight gain and linear growth in the first 24 months largely because anthropometric processes tend to stabilize beyond 24 months (Victora et al. 2009) but also because we lack dietary indicators for children older than 24 months.

We also test whether rice productivity growth explains changes in some basic DHS indicators of child-feeding practices among children aged 4 to 18 months.6 The first such indicator is the timely introduction of complementary foods, which refers to the proportion consuming any solid or semisolid foods. A second child-feeding indicator is minimum dietary diversity (as defined by the World Health Organization), which is the percentage of children consuming more than four food groups out of a maximum of seven. Dietary diversity indicators have previously been shown to be a robust predictor of linear growth in children (Arimond and Ruel 2006). Last, we look at dairy consumption (specifically, milk and yogurt consumption) as a diversity indicator. This has previously been shown to be a strong predictor of child growth outcomes in rural areas of other developing countries (Hoddinott et al. 2015; Rawlins et al. 2014) and, in a number of more experimental studies, in both developed and developing countries (Hoppe et al. 2006; Iannotti, Muehlhoff, and Mcmahon 2013; de Beer 2012).

In terms of rice productivity, we focus on rice yields in the 12 months just prior to the start date of each DHS round. Yields are likely the best summary indicator of productivity in rice production, reflecting the benefits of HYVs, irrigation, and other modern inputs such as fertilizers, pesticides, and

6 The timely introduction of complementary foods is normally measured in the 6 to 8 month window per World Health Organization guidelines. However, restricting the sample to 6 to 8 months would not provide us with enough children at the district level to ensure sufficiently minimal measurement error. Moreover, earlier rounds of the Bangladesh DHS suggest that reasonably large numbers of children were not even given solid or semisolid foods in late infancy (12–18 months). In the past, some nutritionists have also advocated earlier introduction of solid or semisolid foods (4–5 months) in conjunction with continued breast-feeding. Hence, our window of measurement for this indicator is somewhat unconventionally large. In any event our results are qualitatively robust to marginally reducing this window, although they are inevitably somewhat less precise.

9

herbicides. But because our weight-based indicators are potentially quite sensitive to short-term fluctuations, we pay particular attention to the timing of measurement. We measure yields in the 12 months prior to each survey since current household rice stocks and local rice prices depend largely on production in recent seasons rather than the still-ongoing current season.7 Given Bangladesh’s multiple seasons, rice yields are the sum of production in different seasons divided by the total crop area harvested in those seasons. This indicator is available at only the district level and is based on agricultural surveys carried out by the Bangladesh Bureau of Statistics (Bangladesh, Bureau of Statistics 2014). Table 4.1 shows descriptive statistics for our key agricultural and nutrition indicators for the 109 district-years in our sample. There is substantial variation in our productivity and nutrition indicators across district-time periods.

Table 4.1 Descriptive statistics Variable Number of district-

year observations Mean Standard

deviation Rice productivity

Yields (metric ton/acre) 109 0.93 0.21

Child nutritiona

HAZ score 109 –1.56 0.33

Moderate stunting, HAZ < –2 (0–1) 109 0.39 0.11

Severe stunting, HAZ < –3 (0–1) 109 0.16 0.08

WHZ score 109 –1.02 0.29

Mild wasting, WHZ < –1 (0–1) 109 0.51 0.10

Severe wasting, WHZ < –2 (0–1) 109 0.19 0.08

Child feedingb

Complementary feeding (0–1) 109 0.68 0.18

Dietary diversity (1–7) 89 2.17 0.78

Dairy consumption (0–1) 109 0.31 0.15

Source: Authors’ estimates from various DHS rounds. Note: HAZ = height-for-age z score; WHZ = weight-for-height z score. a. Child nutrition indicators are measured for children

aged 0 to 24 months. b. Child feeding indicators are measured for children aged 4 to 18 months. Note that the dietary diversity indicator is not available for the 1996/1997 round.

We model the relationship between nutrition outcomes and yields as

𝑁𝑖,𝑡 = 𝛼 + 𝛽. 𝑦𝑖,𝑡��𝜸𝑫𝑖,𝑡 + 𝜹𝑿𝑖,𝑡 + 𝑻𝑡 + 𝝁𝑖 + 𝜀𝑖,𝑡, (1)

where N refers to child nutrition outcomes for district i and time t; y to rice yields;�𝛽 to the impact of yields on nutrition; D to child age and its square; X to various other determinants of nutrition outcomes such as education, health, sanitation, and demographic variables; T to time dummies for each survey round; 𝝁�to district fixed effects; and 𝜀 to an error term. As such, we control for time-invariant factors that could be correlated with both yields and nutritional quality (for example, water availability, soil quality,

7 The availability of seasonal production data at the district level allows us to adjust our agricultural measurement dates to the DHS survey timing. For example, in the 1996/1997 DHS round the survey dates ranged from October 1996 to April 1997. Thus we measured yields in the 12 months from July 1995 to June 1996, which includes the 1995 aman yield (harvested mostly in November and December 1995) and the boro and aus yields from April to June 1996. In another round (2004) the DHS began in January and finished in May, so we measured yields in the previous calendar year (2003).

10

climate, cultural practices), aggregate temporal effects that might simultaneously influence agricultural productivity and nutrition (for example, changes in international food prices), and a number of nutrition-relevant control variables that could also potentially be correlated with rice productivity/technology growth.

A limitation of equation 1 is that yields, though predetermined, may not be strictly exogenous. Unobservable shocks that cause yields to vary may be correlated with other factors that influence nutrition outcomes, including both positive unobservable shocks (such as transitory increases in nonfarm employment) and negative unobservable shocks (such as localized flooding). Past realizations of these shocks might also have persistent effects on both yields and nutrition outcomes. This can be addressed by including lagged nutrition outcomes as an additional explanatory variable, but of course this will be correlated with the error term to yield a dynamic panel bias (Arellano and Bond 1991). The solution to both concerns is to use lagged levels or differences of the endogenous or predetermined variables as instruments. Hence, in addition to the conventional fixed-effects model in equation 1, we estimate a first-differenced version of equation 1 with lagged nutrition and yields dynamically instrumented using the GMM estimator proposed by Blundell and Bond (1998). Note also that we use a collapsed instrument set to prevent overidentification of the endogenous/predetermined variables, as recommended by Roodman (2009).

11

5. RESULTS

Table 5.1 reports the results of regressing the various chronic and acute child nutrition outcomes against rice yields in models that control for fixed effects and time period effects as well as child age and its square. In regressions 1, 2, and 3 we find that rice yields have no significant effect on the three chronic nutrition indicators. One explanation may be that current stunting rates are in fact related to cumulative nutrition processes and hence to several years of lagged rice productivity outcomes. Another explanation may be that linear growth benefits more from dietary diversification than calorie availability. We explore these explanations below.

Table 5.1 Fixed-effects estimates of the impact of rice yields on chronic and acute child nutrition indicators

1 2 3 4 5 6

Variable

HAZ score

Moderate stunting

(HAZ < –2)

Severe stunting

(HAZ < –3)

WHZ score

Mild wasting

(WHZ < –1)

Moderate wasting

(WHZ < –2) Rice yields –0.16 0.15 –0.03 0.83** –0.26** –0.14 (0.34) (0.16) (0.10) (0.31) (0.12) (0.14) Age control Yes Yes Yes Yes Yes Yes Time effects Yes Yes Yes Yes Yes Yes

Fixed effects Yes Yes Yes Yes Yes Yes

R-squared .42 .33 .42 .28 .33 .39 Sample size 109 109 109 109 109 109

Source: Authors’ estimates. Note: HAZ = height for age z score; WHZ = weight for height z score. Robust standard errors are included in parentheses. All

indicators are measured for children 0 to 24 months of age. *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level.

In contrast, rice yields have a large, well-measured impact on acute undernutrition. The effects of yields on the WHZ score are significant at the 5 percent level and are meaningful in magnitude. The 0.5 metric ton increase in yields per acre from 1997 to 2011 predicts a 0.41 standard deviation improvement in WHZ scores. The same increase in yields predicts a 13 percentage point decline in mild wasting (WHZ < –1). But in regression 6 we observe much more modest effects on moderate wasting, with the coefficient’s being insignificant though still negative and reasonably large in magnitude.

Our results could be confounded by location-specific time-varying characteristics. Fortunately, DHS also provide a wealth of nutrition-relevant control variables, such as education, sanitation, access to health services, fertility rates and household demographics, and proxies for women’s empowerment. In Table 5.2 we assess the robustness of our results on WHZ scores and mild wasting by adding a set of control variables to the relevant models from Table 5.1. Our selection of variables is guided by the existing literature on nutritional change in Bangladesh, particularly Headey et al. (2015). That study showed that parental education was one of the main drivers of nutritional change from 1997 to 2011, particularly maternal education, though several other important factors were also identified, including availability of toilets, basic health services, and reductions in fertility rates. We also add an admittedly crude indicator of women’s decisionmaking (whether a woman can go to a health clinic alone) on the basis of Bangladesh’s pervasive emphasis on female empowerment initiatives as well as the percentage of women surveyed who said they participated in NGO programs. The latter control could be important because several NGOs, particularly BRAC, engage in both agricultural programs and nutrition and health initiatives.

12

Table 5.2 Adding control variables to the WHZ and mild wasting fixed-effects models 1 2 3 4 Variable WHZ Mild wasting (WHZ < –1) Rice yields 0.83** 0.70*** –0.26** –0.22*

(0.31) (0.24) (0.12) (0.12)

Maternal education (years) 0.18** –0.03

(0.09) (0.02)

Paternal education (years) –0.01 0.01

(0.05) (0.02)

Households with toilets (%) 0.80** –0.21*

(0.36) (0.12)

Prenatal care (%) 0.29 –0.16

(0.35) (0.11)

Fertility rates 0.13 –0.02

(0.09) (0.03)

Women can visit clinics (%) 0.11 –0.08

(0.28) (0.09)

Nongovernmental organization members (%)

0.26 –0.07

(0.56) (0.16)

Age controls Yes Yes Yes Yes

Time effects Yes Yes Yes Yes

Fixed effects Yes Yes Yes Yes

R-squared .28 .44 .33 .40

Sample size 109 109 109 109

Source: Authors’ estimates. Note: WHZ = Weight for height z scores. Robust standard errors are included in parentheses. All indicators are measured for

children 0 to 24 months of age. *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at the 1 percent level.

Table 5.3 shows that adding time-varying control variables to the models reduces the effects of yield growth only very slightly. In the case of WHZ scores the coefficient drops from 0.83 to 0.70, while in the case of mild wasting it falls from –0.26 to –0.22. The results from Table 5.1 therefore appear quite robust to incorporating some of the alternative explanations of nutritional change in Bangladesh.

Yield-induced improvements in child weight gain suggest that children are given more food or possibly also better food (since micronutrients can also build up immunity to diseases, which can affect weight gain). The amount of food given to children is not measured in DHS, though as noted above, we do have an indicator of the timely introduction of solid or semisolid foods (complementary feeding). We also have a dietary diversity indicator from the 2000 DHS round onwards, as well as an indicator of dairy consumption. As we noted above, however, only the complementary feeding indicator has significantly improved over time. Specifically, the proportion of children in the 4 to 18 month age bracket who were given any solid or semisolid foods rose from a paltry 42 percent in 1996/1997 to 74 percent in 2011, while dietary diversity and milk consumption improved marginally up to 2007 but declined significantly from 2007 to 2011. A priori, it seems unlikely that yield growth has contributed much to the dietary diversification of children.

13

In Table 5.3 we regress these three indicators against rice yields while controlling for a larger set of time-varying variables discussed in the previous section. Consistent with expectations we observe a large and significant effect of rice yields on complementary feeding but no significant effects on dietary diversity or milk consumption. The results for complementary feeding in regression 1 imply that the 0.5 metric ton per acre increase in yields from 1997 to 2011 increased the prevalence of complementary feeding by around 10 percentage points, which amounts to about one-third of the actual change observed during that period. This suggests that—in the past, at least—poor rural Bangladeshi households would delay the introduction of complementary foods and instead extend their reliance of breast-feeding as the principal source of nutrition for young children.

Table 5.3 Fixed-effects estimates of the impact of yield growth on child-feeding indicators (in percentages) 1 2 3

Variable Complementary foods Minimum dietary

diversity Dairy consumption

Rice yields 0.21** –0.04 –0.17

(0.09) (0.13) (0.17)

Child age controls Yes Yes Yes

Additional time-varying controls Yes Yes Yes

Time effects Yes Yes Yes

Fixed effects Yes Yes Yes

R-squared .87 .88 .44

Sample size 109 89 109

Source: Authors’ estimates. Note: Robust standard errors are included in parentheses. Complementary feeding is the proportion of children aged 4 to 18

months who were given any solid or semisolid foods in the past 24 hours. Minimum dietary diversity is the proportion of children given more than four food groups. Dairy consumption refers to milk and yogurt consumption. Additional time-varying controls are maternal education (years), percentage of women with access to prenatal care, fertility rates, whether women can visit health clinics unaccompanied (percentage), and whether women are nongovernmental organization members (percentage). **Significant at the 5 percent level.

Finally, Table 5.4 reports associations between regressions of mild wasting and moderate stunting rates against the three indicators of feeding practices used in Table 5.3, with fixed effects, time period effects, and child age controls again included. Regression 1 shows that the introduction of complementary foods is strongly associated with child weight outcomes, with a coefficient of –0.19. A 30-point increase in complementary feeding—as observed from 1996/1997 to 2011—is associated with a decrease in wasting of almost 6 percentage points. Minimum dietary diversity and milk consumption have highly insignificant coefficients in the wasting model (regressions 2 and 3, respectively). In regression 4 we observe no significant association between complementary feeding and stunting, consistent with the lack of a significant impact of yields on stunting. However, minimum dietary diversity and dairy consumption are strongly and negatively associated with stunting rates, with comparable point estimates (–0.29).

14

Table 5.4 Fixed-effects estimates of the associations between child feeding indicators and child nutrition outcomes

1 2 3 4 5 6 Variable Mild wasting (WHZ < –1) Moderate stunting (HAZ < –2) Complementary feeding (%) –0.20** 0.08 (0.09) (0.14) Minimum diet diversity (%) 0.06 –0.29* (0.11) (0.15)

Dairy consumption (%) 0.07 –0.29** (0.09) (0.14) Age controls Yes Yes Yes Yes Yes Yes Time effects Yes Yes Yes Yes Yes Yes Fixed effects Yes Yes Yes Yes Yes Yes R-squared .28 .27 .26 .31 .24 .36 Sample size 107 88 107 109 89 109

Source: Authors’ estimates. Note: Robust standard errors are included in parentheses. Complementary feeding is the proportion of children aged 4 to 18

months who were given any solid or semisolid foods in the past 24 hours. Minimum dietary diversity is the proportion of children given more than four food groups. Dairy consumption refers to milk and yogurt consumption. In these regressions anthropometrics are also measured for children aged 4 to 18 months. Also note that two outliers were removed for both the WHZ and HAZ scores. However, the results are qualitatively robust to inclusion of these outliers. *Significant at the 10 percent level. **Significant at the 5 percent level.

Together, the results in Tables 5.3 and 5.4 suggest that rice productivity growth has significantly influenced child weight gain through the earlier introduction of solid and semisolid food and perhaps also greater quantities of calories. However, the insignificant effects of yield gains on linear growth in children seem related to the limited impacts of these gains on dietary diversification.

Our results are qualitatively robust to inclusion of outliers, to the exclusion of the Chittagong Hill districts (which have smaller DHS samples and somewhat less dependence on rice production), and to the inclusion of additional control variables (such as alternative indicators of access to health services). However, the results in Table 5.2 do not extend to older children aged 24 to 59 months (results available on request). This likely stems from the widely observed stabilization of anthropometric outcomes from 24 months of age onwards, including—perhaps—reduced sensitivity to household food insecurity for older children (certainly infants and younger children are especially vulnerable to combinations of poor care practices, low nutrient intake, and a harmful disease environment). Another explanation may be that older children are universally given complementary foods, such that substitution between prolonged breast-feeding and solid food consumption (our rationing hypothesis) is not a relevant mechanism linking agricultural production and weight gain.

Our final set of robustness checks are the Blundell-Bond GMM estimates reported in Tables 5.5 and 5.6. The Blundell-Bond estimates do not change the core conclusions of the fixed-effects models; specifically, they are never smaller than the fixed-effects estimates. In many cases, they are much larger in magnitude. The point estimate of the coefficient on yields in the WHZ regression, for example, rises from 0.83 in the fixed-effects model to 2.55 in the Blundell-Bond model. The analogous point estimates in the mild wasting model more than double, and unlike the fixed-effects estimate, the impact of yields on moderate wasting is statistically significant in the Blundell-Bond model. We note, however, that because these point estimates are more imprecisely measured in the Blundell-Bond model, we cannot reject the null hypothesis that they equal the fixed-effects estimates.

15

Table 5.5 Assessing the robustness of the impact of yields on child weight: Fixed-effects and Blundell-Bond generalized methods of moments estimates

1 2 3 4 5 6 WHZ Mild wasting

(WHZ < –1) Moderate wasting

(WHZ < –2) Variable FE Blundell-Bond FE Blundell-Bond FE Blundell-Bond Rice yields 0.83** 2.55*** –0.26** –0.62** –0.14 –0.47* (0.31) (0.90) (0.12) (0.30) (0.14) (0.27)

Lag-dependent variable –0.82*** –0.46** –0.25 (0.26) (0.18) (0.23) R-squared .28 .33 .39 Sample size 109 84 109 84 109 84 Arellano-Bond test 0.8 0.26 0.47 Sargan overid test 0.36 0.47 0.08** Hansen overid test 0.64 0.84 0.74 Number of instruments 27 27 27

Source: Authors’ estimates. Note: FE = fixed effects. *Significant at the 10 percent level. **Significant at the 5 percent level. ***Significant at

the 1 percent level.

In Table 5.6, the Blundell-Bond estimate of the effects of yields on complementary feeding is unchanged though marginally insignificant at the 10 percent level. Unlike the fixed-effects estimates, the Blundell-Bond estimate of yields on milk consumption is significant and positive, though in this case the model does not pass the Sargan overidentification test, so we treat this result cautiously. Overall, however, the Blundell-Bond models suggest that our fixed-effects results are lower-bound estimates of the impact of rice yields on child weight.

Table 5.6 Assessing the robustness of effects of yields on child-feeding indicators: Fixed-effects and Blundell-Bond generalized methods of moments estimates (in percentages)

1 2 3 4 5 6 Complementary feeding Minimum diet diversity Dairy consumption Variable FE Blundell-Bond FE Blundell-Bond FE Blundell-Bond Rice yields 0.28** 0.27 0.06 0.05 –0.02 0.57* (0.13) (0.22) (0.12) (0.21) (0.16) (0.30) Lag-dependent variable –0.14* –0.24 –0.06 (0.08) (0.16) (0.13) R-squared .28 .33 .39 Sample size 109 84 109 84 109 84 Arellano-Bond test 0.80 0.26 0.47 Sargan overid test 0.36 0.47 0.08** Hansen overid test 0.64 0.84 0.74 Number of instruments 27 27 27

Source: Authors’ estimates. Note: FE = fixed effects. Robust standard errors are included in parentheses. Complementary feeding is the proportion of

children aged 4 to 18 months who were given any solid or semisolid foods in the past 24 hours. In these regressions anthropometrics are also measured for children aged 4 to 18 months. Also note that two outliers were removed for both the regressions. *Significant at the 10 percent level. **Significant at the 5 percent level.

16

6. CONCLUSIONS

In this paper, we assess the impact of productivity growth in staple food production on preschool children’s nutritional status. We find that increases in rice yields have large and statistically significant impacts on child weight gain. The result appears to be explained by increased food consumption for young children, particularly the timelier introduction of solid and semisolid foods in the critical early window of child development. Conversely, we find no substantial evidence that growth in rice productivity has significantly improved linear growth of young children or the dietary diversification processes that are significantly associated with linear growth. Thus, while the estimated effects of yield growth on WHZ scores are large enough to account for much of the improvement in weight gain observed in rural areas of Bangladesh from 1997 to 2011, we find no evidence that yield growth explains much of the rapid improvement in stunting observed during this period. Of course, this does not mean that no such effects exist. Rice productivity growth might have had general equilibrium price effects—which our model is not suited to identifying—that favorably affected linear growth in children. Another possibility is that rice productivity growth had improvements on maternal nutrition, which heavily determines a child’s birth size. While there are some indirect indications that birth size improved quite substantially from 1997 to 2011 (Headey et al. 2015), the DHS do not directly measure birth size, meaning that we are unable to directly test this hypothesis.

Bearing these caveats in mind, our findings yield some important policy implications. In particular, while agricultural productivity growth in Bangladesh has made a significant contribution to alleviating food insecurity and acute undernutrition, its contribution to reducing chronic undernutrition has been more limited because of the limited effects of rice productivity on the dietary diversification of young children. A major challenge going forward, then, is to understand the determinants of dietary diversification in Bangladesh and policy options for promoting diversification. Examples include a rebalancing of Bangladesh’s agricultural research and development portfolio toward more micronutrient-rich crops and livestock products, expansion or improvements in homestead gardening programs (Iannotti, Cunningham, and Ruel 2009), and more effective use of trade mechanisms to bolster domestic production with imports. Linking these to improvements in women’s schooling and autonomy, with improved hygiene, sanitation, and child care practices (particularly around the introduction of more diverse diets), may well have especially large impacts. Further research, however, is needed in this area.

17

REFERENCES

Ahmed, A. U. 1993. Food Consumption and Nutritional Effects of Targeted Food Interventions in Bangladesh. Bangladesh Food Policy Project Manuscript 31. Washington, DC: International Food Policy Research Institute.

Ahmed, R., S. Haggblade, and T. Chowdhury. 2000. Out of the Shadow of Famine: Evolving Food Markets and Food Policy in Bangladesh. Baltimore: Johns Hopkins University Press.

Arellano, M., and S. Bond. 1991. “Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Situations.” Review of Economics and Statistics 58: 277–297.

Arimond, M., Ruel, M., 2006. “Dietary Diversity Is Associated with Child Nutritional Status: Evidence from 11 Demographic and Health Surveys.” The Journal of Nutrition 134: 2579–2585.

Balagtas, J. V., H. Bhandari, E. R. Cabrera, S. Mohanty, and M. Hossain. 2014. “Did the Commodity Price Spike Increase Rural Poverty? Evidence from a Long-run Panel in Bangladesh.” Agricultural Economics 45: 303–312.

Bangladesh Bureau of Statistics. 2014. Yearbook of Agricultural Statistics of Bangladesh. Dhaka.

Barrett, C. B., and M. R. Carter. 2010. “The Power and Pitfalls of Experiments in Development Economics: Some Non-random Reflections.” Applied Economic Perspectives and Policy 32: 515–548.

Behrman, J.R., Deolalikar, A.B., 1988. “Health and Nutrition” In Handbook of Development Economics, edited by H. Chenery and T. N. Srinivasan, 631–711. New York: Elsevier.

Berti, P. R., J. Krasevec, and S. FitzGerald. 2004. “A Review of the Effectiveness of Agriculture Interventions in Improving Nutrition Outcomes.” Public Health Nutrition 7: 599–609.

Bezemer, D., and D. D. Headey. 2008. “Agriculture, Development and Urban Bias.” World Development 36: 1342–1364.

Bhagowalia, P., D. D. Headey, and S. Kadiyala. 2012. Agriculture, Income, and Nutrition Linkages in India: Insights from a Nationally Representative Survey. Washington, DC: International Food Policy Research Institute.

Blundell, R., and S. Bond. 1998. “Initial Conditions and Moment Restrictions in Dynamic Panel Data Models.” Journal of Econometrics 87: 115–143.

Bouis, H. E. 2000. “Special Issue on Improving Nutrition through Agriculture.” Food and Nutrition Bulletin 21: 1–6.

Brainerd, E., and N. Menon. 2014. “Seasonal Effects of Water Quality: The Hidden Costs of the Green Revolution to Infant and Child Health in India.” Journal of Development Economics 107: 49–64.

Chowdhury, A. M., A. Bhuiya, M. E. Chowdhury, S. Rasheed, Z. Hussain, and L. C. Chen. 2013. “The Bangladesh Paradox: Exceptional Health Achievement Despite Economic Poverty.” Lancet 382: 1734–1745.

Dasgupta, S., Meisner, C., Huq, M., 2005. Health Effects and Pesticide Perception as Determinants of Pesticide Use: Evidence from Bangladesh. Policy Research Working Paper No. WPS3776. Washington DC: The World Bank.

de Beer, H. 2012. “Dairy Products and Physical Stature: A Systematic Review and Meta-analysis of Controlled Trials.” Economics and Human Biology 10: 299–309.

Diao, X., P. Hazell, and J. Thurlow. 2010. “The Role of Agriculture in African Development.” World Development 38: 1375–1383.

Diao, X., D. Headey, and M. Johnson. 2008. “Toward a Green Revolution in Africa: What Would It Achieve, and What Would It Require?” Agricultural Economics 39: 539–550.

Dillon, A., K. McGee, and G. Oseni. 2014. Agricultural Production, Dietary Diversity, and Climate Variability. Policy Research Working Paper No. 7022. Washington, DC: World Bank.

18

Dorosh, P. A., and Q. Shahabuddin. 2002. Rice Price Stabilization in Bangladesh. MTID Discussion Paper No. 46. Washington, DC: International Food Policy Research Institute.

Elbers, C., and J. W. Gunning. 2013. “Evaluation of Development Programs: Randomized Controlled Trials or Regressions?” World Bank Economic Review 28: 432–445.

Evenson, R. E., and D. Gollin. 2003. “Assessing the Impact of the Green Revolution, 1960 to 2000.” Science 300: 758–762.

FAO (Food and Agriculture Organization of the United Nations). 2014. AGROSTAT. Rome.

Hanif, M. H. 2013. “Trends in Infant and Young Child Feeding Practices in Bangladesh, 1993–2011.” International Breastfeeding Journal 8: 1–6.

Hazell, P. B. 2009. The Asian Green Revolution. IFPRI Discussion Paper 00911.Washington, DC: International Food Policy Research Institute.

Headey, D., A. Chiu, and S. Kadiyala. 2012. “Agriculture’s Role in the Indian Enigma: Help or Hindrance to the Malnutrition Crisis?” Food Security 4: 87–102.

Headey, D., Hoddinott, J., Ali, D., Tesfaye, R., Dereje, M., 2015. The Other Asian Enigma: Explaining the Rapid Reduction of Undernutrition in Bangladesh. World Development 66: 749–761.

Hoddinott, J., D. Headey, and M. Dereje. 2015. “Cows, Missing Milk Markets and Nutrition in Rural Ethiopia.” Journal of Development Studies, forthcoming.

Hoppe, C., C. Mølgaard, and K. F. Michaelsen. 2006. “Cow’s Milk and Linear Growth in Industrialized and Developing Countries.” Annual Review of Nutrition 26: 131–173.

Hossain, M. 2004. Rural Non-farm Economy in Bangladesh: A View from Household Surveys. CPD Occasional Paper Series No. 40. Dhaka, Bangladesh: Centre for Policy Dialogue.

Hossain, M., M. L. Bose, and B. A. A. Mustafi. 2006. “Adoption and Productivity Impact of Modern Rice Varieties in Bangladesh.” Developing Economies 44: 149–166.

Iannotti, L., K. Cunningham, and M. Ruel. 2009. Improving Diet Quality and Micronutrient Nutrition: Homestead Food Production in Bangladesh. IFPRI Discussion Paper No. 928. Washington, DC: International Food Policy Research Institute.

Iannotti, L., E. Muehlhoff, and D. Mcmahon. 2013. “Review of Milk and Dairy Programmes Affecting Nutrition.” Journal of Development Effectiveness 5: 82–115.

Jayachandran, S., and R. Pande. 2013. Why Are Indian Children Shorter than African Children? Department of Economics Working Paper. Cambridge, MA, US: Harvard University.

Kabir, I., M. Khanam, K. E. Agho, S. Mihrshahi, M. J. Dibley, and S. K. Roy. 2012. “Determinants of Inappropriate Complementary Feeding Practices in Infant and Young Children in Bangladesh: Secondary Data Analysis of Demographic Health Survey 2007.” Maternal & Child Nutrition 8: 11–27.

Kar, K. 2003. Subsidy or Self-respect? Participatory Total Community Sanitation in Bangladesh. IDS Working Paper 184. Brighton, UK: Institute for Development Studies.

Khandker, S. R. 2005. “Microfinance and Poverty: Evidence Using Panel Data from Bangladesh.” World Bank Economic Review 19: 263–286.

LaFave, D., Thomas, D., 2014. Farms, Families, and Markets: New Evidence on Agricultural Labor Markets. NBER Working Paper Series No. 20699. Cambridge, MA: National Bureau of Economic Research.

Leroy, J. L., and E. A. Frongillo. 2007. “Can Interventions to Promote Animal Production Ameliorate Undernutrition?” Journal of Nutrition 137: 2311–2316.

Masset, E., L. Haddad, A. Cornelius, and J. Isaza-Castro. 2012. “Effectiveness of Agricultural Interventions That Aim to Improve Nutritional Status of Children: Systematic Review.” BMJ 344. doi:http://dx.doi.org/10.1136/bmj.d8222.

19

Mellor, J. W. 1976. The New Economics of Growth: A Strategy for India and the Developing World. Ithaca, NY: Cornell University Press.

Menon, P., 2012. “The Crisis of Poor Complementary Feeding in South Asia: Where Next?” Maternal & Child Nutrition 8: 1–4.

Hoppe, C., C. Molgaard, K. F. Michaelsen. 2006. “Cow's Milk and Linear Growth in Industrialized and Developing Countries.” Annual Review of Nutrition 26: 131–173.

Molini, V., M. Nubé, and B. van den Boom. 2010. “Adult BMI as a Health and Nutritional Inequality Measure: Applications at Macro and Micro Levels.” World Development 78: 1012–1023.

Naher, F. 1997. “Green Revolution in Bangladesh: Production Stability and Food Self-sufficiency.” Economic and Political Weekly 32: A84–A89.

Nazneen, A., Z. Bakht, P. Dorosh, and Q. Shahabuddin. 2007. Distortions to Agricultural Incentives in Bangladesh. Agricultural Distortions Working Paper 32. Washington, DC: World Bank.

Orr, A., B. Adoplh, R. Islam, M. Rahman, B. Barua, and M. K. Roy. 2009. Pathways from Poverty: The Process of Graduation in Rural Bangladesh. Dhaka, Bangladesh: University Press.

Pinstrup-Andersen, P., 2013.” Nutrition-sensitive Food Systems: From Rhetoric to Action.” The Lancet 38: 375–376.

Pinstrup-Andersen, P., and M. Jaramillo. 1991. “The Impact of Technological Change in Rice Production on Food Consumption and Nutrition.” In The Green Revolution Reconsidered: The Impact of the High Yielding Rice Varieties in South India, edited by P. B. Hazell and C. Ramaswamy. Baltimore: Johns Hopkins University Press; New Delhi: Oxford University Press.

Ravallion, M. 1990. “Rural Welfare Effects of Food Price Changes under Induced Wage Responses: Theory and Evidence for Bangladesh.” Oxford Economics Papers 42: 574–585.

Rawlins, R., S. Pimkina, C. B. Barrett, S. Pedersen, and B. Wydick. 2014. “Got Milk? The Impact of Heifer International’s Livestock Donation Programs in Rwanda on Nutritional Outcomes.” Food Policy 44: 202–213.

Roodman, D. 2009. “How to Do Xtabond2: An Introduction to Difference and System GMM in Stata.” Stata Journal 9: 86–136.

Ruel, M. T., and H. Alderman. 2013. “Nutrition-sensitive Interventions and Programmes: How Can They Help to Accelerate Progress in Improving Maternal and Child Nutrition?” Lancet 382:536–551.

Senarath, U., K. E. Agho, D.-e.-S. Akram, S. S. P. Godakandage, T. Hazir, H. Jayawickrama, N. Joshi, I. Kabir, M. Khanam, A. Patel, Y. Pusdekar, S. K. Roy, I. Siriwardena, K. Tiwari, and M. J. Dibley. 2012. “Comparisons of Complementary Feeding Indicators and Associated Factors in Children Aged 6–23 Months across Five South Asian Countries.” Maternal & Child Nutrition 8:89–106.

Shahabuddin, Q., and P. A. Dorosh. 2002. Comparative Advantage in Bangladesh Crop Production. MTID Discussion Paper No. 47. Washington, DC: International Food Policy Research Institute.

Singh, I., Squire, L., Strauss, J., 1986. Agricultural Household Models: Extension, Application and Policy. Baltimore: Johns Hopkins University Press.

Smith, L. C., U. Ramakrishnan, A. Ndiaye, L. Haddad, and R. Martorell. 2003. “The Importance of Women’s Status for Child Nutrition in Developing Countries.” In Research Report No. 131. Washington, DC: International Food Policy Research Institute.

Torlesse, H., L. Kiess, and M. W. Bloem. 2003. “Association of Household Rice Expenditure with Child Nutritional Status Indicates a Role for Macroeconomic Food Policy in Combating Malnutrition.” Journal of Nutrition 133: 1320–1325.

Victora, C. G., M. de Onis, P. Curi Hallal, M. Blössner, and R. Shrimpton. 2009. “Worldwide Timing of Growth Faltering: Revisiting Implications for Interventions.” Pediatrics 2010: 473–480.

20

World Bank. 2005. Maintaining Momentum to 2015? An Impact Evaluation of Interventions to Improve Maternal and Child Health and Nutrition Outcomes in Bangladesh. Washington, DC: Operations Evaluation Department, World Bank.

Zhang, X., S. Rashid, K. Ahmad, V. Mueller, H. L. Lee, S. Lemma, S. Belal, and A. U. Ahmed. 2013. Rising Wages in Bangladesh. Washington, DC: International Food Policy Research Institute.

RECENT IFPRI DISCUSSION PAPERS

For earlier discussion papers, please go to www.ifpri.org/pubs/pubs.htm#dp. All discussion papers can be downloaded free of charge.

1422. Rural finance and agricultural technology adoption in Ethiopia: Does institutional design matter? Gashaw Tadesse Abate, Shahidur Rashid, Carlo Borzaga, and Kindie Getnet, 2015.

1421. Is more inclusive more effective? The “new-style” public distribution system in India. Avinash Kishore and Suman Chakrabarti, 2015.

1420. Explicitly integrating institutions into bioeconomic modeling. Kimberly A. Swallow and Brent M. Swallow, 2015.

1419. Time allocation to energy resource collection in rural Ethiopia: Gender-disaggregated household responses to changes in firewood availability. Elena Scheurlen, 2015.

1418. Communities’ perceptions and knowledge of ecosystem services: Evidence from rural communities in Nigeria. Wei Zhang, Edward Kato, Prapti Bhandary, Ephraim Nkonya, Hassan Ishaq Ibrahim, Mure Agbonlahor, and Hussaini Yusuf Ibrahim, 2015.

1417. 2011 social accounting matrix for Senegal. Ismaël Fofana, Mamadou Yaya Diallo, Ousseynou Sarr, and Abdou Diouf, 2015.

1416. Firm heterogeneity in food safety provision: Evidence from Aflatoxin tests in Kenya. Christine Moser and Vivian Hoffmann, 2015.

1415. Mechanization outsourcing clusters and division of labor in Chinese agriculture. Xiaobo Zhang, Jin Yang, and Thomas Reardon, 2015.

1414. Conceptualizing drivers of policy change in agriculture, nutrition, and food security: The Kaleidoscope Model. Danielle Resnick, Suresh Babu, Steven Haggblade, Sheryl Hendriks, and David Mather, 2015.

1413. Value chains and nutrition: A framework to support the identification, design, and evaluation of interventions. Aulo Gelli, Corinna Hawkes, Jason Donovan, Jody Harris, Summer Allen, Alan de Brauw, Spencer Henson, Nancy Johnson, James Garrett, and David Ryckembusch, 2015.

1412. Climate change adaptation assets and group-based approaches: Gendered perceptions from Bangladesh, Ethiopia, Mali, and Kenya. Noora Aberman, Snigdha Ali, Julia A. Behrman, Elizabeth Bryan, Peter Davis, Aiveen Donnelly, Violet Gathaara, Daouda Kone, Teresiah Nganga, Jane Ngugi, Barrack Okoba, and Carla Roncoli, 2015.

1411. Information networks among women and men and the demand for an agricultural technology in India. Nicholas Magnan, David J. Spielman, Kajal Gulati, and Travis J. Lybbert, 2015.

1410. Measurement of agricultural productivity in Africa South of the Sahara: A spatial typology application. Bingxin Yu and Zhe Guo, 2015.

1409. Eliciting farmers’ valuation for abiotic stress-tolerant rice in India. Anchal Arora, Sangeeta Bansal, and Patrick S. Ward, 2015.

1408. Understanding the policy landscape for climate change adaptation: A cross-country comparison using the net-map method. Noora Aberman, Regina Birner, Eric Haglund, Marther Ngigi, Snigdha Ali, Barrack Okoba, Daouda Koné, Tekie Alemu, 2015.

1407. Evolving public expenditure in Chinese agriculture definition, pattern, composition, and mechanism. Bingxin Yu, Kevin Chen, Yumei Zhang, and Haisen Zhang, 2014.

1406. Fertility, agricultural labor supply, and production. Bjorn Van Campenhout, 2014.

1405. Impact simulation of ECOWAS rice self-sufficiency policy. Ismaël Fofana, Anatole Goundan, and Léa Vicky Magne Domgho, 2014.

1404. Political economy of state interventions in the Bangladesh food-grain sector. Nurul Islam, 2014.

1403. Loan demand and rationing among small-scale farmers in Nigeria. Aderibigbe S. Olomola and Kwabena Gyimah-Brempong, 2014.

1402. Revisiting the labor demand curve: The wage effects of immigration and women’s entry into the US labor force, 1960–2010. Alan de Brauw and Joseph R. D. Russell, 2014.

INTERNATIONAL FOOD POLICY RESEARCH INSTITUTE

www.ifpri.org

IFPRI HEADQUARTERS 2033 K Street, NW Washington, DC 20006-1002 USA Tel.: +1-202-862-5600 Fax: +1-202-467-4439 Email: ifpri@cgiar.org