WPS7200 Policy Research Working Paper 7200 Drinking Water Salinity and Infant Mortality in Coastal Bangladesh Susmita Dasgupta Mainul Huq David Wheeler Development Research Group Environment and Energy Team February 2015 Policy Research Working Paper 7200 Abstract Bangladesh, with two-thirds of its land area less than five survey information from the Bangladesh Demographic and meters above sea level, is one of the most climate-vulnerable Health Surveys for 2004 and 2007. It uses probit and logit countries in the world. Low-lying coastal districts along analyses to estimate mortality probability for infants less the Bay of Bengal are particularly vulnerable to sea level than two months old. Controlling for many other determi- rise, tidal flooding, storm surges, and climate-induced nants of infant mortality, the analysis finds high significance increases in soil and water salinity. This paper investigates for salinity exposure during the last month of pregnancy the impact of drinking water salinity on infant mortality and no significance for exposure during the preceding in coastal Bangladesh. It focuses on the salinity of drinking months. The estimated impact of salinity on infant mor- water consumed during pregnancy, which extensive medical tality is comparable in magnitude to the estimated effects research has linked to maternal hypertension, preeclamp- of traditionally-cited variables such as maternal age and sia, and post-partum morbidity and mortality. The study education, gender of the household head, household wealth, combines spatially-formatted salinity measures for 2001–09 toilet facilities, drinking water sources, and cooking fuels. provided by Bangladesh with individual and household This paper is a product of the Environment and Energy Team, Development Research Group. It is part of a larger effort by the World Bank to provide open access to its research and make a contribution to development policy discussions around the world. Policy Research Working Papers are also posted on the Web at http://econ.worldbank.org. The authors may be contacted at Sdasgupta@worldbank.org. The Policy Research Working Paper Series disseminates the findings of work in progress to encourage the exchange of ideas about development issues. An objective of the series is to get the findings out quickly, even if the presentations are less than fully polished. The papers carry the names of the authors and should be cited accordingly. The findings, interpretations, and conclusions expressed in this paper are entirely those of the authors. They do not necessarily represent the views of the International Bank for Reconstruction and Development/World Bank and its affiliated organizations, or those of the Executive Directors of the World Bank or the governments they represent. Produced by the Research Support Team Drinking Water Salinity and Infant Mortality in Coastal Bangladesh Susmita Dasgupta1, Mainul Huq2, David Wheeler3 Keywords: Bangladesh, Coastal Areas, Infant Mortality, Water Salinity JEL Classification: I00, Q54 * Authors’ names are in alphabetical order. This research was funded by the Knowledge for Change Program. We would like to extend our special thanks to Dr. Moqbul Hossain and Dr. Zahirul Huque Khan for the soil salinity and river salinity data in coastal Bangladesh. We are thankful to Dr. Abbas Bhuyia, Acting Executive Director, ICDDRB and Dr. Michael Toman, Research Manager, World Bank for their comments and suggestions. 1 Lead Environmental Economist, Development Research Group, World Bank 2 Consultant, World Bank 3 Senior Fellow, World Resources Institute Introduction The potential impacts of climate change on coastal regions include progressive inundation from sea level rise, heightened storm damage, loss of wetlands, and increased salinity from saltwater intrusion. Worldwide, about 600 million people currently inhabit low-elevation coastal zones and will be affected by progressive salinization (Wheeler 2011; CIESIN 2010). Recent research suggests that the sea level may rise by one meter or more in the 21st century, which would increase the vulnerable population to about one billion by 2050 (Hansen and Sato 2011; Vermeer and Rahmstorf 2009; Pfeffer, et al. 2008; Rahmstorf 2007; Dasgupta, et al. 2009; Brecht, et al. 2012). While most research has focused on inundation and losses from heightened storm surges, increased salinity from saltwater intrusion may pose another serious threat through its impact on household water supplies. Understanding the significance and magnitude of this threat may be critical for long-term development and poverty alleviation in countries with vulnerable coastal regions (Brecht, et al. 2012). Bangladesh provides an excellent setting for investigation; with two-thirds of its land area less than 5 meters above sea level, it is one of the countries most threatened by sea level rise and saltwater intrusion. About 30% of the cultivable land of Bangladesh is in coastal areas where salinity is affected by tidal flooding during the wet season, direct inundation by storm surges, and movement of saline groundwater during the dry season (Haque, 2006). In consequence, the potential impact of salinity has become a major concern for the Government of Bangladesh and affiliated research institutions.4 Recently, the Bangladesh Climate Change Resilience Fund (BCCRF) Management Committee and the 4 Prior research on salinity intrusion has been conducted or co-sponsored by the Ministry of Environment and Forests (World Bank 2000) and two affiliated institutions: the Center for Geographic and Environmental Information Services (Hassan and Shah 2006) and the Institute of Water Modeling (IWM 2003; UK DEFR 2007). Additional research has been conducted by the Bangladesh Center for Advanced Studies (World Bank 2000; Khan, et al. 2011), the Bangladesh Agricultural Research Council (Karim, et al. 1982, 1990), and the Bangladesh Soil Resources Development Institute (SRDI 1998a,b; Peterson and Shireen 2001). 2 World Bank have highlighted salinity intrusion in coastal Bangladesh as a critical part of adaptation to climate change (Dasgupta et al. 2014a and Dasgupta et al. 2014b).5 Resources will remain scarce, and mobilizing a cost-effective response will require an integrated spatial analysis of salinity diffusion, its impacts, and the costs of prevention, adaptation and remediation. This paper attempts to contribute by addressing two critical components of the problem: salinity changes in the coastal region of Bangladesh, and their consequences for maternal and child health in the region. The paper focuses on the salinity of drinking water consumed during pregnancy, which extensive medical research has linked to maternal hypertension, preeclampsia and post-partum morbidity and mortality.6 Methods Previous Research Recent micro-level research on drinking water salinity in the coastal region of Bangladesh and its impact on hypertension in pregnant women (Khan, et al. 2008, Khan, et al. 2011) draws on numerous international observational studies and clinical trials that establish a strong link between higher salt intake and elevated blood pressure (Alderman 2000; Calabrese and Tuthill 1981; Hallenbeck, et al. 1981; He and MacGregor 2007; Midgley, et al. 1996; Welty, et al. 1986). The current state of the art has been established by Khan, et al. (2011), who focus on the links between drinking water salinity and 5 Sarwar (2005) and SRDI (1998a,b, 2000, 2010) have documented changes in salinity that have accompanied coastal subsidence and thermal expansion of the ocean. Detailed assessments of salinization have employed two principal methods. One approach focuses on simulation of salinity change in rivers and estuaries, using hydraulic engineering models whose results are compared with actual measures (Bhuiyan and Dutta 2011; Mohal et al. 2006; Aerts, et al. 2000; Nobi and Das Gupta 1997). Another approach focuses on local salinity impacts, using surveys and descriptive statistics (Mahmood, et al. 2010; Khan, et al. 2008, 2011; Haque 2006; Sarwar 2005; Rahman and Ahsan 2001; Hassan and Shah 2006; Karim, et al. 1982, 1990). In the most comprehensive study to date, Dasgupta, et al. (2014a) develop detailed projections of river salinity through 2050 in Bangladesh’s coastal region. 6 This is a serious public health issue because hypertension in pregnancy is associated with numerous maternal and infant health threats, including impaired liver function, intrauterine growth retardation, preterm birth, and maternal and perinatal deaths (Sibai, 2002). 3 hypertension during pregnancy in Dacope Upazilla, Khulna. The study infers the impact of salinity on maternal health from three information sources: monthly measurements of salinity in shallow and deep groundwater tubewells at various sites in the Khulna region and along the Passur River during the period 1998-2000; urine samples and blood pressure measures for 343 pregnant Dacope women during the 2009-2010 dry season (October 2009 through March 2010); and hospital-based measures of the prevalence of hypertension for 969 pregnant women during the period July 2008 - March 2010. The study has three principal findings: (1) The salinity of tubewell water in the study area was far higher during dry season months in 1998-2000. (2) Pregnant women who drank tubewell water during the dry season of 2009-2010 had significantly higher urinary sodium than women from the sample group who drank rainwater during that period. However, the study does not establish a significant link between urinary sodium and hypertension. (3) Hospital-based blood pressure measures during wet and dry seasons during the period 2008-2009 revealed a significantly higher incidence of hypertension during the dry season. This pioneering inferential study yields valuable insights into the relationship between salinity and maternal health in Bangladesh. However, its restricted geographic coverage, limited temporal span, and temporal mismatch of data sources have left a significant gap between its findings and the results that could be delivered by the longitudinal study that we described above. Extensions in This Paper In this study, we attempt to narrow the research gap by mobilizing information that features a larger sample population, broader geographic coverage, a longer temporal span, and temporally-matched data. Monitoring household-specific drinking water and health indicators was beyond the scope of our research. Instead, we relied on (1) proxies for drinking water salinity: monthly spatially-formatted data from 41 soil salinity monitoring stations for the period 2001-2009, provided by the Bangladesh Soil 4 Research Development Institute; (2) spatially-formatted Bangladesh Demographic and Health Surveys (DHS) conducted during the same period. The 2004 Bangladesh DHS includes information on 11,440 women age 10-49 and 4,297 men age 15-54 from 10,500 households, covering 122 urban and 239 rural clusters (NIPORT, et al. 2005). The 2007 Bangladesh DHS includes information on 10,996 women age 15-49 and 3,771 men age 15-54 from 10,400 households in 134 urban and 227 rural clusters (NIPORT, et al. 2009). The Bangladesh DHS surveys do not include hypertension measures for the women surveyed. However, their information on cluster location allows us to construct monthly indicators of drinking water salinity for each surveyed woman during a nine-year period (2001-2009). The surveys also include temporally-specific information on mortality events for children in the surveyed households. The data therefore permit estimation of a reduced-form model that relates mothers’ consumption of saline water during pregnancy to post-natal mortality outcomes for their children.7 Model Specification We focus on the relationship between drinking water salinity during pregnancy and mortality during the first two months of life. Using data for individual children, we estimate equations for mortality events with the following general form: (1) = 0 + ∑ =1 + ∑=1 + ∑=1 + ∑ℎ=1 ℎ ,−ℎ + where, for child i, mother j, household k, DHS cluster l and birth month t: P = Event probability C = Child characteristics M = Mother characteristics H = Household characteristics St-h = Drinking water salinity index, lagged h months from the child’s birth month Ɛ = Random error term 7 We should emphasize that this is an exercise in statistical control, not direct experimental control. A quasi-experimental approach might test for significance using an arbitrary segmentation of sample-population mothers into “exposed” and “unexposed” groups residing in “low-salinity” and “high-salinity” areas. In contrast, our regression-based approach estimates the size and significance of incremental effects using continuous measures of exposure. Although our sample population resides solely in the coastal region, our salinity monitoring data indicate that salinity exposure for this population varies from negligible to very high. 5 Database Construction Cluster Estimates of Soil Salinity Figure 1 displays our study area, which comprises four regions of southern Bangladesh: Barisal, Chittagong, Dhaka and Khulna. Monitoring stations are color-coded by their mean soil salinity measure in 2001 - 2009.8 Figure 1 and Table 1 illustrate the substantial variation in measured salinity across monitors, years and seasons within years. In Figure 1, clusters of monitors with low and high mean readings are particularly noticeable in north and central Khulna. Seasonal variation is clear in Table 1, with minimum and maximum measures often differing by several multiples in the same year. Cross- monitor variations are also evident, particularly for maximum readings. Figure 1: Soil salinity monitoring stations and DHS survey clusters in Khulna, Dhaka, Barisal and Chittagong 8 The standard sample-based measure for soil salinity is electrical conductivity (in dS/m -- deciSiemens per meter). 6 Table 1: Soil salinity monitor measures (dS/m): 2001, 2005, 2009 2001 2005 2009 Local District Upazila Union Min Max Min Max Min Max No. Bagerhat Mongla Burirdanga 1 7.50 13.10 4.70 23.10 5.70 23.90 Bagerhat Mongla Burirdanga 2 6.80 16.40 1.70 14.70 4.80 22.20 Bagerhat Morrelganj Boloibunia 1 2.50 5.60 2.00 17.20 1.10 9.10 Barguna Amtali Amtali Paurashava 1 0.59 2.40 0.63 2.46 1.09 2.85 Barguna Amtali Amtali Paurashava 2 0.71 1.89 0.85 1.86 0.96 2.10 Bhola Bhola Sadar Illisha 1 0.99 6.56 0.18 4.00 0.21 1.09 Bhola Charfesson Aslampur 1 0.81 10.07 0.50 8.42 0.36 2.28 Bhola Charfesson Betua 1 0.90 2.80 0.40 11.54 1.24 12.18 Bhola Tazumuddin Shambhupur 1 0.64 2.38 0.26 3.94 0.25 1.93 Chittagong Sadar Patenga 1 1.12 4.13 0.35 3.13 0.80 3.00 Jessore Abhoynagar Noapara Paurshava 1 0.40 0.60 0.30 0.90 1.04 2.31 Jessore Keshabpur Sagardari 1 0.70 1.00 0.40 0.90 1.18 2.78 Khulna Batiaghata Batiaghata 1 2.20 6.80 2.20 11.50 1.40 22.50 Khulna Batiaghata Batiaghata 2 1.00 8.30 2.10 8.90 1.50 17.80 Khulna Batiaghata Batiaghata 3 2.00 11.10 2.20 21.90 1.20 14.60 Khulna Batiaghata Batiaghata 4 2.20 17.40 2.50 17.20 1.20 13.50 Khulna Batiaghata Jalma 1 1.60 10.10 1.80 11.30 1.10 19.00 Khulna Batiaghata Jalma 2 1.60 9.80 1.80 8.40 1.10 12.80 Khulna Dumuria Gutudia 1 1.20 25.20 4.70 28.90 1.40 19.70 Khulna Dumuria Gutudia 2 1.30 6.80 9.60 29.10 2.20 15.30 Khulna Paikgacha Paikgacha Paurshava 1 5.40 37.40 2.60 22.50 1.30 43.30 Narail Lohagara Kotakul 1 0.60 0.90 0.70 1.00 0.80 2.78 Narail Sadar 1 0.60 0.80 0.50 1.10 0.92 1.18 Narail Sadar Tularampur 1 0.60 0.90 0.70 1.30 0.92 2.04 Narail Kalia Babra Hachla 1 0.50 0.80 0.40 1.50 1.04 2.78 Narail Sadar Singasolpur 1 0.60 0.90 0.60 1.20 0.92 1.18 Noakhali Sadar 1 1.60 11.50 0.60 14.20 0.20 4.60 Noakhali Sadar 2 1.40 3.40 0.30 2.70 0.10 1.20 Noakhali Sadar Noakhali Paurashava 1 1.50 6.60 0.20 15.20 0.10 1.20 Noakhali Subarnachar Char Bata 1 1.90 24.50 0.50 19.90 0.20 9.50 Noakhali Subarnachar Char Jubilee 1 1.40 8.80 0.10 15.30 0.10 3.20 Patuakhali Dumki Lebukhali 1 0.86 1.98 0.65 2.20 0.91 2.49 Patuakhali Dumki Lebukhali 2 0.86 2.00 0.73 2.30 0.87 2.24 Patuakhali Kalapara Kalapara Paurashava 1 1.10 4.11 1.00 4.14 1.00 5.31 Patuakhali Kalapara Kalapara Paurashava 2 1.17 4.39 2.00 4.76 2.22 4.10 Patuakhali Kalapara Nilganj 1 1.25 8.47 1.60 5.73 1.71 6.74 Pirojpur Bhandaria Nudmulla 1 0.22 0.86 0.24 1.82 0.18 1.14 Pirojpur Mathbaria Sapleza 1 0.50 5.78 0.47 3.14 0.24 2.83 Pirojpur Mathbaria Tushkhali 1 0.79 1.73 0.36 3.10 0.24 9.80 Pirojpur Nazirpur Nazirpur 1 0.61 4.17 0.72 4.50 Pirojpur Nazirpur sekhmatia 1 1.24 3.86 0.70 3.90 0.54 6.46 7 Cluster locations for the DHS surveys are identified in Figure 1 by circles for 2004 and stars for 2007.9 They are sited at widely-varying distances from salinity monitors, with some clusters close to several monitors. Using the information illustrated by Figure 1, we estimate salinities for DHS clusters using two basic assumptions. First, we assume that the soil salinity measure for an area provides a useful proxy indicator for drinking water salinity. This seems reasonable on two grounds. First, runoff from fields and subsurface diffusion should produce higher water source salinity in areas with more saline soils. As Yu (2010) shows, soil composition in coastal Bangladesh is particularly conducive to vertical diffusion of salinity from the surface to groundwater. Second, in cases where soil monitors are near more scattered river monitors, Dasgupta, et al. (2011a) show that neighboring soil and water salinity readings are highly correlated.10 This provides further evidence of diffusion from runoff. Although we believe that local soil salinity provides a useful proxy indicator for drinking water salinity, we recognize that this assumption introduces random measurement error of unknown magnitude. Our second assumption is that we can use spatial interpolation to compensate for the locational mismatch between monitors and DHS clusters in our database. We apply spatial interpolation in two steps. First, we select monitors within a constant distance from each DHS cluster. Then, for each month and year, we estimate cluster salinity as the weighted average of monitor salinity readings, where the weights are inversely proportional to squared distances. To test for robustness, we create three sets of cluster estimates for bounding distances of 20, 30 and 40 km. We face an unavoidable tradeoff in setting the bound: Greater distance permits salinity estimation for more DHS clusters, but it may also 9 The indicated locations are cluster centroids supplied by the DHS, which randomly displaces GPS latitude/longitude positions to ensure confidentiality for respondents. DHS-supplied urban centroids are within 2 km of actual centroids; 99% of rural cluster centroids are within 5 kilometers of actual centroids and 1% are within 10 km. 10 This study uses the soil salinity monitors because they provide many more observations for estimation. Soil monitors substantially outnumber river monitors in the databases available for this study. The soil monitors are also more widely scattered geographically, so they are relatively close to many more DHS clusters than the river monitors. 8 introduce more spatial interpolation error. We explore the implications in our discussion of the estimation results. DHS Variables Many studies have investigated the determinants of mortality risk for infants in Bangladesh and other low-income countries (Amouzou and Hill 2004; Bicego and Ahmad 1996; Chowdhury 2013; Das Gupta 1990; Hobcraft, et al. 1984, 1985; Jain 1985; Majumder, et al. 1997; Martin, et al. 1983; Murphy and Wang 2001; World Bank 1993). This paper draws on their findings to identify potentially- significant variables that are provided by the Bangladesh DHS. They include numerous variables for children (e.g., gender); parents (e.g., age, education); and households (e.g., gender of household head, wealth status, sanitation facilities, drinking water sources, cooking fuels). Model Estimation To maximize degrees of freedom, we estimate our model using a combined sample from DHS 2004 and 2007 that contains information on 39,150 children. We perform all estimations using standard and robust estimators for both probit and logit models. The results across all four estimators are effectively identical, so we report only the standard probit estimates in this paper.11 DHS Variable Selection We identify potentially-significant child, parent and household variables in the DHS data for full sample estimation for infant mortality, excluding salinity because inclusion drastically reduces degrees of freedom by restricting the sample to clusters within 40 km of salinity monitors in the coastal region. Variables included in this step are gender of child; age and education of the mother; gender of household head and household wealth; and selected dummy variables for toilet facilities (flush toilets, pit latrines), water sources (tubewells, springs and other surface water) and cooking fuels (clean fuels - electricity and 11 The other estimates are available from the authors on request. We have not estimated spatial variants of these models because the appropriate estimators are not yet available in Stata. 9 gas). The dummy variables provide tests against excluded dummies for inferior toilet facilities (none), superior water sources (pipes and public taps), and inferior cooking fuels (principally charcoal, wood, straw and dung).12 Results Results for DHS Variables We present our full-sample results for the DHS variables in the first column of Table 2.13 These results are of independent interest, since they provide new insight into the determinants of infant mortality in Bangladesh.14 All of the significant parameters have the expected signs: The probability of death is significantly lower for infants whose mothers are older, more highly educated, and live in female-headed households with access to piped or tap water and flush toilets or pit latrines. We find no significance in the full sample for child gender, household wealth, tubewell water sources or clean cooking fuels.15 Results for Salinity Columns (2) - (4) of Table 2 summarize an extensive exploration of lagged salinity impacts that is illustrated for the 40 km distance bound in Table 3. Using the selected DHS variables, we successively introduce lagged values of our salinity indicator, starting with the first month before each child’s birth. Our standard and robust probit and logit estimators all yield the same strong result: Only salinity in the 12 Appendix A provides more detailed information on variable selection for toilet facilities, drinking water sources and cooking fuels. 13 It is important to note that these are regression results for data on individuals, not statistics for sub-groups in the population. Employment of the latter would require sample weighting. As the illustration in Table 4 shows, our microeconometric results can be used to construct estimates for arbitrarily-segmented sub-groups. 14 The term “infant mortality” is commonly understood to mean mortality of infants less than twelve months old. For expositional simplicity in this paper, we use the term for mortality of infants less than two months old. 15 Previous micro-level research by the authors on fuel use and indoor air pollution in Bangladesh (Dasgupta, et al. 2006) finds a significant relationship between cooling fuels and indoor levels of harmful particulate air pollution. However, the results also show that structural conditions (e.g., building materials, kitchen locations) play a more significant role. In addition, the previous work does not test the impact of indoor particulate pollution on infant survival probability. Our weak results in the current research suggest that robust inferences about the impact of cooking fuel choice on infant health require much more detailed structural and environmental data than are available for this exercise. 10 first month before birth is significant. Accordingly, we include this variable in the final estimates that are reported in Table 2. All of our estimates include monthly dummy variables (excluded from the tables for brevity) to ensure that our results for salinity (which varies greatly between wet and dry seasons) are not simply reflections of other seasonal factors that affect the incidence of infant mortality. We obtain strong results for salinity despite our monthly controls because, as Table 1 shows, salinity exhibits substantial independent variation, both over time within locations and across locations at each point in time. In any case, the risk of seasonal correlation is minimal, since our salinity variable is separately lagged for each child relative to birth month, not calendar month. To test robustness and parameter stability, we present results for cluster salinity generated from distance bounds of 40, 30 and 20 km.16 As Table 2 shows, reducing the bound has a very large impact on degrees of freedom: Available observations fall from 8,183 at 40 km to 4,847 at 30 km and 2,072 at 20 km. These sharp reductions have consequences for the statistical significance of all estimated parameters, as would be expected. In the 40 km estimate, salinity in the month before birth is one of the five variables that have statistical significance of 95% or higher (the others are male household head, mother’s age, wealth17 and spring water source). In the 30 km estimate, salinity is one of three variables with significance of 95% or higher. In the 20 km estimate, none of the variables has 95% significance. In the transition from 40 km to 20 km, the parameter estimates for salinity remain stable at .02 or slightly higher. 16 These three distance bounds are sufficient to establish the clear relationship linking geographic proximity, sample size and estimation robustness. Our impact estimates for salinity are effectively identical for bounds of 40, 30 and 20 km. However, as Table 2 shows, estimated standard errors rise (and t-statistics fall) as degrees of freedom fall sharply with progressive reduction of the distance bound. 17 The DHS surveys include self-reported wealth in five twenty-percentile classes. Prior experimentation revealed significant effects for the top two classes separately, but no significant difference in magnitudes. We have therefore combined the two classes into “Top 40%” for the final estimates reported in this paper. 11 Table 2: Probit estimation results (controlling for birth month) Dependent variable: Child death before age 2 months (1) (2) (3) (4) Full Cluster Distance Bound (km) Sample 40 30 20 Observations 39,150 8,183 4,847 2,072 Pre-natal salinity 0.020 0.023 0.024 (month before birth) (2.21)* (2.07)* (1.52) Female child -0.025 -0.024 -0.062 -0.083 (1.09) (0.43) (0.86) (0.74) Male household head 0.123 0.400 0.462 0.484 (2.71)** (3.10)** (2.58)** (1.71) Age of mother -0.012 -0.014 -0.011 -0.005 (6.10)** (2.97)** (1.79) (0.54) Mother education (years) -0.019 -0.007 -0.010 -0.016 (5.63)** (1.05) (1.02) (0.97) Wealth - top 40% -0.001 -0.183 -0.188 -0.206 (0.05) (2.50)* (1.94) (1.36) Sanitation (no toilet facility excluded) Flush toilet -0.100 -0.226 -0.188 -0.095 (2.14)* (1.95) (1.26) (0.41) Pit latrine -0.061 -0.106 -0.069 -0.012 (2.06)* (1.43) (0.71) (0.08) Water source (tap and piped water excluded) Tubewell 0.069 -0.041 -0.065 -0.084 (1.85) (0.48) (0.58) (0.49) Spring 0.386 0.425 0.482 0.503 (5.01)** (2.69)** (2.38)* (1.62) Surface 0.235 -0.187 -0.168 -0.101 (2.32)* (0.72) (0.52) (0.22) Cooking fuel (dirty fuels excluded) Electricity or gas -0.067 0.196 0.204 0.202 (1.23) (1.66) (1.36) (0.88) Constant -1.459 -1.803 -1.957 -2.158 (17.31)** (8.52)** (6.94)** (4.89)** Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1% 12 Table 3: Probit estimates (40 km): Impact on incidence of infant mortality - salinity exposure by month before birth 1 2 3 4 5 6 7 8 9 Pre-natal salinity 0.020 0.007 0.005 -0.003 0.003 -0.017 -0.005 0.000 -0.002 (2.21)* (0.64) (0.45) (0.27) (0.33) (1.48) (0.58) (0.01) (0.23) Female child -0.024 -0.020 -0.018 -0.021 -0.021 -0.062 -0.038 -0.040 -0.006 (0.43) (0.36) (0.33) (0.37) (0.37) (1.08) (0.67) (0.69) (0.10) Male household head 0.400 0.406 0.406 0.412 0.410 0.408 0.397 0.388 0.497 (3.10)** (3.15)** (3.14)** (3.19)** (3.17)** (3.13)** (3.03)** (2.97)** (3.34)** Mother age -0.014 -0.015 -0.015 -0.015 -0.015 -0.021 -0.020 -0.019 -0.022 (2.97)** (3.04)** (3.05)** (3.09)** (3.05)** (4.07)** (3.85)** (3.73)** (4.12)** Mother education (years) -0.007 -0.007 -0.007 -0.008 -0.008 -0.004 -0.003 -0.004 -0.001 (1.05) (1.05) (1.05) (1.08) (1.10) (0.62) (0.48) (0.60) (0.18) Wealth index - top 40% -0.183 -0.167 -0.167 -0.158 -0.162 -0.159 -0.154 -0.163 -0.137 (2.50)* (2.28)* (2.28)* (2.16)* (2.20)* (2.14)* (2.07)* (2.16)* (1.79) Flush toilet -0.226 -0.227 -0.224 -0.224 -0.217 -0.190 -0.209 -0.196 -0.316 (1.95) (1.96) (1.93) (1.93) (1.86) (1.62) (1.77) (1.66) (2.53)* Pit latrine -0.106 -0.103 -0.100 -0.096 -0.087 -0.110 -0.126 -0.123 -0.173 (1.43) (1.38) (1.33) (1.28) (1.17) (1.45) (1.65) (1.60) (2.22)* Tubewell water -0.041 -0.034 -0.030 -0.025 -0.037 -0.050 -0.060 -0.068 0.015 (0.48) (0.40) (0.35) (0.29) (0.43) (0.58) (0.69) (0.77) (0.16) Spring water 0.425 0.437 0.436 0.455 0.463 0.474 0.482 0.399 0.334 (2.69)** (2.75)** (2.74)** (2.85)** (2.87)** (2.92)** (2.95)** (2.35)* (1.78) Surface water -0.187 -0.151 -0.142 -0.118 -0.152 -0.072 -0.111 -0.140 -0.048 (0.72) (0.58) (0.55) (0.45) (0.59) (0.28) (0.43) (0.54) (0.18) Clean cooking fuel 0.196 0.187 0.185 0.183 0.185 0.089 0.105 0.114 0.106 (1.66) (1.59) (1.57) (1.55) (1.56) (0.71) (0.83) (0.90) (0.79) Constant -1.803 -1.773 -1.769 -1.754 -1.765 -1.541 -1.637 -1.642 -1.715 (8.52)** (8.36)** (8.32)** (8.26)** (8.30)** (7.12)** (7.49)** (7.44)** (7.36)** Observations 8,183 8,108 8,020 7,956 7,881 7,816 7,730 7,634 7,526 Absolute value of t statistics in parentheses * significant at 5%; ** significant at 1% 13 Overall, we opt for the 40 km results because they incorporate many more degrees of freedom, while exhibiting no deviation from the 30 and 20 km results in signs and little deviation in magnitudes. The results for salinity are particularly stable. Assessment of Implications for Public Health in the Coastal Region To illustrate the implications, we use our 40 km results to predict mortality probabilities for the case presented in Table 4: a female child, one month old, of a mother with no formal education living in a poor, male-headed household with no toilet facilities, access to a community water tap and wood fuel for cooking. We predict probabilities with salinity at the 5th and 95th percentile measures for the sample (0.52 and 7.83 dS/m, respectively), and for maternal ages 30 and 17. As the table shows, increasing salinity from the 5th to 95th percentile increases the child’s mortality probability by about 30% (33.3% for maternal age 30 and 29.1% for age 17). In comparison, the 13-year difference in age accounts for a change of about 40% in mortality probability (43.3% for 5th-percentile salinity; 38.8% for 95th- percentile salinity). Table 4: Salinity impact illustration Child Household Gender: Female Head: Male Age: 1 month Wealth class: Bottom 60% Birth month: June Toilet facility: None Water access: Community tap Mother Cooking fuel: Wood Education (years): 0 Mortality Probability (%) Mother Age Salinity Salinity 30 17 Diff. % Ch. Percentile (dS/m) 5 0.52 6.0 8.6 2.6 43.3 95 7.83 8.0 11.1 3.1 38.8 Diff 2.0 2.5 % Ch. 33.3 29.1 14 Experimentation with a wide variety of individual, maternal and household characteristics reveals similarly-large percent increases in infant mortality when salinity is increased from the 5th to the 95th percentile. Overall, our results suggest that the impact of salinity variation on infant health in coastal Bangladesh is comparable to the impact of other variables that have been assigned major importance by previous research (e.g., maternal age and education, gender of household head, wealth, toilet facilities, drinking water sources and cooking fuel). Conclusions In this paper, we have used new monitoring data to investigate the impact of drinking water salinity on infant mortality in coastal Bangladesh. Our work also incorporates many determinants of mortality and morbidity that have been identified by previous research. We use probit and logit analysis to estimate mortality probability models for infants whose age is less than two months. Our database combines spatially-formatted individual and household survey information from the Bangladesh Demographic and Health Surveys for 2004 and 2007 with spatially-formatted salinity measures for 2001-2009 provided by the Bangladesh Soil Research Development Institute. Our approach is based on extensive medical research that has established significant links between the salinity of drinking water consumed during pregnancy, hypertension and preeclampsia in pregnant women, and postpartum infant morbidity and mortality. We focus particularly on testing the post-natal impact of pre-natal salinity exposure during each month of pregnancy. Controlling for many other determinants, our probability model estimation for infant mortality finds high significance for salinity exposure during the last month of pregnancy and no significance for exposure during the preceding months. The implied impact is comparable in magnitude to the estimated 15 effects of traditionally-cited variables such as maternal age and education, gender of the household head, household wealth, toilet facilities, drinking water sources and cooking fuels. To summarize, our results strongly suggest that drinking water salinity is a significant determinant of infant mortality in coastal Bangladesh. They also provide some new insights about the relationship between post-natal impacts and the timing of pre-natal ingestion of saline drinking water. However, we should add a cautionary note: Although our research has benefitted from unprecedented access to spatially-formatted salinity monitoring information, our estimation exercise is not based on direct, household-specific measures of drinking water salinity. We have constructed salinity estimates for each DHS cluster using spatial interpolation of information from salinity monitors that lie within 40 km of the cluster. In addition, our monitoring information is for soil salinity, not the salinity of drinking water sources. We provide a rationale for our approach in the paper, but we readily acknowledge the possibility that the resulting measures proxy actual drinking water salinity with substantial random error. In closing, we believe that our results are sufficiently strong to encourage further empirical work on this issue. An ideal longitudinal study would track health outcomes for women and their children over an extended period, while measuring the salinity of their drinking water at frequent intervals. The study would also track other determinants of maternal and child health outcomes (e.g., mother’s education, household wealth, access to sanitation, drinking water sources and cooking fuels). Such a study might well seem infeasible as a stand-alone exercise, but we believe that a relatively low-cost expedient could contribute valuable information. Once sampling clusters are identified for future Bangladesh Demographic and Health Surveys, an independently-financed team could undertake monthly sampling of drinking water salinity in those clusters, both before and during the survey period. This would provide sufficient information to support more robust and detailed short-run analyses of health impacts and, if repeated in successive surveys, would ensure the growth of a spatially-formatted database for 16 long-term analysis. In light of the apparent strength of our results, we believe that the potential value of such an exercise would justify the cost of fielding the requisite team. 17 References Aerts L.J., A. Hassan, H. Savenije and M. Khan. 2000. Using GIS tools and rapid assessment techniques for determining salt intrusion: STREAM - a river basin management instrument. Physics Chem Earth B Hydrol Oceans Atmos, 25: 265–273. Alderman, M.H. 2000. Salt, blood pressure, and human health. Hypertension, 36:890–893. Amouzou, A. and K. Hill. 2004. Child mortality and socio economic status in Sub-Saharan Africa. African Population Studies, 19(1): 1-11. Bhuiyan, Md. J.A.N and Dushmanta Dutta. 2011. Assessing impacts of sea level rise on river salinity in the Gorai river network, Bangladesh. Estuarine, Coastal and Shelf Science, 96(1): 219-227. Bicego, G. and O. B. Ahmad. 1996. Infant and Child Mortality. Demographic and Health Surveys, Comparative Studies No. 20. Calverton, MD: Macro International. Brecht, Henrike, Susmita Dasgupta, Benoit Laplante, Siobhan Murray and David Wheeler. 2012. Sea-Level Rise and Storm Surges: High Stakes for a Small Number of Developing Countries. The Journal of Environment & Development, 21: 120-138. Calabrese, E.J. and R.W. Tuthill. 1981. The influence of elevated levels of sodium in drinking water on elementary and high school students in Massachusetts. Science of the Total Environment, 18:117–133. Chowdhury, Abdul H. 2013. Determinants of Under-Five Mortality in Bangladesh. Open Journal of Statistics, 2013(3): 213-219 CIESIN. 2010. Low-Elevation Coastal Zone (LECZ) Rural-Urban Estimates. http://sedac.ciesin.columbia.edu/gpw/lecz.jsp Dasgupta, Susmita, Farhana Akhter Kamal, Zahirul Huque Khan, Sharifuzzaman Choudhury and Ainun Nishat. 2014a. River Salinity and Climate Change: Evidence from Coastal Bangladesh. Policy Research Working Paper No. 6817, Development Research Group, World Bank. March Dasgupta, Susmita, Md. Moqbul Hossain, Mainul Huq and David Wheeler. 2014b. Climate Change, Soil Salinity and the Economics of High-Yield Rice Production in Coastal Bangladesh. Policy Research Working Paper (forthcoming), Development Research Group, World Bank. Dasgupta, Susmita, Benoit Laplante, Craig Meisner, David Wheeler and Jianping Yan. 2009. The impact of sea level rise on developing countries: a comparative analysis. Climatic Change, 93(3-4): 379- 388. Dasgupta S., M. Huq, M. Khaliquzzaman, K. Pandey and D. Wheeler. 2006. Indoor air quality for poor families: new evidence from Bangladesh. Indoor Air, 16(6):426-44. 18 Das Gupta, M. 1990. Death Clustering, Mothers’ Education and the Determinants of Child Mortality in Rural Punjab, India. Population Studies, 44(3): 489-505. Hallenbeck, W.H., G.R. Brenniman and R.J. Anderson. 1981. High sodium in drinking water and its effect on blood pressure. American Journal of Epidemiology, 114:817–826. Hansen, J.E., and M. Sato. 2011. Paleoclimate Implications for Human-made Climate Change. NASA Goddard Institute for Space Studies and Columbia University Earth Institute. Haque, S.A. 2006. Salinity Problems and Crop Production in Coastal Regions of Bangladesh. Pakistan Journal of Botany, 38(5): 1359-1365. Hassan, A. and Md. A.R. Shah. 2006. Impact of Sea Level Rise on Suitability of Agriculture and Fisheries. Center for Environmental and Geographic Information Services, Dhaka, Bangladesh. He, F.J. and G.A. MacGregor. 2007. Salt, blood pressure and cardiovascular disease. Current Opinion in Cardiology, 22:298–305. Hobcraft, J.N., J. W. McDonald and S. O. Rutstein. 1984. Socio-Economic Factors in Infant and Child Mortality: A Cross-National Comparison. Population Studies, 38 (2): 193-223. Hobcraft, J.N., J. W. McDonald and S. O. Rutstein. 1985. Demographic Determinants of Infant and Early Child Mortality: A Comparative Analysis. Population Studies, 39(3): 363-385. IWM. 2003. Sundarban Biodiversity Conservation Project: Surface Water Modeling, Final Report. Institute of Water Modeling, Ministry of Environment and Forests, Government of Bangladesh. Jain, A.K. 1985. Determinants of Regional Variations in Infant Mortality in Rural India. Population Studies, 39(3): 407-424. Karim, Z., S.M. Saheed, A.B.M. Salauddin, M.K. Alam and A. Huq. 1982. Coastal saline soils and their management in Bangladesh. Soils publication No. 8, Bangladesh Agricultural Research Council. Karim, Z., S.G. Hussain and M. Ahmed. 1990. Salinity problems and crop intensification in the coastal regions of Bangladesh. Soils publication No. 33. Bangladesh Agricultural Research Council. Khan, Aneire, Santosh Kumar Mojumder, Sari Kovats and Paolo Vineis. 2008. Saline contamination of drinking water in Bangladesh. The Lancet, 371(9610):385. Khan, Aneire, Andrew Ireson, Sari Kovats, Sontosh Kumar Mojumder, Amirul Khusru, Atiq Rahman and Paolo Vineis. 2011. Drinking Water Salinity and Maternal Health in Coastal Bangladesh: Implications of Climate Change. Environmental Health Perspectives, 119(9): 1328-1332. 19 Mahmood, S. M. Shah, Farhana Najneen, Kazi Samiul Hoque, Saidur Rahman and Mahamud Shamim. 2010. Climate Change: a study on impact and people’s perception (a case study on Mongla Upazila , Bagerhat District , Bangladesh). Bangladesh Research Publications Journal, 4(2): 153-164. Majumder, A.K., M. May and P. D. Pant. 1997. Infant and Child Mortality Determinants in Bangladesh: Are They Changing? Journal of Biosocial Science, 29(4): 385-399. Martin, L.G., J. Trussell, F. R. Salvail and N. M. Shah. 1983. Co-Variates of Child Mortality in the Philippines, Indonesia, and Pakistan: An Analysis Based on Hazard Models. Population Studies, 37(3): 417-432. Midgley, J.P., A.G. Matthew, C.M. Greenwood and A.G. Logan. 1996. Effect of reduced dietary sodium on blood pressure: a meta-analysis of randomized controlled trials. Journal of the American Medical Association, 275:1590–1597. Mohal, N., Z.H. Khan, and N. Rahman. 2006. “Impact of Sea level Rise on Coastal Rivers of Bangladesh.” Institute of Water Modelling, Dhaka, Bangladesh. NIPORT (National Institute of Population Research and Training), Mitra and Associates, and ORC Macro. 2005. Bangladesh Demographic and Health Survey 2004. Dhaka, Bangladesh and Calverton, Maryland [USA]: National Institute of Population Research and Training, Mitra and Associates, and ORC Macro. NIPORT (National Institute of Population Research and Training), Mitra and Associates, and Macro International. 2009. Bangladesh Demographic and Health Survey 2007. Dhaka, Bangladesh and Calverton, Maryland, USA:National Institute of Population Research and Training, Mitra and Associates, and Macro International. Nobi, N. and A. Das Gupta. 1997. Simulation of Regional Flow and Salinity Intrusion in an Integrated Stream-Aquifer System in Coastal Region: Southwest Region of Bangladesh. Ground Water, 35(5): 786–796. Peterson, L. and S. Shireen. 2001. Soil and water salinity in the coastal area of Bangladesh. Bangladesh Soil Resource Development Institute. Pfeffer, W. T., J. T. Harper and S. O’Neel. 2008. Kinematic constraints on glacier contributions to 21st- century sea-level rise. Science, 321: 1340-1343. Rahman, M.M. and M. Ahsan. 2001. Salinity constraints and agricultural productivity in coastal saline area of Bangladesh. Soil Resources in Bangladesh: Assessment and Utilization. Rahmstorf, Stefan. 2007. A semi-empirical approach to projecting future sea-level rise. Science, 315: 368-370. Sibai, B.M. 2002. Chronic hypertension in pregnancy. Obstetrics and Gynecology, 100:369–377. 20 SRDI. 1998a. Soil salinity map of Bangladesh (1973). Soil Resources Development Institute, Dhaka. SRDI. 1998b. Soil salinity map of Bangladesh (1997). Soil Resources Development Institute, Dhaka. SRDI. 2000. Soil salinity of Bangladesh 2000. Soil Resources Development Institute, Dhaka. SRDI. 2010. Saline Soils of Bangladesh. Soil Resources Development Institute, Dhaka. UK DEFR (U.K. Department of Environment, Food and Rural Affairs). 2007. Investigating the Impact of Relative Sea-Level Rise on Coastal Communities and Their Livelihoods in Bangladesh. Dhaka:Institute of Water Modelling in Bangladesh and Center for Environment and Geographic Information Services in Bangladesh. Vermeer, M. and S. Rahmstorf. 2009. Global Sea Level Linked to Global Temperature. Proceedings of the National Academy of Sciences, 106(51): 21527-32. Welty, T.K., L. Freni-Titulaer, M.M. Zack, P. Weber, J. Sippel, N. Huete N, et al. 1986. Effects of exposure to salty drinking water in an Arizona community. Cardiovascular mortality, hypertension prevalence, and relationships between blood pressure and sodium intake. Journal of the American Medical Association, 255:622–626. Wheeler, David. 2011. Quantifying Vulnerability to Climate Change: Implications for Adaptation Assistance. Center for Global Development Working Paper No. 240. World Bank. 1993. World Development Report 1993: Investing in Health. New York: Oxford University Press. World Bank. 2000. Bangladesh: Climate Change and Sustainable Development. Report No. 21104-BD. Rural Development Unit, South Asia Region. December. Yu, Winston. 2010. Implications of climate change for fresh groundwater resources in coastal aquifers in Bangladesh. Washington, DC: World Bank. 21 Table A1: Drinking water sources (Bangladesh DHS 2004, 2007) Water Sources DHS Description Freq. % Dummy variables included Tubewell Tube well or borehole 33,369 85.28 Spring Protected spring 441 1.13 Unprotected spring 270 0.69 Surface River/dam/lake/pond/stream/canal/irrigation 444 1.13 Dummy variables excluded (for comparison) Piped Piped into dwelling 1,548 3.96 Piped to yard/plot 930 2.38 Tap Public tap/standpipe 150 0.38 Excluded (Insufficient data) Well Protected well 21 0.05 Unprotected well 126 0.32 Other Rainwater 3 0.01 Cart with small tank 6 0.02 Bottled water 3 0.01 Other 18 0.05 Not a de jure resident 1,794 4.59 Total 39,123 100 22 Table A2: Toilet facilities (Bangladesh DHS 2004, 2007) Facility Type DHS Description Freq. % Dummy variables included Flush toilet Flush to piped sewer system 2,496 6.38 Flush to septic tank 2,646 6.76 Flush to pit latrine 870 2.22 Flush to somewhere else 690 1.76 Flush, don't know where 36 0.09 Pit latrine Ventilated improved pit latrine 3,192 8.16 Pit latrine with slab 9,846 25.16 Pit latrine without slab/open pit 11,466 29.31 Pit latrine, other 465 1.19 Dummy variables excluded (for comparison) No toilet facility No facility/bush/field 3,453 8.83 Excluded (ambiguous definition or insufficient data) Bucket toilet 12 0.03 Hanging toilet/latrine 2,160 5.52 Not a de jure resident 1,794 4.59 Total 39,126 100 23 Table A3: Cooking fuels (Bangladesh DHS 2004, 2007) Cooking Fuel Type DHS Description Freq. % Dummy variables included Electricity or gas Electricity 24 0.06 Lpg 1,761 4.5 Natural gas 1,449 3.7 Biogas 102 0.26 Dummy variables excluded (for comparison) Charcoal Charcoal 7,452 19.04 Biomass Wood 9,780 24.98 Straw/shrubs/grass 8,559 21.87 Agricultural crop 5,037 12.87 Dung Animal dung 1,248 3.19 Excluded (insufficient data) Kerosene Kerosene 9 0.02 Coal Coal, lignite 9 0.02 Other 123 0.31 Not a de jure resident 3,591 9.17 Total 39,144 100 24