Abstract
INTERNATIONAL JOURNAL OF URBAN SUSTAINABLE DEVELOPMENT 2021, VOL. 13, NO. 2, 233–247 https://doi.org/10.1080/19463138.2020.1868475 ARTICLE Determinants of the severity of household food insecurity among the slums of Dhaka city, Bangladesh Poushali Bhattacharjee and Maria Sassi Department of Economics and Management, University of Pavia, Pavia, Italy ABSTRACT ARTICLE HISTORY Received 21 January 2020 According to the National Food Policy of Bangladesh, the urban slum dwellers are the Accepted 13 December 2020 most vulnerable group to ensure food security. Their condition has not been materi- ally improved over three decades. In our study, we present the current food insecurity KEYWORDS scenario in the slums of Dhaka city using the most recent ‘Bangladesh – Urban Food insecurity; food calorie Informal Settlements Baseline Survey’ dataset of the World Bank. Afterwards, we gap; urban slums; IHS analyse the determinants of the household food calorie gap by applying the transformed DH model; Inverse Hyperbolic Sine transformed Double Hurdle model. The determinants are Bangladesh organised to represent the three pillars of food security (food availability, access, and utilisation) and all three pillars have emerged as significant factors in determining the food calorie gap. From our empirical results, we highlight some vital gaps in the National Food Policy and recommend broad areas of interventions for the betterment of the food security status among the slum dwellers of Dhaka. 1. Introduction security (Ministry of Food and Disaster Management 2006). By 2030, a determined goal has been set by According to the World Bank, mega-urbanisation Bangladesh to reach the United Nations Sustainable has become the most dominant demographic process Development Goal number two by ending hunger in Bangladesh which has resulted in an increase in the and malnutrition and ensuring access to substantial urban population from 5% to 36% between 1950 and safe and nutritious food for all (General Economic 2017 (The World Bank 2019). In the capital city of the Division (GED), Bangladesh Planning Commission, G. country, Dhaka, the mega-urbanisation process often 2018). Since independence, the country has come a cannot be controlled spatially, demographically, or long way to achieve substantial food self-sufficiency structurally, as it has limited infrastructural and public through agricultural innovation, nutrition pro- service provision capacity. On top of that, the city is grammes, and food production (USAID, 2019). being crowded with the influx of poor migrants from Despite this effort, in 2019 almost 25% of the popula- rural areas due to unemployment, river erosion, and tion in Bangladesh remained food insecure and 36% limited access to quality education and health of children younger than 5 years of age suffered from (Agyeman, 2013; Ahmed and Meenar 2018). This mar- stunting, a common measure of chronic malnutrition ginalised section of the city generally fail to enter into (USAID, 2019). According to the National Food Policy the formal job market and end up being the low-paid (NFP) of Bangladesh, the undernutrition rates are labour. They do not generate enough income to pay higher among the urban slum and rural landless for formal housing and create slums or informal set- households. A balanced diet is still beyond the reach tlements as their shelter (Mohit 2012; Ahmed and of this section of the country. Therefore, slum and Meenar 2018). rural landless households are identified as the most The Bangladesh Bureau of Statistics Bangladesh distressed population in terms of achieving food Bureau of Statistics, Ministry of Populations (2015) CONTACT Poushali Bhattacharjee, poushalibuet@gmail.com 109 Merrimac St, Buffalo, NY 14214, USA © 2021 Informa UK Limited, trading as Taylor & Francis Group 234 P. BHATTACHARJEE AND M. SASSI defines a slum as a compact settlement of five or policy because of their divergent physical and social more households characterised by very poor hous- environments (Raju et al. 2017). ing; high population density and room crowding; Our study put effort to assist in designing effective poor environmental services (water and sanitation); policies and programs for improving the food security low socioeconomic status and lack of security of status of slum households. More precisely, it analysed tenure. The most recent slum census of 2014 the determinants of the status and depth of food recorded 2.23 million slum populations covering insecurity among the urban slum households of the 6.33% of the country’s urban population, living Dhaka city through implying the Inverse Hyperbolic across 13,935 slums in Bangladesh, whereas Dhaka Sine (IHS) transformed Double Hurdle (DH), model. It alone is bearing the pressure of 1.06 million slum used 580 sample slum households for the analysis dwellers (Bangladesh Bureau of Statistics, Ministry of from the most recent detailed slum Household Populations 2015). The mean growth of the children Income and Expenditure Survey conducted in 2016 residing in these slum is generally poorer than both by the World Bank ([dataset] Yanez-Pagans, M. Y non-slum and rural children (Raju et al. 2017). As 2017). Furthermore, to compute the Kcal/Adult stated by Bangladesh Urban Slum Survey 2013 by Equationuivalent/Day for each household, we referred WFP, the rates of stunting and wasting of children to the latest food consumption table and adult equa- under five years in slums were 44% and 16%, respec- tionuivalent values for Bangladesh by using the study tively, which was higher than the rate found in all of Shaheen et al. (2013) and Waid et al. (2017) respec- non-slum urban areas of the country (World Food tively. Therefore, our study presents an updated sce- Programme 2015). nario of the food insecurity condition of the slum It was observed comparing the World Food households of the Dhaka city. Programme (WFP) survey results of 2006 and 2013 The study makes a three-fold contribution to the that more than 50% of slum households were food literature. First of all, it has tapped into a partly uncov- insecure based on their calorie consumption (IFPRI ered issue, the determinant analysis of the depth of 2007; World Food Programme 2015). Moreover, by food insecurity, by using an impact variable that using the most recent slum household food consump- accounts for the severity of food insecurity. These tion dataset of the Dhaka city collected by the World types of studies are very scarce and geographically Bank in 2016, our study also identified 51.73% of limited. In Bangladesh, except for two extensive stu- households as food insecure due to not consuming dies conducted by the WFP, there has been lacking the minimum energy requationuirement of 2,430 prominent and detailed empirical research work done Kcal/Adult Equationuivalent/Day. All these shreds of on the depth of food insecurity among the slum evidence proved that the food security status of slum households especially in the capital city Dhaka. In households in Bangladesh has not materially the first one, 1,900 slums households in Dhaka, improved over time. Chittagong, Khulna, and Rajshahi were surveyed in Even with this severe situation, there is no mention 2006 to analyse the determinants of food security of specific measures adopted in the Plan of Action status using calorie consumption sufficiency, Diet (PoA) of the NFP for the urban poor living in the Diversity Score (DDS), and Household Food slums accommodating their unique features and chal- Insecurity Access Scale as impact variables (IFPRI lenges. The policies and programs targeting to 2007). The second study conducted in 2013 only col- improve the food security status of the people are lected detail gender-disaggregated data on various mostly concentrated in rural areas (Ministry of Food aspects related to food consumption with no further and Disaster Management 2006, 2008). In the food insecurity determinants analysis (World Food Government’s Poverty Reduction Strategy, NFP and Programme 2015). PoA, there is an emphasis on intensifying agricultural Moreover, most of the other existing studies on research on crop diversification; improving rural peo- the slum food insecurity determinants, mainly used ple access to market and asset; and expanding the discrete choice logit model to assess the determi- coverage of social safety net programs mostly in rural nants of whether a household is food secure or not areas (Food Security Portal facilated by IFPRI 2012). (Dharmaraju et al. 2018; Joshi et al. 2019; Chinnakali The slum nutrition policy should be treated as distinct et al. 2014; Agarwal et al. 2016; Gebre 2012; Akinloye from the study of urban and rural food and nutrition et al. 2016). However, these studies could not FOOD ADDITIVES & CONTAMINANTS: PART B 235 capture the variability in the depth of the food cal- section VI by linking with policy implications and orie gap among the food insecure households. In our finally, Section VII concludes the study. study, we computed the Foster-Greer-Thorbecke (FGT) index where the depth of food security 2. Dataset revealed that each food-insecure household needs 448.482 Kcal/Adult Equationuivalent/Day to bring This study uses the ‘Bangladesh – Urban Informal them in the food-secure state. The squared food Settlements Baseline Survey (BUISBS) 2016’, a second- insecurity gap expressed that the severity of food ary dataset collected by the World Bank ([dataset] insecurity is 5.12% among the urban slum house- Yanez-Pagans, M. Y 2017). The target population was holds. Therefore, our study focused on analysing the identified slum households in the Dhaka City not only the determinants of food insecurity status Corporation (DCC) area in the BBS slum census 2014. but also the factors which contributed to the varia- A total of 580 households were used in this study for tion in the severity of food insecurity among the all the analysis by excluding eight outliers from the households. original dataset based on the value of the dependent We also improved the literature using the Limited variable. The multipurpose survey dataset recorded Dependent Variable (LDV) model in the field of food household food consumption data of 14 days at a insecurity depth analysis. In the context of urban stretch based on a small recall period of two-days. slums households, the use of the LDV model is almost The detailed food consumption list in the question- negligible. Some studies used the LDV models like naire contained 148 food items with their respective Tobit and Heckman model to analyse the depth of quantity (in gram, millimetre, number, and cup), the food insecurity (Kamau et al. 2011; Sani and Kemaw monetary value in Bangladesh Taka (BDT), and the 2019). All these two-stage LDV models assume the source. However, it did not capture the seasonal var- bivariate normal distribution of the error terms. If iation of food consumption which created an obstacle this assumption is violated, the estimation will be in considering the food stability pillar in the analysis. inconsistent. This possibility is not tested in the The dataset also did not record the intra-household above-mentioned studies. Yen and Jones (1997) intro- food consumption and so the actual calorie consump- duced the IHS transformation of the dependent vari- tion of each household member has a certain margin able in the DH model to relax the restrictive of error. Nonetheless, while calculating the calorie assumption. This study has applied IHS transformation consumption, we followed specific guidelines to mea- in the DH model to understand the factors determin- sure calorie consumption/Adult Equationuivalent as ing the depth of food insecurity. precisely as possible (discussed in section 3). Lastly, the study contributed to identifying some gaps in the NFP of Bangladesh and its PoA regarding 3. Variable selection the adoption of effective strategies to improve the food security status of slum households. It also tried Table 1 presents the descriptive statistics of all the to suggest some possible solutions exclusively for the selected variables which we used in our study for the slum settings which can assist to design specific poli- determinant analysis. In the household level food cies and special safety net programs with better tar- security analysis, calorie intake (Kcal/Adult geting of the distressed urban poor to improve their Equationuivalent/Day) has been considered as ‘gold food security status. standard’ for food security measurement (Hoddinott The remainder of this study is organised as follows: and Yohannes 2002). As a result, we used the food Section II introduces the dataset with the sampling calorie gap of each household with respect to the technique and sample size. The methodology used in minimum energy requationuirement set by FAO selecting and constructing the variables for food inse- (2010) for Bangladesh (2430 Kcal/Adult curity determinant analysis is described in section III. Equationuivalent/Day) as our dependent variable. Section IV elaborates on the empirical framework The variable is continuous ranging from −1816.502 used for analysing the determinants of food insecurity to 1564.628 Kcal/Adult Equationuivalent/Day with a status and depth. The results of the regression model mean of 319.57 Kcal/Adult Equationuivalent/Day, are mainly presented in section V. Further discussion where positive values denote food-insecure house- on the outcome of empirical results is conducted in holds. The construction of our dependent variable 236 P. BHATTACHARJEE AND M. SASSI Table 1. Descriptive statistics of the independent variables. Food secure Food insecure Variables Total households households households Food Access Factors DD/household/Day 7.33 7.68 7.01 Monthly expenditure (ME)/Capita (in BDT) 2877.213 ± 1414.45 3414.515 ± 1514.12 2375.731 ± 1101.07 Monthly education cost (in BDT) 589.032 492.4869 679.1417 Number of unemployed household member within 15 to 0.8172 0.7214 0. .9066 64 years Occupation of household head 36(6.21%) 14 (5.00%) 22 (7.33%) Not working 223 (38.45%) 104 (37.14%) 119 (39.67%) Own business and service 321 (55.34%) 162 (57.86%) 159 (53.00%) Engaged in Employed work Food Utilisation Factors Number of adult female member 1.251 1.25 1.253 Education level of household head 276 (47.59%) 121 (43.21%) 155 (51.67%) Never attended school 304 (52.41%) 159 (56.79%) 145 (48.33%) Attended school Drinking water source 547 (94.31%) 272 (97.14%) 275 (91.67%) Pipe water 33 (5.69%) 8 (2.86%) 25 (8.33%) Other source except pipe Food Availability Factors Wealth Index −0.0066884 0.0926 −0.099417 Previous Place of living 253 (43.62%) 123 (43.93%) 130 (43.33%) Anywhere except slum (Base level) 327 (56.38%) 157 (56.07%) 170 (56.67%) Slum Security of household 349 (60.17%) 183 (65.36%) 166 (55.33%) Sufficiently Secure 196 (33.79%) 83 (29.64%) 113 (37.67%) Moderately secure 35 (6.03%) 14 (5.00%) Insecure (1) (7.00%) Household Factors Sex of household head 497 (85.69%) 240 (85.71%) 257 (85.67%) Male 83 (14.31%) 40 (14.29%) 43 (14.33%) Female household size 4.296 3.94 4.62 Age of household head 68 (26.38%) 85 (30.36%) 68 (22.67%) 18–30 years 101 (33.79%) 95 (33.93%) 101 (33.67%) 31–40 years 72 (22.41%) 58 (20.71%) 72 (24.00%) 41–50 years 59 (17.41%) 42 (15.00%) 59 (19.67%) Above 50 years Source: BUISBS, 2016. followed several steps. The calorie intake (Kcal/Adult Maximum food items consumed by the slum Equationuivalent/Day) of the households was com- households were recorded into gram. Some food puted by following the guidelines of the World Bank items were in millilitre (ml). We converted them into and FAO (Moltedo et al. 2014). More precisely, the gram by following the formula: calorie intake (in Kcal) of households from each food FQ ¼ FQ � D (2) gi mi i item (E was calculated by the general formula: Ki) where FQ and FQ are the quantity of ith food item gi mi E ¼ ðEQ � C Þ=100 (1) Ki gi Ki in gram and ml, respectively, and D is the density of that food item in gram/ml. After converting all the where EQUATION is the edible quantity in gram and gi food quantities into grams, the edible quantity of C is the calorie per 100 gm edible portion in Kcal of Ki each food item consumed was found by following the ith food item taken from the latest food consump- the formula: tion table for Bangladesh, where the calorie of each food item is expressed in terms of per 100 g edible EQ ¼ FQ � EC (3) gi gi i portion (Shaheen et al. 2013). FOOD ADDITIVES & CONTAMINANTS: PART B 237 where EC is the edible coefficient of the ith food item. access to safe water, sanitation and health services; This value was finally used in equation (1). and weak coverage of government safety net pro- In the food item list, there was the option of grams (Maxwell et al. 1998; Ruel et al. 1998; World ‘others’ under each food group which can create an Food Programme 2002). These attributes were also error in the precise calorie consumption. We adopted taken into consideration while selecting the indepen- special treatment for only 5.79% ‘others’ responses to dent variables. minimise the error. For ensuring the lowest dispersion In the second step of the variable selection, we in calorie calculation, we replaced each ‘others’ from a performed statistical tests on the shortlisted indepen- food group, by the mean calorie acquired by the dent variables to check for the significant association household from that particular food group (Moltedo with the dependent variable and multicollinearity et al. 2014). among themselves. Each considered independent The adult equationuivalent for each household variable was regressed against the dependent vari- was calculated based on age, the gender of its mem- able on an individual basis and if anyone of them ber, mean body weight for a particular age in had a p-value higher than 0.25, it was considered for Bangladesh and a moderate level of physical activity the exclusion (Chinnakali et al. 2014). Before removing by using the study of Waid et al. (2017). Finally, the any categorical variable, we performed the ANOVA mean calorie intake (Kcal) of 14 days was divided by test to observe the significance in the association the total adult equationuivalent for each household to between dependent and independent variables in all get the Kcal/Adult Equationuivalent/Day. possible combinations. We verified the possible exis- To construct our dependent variable food calorie tence of multicollinearity among continuous and gap, the obtained Kcal/Adult Equationuivalent/Day of categorical variables. For the continuous variables, each household was subtracted from the minimum we excluded those having correlation values more energy requationuirement threshold 2,430 Kcal/Adult than 0.5 in the correlation matrix and also as a compli- Equationuivalent/Day. Among the respondent house- ment, crosschecked the exclusion list with variables holds, 51.73% of them consumed less than 2,430 Kcal/ having the Variance Inflation Factor (VIF) value more Adult Equationuivalent/Day and were identified as than 10 (Yoo et al. 2014). In order to select the cate- food insecure households with positive food cal- gorical variables, we performed the chi-square test orie gap. among the related group of variables with a 95% confidence level to observe their association with each other. In total, 65 variables were considered in 3.1 Selection strategy of the independent the first stage of the variable selection which even- variable tually came down to 14 variables to incorporate in the We organised the independent variable selection final model. strategy into two parts. At first, the determinants of food insecurity were selected in combination with 3.2 Food access factors empirical literature reviews, following the conceptual framework prepared for urban poor household food We selected five variables to represent the food security determinants analysis in Bangladesh (IFPRI access pillar of the food security which were Dietary 2007) and the framework linking fundamental pillars Diversity Score (DDS)/household/Day, monthly of food and nutrition security at the household and expenditure/capita, monthly education cost, number individual level by Sassi (2017). The variables were of the unemployed household member within 15 to organised representing the three pillars of food secur- 64 years, and occupation of the household head. The ity introduced by FAO (2008): food availability, food DDS/household/Day reflects the household’s access access, and food utilisation. Urban slum households to a variety of food groups as a proxy of their nutri- hold some distinctive characteristics such as greater tional adequationuacy. We formulated the DDS score dependency on cash income and less reliance on for each household following the guidelines of FAO agriculture and natural resources; low wages from (2010) based on selected 12 food groups. The mostly informal jobs; a large number of women work- monthly expenditure/capita (in BDT) was included as ing outside the home; legal obstacles including inse- an independent variable because expenditure is a cure land and housing tenure; inadequationuate better representation of purchasing power; and less 238 P. BHATTACHARJEE AND M. SASSI vulnerable to under-reporting and recall error (Meyer dummy variable which denoted ‘0ʹ for the households and Sullivan 2003). To capture the effect of the non- using a piped water into dwelling, yard, and com- food spending on food calorie intake, we took one of pound and ‘1ʹ for using other sources like a public the most important expenditures which is monthly tap, borehole, and tube well. education cost (in BDT) as a continuous variable. The number of earning members from the working-age 3.4 Food availability factors group in each household is included in the model as a discrete variable to represent the economic access to The availability of food for the households in terms of food. The occupation of household head was a cate- quantity and quality was expressed by the Wealth gorical variable where ‘0ʹ was assigned for the house- Index (WI), the previous living place of the household hold heads who were not working, ‘1ʹ for the head, and overall security of the household. The household heads engaged in own business and ser- Wealth Index is considered a reliable measure of living vices, and ‘2ʹ for all types of employed work. For the standard and long-term economic condition, espe- slum dwellers, due to fundamental differences, the cially for households from the low-income group self-owned business and service sector is generally (Moser and Felton 2007). In this study, we used the considered as a distinctive category than other Polychoric Principal Component Analysis (PCA) to engaged occupation (Hasam et al. 2017). Therefore, compute the WI, as it allows ordinal categorical vari- apart from differentiating the working and non-work- ables to reduce the loss of information instead of ing household head, we further distinguished using only the binary variable (Kolenikov and between those who were self-employed and working Angeles 2004). The WI was formed with 14 variables, as an employee to understand the difference in their namely: possession of boat, agricultural land, TV, association with food calorie gap. fridge, watch, home, drawing and dining room furni- ture; type of housing, floor and wall material; the number of rooms; and sharing of toilet and water. 3.3 Food utilisation factors The correlation value among these variables was The food utilisation pillar of food security was repre- within the recommended 0.1 to 0.7 range (Hjelm et sented through three variables, namely, the number al. 2017). The factor test was also performed to con- of adult female members present in the household, firm that the retained variables are correlated enough education level of household head, and drinking to run factor analysis. Moreover, the computed Kaiser- water source. The adult female generally manages Meyer-Olkin (KMO) value for the dataset was 0.755 the daily household diet plan and all the needed which is above the recommended minimum value of arrangements to prepare them. Their presence can 0.6 (Hjelm et al. 2017). The KMO value justified the use create awareness and improve the diet quality of of Polychoric PCA to formulate the WI. In the factor households (World Food Programme 2002). analysis process, factors having eigenvalue more than Therefore, the number of adult women present in one were able to explain very high variance (73.13%) each household was considered to express the food in the wealth of the households. utilisation pillar. Most of the household heads never In the slums, only 4.4% of households received attended school. Hence, the education level of house- formal aid or relief and so we could not consider this hold head was divided into two categories where the important variable directly in our regression model. value ‘0ʹ refers to the household head who never The slum households are still dependent on family, attended school and ‘1ʹ for those household heads neighbours, and local shop as a form of informal who attended school at any level of study. The poor insurance to smoothen any consumption shock. The quality of drinking water can result in water-borne households who were born or came from the slum diseases that hamper the proper utilisation of food were supposed to be more accustomed to the slum (FAO 2015). According to UNICEF and WHO, pipe social network and rely more on informal social insur- water is considered a safe and improved drinking ance (Zingel et al. 2011). Therefore, we introduced a water source (Hjelm et al. 2017). By following the dummy variable – ‘previous place of living of the study of IFPRI (2007), we distinguished between household head’ to capture this aspect. It takes the household using an improved and unimproved value ‘0ʹ for those household heads, who came in the source of the drinking water source by introducing a current slum from different districts (not slum) and ‘1ʹ FOOD ADDITIVES & CONTAMINANTS: PART B 239 � � 0 0 for those household heads who were born in the Truncated : EðYjY > 0Þ ¼ βþ λσ β=σ (6) X X currently living slums or came from any other slum. When there is a lack of feeling of security in the where Φ is the standard normal cumulative distribu- slums, there is distrust among neighbours, which lim- tion function (cdf), β is the coefficients of independent its the social reliance and informal safety nets (Zingel variable x and σ is the standard deviation. However, et al. 2011). Based on this consideration, we further this model is restrictive, as the independent variables included a categorical variable presenting the overall x and its coefficients β is bounded to be the same in feeling of security among the households. This vari- both probit and truncated regression model. As an able consists of three categories – sufficiently secure, alternative to this model, in 1971 Cragg proposed a moderately secure, and insecure. Along with the vari- more flexible DH model (Lin and Schmidt 1984; Smith ables representing the three pillars of food security, and Brame 2003). The DH model can be expressed as some household factors like sex and age of the house- � � hold head and household size were also included as Probit : Prob ðy > 0Þ ¼ Φ γ (7) control variables in the final model. � � 0 0 Truncated : Eðyjy > 0Þ ¼ βþ λσ β=σ (8) Z X 4. Empirical strategy As aforementioned, in our sample, 51.73% of slum Unlike equation (5) and (6), the DH model increases households were identified as food insecure. To the flexibility by accommodating a different set of understand the determinants of the severity of food variables (x and z) and coefficients (γ and β) for each insecurity, the focus needs to be put only on the food stage. In our study, it is not guaranteed that the sig- insecure households. In this case, OLS will produce a nificant independent variables and their coefficients biased estimator, as the sample will not be any more would be the same determining both the status and random. As the dependent variable needs to be left- depth of food insecurity among households. censored with the lower limit of zero (0) to consider Therefore, the DH model was preferred over Tobit only the food insecure households, LDV models are due to its flexible nature. We also performed the log- more appropriate instead of OLS (Wooldridge 2003; likelihood test to assure the better suitability of the Kamau et al. 2011; Katchova 2013). In this study, at DH model over the Tobit model (Lin and Schmidt first, we considered the Tobit model, the most com- 1984; Martínez-Espiñeira 2006). monly used LDV model in social science, where the In the first stage of the DH model, a participation dependent variable (Y) became the incompletely equationuation of the ith household in food insecure observed value of the latent dependent variable y* group or not is presented as: (Smith and Brame 2003). The Y variable can be pre- sented as: � y þ β μ (9) 1i ¼ 1 þ 1i 1i y� ; y� > L Y ¼ (4) where y* denotes the latent participation indicator; 1i L; y� � L x is a vector of explained variables of the determi- 1i where L is the lower limit with a value of zero. Tobit nants of household food insecurity; β is a vector of model is a combination of probit and truncated parameters to be evaluated; µ is the error term. In 1i regression model. The probit model analyses the vari- the second stage of DH model, the depth of food able determining the probability of Y being zero or insecurity equationuation of the ith food insecure positive and truncated regression analyses the vari- household is presented as: ables contributing the increase or decrease of the positive continuous values of Y by imposing a lower y þ β μ (10) 2 þ 2i 2i ¼ 2i limit of zero on Y. The probit and truncated model can be presented as: where y* denotes the latent depth indicator; x is a 2i 2i � � vector of explained variables of the determinants of Probit : Prob ðY > 0Þ ¼ Φ x β (5) the depth of food insecurity; β is a vector of para- meters to be evaluated; µ is the error term. Generally, 2i 240 P. BHATTACHARJEE AND M. SASSI the DH model is estimated by a maximum likelihood simple logistic model, the results become more robust estimation (MLE). However, this estimation is based and accurate (Yen and Jones 1997). on the assumption of bivariate normal distribution and non-correlation of the error terms µ and µ 1i 2i 5. Results from equation (9) and (10) (Chen et al. 2018) which can be presented as: Table 2 presents the estimates of the factors affecting � � � � household food insecurity determined by the IHS � � X X μ1i 1 ρσ transformed DH model. The second column of the BVN 0; ; ¼ (11) μ2i ρσ 1 table illustrates the marginal effect of the factors determining the status of food insecurity through where ρσ is the covariates between µ and µ . 1i 2i the probit model, which is the first stage of the DH Once the assumption expressed in equation (11) is model. Whereas the third column shows the marginal violated, the MLE is inconsistent. In our study, these effect of the factors explaining the depth of the food assumptions were tested and the error terms were calorie gap among the food insecure households found to be negatively correlated with violation of using the truncated regression model on the IHS normal distribution. To solve this problem, we per- transformed dependent variable. The associations of formed the IHS transformation of the dependent vari- all the variables with the food calorie gap were able which was introduced by Yen and Jones (1997). reported up to the 90% confidence level. The proposed solution allows flexible parameterisa- The significant factors and their estimated coeffi - tion and accommodates dependence, heteroscedasti- cients were different in the probit and truncated city, and non-normality of error terms. After IHS regression model. From the first tier of probit model, transformation, the dependent variable y can be pre- it was observed that household size, DDS/Day, sented as: Monthly expenditure/capita, Monthly education cost, h i 1=2 T 2 2 1 y ¼ ln fθy þ θ y þ 1 g =θ ¼ sinh ðθyÞ=θ number of unemployed member between age group i i i 15 to 64 years, number of the adult female member, (12) drinking water source and overall security were the where θ is an unknown parameter that can be esti- significant factors to determine whether the urban mated from the data and if the value is near 0, equa- slum households were food insecure or not. In the tion (12) becomes linear in form. Generally, the value second tier of the truncated regression model, along is assumed to have a unit value in most cases with the factors of the probit model, also the occupa- (Chiwaula 2018; Chen et al. 2018). The sample like- tion of the household head and Wealth Index were lihood function of the IHS transformed DH equatio- emerged as significant factors to determine the food nuations formulated from (9) and (10) can be gap among the food insecure households. The direc- expressed as tion of the coefficients of these significant factors was Y � ; ; 1 d as per expectation and similar to the other empirical L ¼ ½1 Ψðx β ; x β =σ ; ρÞ� 1i 1 2i 2 i¼1 h n oi pffiffiffiffiffiffiffiffiffiffiffiffiffiffi � � � �� d 1=2 2 2 T ; ; T ; literature reviews of Dharmaraju et al. (2018), Joshi et � 1þ θ y � ð1=σÞϕ y x β =σ Φ x β þðρ=σÞ y x β = 1 ρ i 2i 2 1i 1 2i 2 (13) al. (2019), Chinnakali et al. (2014), Agarwal et al. (2016), Gebre (2012) and Akinloye et al. (2016). where Ψ is the bivariate standard normal cumulative However, our study focused on one striking fact that density function with a covariance of ρσ, φ is the the occupational status of the household head and univariate standard normal probability density func- the value of the Wealth Index were not a significant tion, Φ is the cumulative density function, respectively, factor to decide the food security status. But if a and d is the dichotomous index, which is equationual household was food insecure, these factors were to 1 if y > 0, and 0 otherwise (Yen and Jones 1997). To responsible to change the severity of the food inse- construct a robust econometric model for food inse- curity depth analysis, we finally used the equation (13) curity status. for our estimation. Though the IHS transformation of Here, * is significant at 90% confidence level, ** the dependent variable makes the model complicated significant at 95% confidence level and *** significant to interpret the coefficients in comparison to the at for 99% confidence level. FOOD ADDITIVES & CONTAMINANTS: PART B 241 Table 2. Marginal effects of the determinants of food insecurity of urban slum HHs. Truncated regression model on HIS transformed Y Probit model (1st variable Variables tier) (2nd tier)) Y variable: Food Calorie Gap/Adult Equationuivalent/Day Food Access Factors DDS/household/Day −0.467*** −1.0089*** Monthly expenditure (ME)/Capita (in BDT) −0.00031** −0.00069** Monthly education cost (in BDT) 0.000217** 0.00045*** Number of unemployed household member within 15 to 0.1875** 0.3712** 64 years Occupation of household head −0.1104 −0.288 Not working (base level) -.3480 –0.8347* Own business and service Engaged in Employed work Food Utilisation Factors Number of adult female member −0.2030* −0.446* Education level of household head −0.027 0.0199 Never attended school (base level) Attended school Drinking water source 0.3704* 0.8352** Pipe water (Base level) Other source except pipe water Food Availability Factors Wealth Index −0.0775 − 0.194* Previous Place of living 0.01232 0.143 Anywhere except slum (Base level) Slum Security of household 0.280* 0. 665** Sufficiently Secure (Base level) 0.029 0.247 Moderately secure Insecure Household Factors Sex of household head −0.126 −0.2169 Male (Base Level) Female household size 0.218*** 0.457*** Age of household head 0.125 0.310 18–30 years (Base level) 0.0505 0.197 31–40 years 0.150 0.311 41–50 years Above 50 Source: BUISBS, 2016. Among all the factors, the DDS/Household/Day Additionally, the households spending more on emerged as one of the most important factors due monthly non-food consumption like education were to its high significance and coefficient value together. more likely to be food insecure with a higher food The higher DDS score significantly decreased the odds calorie gap. Investigating closely we further found, the of being food insecure and the food calorie gap presence of a high number of unemployed members among the food insecure households. Another factor from working age in the household increased the which also turned out to be highly significant was the odds of being food insecure. If there is an increase of per capita monthly expenditure in both the probit one unemployed member, the food gap among food and truncated regression model. To elaborate more insecure households will increase by 37.12%, keeping on this factor, households with higher monthly other variables constant. Among the food insecure expenditures were more likely to be food secure and households, the household where the head was had less food gap if they were food insecure. engaged in any employed work, had a lower food 242 P. BHATTACHARJEE AND M. SASSI calorie gap than the households with unemployed having a higher food gap among the food insecure heads. households than the sufficiently secure households. The presence of a higher number of females aged Among the control variables, the large household 18 years and above in each household increased the had a significant positive effect on both increasing the odds of being food secure and low-calorie gap among odds of being food insecure and the food calorie gap food-insecure households. The empirical result amidst the food insecure households. The large showed that households using improved source households were more likely to be food insecure. pipe water for drinking purposes were more likely to To examine the robustness of the above-men- be food secure than those households not using pipe tioned empirical results, table 3 presents the out- water. Also, the food calorie gap among the food comes of various alternative models like OLS and insecure households increased significantly for those Tobit to check the sensitivity of the marginal effects who were not using piped water for drinking. of the determinants on the food insecurity depth. If With each additional unit of wealth score, the food the results are compared with the base model in the gap among food insecure households was expected third column of table 2, it is observed that the esti- to decrease by 19.4%. The households with moderate mates did not change significantly. In the regression security were more likely to be food insecure and model, the primary sampling unit (PSU) was Table 3. Comparison of alternate models and the base model for robustness check. Variables OLS (marginal effect) Tobit model (marginal effect) Y variable: Food Calorie Gap/Adult Equationuivalent/Day Food Access Factors DDS/Household/Day −234.52*** −189.47*** Monthly expenditure (ME)/Capita (in BDT) −0.13847** −0.144** Monthly education cost (in BDT) 0.0559*** 0.060*** Number of unemployed household member within 15 to 64 years 24.947 36.91 Occupation of household head −172.45 −36.569 Not working (base level) –258.26** –148.61* Own business and service Engaged in Employed work Food Utilisation Factors Number of adult female member −97.442** −70.16* Education level of household head 42.88 29.556 Never attended school (base level) Attended school Drinking water source 75.68 106.37 Pipe water (Base level) Other source except pipe water Food Availability Factors Wealth Index −2.586 −19.81 Previous Place of living 44.758 8.055 Anywhere except slum (Base level) Slum Security of household 72.16 101.44** Sufficiently Secure (Base level) 24.98 88.938 Moderately secure Insecure Household Factors Sex of household head −13.78 −16.5 Male (Base Level) Female household size 103.07*** 78.67*** Age of household head 29.45 88.85* 18–30 years (Base level) 88.77 84.19* 31–40 years –31.796 50.07 41–50 years Above 50 years Source: BUISBS, 2016. FOOD ADDITIVES & CONTAMINANTS: PART B 243 considered as a cluster variable to produce more intergenerational poverty and food insecurity by limit- robust marginal effects. To observe the locational ing their capability to attain standard income genera- effect of the slums on the food insecurity depth, all tion opportunities in the future (Bird 2013; Chilton et the 63 slums were divided into eight large areas al. 2017). This risk is especially high in the analysed according to their ward number in the DCC area. If slums, where almost 33% of the respondents never we control for the location of the slums, still there is attended school, and around 19% completed only the not any remarkable change in the significance, direc- primary level of education. Therefore, the low educa- tion, and value of the factors. tional level deserves specific policy attention. Here, * is significant at 90% confidence level, ** For unemployed people with no formal education, significant at 95% confidence level and *** significant the public sector and NGOs can work hand in hand for at for 99% confidence level. imparting a minimum level of financial literacy, voca- tional and skill-based training emphasising the market demand to secure an earning source for them. These 6. Discussion interventions are coherent with the PoA of the NFP This section has been divided into the three consid- which also emphasises improving access, relevance, ered food security pillars of our analysis to discuss the quality, coordination, efficiency, and sustainability of linkages between the empirical results and its possi- technical and vocational education (Ministry of Food ble implications in policy formulation. and Disaster Management 2008). There are strong linkages among the informal economy of the slums, urban economy, and the 6.1 Food access national economy (Peattie and Aldrete-Haas 1981). If Our results highlighted that the variables represent- the informal economies of the slum dwellers can be ing the food access pillar, more specifically economic brought under a certain extent of formalisation access, were identified as an important area for through participatory slum up-gradation, the produc- further discussion. This result was partly expected as tivity and chain of employments might increase ensuring the economic stability is one of the major (Sheng and Brown 2018). It is undeniable that the keys to improving the health and nutrition of slum government and development partners can play a dwellers (Kumar 2016). According to our dataset, vital role to provide legal and social protection to almost 63% of households had an unemployed mem- the informal economy in such a way that the benefits ber from the working-age (15 to 64 years) in the slums of formalisation exceed its cost for the slum dwellers. of the DCC area. Consequationuently, the mean per In addition, the possibility of designing effective capita monthly expenditure of the slum households microcredit programs for ensuring sustainable liveli- was limited at only 2877 BDT (34.06 USD) which is hoods and stimulating the small business of the slum much lower than the national average (49.32 USD) of dwellers should be evaluated to improve their access that particular year 2016 (CEIC 2018). Moreover, the to credit (Mohapatra 2015). slum households on average spent 72.77% of their To control the negative effect of household educa- monthly expenditure only on food. The traditional tional expenses on food calorie intake, policy priority cereals are sold at a higher price in an urban area should be given to increase the coverage of free than the rural market (Ruel et al. 1998) and the slum schools inside the slums and school feeding pro- households generally spend 30% or even more than gramme in those schools. Previously the intervention the rural households for consuming the same food of free school installation was mostly rural area items (Argenti 2000). From our result, we observed focused and only recently the NGOs have started to that due to dedicating a huge portion of their limited expand such schools inside the slum area (Kabir 2014). monthly expenditure on food, they ended up com- There are only 295 free of cost government primary promising with the necessary non-food expenditure schools in the DCC area which is too small number to like education. The households expending more on serve the city’s large population (Cameron 2010). In monthly education were more likely to be food inse- 2002, the government of Bangladesh and the WFP cure with higher food calorie gap. However, the litera- launched the school feeding program in chronically ture underlines the high risk for uneducated food-insecure areas which provides a mid-morning households to enter into the vicious cycle of snack of 300 Kcal. However, the program has covered 244 P. BHATTACHARJEE AND M. SASSI rural areas of 32 sub-districts and in contrast, only other nutritious food groups in the daily diet plan. The urban slums of 4 sub-districts in Dhaka City by 2003 NFP has already focused on increasing the availability (Ahmed and Babu 2007). The emphasis should be put of the non-cereal nutritious food products and prepar- to expand the coverage of the school feeding pro- ing of a low-cost diet chart based on locally available gram in the free schools for slum children with mini- nutritious food. The government is running an Area mum targeting errors of exclusion-inclusion and Based Community Nutrition program currently active leakages. in 105 sub-districts. However, this program has a future expansion plan only in rural areas (Ministry of Food and Disaster Management 2008). To narrow down the 6.2 Food availability food calorie gap effectively, the possibility to introduce From the empirical model, we observed that the over- door to door campaigns, programs, and training on all feeling of security played a positive role in being nutrition and behavioural change focusing on women food secure. The feeling of security among neighbour- in the slum areas should be evaluated. hoods creates trust in the society which results in a Safe drinking water and hygienic sanitation are the better reliance on the informal safety nets (Zingel et requationuirements for the effective utilisation of food al. 2011). In the NFP, the government of Bangladesh (Rahman 2019). From our empirical results, we can see admitted the need for a well-targeted formal safety that the households using improved pipe water for net programs for the vulnerable slum households drinking purpose had less food calorie gap than the through food distribution and public works programs others. In our sample slum households, the water and (Ministry of Food and Disaster Management 2006). sanitation facilities were mostly managed by the com- However, the interventions targeting the emergency munity leaders and landlords. They usually set up period and seasonal variation are mostly concen- illegal connections of the water network and sell this trated in rural areas (World Food Programme 2015). water at high prices to slum households (Simavi. 2015). Sometimes, the safety net programs adopted in urban Policy intervention is needed to make the landlords areas failed due to the unsuitable design. For exam- accountable for providing adequationuate infrastruc- ple, to tackle the price hike of 2008 in the slums of tural services to slum households for improving their Dhaka, rice was procured by the government and sold access to improved water sources. Any future initiative by the sales unit of the Bangladesh Rifles (BDR) with a of entrepreneurship should be welcomed in the slum subsidised price. This public food distribution scheme with the cooperation of the Community Based did not succeed to reach all the slum households, as Organisations to improve WASH services. they could not afford to lose their working hours by Nonetheless, the piped water provided by the standing in a queue for hours to get 3 kg of rice per Dhaka Water Supply and Sewerage Authority person (Zingel et al. 2011). Therefore, the traditional (DWASA) was reported by people not being able to safety net programs should be redesigned and mod- drink directly, due to its poor quality (Rahman 2019). ified to make them suitable and effective for the Our data revealed that 72.6% of slum households did urban slum dwellers. not perform any water treatment process before drinking it. In this situation, the urban community clinic workers may motivate slum households to 6.3 Food utilisation bring positive behavioural changes for treating their Women are considered as effective vehicles for the drinking water effectively. better utilisation of food through practicing proper care behaviour in households (Ministry of Food and Disaster Management 2006). The study also ascer- 7. Conclusion tained the positive effect of the presence of adult According to our knowledge, this study is one of the women in the household on improving their food very first studies which tried to analyse the determi- security status and calorie intake. Our analysis high- nants of the depth of food calorie gap among the lighted the prevalent knowledge gap among house- food-insecure households in the urban slums of the holds especially about the diet plan and WASH practices. The calorie shares of cereals, oil, and outside Dhaka city. Our results found the significance of vari- meals were significantly higher in comparison with the ables related to all three pillars of food security. FOOD ADDITIVES & CONTAMINANTS: PART B 245 Additionally, the empirical evidence also highlighted University of Pavia, Italy where she teaches Food Economics and Agricultural Development and Quantitative Policy Analysis for the importance of undertaking the sustainable urban Development. slum up-gradation plan ensuring better access to education, employment, housing, WASH services, and overall security for creating a healthy slum com- ORCID munity. Therefore, to improve the food security con- Poushali Bhattacharjee http://orcid.org/0000-0001-6998- dition of the slum households, a multidimensional policy framework should be formulated with associa- Maria Sassi http://orcid.org/0000-0002-6114-6826 tion with different ministries and departments of the government of Bangladesh. Understanding the References empirical results, we highlighted some relevant areas of interventions aiming at improving access to diver- Agarwal S, Sethi V, Gupta P, Jha M, Agnihotri A, Nord M. 2016. Experiential household food insecurity in an urban under- sified nutritious food; employment generation; served slum of North India. SSRN. doi:10.2139/ssrn.2789898 expanding coverage of free schools and school feed- Agyeman J. 2013. Introducing just sustainabilities: policy, plan- ing program for the slum children; redesigning effec - ning, and practice. London (UK and New York, NY): Zed Books. tive safety net programs suitable for urban slums, and Ahmed AU, Babu SC. 2007. The impact of food for education providing training on a balanced diet and wash prac- programs in Bangladesh. In: Food policy for developing tices focusing on women. Our results and policy out- countries: the role of government in the global food system. comes are drawn from the empirical evidence based Cornell University Library Initiatives in Publishing (CIP) Ithaca, New York; p. 3–8, 12. on a limited sample of a particular city. Therefore, the Ahmed S, Meenar M. 2018. Just sustainability in the global effects of the determinants cannot be generalised for South: a case study of the megacity of Dhaka. J Dev Soc. 34 all kinds of contexts and circumstances. In general (4):1–24. doi:10.1177/0169796X18806740. terms, it should be noted that the food and nutrition Akinloye O, Putuma M, Adeyemi ADEYEFA S. 2016. Determinants of food insecurity among the urban poor in the city of security policies implemented in Bangladesh mostly Tshwane, South Africa. J Econ Dev Stud. 4(2):101–114. focused on the rural areas for their intervention. doi:10.15640/jeds.v4n2a9. However, the vulnerable slum households have not Argenti O (2000). Achieving urban food and nutrition security in experienced any positive change in terms of food the developing world: feeding the Cities: food Supply and Distribution. Washington (IFPRI). security for the last two decades (World Food Bangladesh Bureau of Statistics, Ministry of Populations. 2015. Programme 2015). Hence, to achieve the target of Census of slum areas and floating population 2014. http:// sustainable development goal of zero hunger, the 203.112.218.65:8008/WebTestApplication/userfiles/Image/ government of Bangladesh should modify the NFP Slum/FloatingPopulation2014.pdf Bird K. 2013. The intergenerational transmission of poverty: an to make it more inclusive by focusing on this vulner- overview. In: Chronic poverty. Palgrave Macmillan UK; p. 60– able urban slum dweller to ensure their fundamental 84. doi:10.1057/9781137316707_4 right to food. Cameron S. 2010. Access to and exclusion from primary educa- tion in slums of Dhaka, Bangladesh. Pathways to Access Research Monograph. Falmer: Consortium for Research on Disclosure statement Educational Access, Transitions and Equationuity. accessed http://www.create-rpc.org/pdf_documents/PTA45.pdf No potential conflict of interest was reported by the authors. [dataset] Yanez-Pagans, M. Y 2017, July 20. Bangladesh - Urban Informal Settlements Baseline Survey (BUISBS) 2016. The World Bank Microdata Library. accesssed https://microdata. Notes on contributors worldbank.org/index.php/catalog/2864/data_dictionary CEIC. 2018, 12 18. Bangladesh household expenditure per Poushali Bhattacharjee worked as a research scholar in the capita: accesssed https://www.ceicdata.com/en/indicator/ Department of Economics and Management at the University bangladesh/annual-household-expenditure-per-capita of Pavia. She completed her master's degree in the Erasmus Chen Y, Zhu W, Chen Z. 2018. The determinants of mutton Mundus Joint Master Degree: International Development consumption-at-home in urban China using an IHS double- Studies and had a bachelor's degree in Urban and Regional hurdle model. Br Food J. 120(5):952–968. doi:10.1108/BFJ-06- Planning. 2017-0337. Maria Sassi, Ph.D. in agricultural economics, is an Associate Chilton M, Knowles M, Bloom SL. 2017. The Intergenerational Professor at the Department of Economics and Management, circumstances of household food insecurity and adversity. J 246 P. BHATTACHARJEE AND M. SASSI Hunger Environ Nutr. 12(2):269–297. doi:10.1080/ https://blogs.worldbank.org/endpovertyinsouthasia/urbani 19320248.2016.1146195. zation-and-its-discontents-educating-dhaka-s-slum-children Chinnakali P, Upadhyay RP, Shokeen D, Singh K, Kaur M, Singh Kamau M, Githuku J, Olwande J. 2011. Food security in urban AK, Goswami A, Yadav K, Pandav CS. 2014. Prevalence of households: an analysis of the prevalence and depth of hun- household-level food insecurity and its determinants in an ger in Nairobi and its relationship to food expenditure (No. urban resettlement colony in north India. J Health Popul Nutr. 680-2016-46716). 32(2):227–236. accessed http://www.pubmedcentral.nih. Katchova A. 2013. Econometrics Academy. Econometrics gov/articlerender.fcgi?artid=PMC4216959 Models. accesssed https://docs.google.com/file/d/ Chiwaula, L. S., Chirwa, G. C., Binauli, L. S., Banda, J., & Nagoli, J. 0BwogTI8d6EEiUHYtZHpDd1dZakU/edit (2018). Gender differences in willingness to pay for capital- Kolenikov S, Angeles G. 2004. The use of discrete data in pca: intensive agricultural technologies: the case of fish solar tent theory, simulations, and applications to socioeconomic dryers in Malawi. Agricultural and Food Economics, 6(1). indices. Chapel Hill: Carolina Popul Center, Univ of North https://doi.org/10.1186/s40100-018-0096–2 Carolina. 1–59. http://www.cpc.unc.edu/measure/publica tions/pdf/wp-04-85.pdf . Dharmaraju N, Mauleshbhai S, Arulappan N, Thomas B, Marconi Kumar S. 2016. Dimensions and determinants of morbidity: D, Paul S, Mohan V. 2018. Household food security in an evidence from slums of Chandigarh. J Int Dev. 28(6):845– urban slum: determinants and trends. J of Family Med and 856. doi:10.1002/jid.3128. Primary Care. 7(4):819. doi:10.4103/jfmpc.jfmpc_185_17. Lin T-F, Schmidt P. 1984. A test of the tobit specification against FAO. 2010. Guidelines for measuring household and individual an alternative suggested by Cragg. Rev Econ Stat. 66 dietary diversity. Fao. 1–60. 613.2KEN (1):17177. doi:10.2307/1924712. FAO. 2015. Towards a water and food secure future: critical Martínez-Espiñeira R. 2006. A box-cox double-hurdle model of perspectives for policy makers. World Water Counc. I4560:62. wildlife valuation: the citizen’s perspective. Ecoll Econ. 58 accessed. www.fao.org/publications%0Ahttp://www.fao.org/ (1):192–208. doi:10.1016/j.ecolecon.2005.07.006. nr/water/docs/FAO_WWC_white_paper_web.pdf Maxwell D, Levin C, Csete J. 1998. Does urban agriculture help FAO FSP. 2008. An introduction to the basic concepts of food prevent malnutrition? Evidence from kampala. Food Policy. security food security information for action. Food and agri- 23(5):411–424. doi:10.1016/S0306-9192(98)00047-5. culture organisation (pp. 1–3). Meyer BD, Sullivan JX. 2003. Measuring the well-being of the Food Security Portal facilated by IFPRI. 2012. Bangladesh. poor using income and consumption. The Journal of Human accesssed http://www.foodsecurityportal.org/bangladesh/ Resources. 38(SUPPLEMENT):1180–1220. doi:10.2307/ resources Gebre G. 2012. Determinants of food insecurity among house- Ministry of Food and Disaster Management. 2006. National Food holds in addisababa city, Ethiopia. Interdiscip Description of Policy. Peoples’ Republic of Bangladesh. Complex Syst. 10(2):159–173. doi:10.7906/indecs.10.2.9. Ministry of Food and Disaster Management. 2008. The national General Economic Division (GED), Bangladesh Planning Commission, G. food policy plan of action (2008-2015). Food Planning and 2018. Sustainable development goals report 2018. United Nations Monitoring Unit (FPMU). Department of Economic and Social Affairs (p. 64). Mohapatra A. 2015, June 28. Urban poor need microfinance for Hasam MA, Arafin S, Naznin S, Mushahid M, Hossain M. 2017. livelihoods. DownToEarth. accesssed https://www.down Informality, poverty and politics in urban Bangladesh: an toearth.org.in/blog/urban-poor-need-microfinance-for-liveli empirical study of Dhaka City. Jl of Econ and Sustainable hoods-6729 Mohit MA. 2012. Bastee settlements of Dhaka City, Bangladesh: a Dev. 8(14):158–182. review of policy approaches and challenges ahead. Procedia Hjelm L, Mathiassen A, Miller D, Twose A, Wadhwa A. 2017. Soc Behav Sci. 36:611–622. doi:10.1016/j.sbspro.2012.03.067. Creation of a Wealth Index. World Food Programme. Moltedo A, Troubat N, Lokshin M, Sajaia Z, editors. 2014. accesssed https://docs.wfp.org/api/documents/WFP- Analyzing food security using household survey data: stream- 0000022418/download/ lined analysis with ADePT software. The World Bank. Hoddinott J, Yohannes Y. 2002. Dietary diversity as a food doi:10.1596/978-1-4648-0133-4 security indicator. Food Consumption and Nutr Div. 136:1– Moser C, Felton A. 2007. The construction of an asset index 94. doi:10.1016/S0306-9192(99)00035-4 measuring asset accumulation in ecuador. SSRN Electronic IFPRI. 2007. Study of household food security in urban slum J. doi:10.2139/ssrn.1646417 areas of Bangladesh, 2006. final report for world food pro- Peattie L, Aldrete-Haas JA. 1981. ‘Marginal’ settlements in devel- gramme – bangladesh. J of Dev Stud. January. 204. accessed. oping countries: research, advocacy of policy, and evolution http://documents.wfp.org/stellent/groups/public/docu of programs. Annu Rev Sociol. 7(1):157–175. doi:10.1146/ ments/ena/wfp176608.pdf?iframe annurev.so.07.080181.001105. Joshi A, Arora A, Amadi-Mgbenka C, Mittal N, Sharma S, Malhotra Rahman M. 2019, May 6. Consumers condemn Dhaka Wasa B, Grover A, Misra A, Loomba M. 2019. Burden of household water, but the trouble goes much deeper. Dhaka Tribune. food insecurity in urban slum settings. PLoS ONE. 14(4): accesssed https://www.dhakatribune.com/bangladesh/ e0214461. doi:10.1371/journal.pone.0214461. dhaka/2019/05/06/consumers-condemn-wasa-water-but- Kabir M. 2014, July 07. Urbanization and its discontents: educat- the-trouble-goes-much-deeper ing Dhaka’s slum children. World Bank Blogs. accesssed FOOD ADDITIVES & CONTAMINANTS: PART B 247 Raju D, Kim KY, Nguyen QT, Govindaraj R. 2017. Cities, slums, Sociol Methods Res. 31(3):364–388. doi:10.1177/ and early child growth: Empirical evidence from Bangladesh 0049124102239080. (Policy Research Working Paper 8094). Washington (DC): The World Bank. 2019. Urban Population. The World Bank Data. World Bank Group. accesssed https://data.worldbank.org/indicator/SP.URB. Ruel MT, Garrett JL, Morris SS, Maxwell D, Oshaug A, Engle P, TOTL.IN.ZS Haddad L. 1998. Urban challenges to food and nutrition USAID. 2019, May 21. Agriculture And Food Security. accesssed security: a review of food security, health, and caregiving in https://www.usaid.gov/bangladesh/agriculture-and-food- the cities. America. International Food Policy Research security Institute (IFPRI). accessed http://ebrary.ifpri.org/utils/getfile/ Waid JL, Bogard JR, Thilsted SH, Gabrysch S. 2017. Estimates of collection/p15738coll2/id/125713/filename/125744.pdf average energy requationuirements in Bangladesh: adult Sani S, Kemaw B. 2019. Analysis of households food insecurity male equationuivalent values for use in analyzing household and its coping mechanisms in Western Ethiopia. Agric Food consumption and expenditure surveys. Data Brief. 14:101– Econ. 7(1). doi:10.1186/s40100-019-0124-x. 106. doi:10.1016/j.dib.2017.07.022 Sassi, M. (2017). Understanding food insecurity: Key features, Wooldridge JM. 2003. Introductory econometrics: a modern indicators, and response design. Understanding Food approach. Econ Anal. 2 nd. 824. doi: 10.1198/jasa.2006.s154 Insecurity: Key Features, Indicators, and Response Design World Food Programme. 2002. Urban food insecurity: strategies (pp. 1–132). Springer International Publishing Switzerland. for WFP food assistance to urban Areas. Rome. https://doi.org/10.1007/978–3–319–70362–6 World Food Programme. 2015. Food insecurity and undernutri- Shaheen, N., Rahim, A.T.M.A., Mohiduzzaman, Md., Banu, C.P., tion in the urban slums of bangladesh a 2013 survey of slum Latiful Bari, Md., 664 Tukun, A.B., Mannan, M.A., households in Dhaka, Barisal and Sirajganj. https://docu Bhattacharjee, L., Stadlmayr, B., 2013. Food composition ments.wfp.org/stellent/groups/public/documents/ena/ table for Bangladesh. Institute of Nutrition and Food wfp282624.pdf?_ga=2.105651329.385441684.1569320124- Science, University of Dhaka, Bangladesh. Retrieved from 600239824.1569320124 http://www.fao.org/fileadmin/templates/food_composition/ Yen ST, Jones AM. 1997. Household consumption of cheese: an documents/FCT_10_2_14_final_version.pdf inverse hyperbolic sine double-hurdle model with depen- Sheng YK, Brown A. 2018. Prosperity for all: enhancing the dent errors. Am J Agric Econ. 79(1):246–251. doi: 10.2307/ informal economy through participatory slum upgrading. 1243958. Cardiff: UN-HABITAT. http://unhabitat.org/prosperity-for-all- Yoo W, Mayberry R, Bae S, Singh K, He QP, Lillard JW. 2014. A study of enhancing-the-informal-economy-through-participatory- effects of multicollinearity in the multivariable analysis. Int. J. Appl. Sci. slum-upgrading/ Technol. 4(5):9–19. accessed http://www.pubmedcentral.nih.gov/arti Simavi. 2015, June 26. The challenge of ensuring safe drinking clerender.fcgi?artid=PMC4318006 water and sanitation in Dhaka’s low-income communities. Zingel WP, Keck M, Etzold B, Bohle HG. 2011. Urban food security https://simavi.org/long-read/safe-drinking-water-sanitation- and health status of the poor in Dhaka, Bangladesh. In: dhaka/ Krämer A., Khan M., Kraas F. (eds) Health in megacities and Smith DA, Brame R. 2003. Tobit models in social science urban areas. Contributions to Statistics Heidelberg: Physica; research: some limitations and a more general alternative. p. 301–319. https://doi.org/10.1007/978-3-7908-2733-0_19
Journal
International Journal of Urban Sustainable Development
– Taylor & Francis
Published: May 4, 2021
Keywords: Food insecurity; food calorie gap; urban slums; IHS transformed DH model; Bangladesh
