Consumption of food away from home in Bangladesh: Do rich households spend more?

Accepted Manuscript Consumption of food away from home in Bangladesh: Do rich households spend more? Khondoker A. Mottaleb, Dil Bahadur Rahut, Ashok K. Mishra PII:

S0195-6663(17)30433-6

DOI:

10.1016/j.appet.2017.03.030

Reference:

APPET 3391

To appear in:

Appetite

Received Date: 9 August 2016 Revised Date:

17 February 2017

Accepted Date: 21 March 2017

Please cite this article as: Mottaleb K.A., Rahut D.B. & Mishra A.K., Consumption of food away from home in Bangladesh: Do rich households spend more?, Appetite (2017), doi: 10.1016/ j.appet.2017.03.030. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Consumption of Food Away from Home in Bangladesh: Do Rich Households Spend More? Khondoker A. Mottaleb1 Dil Bahadur Rahut2 and Ashok K. Mishra3

RI PT

SC M AN U TE D EP

2. 3.

Agricultural Economist, Socioeconomics Program, International Maize and Wheat Improvement Center (CIMMYT), Carretera México-Veracruz Km. 45, El Batán, Texcoco, México, C.P. 56237. Email: [email protected] Program Manager, Socioeconomics Program, CIMMYT. Corresponding author: Kemper and Ethel Marley Foundation Chair, Morrison School of Agribusiness, W.P. Carey School of Business, Arizona State University, 235M Santan Hall, 7231 E. Sonoran Arroyo Mall, Mesa, AZ 85212, Ph: 480.727.1288, Fax 480.727.1961, Email: [email protected]

AC C

1.

ACCEPTED MANUSCRIPT

Consumption of Food Away from Home in Bangladesh: Do Rich Households Spend More? Abstract

4

While consumption of food away from home (FAFH) is an established phenomenon among

5

households in the developed countries, FAFH is a growing phenomenon in many middle-income

6

and rapidly growing developing countries. Although, studies are available on the factors

7

affecting consumption of FAFH in developed countries, there is a paucity of such studies in

8

developing countries. This study examines households’ choice of and expenditures on FAFH.

9

We used information from Bangladeshi households and applied a double-hurdle regression

10

model estimation procedure. Findings show that, in general, rich households are spending

11

proportionately less on FAFH and, over time, the trend is continuing. Although households with

12

female members who work in the non-farm sector are more likely to consume FAFH, educated

13

household heads and spouses, and particularly urban households are less likely to consume and

14

spend on FAFH. As the problem of food adulteration by dishonest sellers is rampant in

15

Bangladesh, perhaps it discourages rich, urban and households headed by educated heads and

16

spouses to consume and spend more on FAFH. Based on the findings, some points of

17

interventions are also prescribed in this study.

18 19

Keywords: food-away-from-home; food safety; household behavior; income; schooling; doublehurdle model

20

JEL Classification: C24, D03, D12, D18, E2

AC C

EP

TE D

M AN U

SC

RI PT

1 2 3

1

ACCEPTED MANUSCRIPT

Consumption of Food Away from Home in Bangladesh: Do Rich Households Spend More?

21 22

Introduction

24

One of the important stylized facts about people’s dietary habits is that, as income increases,

25

people initially shift their dietary preferences gradually from cereal-based diets to high food

26

value enriched items, such as fish, meat, fruit and vegetables (Barker et al., 1985; Ingco, 1991;

27

Hossain, 1998; Pingali and Khwaja, 2004; Pingali, 2006; Kearney, 2010). With a further increase

28

in income, peoples’ dietary preferences shift further towards processed foods and particularly

29

towards ‘food away from home’ (FAFH) (Yen and Huang, 1996; Byrne et al., 1996; Stewart et

30

al., 2004; Pingali, 2006;). For example, in 1970, on average, the expenditure on FAFH in the

31

USA was 25.9% of the total food expenditure of a household; in 2012 it increased to 43.1%

32

(USDA, 2016). Importantly, now FAFH in the USA is treated as a necessary food item (Byrne et

33

al., 1998). In a recent study, D’Addezio et al. (2014) show that 9.5% of the total households in

34

the UK and 21.3% in Italy, 20.9% in Belgium, 17.7% in Poland and 23.1% in Demark take lunch

35

away at least three times a week. The increasing expenditure on FAFH can also be observed in

36

the middle income and rapidly-growing developing countries. For example, in 1973, the average

37

expenditure on FAFH by a Malaysian household was only 4.6% of their total food expenditure;

38

whereas, in 1999, the share increased to 10.9% (Lee and Tan, 2006). A question arises as to what

39

the factors are that affect the decision to take FAFH and expenditure on it, particularly in

40

developing countries.

SC

M AN U

TE D

EP

AC C

41

RI PT

23

Becker (1965) explained classical demand theory, in which consumption is treated as the

42

output of a home production function that employs both time and expenditure as inputs. Based

43

on Becker’s classical demand theory, empirical literature often pointed out that income, prices,

44

demographics, opportunity costs and time constraints can affect a household’s decision to spend 2

ACCEPTED MANUSCRIPT

on FAFH. For example, households with higher income may take more FAFH, not only because

46

of their capability to do so (e.g., Liu et al., 2012), but also because the household members tend

47

to enjoy leisure time, free from preparing food and cleaning the kitchen and dishes at home

48

(McCracken and Brandt, 1987; Byrne et al., 1998). The working hours of a household head,

49

spouse, family members and farm and non-farm affiliations can also significantly affect the

50

decision to take FAFH because household members who are involved in laborious jobs and work

51

long hours outside the house could prompt them to spend more on FAFH (e.g., Byrne et al.,

52

1998). For example, in India, the larger the share of salary and business income into the total

53

household income, the higher the probability of taking food and expenditure on FAFH is (Gaiha

54

et al., 2013). A few studies indicate that age is also a decisive factor in determining FAFH, as at

55

retirement, although the overall food expenditure declined, this decline is, however,

56

accompanied by an increase in the time spent in food preparation at home (e.g., Aguiar and

57

Hurst, 2005).

SC

M AN U

TE D

58

RI PT

45

The non-farm employment of female household members and female education can affect the decision to spend on FAFH (e.g., Liu et al., 2012) because, in developing countries, the

60

major role of female members in a household is to prepare and cook food. If the spouse and other

61

female members are involved in non-farm employment, the spouse and other female members

62

might have less time or may have a higher opportunity cost of preparing food at home. It can

63

positively influence taking FAFH. The size of a household can also affect the decision to spend

64

on FAFH because of the economies of scale in food preparation, in which smaller households

65

may have more cost advantages of taking FAFH than a household with more family members

66

(e.g., McCracken and Brandt, 1987; Deaton and Paxson, 1998; Gan and Vernon, 2003). This is

67

because, while there is little difference in time spent on cooking for a few or many members,

AC C

EP

59

3

ACCEPTED MANUSCRIPT

68

larger households can buy food and cook items at a cheaper rate as they can buy these in bulk.

69

Thus, the preparation of food at home for a household with more family members tends to be

70

more economical than for smaller households and vice versa. Although the above-mentioned studies and findings have added many important insights

RI PT

71

to household behavior on FAFH, these studies are mostly based on developed and middle-

73

income country cases. Existing studies, therefore, do not necessarily reflect the issues related to

74

FAFH in developing countries; consequently, they do not necessarily capture the dynamics of

75

FAFH in developing countries. Importantly, in many of the rapidly-growing developing

76

countries, the food quality and the safety of FAFH is a major concern. For example, in

77

Bangladesh and India, two of the rapidly-developing economies in South Asia, food adulteration

78

with the application of hazardous chemicals is a major concern (e.g., Sudershan et al., 2009;

79

Khan, 2009, Sabet 2013; Nasreen and Ahmed, 2014; Gahukar, 2014; Sobhani, 2015; Rahman et

80

al., 2015). In-depth studies in developing countries using household-level data sets can provide

81

deep insights into the behavior of the households on the decision to spend on FAFH. The

82

household’s decision to take FAFH depends on local culture and tradition as well as household-

83

specific characteristics, such as income and education of the household head and spouse. It

84

indicates that country-specific case studies based on household-level information can unveil the

85

important factors that influence a household’s decision to take FAFH.

M AN U

TE D

EP

AC C

86

SC

72

To fill the above-mentioned gaps the present study investigates the underlying factors

87

that affect the decision to spend on FAFH by households in developing countries, using

88

Bangladesh as a case. Since 2000, the economy of Bangladesh has grown annually between

89

4.0%-7.1% (World Bank, 2016). The per capita nominal Gross Domestic Product (GDP) has

90

increased from $363 in 2000 to $1,115 in 2014, a 52% increase in a short time (GOB, 2015).

4

ACCEPTED MANUSCRIPT

Table 1 presents the occupational distribution, by sector, of economically active people (15 years

92

or older) in Bangladesh; it shows that the active labor force in Bangladesh has increased from

93

about 35 million to more than 54 million from 1990-2010. Secondly, Bangladesh is observing

94

rapid urbanization. For example, in 2001, about 20% of the population lived in urban areas, and

95

that increased to 32% in 2012 (World Bank, 2016). Reardon and Timmer (2014) stress that in the

96

more urbanized countries of South Asia, East and Southeast Asia urban consumers are

97

responsible for roughly two-thirds, even up to three-quarters, of all food expenditures.

98

Furthermore, the majority of the population is engaged in agriculture, forestry, and fishery. Note

99

that the hospitality sector like the hotel and restaurant industry often is largely based on the

M AN U

SC

RI PT

91

economic status of the population. One would expect a direct correlation between revenues of

101

the hotel and restaurant industry and per capita household income.1 As noted by Mihalopoulos

102

and Demoussis (2001) and Chang and Mishra (2008) an increase in per capita household income

103

has a direct effect on household expenditures and, in particular, food consumption away from

104

home (Mihalopoulos and Demoussis, 2001; Chang and Mishra, 2008). The issue warrants careful

105

investigation to understand how the increase in per capita income together with speedy

106

urbanization can shift the dietary preferences of consumers towards processed and ready-to-eat

107

foods and in particular towards consumption of FAFH.

108

[Insert Table 1 here]

AC C

EP

TE D

100

Empirical analysis of income and the demographic factors affecting consumption of

109 110

FAFH has become increasingly important, not only for consumers and sellers, but also for 1

Table 2 shows that yearly net revenue and monetary value of the hotel and restaurant business enterprises has decreased significantly during 2000-2010; the number of salaried workers in the sector has also decreased, especially during 2005-2010 period.

5

ACCEPTED MANUSCRIPT

policymakers and researchers, who seek policy incentives to increase spending in the hotel and

112

restaurant food industry, increase employment, and greater awareness of food safety and public

113

health security. Importantly, similar to Bangladesh, FAFH is a growing phenomenon in rapidly-

114

emerging developing countries, and the policy implications generated from the present study can

115

be generalized for other developing countries as well. Herein lies the objective of this study: we

116

examine the factors affecting a Bangladeshi household’s decision to consume FAFH and the

117

expenditures on FAFH. To accomplish this aim, we use a larger sample than previously reported,

118

of Household Income and Expenditure Survey (HIES) data (HIES 2000, HIES 2005, and HIES

119

2010) and double-hurdle (DH), two-limit Tobit (2LT) and truncated regression (TR) estimation

120

procedures.

SC

M AN U

121

RI PT

111

This study shows that although FAFH is a growing phenomenon in Bangladesh, the level of physical as well as human capital of the household head and spouse significantly influences

123

the likelihood of consuming FAFH and expenditures on FAFH. In general, relatively wealthy

124

households are less likely to consume FAFH, and as a result, tend to spend less on FAFH.

125

Relatively highly educated household heads and spouses are also less likely to consume FAFH

126

and spend less on it. A possible explanation, in the presence of pervasive food adulteration by

127

dishonest sellers, could be that educated and economically affluent households are more aware

128

and concerned about food safety and health issues related to FAFH. Finally, we find that, during

129

the years sampled, the real net revenue earned by the hotel and restaurant businesses in

130

Bangladesh has also declined significantly.

132

EP

AC C

131

TE D

122

The next section presents the data sources, and characteristics of sampled households,

and their behavior regarding FAFH over the sampled years. Section 3 specifies the model used to

6

ACCEPTED MANUSCRIPT

133

examine the factors that affect the decision to eat FAFH and the amount of money spent on

134

FAFH. The concluding section draws some policy implications based on the findings.

135

2.0 Data sources and sampling This study uses Household Income and Expenditure Survey (HIES) data from

RI PT

136

Bangladesh collected in 2000, 2005 and 2010 by the Bangladesh Bureau of Statistics (BBS). The

138

BBS uses a two-stage stratified random sampling method in which in the first stage, the BBS

139

selects primary sampling units (PSUs) consisting of specific geographical areas. In the second

140

stage, it randomly selects 20 households from each PSU that represent rural, urban, and

141

statistical metropolitan areas (SMAs). For example, in the 2000 HIES survey, 7,440 households

142

were randomly selected from seven divisions, 64 districts, and 303 sub-districts. In the HIES

143

2005, 10,080 households were randomly chosen from seven divisions, 64 districts, and 355 sub-

144

districts. Finally, in the HIES 2010, 12,240 households were randomly selected from seven

145

divisions, 64 districts, and 384 sub-districts. Since information on food prices and non-food

146

expenditure was missing for the 122 households, this study uses information from 29,648

147

households, of which 9,182 and 20,466 were from urban and rural areas, respectively.

148

3.0 Model specification and estimation procedure

M AN U

TE D

EP

149

SC

137

Among the sampled 29,648 households, 64% reported positive expenditure on FAFH and the other 36% of the sampled households did not spend on FAFH. It means that a significant

151

proportion of households reported zero expenditure on FAFH. In this case, the use of a Tobit

152

model to estimate the function explaining FAFH is recommended on theoretical grounds in

153

preference to the OLS models for datasets with censored samples (Gujarati, 1995). A major

154

drawback of the Tobit model, however, is that it is restrictive in its parameterization because the

155

factors that affect expenditures on FAFH are assumed to be the same as those that also affect its

AC C

150

7

ACCEPTED MANUSCRIPT

156

probability (of choosing FAFH). Furthermore, Tobit estimation may not provide robust results

157

across distributional assumptions of the error term (Arabmazar and Schmidt, 1981, 1982). We explored a model popularly known as the “double-hurdle” model, originally

159

formulated by Cragg (1971) and later developed into a user interface in the STATA program by

160

Burke (2009). Based on the DH model, it is assumed that a household’s decision to consume (yes

161

=1, no = 0), which is the first hurdle, and expenditures on FAFH, which is the second hurdle, are

162

determined separately by two sets of explanatory variables. Note, in order to obtain a positive

163

proportion of spending on FAFH, two hurdles that are ‘yes’ in the first hurdle which is a decision

164

to consume FAFH, and positive expenditure in the second hurdle, which means the proportion of

165

expenditure on FAFH to total food expenditure must be passed separately. Interestingly, if the

166

explanatory variables in the two hurdles are the same, the Tobit model is nested within Cragg’s

167

alternative approach (Burke, 2009). In our case, the empirical model can be specified as:

168

∗
 =  ∝ + decision for FAFH (yes=1, no=0);  ~(0, 1)

(1)

169

∗
 =   +  proportion of expenditures on FAFH;  ~(0, 2)

(2)

170

∗ ∗
 =   +  only if
 =  ∝ + > 0 and
 =   +  > 0;

(3)

171


 = 0 otherwise

172

∗ where
 is a latent variable presenting the household’s decision to consume FAFH (yes = 1, 0

173

∗ otherwise);
 is a latent variable presenting the expenditures on FAFH;
 is the observed

174

dependent variable, expenditures on FAFH by a household;  is a vector of explanatory

175

variables explaining the decision to consumer FAFH;  is a vector of explanatory variables

176

explaining expenditures on FAFH; vi and ui are the respective error terms assumed to be

177

independently and normally distributed. Thus, the stochastic specification of the error terms in

178

participation equation and proportion of expenditure equations can be written as:

AC C

EP

TE D

M AN U

SC

RI PT

158

8

ACCEPTED MANUSCRIPT

  ~[

179

0 1 0 , ] 0 0 2

(4)

The DH model with independent error terms can be estimated by the following log-likelihood

181

function (Moffatt, 2005; Aristei et al., 2007):

182

!! = " ln %1 − Φ(w) α)+ (

183

In the empirical estimation approach, to overcome the inconsistency in estimating the empirical

184

model in the presence of heteroscedasticity and following Yen and Huang (1996), we allowed

185

the variance of the error term to vary across observations by specifying it as a function of the

186

selected continuous variables. Specifically:

187

 = exp (: ℎ)

188

where zi are some elements of  and ; however, only the total expenditure on all food items,

189

years of schooling for spouse, and the intercept were statistically significant and were thus

190

included in the final variance equation. The empirical model was specified as follows:

191


 = >2 +  (?@ABC  DEBFC) +  (GCH IJK CK. ) + +(MMFCNAOHJKℎPID)  +

192

∑TU 3F (RR)S +

193

∑XU V@ (WCJH)X +  (YHZJB F NN
)[X + ∑_^U \ ( ]KCBFP@ HC OHA K)^ + _ (`2 a WCJH 2005) +

194

c (`3 a WCJH 2005) + e (`4 a WCJH 2005) + T (`2 a WCJH 2010) +

195

g (`3 a WCJH 2010) + h (`4 a WCJH 2010) + i

196

Where
 is a vector of dependent variables that assumes a value of 1, when a household i chose

197

to consume FAFH (the first hurdle) and assumes any positive value when a household actually

198

spent a portion of their food expenditure on FAFH. Among the independent variables, per capita

199

expexpnditures. is real yearly per capita total food and non-food expenditure;

45,-. /

6

0

)1

(5)

SC

0



)1 + " ln %Φ(w) α) 0 3 (

M AN U

2

,-. /

RI PT

180

AC C

EP

TE D

(6)

9

(7)

ACCEPTED MANUSCRIPT

MMFCNAOHJKℎPID is a vector of variables that includes the age and sex of the head of the

201

household; the sex dummy assumes a value of 1 if a household head is female, and 0 otherwise;

202

years of schooling of the household head and spouse, size of the household (numbers of

203

household member), DD, representing six divisional dummies by location of the household,

204

where Barisal division is the base division (Barisal = 0), year is two year dummies for three

205

sampled years in which year 2000 is the base; urban dummy assumes a value of 1 if a household

206

is located in the urban area, 0 otherwise; Expenditure group is three dummies for four

207

expenditure (income) groups. In our case, we have expenditure quartiles: Q1, Q2, Q3, and Q4.

208

Q1 is the base (Q1=1) where the poorest households belong to; >2 is a scalar,  3S , P , V

209

\P , are the parameters to be estimated, and i is the random error term. To examine the validity

210

of the DH model against the two-limit Tobit model (2LT), and truncated regression (TR), this

211

study also reports 2LT and TR results. Finally, the Chi-square test was performed to validate the

212

suitability of the DH model against 2LT and TR model. The formula used in this case is as

213

follows:

214

λ= 2* (LLprobit+ LLtruncreg- LLTobit)

215

where LLprobit, LLtruncreg, and LLTobit are log likelihoods after probit, truncated regression

216

and Tobit. The test statistics (λ= 9167.6) show a rejection of the 2LT model in favor of DH

217

model. Therefore, in the next section, we mainly report the results of the DH estimation

218

approach.

219

4.0 Major findings

220

AC C

EP

TE D

M AN U

SC

RI PT

200

Tables 3-5 characterize the sample households and present the general findings. Table 3

221

presents demographic information on the sampled households, in which the sampled households

222

are divided into four expenditure quartiles based on the yearly per capita total food and non-food

10

ACCEPTED MANUSCRIPT

expenditure in real Bangladesh Taka (BDT)2. Following Deaton and Zaidy (2002), this study

224

normalized the total yearly expenditure of the sampled households by dividing it with the family

225

members of the households. On average, households spent BDT11.27 thousand yearly per capita

226

on food and non-food items, in which BDT5.99 thousand were spent on non-food items and the

227

other BDT5.28 thousand were spent on food items. However, the top two richest households

228

groups, Q3, and Q4 spent BDT 10.46 thousand and BDT22.70 thousand yearly per capita on

229

food and non-food items, whereas the poorest group (Q1) spent BDT 4.68 thousand. Table 3

230

shows that, on average, a household is most likely to be headed by a 45-year-old male with four

231

years of schooling, and nearly 70% of them are likely to reside in the rural areas. On average, a

232

household is comprised of nearly five members. On the other hand, spouses have three years of

233

schooling, and 7% of them are employed in the non-farm sector.

234

[Insert Table 3 here],

M AN U

SC

RI PT

223

Table 3, however, shows that rich households have more human capital and fewer family

236

members than poor families. On average, the educational attainment of the head and spouse of a

237

household in Q4 is more than six and five years, respectively. By contrast, on average, the level

238

of education of the head and spouse of a household in Q1 is less than two years. Table 3, also

239

shows that compared to households in Q1 and Q2, households in Q4 are more likely to reside in

240

the urban areas. It is interesting to note that despite the fact that spouses of the rich households

241

tend to have more schooling, the participation in the non-farm employment by the spouses is

242

slightly higher over the expenditure quartiles. Table 4 indicates that the Stone’s price index3 is

243

greater for families in the higher income quartiles which reflect that rich households spend more

244

on the sampled food items than the poor families.

AC C

EP

TE D

235

2

USD1= BDT 78 approximately. Defined by jB J (K) ≈ ∑l l jBKl , where pj is the price of the jth item and wj is the share of expenditure on item j. To calculate Stone’s price index this study considered all commodities consumed by a sampled household. 3

11

ACCEPTED MANUSCRIPT

245

Table 4 presents information in such a way as to capture the trends of consumption and expenditure on FAFH over the expenditure quartiles and years sampled. It shows that nearly

247

64% of the households consumed FAFH, and the average household spent about BDT880 on

248

FAFH in a year, which is about 4% of total food expenditures. In the HIES data sets, the

249

expenditure of FAFH included expenditures on 13 food items including rice, fish, meat, cake,

250

sandwich, burger, hotdog, pizza, tea, coffee, soft drinks and other food items. Table 4, however,

251

reveals that relatively rich households consumed more FAFH and spent more on FAFH, both in

252

absolute terms and the share of expenditures on FAFH to total food expenditures. A scrutiny of

253

Table 4, however, shows that over the years the rich households progressively spend less on

254

FAFH. For example in 2000, the households in Q4 spent BDT1720 on FAFH, whereas between

255

2005 and 2010 it was reduced to BDT 1420 and BDT 1120, respectively. Figure 1 also confirms

256

the findings of Table 4 that the expenditure on FAFH has remained stagnant or declined

257

marginally, over the period sampled. The findings in Table 4 confirm that although a relatively

258

higher share of rich households consumes and spend more on FAFH in general, relatively rich

259

households increasingly spent less on FAFH. A possible explanation may be that relatively rich

260

households are more concerned about, and aware of, rampant food adulteration with hazardous

261

chemicals along with poor government monitoring systems.

262

[Inset Table 4 here]

263

[Inset Figure 1 here]

SC

M AN U

TE D

EP

AC C

264

RI PT

246

Table 5 shows that households where spouses’ work outside the home, consume and

265

spend more on FAFH, and spend more on food items than others. In the next section, we develop

266

an empirical model to examine the impacts of household income, spouse’s non-farm

267

employment, and other factors that determine the decision to consume and spend on FAFH.

12

ACCEPTED MANUSCRIPT

[Insert Table 5 here]. Table 6 presents the empirical findings explaining FAFH by the sampled

269

households in Bangladesh. The first panel of Table 6 presents the estimated functions explaining

270

the decision to consume (yes =1, 0 =otherwise) and expenditures on FAFH.

271

[Insert Table 6 here]

RI PT

268

We find that both physical capital in terms of yearly per capita total food and non-food

272

expenditure, human capital in term of years of schooling of the household head and spouses, and

274

age, have a negative and significant impact on the likelihood of consuming, and the amount

275

spent on FAFH. Recall that educated household heads and spouses are more likely to be

276

financially solvent and more aware of the dangers of consuming FAFH.4 Consequently, these

277

families tend to spend less on FAFH. The estimated DH model in Table 6, however, shows that

278

families with a spouse working outside the home, in the non-farm sector, for instance, are more

279

likely to consumer FAFH. Findings in Table 6 reveal that female heads of households are less

280

likely to consume FAFH. In Bangladesh, compared to the male head of households, female

281

heads of households are less likely to work outside the home, and the opportunity cost of food

282

preparation at home tends to be low. Additionally, a female household head may consider family

283

matters, such as caring for the elderly and children, family members’ health as it relates to food

284

and food safety more important than FAFH.

EP

TE D

M AN U

SC

273

The likelihood of consuming FAFH by a household increases with an increase in the

AC C

285 286

number of family members, however; we also find a negative and significant effect of the

287

number of family members on the expenditures on FAFH. In other words, households with more

288

family members are more likely to consume FAFH but spend less proportion of their food 4

In Bangladesh, sellers apply formalin (formaldehyde, 40% by volume; methanol, 6 to 13%; and water) on various food items, in order to increase the shelf life of farm products and fish (Sabet, 2013). Media, including television news and newspapers, report adulterated food and fruits in stores and shops by sellers (Khan, 2012). While the government has taken some projects to prevent the widespread use of formalin in fish and other food items (GOB, 2012), these steps are insufficient.

13

ACCEPTED MANUSCRIPT

expenditure on FAFH. Households with more family members generally enjoy the economies of

290

scale in preparing food at home. Probably, this is why a household with more family members

291

spend relatively less on FAFH. This finding is consistent with previous studies (e.g., McCracken

292

and Brandt, 1987; Deaton and Paxson, 1998; Gan and Vernon, 2003). Compared to households

293

in the Barisal division, households located in the Dhaka, Rajshahi, and Sylhet divisions are less

294

likely to consume and spend on FAFH. Dhaka is the capital city of Bangladesh, and the largest

295

industrial belt is located around Dhaka. On the other hand, households in the Sylhet division

296

receive the sheer portion of remittance income in Bangladesh. A large number of migrants from

297

Sylhet reside in the United Kingdom and other European countries. Note that households in the

298

Dhaka and Sylhet divisions are economically affluent; they are also more likely to be aware of

299

the quality of food being served in restaurants and hotels in the presence of rampant food

300

adulteration in Bangladesh. However, we find that households located in other regions

301

(Chittagong, Rangpur, and Khulna divisions) are more likely to consume FAFH. Table 6 shows

302

that, compared to rural households, urban households are less likely to consume and spend on

303

FAFH. There may be two plausible reasons. First, urban households are relatively more affluent

304

both in terms of physical and human capital. Probably, it is more economical for an urban

305

household to prepare hygienic food in their home than to consume FAFH in the presence of

306

rampant food adulteration. The negative and significant coefficient on a 2010-year dummy in

307

both hurdles (decision and expenditures on FAFH) indicates that the overall demand FAFH in

308

Bangladesh is decreasing, probably due to food quality and food safety issues.

SC

M AN U

TE D

EP

AC C

309

RI PT

289

Table 6 shows that the likelihood of consuming FAFH and expenditures on FAFH are

310

significantly and positively determined by Stone’s food price index. The findings support the

311

hypotheses that consuming FAFH is a form of leisure for a wealthy household that usually

14

ACCEPTED MANUSCRIPT

spends more on food (e.g., McCracken and Brandt, 1987; Byrne et al., 1998). For example,

313

McCracken and Brandt, (1987) found that the conditional income elasticity for FAFH increases

314

from 0.023 with an income less than USD5000/year to 0.178 with an income more or equal to

315

USD20,000 at the household level. Similarly, Byrne et al., (1998) demonstrated that income

316

increased the likelihood of taking FAFH positively and significantly from 1982 to 1989 in the

317

USA. In the case of Bangladesh, on average, relatively rich households as measured by Stone’s

318

price index are more likely to consume and proportionately spend more on FAFH (α1 =

319

0.49, 1 = 064, G < 0.001). Table 6, however, shows that the total food and non-food

320

expenditures are, in fact, negatively correlated with the consumption of FAFH, and importantly,

321

wealthy households actually spend less on FAFH after controlling for other effects, once they

322

decide to consume FAFH. Finally, Table 6 shows that rich households (income quartile Q3 and

323

Q4) are less likely to choose consumption of FAFH, and progressively reduced their

324

expenditures on FAFH, in 2005 and 2010 on it, compared to 2000. Figure 2, panel a presents the

325

parameter estimates explaining the likelihood of consumption of FAFH. The panel shows that, in

326

2005, the coefficients of the interaction terms between expenditure quartiles and year dummies

327

were 0.05, 0.02, and 0.002 for Q2, Q3 and Q4 expenditure quartile, respectively. In 2010, the

328

coefficients of the interaction terms between expenditure quartile and years dummies decreased

329

to -0.07, -0.06 and -0.03 for Q2, Q3, and Q4 expenditure quartiles, respectively. On the other

330

hand, panel b of Figure 2 presents the parameter estimates explaining the share of expenditure on

331

consumption of FAFH. The figure reveals that in 2005 the coefficients of the interaction terms

332

between expenditure quartiles and year dummies were 0.019, -0.01 and -0.09 for Q2, Q3 and Q4

333

expenditure groups, respectively. In 2010, the same coefficients decreased to -0.04, -0.02 and -

334

0.13 for Q2, Q3 and Q4 expenditure groups, respectively. This finding demonstrates that rich

AC C

EP

TE D

M AN U

SC

RI PT

312

15

ACCEPTED MANUSCRIPT

335

households, over the years have spent less on FAFH, which is in contrast to the trend in the USA

336

(McCracken and Brandt, 1987) and Malaysia (Lee and Tan, 2006).

337

[Insert Figure 2 here] The estimated functions explaining the proportion of food expenditure on FAFH

RI PT

338

estimated by applying 2LT and TR in Table 7 also vividly demonstrate that relatively wealthy

340

and urban households and particularly the households headed by relatively more educated heads

341

and spouses are less like to consume and spend less on FAFH. Both 2LT and TR regression

342

results reveal that the trends of consuming and spending on FAFH in Bangladesh are negative

343

and wealthy households progressively consume and spend less on FAFH. The similarities in the

344

findings across different models indicate the robustness of the findings.

345

[Insert Table 7 here]

346

5.0 Conclusion and policy implications

M AN U

While consuming FAFH is a growing phenomenon among households in many rapidly

TE D

347

SC

339

growing developing countries, existing studies on the determinants of a households’ behavior on

349

FAFH are mostly available for developed countries. Using the Household Income and

350

Expenditure Survey (HIES) data and a double-hurdle estimation approach, this study examines

351

the factors that affect Bangladeshi households’ decision to consume and spend on FAFH.

352

Findings from this study reveal that urban, relatively wealthy households, and households headed

353

by a relatively more educated head and spouse are more likely consuming less and spending less

354

on FAFH. The households from relatively affluent divisions such as Dhaka and Sylhet are less

355

likely to consume and spend on FAFH. Contrary to the findings in the developed world,

356

educated heads and spouses and affluent households in Bangladesh are less likely to consume

357

FAFH. Additionally, wealthy households are progressively reducing their expenditures on

AC C

EP

348

16

ACCEPTED MANUSCRIPT

FAFH. A plausible explanation could be the widespread incidence of food adulteration through

359

the application of hazardous chemicals in Bangladesh. As a result,wealthy and educated

360

households tend to avoid FAFH. It might be the case that the urban, rich and the households

361

headed by more educated heads and spouses tend to avoid oily and junk food. Consequently,

362

their likelihood of consumption and the proportion of expenditure on FAFH are low. This may

363

indicate a market failure stemming from an information asymmetry problem between hotel and

364

restaurant owners (on food quality) and consumers (households) in Bangladesh. Based on the

365

findings, we suggest both market-based and government-led initiatives to address the food safety

366

problems.

SC

M AN U

367

RI PT

358

First, it is necessary to strengthen both equipment and human resources needed in monitoring food quality. The import of hazardous chemicals, such as formalin, can also be

369

strictly monitored to prevent any misuse of it, especially on food items. Public awareness

370

programs, including advertisement on television and community radio about the harmful effects

371

of food adulteration, and the penalties should be broadcasted. Above all, a hazardous chemical

372

detection kit, which is easy to use and maintain, should be developed and made publicly

373

available at an affordable price. Among market-based solutions, the development and declaration

374

of the hazardous-chemical-free kitchen and food markets could be introduced in the major cities

375

and semi-urban areas across the country. Civil society, including merchant groups, could be

376

educated and encouraged to take the necessary steps in establishing such markets. Finally, the

377

motivation (mainly for storage and shelf life) behind the application of hazardous chemicals to

378

food items by dishonest sellers must be addressed. A public-private partnership could be initiated

379

to establish low-cost cold-storage facilities for fish and seasonal fruits. Donor agencies, such as

380

the World Bank and other developed countries, could help to establish storage facilities.

AC C

EP

TE D

368

17

ACCEPTED MANUSCRIPT

Although the present paper, using large nationally-representative data sets and applying

382

econometric estimation procedures demonstrated the consumption decision and expenditure on

383

FAFH in Bangladesh, this study does not necessarily focus on whether or not relatively wealthy,

384

educated and urban households in Bangladesh are spending more on semi- processed or high-

385

value-added packaged food items in the superstore. It might be the case that, relatively wealthy,

386

educated and urban households in Bangladesh are spending less on FAFH due to concerns about

387

food adulteration and hygienic issues, and that they are actually substituting FAFH by spending

388

more on semi-processed food items in the superstores. Since the present study does not focus on

389

the food substitution between FAFH and semi-processed food items in the superstores, it is

390

difficult to know from the current results what has driven the reduction in away-from-home

391

spending in addition to the rampant food adulteration problem. This study does not shed light on

392

the food ingredients and food quality in the hotel and restaurants in Bangladesh due to the

393

paucity of such data from the secondary sources. As there is a strong correlation between fast

394

food and obesity in the developed world (Dunn, 2010), future research should be directed to

395

revealing the substitution effects between FAFH and semi-processed food, and food quality of

396

the hotel and restaurant sectors in developing countries.

397

Acknowledgements

398

The authors wish to thank the reviewers and the editor for their suggestions in the preparation of

399

the current manuscript. Mishra’s time on this project was supported by the Kemper and Ethel

400

Marley Foundation.

SC

M AN U

TE D

EP

AC C

401

RI PT

381

402

References

403 404 405

Aguiar, M., and Hurst, E. (2005). Consumption versus Expenditure. Journal of Political Economy, 113(5), 919-948. 18

ACCEPTED MANUSCRIPT

406 407 408 409 410 411

Arabmazar, A., & Schmidt, P.(1981). Further evidence on the robustness of the Tobit estimator to heteroscedasticity. Journal of Econometrics, 17(2), 253-258.

412 413 414 415 416 417 418 419 420 421 422 423 424 425 426 427 428 429 430 431 432 433 434 435 436 437 438 439 440 441 442 443 444 445 446 447 448 449

Aristei,D., Perali. F., & Pieroni, L. (2007). Cohort, age and time effects in alcohol consumption by Italian households: a double-hurdle approach. Empirical Economics, 35(1): 29-61.

RI PT

Arabmazar,A., & Schmidt, P. (1982). An investigation of the robustness of the Tobit estimator to non-normality. Econometrica, 50(4), 1055-1063.

Barker, R., Herdt, R. W., & Rose, B. (1985). The Rice Economy of Asia. Washington, D.C.: Resources for the Future, Inc., The Johns Hopkins University Press. Pp.159-171.

SC

Becker, G. S. (1965). A theory of the allocation of time. The Economic Journal, 75 (299), 493517.

M AN U

Bellante, D., & Foster, A.(1984). Working wives and expenditures on services. Journal of Consumer Research 11(2), 700-707. Burke, W. J. ( 2009). Fitting and interpreting Cragg’s tobit alternative using STATA. The Stata Journal, 9(4): 584-592. Burke, W. J. (2009). Fitting and interpreting Cragg’s tobit alternative using STATA. The Stata Journal, 9(4), 584-592.

TE D

Byrne, P. J., Capps, Jr., O.,& Saha, A.(1996).Analysis of food-away-from-home expenditure patterns for U.S. households, 1982-89. American Journal of Agricultural Economics, 78(3), 614-627.

EP

Byrne, P. J., Capps, Jr., O.,& Saha, A.(1998). Analysis of Quick-serve, Mid-scale, and Up-scale Food Away from Home Expenditures. The International Food and Agribusiness Management Review, 1(1998): 51-72.

AC C

Chang, H., & Mishra, A. K. (2008). Impact of off-farm labor supply on food expenditures of the farm household. Food Policy33(6), 657-664. Cragg, J. G. (1971). Some statistical models for limited dependent variables with application to the demand for durable goods. Econometrica, 39 (5): 829-844. Deaton, A., & Paxson. C. (1998). Economies of scale, household size, and the demand for food. Journal of Political Economy, 106(5), 897-930. Deaton, A., & Zaidi, S. (2002). Guidelines for Constructing Consumption Aggregates for Welfare Analysis. Washington D.C.: The World Bank: 46.

19

ACCEPTED MANUSCRIPT

450 451 452 453 454 455 456 457 458 459 460 461 462 463 464 465 466 467 468 469 470 471 472 473 474 475 476 477 478 479 480 481 482

Dunn, R.A. (2010). The Effect of Fast-Food Availability on Obesity: An Analysis by Gender, Race, and Residential Location. American Journal of Agricultural Economics, 92(4): 1149-1164.

483 484 485 486 487 488 489 490 491 492 493 494

Kearney, J.(2010). Review of food consumption trends and drivers. Philosophical Transactions of the Royal Society, B, Vol. 365 (1554), 2793-2807. Available from URL:http://rstb.royalsocietypublishing.org/content/365/1554/2793.full.pdf+html, [Accessed June 20, 2016].

RI PT

Gahukar, R.T. (2014). Food adulteration and contamination in India: occurrence, implication and safety measures. International Journal of Basic and Applied Research, 3(1): 47-54. Gaiha, R., Jha, R., & Kulkarni, V. S. (2013). How pervasive is eating out in India? Journal of Asian and African Studies, 48, 370-386.

SC

Gan, L., & Vernon, V. (2003). Testing the Barten model of economies of scale in household consumption: toward resolving a paradox of Deaton Paxson. Journal of Political Economy, 111(6), 1361-1377.

M AN U

Government of Bangladesh (GOB). (2012). National Food Policy Plan of Action and Country Investment Plan: Monitoring Report 2012. Dhaka: Ministry of Food and Disaster Management, Government of the Peoples Republic of Bangladesh. Available from URL: http://www.nfpcsp.org/agridrupal/content/national-food-policy-plan-action-and-countryinvestment-plan-monitoring-report-2012, [Accessed July 16, 2016]. Gujarati, D. N. (1995). Basic Econometrics, International 3rd Edition. New York: McGraw Hill Inc.

TE D

Hossain, M. (1998). “Sustaining food security in Asia: economic, social, and political aspects.” in Dowling, N.G., Greenfield, S.M., & Fischer, K.S. (eds.). Sustainability of Rice in the Global Food System. Pp. 19-43. Los Baños (Philippines): International Rice Research Institute (IRRI).

EP

Ingco, M. D. (1991). Is Rice Becoming an Inferior Good: Food Demand in the Philippines. Policy, Research and External Affairs, Working Paper WPS 722. Washington, D.C.: World Bank. Available from URL: http://documents.worldbank.org/curated/en/493481468776045279/pdf/multi-page.pdf, ,

AC C

[Accessed July 29, 2016].

Khan, M. I. (2012). Food adulteration: a review of the prevailing legislation. Dhaka Courier, May 17, Thursday 2012.Available from URL:http://www.dhakacourier.com.bd/?p=6090, [Accessed June 20, 2016]. Kinsey, J.(1983). Working wives and the marginal propensity to consume food away from home. American Journal of Agricultural Economics65 (1):10-19.

20

ACCEPTED MANUSCRIPT

Lee, H., & Tan, A. (2006). Determinants of Malaysian household expenditures of food-awayfrom-home. Contributed paper prepared for presentation at the International Association of Agricultural Economists Conference, Gold Coast, Australia, August 12-18, 2006.Available from URL:http://ageconsearch.umn.edu/bitstream/25430/1/cp060076.pdf, [Accessed June 20, 2016].

SC

RI PT

Liu. H., Wahl. T., Seale, J.L., Jr., Bai, J. (2012). Household composition and food away from home expenditures in urban China. Selected Paper prepared for presentation at the Agricultural & Applied Economics Association’s 2012, AAEA Annual Meeting, Seattle, USA August 12-14, 2012. Available from URL:http://ageconsearch.umn.edu/bitstream/131057/3/Household%20Composition%20a nd%20Food%20Away%20From%20Home%20Expenditures%20in%20Urban%20China. pdf,[Accessed 06, April 2016].

M AN U

McCracken, V. A.,& Brandt, J. S. (1987). Household consumption of food-away-from-home: total expenditure and by type of food facility. American Journal of Agricultural Economics, 69(2), 274-284. Mihalopoulos, V. G., & Demoussis, M. P. (2001). Greek household consumption of food away from home: a microeconometric approach. European Review of Agricultural Economics 28(2),421-432. Moffatt, P. G.(2005).Hurdle models of loan default. The Journal of the Operational Research Society, 56(9):,1063-1071.

TE D

Nasreen, S., & Ahmed, T. (2014). Food Adulteration and Consumer Awareness in Dhaka City, 1995-2011. Journal of Health, Population, and Nutrition, 32(3), 452-464.

EP

Nayga. R.(1996). Wife’s labor force participation and family expenditures for prepared food, food prepared at home, and food away from home. Agricultural and Resource Economics Review 25, 179-186. Pingali, P. (2006). Westernization of Asian diets and the transformation of food systems: implications for research and policy. Food Policy, 32(2006), 281-298.

AC C

495 496 497 498 499 500 501 502 503 504 505 506 507 508 509 510 511 512 513 514 515 516 517 518 519 520 521 522 523 524 525 526 527 528 529 530 531 532 533 534 535 536 537 538 539

Pingali, P., & Khwaja, Y. (2004). “Globalization of Indian Diets and the Transformation of Food Supply Systems.” ESA Working Paper No. 04-05. Agricultural and Development Economics Division, The Food and Agriculture Organization of the United Nations. Available from URL:http://www.eldis.org/vfile/upload/1/document/0708/DOC15357.pdf, [Accessed 06, April 2016]. Prochaska, F.,& Schrimper, R.(1973). Opportunity cost of time and other socioeconomic effects on food-away-from-home consumption. American Journal of Agricultural Economics55(4),595-603.

21

ACCEPTED MANUSCRIPT

Rahman, M.A., Sultan, M.Z., Rahman, M.S., & Rashid, M.A. (2015). Food Adulteration: A Serious Public Health Concern in Bangladesh. Bangladesh Pharmaceutical Journal, 18(1): 1-7.

RI PT

Reardon, T., & Timmer, C. P. (2014). Five Inter-Linked Transformations in the Asian Agrifood Economy: Food Security Implications. Global Food Security. 3(2), 108-117. Redman, B.(1980). The impact of women’s time allocation on expenditure for meal away from home and prepared foods. American Journal of Agricultural Economics62(2), 234-237.

SC

Sabet, D. M. (201. Formalin: How to Address a Market Failure? Dhaka Tribune, June 27, 2013.Available from URL:http://www.dhakatribune.com/longform/2013/jun/27/formalin-how-address-market-failure, [Accessed 06, April 2016].

M AN U

Sexauer, B.(1980). Major changes affecting consumers and consumption. Proceedings of the 29th Annual ACCI conference, Kansas City, Missouri, ed. K. Gould, Columbia, MO: ACCI. Sobhani, M.M. (2015). Food Adulteration: The Bangladesh Paradox. Law Journal Bangladesh, 2 (January-June). Available from URL: http://www.lawjournalbd.com/2015/06/foodadulteration-the-bangladesh-paradox-3/ [Accessed 06, April 2016].

TE D

Stewart, H. Blisard, N., Bhuyan, S., & Nayga, R.M. (2004). “The Demand for Food Away from Home: Full-Service of Fast Food?” Agricultural Economic Report Number 829, Economic Research Service, Washington D.C.: United States Department of Agriculture. Available from URL: http://www.ers.usda.gov/media/306585/aer829_1_.pdf, [Accessed July 29, 2016]. Sudershan, R.V., Rao, P., & Polasa, K. (2009). Food safety research in India: a review. Asian Journal of Food and Agro-Industry, 2(3): 412-433.

EP

United States Department of Agriculture. (2016). Food Expenditures. Washington, D.C.: Economic Research Service, USDA. Available from URL: http://www.ers.usda.gov/data-products/food-expenditures.aspx#26636, [Accessed 06, April 2016].

AC C

540 541 542 543 544 545 546 547 548 549 550 551 552 553 554 555 556 557 558 559 560 561 562 563 564 565 566 567 568 569 570 571 572 573 574 575 576 577 578 579 580 581

World Bank. (2016). The World Development Indicators 2016. Washington, D.C.: World Bank. Available from URL:http://data.worldbank.org/data-catalog/world-developmentindicators, [Accessed 06, April 2016]. Yen, S. T., & Huang, C. L. (1996). Household demand for finfish: a generalized double-hurdle model. Journal of Agricultural and Resource Economics 21 (2), 220-234.

582

22

ACCEPTED MANUSCRIPT

583

Figure 1: Relationship between expenditure on FAFH and total food expenditures, in real BDT.

584 2000

0

589

SC

600 400

0

594

M AN U

593

200

592

Expenditure on FAFH (real BDT)

591

5000

2010

0

590

RI PT

200

400

Expenditure on FAFH (real BDT)

588

0

587

Expenditure on FAFH (real BDT)

586

2005

600

585

5000

Expenditure on food in two weeks (real BDT)

595

Expenditure on FAFH Ga r h p s yb y a e r

596

TE D EP

598

Sources: HIES, 2000, 2005 and 2010

AC C

597

Fitted values

23

ACCEPTED MANUSCRIPT

Figure 2: Estimated coefficients on the likelihood of FAFH and the share of expenditures on FAFH, by year and expenditure quartiles.

601

604 605 606 607 608

0.05 Year 2005

0.04

0.02 0.02

0.002 0 -0.02

-0.03

-0.04 -0.06

-0.06 -0.07

-0.08

Q3

610

Expenditure quartiles

611

Panel A: Estimated coefficients on the likelihood of consumption of FAFH and expenditure quartiles. Note: Expenditure quartile, Q1 is the base group. Source: Authors’ estimation. 0.04

622 623 624 625 626 627 628 629 630 631

Year 2010

-0.01

-0.02 -0.04

-0.02

-0.04

-0.06 -0.08

EP

621

0

Year 2005

AC C

620

Estimated coefficients on share of expenditures on FAFH

619

0.02

TE D

0.019

617 618

Q4

M AN U

Q2

609

612 613 614 615 616

Year 2010

RI PT

603

Estimated coefficients, likelihoood of consumption of FAFH

602

0.06

SC

599 600

-0.09

-0.1

-0.12

-0.13

-0.14

Q2

Q3

Q4

Expenditure quartiles Panel B: Estimated coefficients on the proportion of expenditure of FAFH and income quantiles. Note: Expenditure quartile, Q1 is the base group. Source: Authors’ estimation.

24

ACCEPTED MANUSCRIPT

Table 1: Occupational distribution of total employed population, by major sectors, Bangladesh

Trade, hotels, and restaurants Manufacturing Community and personal services Other a

1999-2000 39.0 19.8 (50.77) 6.1 (15.64) 3.7 (9.49) 5.1 (13.08) 4.3 (11.03)

2002-2003 2005-2006 2010 44.3 47.4 54.1 22.9 22.8 25.7 (51.69) (48.10) (47.50) 6.7 7.8 8.4 (15.12) (16.46) (15.53) 4.3 5.2 6.7 (9.71) (10.97) (12.38) 2.7 2.6 3.4 (6.09) (5.49) (6.28) 7.5 9.1 10.1 (16.93) (19.20) (18.67)

Notes:a The other category includes: mining and quarrying, electricity, gas and water, construction, transport, storage and

EP

TE D

M AN U

communication, finance, business services and real estates, health education, public administration and defense. b Values in parentheses are percentages. Sources: BBS (Bangladesh Bureau of Statistics). 2004. Report on Labor force survey (LFS). 2002-2003 and BBS. 2013. Report on Labor force survey (LFS) 2010.

AC C

633 634 635 636 637 638

1995-1996 34.8 17.0 (48.85) 6.0 (17.24) 3.5 (10.06) 4.8 (13.79) 3.8 (10.92)

RI PT

Economic sectors Total employment (>15, in millions) Agriculture, forestry, and fisheries

SC

632

25

ACCEPTED MANUSCRIPT

Table 2: Performance indicators of the hotel and restaurant businesses, Bangladesh. Year

No. of sampled households owning hotel and restaurants

2005

2010

39

55

63

368.25

Net revenue (000, BDT)

66.27

Monetary value of the present business (000, BDT)

735.08

Years in present business

13.59

Nos. of household members working

1.69

Nos. of salaried workers

1.49

319.21 220.69 91.22

53.73

123.97

64.59

9.76

9.54

1.50

2.62

2.47

1.44

SC

Gross revenue (000, BDT)

EP

TE D

M AN U

Note: All monetary values computed in terms of real BDT using consumer price index, 1995-96 = 100. Sources: Household Income and Expenditure Surveys 2000, 2005, and 2010.

AC C

640 641

2000

RI PT

639

26

ACCEPTED MANUSCRIPT

Table 3: Demographic Information of the sampled households’ by total expenditure (yearly/food and non-food) quartiles.

Stone’s price index household faced

SC

Yearly/per capita total food and non-food expenditure (000,BDT)

M AN U

Yearly/per capita non-food expenditure (000, BDT) Yearly/per capita food expenditure (000, BDT) Years of schooling of the household head

No. of family members % Rural households

AC C

% Spouse works in the non-farm sector

EP

Years of schooling of the spouse

TE D

Age of the household head % Female headed households

All 29,648 3.54 (0.51) 11.27 (10.40) 5.99 (8.98) 5.28 (2.65) 3.74 (4.45) 45.40 (13.63) 11.5 (68.04) 2.96 (3.91) 7.39 (26.16) 4.81 (2.05) 69.03 (46.24)

Q1 7,412 3.13 (0.39) 4.68 (0.092) 1.62 (0.07) 3.06 (0.07) 1.80 (3.23) 43.69 (13.20) 8.57 (72.01) 1.30 (2.63) 6.25 (24.20) 5.37 (2.05) 81.62 (38.73)

RI PT

No. of observations

Expenditure quartiles Q2 Q3 7,412 3.45 (0.41) 7.24 (0.07) 2.92 (0.09) 4.33 (0.09) 2.58 (3.75) 45.11 (13.76) 9.96 (70.06) 2.03 (3.19) 6.06 (23.86) 4.91 (2.03) 75.15 (43.22)

7,412 3.67 (0.42) 10.46 (1.21) 4.79 (1.54) 5.67 (1.43) 3.96 (4.34) 45.91 (13.88) 12.30 (67.15) 3.17 (3.83) 7.53 (26.39) 4.68 (2.03) 67.98 (46.46)

Q4 7,412 3.91 (0.45) 22.70 (15.46) 14.64 (14.62) 8.06 (3.33) 6.60 (4.76) 46.88 (13.48) 15.35 (63.95) 5.35 (4.48) 9.71 (29.62) 4.31 (1.93) 51.36 (49.98)

Note:All monetary values are computed in terms of real BDT using consumer price index 1995-96=100. Sources: Household Income and Expenditure Surveys 2000, 2005, and 2010.

27

ACCEPTED MANUSCRIPT

Table 4: Patterns of food and non-food expenditures including expenditure on FAFH, by the sampled years and income quartiles.

Yearly/per capita food expenditure (000, BDT) Yearly total expenditure on FAFH (000, BDT) Proportion of total food expenditure on FAFH (expenditure on FAFH/Total food expenditure)

Q4 1,273 71.17 (45.31) 23.41 (16.57) 16.37 (16.62) 7.05 (3.84) 1.72 (2.92) 5.41 (8.09)

Q1 3,063 56.22 (49.62) 4.71 (0.88) 1.58 (0.56) 3.13 (0.69) 0.44 (0.73) 2.63 (4.58)

2005 Q2 Q3 2,768 2,202 72.51 76.52 (44.66) (42.40) 7.18 10.39 (0.72) (12.19) 2.77 4.65 (0.84) (1.40) 4.41 5.74 (0.81) (1.25) 0.88 1.16 (1.21) (1.50) 4.26 4.49 (6.07) (6.38)

RI PT

2000 Q2 Q3 1,766 1,438 65.18 71.07 (47.65) (45.36) 7.18 10.38 (0.74) (11.99) 2.86 4.80 (1.22) (2.10) 4.32 5.58 (1.20) (1.95) 0.93 1.32 (1.48) (2.03) 3.56 4.49 (4.92) (6.38)

Q1 2,930 51.47 (49.99) 4.46 (0.97) 1.50 (0.76) 2.97 (0.84) 0.43 (0.83) 2.22 (3.82)

SC

% households reporting expenditures on FAFH Yearly/per capita total food and non-food expenditure (000,BDT) Yearly/per capita non-food expenditure (000, BDT)

All 29.648 63.45 (48.16) 11.27 (10.400) 5.99 (8.98) 5.28 (2.65) 0.88 (1.69) 3.45 (5.64)

M AN U

Year Income quartiles No. of observations

AC C

EP

TE D

Note: All monetary values are computed in terms of real BDT using consumer price index, 1995-96 = 100. Sources: Household Income and Expenditure Surveys 2000, 2005, and 2010.

28

Q4 2,021 74.22 (43.75) 22.93 (15.71) 14.44 (14.35) 8.49 (3.74) 1.42 (2.04) 4.06 (5.60)

Q1 1,419 46.51 (49.90) 5.06 (0.74) 1.94 (0.57) 3.12 (0.61) 0.33 (0.63) 2.10 (3.86)

2010 Q2 Q3 2,878 3,772 56.39 61.45 (49.60) (48.68) 7.34 10.52 (0.74) (1.21) 3.09 4.87 (0.84) (1.35) 4.26 5.65 (0.80) (1.27) 0.57 0.87 (1.06) (1.64) 2.85 3.39 (5.10) (5.94)

Q4 4,118 65.83 (47.43) 22.37 (14.96) 14.21 (14.04) 8.16 (2.84) 1.12 (2.48) 3.43 (59.5)

ACCEPTED MANUSCRIPT

Table 5: The impact of female members’ non-farm employment on FAFH. Status of any female member working in non-farm sector

No. of observations

27,458

67.08

873.87

M AN U

Real expenditure on FAFH(yearly/BDT)

Yearly per capita total food expenditure (BDT)

Share of expenditure on FAFHb

63.17

2,190

SC

% Households reporting expenditures on FAFH

Yes (B)

RI PT

No (A)

5248.05

3.38

1018.94

5659.95

TE D

EP AC C

-3.91*** (-3.66)

-145.07*** (-3.86)

-411.89*** (6.99) -0.78***

4.17

Note: All monetary values are computed in terms of real BDT using consumer price index, 1995-96 = 100. a Differences = Mean (A) – Mean (B). H0: Diff = 0 , Ha: Diff < 0 (one-sided t-test) b Defined as a share of expenditures on FAFH to total food expenditures.

29

Mean differences and the level of significance a

(-6.22)

ACCEPTED MANUSCRIPT

Table 6: Maximum likelihood estimates of the generalized double-hurdle model estimation procedure explaining the choice and the proportion of expenditure on FAFH.

Age, household head (α4, 4) Female-headed household dummy (yes=1) (α5, 5) Years of schooling, spouse (α6, 6)

Urban household dummy (yes=1) (α8, 8) Chittagong division dummy (yes=1) (α9, 9) Dhaka division dummy (yes=1) (α10, 10)

AC C

Khulna division dummy (yes=1) (α11, 11)

EP

No. of family members (α7, 7)

TE D

Spouse was employed in non-farm sector (dummy yes=1) (α7, 7)

Rajshahi division dummy (yes=1) (α12, 12) Rangpur division dummy (yes=1) (α13, 14) Sylhet division dummy (yes=1) (α14, 14) Year 2005 dummy (yes=1) (α15, 15) Year 2010 dummy (yes=1) (α16, 16)

RI PT

Yearly per capita total food and non-food expenditure (000, BDT) (α2, 2) Years of schooling, household head (α3, 3)

SC

Stone’s price index for all food items consumed (α1, 1)

Proportion of expenditure on FAFH to total food expenditure(P) 0.64*** (0.12) -0.047** (0.02) -0.01 (0.01) -0.01*** (0.00) 0.01 (0.07) -0.07*** (0.02) 0.07 (0.05) -0.17*** (0.04) -0.46*** (0.10) 0.34*** (0.09) -0.14** (0.06) -0.44*** (0.11) -0.27*** (0.09) 0.24*** (0.07) -0.61*** (0.17) -0.15** (0.07) -0.48***

M AN U

Independent variables

Dummy for positive expenditure on FAFH ((αi)) 0.49*** (0.03) -0.001 (0.00) -0.01*** (0.00) -0.04*** (0.00) -0.60*** (0.03) -0.0002 (0.00) 0.16*** (0.03) 0.06*** (0.00) -0.09*** (0.02) 0.45*** (0.03) -0.34*** (0.03) 0.29*** (0.03) -0.06* (0.03) 0.62*** (0.04) -0.37*** (0.04) 0.03 (0.03) -0.40***

30

Heteroscedasticity ( )

0.004** (0.00) -0.001 (0.00)

0.01*** (0.00)

ACCEPTED MANUSCRIPT

Wealthy group 4 (α19, 19) Wealthy group 2 X Year 2005 dummy (α20, 20) Wealthy group 3 X Year 2005 dummy (α21, 21) Wealthy group 4 X Year 2005 dummy (α22, 22)

M AN U

Wealthy group 2 X Year 2010 dummy (α23, 23)

RI PT

Wealthy group 3 (α18, 18)

Wealthy group 3 X Year 2010 dummy (α24, 24) Wealthy group 4 X Year 2010 dummy (α25, 25) Constant (α0, 0)

(0.13) 0.16** (0.08) 0.17* (0.09) 0.36*** (0.13) 0.019 (0.04) -0.01 (0.03) -0.09*** (0.03) -0.04 (0.05) -0.02 (0.04) -0.13*** (0.04) -1.88*** (0.49)

SC

(0.05) 0.36*** (0.04) 0.52*** (0.05) 0.54*** (0.05) 0.05* (0.03) 0.02 (0.02) 0.002 (0.02) -0.07** (0.03) -0.06*** (0.02) -0.03* (0.02) -2.13*** (0.11)

Wealthy group 2 (α17, 17)

0.27*** (0.04)

AC C

EP

TE D

Observations Log pseudo-likelihood Sigma Wald Chi2 (26) Prob>Chi2 Note: All monetary values are computed in terms of real BDT using consumer price index 1995-96=100. Numbers in parentheses are standard errors calculated based on the intragroup correlation. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

31

ACCEPTED MANUSCRIPT

Table 7: Estimated function explaining proportion of expenditure on FAFH, two-limit-Tobit, and truncated regression.

Stone’s price index for all food items consumed (α1, 1) Yearly per capita total food and non-food expenditure (000, BDT) (α2, 2)

SC

Years of schooling, household head (α3, 3)

Female-headed household dummy (yes=1) (α5, 5) Years of schooling, spouse (α6, 6) Spouse was employed in non-farm sector (dummy yes=1) (α7, 7)

Dhaka division dummy (yes=1) (α10, 10) Khulna division dummy (yes=1) (α11, 11)

AC C

Rajshahi division dummy (yes=1) (α12, 12)

EP

Chittagong division dummy (yes=1) (α9, 9)

TE D

No. of family members (α7, 7)

M AN U

Age, household head (α4, 4)

Urban household dummy (yes=1) (α8, 8)

Rangpur division dummy (yes=1) (α13, 14) Sylhet division dummy (yes=1) (α14, 14) Year 2005 dummy (yes=1) (α15, 15)

2LT TR Proportion of expenditure on FAFH (expenditure on FAFH/Total food expenditure) 0.05*** 0.72*** (0.00) (0.15) -0.0002*** -0.04* (0.00) (0.00) -0.001*** -0.02** (0.00) (0.01) -0.0003*** -0.01*** (0.00) (0.00) -0.03*** -0.06 (0.00) (0.07) -0.001*** -0.02*** (0.00) (0.01) 0.01*** 0.07 (0.00) (0.06) -0.001*** -0.22*** (0.00) (0.05) -0.01*** -0.55*** (0.00) (0.13) 0.03*** 0.44*** (0.00) (0.11) -0.02*** -0.16** (0.00) (0.08) 0.01*** -0.49*** (0.00) (0.13) -0.01*** -0.30*** (0.00) (0.11) 0.03*** 0.31*** (0.00) (0.10) -0.03*** -0.73*** (0.00) (0.22) -0.01*** -0.17 (0.00) (0.10) -0.04*** -0.57*** (0.00) (0.18) 0.02*** 0.30**

RI PT

Models Dependent variable

Year 2010 dummy (yes=1) (α16, 16) Wealthy group 2 (α17, 17)

32

ACCEPTED MANUSCRIPT

Wealthy group 3 (α18, 18)

RI PT

Wealthy group 4 (α19, 19)

(0.00) 0.02*** (0.00) 0.03*** (0.00) 0.003** (0.00) 0.003 (0.00) -0.003*** (0.00) -0.003* (0.00) -0.003** (0.00) -0.004*** (0.00) -0.14*** (0.01) 29648 15158.90 0.07*** (0.00)

Wealthy group 2 X Year 2005 dummy (α20, 20) Wealthy group 3 X Year 2005 dummy (α21, 21)

SC

Wealthy group 4 X Year 2005 dummy (α22, 22) Wealthy group 2 X Year 2010 dummy (α23, 23)

M AN U

Wealthy group 3 X Year 2010 dummy (α24, 24) Wealthy group 4 X Year 2010 dummy (α25, 25) Constant (α0, 0)

TE D

Observations Log pseudo-likelihood Sigma

(0.12) 0.30** (0.13) 0.37*** (0.14) 0.07 (0.06) -0.02 (0.04) -0.09** (0.04) -0.07 (0.08) -0.04 (0.05) -0.12*** (0.04) -2.87*** (0.65) 18,813 37027.31 0.35*** (0.04) 29.07 0.31

AC C

EP

Wald Chi2 (26) Prob>Chi2 F (26, 29622) 154.18 Prob>F 0.00 Left-censored observations 10,835 10,835 Uncensored observations 18,812 Right censored observations 1 Note: All monetary values are computed in terms of real BDT using consumer price index 1995-96=100. Numbers in parentheses are standard errors calculated based on the intragroup correlation. *Significant at the 10% level; **significant at the 5% level; ***significant at the 1% level.

33

ACCEPTED MANUSCRIPT

Highlights

RI PT

SC M AN U TE D EP

• •

Food away from home is an established phenomenon in developed countries There is a paucity of studies in the developing countries on this issue Using Bangladesh as a case this study examines food away from home in developing countries Rich, urban and educated households are less like to consume food away from home Rampant food adulteration problem may be influencing such decision by the households

AC C

• • •