12 Round I survey of CFPR phase II

Food Consumption Pattern and

Dietary Diversity
Chowdhury SB Jalal, Nuzhat Choudhury and Munshi Suliman

INTRODUCTION

This chapter provides information about the quantities of food items consumed
and the dietary diversity of the survey households. Household food consumption
has been defined as the total amount of food available for consumption in the
household, generally excluding the food taken outside unless prepared at home
(Klaver, Knuiman et al. 1982). It serves as a direct indicator of food security as
well as a distal proxy for a poverty indicator (WFP 2007). Research in
developing countries show that as income increases, the poorest households
spend a major share of their additional income on food expenses. This increase in
the food budget resulting from rise in income is manifested by increased quantity
as well as improved quality of the food (Subramanian and Deaton 1996). Further
the share of food expenses increase in their budget, more it is characterized by
the diversity in the type of food they acquire and consume, although not
necessarily altering their calorie intakes (Behrman and Deolalikar 1989).

Dietary diversity is the sum of the number of different food groups consumed
over a given reference period (Hoddinott and Yohannes 2002). It is considered as
a proxy to household food security. Diversity in diet is an important outcome in
and of itself. A more diversified diet is associated with a number of improved
outcomes in areas such as, birth weight, child anthropometric status, and
improved hemoglobin concentrations (Swindale and Bilinsky 2006). Diversity in
the diet is highly correlated with factors such as caloric and protein adequacy,
percentage of protein from animal sources, and household income.

199
 Food consumption pattern and dietary diversity

Information about food consumption and diversity in diet is important from the
programmatic point of view as it has the potential to be used to effectively
change, modify or improve programme activities. Understanding the baseline
consumption pattern of the poorest of the poor households and the extent of their
dietary diversity is important to assess the impact of the programme in terms of
poverty alleviation as well as improvement in their food security, and health and
nutritional wellbeing. Also, it will help design policies or programmes targeting
specific population which depends on geographical or household characteristics.

METHODOLOGY

The STUP baseline survey was conducted on 29,140 households from 19 districts
of Bangladesh. This research, however, included households that only had
complete dietary information. Therefore, after necessary cleaning of the data,
only 21,868 households were finally included in this study, of which 18,956
households were from the STUP I areas while the remaining 2,912 households
from the STUP II areas. A structured questionnaire, based on the three-day recall
method was applied to gather dietary information. Data was collected from the
female members of the households, who are usually more informed about food
purchases, intra-household food allocation, cooking and child feeding. The
respondents were asked to recall all food items that they had consumed within
the last three days prior to the interview. A checklist of food items was used by
the enumerators to help the respondents recall the names and amount of the food
consumed. The checklist also helped them calculate the number of household
members who had eaten during those days.

Per capita calorie consumption was derived by dividing the total household
consumption of three days by the number of persons (including guests) in that
household for the same time. The quantity of food consumed at household level
was first estimated in household measures (i.e., cup, spoon, bowl etc). The
enumerators then converted those measures into their raw weight in grams. The
amounts of ingredients of cooked food were calculated using a conversion Table
that had been provided to the enumerator. The food items were pooled into six
basic groups for programmatic use as shown in Figure 2. Conversion Factors
(CF) were not used during conversion of amount in grams to calorie for any of
the food items. Considering this overestimated the actual calorie derived from the
food by 10%, the analysis was done after the adjustment for this increment
(Gibson 2005).

The number of persons per day was calculated based on the number of persons
who ate at least one meal during any specific day. The total number of persons
having a meal in each day was then compared to the other two days and a
maximum variability of 3 persons between any of the three days was only
considered in the analysis. To standardize the consumption of individuals within

200
Round I survey of CFPR phase II

households, all children below age 10 years were weighted 0.5 to convert them to
adult equivalent (Gibson 2005; BBS 2007).

The enumerators also recorded the amount of money spent on the food consumed
during the 3 days prior to data collection. In terms of food produced, received in
kind or collected otherwise, where the households did not have to spend money,
the expense equivalent for that food item was calculated and used in the analysis.
The food expenditure was calculated based on the local market price for the food.

Two dietary diversity scores were used in the analysis based on food groups. The
first was based on the six basic food groups, i.e., cereal, pulse, vegetables, fruits,
animal products, and oil. The second was based on more diverse food groups
created by separating leafy vegetables from ‘vegetables’, and splitting animal
products into meat, fish, egg and milk as suggested by Helen Keller International,
Bangladesh and Household Dietary Diversity Indicator Guide (Damton-Hill,
Hassan et al. 1988; Swindale and Bilinsky 2006.

All analyses for this chapter have been done using STATA version 9 and SPSS
WIN version 15.

RESULTS

As expected in the context of rural Bangladesh, the total amount (g) of food
intake was significantly higher (p<0.001 and p=0.034) in economically better-off
households compared to the poorer households (NTP vs. TUP) within same
STUP areas, as well as between households (TUP vs. TUP) (p<0.001) of the two
areas (Table 1). Further, the between-area comparison shows that the total
amount (g) of food consumed by the households of STUP I areas was
significantly higher (p<0.001) than the amount consumed by households of
STUP II areas. On average about two thirds (62.3%) of the amount consumed by
the households had been from cereal-based foods. Within the same area, the
share of cereal-based food in the diet (percentage of total intake) was
significantly higher (p<0.001 and p=0.028) in TUP households than that of the
NTP households. However, there was no difference found between households of
the two areas. Marked differences (p<0.001) were observed in the amount intake
from pulses between TUP households of the two areas (e.g., 3 vs. 6). Significant
(p<0.001) differences were also observed between same categories of households
from STUP I and STUP II areas.

The intake of vegetables, which included potato and other roots and tubers, was
higher in the non-poor households compared to the poorer households (Table 1).
Contrary to the general trend in consumption, however, intake of green leafy
vegetables (shak) was higher in the TUP households compared to the NTP
households (p<0.001). The average amount (83.6g) of animal products consumed

201
 Food consumption pattern and dietary diversity

by the survey households is identical to national average intake (88.3g) (BBS

2007).

Table 1. Mean per capita per day quantity of food intake (g) by the survey
households

Variable STUP I STUP II p values

NP NTP TUP NP NTP TUP 2 vs. 3 5 vs. 6 3 vs. 6
(1) (2) (3) (4) (5) (6)
Total (g) 999.9 921.1 868.8 908.3 849.1 811.5 <0.001 0.034 <0.001
Cereal (g) 587.7 577.6 560.3 538.2 539.5 528.1 <0.001 ns <0.001
Pulses (g) 12.4 9.6 10.6 22.3 17.8 21.8 ns ns <0.001
Vegetables 198.2 194.3 189.3 165.3 158.4 151.2 ns ns <0.001
Leafy (g) 40.2 49.2 61.5 25.5 30.1 33.6 <0.001 ns <0.001
Others (g)* 158.0 145.1 127.8 139.8 128.3 117.6 <0.001 ns 0.033
Fruit (g) 55.2 41.2 30.3 45.0 30.0 22.4 <0.001 ns ns
Animal product 111.1 70.6 50.9 93.6 68.6 57.9 <0.001 ns ns
Fish (g) 52.7 39.7 33.7 54.5 43.3 38.8 <0.001 ns ns
Meat (g) 11.9 5.7 3.4 13.9 11.6 8.3 0.004 ns 0.017
Egg (g) 4.6 2.7 2.6 6 4.1 4.2 ns ns 0.005
Milk (g) 41.9 22.5 11.1 19.1 9.6 6.6 <0.001 ns ns
Oil (g) 13.3 10.7 10.3 18.4 15.2 14.4 ns ns <0.001
Others (g) 21.9 17.1 17.1 25.5 19.4 15.6 ns ns ns
% from cereal 58.8 62.7 64.5 59.3 63.5 65.1 <0.001 0.028 ns
n 4268 8922 5766 998 1268 646
*includes potato
ns: Not significant at the 5% level

Consistent with the amount of food consumption shown in Table 1, the total
calorie intake within STUP areas was higher in economically better-off
households compared to the poorer households (i.e., NP vs. NTP, NTP vs. TUP
etc.) (Table 2). This trend, however, is not consistent in consumption of other
types of food. The households of STUP I areas in general, consumed
significantly (p<0.001) more calories (2264.1 Kcal) than the households of STUP
II (2203.1 Kcal) areas. On average, about four-fifth (80.1%) of the calories
consumed by the households were from cereal-based foods. Within the same
area, poorer households gained more percentage of energy from cereal-based
foods compared to the economically better-off households.

Calorie intake from vegetables including potato and other roots and tubers, was
significantly (p<0.001) higher in TUP from STUP I compared to STUP II
households. In contrary to the general trend in Table 2, calorie intake from green
leafy vegetables (shak) was higher in the TUP households.

202
Round I survey of CFPR phase II

The calorie consumed from oil and animal products are significantly higher
(p<0.001 and p<0.01 respectively) in households of STUP II areas compared to
households of STUP I areas. The calories consumed from oil in households of
STUP II areas is about 40% higher than that of STUP I households, however, the
consumption of animal products was only found to be 10% higher.

Table 2. Mean per capita per day calorie intake (Kcal) by the survey
households

Variable STUP I STUP II p values

NP NTP TUP NP NTP TUP 2 vs. 3 5 vs. 6 3 vs. 6
(1) (2) (3) (4) (5) (6)
Total (Kcal) 2346.4 2210.6 2126.9 2254.4 2159.5 2095.6 <0.001 ns ns
Cereal (Kcal) 1881.9 1849.9 1794.8 1722.6 1727.4 1690.6 <0.001 ns <0.001
Pulse (Kcal) 38.5 29.7 32.9 69.1 55.2 67.9 ns ns <0.001
Vegetables 114.4 107.0 102.1 98.4 93.3 88.6 <0.001 ns <0.001
Leafy (Kcal) 15.4 19.6 25.1 9.4 11.0 12.9 <0.001 ns <0.001
Others
99.0 87.4 77.0 89.0 82.3 75.7 <0.001 ns ns
(Kcal)*
Fruit (Kcal) 36.1 25.7 19.2 30.6 19.9 16.0 <0.001 ns ns
Animal product 104.0 68.3 53.1 103 76.6 64.1 <0.001 0.043 0.037
Fish (Kcal) 57.0 43.7 37.8 66.0 51.2 44.4 <0.001 ns ns
Meat (Kcal) 12.3 5.8 3.5 14.2 12.0 8.5 0.005 ns 0.023
Egg (Kcal) 7.3 4.3 4.1 9.6 6.6 6.8 ns ns 0.004
Milk (Kcal) 27.4 14.5 7.6 13.0 6.8 4.4 <0.001 ns ns
Oil (Kcal) 108.1 86.8 84.3 148.9 123.2 116.6 ns ns <0.001
Others (Kcal) 63.1 42.9 39.4 77.7 59.7 46.0 ns ns ns
% from cereal 80.2 83.7 84.4 76.4 80.0 80.7 <0.001 ns <0.001
n 4268 8922 5766 998 1268 646
*includes potato
ns: Not significant at the 5% level

The mean food expenditure of the households from STUP II areas (Tk. 27.23) is
significantly (p<0.001) higher than that of households of the STUP I areas (Tk.
24.04) (Table 3). The within-area NTP-TUP difference (p<0.001 and p=0.013)
and the between-area difference among TUP households (p<0.001) was also
found to be significant. Although, the total amount of per capita calorie gained
from cereal-based foods is almost 80% over the areas, the households only spent
about half (54.7%) of their food expenses on cereal-based food.

The amount spent on purchasing fish by households across areas was about half
of the total amount spent on animal products. This does not necessarily indicate

203
 Food consumption pattern and dietary diversity

the preference of the households for fish over other animal products, but perhaps
highlights the wider availability of fish in villages.

Table 3. Mean per capita per day food expenditure (Taka) by survey
households

STUP I STUP II p values

NP NTP TUP NP NTP TUP 2 vs. 3 5 vs. 6 3 vs. 6
(1) (2) (3) (4) (5) (6)
Total (Tk.) 26.52 22.35 20.40 30.17 25.52 23.96 <0.001 0.013 <0.001
Cereal (Tk.) 12.79 12.35 11.97 13.68 13.57 13.18 <0.001 ns <0.001
Pulse (Tk.) 0.65 0.49 0.59 1.07 0.88 1.15 <0.001 0.053 <0.001
Vegetables 2.82 2.42 2.1 2.72 2.25 2.12 <0.001 ns ns
Leafy (Tk.) 0.26 0.31 0.37 0.31 0.33 0.34 <0.001 ns ns
Others (Tk.)* 2.56 2.11 1.73 2.41 1.92 1.78 <0.001 ns ns
Fruit (Tk.) 0.99 0.60 0.45 0.98 0.46 0.41 <0.001 ns ns
Animal
6.48 4.18 3.12 7.23 5.4 4.22 <0.001 ns ns
product
Fish (Tk.) 3.62 2.68 2.16 4.22 2.98 2.41 <0.001 0.041 ns
Meat (Tk.) 1.55 0.77 0.47 1.80 1.72 1.13 0.014 ns 0.045
Egg (Tk.) 0.47 0.29 0.27 0.61 0.39 0.44 ns ns <0.001
Milk (Tk.) 0.84 0.44 0.22 0.60 0.31 0.24 <0.001 ns ns
Oil (Tk.) 1.15 0.94 0.94 2.55 1.33 1.36 ns ns <0.001
Others (Tk.) 1.64 1.37 1.23 1.94 1.63 1.52 <0.001 0.036 ns
% from cereal 48.2 55.3 58.7 46.9 53.2 55.3 <0.001 0.014 0.011
n 4268 8922 5766 998 1268 646
*includes potato
ns: Not significant at the 5% level

COMPARISON WITH NATIONAL DATA

We compared the amount of food consumed by households of the STUP areas

with the national rural consumption as reported by Household Income and
Expenditure Survey (HIES) by Bangladesh Bureau of Statistics (BBS 2007). The
method suggested by HKI was followed in defining ‘vegetables’ and ‘other’ type
of foods which may have varied from the definition that has been used in HIES.
We therefore, excluded these two food groups from comparing with the HIES
data.

The total amount of food intake by households of the STUP I areas (952.0g) was
higher than that of national rural consumption (946.3g). The mean intake of
households of STUP II areas (876.3g), however, was much lower than the
national rural mean. The share of cereal-based foods is much higher in the

204
Round I survey of CFPR phase II

households of the STUP areas as percentage of total intake (62.3% vs. 51.3%).
The STUP II households consumed a greater amount of pulses, fruits, and oil
compared to the national rural average.

Table 4. Comparison of per capita mean amount of food intake of the survey
households with Household Income and Expenditure Survey

STUP I STUP II Total 2005 HIES Rural

Total (g) 952.0 876.3 914.2 946.3
Cereals (g) 580.8 538.5 559.7 485.6
Pulse (g) 10.9 20.1 15.5 12.7
Vegetables (g)* 195.7 161.5 178.6 218.4
Fruit (g) 46.6 37.0 41.8 32.4
Animal product (g) 86.9 80.2 83.6 88.3
Oil (g) 11.9 16.7 14.3 14.3
Others (g) 19.2 22.3 20.7 94.6
% from cereal 62.2 62.5 62.3 51.3
*includes potato

The amount of major food groups (i.e., cereal, pulse, vegetables, fruits, animal
product, and cooking oil) consumed by the survey households have been
compared to the recommended intake1 for a Bangladeshi individual (Figure 1 and
Table 5). Expectedly, the average total amount (i.e., 891.2 g/person/day)
consumed from the major food groups by all categories of households is much
lower compared to the recommended intake (i.e., 1015 g/person/day). The quality
of their diet is also compromised by adding more of cheaper cereal-based foods
to achieve fulfilling volumes of food. The bulk of the cereal-based food took
shares of the other food groups, further compromising a balanced diet that is
important for a healthy life.

1
National Food Policy Capacity Strengthening Project, Food Planning and Monitoring Unit
Ministry of Flood and Disaster Management/ Ministry of Agriculture, Government of Bangladesh
expert consultation, August 2007.

205
 Food consumption pattern and dietary diversity

Figure 1. Consumption (gram) of selected food groups of the survey

households compared to the recommended intake for Bangladeshi
individuals

1200.00 Food items

Cereal
Pulse
1000.00 Veg and potato
Fruits
gm / person / day

Animal product
800.00 Cooking oil

600.00

400.00

200.00

0.00
NP1 NTP1 STUP1 NP2 NTP2 STUP2 Desirable
intake
2007
Household category

To achieve the recommended intake, consumption of cereal-based foods need to

be reduced to about two-thirds (67.7%) of the current consumption of the STUP
households (Figure 1 and Table 5). Conversely, the amount consumed from all
other food groups need to be increased by varying degrees. Pulse consumption
should be increased by four times, while vegetables by 1.5 times of the current
intake. In addition, as much as three times more fruits, animal products, and oil
need to be added to the diets of the survey households to achieve the
recommended intake amount.

Table 5. Consumption (gram) of selected food groups of the survey

households compared to the recommended intake for Bangladeshi
individuals

STUP I STUP II Desirable intake

NP NTP TUP NP NTP TUP 2007
Cereals (g) 587.7 577.5 560.2 538.2 539.5 528.1 375
Pulse (g) 12.4 9.6 10.6 22.3 17.8 21.8 60
Vegetables* (g) 198.2 194.3 189.3 165.3 158.4 151.2 260
Fruit (g) 55.2 41.2 30.3 45.0 30.2 22.5 100
Animal pro 111.1 70.6 50.9 93.6 68.6 57.9 180
Oil (g) 13.4 10.7 10.4 18.4 15.2 14.4 40
Total 978 903.9 851.7 882.8 829.7 795.9 1015
*includes potato

206
Round I survey of CFPR phase II

Similar to the calorie consumption pattern, we observed much the same

monotonic trend in dietary diversity within households of the STUP I and STUP
II areas. In general, the households of STUP I areas consumed fewer varieties of
food compared to the households of STUP II areas (Figure 2). The differences
between NTP and TUP households within the same areas were also pronounced.
Across STUP I and STUP II areas, about two-thirds (66.8% and 69.5%) of the
non-poor households and half (44.3% and 54.0%) of the poorer households
consumed 5 or more major food groups. More than a fifth (23.4% and 21.9%) of
the non-poor households and about a tenth of the poorer households (8.9% and
10.5%) in both areas consumed all 6 food groups.
Figure 2. Proportion of households achieving dietary diversity (six food
groups)

HH category
50.00
NP1
NTP1
TUP1
NP2
NTP2
40.00 TUP2
Percent HH with food diversity

30.00

20.00

10.00

0.00
3 or less types of food 4 types of food 5 types of food 6 types of food
DIVERSITY

We explored the food groups that lacked most in achieving a completely diverse
diet covering the major (i.e., macro-nutrient) six food groups (Figure 3). Results
show that fruits and pulses were the food groups that were most deficient in all
types of households. Very small percentages of households were found to be
lacking oil or animal protein in their diet, while only few were deficient in
vegetables. As expected, none of the households were found deficient in cereal-
based food in their diet. More than half (ranging from 56.2% to 80.3%) of all
household categories fell short of only fruits in their diets across areas with more
percentage of STUP II households lacking fruits compared to the STUP I
households. The NTP-TUP household differences in both areas were also found
significant (p<0.01) in terms of deficiency only in fruits to achieve a completely

207
 Food consumption pattern and dietary diversity

diverse diet. Conversely, fewer STUP II households lacked pulse in their diet
compared to the STUP I households. Similar NTP-TUP differences were
observed in both areas although in a reverse direction.
Figure 3. Proportion of household by categories lacking only one group of
food in achieving diversity in their diet (six groups)

100.00
Food group
Pulse
Fruits
Oil
Animal product
80.00
Percent lacking

60.00

40.00

20.00

0.00
NP-1 NTP-1 TUP-1 NP-2 NTP-2 TUP-2
HH category

Further exploration of dietary diversity shows that about 80% of all households
consumed six or more out of 12 food groups (as mentioned in the methodology
section). Only a small percentage (0.7%) of households consumed all food
groups. These household were therefore, not included in the graphical
presentation. The percentage of households decreased with the increase in dietary
diversity score.

208
Round I survey of CFPR phase II

Figure 4. Proportion of households achieving dietary diversity (12 food

groups)

100.00 Category
NP1
NTP1
STUP1
NP2
80.00 NTP2
Percent HH with dietary diversity

STUP2

60.00

40.00

20.00

0.00

DDS <5 DDS <6 DDS <7 DDS <8 DDS <9 DDS <10
Items

DISCUSSION AND CONCLUSION

The major purpose of this study was to create a benchmark profile of the TUP
households to evaluate the impact of the programme after a certain period of
intervention. Another aim was to suggest the programme implementers on
specific issues for strengthening the intervention components. We focused on
four key findings of the study. First, the food consumed by the TUP households
was much lower in amount (lower than the mean national rural intake) compared
to the recommended intake for Bangladeshis as set by the national expert
committee. Second, although the households of the STUP II areas consumed
lesser calories, their food expenditure was significantly higher than those of the
STUP I areas.. Third, in both the areas, percentage of calorie intake from cereal-
based foods was much higher than the recommended intake and the national
average. Fourth, the diet of the TUP households was far from reaching the
desirable diversity in major food groups.

209
 Food consumption pattern and dietary diversity

Within the same area, the higher calorie intake of the non-poor (NP) households
compared to the poorer (TUP) households supports findings from other study that
the calorie intake increases with a rise in income in the developing countries
(Subramanian and Deaton 1996). On the other hand, the households of STUP II
areas consumed less calories than households of the STUP I areas, although they
spent more money in buying food. This suggests that the households of the
economically better-off areas (i.e., STUP II) may have had consumed relatively
higher priced food items such as pulse, fish, meat, and egg, which added quality
in their diet but not necessarily increased the total calorie intake (Behrman and
Deolalikar 1989; BBS 2004). Adding these non-inferior food items, however,
improved the diversity of their diet. This phenomenon is also consistent with the
characteristics of the households from poorer socioeconomic areas where the
cheaper cereal-based food adds to the bulk of the food volume, thus fulfilling the
demand for adequate meals. In households of the STUP areas, increased share of
income was used to buy relatively higher priced food. Such increased calorie
consumption of oil and animal product may have been also due to the relief
packages that composed mainly of rice, pulse, and oil.

The difference between STUP areas in the consumption of pulses, animal

protein, and oil may have been due to the variation in the availability of animal
protein and pulses in the southern areas. The southern districts may have a
different food culture with higher dependency on pulses compared to the
northern districts. Another likely explanation could be based on the composition
of food relief that had been provided to the households affected by hurricane Sidr
in some of the southern districts (STUP II areas) of Bangladesh. The high amount
of pulses and oil that had been provided to the households to meet their protein
needs may have allowed the households to spare money to buy more animal
foods which leveraged the consumption of quality proteins by households of
STUP II areas. It is likely that the income generating activities of the program
particularly relating to poultry, goat and cow rearing also have the potential to
improve the quality of diets of the ultra poor. Efforts should continue, to find out
way to sustainably include animal protein in the diets of TUP households not
involved in IGA related to poultry and livestock. At the same time the pulse
supplementation currently provided to the TUP households should also continue
as it has the potential to improve the quality as well as add more diversity to their
diets.

210