s- s ss 4 / o%f - world bank · 2/23/1981 · the income concept used in the survey is broadly...
TRANSCRIPT
s- s SS 4 / o%f
DRAFT
EPD INCOME DISTRIBUTION PROJECT
DATA ON INCOME DISTRIBUTION IN BANGLADESH
Satish Kansal
Division Working Paper No. 1981-8July 1981
Economic and Social Data DivisionEconomic Analysis and Projections DepartmeutDevelopment Policy StaffThe World Bank
Division Working Papers report on work in progress and are circulatedfor Bank staff use to stimulate discussion and comment. The viewsand interpretations in a Working Paper are those of the author andmay not be attributed to the World Bank or its affiliated organizations.
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
Pub
lic D
iscl
osur
e A
utho
rized
- DATA ON INCOME DISTRIBUTION IN BANGLADESH
This paper evaluates the 1973-74 Household Expenditure
Survey in Bangladesh and the estimates of income distribution
derived from it. Based on this evaluation, the survey data
were adjusted for biases in sample composition, and a more
representative distribution of household income was derived.
- .
TABLE OF CONTENTS
Page
I. Introduction. ..................... ........ 1
II. 1973-74 Household Expendituze Survey:A Brief Description ............................... 2
III. 1973-74 Household Expenditure Survey:An Evaluation ... ..................... 5
IV. Data Adjustment and the Derivation of Income Distri-bution ..... .. . .. . * . . . . . . . . . . . . .. 13
V. Comparison with Other Estimates ................. 20
VI. Concluding Remarks . ........................ 22
References .............................. ........... 23
Appendix: Tables ............................................ 24
DATA ON INCOME DISTRIBUTION IN BANGLADESH
I. Introduction
Bangladesh emerged as an independent nation in late 1971. The
first household survey was conducted in 1973-74 by the Bangladesh Bureau
of Statistics and coverted the period July to September 1973. Since then,
the household surveys have been conducted regularly on a quarterly basis,
separately for rural and urban areas. A published report is available only
for 1973-74; it includes the first four quarterly surveys, covering the
period July 1973 to June 1974. Prior to 1971, the Central Statistical Office
of Pakistan had conducted periodic annual household income and expenditure
surveys in both West and East Pakistan (now Bangladesh). However, the latest
published report for the eastern section of Pakistan is only for 1966-67.
The Bangladesh Bureau of Statistics, in one of its publications,/l
estimated household income distribution by deciles of households for 1966-67
and 1968-69, but did not assess the quality of the data on which the estimate
was based. Shail Jain also used the 1966-67 survey data to derive income
distribution, which was published in S. Jain's Size Distribution of Income
(1975).
The present study evaluates the 1973-74 Household Expenditure Survey
data and then derives income distribution by household and per capita in 1973-74
after making the necessary adjustments so that the distribution estimate could
be made for all of Bangladesh. It also broadly compares the 1973-74 household
income distribution estimates with those of 1966-67.
/1 A. K. M. Ghulam Rabbani and Shadat Hussain, Rural and Urban ConsumptionPatterns in Contemporary Bangladesh, Bangladesh Bureau of Statistics,Statistics Division, Ministry of Planning, Dacca, May 1978, pp. 2-6.
-2-
II. 1973-74 Household Expenditure Survey: A Brief Description
The 1973-74 survey was the first nationwide sample survey of
household income and expenditures in Bangladesh after its independence.
The annual results were obtained through the sample survey conducted on
a quarterly basis by the Bangladesh Bureau of Statistics in rural and
urban areas of the country during 1973-74. The objectives of the survey
were "to obtain data on household income and expendi.ture to ascertain
seasonal variatioi in expenditure patterns," "to provide data for deter-
mining weights for computing the consumer price index," "to collect data
about the standard of living," "'to collect data for calculating national
income by the consumption method," "to collect data related to nutrition
analysis," and "to collect data for the calculation of demand function."
a. Coverage
The survey covered all geographical areas of Bangladesh and all
households excluding those located in institutions such as hotels, hostels,
boarding houses, etc. The survey also excluded those households which
depended entirely on charity for their living.
b. Definitions
A household consisted of one or more persons, related. or unrelated,
who normally lived in the same dwelling and ate together with common cooking
arrangements. Thus household members included all persons present or tem-
porarily absent whose usual place of residence at the time of survey was
the sample household. Since living and eating together were essential
requirements for being counted as household members, a lodger livinig in
the household but caking meals outside and a servant taking his meals from
-3-
the employer's kitchen but whose usual place of residence was somewhere
else were not counted as household members. The concept of household,
used in the survey is similar to the de jure approach on which the house-
keeping concept of the United Nations' guidelines is based.
Household income was defined as the "material return in kind or
in cash, in exchange for goods and services, or from gifts, assistance,
etc., obtained by the household earners other than boarders, lodgers and
servants." For recording purposes; household income was classified into
monthly and annual income in cash or in kind. Monthly income included
wages and salaries, pensions, contributions made by boarders and lodgers
and professional fees, etc.; yearly income included interest, dividends,
e.'rnings from agricultural activities, commercial and industrial enter-
prises, rents from land and property, gifts and assistance, insurance
benefits, etc., and other regular receipt:. Imputed income was taken as
the estimated value at current market prices of the goods and services
received by households for which no cash payment was made. It included
home-produced goods consumed by the household, rent from owner-occupied
and rent-free houses, gifts and assistance received in kind, free meals
supplied by the employer, etc. The survey excluded receipts such as with-
drawals from savings and working capital, sales of assets, borrowings, etc.
The income concept used in the survey is broadly similar to the total house-
hold income concept in the United Nations' guidelines./- It did, however,
exclude the incomes of boarders and servants.
/l UN, Provisional Guidelines on Statistics of the Distribution of IncomeConsumption and Accumulation of Households, Series M; No. 61, New York,1977.
* ,, ,-J-, .- ,. ,,- - - ,--. . .-:-- --. :w
-4-
c. Samp14f Design
The whole country was divided into rural and urban areas. For
each, an independent two-stage stratified sample design was used.
Rural Bangladesh was divided into 60 strata. The sampling frame
for the first stage unit was a list of villages with information on
location, population and households according to the 1961 census. Eight
hundred villages (out of 64,500) were selected with a probability proportional
to the population size for the whole year. In each quarter, 200 villages,
selected at random, were covered. The second stage sampling frame comprised
a list of households prepared within the selected villages at the time of
enumeration. The households were selected by systematic sampling from each
sampled village using a constant overall sampling fraction of 1/1,190 for
all strata.
The urban areas of Bangladesh were divided into 19 strata on the
basis of the 1961 population. The sampling frame for the first stage sampling
unit was the stratum-wise list of electoral units (covering part of a city
or a group of towns) designated by the Election Commission for the 1973
elections. The number of electoral units to be sampled was fixed at 216,
covering 54 units in each quarter. The second stage sampling frame was
constructed in the same way as for the rural areas--by listing all households.
in the selected electoral units. Households were selected systematically
using a constant overall sampling fraction of 1/356 for all s trata.
Data were processed for 9,536 rural households and 2,237 urban house-
holds.
d. Survey Methodology and Reference Period
The survey was conducted between July 1973 and June 1974. Data
were collecteu through interviews, with the enumerator filling out the
-5-
questionnaire with the help of the head of the household and other household
members. The reference period for wages and salaries, pensions, professional
fees, etc. was the one month prior to the date of enumeration; for interest,
dividends, agricultural incomes, business and commercial incomes, property
income, gifts, assistance, etc. it was the previous year. The tabulated
data gave average monthly household income during the survey period.
III. 1973-74 Household Expenditure Survey: An Evaluation
a. Non-Enumeration
The report on the survey did not give any data about non-enumeration
rates. It only mentioned that data were processed for 9,536 rural and 2,237
urban households. If it is assumed that non-enumeration was zero, it shouLd
be possible to work out the total number of households in Bangladesh by using
overall sampling fractions and the actual number of sample households for
which data were processed in the rural-urban areas. Since the 1973-74 survey
period coincided with the 1974 census period (March 1974), the estimate of
total number of households based on the survey should be quite close to the
census figures for the total number of households./- Table 1 gives the total
-number of households in rural and urban areas of Bangladesh based on the census
and survey data.
For rural areas, the survey estimate was quite close to the census
estimate of total number of households (about 98%). However, for urban areas
the survey figure was only 74 percent of the census figure. Apart from
sampling errors, non-enumeration might be the main reason for such a big
difference in the two estimates of urban households.
/1 Bangladesh Population Census, 1974, Bulletin No. 3, Union PopulationStatistics, Bangladesh Bureau of Statistics, Statistics Division, Ministryof Planning, Dacca, April 1976.
-6-
Table 1: ESTIMATED NUMBER OF TOTAL HOUSEHOLDS, BANGLADESH, 1973-74
Unit Rural Urban
1. Sample households (number) 9,536 2,237
2. Overall sampling fraction - 1/1190 1/356
3.. Survey estimate of-total households - (thousand(1 divided by 2) numbers) 11,348 796
4. Census estimate of total households " 11,601 1,075
5. Survey es.timate of-total householdsexpressed as % of census households (%) 97.8 74.0
b. Bias in the Sample Composition
The survey report gave the distribution of sample households by
size of households in great detail. Similar detailed information regarding
the distribution of households by size was not available from the census
data. However, the housing census undertaken in 1973 does give the dis-
tribution of households by one- and two-person households; this can
reasonably be compared with the survey data (Table 2).
Table 2: PERCENTAGE DISTRIBUTION OF HOUSEHOLDS BY SIZE OF HOUSEHOLD
Rural /a Urban /aSurvey Centsus- Survey Census-
1-person household 1.56 2.80 2.59 5.50
2-person household 6.92 8.50 4.65 7.90
3-or more person household 91.52 88.70 92.76 86.60
All households 100.00 100.00 100.00 100.00
/a Derived from 1979 Statistical 'Fearbook of Bangladesh, Bangladesh Bureauof Statistics, Statistics Division, Ministry of Planning, Dacca, April1979, p. 108.
-7-
The above comparison suggests that one- and two-person households
were under-represented in the survey, while large size households were over-
represented. The under-representation of small-size households was relatively
higher in urban areas. This might be the result of a large non-response rate
in urban areas, where there are more small-size households, which are
more prone to non-response as compared with large-size households.
Similar data on the distribution of households by size were available
for Pakistan for 1970-71. Remembering that both Pakistan and Bangladesh are
Moslem countries, it was expected that the distribution of households by size
should be similar in both. However, the comparison of the Bangladesh survey
data and the Pakistan data shows that large-size households (especially 10
or more person households) had a much higher weight in Bangladesh than in
Pakistan (Table 3).
Table 3: PERCENTAGE DISTRIBUTION OF HOUSEHOLDS BY SIZE
Household Rural a Urbansize Bangladesh Pakistar Bangladesh Pakistan
1 1.6 2.1 2.6 3.8
2 6.9 9.9 4.6 8.9
3 12.0 14.0 8.6 11.5
4 .15.1 16.7 11.7 13.8
5 16.3 16.9 14.4 15.6
6 14.8 14.0 14.0 14.0
7 11.8 10.3 - 12.4 10.7
8 8.0 6.7 9.6 8.7
9 4.4 4.5 6.3 5.7
10 or more 9.1 4.9 15.8 7.3
All households 100.0 100.0 100.0 100.0 4
/a Taken from housenoia-<ncome ana :xpenciture Survey 17u-71, StatisticsD-.i -sf - - -.-F-
-8-
The above comparison reinforces the suspicion thaL the sample
composition of the 1973-74 survey was biased toward large-size households.
And since average household income increased with an increase in hcusehold
size (see Appendix Tables A-1 and 2), the survey estimate of average household
income for all households was likely to be overestimated in both the rural and
urban areas of Bangladesh.
The 1973-74 survey report gave household income distribution for
rural and urban areas separately. It also gave combined household income
distribution, which had been derived by taking the weighted average of rural
and urban distributions, the weights being the survey-based proportions of
total households in the two areas. According to the survey data, the ratio
for the distribution of urban-rural households was 6.6 to 93.4, as compared
to the census ratio of 8.5 to 91.5. Thus in the survey report, the combined
income distribution was worked out by attaching a higher weight to rural
areas, creating a bias toward rural income distribution.
In summary, the bias in the sample composition affected the survey
results in two ways. First, the sample over-;represented large-size households;
average household income was therefore overestimated. Second, in the survey
the weight of urban households (which had a higher average income than rural
households) was smaller; hence, the average household income for all of
Bangladesh was underestimated. Thus the two biases worked in opposite directions.
However, the net effect appears to be an overestimation of average household
income by about 5.4 percent (see Section IV).
c. Consistency of the Survey Data on Income and Expenditures
The survey report gave data on average household income and consumption
expenditures by different income classes. Consumption expenditures exceeded
- I
-9-
income for all income classes except the highest one (Table A-3). This
result confirms the general belief that respondents normally under-report
income. However, there is reason to suspect that the consumption expend-
itures in the survey were overestimated. In part this might be due to an
overestimation of the imputed value of home-grown consumption and the imputed
rent of owner-occupied and rent-free houses in the higher expenditure group
households, which were evaluated at the current market prices./- It is ex-
pected that the proportion of househGld (or percapita) expenditures for food
should decline along with an increase in total household (or per capita)
expenditures. However, the survey data did not show any such decline, although
total household expenditures increased manyfold (Table 4). For example, average
monthly household expenditures increased from 38 Taka in the lowest expenditure
class to 1706 Taka in the second highest expenditure class, but the share of
expenditures for food did not decline, remaining at around 74 percent. In fact,
it continued to increase in most of the expenditure classes.
As an independent check on the overall magnitude of total housohold
income, the survey-based estimate of total household income was compared with
a similar aggregate derived from the national accounts data. A review of the
national accounts data revealed that data were scanty, and it was not possible
to derive precise estimates of total household income from the national accounts
statistics. However, tentative estimates of total household income, derived
indirectly by adding private consumption and private savings, worked out at
around 69.0 billion Taka in 1973-74 (Table 5).
/1 In view of the non-availability of a standard rate of evaluation at currentmarket prices, especially in rural areas, the problem of imputation ofreceipts in kind and home-grown consumption was mentioned in the surveyreport itself.
- 10 -
Table 4: SHARE OF FOOD EXPENDITURES IN TOTAL EXPENDITURES BY EXPENDITURECLASSES, BANGLADESH, 1973-74
Average expenditure perHousehold expenditure Household Person Percentage of
class (Taka) (Taka) (Taka) expenditures on food
Less than 50 38.4 15.1 71.8
50 - 99 82.7 29.6 70.8
100 - 149 127.1 37.7 73.1
150 - 199 175.6 45.5 74.7
200 - 249 224.5 53.2 75.3
250 - 299 273.1 58.0 76.0
300 - 399. 347.4 66.2 76.0
400 - 499 446.9 74.7 75.5
500 - 749 604.6 88.0 75.3
250 - 999 857.2 104.2 74.3
1,000 - l1499 1,186.0 118.8 72.3
1,500 - 1,999 1,705.71 156.3 73.8'
2,000 - above - '3,728.6 345.2 68.6
Alt classes . . 500-.6. 85..9 74.2
Source: A Report on the Household Ex=enditure Survey of Bangladesh, 1973-74,Volume II, Bangladesh Bureau of Statistics, Statistics Division,Ministry of Planning, Dacca, December 1980.
Table 5: ESTIMATION OF HOUSEHOLD INCOME IN BANGLADESH, 1973-74
1973-74(Taka millions)
1. Gross Domestic Product at market prices 71,086
2. Imports of goods and services 7,320
3. Exports of goods and services 2,983
4. Gross investment 7,570
5. Government consumption expenditures on goodsand services .2,555
6. Private consumption expenditure-/a 65,298
7. Domestic savings (1-5-6) 3,233
a) Government saving b on current accounts -486b) Corporate savings- n.a.c) Household savings 3,719
8. Household income (6 +7c) 69,017
/a Inclusive of the consumption expenditures of non-profit privateinstitutions.
/b Corporate savings were likely to be very small, as corporate taxeson income were only 9 million Taka in 1973-74.
Sources: 1979 Statistical Yearbook of Bangladesh, Bangladesh Bureauof Statistics, Ministry of Planning, Dacca, April 1979.
Bangladesh: Current Economic Situation and Review of theSecond Plan, Volume I, World Bank, ReDort No. 3309-BD,Washington, D.C., February 23, 1981.
- 12 -
As compared to the national accounts-based estimate of 69.0 billion
Taka, the survey estimate of total household income of about 72.7 billion Taka
was about 5 percent higher. Normally,, a survey estimate of household income
is lower than a similar estimate derived from the national accounts, as there
is general tendency on-the part of respondents to under-report income. However,
in the case of the Bangladesh household survey, the sample was biased toward
large-size households (with larger household income); thus average household
income was over-estimated.
If the sample is adjusted for the biases in the sample composition,
total household income works cut to 68.8 billion Taka, which is quite close
to the national accounts- estimate (Table 6).
Table 6: SURVEY ESTIMATE OF TOTAL HOUSEHOLD INCONE
Unit . Rural Urban Total
1. Average annual household income Taka 5,565 7,555 -
2. Number of households thous. 11,601 1,075 12,676
3. Total household income mil Taka 64,560 8,122 72,682
4. Adjusted annual household income-/ Taka 5,292 6,900 -
5. Adjusted total household inco /1 mil Taka 61,392 7,418 68,810
/1 Adjusted for the bias in the sample composition.
The revised estimate of average household income is derived by making
adjustments for the bias in the sample composition resulting from the under-
representation of one- and two-person households and for the bias toward rural
households. There might still be an over-estimate if the three- or four-person
- 13 -
households were also under-represented in the sample. However, there is
no comparable census data. On the other hand, household income might be
under-estimated, as survey respondents normally under-report income. Since
the above two factors.work in opposite,directions, the adjusted figure for
average household income should reasonably reflect the prevailing level of
average household income in 1973-74.
IV. Data Adjustment and the Derivation of Income Distribution
Table 2 showed that the proportions of one- and two-person households
in the sample were lower than in the census. Thus the sample proportions for
one- and two-person households were increased to the census levels, the latter
being 2.8 percent and 8.5 percent respectively for rural areas and 5.5 percent
and 7.9 percent respectively for urban areas. Since "ten or more person"
households were over-represented in the sample (see Table 3), they were reduced
from 9.1 percent to 6.3 percent in rural areas and from 15.8 percent to 9.6
percent in urban areas. Table 7 gives the adjusted distribution of households
and average monthly househo,ld income by size of household.
Using the adjusted distribution of households by size and the cor-
responding average monthly household incomes, the adjusted overall average
monthly household incomes were derived. They were 441 Taka for rural areas
(5,292 Taka yearly) and 575 Taka for urban areas (6,900 Taka yearly). From
these, total household income worked out to 68.8 billion Taka, very close
to the 69.0 billion Taka derived from the national accounts (see Table 6).
The adjusted number of sample households in each household size
group were distributed among various income classes in the same proportion
as was observed in the original sample. Finally, the adjusted distributions
- 14 -
Table 7: ADJUSTED DISTRIBUTION OF HOUSEHOLDS AND AVERAGE HOUSEHOLDINCOME BY HOUSEHOLD SIZE
- Rural UrbanAverage Average
- monthly monthlyHousehold size % of household % of household
households income households income(Taka) (Taka)
One 2.8 184 5.5 252
TWO 8.5 223 7.9 305
Three 12.0 285 8.6 342
Four 15.1 333 11.7 421
Five 16.3 393 14.4 489
Six 14.8 446 14.0 555
Seven 11.8 525 12.4 630
Eight 8.0 577 9.6 666
Nine 4.4 710 6.3 866
Ten or more 6.3 1,004 9.6 1,180
All households 100.0 441 100.0 575
No. of sample house-holds 9,536 - 2,237 -
Source: Derived from Tables A-1 and A-2 after adjusting for the bias inthe sample composition.
- 15 -
of households and income were obtained by aggregating revised sample house-
holds and their income shares over the household size groups for each
income class, separately for rural and urban areas.
The combined household income distribution for all of Bangladesh
was derived by-aggregating the-rural and-urban- income distributions, using
the total number of census households in the two areas as weights.
Table 8 gives the adjusted household income distributions for
rural, urban and all-Bangladesh by houshold income levels. lncnme inequality,
as measured by the Gini ratio, was 0.358 for- the whole country. It was
slightly higher in urban areas (0.379) than in rural areas (0.354).
Table 9 gives the distribution of income by deciles.of households.
The income shares of the lowest 20 percent of households were 6.9 percent
in rural areas and 6.7 percent in urban areas; for the top 10 percent of
households, the income shares were 27.1 percent for rural areas and 29.3 percent
for urban areas.
It is interesting to note that the level of income inequality in
Bangladesh is considerably lower than in other developing countries in the
region (Appendix Table A-5).
The estimates of household income distributions as shown in
Tables 8 and 9 did not take into account the effects of household size on
the pattern of-income distribution. To eliminate such effects, households
(and household members) were reclassified according to per capita income.
Appendix Tables A-l and A-2 provided the distribution of sample households
by household size for each income class. They also provided average per
capita income for different household size groups, computed by dividing the
average household income with the different sizes of hiouseholds. The dis-
tribution of household members according to per capita income was derived
- 16 -
Table 8 : HOUSEHOLD INCObE DISTRIBUTION, ADJUSTED, 1973-74
Monthlyhousehold Rural Urban Combined
income class Households Income Households Income Households Income(Taka) ()( ) ( ( (
up to 50 0.19 0.01 - - 0.17 0.01
50 - 99 2.15 0.39 0.98 0.13 2.05 0.36
100 - 149 6.29 1.80 2.64 0.59 5.98 1.67
150 - 199 9.62 3.79 5.54 1.67 9.27 3.57
200 - 249 11.07 5.63 9.12 3.53 10.91 5.40
250 - 299 10.35 6.41 9.25 4.35 10.26 6.18
300 - 399 17.69 13.88 17.30 10.35 17.66 13.50
400 - 499 12.78 12.89 13.14 10.17 12.81 12.60
500 - 749 18.19 24.87 20.70 21.78 18.40 24.54
750 - 999 6.26 12.22 8.99 13.32 6.49 12.34
1,000 - 1,499 4.12 11.15 8.23 16.98 4.47 11.77
1,500 - 1,999 0.71 2.74 2.23 6.54 0.84 3.15
2,000 & above 0.58 4.22 1.88 10.59 0.69 4.91
All classes 100.00 100.00 100.00 100.00 100.00 100.00
Gini ratio 0.354 0.379 0.358
Source: Derived from Appendix Table A-4 after adjusting for bias in samplecomposition.
J
- 17 -
Table 9: HOUSEHOLD INCOE DISTRIBUTION BY DECILES, ADJUSTED,1973-74
Cumulative Cumulative share of incomedeciles of Rural Urban Combined__households
10 2.74 2.71 2.73
20 6.88 6.73 6.86
30 12.04 11.75 11.99
40 18.29 17.73 18.23
50 26.14 24.62 25.87
60 34.57 32.93 34.33
70 44.66 43.45 44.48
80 58.28 54.54 57.81
90 72.92 70.72 72.56
100 100.00 100.00 100;.00
Source: Derived from Table 8.
by reclassifying the household members in different per capita income classes
(Table A-6). Per capita income distribution was adjusted for biases in the
sample composition in the same way as was done for household income distri-
bution. The adjusted per capita income distribution was then derived (Table 10).
Income distribution by deciles of population show that the income share of the
lowest 20 percent of population was about 8.9 percent in rural areas and 8.6
percent in urban areas (Table 11). For the top 10 percent of the population,
income shares in the rural and urban areas were 22.7 percent and 25.3 percent
respectively.
-18 -
Tablel0: PER CAPITA INCOME DISTRIBUTION BY HOUSEHOLD MEMBERS,ADJUSTED, 1973-4
Monthlyper capita Rural Urban Combined
income class Popuiiltio&- Iicome- Population Income Population IncomeCTaka) (M) (%) (M) (%) (x)
below 25- 2.58 - 0.65 2.00 0.40 2.42 0.58
25 - 34 6.79 2.60 3.53 1.13 5.87 2.18
35 - 44 9.77 5.05 7.51 3.16 9.13 4.51
45 - 54 10.28 6.51 8.58 4.82 9.80 6.03
55 - 64 12.10 8.84 9.44 5.61 11.35 7.93
65 - 74 14.88 13.18 10.37 7.34 13.60 11.53
75 - 89 14.15 14.72 16.45 13.65 14.80 14.41
90 - 104 10.31 12.32 12.28 12.09 10.87 12.26
105 - 119 4.70 6.50 4.54 5.06 4.65 6.09
120 - 139 6.09 9.46 8.50 10.72 6.77 9.82
140 - 159 3.36 6.20 4.87 7.26 3.79 6.50
160 - 199 2.28 5.05 4.26 7.70 2.84 5.80
200 - 249 2.03 5.37 5.42 12.08 2.99 7.27
250 & above 0.68 3.55 2.25 8.98 1.12 5.09
All classes 100.00 100.00 100.00 100.00 100.00 100.00
Gini ratio 0.280 0.310 0.284
Source: Derived from Appendix Table A-6 after adjusting for the bias in the samplecomposition.
- 19 -
Table 11: INCOME DISTRIBUTION BY DECILES OF POPULATIONADJUSTED , 1973-74
Cumulative Cumulative shares of incomedeciles of Rural Urban Combinedpopulation
10 3.58 3.41 3.60
20 8.84 8 60 8.86
30 15.23 14.49 15.24
40 22.54 21.45 22.44
50 31.16 29.57 30.92
60 40.57 38.20 40.38
70 50.97 48.04 50.58
80 62.84 59.94 62.26
90 77.27 74.73 76.61
100 100.00 100.00 100.00
Source: Derived from Table 10.
A comparison of household and per capita income distributions
shows that the income inequality was less on a per capita basis than on
a household basis. A similar trend was noted in the income shares of the
bottom 20 percent and top 10 percent of population, indicating less income
inequality in per capita income distribution than in household income dis-
tribution (Tab.le 121.
Table 12: INCOME INEQUALITY MEASURES FOR HOUSEHOLD ANDPER CAPITA INCOME DISTRIBUTIONS, 1973-74
Household Population
Share of lowest 20% 6.9 8.9
Share of top 10% 27.4 23.4
Gini ratio 0.36 0.28
- 20 -
It is difficult. to explain the lower income inequality in per
capita income distribution as compared to household income distribution.
Presumably it decreased, mainly for the following reason. The average size
of households increased -with an increase in household income, and the lower
deciles of households, which accounted for a smaller share of income,
included a smaller percentage of the population, whereas the higher deciles
of households included a larger percentage of the population. However,
part of the effect might have been cancelled out as some of the households
changed their positions when reclassified by per capita income.
The per capita income distribution confirms the low degree of
income inequality in Bangladesh as compared with other developing countries.
V. Comparison with Other Estimates
The-World-Bank- Social- Indicator Data Sheets-give some data on
household income distribution in Bangladesh for the year 1966-67. They were
taken from Shail Jain's compilation of the size distribution of income, which
in turn, as noted, was based on the 1966-67 Survey of Household Income and
Expenditure undertaken by the Central Statistical Office of Pakistan.-/
We reviewed the 1966-67 survey report and found it reasonably
comparable with the 1973-74 survey results in terms of concepts, definitions,
and coverage. In fact, the 1966-67 survey had a more representative coverage
of small-size households than the 1973-74 survey. In both surveys, households
were classified by similar household iricome classes, so that two income dis-
tributions were fairly comparable (Appendix Table A-7). Table 13 gives the
inequality measures for the 1966-67 and 1973-74 household income distribu-
tions for the whole country.
/1 Report on the Quarterly Survey of Current Economic Conditions in Pakistan(Household Income and Expenditure,Jl 1966 to June I97,Central Statis-
-21-
Table 13: HOUSEHOLD INCONE INEQUALITY MEASURES, BANGLADESH,1966-67 AND 1973-74
Gini Income share in %coefficient Top 10% Lowest 20% Lowest 40%
households households households
1. 1973-74 Survey 0.36 27.4 6.9 18.2/a
2. 1966-67 Survey-
Jain's estimate 0.34 26.7 7.9 19.6
Rabbani's/b 0.34 29.5 8.7 19.6estimate
/a Jain's figures are slightly different than Rabbani's figures, as the formerwere derived by fitting a Lorenz curve to the observed data.
/b Taken from Rural and Urban Consumption Patterns in Contemporary Bangladesh,by A. K. M. Ghulam Rabbani and Shadat Hussain, Bangladesh Bureau of Statis-tics, Statistics Division, Ministry of Planning, Dacca, May 1978, pp. 2-6.
Table 13 shows a slight increase in income inequality from 1966-67
to l973- 4. However, in the 1966-67 survey report, about 60 percent of the
sample households were grouped in only two classes, and the summary inequality
measure3 as calculated from these data would be subject to a wider margin of
error. In such a case, income inequality is normally under-estimated. Thus,
we feel that the income distributions in the two periods were not very dif-
/1ferent.- The 1966-67 survey also confirmed the low degree of income in--
equality in Bangladesh.
/1 Shail Jain also gave per capita income distribution for Bangladesh. It,however, is not comparable with our per capita income distribution, asshe classified household members by household income levels instead ofby per capita income levels.
- 22 -
VT.- Concluding Remarks .
We derived the distribution of income by-househoid in E1langladesh for
the period July 1973 td June 1974, yt,ing'data from the country's fi.rst nati6oal
household survey. We also derived per capita distribution by reclassifying
household members in the per capita income classes. The income in the survey
referred to total household income before the deduction of direct taxes.
However, the distribution of before-tax income and that of after-tax income
should not differ very much, as the direct tax base is very small-in Bangladesh.
We compared our estimate of household income distribution with
other estimates for 1966-67. All confirm the low degree of income inequality
in Bangladesh. The pattern of income distribution did not show any signifi-
cant change in the two-periods (1966-67 and 1973-74), and income inequality
remained more or less unchanged.
The survey data suffered from both high non-response rates and biases
in the sample compositions. Still, the survey was national and used sound
concepts and definitions. It provided a reasonably good data base for deriving
estimates of income distribution after making the necessary adjustments.
J
- 23 -
REFERENCES
1. A Report on the Household Expenditure Survey of Bangladesh, 1973-74,Volume I, Bangladesh Bureau of Statistics, Statistics Division, Ministryof Planning, Dacca, August 1978.
2. A Report on the Household Expenditure Survey of Bangladesh, 1973-74,Volume II, Bangladesh Bureau of Statistics, Statistics Division,Ministry of Planning, Dacca, December 1980.
3. Rabbani, A. K. M. Ghulam, and, Shadat Hussain, Rural and Urban ConsumptionPatterns in Contemporary Bangladesh, Bangladash Bureau of Statistics,Statistics Division, Ministry of Planning, Dacca, May 1978.
4. JN, Provisional Guidelines on Statistics of the Distribution of Income,Consumption and Accumulation of Households, Series M, No. 61, New York,1977.
5. Bangladesh Population Census: 1974, Bulletin No. 3, Union PopulationStatistics, Bangladesh Bureau of Statistics, Statistics Division,Ministry of Planning, Dacca, April 1976.
6. 1979 Statistical Yearbook of Bangladesh, Bangladesh Bureau of Statistics,Statistics Division, Ministry of Planning, Dacca, April 1979.
7. Household Income and Expenditure Survey, 1970-71, Statistics Division,Ministry of Finance, Government of Pakistan, May 1973.
8. Ban&ladesh: Current Economic Situation and Review of the Second Plan,Volume I, World Bank, Report No. 3309-BD, February 23, 1981.
9. Report on the Quarterly Survey of Current Economic Conditions in Pakistan(Household Income and Expendit.ure), July 1966 to June 1967, CentralStatistical Office, Pakistan, Karachi 1968.
10. Jain Shail, Size Distribution of Income: A compilation of Data, TheWorld Bank, Washington, 1975.
Table A-1: DISTRIBUTION OF SAHPLE IHOUSEIIOLDS BY tMONTIHLY IIOUSEIHOLD INCOME AND HOUSEHOLD SIZE, RURAL BANGLADESH, 1973-74
Honthly Average Numberhousehold household of Size of household
income class income householde One Two Three Four Five Six Seven Eight Nine Ten or more
(Taka) (Taka)
Less than 50 29.95 12 a 29.95 14.98 9.98 7.49 -- -- -- -- i-- --
b 6 3 1 2 -- -- -- -- r-- --
50- 99 79.85 164 a 79.85 39.93 26.62 19.96 15.97 13.3'. 11.41 9.98 -- 6.88
b 42 41 30 21 10 11 5 2 2
100 - 149 126.49 537 a 126.49 63.25 42.16 31.62 25.30 21.08 18.07 15.81 j14.05 16.20
b 41 136 132 101 74 26 15 9 2 1
150 - 199 174.20 868 a 174.20 37.10 58.07 43.55 34.84 29.03 24.89 21.78 19.36 11.69
b 22 144 211 186 142 85 53 15 7 3
200- 249 224.41 1,016 a 224.41 112.20 7h.80 56.10 44.88 37.40 32.06 28.05 , 24.93 22.67
b 15 137 219 239 183 124 . 50 33 5 11
250 - 299 273.19 971 a 273.19 136.60 91.06 68.30 54.64 45.53 39.03 34.15 30.35 25.06
b 4 76 174 215 208 150 77 40 j5 12
300- 399 346.31 1,673 a 346.31 173.16 115.44 86.58 69.26 57.72 . 49.47 43.29 k3P.48 30.65
b 7 88 191 304 356 333 208 104 i'43 39
400 - 499 445.24 1,228 a 445.24 222.62 148.41 111.31 89.05 74.21 63.61 55.66 *49.47 40.85 >
b 6 19 100 180 248 247 196 130 142 60 a
500 - 749 603.74 1,791 a 603.74 301.87 201.25 150.94 120.75 100.62 86.25 75.47 67.08 53.43
b 4 12 66 158 257 327 344 264 154 205
750 - 999 861.77 659 a -- 430.89 287.26 215.44 172.35 143.63 123.11 107.72 .95.75 74.29
b -- 2 9 26 51 67 106 104 92 202
1,000 - 1,499 1,193.88 464 a 1,193.88 596.94 397.96 298.47 238.78 198.98 170.55 149.24 i32.65 93.27
b 1 2 5 11 16 38 55 53 *47 236
1,500 - 1,999 1,694.89 83 a 1,t6A4.89 -- 564.96 423.72 338.98 282.48 242.13 211.86 188.32 126.48
b 1 -- 3 1 6 3 3 6 '8 52
2,000 aud above 3,232.59 70 a -- -- 1,077.53 808.15 646.52 538.77 461.80 -- 359.18 209.91
b 1 1 5 4 4 -- 5 50
All clas6es 463.73 9,536 a 183.96 111.70 94.90 83.14 78.54' 74.40 74.99 72.14 78.94 82.94
b 149 660 1,142 1,445 1,556 1,415 1,116 760 420 873
Average hiousehold 463.73 - 183.96 223.41 284.69 332.57 392.70 446.37 524.91 577.10 710,42 1,003.60 0
Income
ote: 'a' denotes average monthly per capita Income in Taka. 'b' denotes number of sample househiolds.
ource: A Report on the lHousehold Expenditure Survey of Bangladesh, 1973-74, Volume I, Bangladesh Bureau of Statistics, Statistics Division, Hinistry
of Planning, Dacca, August 1978.
4L.
Table A-2: DISTRIBUTION OF SAMPLE HOUSEHOLDS BY tIONTIILY IhOUSEHiOLD INCOtME AND hIOUSEIIOLD SIZE. URBAN BANGLADESII, 1973-74
Monthly Average Numberhousehiold hiousehold of Size of houselholdincome class income houselholds One Two Three Four Five Six Seven Eight Nine Ten or more
(Taka)
Less than 50 --- a -b -- -- -- -- -- -- -- -- -- --
50 99 77.90 15 a 77.90 38.95 25.97 19.48 15.58 12.99 -- -- -- 7.79b 5 2 2 1 1 3 -- -- -- 1
100 - 149 127.72 43 a 127.72 63.86 42.57 31.93 25.54 21.29 18.25 -- 14.19 --
b 6 13 15 3 2 2 1 -- 1 --
150- 199 172.67 107 a 172.67 86.33 57.56 43.17 34.53 28.78 24.67 21.58 19.19 13.93b 8 13 24 23 19 8 5 4 1 2
i200 - 249 222.42 175 a 222.42 111.21 74.14 55.61 44.48 37.07 31.77 27.80 24.71 18.23b 12 26 29 36 29 18 9 6 4 6
250 - 299 270.01 181 a 270.01 135.00 90.00 67.50 54.00 45.00 38.57 33.75 30.00 24.33b 17 13 26 28 40 24 14 9 6 4
300 - 399 343.72 372 a 343.72 171.86 114.57 85.93 68.74 57.29 49.10 42.97 38.19 33.37 a; b 5 18 45 71 76 60 47 34 7 9
400 - 499 444.86 291 a 444.86 222.43 148.29 111.22 88.97 74.14 63.55 55.61 49.43 41.97b 3 12 22 34 46 58 47 35 11 23
500 - 749 604.82 483 a 604.82 302.41 201.61 151.21 120.96 100.80 86.40 75.6Q 67.20 53.05b 1 6 23 44 65 84 86 63 47 64
750 - 999 852.28 228 a -- -- 284.09 213.07 170.46 142.05 121.75 106.54 94.70 76.78b -- -- 5 10 24 32 35 29 24 69
1.000 - 1.499 1,186.68 224 a 1,186.68 -- 395.56 296.67 237.34 197.78 169.53 148.34 131.85 96.48b 1 -- 1 9 14 19 23 25 26 106
1,500 - 1,999 1,682.21 62 a -- -- 560.74 420.55 336.44 280.37 240.32 210.28 186.91 123.69b -- -- 1 3 2 3 7 7 9 30
2,000 and above 3,242.71 56 a -- 1,621.36 -- -- 648.54 540.45 463.24 405.34 360.30 219.10b -- 1 -- -- 3 3 3 2 5 39
All classes 629.59 2,237 a 252.43 152.65 113.96 105.34 97.77 92.45 89.98 83.21 96.19 97.49b 58 104 193 262 321 314 277 214 140 353
*Average household 629.59 -- 252.43 305.30 341.87 421.34 488.86 554.71 629.88 665.70 865.70 1,179.60
Note: 'a' denotes average monthly per capita income in Taka. 'b' denotes number of sample households.
Source. Same as In Table A-1.
4.
APPENDIX: TABLES
- 26 -
Table A-3: MONTHLY HOUSEHOLD INCOME AND EXPENDITURES BYHOUSEHOLD INCOME LEVELS, BANGLADESH, 1973-74
Household Average monthly Ratioincome class Income (y) Expenditure (c) y/c
(Taka) (Taka) (Taka)
Lass than 50 29.95 58.92 0.51
50 - 99 79.80 128.56 0.62
100 - 149 126.52 159.43 0.79
150 - 199 174.14 196.83 0.88
200 - 249 224.31 259.81 0.86
250 - 299 273.02 294.96 0.93
300 - 399 346.15 365.71 0.95
400 - 499 445.21 483.09 0.92
500 - 749 603.82 635.90 0.95
750 - 999 860.88 902.72 0.95
1000 - 1499 1192.97 1242.11 0.96
1500 - 1999 1692.58 1952.87 0.87
2000 and above 3234.55 2519.44 1.28
All classes 474.60 501.57 0.95
Source: Same as in Table A-1.
APPENDIX: TABLES
- 27 -
Table A-4: HOUSEHOLD INCOME DISTRIBUTION, UNADJT1TSTTEn, 1973-74
Monthlyhousehold Rural Urban Combined
income class Households Income Households Income Households Income(.Taka) (%) (%) (%) (%) (%) (%)
Less than 50 0.13 0.01 0.00 0.00 0.12 0.01
50 - 99 1.72 0.30 0.68 0.09 1.65 0.28
100 - 149 5.63 1.54 1.93 0.39 5.39 1.44
150 - 199 9.10 3.42 4.78 1.31 8.82 3.24
200 - 249 10.65 5.16 7.82 2.76 10.47 4.95
250 - 299 10.18 6.00 8.09 3.47 10.05 5.78
300 - 399 17.54 13.10 16.63 9.08 17.47 12.75
400 - 499 12.88 12.36 13.01 9.19 12.89 12.09
500 - 749 18.79 24.44 21.59 20.74 18.96 24.12
750 - 999 6.91 18.84 10.19 13.80 7.13 12.92
1,000 - 1,999 0.87 3.18 2.77 7.41 1.00 3.55
2,000 & above 0.73 5.12 2.50 12.89 0.85 5.79
All classes 100.00 100.00 100.00 100.00 100.00 100.00
Gini ratio 0.358 0.384 0.362
Source: A Report on the Household Expenditure Survey of Bangladesh, 1973-74, Volume I,
Bangladesh Bureau of Statistics, Statistics Division, Ministry of Planning,
Dacca, August 1978.
) ,
APPENDIX: TABLES
Table A-5: HOUSEHOLD INCOME INEQUALITY MEASURESIN SELECTED COUNTRIES
Gini Income share in %Country Year coefficient Lowest Lowest Top
20% 40% 10%households households households
Bangladesh /a 1973-74 0.36 6.9 18.2 27.4
India /a 1975-76 0.42 7.0 16.2 33.6
Nepal /a 1976-77 0.53 4.6 12.6 46.5
Indonesia/- 1976- 0.44 6.6 14.4 34.0
Philippines/a 1970-71 0.47 5.2 14.2 38.5
Malaysia/b 1970 0.50 3.3 10.6 39.6
Sri Lanka /b 1969-70 0.36 7.5 19.2 28.2
/a Obtained from the EPD Income Distribution Project, Division Working Papers, The WorldBank, Washington, D.C.
/b Obtained from World Development Report, 1980, The World Bank, Washington, D.C.,August 1980, pp. 156-157.
, s.
APPENDIX: TABLES
-29-
TableA-6: PER CAPITA INCOME DISTRIBUTION BY HOUSEHOLD MEMBERS,UNADJUSTED, 1973-74
Monthlyper capita Rural Urban Combined
income class Population Income Population Income Population Income(in Taka) (%) (x) (%) (%) (%) (%)
Below 25 2.54 0.64 2.19 0.45 2.44 0.59
25 - 34 6.76 2.59 3.44 1.13 5.82 2.18
35 - 44 9.59 4.97 7.35 3.17 8.96 4.46
45 - 54 11.04 7.06 10.30 5.47 10.83 6.61
55 - 64 11.37 8.32 8.31 5.03 10.50 7.39
65 - 74 15.42 13.75 9.26 6.70 13.68 11.75
75 - 89 13.24 13.82 16.58 13.95 14.19 13.86
90 - 104 11.48 13.70 14.42 14.49 12.31 13.92
105 - 119 4.35 6.03 3.81 4.34 4.20 5.55
120 - 139 6.05 9.43 8.53 10.99 6.75 9.87
140 - 159 3.19 5.90 4.35 6.63 3.52 6.11
160 - 199 2.06 4.59 3.57 6.60 2.48 5.16
200 - 249 2.31 6.11 6.15 13.99 3.40 8.34
250 & above 0.60 3.09 1.74 7.06 0.92 4.21
All classes 100.00 100.00 100.00 100.00 100.00 100.00
Gini ratio 0.277 0.307 0.282
Source: Derived from Appendix Tables A-1 and A-2.
..
APPENDIX: TABLES
-30-
Table A-7: PERCENTAGE DISTRIBUTION OF HOUSEHOLDS AND INCOME BY HOUSEHOLDINCOME CLASS, 1966-67, EAST PAKISTAN, BANGLADESH
Monthly Rural Urban Combinedhousehold income Households Income Households Income Households Incomeclass (Rupees) % % % % % %
Less than 50 2.8 0.8 0.9 0.2 2.7 0.7
50 - 99 35.0 17.9 18.2 6.6 34.3 17.2
100 - 149 25.9 21.1 27.2 14.9 26.0 20.7
150 - 199 15.6 17.8 16.6 12.7 15.7 17.5
200 - 249 8.9 13.1 10r7 10.6 8.9 12.9
250 - 299 5.1 9.2 7.2 8.7 5.2 9.2
300 - 399 3.7 8.5 8.7 13.1 3.9 8.8
400 - 499 1.4 4.0 4.0 7.9 1.5 4.4
500 - 749 1.2 4.9 3.5 8.8 1.3 5.1
750 - 999 0.2 1.2 1.2 4.8 0.3 1.4
1000 - 1499 0.2 1.5 1.3 6.7 0.2 1.8
1500 - 1999 0.0 0.0 0.3 2.3 0.0 0.1
2000 and above 0.0- 0.0 0.2 2.7 0.0 0.2
All classes 100.0 100.0 100.0 100.0 100.0 100.0
Source: A Revort on the Ouarterly Survey of Current Economic Conditions in.,Pakistan (Household Income and Expenditure), July 1966 to June 1967,Central Statistical Office, Pakistan, Karachi, 1968, pp. 100-102.
J ',D0