ASSESSING INTRA-HOUSEHOLD VARIATIONS

IN CONSUMPTION, EDUCATIONAL AND

ECONOMIC ATTAINMENTS IN DIFFERENT
CATEGORIES OF HOUSEHOLDS IN
UTTAR PRADESH

Technical Proposal

Rajendra P. Mamgain

Professor of Economics

GIRI INSTITUTE OF DEVELOPMENT STUDIES

Sector-O, Aliganj, Lucknow
Uttar Pradesh
 Contents

Page No

Motivation for the Study 4

Objectives 6

Approach of the Study: Data and Methodological Strategy 6

Empirical Outline of Secondary Data Analysis 7

Primary Survey: Sampling Design and Sample Size 13

Questionnaire Design 15

Work Plan 15

2
 ASSESSING INTRA-HOUSEHOLD VARIATIONS IN CONSUMPTION,
EDUCATIONAL AND ECONOMIC ATTAINMENTS IN DIFFERENT
CATEGORIES OF HOUSEHOLDS IN UTTAR PRADESH

1. Motivation for the Study

The Directorate of Economics and Statistics (DES), Department of Planning, Government of

Uttar Pradesh has been making its best efforts towards strengthening the statistical system in the
state. One of such endeavors relate to improving Gross State Domestic Product (GSDP)
estimates periodically. It has clearly identified several sectors of the state economy of Uttar
Pradesh which are not being captured adequately in the estimation of GSDP. Accordingly, the
DES rightly aims at initiating a series of studies for understanding and strengthening its GSDP
estimates. One of such sectors which assumes importance in the measurement and distribution of
income relates to intra-household inequalities in various indicators of development that are likely
to effect the overall magnitude of state income. We intend to study this important issue which is
scantly explored in the Indian context due to paucity of data at household level for its individual
members.

Inequality in distribution of income or resources and its impact on economic well-being of

individual at micro level as well as at the macroeconomic level has been a matter of intense
inquiry over time and again. In the last two decade or so there has been ample empirical evidence
regarding rising inequality both in income and consumption expenditure. It has been found true
for both developed as well as developing countries. In all these evidence the unit of inequality
analysis has been the household. However, such studies could hardly focus on intra-household
dynamics of consumption, income and resource allocation, which have far reaching impact on
the overall household’s well-being.

Although theoretically in the field of utility demand analysis an individual has been the core unit,
but due to asymmetries of information within household empirical evidences on intra-household
inequalities are very less. There has been large body of theoretical literature developed since
1960s (starting with Becker) but the assumption are validated and rejected again and again.
There has been evidence of intra-household discrimination and pro male bias in economic

3
decision making like nutrients allocation, expenditure particularly schooling, health, migration
and command over resources (Sen and Sengupta, 1989; Behrman, 1992 etc).Despite this, the
household remained as allocation unit with some advancement of eliminating such
discrimination in the form of conditional transfers1.

Theoretically there are different approaches to analyse the intra-household variation in terms of
allocation of the resources. Most popular is the debate between unitary versus collective
approach of intra-household resource allocation which is reflected and measured through income
and expenditure information of the household and its members. Time and again theoretical
approaches provide justification about the discrimination and discrepancies in intra-household
consumption variations.

The other major needs of intra household analysis is the advancement in the level of education of
(both male and female), increasing female participation in formal employment and their
contribution in earnings. With the economic empowerment of women the decision making
within household may change which has definite impact on resource allocation. Improving
education level and participation in society may break what Sen (1999, 62) calls as endogenous
or adaptive preference of female to adjust with the situation. The change in source of income is
strong variable to effect within household allocation. As Lise and Seitz (2011) has noted:

…. “if consumption allocation depends on the source of income and the sources of
income within household have changed over time, then adult equivalence scales will
produce an inaccurate picture of the trends in consumption inequality..”

Developing country like India has still large proportion of population living in absolute poverty.
Among different strategies, state transfers through various schemes to improve individual
standard of living has a crucial role in the economic development. In this strategy individual
welfare is assumed to be the average of the household based on adult equivalence. However,
between the welfare policies of the state and the individual well-being family plays an
intermediary role. The household decision making has an important implication on the resource
allocation for the individual and thus generates intra-household variation in different indicators
including consumption, educational attainment, employment, income, etc.
1
In India cash transfer in JSY for delivery in government hospital and 33 percent of female participation in
MNREGA has been an attempt to improve delivery of gain at required individual level.

4
Uttar Pradesh being the largest state in terms of population size has also relatively higher number
of poor as compared to other states. At the same time there is considerable level of disparity in
economic and social development across its different agro-climatic regions as well as across
various social groups. To achieve the development of the state improving the standard of living
of the individual will have synergic effect. However, the effort directed at the household ignores
the crucial discrepancies that exist within household which have larger impact on achieving
overall desired outcome(s).

2. Objectives

The major objective of the study will be to measure intra-household variations in consumption
expenditure, education and other economic indicators such as employment and income. Effort
shall be to calculate certain indicators that can be used for prospective intra-household variations.
To monitor certain indicators at individual level may be helpful in designing and tracking policy
in a better manner. Thus, it will not be a kind of work to do impact analysis of any discrimination
or intra household inequality on certain outcome variables (like impact of nutrient allocation on
anthropometric measures of health).

3. Approach of the Study: Data and Methodological Strategy

As mentioned above, despite the cruciality of the issue one of the major hurdles in intra
household empirical research is the paucity of data at the individual level. National Sample
Survey Organisation (NSSO) under Ministry of Statistics and Program Implementation (MoSPI)
collects detailed information on consumption expenditure and other indicators like education
activity status, employment features etc in thick rounds of Consumption Expenditure Survey
(CES) and Employment Unemployment Survey (EUS). There are certain limitations of the data
which inhibits the analysis of intra household allocation. Listing the few includes:

 First, questions in CES are framed at the household level and not at the individual level.
Per capita expenditure is calculated on the basis of adult equivalence scale.
 Second, questions regarding individual activity status have not been asked in the CES for
every members of the household.
 Third, although expenditure on education and health is asked in CES but it contains no
clue of type of institutions and also for each individuals within a household.

5
  Fourth, although in EUS information on activity status of every household member is
collected but the expenditure data is not a detailed expenditure across each individual,
which restricts the even tentative intra household analysis.
 Fifth, the question regarding educational attainment of each household member is
probed in CECs, but hardly reflects other information such as quality of institution,
expenditure on each individual information. Such details on education are collected in
special rounds on education and thus can be used for calculating intra-household
inequalities. But these variations can be measured not for all household members but only
for those currently attending educational institutions.
 Sixth, certain questions like decision making in resources allocation is not available
which is very crucial for examining the possible indicator(s) of variations in intra
household consumption and other indicators.

NSSO’s unit record data on CES and EUS of various thick rounds will be utilized to study the
different household indicators for examining the intra household variations. Given the objectives
of the study and the limitations of the available secondary information, it becomes essential to
collect data through an in depth primary survey for measuring intra-household variations in
consumption, education and economic empowerment among others from a same set of sample
households. The sample design and sample selection of the primary survey has been discussed in
Section 3.2 below.

3.1 Empirical Outline of Secondary Data Analysis

In the last two decades, the availability of unit record data and advancement in the software
analysis has strongly revived the interest in micro econometric analysis of human behavior.
However, still the ultimate unit of analysis and policy design has been households as the
questions are probed at the household level. Even the target of state welfare transfer also
remained the household given the intra household information asymmetries and difficulties in
individual targeting. In this study we will try to analyse within household dynamics and its
impact on intra-household variations in consumption and other indicators that includes
educational attainment and employment outcome.

For consumption and employment, we plan to use NSSO 68 th Round; for educational
development, NSSO 71st Round shall be used. We also plan to pool central and state samples of

6
above mentioned rounds for better representation of regions, gender, social groups and economic
categories. For this, we would seek help of DES for making available pooled data.

3.1(a) For Consumption Variation

Technically, intra-household inequality analysis warrants individual specific information of the

household. But the drawbacks of NSS consumption expenditure data is that it is framed and
collected at the household level and thus provided individual specific information only for single
member household. However, borrowing theoretical insights variations in the household
consumption will be examined. As mentioned above, one of the important observations
regarding intra household inequality is the prevalence of Kuznets curve relationship. It implies
increase in intra-household inequality initially with increase in income/resources and then
decline after reaching a limit. Haddad and Kanbur (1990) have found empirical support for this
hypothesis empirically from Philippines. Using this insight, comparative analysis of expenditure
of the household will be analysed to see the various expenditures of intra-household variations
across similar typologies of households. The role of source of income on household consumption
decision (Lise and Seitz, 2007) as well as impact of decision making on pattern of consumption
expenditure (Iverson, 2011) can be prior for intra household inequality analysis.

Figure: Kuznets Curve Effect in Intra Household Inequality

Intra
Hous
ehol
Thus,
d information on family
Ineq demographic
ualit
y characteristics can be used
to design a household indicator
which can be used for grading of
households. Information
Household Income/Resource
available for individual member of
the household includes age,
sex, marital, status, educational level, relationship with head etc. Apart from this certain

7
household level characters include social groups, principal livelihood source, assets and place of
residence, etc. Empirically,

HIin = f (hh_size; demo_chrcts; ear_memg)

Where HI is the household indicator which is calculated from the various combinations of the
three indicators includes: household size (hh_size), demographic characteristics (dem_chrtcs)
and working members whether male or female (ear_mem g)2. Since there has been considerable
variations in level of development of different regions of the state and social group it will also be
accounted. The subscript ‘i’ denotes the regions of UP and ‘n’ is for the social and economic
category of the household. Grading of the household can be done on different combinations of
household characteristics as given below.
Table: Tentative Combination of Different Indicators for Household Grading
Household Size Income source Income Level Youth/Adult
Agriculture/Non- Age composition of
  Agriculture & others Different Quintile Children
Single      
Single + 1 (couples or children)      
Couples + children (1 or 2)      
Couples + Children + Elder      
.      
.      
.      
.      
.      

For different households based on the HI, expenditure share on various food and non-food items
including education and medical expenditure will be examined to see variations with household
covariates. Here the notion of public and private expenditure discussed in literature will be

applied for various types of expenditure items. Public expenditures are those which are not
mutually exclusive and can be used by all household members like durable goods. Thus there is
very least possibility of any discrimination in different category of household will be chosen as
the base for analysis.
2
In demographic characteristics of the household

8
Since the expenditure is collected at household level, it could be function of both household
needs reflected by the demand as well as the income of the household. Expenditure of household
includes both food and non-food which includes education, health and other durables goods.
These durable goods are not mutually exclusive and can be used by every member and so
regarded here as public goods. Further expenditure on the remaining goods is competing
demands. So,

Ehh= Epub + Epriv

Three important component of private expenditure includes food, education and health. Demand
for food at the individual is determined apart from availability/accesability by type of
occupation, body metabolism etc. Precisely this is the reason to see the nutrients and its impact
on health outcome of individual which may not be posssible to be analysed from NSS CES data.
So here, the expenditure on non-food items like on health and education will be examined to see
the intra household variations. Private household expenditure (Epriv) is determined by many
factors like income of household, gender preference, decision making in the household etc.
Specifically,

Epriv = f (y, p, b)

Where y is household income/resource, p preference for boy or girl and b is the decision making
or bargaining power of different individual. According to different typology of the households,
expenditure composition will be calculated. Suppose there are n types of households based on
the criteria HI1, HI2, HI3….HInand so each household will be having distinct consumption
expenditure yi1, yi2, yi3, yi4 ….yinwhere i shows the different category of expenditure discussed
above.

9
 Expenditure distribution on different items for a household category (HI1)
HouseholExpenditure on Expenditure on Expenditure on . .
d category
items in category items in category items in category
‘1’ ‘2’ ‘3’
HI1 y11 y21 y31 . .
HI1 y12 y22 y32 . .
HI1 y13 y23 y33 . .
. . . . . .
HI1 y1n y2n y3n . .
Now given the distribution of expenditure on various item of expenditure mean can be
computed. Specifically,
E ( Y 1 i )=∫ f ( y )dy
Similarly, different inequality measure of the household like coefficient of variation for different
distribution will be calculated:
E ( y 1 i)
CV ( y 1i ) =
σ1i
E ( y 2 i)
CV ( y 2i ) =
σ2i
.
.
.
E( y jn )
CV ( y jn )=
σ jn
Where j is the number of household category and n is the category of expenditure items.
Coefficient of variation of different distribution for different typology of household will show
the level of inequality as compared to aggregate inequality measured by similar indicator. It will
also be insightful in verifying the theoretical predications of Kuznets curve in intra household
inequality.

3.1(b) Educational and others economic attainment

For analyzing intra-household variations in educational attainment data from EUS 68 th round as
well as 71st round NSS data on education expenditure will be used. EUS unit record data
provides information on current attendance in educational institution of individual aged between
0 to 29 years and educational level of all the members of the household. However, it does not

10
collect the information on type of institution which is crucial for understanding the intra-
household allocation. The 71st round NSS data collects and provides information on the type of
educational institution and some indicators to define quality of education. As we know, gender
within the household has remained one of the most important indicators of inequality in intra-
household resource allocation. There has been certain theoretical explanation like long-run utility
maximization strategy of household for differential investment on education and health of male
and female children. Given the resource constraints and cost of education, parents will invest in
the children education keeping in mind also the long run return (Sen and Sengupta, 1983).
Against this under common preference assumption of the unitary model of resource allocation
there will not be any difference in the education expenditure of the children irrespective of the
gender.

Here the attempt will be to see variation in intra-household educational attainment which will be
measured by ‘mean year of schooling of individual’ (Sm). Sm will be calculated for different type
of household graded on the basis of indicators discussed above. At individual level, S m will be
computed separately for male and female in the household and this will be adjusted by the age of
the individual. Specifically,

Member of HH Mean schooling (Sm) Age (a) (Smadj) = Sm/age

1 Sm1 a1 Smadj1
2 Sm2 a2 Smadj2
3 Sm3 a3 Smadj3
. . . .
. . . .

Thus adjusted mean schooling (Smadj) of individuals in different categories of the household will
be computed to find the mean gender difference (Dav). Empirically,

Dav.hhi = E (Smadj) m – E (Smadj) f

Since educational attainment is the outcome variable and pro-male bias preference may affect the
welfare of an individual with different gender over the whole life cycle. Welfare may be in the
form of earnings from the employment opportunities as labour market outcome in the form of
quality employment is highly correlated with the educational level of the individual (Schultz,

11
1961). So to see the intra household variations in employment opportunities given the
educational level particularly from gender perspective the data EUS will be source of analysis.

Here also the strategy of analysis will be same as discussed above. Household grading will be
done on the basis of certain combination of indicators which will further be used for intra
household analysis.

3.2 Primary Survey: Sampling Design and Sample Size

For achieving precision in the measurement of intra-household variation in consumption,

education, employment and income, primary data shall be collected by using multi-staged
stratified random sampling. The sample will be representative of the state level population and
due care will be taken to ensure its representativeness of different social groups that includes
SCs, OBC and Others. Since Uttar Pradesh has very less ST populations that too concentrated in
few places so they may be taken only if they are staying in the sample village and urban block.

There are four economic regions in UP, namely, Western, Central, Eastern and Bundelkhand.
Region-wise equal sample villages and urban block will be selected. From each region, one
representative district will be selected and from the selected district six villages and two urban
blocks will be selected on the basis of probability to population size (PPS) sampling design.
Villages and urban block/ward will be arranged in ascending order on the basis of population. An
interval (N/n) factor will be calculated by dividing the number of total villages/urban ward (N) in
the district by the number of village/ward (n) to be selected. After selecting the first village/urban
ward randomly, every (N/n)th village/urban is selected until we get the required number of
village/urban.

12
 Figure: Sample Design Framework

Uttar Pradesh

Western Region Central Region Eastern Region Bundelkhand

District District District District

Village &Urban Village &Urban Village &Urban Village

HH HH HH HH

To say with 95% level of confidence that our estimates fall within the 95% of confidence
interval, we have used a scientific method to calculate the minimum sample size for our lowest
unit of analyses that is region.

n = {z2 * σ2 * [N / (N - 1)]} / {ME2 + [z2 * σ2 / (N - 1)]}

Where z=1.96, N= Population Size of the Regional, ME=0.05, σ =0.5 (heterogeneity of the
sample).

Based on the above equation, the minimum sample sizes for the each region vary from 398 to
400 households. With this regional sample, the total sample for four regions comes around 1800
households. However, taking into consideration the non-response rate as well as the
representation from all social and economic classes as well as diverse household size, a
minimum of 450 samples will be collected from each region. In the district the rural urban
sample may vary depending on the population proportion of the district between rural and urban
area. Thus, sample size for primary survey shall consist of four districts, 24 villages, eight urban
enumeration blocks, 1200 rural households and 600 urban households.

13
3.3. Questionnaire Design

Since one of important task of the study is to collect detailed individual level information from
sample households, questionnaire designing will be crucial task. As the study is an attempt to
identify certain indicators that are predominant in intra-household variations, questionnaire will
gather following information for each individual of household of a household, wherever
possible:

Tentative variables on which information will be collected:

 Household level information that will include: social groups, principal livelihood source,
household size, assets etc.
 Demographic characteristics of household members. It will include demographic
characteristics like sex, age, activity status, education etc.
 Detailed expenditure of household on food and non-food items which will be decided on
the basis of different NSS schedules.
 Detailed expenditure of individual member of the household on education and health
depending on the type of institution.
 Detailed information on expenditure of individual members on household on transport,
personal electronic gadgets such as computer, mobile phone, clothes intoxicants, etc.
 As preference and decision making has been an important indicator of intra household
variations, a schedule will be designed to analyse the household head gender preference
and role of female in household decision making.
 Households source(s) of Income: As consumption allocation depends on the source of
income, changes in the sources of income will affect consumption inequality. So multiple
source of income in the household will be probed at individual and household level
specifically apart from principal source of income.

II. Work Plan

Keeping in view the magnitude of work for collecting primary as well as secondary data, the
study shall be completed in seven months. The details of the work plan along with tentative
milestones are given in the following figure.

14
For completing the study, there shall be three core researchers including one Team Leader and
Principal Researcher. For primary data collection, a team of 20 well trained post-graduate
Research Investigators shall be hired for a period of one and half month. Another 10 persons
shall be hired for data entry job. In all, about 1600 person days shall be required to complete the
study. The composition of Team is provided in Form Tech-3.

We plan to organize a Consultation Workshop to share the preliminary results of the study. For
this Consultation Workshop we shall invite experts working on the area of inter- and- intra-
household inequalities along with senior officers from DES and CSO, Government of India. The
aim is to make our results and methodology robust, which then can be applied by other states in
calculating intra-household variations on select items, such as consumption, education and
employment.

15
 First fortnight  Seventh month
 Fourth month

 Sixth month
 Fifth month
 2nd Month

 3rd Month
 1st Month

 Tentative person days

 Average work hours

Sl.
No Work Plan

fortnightSecond
First fortnight

fortnightSecond

fortnightSecond

fortnightSecond

fortnightSecond

fortnightSecond

fortnightSecond
First fortnight

First fortnight

First fortnight

First fortnight
First fortnight
.

Literature review and

finalisation of research
1 methodology                             30 240
Secondary data collection
2 and processing                             45 360
3 Analysis of secondary data                           90 720
Preparation of survey
4 instruments                             15 120
Pre-testing and finalisation
5 of survey instruments                             10 120
6 Data collection                             600 6000
Data scrutiny and
7 consistency checks                             300 2400
Data entry and data
8 processing                             300 2400
Primary survey data
9 analysis                             60 480
10 Preparing first draft report                         90 720
Expert consultation
workshop on results of the
11 study                              
12 Finalisation of report                             45 360

16
17