Title: Shocks, household choices, and children
1Shocks, household choices, and childrens
education
Francesca Modena Ph.D. in Economics and
Management - CIFREM
2OUTLINE
- Preliminary analysis of the data
- Model the household choices to cope with shocks
- Shocks and implications for household
vulnerability - Shocks and implications for childrens education
3PRELIMINARY ANALYSIS OF THE DATA
4INTRODUCTION
Microeconomic shocks may push households into
poverty or exacerbate their existing poverty
status. Income shocks reduce household wealth not
only directly, but also indirectly as a
consequence of the measures used by the household
to overcome or insure against them. Hence it is
important to understand how households cope with
actual shocks and the possibility of future
shocks, and to evaluatewhich responses are
costlier for the household.
5Some issues to be considered
6Different types of shocks
- Idiosyncratic versus common shocks
- - Idiosyncratic shocks are household specific,
for example the death or sickness of a
householder. - - Aggregate shocks are common across any given
area this may be the case of crop loss or price
falls.
- Demographic versus non-demographic shocks
- Demographic shocks are for example death and
sickness - of a householder
7Different risk coping strategies
- Risk-sharing strategies
- formal institutions (insurance), and informal
mechanisms such as transfers between friends and
neighbours, and in general family/friend
assistance - Intertemporal consumption smoothing
- households smooth consumption through savings,
borrowing, accumulating and selling assets
8DATA
- Indonesia Family Life Survey Data (IFLS), 1993
- Data on household economic shocks
Type of shocks
Death of a householder
Sickness of a householder
Crop loss
Price falls
Business loss
Unemployment
Type of measures
Extra job
Loan
Asset sales
Family assistance
Use savings
Cut down on household expenses
9DESCRIPTIVE ANALYSIS OF DATA
Sample 6280 households, 3454 rural (after
dropping income outliers)
Nr shocks Nr shocks Nr households Nr households Perc Perc
0 0 4319 4319 68.77 68.77
At least one shock 1 1961 1509 31.22 24.03
At least one shock 2 1961 368 31.22 5.86
At least one shock 3 1961 72 31.22 1.15
At least one shock 4 1961 7 31.22 0.11
At least one shock 5 1961 5 31.22 0.08
6280 6280 100 100
10(No Transcript)
11COSTS OF DIFFERENT SHOCKS (Rural Sample)
Type of shock Cost (thousand rupiah) Cost (thousand rupiah) Cost (thousand rupiah) Cost ( HH income) Cost (tot village cost/tot village income)
Obs mean median median median
death 247 471.37 230 41.82 1.32
sickness 336 547.18 200 25.86 2.07
crop loss 507 311.60 135 31.75 2.91
price falls 208 248.04 100 20.40 0.13
- Demographic shocks are costlier than non
demographic shocks - Looking at the percentage on household income,
crop loss becomes costlier than sickness - Shocks have a great impact on household income
- At the aggregate level, crop loss has the
largest impact, even if the aggregate effects of
all these shocks are rather small
12Focusing on rural sample, this table shows the
percentage of households that used each measure
Type of shock Type of shock Type of shock Type of shock Type of shock Type of shock
Measure Taken death sickness crop loss price falls business loss Unemployment
Extra job 12.92 12.36 45.86 38.96 23.44 42.37
loan 26.2 31.18 21.03 18.88 23.44 28.81
Asset sales 28.41 33.43 18.97 15.66 31.25 11.86
Family assistance 33.58 21.63 7.07 4.02 7.81 25.42
Used savings 14.76 16.57 4.66 4.42 6.25 3.39
Cut down on household expenses 5.54 7.02 20.52 31.33 21.88 13.56
121.41 122.19 118.11 113.27 114.07 125.41
Business loss and unemployment affect only a
few households
13INFORMAL INSURANCE MECHANISMS
Type of shock Type of shock Type of shock Type of shock
Measure Taken death sickness crop loss price falls
Family assistance 33.58 21.63 7.07 4.02
Cut down on expenses 5.54 7.02 20.52 31.33
- Our data confirm what suggested in the literature
(see Alderman and Paxson, 1992) - - informal insurance mechanisms, such as family
and community assistance, may protect households
against idiosyncratic shocks thereby smoothing
consumption across households through
risk-sharing - the community may not provide an insurance if
all households in the same area face the same
shock (this could be the case of crop loss)
therefore households would be obliged to use
other means. In these cases, the consumption
reduction is more severe than in the case of an
idiosyncratic shock in response to which
households seem more able to smooth consumption
(see the percentage of households that cut down
on expenditures)
14ASSETS SALE
Type of shock Type of shock Type of shock Type of shock
Measure Taken death sickness crop loss price falls
Asset sales 28.41 33.43 18.97 15.66
Another measure that may be more useful in coping
with idiosyncratic shocks than with common shocks
is asset sale it gives less protection against
common shocks because when the majority of
households try to sell assets, their prices fall
(Morduch, 1994 Frankenberg, Smith and Thomas,
2002).
15LABOUR AS INSURANCE
Type of shock Type of shock Type of shock Type of shock
Measure Taken death sickness crop loss price falls
Extra job 12.92 12.36 45.86 38.96
- Labour supply response plays an important role in
the face of crop loss and price falls (see
Cameron and Worswick, 2003) - This is not the case for those shocks
(demographic shocks) that affect the households
labour force and may induce the household to use
alternative, possibly costlier, methods of
insurance (Kochar, 1995) - Sickness may increase the need for domestic labour
16SAVINGS
Type of shock Type of shock Type of shock Type of shock
Measure Taken death sickness crop loss price falls
Used savings 14.76 16.57 4.66 4.42
the percentage of households that use savings to
overcome shocks is, in general, low, in
particular for non-demographic shocks
Separating households into groups according to
the level on income
Type of shock Type of shock Type of shock Type of shock
death sickness crop loss price falls
Use savings TOP 20 23.81 35.71 10.98 14.29
Use savings BOTTOM 20 6.67 7.35 1.54 2.13
savings is one of the most used measures by the
richest 20 to cope with demographic shocks.
17Model the household choices to cope with shocks
18EMPIRICAL STRATEGY
- The aim of this work is to model the probability
that a household chooses a given measure to
overcome a given shock - I would like to control for some income
indicators - household income and household
expenditures are endogenous -
-
- estimate the permanent and transitory income
components allow us to solve the endogeneity
problem and to examine the different roles of
different income components
19MEASUREMENT OF PERMANENT AND TRANSITORY INCOME
Cameron and Worswicks approach (2003)
- Cameron and Worswick (2003), following Paxon
(1992), propose a different approach that treats
transitory income not as a residual but as a
function of a set of variables. - Where is household income, and
are vectors of variables that may affect
permanent and transitory components of income.
20EXTENSION OF THE CAMERON AND WORSWICKS APPROACH
Our approach extends the Cameron and Worswick
(2003) method including other shocks (not only
crop loss) in the estimation of the transitory
component
What Cameron and Worswick call transitory
component captures only certain negative shocks,
and the residual includes important income
information. Therefore, I prefer not to
distinguish between permanent and transitory
income, but to decompose income in three
components y1, y2 and y3, where y2 captures
negative hardships and y3 is the residual of the
income regression, i.e. the non-identified income
component.
21Logit equation probability that household h uses
the mode m to overcome shock s
- is the weighted average of households that
experienced the shock in a given area, this
variable controls for the commonality of shock in
the same village - are demographic variables (i.e. sex and
education of the household head) - , and are estimates of
household income components permanent component,
a component that reflect transitory negative
shocks, and the residual component
22Estimates of income components
- Variables used to identify permanent income (X1)
- - demographic (the number of household members in
each age categories, the number of adult members
in a number of education/gender categories) - - wealth (occupation of the household head, the
number of householders that earn a wage, a dummy
that identifies if there is a householder who has
a non-farm business, the value of land - - location dummy (Java-Sumatra or the rest of
Indonesia). - Variable used to identify negative transitory
component (X2) - dummy variables that indicate whether a
household experienced a shock in 1992-93 - Some shocks may affect income in a permanent way
- death of a householder is included in the
estimation of permanent income (y1) -
23INCOME EQUATION RESULTS
- Death and unemployment of a householder have a
significant effect on income - The dummy if a householder works in a non-farm
business is very relevant and significant (nt01),
such as the number of household members that earn
a wage/salary (N_empl) - Land value is significantly and positively
correlated to income - Secondary and higher education play an important
role in the income equation, both for male and
for female - The coefficients of household heads employment
type strengthen the positive correlation between
wages/salaries and income
24- The data we are examining have two important
features - Households have a number of possible ways of
responding to shocks (we aggregate to 6
responses). - Multiple responses are possible responses are
non-exclusive - Multivariate logit allows modelling of multiple
responses, but does not accommodate
non- exclusivity.
25Theoretical model
- Households respond in such a way as to minimize
the cost of shock adjustment - By analogy with the McFadden random utility
model, adjustment shocks chms (for household h
adjustment mode m to shock-type s) have a
deterministic component fhms and a random
component ?hms - where xh is a vector of household h
characteristics.
26Choice of adjustment mode
- We first consider the standard multinomial logit
case in which choices are exclusive. - Write household hs response to a shock of type s
as rhs and let phms be the probability rhs m. - Focusing on adjustment mode 1,
-
- Under the standard logit assumption that the
stochastic component ?hms follow an extreme value
(Gnedenko) distribution, the probabilities phms
are logistic -
27Non-exclusive responses
- We generalize the random adjustment cost
framework by introducing a household-specific
threshold th t(xh). - Suppose the cost minimizing adjustment mode is 1.
The household reports an adjustment mode set - The probability of choosing mode 1 is now
-
- Under the standard logit assumptions
28Shocks and implications forhousehold
vulnerability(Future research)
29SHOCKS AND VULNERABILITY
Different measures have different implications
for vulnerability, in the sense of increasing the
risk of entering poverty when faced with future
shocks. Some responses may destroy or reduce the
physical, financial, human or social capital of
the household more that others (as suggested by
Dercon, 2005).
Reducing childrens education or health
expenditure affects childrens human capital and
may worsen the effect of transitory shocks
Other measures, such as asset sale, may destroy
the household physical capital. Gold, usually in
the form of jewellery, is seen as an important
way to save money (Frankenberg, Smith and Thomas,
2003) and using it to mitigate transitory shocks
reduces the possibility for a household
subsequently to overcome long term shocks such as
the 1998 financial crisis.
30- The effects of shocks on future vulnerability
are relevant in particular for poor households. -
- e.g. the percentage of households that take a
loan is higher for the poorest 20 than for the
richest 20, in particular for a death of a
householder (this is not true for price falls,
probably because in this case poor people do not
have collaterals). - This may push poor households into a
poverty trap.
31There are three principal approaches to assessing
vulnerability (Hoddinott and Quisumbing, 2003)
- Vulnerability as expected poverty (VEP)
- vulnerabilityprobability that a household will
fall - into poverty in the future
- where z is the consumption poverty line
-
32- Vulnerability as low expcted utility (VEU)
- vulnerability the difference between the utility
derived from - some level of certainty-equivalent consumption,
zCE, at - and above which the household would not be
considered - vulnerable, and the expected utility of
consumption
33- Vulnerability as uninsured exposure to risk (VER)
- This approach asses whether observed shocks
causes a - household to deviate from expected welfare
- ? Estimates the effects of common and
idiosyncratic shocks on household consumption net
of the mitigating role played by private coping
strategies and public responses
34Shocks and implications for childrens education
35The effects of shocks on child labour/child
education
- Do households use child labour as an insurance
(by taking child out of school and sending them
to work)? - Link between shocks, uncertainty, credit
constraints and child labour/child schooling
(Beegle, Dehejia and Gatti, 2003 Duryea, Lam and
Levison, 2003 Fitzsimons, 2002 Jacoby and
Skoufias, 1997)
36- Child labour as the outcome of an
intra-household bargaining process (Basu, 1999)
collective models
Household allocation decisions are the outcome
of a bargaining process between parents or
parents and child there is no income pooling,
but householders agree on how to divide the
household income according to the individual
bargaining power. The only assumption is that
outcomes are Pareto efficient.
37- I will outline a simple one period cooperative
model to analyse the effect of shocks on
household human capital decisions.
38DATA
- taking child out of school as response to a shock
- why child has never been to school / stop going
to school (help parents earn money, help at
home, ) - child works during the school year / how many
hours per week - hours per day/per week spent in school and for
studying outside the school - For older than 10
- whether they receive a salary/income
- what primary work (professional, management,
farm-forestry, production line, transportation,
blue collar, etc.) work field
39HOUSEHOLD DECISION-MAKING (1)
Are you free to spend your money for household
expenses?
Male Female
Yes, all 43 53,7
Yes, some/daily 42 42,5
No 14,4 3,6
Apart from hh exp., can you spend your income
without consulting your partner?
Male Female
Yes 16,3 22,2
No 83,5 77,8
40- From a preliminary analysis (probit)
- the probability the mother keeps some income
depends mainly on households wealth (household
tot expend.) - the probability the father keeps some income
depends on HH wealth, but also on his education
(), and wifes education (-)
41HOUSEHOLD DECISION-MAKING (2)
CLOTHES Perc.
Father 1,9
Mother 19
Both 24,12
Both, mother 19
Both,father 4,5
Child 26,3
EDUCTION Perc.
Father 3,87
Mother 6,55
Both 51,68
Both, mother 13,38
Both,father 12,35