Income Shocks and Investments in Human Capital1
(Job Market Paper)
January 2010
Abstract
How well can parents insure their children’s future? This paper aims at answering this question by
studying the link between income shocks and parental investments in children in terms of time and goods.
The paper presents three main contributions: (1) it estimates the degree of response to income shocks in
families with young children, without imposing an a priori insurance setup; (2) it analyzes empirically the
mechanism behind the degree of insurance found, in particular, the role of wealth and public transfers, and
heterogeneity in responses to shocks by education and family structure; (3) finally, it proposes a useful way
to use common information in the NLSY79 and the Consumer Expenditure Survey (CEX) and the American
Time Use Survey (ATUS) to combine these three data sets and construct a panel of income, expenditures
and time use.
I use local business cycles as exogenous variation to families’ resources. These are an unpredictable
component of county unemployment rate, which I obtain after removing year and county effects from the
time-series of county unemployment rate.
I find that (1) families only partially insure against income shocks, but expenditures in education of
children respond less to shocks than household consumption, as parents try to shield them against shocks
because investments may be complements across children’s life-cycle; (2) income elasticity of investments
in terms of time is larger in families with young children than in families where there are only school-age
children, because at early ages there is a larger substitutability between different uses of time; and (3)
better off families use savings to buffer against shocks whereas poor families resort on public transfers.
JEL Codes: D12, D91, I30.
Key words: Insurance, human capital, consumption.
1 Introduction
Parents influence their children through genetic inheritance but also by the time and financial resources
dedicated to them. While genes are hard to change, resources may vary over time. The main question
addressed in this paper is the following: how well do parents shield children from fluctuations in family
resources? This involves understanding whether time investments and goods expenditures in children change
substantially with income shocks; whether the effects on child specific expenditures are different than effects
on nondurable consumption; and whether income shocks are transferred to a child’s human capital.
Understanding how parental investments in children respond to income shocks is important because
parents may face imperfect insurance against shocks (see Cochrane, 1991, Blundell, Pistaferri and Preston,
2008, for example). Furthermore, if imperfect insurance is coupled with a technology of skill formation where
the timing of investments matters (Cunha and Heckman, 2007), then income shocks at the beginning of a
1
,child’s life can have irreversible effects on her human capital. Therefore, learning about households’ reaction
to shocks is informative for the design of policies targeting more disadvantaged families with young children.
Although there has been work documenting the relation between changes in income distributions and
consumption, and substantial evidence on differences in the educational attainment of children from different
socioeconomic backgrounds1, there are virtually no studies on the effects of changes in income on parental
investments in children2. One of the reasons for this gap in the literature is the lack of data sets that include
simultaneously information on family income and use of financial and time resources (respectively,
consumption and time use) and measures of human capital at several stages of a child’s development 3. In this
sense, this paper has a dual contribution for the literature: (1) it evaluates the degree of insurance of parents
with respect to investments in their children’s future, and (2) presents a practical method to combine three
widely used American data sets: the National Longitudinal Survey of the Youth 1979 (NLSY79), the Consumer
Expenditure Survey (CEX) and the American Time Use Survey (ATUS).
The role of imperfect insurance is well studied in the literature of consumption 4, but the addition of
parental investments in children to the model poses new challenges. First, investment decisions have
important dynamic implications. Parents are forward-looking and anticipate the effects of the allocation of
time and expenditures on their children adult behaviors and human capital; childhood experiences
accumulate over the life cycle and evolve into skills, work habits, or engagement in risky behaviors when
individuals reach adulthood. The relevant theoretical model has features of a life-cycle model of consumption
with nonseparability of utility over time, such as in models with habit persistence and durable goods 5. Those
investments that are complements over time have characteristics of habit persistence; investments that are
substitutes have characteristics of durable goods (see Heaton, 1993, Cunha and Heckman, 2007).
Second, investments in human capital can take the form of expenditures (in the form of school tuition,
1
See Duncan and Brooks-Gunn (1997) or Carneiro and Heckman (2003) for evidence. 2 Leibowitz, 1974, is one of the first papers
studying parental investment in children. It uses indicators of time instructing children and reading, finding a positive relation
between investments and children’s IQ. 3
Todd and Wolpin, 2003, develop a framework for estimating the relation between child achievement and family and school
investments under different levels of data availability. In this paper I focus on how changes in family resources change family inputs
and try to assess the extent to which these changes are passed onto children outcomes. 4
The hypothesis of complete markets has been rejected by data: see Attanasio and Davis, 1996, and Hayashi, Altonji and Kotlikoff,
1996. Cochrance, 1991, presents mixed evidence on the rejection of full insurance hypothesis. 5
Becker and Murphy, 1988, analyze a model for addictive behavior to rationalize the consumption of substances. In their model, as
in the context of skill formation, there is a large effect of past consumption of the good on current consumption.
books, clothing or toys) or time (spent reading or teaching children, helping with homework or trips to
museums and theaters). These different investments may generate different returns 2 and respond to different
incentives. The opportunity cost of time spent in recreational or educational child care is market wage;
children’s goods can be acquired in the market.
The desirability of social policies (e.g., cash transfers for families with children, free preschool school
programs or food assistance programs) depends crucially on how well households can privately insure
against idiosyncratic income shocks, which in turn depends on the access to financial markets. For example, if
parents cannot secure the resources to invest in their children early in their life, effects of negative
idiosyncratic shocks may be transferred to the following periods. Policies can be designed to overcome, at
2 Guryan, Hurst and Kearney, 2008, show that high educated parents spend more time with their children.
1
,least partially, the effects of negative shocks 3. However, it is important to study empirically what actual
households do when they receive income shocks for the effectiveness of policies. This paper is, to the best of
my knowledge, the first attempt of evaluating how families respond to income shocks using data on changes
in income, consumption, time dedicated to children and measures of child human capital.
To study the link between income shocks and parental investments in children I construct a new panel
data set combining information on family income from the Children of the NLSY79 (CNLSY) with expenditures
from CEX and time use measures from ATUS. I match multiple measures of parenting behavior, materialized
in financial and time investments in children available across each child’s life cycle and family characteristics
on the CNLSY with expenditures and time use measures obtained from cross-sectional data. The method is
based on the use of two data sets: (i) a primary data set where incomplete measures of investment in human
capital are observed, (ii) an auxiliary data which contains both the incomplete and aggregated measures of
investments. I, then, indexes that are interpretable in terms of uses of financial resources and time of parents.
Idiosyncratic income shocks are identified through local business cycles. More specifically, shocks are
constructed using the Bureau of Labor Statistics’ county unemployment rate after accounting for year and
county effects. The persistence of the shock is inferred by studying its time series properties. The use of this
variation has several advantages over statistical decomposition of income residuals. First, idiosyncratic
variation in income is identified by unpredicted shocks on county’s labor demand and does not rely on
specification assumptions. Second, most of the evidence of responses to income shocks relying in
decomposition of income residuals using U.S. data is based on samples of annual earnings or average hourly
wages for continuously working, continuously married males, ignoring risk associated with job loss or illness 4.
However, using narrowly defined samples is likely to understate effects of the shocks confronted by agents,
limiting the scope to study effects of policies to alleviate negative effects of shocks among poorer families.
Finally, this method allows to distinguish between the effect of positive and negative shocks. This distinction
is useful to study nonseparabilities in investments across periods. In particular, if life-cycle/permanent
income model (LC/PIH) fails because of liquidity constraints, then households will be more likely to violate
the LC/PIH when income is expected to growth (see Altonji and Siow, 1987, and Deaton, 1991): temporary
high income draws are smoothed by saving but negative shocks are not smoothed unless household has
wealth. If early investments complement later investments then parents increase investment in children if
they face a positive shock whereas smooth effects of temporary income declines (which is similar to behavior
in savings). Then, if families face negative shocks it is expected larger sensitivity on nondurable consumption
(if credit constrained) than in investments in human capital (unless some investments can be substituted by
others).
The identification strategy used does not come without costs, in particular, it does not allow to study the
effect of shocks with different persistency and the instrument used has a larger predictive power for changes
in earnings of more disadvantaged groups in the population.
My main findings can be summarized as follows. When there are surprise increases in the local
unemployment rate (1) there are little changes on expenditures in children’s education (even though families
3 For example, using the same data of the current paper, Currie and Thomas, 1995, and Carneiro and Ginja, 2008, show that Head
Start (a U.S. preschool program for poor children) may partially compensate effects of early deprivation. The first paper finds positive
effects of the program on measures of cognitive skills; the later shows that the effects on schooling achievement and crime persist
until later adolescence.
4 See for example, Lillard and Weiss, 1979, Macurdy, 1982, Meghir and Pistaferri, 2004, Blundell, Pistaferri and Preston, 2008
2
, can only partially insure the effects of income shocks), (2) families substitute time spent in children’s
educational activities for leisure activities; and, (3) the effects of shocks on measures of child human capital
(are imprecisely estimated but) suggest that effects of shocks are more likely to be transferred to
noncognitive skills than to cognitive skills. I study different responses to shocks by type of shock (positive or
negative), structure of age of children in family and mothers’ education. In particular, (i) families of college of
educated mothers rely on accumulated assets as buffer to shocks, whereas the non-college group uses welfare
income, and (ii) transmission of shocks to human capital only occurs if shock takes place before child turns 10
and in families of less educated mothers. When facing a negative shock parents spend more time in leisure
activities with their children, however, there is no evidence of changes in time spent in education related
activities. If I allow the effects of shock to vary with the age of child I find that parents of children under age 5
are more likely to change their allocation of time in response to shocks, substituting time in education by
leisure with their children.5 This reaction is driven by the group of families of non-college mothers and can be
explained by a larger substitutability in parents’ use of time with children when these are younger: school age
children have a more rigid distribution of their daily time. In terms of policy, parents re-allocation of time
when facing negative suggests that cash-transfers may be insufficient to compensate for the effects of negative
shocks in early childhood, so that they should be coupled with in-kind programs such as Head Start or Perry
Preschool Program (which have been shown to have lasting effects 10).
The paper proceeds as follows. Section 2 presents a brief review of relevant literature. Section 3 includes
the predictions of a life-cycle model augmented to allow for altruistic parents that invest in their children.
Section 4 develops a unique panel data of children’s family income, labor supply, expenditures, time allocation
and measures of child human capital to quantify the effects of unexpected changes in family income. Section 5
describes the empirical approach to analyze the link between income shocks and investment decisions. I
discuss the econometric assumptions on families’ information set that allow the use of local labor market
shocks as exogenous variation for idiosyncratic shocks. Section 6 carries out several tests of formal tests of
consumption insurance. Section 7 concludes.
2 Related Literature
There is a growing literature that tries to assess how poverty affects children’s well-being and the role of anti-
poverty programs ameliorating the effects on negative shocks suffered by their parents. In parallel, there has
been an increasing amount of empirical work that rejects the hypothesis of full insurance by testing
implications of the life-cycle model (e.g., Cochrance, 1991, Mace, 1991, Townsend, 1994, Hayashi et al., 1996,
Blundell, Pistaferri and Preston, 2008). This paper links these two streams of the literature by analyzing one
possible channel that links family’s resources to children development: how well can parents insure their
children when they face shocks to their resources? And, try to analyze to which extent are these shocks
transmitted to child’s human capital.
There has been work trying to income to children outcomes. Most studies suggest that income at early
years have effects on adult ability and schooling outcomes. Permanent income has strong effect on children
outcomes, and income is specially important for disadvantage children in their early years (see Duncan and
5 The measure of time spent instructing children is broad, and varies across children’s life cycle. See Appendix A. 10See
evidence on the effects of early interventions surveyed in Cunha, Heckman, Lochner and Masterov, 2006.
3
(Job Market Paper)
January 2010
Abstract
How well can parents insure their children’s future? This paper aims at answering this question by
studying the link between income shocks and parental investments in children in terms of time and goods.
The paper presents three main contributions: (1) it estimates the degree of response to income shocks in
families with young children, without imposing an a priori insurance setup; (2) it analyzes empirically the
mechanism behind the degree of insurance found, in particular, the role of wealth and public transfers, and
heterogeneity in responses to shocks by education and family structure; (3) finally, it proposes a useful way
to use common information in the NLSY79 and the Consumer Expenditure Survey (CEX) and the American
Time Use Survey (ATUS) to combine these three data sets and construct a panel of income, expenditures
and time use.
I use local business cycles as exogenous variation to families’ resources. These are an unpredictable
component of county unemployment rate, which I obtain after removing year and county effects from the
time-series of county unemployment rate.
I find that (1) families only partially insure against income shocks, but expenditures in education of
children respond less to shocks than household consumption, as parents try to shield them against shocks
because investments may be complements across children’s life-cycle; (2) income elasticity of investments
in terms of time is larger in families with young children than in families where there are only school-age
children, because at early ages there is a larger substitutability between different uses of time; and (3)
better off families use savings to buffer against shocks whereas poor families resort on public transfers.
JEL Codes: D12, D91, I30.
Key words: Insurance, human capital, consumption.
1 Introduction
Parents influence their children through genetic inheritance but also by the time and financial resources
dedicated to them. While genes are hard to change, resources may vary over time. The main question
addressed in this paper is the following: how well do parents shield children from fluctuations in family
resources? This involves understanding whether time investments and goods expenditures in children change
substantially with income shocks; whether the effects on child specific expenditures are different than effects
on nondurable consumption; and whether income shocks are transferred to a child’s human capital.
Understanding how parental investments in children respond to income shocks is important because
parents may face imperfect insurance against shocks (see Cochrane, 1991, Blundell, Pistaferri and Preston,
2008, for example). Furthermore, if imperfect insurance is coupled with a technology of skill formation where
the timing of investments matters (Cunha and Heckman, 2007), then income shocks at the beginning of a
1
,child’s life can have irreversible effects on her human capital. Therefore, learning about households’ reaction
to shocks is informative for the design of policies targeting more disadvantaged families with young children.
Although there has been work documenting the relation between changes in income distributions and
consumption, and substantial evidence on differences in the educational attainment of children from different
socioeconomic backgrounds1, there are virtually no studies on the effects of changes in income on parental
investments in children2. One of the reasons for this gap in the literature is the lack of data sets that include
simultaneously information on family income and use of financial and time resources (respectively,
consumption and time use) and measures of human capital at several stages of a child’s development 3. In this
sense, this paper has a dual contribution for the literature: (1) it evaluates the degree of insurance of parents
with respect to investments in their children’s future, and (2) presents a practical method to combine three
widely used American data sets: the National Longitudinal Survey of the Youth 1979 (NLSY79), the Consumer
Expenditure Survey (CEX) and the American Time Use Survey (ATUS).
The role of imperfect insurance is well studied in the literature of consumption 4, but the addition of
parental investments in children to the model poses new challenges. First, investment decisions have
important dynamic implications. Parents are forward-looking and anticipate the effects of the allocation of
time and expenditures on their children adult behaviors and human capital; childhood experiences
accumulate over the life cycle and evolve into skills, work habits, or engagement in risky behaviors when
individuals reach adulthood. The relevant theoretical model has features of a life-cycle model of consumption
with nonseparability of utility over time, such as in models with habit persistence and durable goods 5. Those
investments that are complements over time have characteristics of habit persistence; investments that are
substitutes have characteristics of durable goods (see Heaton, 1993, Cunha and Heckman, 2007).
Second, investments in human capital can take the form of expenditures (in the form of school tuition,
1
See Duncan and Brooks-Gunn (1997) or Carneiro and Heckman (2003) for evidence. 2 Leibowitz, 1974, is one of the first papers
studying parental investment in children. It uses indicators of time instructing children and reading, finding a positive relation
between investments and children’s IQ. 3
Todd and Wolpin, 2003, develop a framework for estimating the relation between child achievement and family and school
investments under different levels of data availability. In this paper I focus on how changes in family resources change family inputs
and try to assess the extent to which these changes are passed onto children outcomes. 4
The hypothesis of complete markets has been rejected by data: see Attanasio and Davis, 1996, and Hayashi, Altonji and Kotlikoff,
1996. Cochrance, 1991, presents mixed evidence on the rejection of full insurance hypothesis. 5
Becker and Murphy, 1988, analyze a model for addictive behavior to rationalize the consumption of substances. In their model, as
in the context of skill formation, there is a large effect of past consumption of the good on current consumption.
books, clothing or toys) or time (spent reading or teaching children, helping with homework or trips to
museums and theaters). These different investments may generate different returns 2 and respond to different
incentives. The opportunity cost of time spent in recreational or educational child care is market wage;
children’s goods can be acquired in the market.
The desirability of social policies (e.g., cash transfers for families with children, free preschool school
programs or food assistance programs) depends crucially on how well households can privately insure
against idiosyncratic income shocks, which in turn depends on the access to financial markets. For example, if
parents cannot secure the resources to invest in their children early in their life, effects of negative
idiosyncratic shocks may be transferred to the following periods. Policies can be designed to overcome, at
2 Guryan, Hurst and Kearney, 2008, show that high educated parents spend more time with their children.
1
,least partially, the effects of negative shocks 3. However, it is important to study empirically what actual
households do when they receive income shocks for the effectiveness of policies. This paper is, to the best of
my knowledge, the first attempt of evaluating how families respond to income shocks using data on changes
in income, consumption, time dedicated to children and measures of child human capital.
To study the link between income shocks and parental investments in children I construct a new panel
data set combining information on family income from the Children of the NLSY79 (CNLSY) with expenditures
from CEX and time use measures from ATUS. I match multiple measures of parenting behavior, materialized
in financial and time investments in children available across each child’s life cycle and family characteristics
on the CNLSY with expenditures and time use measures obtained from cross-sectional data. The method is
based on the use of two data sets: (i) a primary data set where incomplete measures of investment in human
capital are observed, (ii) an auxiliary data which contains both the incomplete and aggregated measures of
investments. I, then, indexes that are interpretable in terms of uses of financial resources and time of parents.
Idiosyncratic income shocks are identified through local business cycles. More specifically, shocks are
constructed using the Bureau of Labor Statistics’ county unemployment rate after accounting for year and
county effects. The persistence of the shock is inferred by studying its time series properties. The use of this
variation has several advantages over statistical decomposition of income residuals. First, idiosyncratic
variation in income is identified by unpredicted shocks on county’s labor demand and does not rely on
specification assumptions. Second, most of the evidence of responses to income shocks relying in
decomposition of income residuals using U.S. data is based on samples of annual earnings or average hourly
wages for continuously working, continuously married males, ignoring risk associated with job loss or illness 4.
However, using narrowly defined samples is likely to understate effects of the shocks confronted by agents,
limiting the scope to study effects of policies to alleviate negative effects of shocks among poorer families.
Finally, this method allows to distinguish between the effect of positive and negative shocks. This distinction
is useful to study nonseparabilities in investments across periods. In particular, if life-cycle/permanent
income model (LC/PIH) fails because of liquidity constraints, then households will be more likely to violate
the LC/PIH when income is expected to growth (see Altonji and Siow, 1987, and Deaton, 1991): temporary
high income draws are smoothed by saving but negative shocks are not smoothed unless household has
wealth. If early investments complement later investments then parents increase investment in children if
they face a positive shock whereas smooth effects of temporary income declines (which is similar to behavior
in savings). Then, if families face negative shocks it is expected larger sensitivity on nondurable consumption
(if credit constrained) than in investments in human capital (unless some investments can be substituted by
others).
The identification strategy used does not come without costs, in particular, it does not allow to study the
effect of shocks with different persistency and the instrument used has a larger predictive power for changes
in earnings of more disadvantaged groups in the population.
My main findings can be summarized as follows. When there are surprise increases in the local
unemployment rate (1) there are little changes on expenditures in children’s education (even though families
3 For example, using the same data of the current paper, Currie and Thomas, 1995, and Carneiro and Ginja, 2008, show that Head
Start (a U.S. preschool program for poor children) may partially compensate effects of early deprivation. The first paper finds positive
effects of the program on measures of cognitive skills; the later shows that the effects on schooling achievement and crime persist
until later adolescence.
4 See for example, Lillard and Weiss, 1979, Macurdy, 1982, Meghir and Pistaferri, 2004, Blundell, Pistaferri and Preston, 2008
2
, can only partially insure the effects of income shocks), (2) families substitute time spent in children’s
educational activities for leisure activities; and, (3) the effects of shocks on measures of child human capital
(are imprecisely estimated but) suggest that effects of shocks are more likely to be transferred to
noncognitive skills than to cognitive skills. I study different responses to shocks by type of shock (positive or
negative), structure of age of children in family and mothers’ education. In particular, (i) families of college of
educated mothers rely on accumulated assets as buffer to shocks, whereas the non-college group uses welfare
income, and (ii) transmission of shocks to human capital only occurs if shock takes place before child turns 10
and in families of less educated mothers. When facing a negative shock parents spend more time in leisure
activities with their children, however, there is no evidence of changes in time spent in education related
activities. If I allow the effects of shock to vary with the age of child I find that parents of children under age 5
are more likely to change their allocation of time in response to shocks, substituting time in education by
leisure with their children.5 This reaction is driven by the group of families of non-college mothers and can be
explained by a larger substitutability in parents’ use of time with children when these are younger: school age
children have a more rigid distribution of their daily time. In terms of policy, parents re-allocation of time
when facing negative suggests that cash-transfers may be insufficient to compensate for the effects of negative
shocks in early childhood, so that they should be coupled with in-kind programs such as Head Start or Perry
Preschool Program (which have been shown to have lasting effects 10).
The paper proceeds as follows. Section 2 presents a brief review of relevant literature. Section 3 includes
the predictions of a life-cycle model augmented to allow for altruistic parents that invest in their children.
Section 4 develops a unique panel data of children’s family income, labor supply, expenditures, time allocation
and measures of child human capital to quantify the effects of unexpected changes in family income. Section 5
describes the empirical approach to analyze the link between income shocks and investment decisions. I
discuss the econometric assumptions on families’ information set that allow the use of local labor market
shocks as exogenous variation for idiosyncratic shocks. Section 6 carries out several tests of formal tests of
consumption insurance. Section 7 concludes.
2 Related Literature
There is a growing literature that tries to assess how poverty affects children’s well-being and the role of anti-
poverty programs ameliorating the effects on negative shocks suffered by their parents. In parallel, there has
been an increasing amount of empirical work that rejects the hypothesis of full insurance by testing
implications of the life-cycle model (e.g., Cochrance, 1991, Mace, 1991, Townsend, 1994, Hayashi et al., 1996,
Blundell, Pistaferri and Preston, 2008). This paper links these two streams of the literature by analyzing one
possible channel that links family’s resources to children development: how well can parents insure their
children when they face shocks to their resources? And, try to analyze to which extent are these shocks
transmitted to child’s human capital.
There has been work trying to income to children outcomes. Most studies suggest that income at early
years have effects on adult ability and schooling outcomes. Permanent income has strong effect on children
outcomes, and income is specially important for disadvantage children in their early years (see Duncan and
5 The measure of time spent instructing children is broad, and varies across children’s life cycle. See Appendix A. 10See
evidence on the effects of early interventions surveyed in Cunha, Heckman, Lochner and Masterov, 2006.
3
