Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Human Capital Development Before Age Five

Douglas Almond and Janet Currie

Columbia University

November 8, 2009

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Presentation outline

1 Introduction

2 Model

3 Methods

4 Empirical Literature on Prenatal Period

5 Empirical Literature on Ages 0 to 5

6 Policy Responses

7 Conclusions and Future Work

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Motivation

Active literature in economics over last fiveyears

Early childhood measures predict a non-trivialportion of adult outcomes

E.G., 1958 British Cohort Study: 20% of variation in wages atage 33 can be explained by child observables during earlychildhood (Currie & Thomas 1999)

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Conventional view of individual human capital over time

Grossman model considers health/human capitalas a stock variable that varies with anindividual’s age

Responds to investments and depreciation

Stocks at earlier ages matter to adult stocks...

... but their e!ect (and health investments they embody)dwindles over time

No special role for early childhood ages

We want to leave open whether there is indeed“fade out” of investments/experiences at earlyages

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Conventional view of individual human capital over time

Grossman model considers health/human capitalas a stock variable that varies with anindividual’s age

Responds to investments and depreciation

Stocks at earlier ages matter to adult stocks...... but their e!ect (and health investments they embody)dwindles over time

No special role for early childhood ages

We want to leave open whether there is indeed“fade out” of investments/experiences at earlyages

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Conventional view of individual human capital over time

Grossman model considers health/human capitalas a stock variable that varies with anindividual’s age

Responds to investments and depreciation

Stocks at earlier ages matter to adult stocks...... but their e!ect (and health investments they embody)dwindles over time

No special role for early childhood ages

We want to leave open whether there is indeed“fade out” of investments/experiences at earlyages

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Conventional view of individual human capital over time

Grossman model considers health/human capitalas a stock variable that varies with anindividual’s age

Responds to investments and depreciation

Stocks at earlier ages matter to adult stocks...... but their e!ect (and health investments they embody)dwindles over time

No special role for early childhood ages

We want to leave open whether there is indeed“fade out” of investments/experiences at earlyages

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Admitting early origins

Defining h as health or human capital at thecompletion of childhood

h = A[!I1 + (1! !)I2], (1)

I1 "= investments during childhood through age 5

I2 "= investments during childhood after age 5.

! can be greater than .5

Perfect substitutability in equation (1) notuncommon assumption in economics, butproblematic for early origins

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Complementarity

Heckman suggests more flexible “developmental”technology:

h = A!!I !

1 + (1! !)I !2

"1/!, (2)

Constant elasticity of substitution (CES)production function

Elasticity of substitution 1/(1! !)

Perfect substitutability of investments stillallowed (when " = 1)

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Shocks in early-childhood health

Holding other determinants of investments fixed,consider investment shock:

I1 + µg

Long-term damage from a negative µg is:

#h

#µg

Magnitude of damage can depend on levels of I1 and I2Relevant for empirical findings of heterogenous “early origins”damage“Biological” e!ect assumes no investment response

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Remediation

Consider a change to second period investmentsafter shock µg in early childhood:

E!ectiveness of remediation depends on ", !and I1Knowing "h

"µgto be big doesn’t say much about

e!ectiveness of remediation

Optimal investment response also depends onutility function, e.g.:

Up = (1! $)logCp + $logh

Where Cp is the consumption of parents

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Optimized investment response

"I #2 · µg < 0: Compensation helps o!set damage

"I #2 · µg > 0: Reinforcement accentuates damage

For " > 0, compensation optimal

For " < 0, reinforcement optimal

To the extent there is a response, then missing the“biological” e!ect.

Can understate total damage from µg byfocussing exclusively on reduced form "h

"µg

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Empirical work on investment responses

Early work arguably proxied for investments withlater-life outcomes (e.g., years of schoolingconsidered investment measure)

Recent work has begun to consider explicitinvestment measures at younger ages

No consensus on direction of response

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Data constraints

Solutions to lack of large-sample longitudinal data:1 Collect more data, but costly and

time-consuming (have to wait for cohorts tobecome adults)

2 Add retrospective questions to existing datacollections

3 Merge new information to existing datasetsE.G. Add ecological info to vital statistics data (e.g. pollutionmeasures)

4 Merge administrative data from several sourcesusing personal identifiers

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Data constraints: power

Looking for measures of both early childhoodand adult outcomes

Variable-rich datasets tend to have smallersample sizes

Less true in NorwayE!ect of 1% increase in birth weight increases HS completionby .1 percentage pointsUnder reasonable assumptions will need a sample size of atleast 4,000 to detect this

Need to be creative to find suitable data

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Biomedical Literature on Fetal Origins

DJ Barker says mom an inconsistent bu!er ofadversity during pregnancy

Insults experienced when development rapidhave bigger e!ects

Growth very rapid in utero

“Brain sparing” mechanisms often studied inepidemiology and biomedicine

Fetal “trunk” gets lower priority when there’s a shortageCauses chronic health conditions during adulthood (e.g.ischemic heart disease)E!ects may be latent during childlhood

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Birth weight and human capital

All observational studies in economics

Human capital assessed in later adolescence oradulthood

Sign of e!ect fairly well established

low birth weight $ low human capital

Relationship persists in sibling and twinscomparisons

Strength of relationship more wide-rangingDi!ering identification strategies across studiesParental responses may be di!erent in di!erent contexts

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Prenatal shocks and human capital

1 Prenatal infections% epidemiological literature finding e!ects of seasonal variationin infections during pregnancy on adult health

1957 ”Asian Flu” just happened to coincide with pregnancy for1958 British Birth Cohort

Kelly (2009) finds impacts on test scoresIndependent of birth weight e!ect of fluDoesn’t detect an investment response

2 Prenatal economic shocks

3 Prenatal air pollutionParticularly active economics literature

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Prenatal pollution shocks

Ambient pollutionLarge e!ects on multiple measures of health at birthFewer studies of long-term e!ectsData constraint

SmokingLarge birth weight e!ectsAnalysis of other outcomes, esp. long-term e!ects, relativelyuncommon

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Long-term e!ects of prenatal shocks often BIG

Example: 1918 Influenza Pandemic

High school graduation rate falls 13-15% for the children ofinfluenza-infected moms (roughly one-in-three moms infected)

Wages 5-9% lower for children of infected

Fig. 3.—1960 average years of schooling: men and women born in the United States

Fig. 4.—1970 high school graduation: by year of birth

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

From 0 to 5

What are the interesting questions?

It is not interesting to show that it is possiblefor an event that occurs between 0 and 5 tohave lifelong consequences(too easy to think of examples).

So overarching question must be how importantevents at this stage are relative to events earlieror later?

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Additional questions -

How much of adult disparity can be explained byevents from 0-5?

What are the connections between health andcognition (or are they mostly separateprocesses)?

What is the role of third factors (e.g. parenteducation and income?) in mitigating e!ects ofshocks?

How do parents respond to shocks?

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Identification Issues:

Not possible to define sharp cuto!s after whichinfluences will not matter.

Di#cult to distinguish prenatal and postnatalexposures.

Many possible early childhood events/influencesand many possible outcome measures whichmay be bundled – which are important? Whichcan be separately identified?

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Some findings:

Reductions in disease environment in earlychildhood associated with better health andhigher cognitive test scores in developedcountries (even given negatively selectedsurvivors).

The occurrence of negative health shocks isstrongly linked to SES internationally.Cumulative e!ects of negative health shockswiden gaps in health as children age. May alsobe linked to lower educational and labor marketattainment (an active research area).

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

More findings...

Mental health is really important relative tophysical health. Big e!ects and high prevalence.

Parent’s mental health (only moms reallystudied) also important.

Mom’s employment extensively andwrong-headedly studied. Production functionapproach suggests that it is the combination ofinputs that matters, so what is substituted formom’s time and quality of mom’s time isimportant.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Returning to overarching question:

We know a lot more than 10 years ago.

Factors from -1 to 5 have significant long-terme!ects on many children.

Prenatal e!ects are generally large relative tomany post-natal e!ects that have beendocumented (especially since it is not clear howmany conditions/events that are noticedpost-natally have prenatal roots).

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

So can we do anything about it?

Second half of the paper reviews literature onremediation.

We discuss:CashNear cashEarly interventionHealth insurance

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

1st half depressing. 2nd half surprisingly hopeful

There are many interventions which appear tohave positive e!ects.

With appropriate intervention it may be possibleto undo or prevent many of the harmfulconsequences of negative prenatal and early lifeevents.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Evidence re: Cash

Some recent studies suggest that relativelymodest transfers of cash to low-income U.S.households (around $10,000) would haveeducationally meaningful e!ects on test scoresand behavior problems.Some suggestion that girls benefit more in termsof health and behavior. Boys benefit more interms of test scores?Then why is so much aid in-kind? Politics.Paternalism. In-kind programs may be the onlyway to reach children in the most dysfunctionalhouseholds.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Evidence re: Cash

Some evidence that cash transfers as small asaround $10,000 may be enough to produceeducationally meaningful di!erences in testscores (especially for boys), and in behavior andhealth (especially for girls).

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Cash vs. In-Kind

Then why is so much aid delivered in kind?

PoliticsPaternalism

In kind may be the only way to reach the children who aremost at risk.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Near Cash: Housing and Food Stamps

MPC food out of food stamps slightly higherthan out of cash. Cash transfer associated withFSP (now SNAP) has had positive impacts onchild outcomes.

Housing programs also have had positive e!ects.Note, reducing crowding seems to benefit boys.Taking kids out of poor neighborhoods seems tobenefit girls.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Early Intervention

Many programs are e!ectiveQuality preschool programs (e!ects bigger for girls?)WIC (e!ects unlikely to be driven by selection)Nurse home visiting (randomized controlled trials)

Other programs are ine!ectiveHome visiting with para professionals (randomized controlledtrials)Average child care programsEarly Head Start?

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

An Economic Approach...

... may be useful in understanding the di!erentresults

Can get positive results with di!erentcombinations of inputs

Need more research from this perspective.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Public Health Insurance

Causes children to get more care when replacesno insurance. But more is not always better!(e.g. hospitalizations, C-sections).

Causes children to get less care when it crowdsout more generous private insurance policies.

Crowd out likely to increase over time (familiarity, increasingcost private HI)

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Health is a stock

Not a flow, so may be hard to see immediatee!ects in terms of better health, but someevidence that coverage from birth is associatedwith better health and higher test scores.

Note, HI typically does not address major threatsto health including injuries, toxic exposures,poor nutrition, poor parenting/abuse/neglect.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Summary and Conclusions

This is an exploding area of research

Economists bring a lot to the table in terms ofboth empirical methods and a simple theoreticalframework for interpreting results.

Substantively, what happens from -1 to 5 is atleast as important for future outcomes as whathappens thereafter.

Introduction Model Methods Prenatal Period Ages 0 to 5 Policy Responses Conclusions

Many questions for Future Work

Which shocks matter most and at which ages?

Why do e!ects di!er for boys and girls?

What are the interactions between shocks tohealth and shocks to cognition?

What is the least cost way to improveoutcomes?

How can Economists interact constructively withpeople from other disciplines (e.g. to usebiomarkers?, conduct RCTs?)