Does Head Start improve children's life chances? Evidence from a regression discontinuity design
Citation: Jens Ludwig, Douglas L Miller (2007) Does Head Start improve children's life chances? Evidence from a regression discontinuity design. Quarterly Journal of Economics (RSS)
DOI (original publisher): 10.1162/qjec.122.1.159
Semantic Scholar (metadata): 10.1162/qjec.122.1.159
Sci-Hub (fulltext): 10.1162/qjec.122.1.159
Internet Archive Scholar (search for fulltext): Does Head Start improve children's life chances? Evidence from a regression discontinuity design
Tagged:
Summary
Head Start might help children's health by increasing parent involvement, nutrition, social services, mental health services, and health services more generally. Ludwig and Miller ask, Does Head Start result in increased life chances for recipients?
Ludwig and Miller take advantage of a natural experiment in the form of a discontiuinity design. Essentially, when Head Start was first launched, counties needed to apply to participate in the program. Because the federal government was concerned that the poorest counties would not participate, they gave grant writing assistance to the 300 poorest countries. Ludwig and Miller compare the experience in these 300 poorest countries to the next 300 poorest countries to identify the effect of head start.
They unpack their research question by asking two more specific questions:
- Does the offer of assistance in in securing Head Start funding result in lower child mortality rates?
- Does the offer of assistance in in securing Head Start funding result in higher degrees of educational attainment?
The authors uses a complicated dataset from a series of sources that include:
- The NARA Census (1965) that the government used to identifying the poorest counties that in order to reproduce which counties were giving assistance with their HS applications.
- NARA data on OEO expenditure (1967-1980 but only used 1968-1972) of other types during the same period which was used to find out if receiving HS was correlated with receiving other types of aid.
- US Census data from 1960 through 2000 to get schooling by age including a special tabulation that includes detailed schooling data by age, race and gender.
- County-level data from Vital Statistics -- a census of all death certificates in the United States with cause of death code.
- Individual level geo-coded "NELS" data that included a a nationally representative sample of 8th graders.
The authors model the underlying effect of poverty on their outcomes using a series of parametric (e.g., linear and quadratic) and non-parametric forms. Their "treatment effect" is a measure of which side of the "300 poorest countries" cutoff a country a group falls.
They found that providing Head Start seems to have put the 300 poorest counties on an equal footing with the national average in terms of child mortality (from causes that that the authors suggested could be prevented by head start). Essentially, this amounted to about one or two fewer deaths per 100,000 four-year-olds.
Their primary threat to validity is issues of selective migration which they use NELS data to help address and which they conclude was not driving their findings.