|
Activity Number:
|
467
|
|
Type:
|
Contributed
|
|
Date/Time:
|
Thursday, August 7, 2003 : 10:30 AM to 12:20 PM
|
|
Sponsor:
|
Section on Survey Research Methods
|
| Abstract - #301318 |
|
Title:
|
Random Sampling for Impact Analysis in a Statistically Representative Set of Sites: The National Head Start Impact Study
|
|
Author(s):
|
Stephen Bell*+ and Mike Puma and Gary M. Shapiro
|
|
Companies:
|
The Urban Institute and Chesapeake Research and Westat
|
|
Address:
|
2100 M St. NW, Washington, DC, 20037,
|
|
Keywords:
|
policy impact ; evaluation design ; control group ; social experiment ; randomized control trial ; child development
|
|
Abstract:
|
This paper describes techniques for studying the impact of social programs using a national probability sample of sites and random assignment of individuals to program and non-program status. The authors sample 5,000 children in 400 communities to measure the impact of Head Start, the primary U.S. system for helping poor families meet the developmental needs of pre-school children. The paper explores both statistical and practical aspects of the challenge: (1) determining measure of size when the pool of "accepted" applicants grows to accommodate an experimental "control group"; (2) expected statistical precision when full compliance with assignment status (participate/do not participate) is unlikely and final Ns unknown; (3) implications of a "with/without" impact design, rather than a "pretest/post-test" design; (4) strategies for gaining local support for randomly-determined program intake decisions; and (5) meeting sample and program enrollment targets when partial sampling occurs at multiple points in time. The authors provide tools for future designers of nationally representative impact evaluations, a type of research increasing required by evidence-hungry policy makers.
|
