Many social programs are evaluated by experiments which offer initial access to a focal program via lottery. In addition to the well-known issue of imperfect compliance, the existence of close alternatives complicates the experimental evaluation, especially if most control group individuals end up receiving similar services to the focal program (Heckman, Hohmann, Smith, and Khoo, 2000). Another challenge, which is important in evaluating programs that span multiple periods, is that people may choose different paths following the initial period of randomization. For example, those in the treatment group are not obliged to stay in the focal program and may switch to an alternative program; those in the control group who initially enrolled in an alternative program may switch to the focal program later.
This paper analyzes returns to the Head Start Program (HS), the largest and longest-running preschool education program for disadvantaged children aged 3–5 in the United States. Using experimental data from the Head Start Impact Study (HSIS), we examine the effects of childcare program pathways on cognitive outcomes. We find differential returns by the sequencing of program participation, and support engaging low-ability children with prolonged center-based care and high-ability children with some home care.
One central question then is how to redesign HS to induce more individuals of heterogenous ability to choose their “socially desirable” program sequence. We analyzed a counterfactual policy reform which had not been considered before but deemed important in the HSIS evaluation report: limiting the maximum duration of participation in HS to one year only (but keeping open the option to enrol at age 3 or 4). This is driven by pre-emptive responses at age 3 to “bank up” HS eligibility for possible enrolment at age 4, which is more common among low-ability children and worsens their outcomes. These results highlight the importance of modelling forward-looking behavior in evaluating public programs.