链接：https://pan.baidu.com/s/19OVUBYTxr0GWgE_zTR2F-g 

提取码：l1a4 

 

In recent years, an increased emphasis has been placed on the use of random assignment studies 

to evaluate educational interventions. Random assignment is considered the gold standard in 

empirical evaluation work, because when implemented properly, it provides unbiased estimates 

of program impacts and is easy to understand and interpret. The recent emphasis on random 

assignment studies by the U.S. Department of Education’s Institute for Education Sciences has 

resulted in a large number of high-quality random assignment studies. Spybrook (2007) identified 55 randomized studies on a broad range of interventions that were under way at the time. 

Such studies provide rigorous estimates of program impacts and offer much useful information 

to the field of education as researchers and practitioners strive to improve the academic 

achievement of all children in the United States. 

However, for a variety of reasons, it is not always practical or feasible to implement a 

random assignment study. Sometimes it can be difficult to convince individuals, schools, or districts to participate in a random assignment study. Participants often view random assignment as 

unfair or are reluctant to deny their neediest schools or students access to an intervention that 

could prove beneficial (Orr, 1998). In some instances, the program itself encourages participants 

to focus their resources on the students or schools with the greatest need. For example, the legislation for the Reading First program (part of the No Child Left Behind Act) stipulated that states 

and Local Education Agencies (LEAs) direct their resources to schools with the highest poverty 

and lowest levels of achievement. Other times, stakeholders want to avoid the possibility of 

competing estimates of program impacts. Finally, random assignment requires that participants 

be randomly assigned prior to the start of program implementation. For a variety of reasons, 

some evaluations must be conducted after implementation of the program has already begun, 

and, as such, methods other than random assignment must be employed. 

For these reasons, it is imperative that the field of education continue to pursue and 

learn more about the methodological requirements of rigorous nonexperimental designs. Tom 

Cook has recently argued that a variety of nonexperimental methods can provide causal estimates that are comparable to those obtained from experiments (Cook, Shadish, and Wong, 

2008). One such nonexperimental approach that has been of widespread interest in recent years 

is regression discontinuity (RD). 

RD analysis applies to situations in which candidates are selected for treatment based 

on whether their value for a numeric rating (often called the rating variable) falls above or below a certain threshold or cut-point. For example, assignment to a treatment group might be determined by a school’s average achievement score on a statewide exam. Schools scoring below 

a certain threshold are selected for inclusion in the treatment group, and schools scoring above