《Propensity Score Matching》由会员分享，可在线阅读，更多相关《Propensity Score Matching（18页珍藏版）》请在金锄头文库上搜索。
1、534. Propensity Score MatchingSummaryPropensity score matching (PSM) constructs a statistical comparison group that is based on a model of the probability of participating in the treatment, using observed characteristics. Participants are then matched on the basis of this probability, or pro-pensity
2、 score, to nonparticipants. The average treatment effect of the program is then calculated as the mean difference in outcomes across these two groups. The validity of PSM depends on two conditions: (a) conditional independence (namely, that unob-served factors do not affect participation) and (b) si
3、zable common support or overlap in propensity scores across the participant and nonparticipant samples.Different approaches are used to match participants and nonparticipants on the basis of the propensity score. They include nearest-neighbor (NN) matching, caliper and radius matching, stratifi cati
4、on and interval matching, and kernel matching and local lin-ear matching (LLM). Regression-based methods on the sample of participants and non-participants, using the propensity score as weights, can lead to more effi cient estimates.On its own, PSM is a useful approach when only observed characteri
5、stics are believed to affect program participation. Whether this belief is actually the case depends on the unique features of the program itself, in terms of targeting as well as individual takeup of the program. Assuming selection on observed characteristics is suffi ciently strong to determine pr
6、ogram participation, baseline data on a wide range of preprogram characteristics will allow the probability of participation based on observed characteristics to be specifi ed more precisely. Some tests can be conducted to assess the degree of selection bias or participation on unobserved characteri
7、stics.Learning ObjectivesAfter completing this chapter, the reader will be able to discuss Calculation of the propensity score and underlying assumptions needed to apply PSM Different methods for matching participants and nonparticipants in the area of common support Drawbacks of PSM and methods to
8、assess the degree of selection bias on unob-served characteristics Use of PSM in regression-based methods54Handbook on Impact EvaluationPSM and Its Practical UsesGiven concerns with the implementation of randomized evaluations, the approach is still a perfect impact evaluation method in theory. Thus
9、, when a treatment cannot be randomized, the next best thing to do is to try to mimic randomizationthat is, try to have an observational analogue of a randomized experiment. With matching methods, one tries to develop a counterfactual or control group that is as similar to the treatment group as pos
10、sible in terms of observed characteristics. The idea is to fi nd, from a large group of nonparticipants, individuals who are observationally similar to participants in terms of characteristics not affected by the program (these can include preprogram char-acteristics, for example, because those clea
11、rly are not affected by subsequent program participation). Each participant is matched with an observationally similar nonpartici-pant, and then the average difference in outcomes across the two groups is compared to get the program treatment effect. If one assumes that differences in participation
12、are based solely on differences in observed characteristics, and if enough nonparticipants are available to match with participants, the corresponding treatment effect can be measured even if treatment is not random.The problem is to credibly identify groups that look alike. Identifi cation is a pro
13、b-lem because even if households are matched along a vector, X, of different character-istics, one would rarely fi nd two households that are exactly similar to each other in terms of many characteristics. Because many possible characteristics exist, a common way of matching households is propensity
14、 score matching. In PSM, each participant is matched to a nonparticipant on the basis of a single propensity score, refl ecting the probability of participating conditional on their different observed characteristics X (see Rosenbaum and Rubin 1983). PSM therefore avoids the “curse of dimensional-it
15、y” associated with trying to match participants and nonparticipants on every possible characteristic when X is very large.What Does PSM Do?PSM constructs a statistical comparison group by modeling the probability of partici-pating in the program on the basis of observed characteristics unaffected by
16、 the pro-gram. Participants are then matched on the basis of this probability, or propensity score, to nonparticipants, using different methods outlined later in the chapter. The average treatment effect of the program is then calculated as the mean difference in outcomes across these two groups. On its own, PSM is useful when only observed characteris-tics are believed to affect program participation. This assumption hinges on the rules governing the targeting of the program, as well
