Palgrave Handbook of Econometrics: Applied Econometrics

610

Table 12.1

Studies that use matching estimators

Study

Outcome

Treatment

Method

Comment

Dano (2005)

Earnings, annualemployment rate,disposable income,public transfer income

Road traffic accidents

Usesdifference-in-differencesmatching estimator withpanel constructed fromregister data. One-to-onematching withoutreplacement based on linearindex from propensity score.Checks for balancing

Finds an impact onearned income for olderand low incomeindividuals. Also showscompensating effects ofpublic transfers in theDanish system

Frolich

et al.

(2004)

Labor market outcomes:reintegration into thelabor force

Vocational rehabilitationprogram in westernSweden

Multiple treatments withmatching based onmultivariate balancingscores computed frommultinomial probit models.Uses nearest neighbourswith replacement. Checksfor balancing of covariatesafter matching

Finds a negative effect ofrehabilitation. Many ofthe effects areinsignificant: manycontrols are usedrepeatedly in thematching

García-Gómez andLópez-Nicolás (2006)

Employment, income,SAH

Health shocks andemployment shocks

Use matcheddifference-in-differencesestimator. Combine exactmatching on pre-treatmentoutcomes with propensityscore matching. Use bothnearest neighbor andkernel-smoothed matching

Find effects of healthshocks on employmentand activity. Also find aneffect of transition tounemployment on SAH