Introduction: Increasing evidence has shown that student mobility is negatively correlated with individual students' and schools' academic performance (Hanushek, Kain, & Rivkin, 2004; Lara, Mizala, & Repetto, 2011; Sun & Pan, 2014). However, the negative impact may be due to the switching behavior or it could be only a side effect of temporary adjustments. For example, students may naturally upgrade into higher level schools (this will be called a Structural switch) which will eventually lead to higher learning outcomes after a short transition. Students may need more help to adjust to a new environment after moving to a different school within the same level due to various reasons, including family migration, attractive school policies or new learning objectives (Frictional switch). School switch is also one of the most frequent schooling behaviors that changes students' memberships to a school. Most school impact models, however, ignore the student school switch behavior (Chung & Beretvas, 2012; Gilbert, Petscher, Compton, & Schatschneider, 2016; Luo & Kwok, 2012). Concerning the complicated nature of school choice/switch behaviors, it is essential to consider qualitatively different types of school switch and evaluate their effects in the long run. Research Aims: This study proposes the use of a cross-classified model to estimate the potential lasting effects on students' outcome of school switch and differentiate the effects among various kinds of the switch behaviors. To understand the impacts of the behaviors, the models focus on the attributes of the school switch, mainly the two forms of school switch behaviors discussed above, and the type of school the student is switching from and to (public to private, charter to public, etc.). Methods: Data: The data for this study is based on the national standardized test SIMCE (Education Quality Measurement System), which is administrated every year to all students in a specific grade. Our analysis will consider the class of 2019 who took the test in 4th to 10th grade from 2011 to 2017. The outcome variable will be the students' test scores in Reading (Spanish) and Math, which range from 110 to 397 in Reading and from 97 to 428 in Math. The school switching rates increases by grade. The identified frictional switches are more frequent than structural switches in 6th grade 8th grade, while over 40% of the students experience a natural update at 10th grade (2017). A summary of the data is shown in Table 1. A Random Coefficient Switch Effect Model: In this model multiple outcome scores are nested within students, while the students are cross-classified in different schools. It assumes the prior score is correlated with the subsequent outcome. Additionally, students may have switched schools between tests. A school switch matrix is specified to detect affiliation changes between outcome measurements. The model can be specified as y[subscript ijt] = [alpha]X[subscript i] + [beta subscript 0]t + [beta subscript 2i]w[subscript it]S[subscript it] + [delta subscript i] + [delta subscript j] [beta subscript 2i] = [beta subscript 1] + [beta subscript 1]ST[subscript it] + [delta subscript j] where, [delta subscript i]~N(0, [delta superscript 2 subscript 1], i=1, ...I) is the student-level random effects, and [delta subscript j]~N(0, [delta superscript 2 subscript 2], j=1, ...J) is the school-level random effects at time t. S[subscript it] is a I*T dummy matrix to indicate if there is a school switch for student i at time point t and ST[subscript it] is a categorical variable that represents the type of the switch. A weighted version of S[subscript it], w[subscript it]S[subscript it] can be specified for modeling the time-based lasting effects of a school switch. Two coding schemes will be compared in the preliminary analysis. The "Instant Switch Effect" assumes the switch only affects the outcome in the year that switch occurs, and the "Constant Switch Effect" extends the effect to all following years. Preliminary Results: The estimations (Table 2 for Reading and Table 3 for Math) indicate that switching schools would lead to negative effects on the testing scores. The effect sizes of an instant switch are -0.117 on Reading and -0.137 on Math testing scores. It is also expected the effect sizes of a constant switch is lower, which are -0.085 and -0.104 for Reading and Math, respectively. The switches effects vary across schools. The variances of the slopes of the instant school switches explained 12.08% of the school effects on Reading and 9.27% on the Math. The constant school switches explained 17.83% and 14.82% of the school effects on Reading and Math, respectively. When comparing the 2 types of switch, the frictional switch appears to have a larger negative effect on test scores than the structural switch. The same is true when the switch effects vary across schools, with the frictional switch having smaller variances than the structural switch. From our results, we also find a negative year effect on Reading and positive year effect on Math. Implications: The proposed model can enable education researchers and practitioners to capture the effects of school switch. Furthermore, the approach allows us to differentiate effects of various complicated switch behavior and incorporate the field experiences into the model to understand the phenomenon in a deeper sense. The products of the model can be used to identify schools with different transition experiences and provide information for both students and parents in their schooling decisions.