There is growing interest in observing teachers' classroom practice, for formative feedback to teachers, teacher performance evaluation and incentives, program impact evaluation, and research on the determinants of student learning. For observations to generate meaningful data, they must be carried out using valid instruments, standardized protocols, and trained observers. A number of different observation instruments have been used in the US, but only two have gained any traction to date in developing countries - the Classroom Assessment Scoring System, CLASS, and the Stallings classroom observation instrument. Although both have been validated in developing countries in connection with student learning outcomes, they have never been used in the same setting. This paper exploits an ongoing research program that used the CLASS instrument to assess 7 th grade mathematics classrooms in Chile. In this research we re--code the same teacher videos using the Stallings instrument. We generate the first global evidence on the comparability and validity of the dimensions of teacher practice they measure. Because CLASS is significantly more expensive and complex to use - its training materials are proprietary and observers require a high degree of skillthere is a policy and research interest in analyzing the degree to which the simpler, open--source, Stallings instrument produces similar results.We ranked the Chilean teachers on their scores in each of the CLASS instrument's three domains (classroom organization; emotional support and instructional support) and analyzed the resulting quartile distributions' correlations with key Stallings measures of teacher performance. We found that teachers' scores on the Classroom organization domain of CLASS were modestly correlated with the most important Stallings measures, in all cases in the directions expected and with statistical significance. Teachers' CLASS scores on classroom organization were positively correlated with their use of time for instruction (r=0.340, p=0.015) and negatively correlated with time spent on classroom management (r=--0.321, p=0.022). Teachers in the bottom quartile of the distribution averaged only 80% of time on instruction, while those in the top quartile averaged 90%. Time on classroom management in the top quartile was a very low 9% of class time, while the lowest quartile spent 17%. All differences were statistically significant at the 5% or 1% level.Teachers' scores on the CLASS classroom organization domain were also positively correlated with the Stallings measures of teachers' ability to keep students engaged. However, Stallings measures were not correlated with teachers' scores on the other two CLASS domains: instructional support and

The number of K–12 students experiencing homelessness is increasing across the country. Schools may serve as sources of support and stability for homeless children, but little is known about the types of schools that homeless students attend or about the communities in which they live. We investigate the context of student homelessness in Los Angeles by analyzing student-level administrative data from the Los Angeles Unified School District and publicly available data on neighborhoods and schools from school years 2008–2009 to 2016–2017. Our findings suggest that homeless students tend to be clustered within lower-achieving schools with higher concentrations of disadvantaged student groups and live in neighborhoods with higher concentrated disadvantage. Despite policy provisions to ensure stability, homeless students have high rates of school and residential mobility in the years they are homeless, and mobile students tend to move to less-disadvantaged schools. We conclude with policy implications to strengthen the implementation of the federal McKinney-Vento Act.

As student homelessness grows, little is known about how the timing and duration of homeless experiences shape K–12 outcomes. Using Los Angeles Unified School District administrative data, the authors examine whether these temporal dimensions of homelessness predict students’ achievement and attendance. The authors find that homeless students scored lower on math tests and missed more school days than students who were never homeless. These associations are largest when students experience homelessness in the year the outcomes are measured. Experiencing homelessness in only 1 year and in 2 or more years both negatively predict outcomes, but associations are larger for students experiencing homelessness in only 1 year.