Projecting and Understanding the COVID Slide Through a COVID Change Calculator

• In Event: Who Gets to Learn? Student Achievement and Academic Growth

Abstract

Research consistently demonstrates that children experience summer learning loss every year whereby they lose proficiency in learned skills during the previous school year (Cooper et al., 1996). An unfortunate reality is that students from low income families fall further behind during the summer months more than students from higher-income families (Burkham et al., 2004). Income’s impact is situated in the faucet theory (Entwisle et al., 2001); that when school resources are off, high income families have resources to support their children’s development whereas low income families are more likely to suffer from declines in learning (Becker & Tomes, 1986). It is likely that COVID-19 related school closures will exacerbate summer slide for all students and especially vulnerable students (e.g., students from low-income families, students with disabilities, and dual language learners). In addition, it is likely that the impact of COVID-19 may further impact children's reading development such that they experience lower levels of gain than during the next school year.
Sample. We propose to leverage a representative, longitudinal database of reading data (called the Reading Archive Database; RAD) from nearly 1.5 million K-12 students in approximately 4,500 public schools that is maintained at the Florida Center for Reading Research (FCRR). We have identified 100,000 students who were followed from K-3. The racial/ethnic, poverty, English language learner, and special education status categories for this sample of 100,000 are representative of the state of Florida’s demographics.
Data Collection and Design. Students in the RAD were administered curriculum-based measures of reading (R-CBM), including letter name fluency, letter sound fluency, phoneme segmentation fluency, nonword fluency, and oral reading fluency, up to three times a year in September, January, and April.
Data Analysis. Post-stratification and geo-weighting methods will be used in all statistical models so that the resulting expected values in the C3 are representative of national demographics and reading achievement. ArcGIS data will be used to pull distance information from the 4,500 schools. Multilevel, individual growth curve analysis will be used to estimate the initial status and growth in each of the R-CBM from K-3. A variety of time-conditional, growth curve models will be used to understand the nature of change over time. Student-level residuals will be used to create 95% confidence intervals to allow for the determination, for example, as to whether students’ summer learning loss was within expectation or greater/less than expected. Cut points may then be applied to the residuals of slope to identify and create a practical index of learning gains. Following the time-conditional model, main effects and interactions among selected student risk factors will be included to project learning changes based on the risk factors. Significant interactions will allow the expected learning changes and algorithms to be calibrated more precisely for the C3 tool so that teachers can be better prepared to find students who are at greater risk for long-term reading deficits. Cross-level interactions with school factors (as well as possible classroom and district factors) will be explored.

Authors

• Yaacov Petscher, The Florida Center for Reading Research

• Alex Torgesen

• Aidan Ferguson

• Tasha Holden

• Sara Hart

• Donald Compton, Florida State University