Two of the most widely used early care and education program quality tools used in the field are the CLASS: Classroom Assessment Scoring System and ERS: Environmental Rating Scales. Is there an advantage to using one versus the other. In the state of Washington as part of their QRIS: Quality Rating and Improvement System they happen to utilize both. In a study validating their Licensing Decision Making System, I had the opportunity to see them used side by side and wanted to report the results here. In other separate studies conducted in Head Start, Georgia, and Pennsylvania I saw similar results but wanted to wait to have the CLASS and ERS side by side in a specific study.
Here is what I found in making that comparison. In comparing the CLASS head to head with the ERS the correlation between the two scales was r = .24; p < .0001; n = 385. So both scales had a statistically significant correlation which one would expect since they are both measuring classroom quality, albeit from different perspectives.
Where it becomes interesting is when one begins to compare the two with the Washington state QRIS correlations. The CLASS and QRIS is r = .12; p < .022; n = 385 while the ERS and QRIS is r = .39; p < .0001; n = 385. It appears that the ERS is more sensitive at discriminating differences in QRIS than the CLASS. I further tested this my running one-way ANOVAs: CLASS x QRIS: F = 10.71; p < .0001; n = 385 while the ERS x QRIS: F = 26.534; p < .0001; n = 385. Both are statistically significant but the ERS again shows a much larger F ratio than is the case with the CLASS. To delve more deeply into these differences required looking at some basic descriptive statistics, such as the mean, standard deviation, skewness, and kurtosis. The following chart shows the results.
As one can see from the descriptive statistics there are some major differences between the CLASS and the ERS in how the data distributions play out. The ERS clearly has more variance in their data distribution than the CLASS does. These results are consistent with other studies in analyzing the respective data distributions. I feel that these results are significant for other early care and education researchers, developmental psychologists, and regulatory scientists as they conduct similar studies utilizing these respective tools.