Page 12

Page 12 Weights were constructed to ensure the sample was representative of the sampling frame. As such, weights were utilized to adjust how much a given child care provider counted relative to others. For example, if small facilities were less likely to participate in the survey (e.g., because they had less free time), the responses of the small facilities that did participate would carry more weight to compensate for those that did not. These adjustments aimed to align the sample responses more closely with what would be expected with a 100% response rate. Specifically, weights were constructed using reported capacity (five levels: 1-40 children; 41-64 children; 65-100 children; 101+ children, or unknown), whether a facility was listed as a CCPP-participating provider (2 levels: yes; no), and 13 county groupings based on previously established Early Childhood Academy service areas (see Figure below). Due to empty cells in the cross-tabulation of these variables, inverse propensity weighting could not be used. Therefore, raking, a common weighting technique, was used to balance the sample on these variables. The weights produced were within a target range (1.25-4.57). Figure 1. Thirteen County Groupings used in Weighting Process Center-based Sample Weighting