The sample was released in three waves so sampling adjustments could be made based on actual response and eligibility rates, described below, and to avoid sampling too many organizations.
Sample release I (November 2007):
The first release consisted of 122 churches and 119 nonprofit human service organizations that were drawn randomly within seven strata (three impacted areas for each of the two states of Louisiana and Mississippi, as well as one impacted area in Texas). Each stratum of churches was sorted in ascending order, beginning with state, county code, impacted area code (1, 2, or 3), denomination type, membership size, and attendance size. The strata of nonprofits were sorted in ascending order, beginning with state, county code, impact area code, organization type (direct service provider, community group, or facility), total expenses, and NTEE (i.e., activity) code. Systematic sampling was then performed with random seed numbers assigned to the strata to randomly select the first observation. From there, between 10 and 20 observations were drawn per stratum.
Sample release II (December 2007):
The second sample release consisted of 140 churches and 106 nonprofit organizations. The number of churches was higher than the sampled nonprofit groups since early indications suggested that the response rate would be lower for churches and congregations, which were harder to reach than nonprofits. Therefore, the decision was made to sample more churches to increase the chance of attaining our original completion goals. The sampling procedures applied were the same as in the first round of sampling, namely seven presorted strata from which 10 to 27 organizations were randomly selected per stratum. New random seed numbers were assigned to each stratum.
Sampling release III (January 2008):
The third sample release consisted of 72 churches and 56 nonprofit organizations. Again, more churches than nonprofits were represented in the sample to increase the probability that despite the lower response rate among churches, their higher proportion in the sample would increase the probability that the number of interviews completed with churches would be close to our goal of 100. The third release applied the same stratified sampling method used for generating the first and second sample releases.
Based on all three sample releases, a total of 615 organizations were drawn for the survey: 334 churches and 281 nonprofits.