Appendix: Cost Analysis for SPARCCS Survey For this analysis, we apply mathematical modeling to estimate the costs of both a one-contact strategy (baseline contact only), and a two-contact strategy (baseline and one recontact) – and then compare these to the four-contact strategy used in SPARCCS. Considering first the twocontact strategy, let ni , ri , and ci be, respectively, the sample size, the response rate, and the unit cost for Contact i , where i 1, 2 . Then, there will be r1n1 respondents and (1 r1 )n1 nonrespondents from Contact 1 (baseline). From Contact 2 (follow-up), we will obtain r2 (1 r1 )n1 respondents, and thus the overall respondent sample size is r1n1 r2 (1 r1 )n1 . The total data collection cost for the two-contact strategy is then CM c1n1 c2 (1 r1 )n1 . To obtain the same sample size for the single contact, the initial sample size should be r1n1 r2 (1 r1 )n1 r1 . Its data collection cost is then given by CS c1 r1n1 r2 (1 r1 )n1 r1 . Assuming that c1 c2 and r1 r2 , we compare the two total costs by computing the difference between them as follows: D c1n1 c2 (1 r1 )n1 c1 r1n1 r2 (1 r1 )n1 r1 c1n1 c1 (1 r1 )n1 c1 r1n1 r2 (1 r1 )n1 r1 if c2 c1 c1n1 r1 r12 r2 r1r2 r1 c1n1 1 r1 r1 r2 r1 0 if r1 r2 From the above inequality, we can deduce that the two-contact strategy is more expensive. For the K-cycle strategy, assuming that c1 ci and r1 ri for i 1, 2,..., K , we obtain the following cost difference expression: D c1n1 1 r1 r1 r2 1 r1 1 r2 r1 r3 ... 1 r1 1 r2 1 rK 1 r1 rK 0 r1 1 Depending on the data collection mode, the data capture cost for respondents should be added to the total cost (for example, sometimes an incentive is promised and provided only to respondents). Let c0 be the unit cost for data capture and other expenses applicable to the respondents only. Then the additional cost for both strategies is CA c0{r1n1 r2 (1 r1 )n1} when the two strategies are compared. In general the additional cost is given by C A c0 m , where m is the target sample size, which is the same for both strategies. Therefore, the cost comparison does not change because CA is cancelled out in D . However, the relative cost saving is affected, and the saving by the one-contact strategy would be diminished when there is an additional cost for the respondents only. In the following table we consider a very simple (and conservative) case, where the unit cost -arbitrarily set at $1.00, for convenience-- is the same for all contacts. Based on this analysis, a one-contact strategy would save 32% compared to the four-contact strategy ($6,291 versus $9,315). (Note that these are not actual dollar costs, but relative values based on our establishment of $1.00 as the unit cost.) One-contact strategy SPARCCS (four contact cycle) strategy First contact Second contact Sample Size Response Rate Resp Sample Size Cost 6,291 0.350 2,202 $6,291 Sample Size Response Rate Resp Sample Size Cost 3,574 0.350 1,251 $3,574 2,323 0.217 504 $2,323 Third contact 1,819 0.121 221 $1,819 Fourth contact 1,598 0.141 226 $1,598 2,202 $9,315 Total Note: Costs listed are not actual survey field cost, but relative values based on setting unit cost to $1.00 at each contact. 2 However, this calculation does not account for incremental costs associated with respondents (but not non-respondents) such as data capture and processing. As a more realistic approach, we recognize that SPARCCS was a mail survey with an incentive provided to all potential respondents (although only 9 percent of non-respondents cashed the incentive check), so the cost ratio between the two strategies depends on the cost associated with respondents only. If the additional unit cost for respondents is again set at $1.00, then the total cost under this scenario is $8,493 (= $6,291 + $2,202) for the single-contact strategy and $11,517 for the SPARCCS fourcycle strategy. In this case, the cost ratio would be 74 percent, resulting in a cost savings for the single-contact strategy of 27 percent. If the additional unit cost for respondents is set at $2.00, then the cost ratio would be 78 percent, and the savings reduced to 22 percent. The overall savings may even be higher than these figures indicate, because the overhead cost for the multicycle strategy is usually higher. However, we conservatively estimate that a single-contact strategy for SPARCCS would have resulted in a savings intermediate between the two values cited above, or (22 + 27)/2 = approximately 25%, relative to the four-cycle strategy that was implemented. As an alternative approach, we can also compare a two-contact strategy (initial contact and one recontact cycle) with the SPARCCS four-cycle strategy. In this case, the equivalent cost reductions, based on assumptions identical to those above, are estimated at 17 and 14 percent, respectively, for $1 and $2 respondent unit costs (see the table below). Conservative, we again take the average of these values and estimate a total savings of 15 percent, relative to the SPARCCS four-cycle strategy. 3 Unit cost = $1.00 for contact Two-contact strategy First contact Second contact SPARCCS (4-cycle) strategy First contact Second contact Third contact Fourth contact Total Sample Size Response Rate Resp Sample Size Cost 4,485 0.350 1,570 $4,485 2,915 0.217 632 $2,915 2,202 $7,400 Sample Size Response Rate Resp Sample Size 3,574 0.350 1,251 2,323 0.217 504 1,819 0.121 221 1,598 0.141 226 2,202 Cost $3,574 $2,323 $1,819 $1,598 $9,315 Respondent cost with unit cost=$1.00 Respondent cost with unit cost=$2.00 $2,202 $4,404 Two-contact total cost with resp unit cost = $1.00 Two-contact total cost with resp unit cost = $2.00 SPARCCS total cost with resp unit cost = $1.00 SPARCCS total cost with resp unit cost = $2.00 Cost ratio with resp unit cost = $1.00 Cost ratio with resp unit cost = $2.00 $9,602 $11,804 $11,517 $13,719 83.4% 86.0% 4