In a previous post I wrote about the slow steady decline in response rate on UM's Survey of Consumer Attitudes. Richard Curtin, Eleanor Singer (an MSI client) and Stanley Presser have just published (Journal of Official Statistics, 23, 91-106) the results of an incentive experiment designed to understand whether alarger incentive can stem the tide.
Over a series of experiments on the same study between 1996 and 1998 Eleanor and two other colleagues showed that a prepaid $5 incentive increased the response rate by about 12 percent, significantly reduced the calls needed to close out a case, and was cost effective when compared to both a promised $5 incentive and a simple advance letter (Public Opinion Quarterly, 64,171-188). In 2004, the present team essentially repeated the earlier experiments to measure the effect of raising the incentive to $10. Their findings are interesting:
- The $5 was as effective in 2004 as it had been in 1998.
- The $10 in 2004 only increased the response rate by about four points, which was not statically significant. This is consistent with other studies showing that increases in incentive amounts are not as important as simply paying an incentive and therefore these increases are often not cost effective.
- The prepaid incentives were especially effective in reducing the refusal rate.
- The effect of the incentives was to get more of the same kind of people to respond. In other words, it did not help to offset any demographic bias.
So what's the bottom line. Well, it seems that incentives are still useful as a response rate enhancer, and the longstanding finding that prepaid is more effective than contingent on completion has been reaffirmed. On this study, it may have not create any sort of nonresponse bias but on other studies it might. So one needs to keep an eye on that and if the early data suggest it is a problem then using refusal conversion incentive can be used as an offset by targeting underrepresented groups.