Appendix to Chapter 3

The results of the survey experiment discussed in this chapter suggest that medical associations have the ability to influence public opinion on health care cost control, but the question remains whether the support or opposition of physician groups would be less (or more) effective when the endorsement or opposition cue explicitly references CER.¹ We carried out a second survey experiment to address this question. Respondents were told,

Some people have suggested that we allow the government and insurance companies to refuse payment for treatments or procedures if their effectiveness has not been demonstrated by rigorous scientific evidence. Suppose you learned that [group cue conditions] and [political cue conditions]. What about you? Would you support this policy?²

The five political cue conditions were almost identical to those used in the experiment reported in the chapter, but with slightly different phrasing that fit the vignette better: (a) “congressional Democrats support this policy but congressional Republicans oppose this policy,” (b) “congressional Republicans support this policy but congressional Democrats oppose this policy,” (c) “both congressional Democrats and Republicans support this policy,” (d) “a bipartisan commission supports this policy,” or (e) no political group cue was given. These five conditions were randomly assigned with equal probability independently of the group cue treatments.

The group cue treatments, however, were different. We randomly assigned respondents to one of four groups—“leading doctors,” “leading patient advocacy groups,” “high-level government administrators,” or “top drug companies”—or to receive no group cue. The support or opposition of the group for those assigned to one of the four groups was also randomly assigned, such that there were nine total group cue conditions.³ As in the first experiment, some respondents were presented with a single cue (e.g., the endorsement of leading doctors) while others were presented with both a political cue and a group cue.⁴

For each of the 45 experimental conditions, table A3.1 reports the average (weighted mean) for the outcome measure. The table also reports the average for each political cue condition, collapsing group cue conditions (in row 10), and the average for each group condition, collapsing political cue conditions (both including the “no political cue” cases [in column F] and not including those cases [in column G]). We focus on three results.

First, the support of doctors increases public support for the CER health care cost-control proposal. Focusing on column G, we find that respondents who received the leading doctors support cue (row 2) had a higher level of support for the proposal (mean=47.8) than respondents who received the leading doctors oppose cue (row 3, mean=42.1). This net difference (of 5.7 units) is statistically significant (p=.09, two-tailed) and is consistent with the results of the first experiment.⁵ Taken together, the results from both experiments provide evidence that public support of a proposal to use CER to help control health care spending is likely to be significantly influenced by the support of physicians.

Second, the influence of doctors is distinctive in that only their support boosts public acceptance of the CER cost-control proposal. The support of other groups either has no effect or else diminishes public support for the proposal (in comparison to opposition from the group). Surprisingly perhaps, the position of patient advocacy groups has no effect on public opinion about the proposal (p=.90 and .80 for the difference between patient advocacy groups oppose [row 4] and patient advocacy groups support [row 5] in columns G and F, respectively). Two groups—top drug companies and high-level government administrators—have so little standing with the public when it comes to CER and cost control that their opposition (not their endorsement) boosts respondents’ support for the proposal. In column G, the -4.7 unit difference between support and opposition of high-level government administrators is statistically significant (p=.09), as is the -5.6 unit difference between the support and opposition of top drug companies (p=.04).⁶ In short, of the (nonpolitical) group cues we tested, only the support of leading doctors increases public support of a CER proposal to help control health care spending. The support of other groups was either inconsequential or counterproductive (compared to the same group’s opposition).

Third, as in the first experiment, we find only small differences across political cue conditions. Collapsing the group cue conditions (row 10 of table A3.1), we find that in the absence of a political cue, average support for the CER proposal is 42.4 (column A). The largest difference from this baseline condition is obtained when Democrats support the proposal but Republicans oppose it (mean=45.6), a statistically significant (p=.09) 3.1 unit difference. When the bipartisan commission cue is given, average support is approximately 44.8 (column C), a 2.4 unit difference from the no political cue condition that is not statistically significant (p=.24).⁷ Thus, we again find that the support of a bipartisan commission does not significantly increase public support for a proposal to help reduce health care spending—in this case, one specifically linked to CER. Moreover, as with the previous experiment, the effect of support from a bipartisan commission does not vary across respondents with differing partisan identities, including those who identify as Independent (p>.10 for all pairwise comparisons). Finally, collapsing across group cue conditions (row 10), there are no statistically significant differences between the four political cue treatment conditions (p>.10 for all six pairwise comparisons).