28
Table 5. Experimental Results:
Expectations about electoral loss and risk aversion
R’s status
net difference in
Stimulus
Winners
Losers
effect of framing as risk
PR for Congress (no risk version)
44.9% (245) 51.4% (72)
PR for Congress (risky version)
42.2% (225) 65.1% (86)
-2.7
+13.7**
16.7%*
National Initiative (no risk version) 77.2% (250) 77.6% (85)
National Initiative (risky version)
66.6% (243) 68.9% (77)
-10.6**
-8.7
1.9%
Term Limits on Congress (no risk) 82.5% (256) 79.6% (88)
Term Limits on Congress (risky)
62.4% (232) 68.0% (75)
-20.1**
-11.6**
8.5%
a
Note: Cell entries are percent who agree strongly or very strongly with the proposal for change, with the
total number of subjects per sub-group in parentheses. Losers are identified by responses to the question on
their expectations about future elections.
** = significant differences within group (reading down, p < .05), based on Z score calculating significance
of difference between proportions.
* = significant difference between winners and losers in response to the risk frame ( reading across, p < .05).
a = significant difference between winners and losers in response to risk frame (reading across, p = .15).
Significance of differences between groups is derived from a logistic regression equation, where the
individual’s response to a question about institutional change (1 = strong agree / agree, 0 = disagree / strong
disagree) is predicted as a function of the question version (B, the risky version = 1, version A = 0), being a
loser (1 = expects to lose in future , 0 = other ), and an interaction term (question version B * loser). The
interaction tests if losers’ sensitivity to the risky version of a question about institutional change is
significantly different than winners’.
Convention