18
TABLE 6: Differences-in-proportions
Treatment vs.
Control
National vs.
Control
Local vs.
Control
Local vs.
National
Democrats
6.1%
†
(p=.085)
2.2%
9.4%
*
(p=.017)
7.2%
†
(p=.058)
Decline-to-States
4.1%
5.2%
2.9%
-2.3%
Republicans
1.2%
4.7%
-2.9%
-7.6%
All
4.4%
†
(p=.065)
3.6%
5.2%
†
(p=.053)
2.6%
†
p = 0.10; * p < 0.05 (two-tailed test)
The non-partisan mobilization experiment was more successful. The phone calls
seem to have had the desired effect, as turnout is generally driven in the proper direction.
However, for the most part, this falls short of traditional statistical significance.
The exception here are phone calls made to Democrats by people who identified
themselves as from the Berkeley Youth Vote Coalition, as opposed to the alternate
treatment of the callers identifying themselves as from the National Youth Vote
Coalition. The positive turnout effect here is quite strong and significant, with a
treatment effect of 9.4% overall and something greater among the treated (depending on
how one defines a successful contact; see Tables 8 and 9). Among Democrats, the local
treatment results in turnout that is not only significantly greater than the turnout of the
control group, but also significantly greater than the turnout associated with the national
treatment. The national treatment does do better than the local among non-Democrats,
but falls short of statistical significance.
Other findings of marginal statistical significance (p < 0.10) are all driven by this
phenomenon of new Democratic registrants responding to a non-partisan prompt from
Berkeley Youth Vote (whereas they were unresponsive to the Cal Dems’ overtly partisan
prompt). We would tend to expect precisely the opposite; that an appeal to party
Convention