Few would doubt that deliberation, defined roughly as serious, informative, civil discussion changes some individual-level opinions and vote intentions. Gross change—by individuals, in one direction or the other—figures to be appreciable. But if about as many people change about as much in each direction, the net change may still be nil. And from the standpoint of democracy it is the net change that is the more important. No doubt it has some normative value for everyone to reach his or her own considered or authentic preferences. But if the polls and election results still read exactly the same, it makes no difference policy or election outcomes, nor to democracy as a decision-making process. This paper reviews the net change in policy attitudes and electoral preferences across a sizable number of Deliberative Polls in a variety of settings. We first show that statistically significant net change is far more the rule than the exception but that it varies across issues and settings. We then estimate a model aimed at explaining the magnitude of net change, in terms of the location (country), salience of the issue, mode (face-to-face versus online), heterogeneity of the sample, and other factors.