We propose adding two questions: one about expected share of voters for different candidates in one’s social circle – family, friends, and others one is in regular contact with; and the other about expected vote shares in the general population. Vote expectation questions currently included in ANES have been shown to add predictive value beyond personal vote intentions. However, they ask only about which candidate one expects to be elected President and not about expected vote shares, so it is not straightforward to use them for share predictions. Furthermore, there is empirical evidence that people make quite accurate judgments about frequency of different beliefs and behaviors in their social circles, but that their judgments about the general population show some systematic biases. Adding the questions about expected vote shares in one’s social circle and in the general population may improve predictive accuracy of polls and shed light on how social environments influence individual voting behavior.