When generating insights from human groups, conversational deliberation is a key method for exploring issues, surfacing ideas, debating options, and converging on solutions. Unfortunately, real-time conversations are difficult to scale, losing effectiveness in groups above 4 to 7 members. Conversational Swarm Intelligence (CSI) is a new technology that enables large human groups to hold real-time conversations using techniques modeled on the dynamics of biological swarms. Through a novel use of Large Language Models (LLMs), CSI enables real-time dialog among small groups while simultaneously fostering content propagation across a much larger group. This combines the benefits of small-scale deliberative reasoning and large-scale groupwise intelligence. In this study, we engage a group of 81 American voters from one political party in real-time deliberation using a CSI platform called Thinkscape. We then task the group with (a) forecasting which candidate from a set of options will achieve the most national support, and (b) indicating the specific reasons for this result. After only six minutes of deliberation, the group of 81 individuals converged on a selected candidate and surfaced over 400 reasons justifying various candidates, including 206 justifications that supported the selected candidate. We find that the selected candidate was significantly more supported by group members than the other options (p<0.001) and that this effect held even after six minutes of deliberation, demonstrating that CSI provides both the qualitative benefits of conversational focus groups and the quantitative benefits of largescale polling.