Truly efficient markets can compensate for bias in participating populations

For example betting markets may be better than polling for predicting elections even though betting markets probably have a more biased population participating.

References

Conversation with Jim Mallernee