equilibrium w/ positive delegation that dominates majority voting w/o delegation. however, over-delegation is costly; should only delegate when own precision close to random
if $\pi = Pr ( \omega_1 ) = \frac{1}{2}$ for any $F$ and for any odd $N$ and $K$ there exists an equilibrium with delegation that strictly improves over $MV$
incorrectly choosing precision threshold at which to delegate incurs high cost; worse for LD vs MVA
LD proposes improvement over representative democracy under the assumption that not all proposals can tractably be taken to referendum (MV does not stale to large numbers of issues)
the authors demonstrate marginal improvement of LD over MV, then illustrate that humans tend to over-delegate under LD, undermining its theoretical equilibrium advantage
the authors conclude that a simpler solution (given common interests) is majority voting with abstention
the paper does not consider:
representational democracy (the failures of which LD is attempting to solve)
mis-aligned values / plurality of interests
common interests is generally a bad model for the types of problems democracies usefully solve: referendum usually involves some contentious issue w/ (real or perceived) potential harm to some subset of the electorate
dynamics observed in common interest studies are unlikely to generalize
assumes one-dimensional assessment of representatives (precision)
improvement: vary degree of contention (mixed objectives by ratio)
reps should to be assessed on 2 axes: expertise and alignment (thoughts on addressing this in bellman delegation)
