Condorcet Inefficiency and the 2016 Presidential Election

The 2016 election was unusual because it seemed to transcend party lines. It was no longer just Democrat vs Republican, but also Populist vs. Political Insider. Because of this distinction, in many key states (especially MI, WI, and PA), 8%-12% of Sanders’ voters defected to Trump in the general election. The transitive preferences of these individuals are maintained because they valued the populist movement over party lines. This defection may have cost Clinton the election, and more importantly lead to a Condorcet Inefficient outcome.

Take for example Michigan: Trump won by only 13,080 votes (47.6% to 47.3%) and won 16 Electoral College votes. In MI, 8% of Sanders’ voters defected to Trump. Using 2016 State Representative Election Data, approximately 50.2% of MI population voted Democrat in their State Rep elections. I assumed a rationally ignorant voter would vote along party lines for State Rep elections because marginal cost of educating yourself on every issue is greater than the marginal benefit received from the most qualified candidate winning. Therefore I assumed if you voted Democrat for a State Rep, you were affiliated with the Democratic Party. I also assumed the proportion of Democrats who voted for State Reps held for Presidential election voters. Therefore if 50.2% of voters were Democrats, and 48.3% voted for Clinton in the MI Democratic Primary and 49.7% voted for Sanders, preferences for MI voters during the 2016 election would look like the table below. Of course this is a simplified version, assuming only 3 possible candidates and strict preferences for first choice Trump & first choice Clinton voters. If we applied the Borda Count to these preferences: Trump = 188.1 points. Clinton=212.1 points. Sanders 169.2 points. In this case Clinton should have won. However, in the specific case of MI, Sanders seems to have been the Condorcet Efficient candidate, because in pair wise elections he would have beat both Clinton and Trump under these assumptions.

	45.6%	24.3%=50.2%*48.3%	23.0%=(50.2%*49.7%)*92%	2.0%=(50.2%*49.7%)*8%
3	Trump	Clinton	Sanders	Sanders
2	Clinton	Sanders	Clinton	Trump
1	Sanders	Trump	Trump	Clinton