Category Archives: simulation

Three-way referendums: Condorcet vs Instant Runoff

Simulating 3-way choice

I’ve been thinking about a potential second Brexit referendum, where a three-option menu is given:

  • 1: Exit the EU with no deal
  • 2: Accept the negotiated deal
  • 3: Remain in the EU

Three-way choices are alien to British experience of referendums, but don’t really pose great difficulties. First Past The Post (FPTP) is clearly undesirable, but instant runoff (IRV) and Condorcet voting are easy to implement and explain to the electorate. So I built a little simulation to help me understand the problem, particularly to look at the relative performance of IRV and Condorcet.

My conclusions in brief: Condorcet is much more likely to go for the compromise solution (the deal), IRV less so and FPTP the least. Condorcet also seems to correspond well with the average of the underlying utilities. When the population is skewed pro-leave or pro-remain, IRV and Condorcet give more similar results (because it becomes more of a two-horse race). Condorcet cycles are rare. Finally, FPTP is crap.

Continue reading Three-way referendums: Condorcet vs Instant Runoff

Logit, Probit and the LPM

Simulating and modelling binary outcomes

When we have a binary outcome and want to fit a regression model,
fitting a linear regression with the binary outcome (the so called Linear
Probability Model) is deprecated, and logistic and probit regression are
the standard practice.

But how well or poorly does the linear probability model function
relative to logistic or probit regression?

Continue reading Logit, Probit and the LPM