Centerfield

Electing the Doge of Venice: analysis of a 13th Century protocol

Miranda Mowbray and Dieter Gollmann, of HP Labs Bristol, have done some fascinating work analyzing the election protocols used to elect the Doges of Venice (hat tip to Bruce Schneier

Venice had a republican election system to select its Doge (Duke), from sometime before 750 AD to 1797 (ending with its conquest by Napoleon), over 1000 years, an incredibly long period of stability, especially for a city-state, as it was between 803 and 1797.

Venice grew very wealthy, and powerful enough to sack Constantinople in 1204 in the Fourth Crusade. It was a triumph of pragmatic rule rather than popular, though. Doges were elected for life, their electorate was limited to the nobility, and the protocol featured plenty of distrust of said electorate.

The protocol analyzed as used between 1268 and 1797, and it worked well enough that Venice's hardest problems were those of success. It involved ten (!) rounds, nine selecting the next electoral college for the next round, the final the doge. Selection methods alternated between election and random lot.

Good properties included a high incidence of reasonable minority-party candidates reaching dogeship, a high chance that negotiation would be needed (meaning more involvement and protection of non-Doge-winning parties than we see in our winner-take-all system). Like our system, it's delivered only two sons of Doges as Doge themselves, in 75 Doge elections vs our 54 Presidential elections. It had all sorts of other good stuff.

It's occurred to me that this protocol would've come in handy a decade ago for solving Internet Domain Name Service top-level governance questions. There was a need for somebody trusted and given broad power, as Jon Postel used to be, but after him, nobody trusted anybody, because of the $millions involved. This kind of protocol could've solved that, I think.