Lessons about Voting Systems from Afghanistan 2009

The Afghanistan 2009 Presidential election recently attracted a lot of media attention due to its blatant fraudulence. It was supposed to be conducted using plurality with additional top-two-runoff round if necessary. In fact, there was massive, blatant, obvious fraud evident to all external observers and widely reported in the press both before, during, and after the election. [We shall not give details since they have by now been widely reported elsewhere.]

H.Karzai declared himself victorious with over 50% "hence" with no need for a runoff. However, under pressure from international observer groups and consequently the USA, he later reconsidered and agreed to allow a runoff, i.e. essentially, admitting his fraud. It seemed likely Karzai would have won even with honest voting, and it also seemed likely that in the runoff he would just re-fraud the election anyhow (this time more subtly). His rival A.Abdullah therefore dropped out of the runoff, claiming it was pointless for him to participate since he saw no way to prevent such fraud.

Some voting systems are more susceptible to fraud (and when that fraud occurs anyway, it is more obvious and easier to detect); some are less.

• Instant Runoff Voting (IRV) would have been a big mistake for Afghanistan which would have made the situation a lot worse. The problem is IRV cannot be counted in precincts; there is no such thing as a "precinct subtotal." It has to be counted centrally. And in Afghanistan with no widespread internet etc, it would literally have to be counted centrally, by Karzai's closest pals, in Kabul. The advantage of precinct-based counting (and the fact that precinct-subtotals exist) from the standpoint of fraud, is that scrutineers can compare the centrally-claimed versus precinct-claimed subtotals. The only ways to fraud the election are then
  1. Fake a lot of precinct subtotals – but this requires a large number of widely-geographically-separated conspirators, which makes the conspiracy less likely to succeed undetected.
  2. Make the precinct and central counts disagree; then we instantly know somebody lied.
• The 2-round (non-instant) runoff Afghanistan actually had, or single-round "approval voting" (where each voter "approves" or "disapproves" each candidate, and the most-approved one wins) both permit precinct countability. Hence they both are preferable to IRV from the standpoint of fraud. They also both are simpler than IRV – key in a country like Afghanistan with tremendous illiteracy.

  Before the war the literacy rate was estimated at 11.4 percent (18.7% male + 2.8% female). In urban settings 25.9% (35.5% male + 14.8% female) of the population six years old and over were literate, but in rural areas only 8.8% (15.7% male + 0.6% female, in some provinces 0.1%).

  Unfortunately the 2-round system is bad because it is two rounds (a severe imposition in a poor country like Afghanistan) and also, that clearly failed in the present case since Karzai initally refused, and then Abdullah refused, to permit the second round. Just using the first round alone (i.e. plain-plurality voting) would also have been unnacceptable since it delivers results which, when there are 41 candidates, are almost random and hugely undemocratic.
• For multiwinner elections where both proportionality of representation and simplicity of voting are paramount, we recommend F.Simmons' simplified form of asset voting. Again this is precinct-countable and very simple hence would have been excellent for Afghanistan. (The system Afghanistan actually used in 2005 to elect parliament was "single non-transferable vote." In settings like Afghanistan, where e.g, they had a 46-candidate election, it leads to near-random results of doubful legitimacy and quality as well as failing to satisfy a "proportionality theorem." We see no advantages for that versus asset voting.)

Both asset, and plurality voting (as in the top-2-runoff system) and especially approval voting permit enhancement via Rivest-Smith antifraud technology, which although it would probably be too complicated for contemporary Afghanistan, could be a useful idea to hold in reserve for the future.

Bottom line

Afghanistan has a lot of problems which may prevent it from becoming a successful democracy anytime soon. But bad choices of voting system would severely-hurt that goal.

• Instant Runoff = big mistake for Afghanistan: too complex, enables fraud.
• Approval voting (one round) and top-2 (non-instant two-round) systems: either are good ideas.
• For multiwinner elections (like the Afghan parliament) asset voting seems the clear choice.

Return to main page