Even though I am co-editing a volume on Mexican elections and lawmaking (which I had once dreamed of having completed before the 2006 election), I have not addressed this Sunday’s election here in some time.

Courtesy of boz’s presentation of the data, I took a look at a series of late polls. Averaging of the nine polls of the week of 23 June gives us the following:

López Obrador (PRD): 33.6%
Calderón (PAN/Fox’s party): 32.7
Madrazo (PRI-asaurus): 24.2

I am not sure of the margin of error on each of these polls, but I know that for at least some of them, it is around +/- 3 percentage points. Applying that margin to each candidate, we get average ranges of:

López Obrador: 31-37
Calderón: 30-36
Madrazo: 21-27

The polls thus pretty clearly agree that the race is too close to call between Calderón and López Obrador (”AMLO”), and that Madrazo is in third place. However, at least two of the polls that boz reports (Mitofsky and Zogby) do not clearly separate second and third place. In fact, Zogby does not clearly separate first and third place. Zogby has the race as Calderón 30, AMLO 27, Madrazo 24. Assuming +/- 3 MOE, that is no different from a three-way dead heat of three candidates at 27%. Zogby also has the largest undecided or “other” (there are two very minor candidates running) of any of these polls (19%), which presumably means they are more conservative in assigning “leaners” to one candidate or the other.

Other polls have a very large gap between first and third, even if none can separate AMLO and Calderón with certainty. For instance Universal has it AMLO 36, Calderón 34, Madrazo 26. Milenio and Marketing Político have Madrazo on 22% and GEA-ISA has him on 20%, but all of these put only two points between the leaders.

While it is hard to concoct a believable scenario in which Madrazo wins, the fact that he has gained a bit in many polls late in the campaign and that his third place is not decisive in several of them is bad news in at least two respects. First, it reduces the pressure on weaker supporters of the PRI to defect, because they may not be convinced that their candidate has no chance. Second, it increases the chance that the candidate who does win might have lost had more PRI voters defected to whichever of the other two is their second choice. My hunch is that Madrazo being in a strong third place (and perhaps second) helps AMLO, on the grounds that AMLO’s image as the rabble-rouser makes him less likely to pick up late PRI defectors than the “conservative” (in almost any sense of that word) Calderón. More PRI voters sticking with Madrazo means fewer votes shifting late to Calderón.

It is possible that the PRI vote is under-sampled, given its strength in rural areas, and out of this possibility come the scenarios (unlikely though they may be) in which Madrazo ekes out a surprising win. But I do think this is indeed unlikely, because I still expect few undecideds to go for Madrazo in the end and many soft supporters of the PRI ultimately to vote tactically, and thus to cancel out any under-sampled PRI vote. The fact that a federation of formerly PRI-affiliated unions endorsed AMLO recently throws further cold water on the chances of a late recovery by the PRI.

I would not bet on this election, as it really could go either way between Calderón and AMLO.

The presidential election is decided by plurality, so there will be no second round even if, as seems likely, the winner has well under 40% of the vote and a very small margin.

The other thing to watch is the congressional race. In 2000, Mexican voters showed some, albeit limited, willingness to split tickets. Thus the PRI in congress could slightly outpoll its presidential result. There is no direct translation of national vote shares to seats in the 500-seat Chamber of Deputies; contrary to some otherwise reputable sources, Mexico’s mixed-member electoral system is not MMP, but parallel.* It is also a rare mixed-member system with a single vote: The voter votes for a candidate in his or her single-seat district, where the plurality prevails, and this vote is also pooled to the separate contest for list seats (closed regional lists).

Because of the single vote for both candidate and list, every party has a strong incentive to run a candidate in every district, even if it clearly can’t win a given district. If it does not have a “face” in a district, it has no way to obtain list votes in that district. And, of course, because of the single vote, there is no incentive for you if you are a voter to defect from your first choice party’s candidate, even if he or she can’t win your district. Thus many districts will be won by narrow margins and much less than 50%, and this will benefit whichever party has the greatest regional spread and harm any party that is regionally concentrated. This aspect of the system is much more likely to give a (small) seat bonus to the PRI than to the PAN, and to harm the PRD.

The system also puts a potential premium on parties’ having attractive candidates, given that the candidates are what the voter must vote for. However, there is not much evidence that voters care much about candidates (there is some, however). The development of a “personal vote” is hampered by the ban on consecutive terms in all elections in Mexico. No incumbents are running for reelection to their seats.

It is very possible that the result could be López Obrador as president with not even one third of congressional seats–the share needed to sustain a veto. If Calderón were to win, he, too, could be short of one third.

In the 128-seat Senate, all of which is also elected Sunday–there is no chance that any party will have a majority.**

Coalition-building will be even more important to Sunday’s winner than it was to Fox–and likely more difficult.

Of course, La Profesora Abstraída has been covering the campaign all along.

* The only senses in which the electoral system is not parallel (MMM) are in two ways that will not be triggered in this election: No party may have an overrepresentation of more than eight percentage points above its nationwide Chamber vote share, nor more than 300 seats. Otherwise, it is a pure parallel system.