THREAD. A number of people have asked me for an opinion on what I will call the “Villegas Report” on the election in Bolivia. For now, let’s focus on Sections 16.3 and particularly 16.4 “Expected results based on TREP0 count” (1/7)

Right off the bat, there are obvious problems. Even though we have precinct level data, the Villegas report aggregates into only three very broad groups. While it is true that MAS performed better in rural areas than urban, this simply does not suffice (2/7)

For example, MAS got clobbered in votes coming from Colombia, with a margin of -55%, compared to +33% in votes coming from other countries. But none of these votes from Colombia were included in the early TREP (3/7)

Likewise, if we restrict ourselves to the early TREP, we can see that El Alto (MAS +29) looks nothing like the capital cities (MAS -16) let along Sucre (-32) or Potosí (-40). Don’t we care which city we are talking about? (4/7)

And there is significant variation even within cities. To take an extreme case, in Cochabamba the early MAS margin was -13 compared to +16 late. And yet the precinct-level data is pretty stable (5/7)

The real difference is that MAS areas of Cochabamba were— on balance— counted later. Nearly all the areas favorable to CC were counted early. MAS precincts were more likely to be counted late (6/7)

This lack of geographic specificity adds up. That’s why we projected the uncounted areas based on the smallest geographies for which we had data. This is almost certainly why we find the late vote predictable— we match closely rather than discarding useful data (7/7)