Important New Violent Death Estimates for the War in Peru with Implications Beyond just Peru: Part 7

Today I return to Peru.  If you’re not up to speed on this discussion please go back to post 1 of the present series.

I could include this post in either this series, this series or this series.  But I think it works best within my first Peru series because the present post covers an attempt by Daniel Manrique-Vallier and Patrick Ball (MVB) to use a (relatively) new dataset to discredit the Shining Path (SP) estimates of Silvio Rendon.  These estimates were the focus of my first series.

MVB really offer just one argument against Rendon’s SP estimates (and no arguments against the rest of his estimates).  First, they stitch the SP-attributed deaths in a new dataset together with the SP-attributed deaths in the old lists that were used by both Peru’s Truth and Reconciliation Commission (TRC) and by Rendon.  MVB then claim that death counts in their new stitched-together list exceed Rendon’s estimates.

There are two main reasons why MVB’s argument is false.  Reason 1 is that, even if we accept MVB’s stitched-together numbers, their argument still fails.  Reason 2 is that the stitched-together numbers simply aren’t credible.  Let’s consider the two reasons in turn.

Table 3 in the appendix to Rendon’s second paper is the best resource for assessing Reason 1.  It provides information on the 9 geographical strata that allow head-to-head comparisons between Rendon’s estimates and MVB’s stitched-together numbers.  Rendon’s central estimates exceed MVB’s new numbers in 5 of these 9 strata.  Moreover, the upper limits of Rendon’s uncertainty intervals exceed the MVB numbers in 8 of the 9 strata.  The totals for the 9 strata combined are 7,974 for Rendon’s central estimates and 7,311 for MVB’s numbers.  Thus, MVB’s argument doesn’t work even if we take their stitched-together numbers at face value.

We now move on to Reason 2; the new data that MVB try to import into the discussion of war deaths in Peru collapse immediately when subject to even a moderate degree of scrutiny.

MVB have so far avoided the topic of how their new data were gathered.  This indifference toward data collection methodologies lands them well within the path established by the statistical work of Peru’s TRC which they participated in.  Yet the burden falls to MVB to make a positive case for their data.  Since they have abdicated this responsibility I attempt here to fill the void they leave behind.

The new data come from the “Census for Peace” (CP) conducted in 5 waves by Peru’s Ministry for Women and Social Development (MIMDES).  The CP was meant to support war reparations policy by identifying rural communities that were victimized by the war.  Several important points flow from this mandate.

  1. “Terrorists,” i.e., members of the SP and other guerrillas groups, are not eligible for reparations. So the CP’s reparations focus excludes many of the people killed by the State.  Indeed,  there are even incentives for CP respondents to misreport people killed by the State as having been killed by guerrillas.  This misdirection can short circuit suspicions that these victims were “terrorists”.
  2. CP data collection was exclusively rural although the war did have an urban component.  This restriction may bias the data collection further against finding deaths caused by the State.
  3. The CP did collect data on individual victims but it focused its efforts on finding affected  communities.  So the quality of victim data is likely to be relatively low.  Data quality is particularly important when stitching together data across multiple sources because data imperfections translate into failures to identify duplicate deaths across sources.  For example, M Szpagat from Oak Brook Illinois may appear, spuriously, to be a different person from Mike Spagat of Oak Park, Illinois.

Peru’s Reparations Council ultimately absorbed the CP data into Peru’s Unique Registry of Victims, the official list of people and places eligible for reparations.  Importantly, the Reparations Council did not classify the CP’s list of victims as a “pre-existing source“, a designation that would have led them to treat the CP list as provisionally valid albeit subject to some checking.  Jairo Rivas, the former Technical Secretary of the Reparations Council told me in an email that the CP “was not rigorous in the individual identification of victims.”  Therefore, the Reparations Council used the CP victims list only for guidance when the Council prepared its own list.  In fact, an official human rights report revealed that in many areas the CP collected information only from local authorities without even interacting with the populations these authorities represented.  Worse, the CP did not even attempt to verify the information it received. 

la información recogida por el Censo por la Paz no ofrece todas las garantías necesarias para que se dé correctamente unproceso de esta envergadura, ya que en muchos casos se tomaba como fuente de información principal los datos dados por las autoridades locales –sin participación de la población– y no necesariamente era contrastada con otras fuentes de información. Hasta el momento, no se ha tenido tomado conocimiento de que se haya empezado con la implementación del Censo ya que se cuentan con los recursos necesarios para dicho fin.  (Source: Informe Anual 2007: La HORA de la JUSTICIA, page 109.  This document doesn’t seem to be freely available online but I can share the pdf with anyone who might be interested.)

Jairo Rivas added that the Reparations Council took the CP list of affected communities more seriously than it took the victims list.  Nevertheless, it seems that the CP’s identification of affected communities was stunningly sloppy.  959 out of 2,539 communities registered as war affected by the CP in its last wave were rejected as such by the Reparations Council.  Reasons for these corrections range from the communities not conforming to pre-agreed (with the CP) inclusion criteria to CP errors in recording community locations.  Indeed, the Reparations Council made an embarrassing climb down in the face of these CP cock ups.  After putting out an April 2009 press release announcing 2,539 new affected communities, based on naive acceptance of CP work, the Reparations Council followed up already in May 2009 with a further release lowering this number to 1,576.

Remember that the identification of affected communities is considered to be the good part of the CP’s work.  The identification of individual victims is the bad part.  And MVB rely exclusively on the identification of individual victims.

When MVB find a name on the CP victims list that does not seem to match with a name on the old lists used by the TRC what are the possible scenarios?

  1. This person could be a true victim who was missed by the earlier sources.
  2. This person could be a true victim who did appear in the earlier sources but who is impossible to match against these sources because his/her rendering on the CP list is so garbled that he/she looks like a different person.  Remember that the CP recorded incorrect locations for many of the whole communities that it put forward to receive reparations.  Moreover, recall that much of the CP’s victim information entered its list only indirectly through accounts of local leaders who could easily get key details wrong.
  3. This person was not a true victim.  The CP’s work was all about reparations.  More victims reported from a community can mean more reparations.  And the CP did not check any of the submitted names.   This is a recipe for fabrication and, of course, fabricated names will not match against real ones.

Speaking of fabrication, anyone inventing victims should probably attribute these deaths to the Shining Path rather than to the State.  This attribution makes the fake victims look, most plausibly, like victims of terrorism rather than like possible terrorists themselves.

In short, the victims list of the Census for Peace is garbage and was recognized as such by Peru’s Reparations Council and other analysts with knowledge of the work.   Stitching crap together, even with the finest thread, produces….well …. something that we shouldn’t integrate into our analysis.  The Census for Peace does not belong in a serious discussion about  Peru’s war losses.




Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s