Bogotá/New York/Brussels, 24 October 2018
The quantitative analysis carried out in this briefing uses open-source data collected from the World Bank’s World Development Indicators archive, available at: http://
databank.worldbank.org/data/reports.aspx?source=world-development-indicators.
In order to carry out a difference-in-differences approach, a few assumptions need to be validated. The first step is to construct a synthetic control case for Guatemala using the country’s neighbours during the pre-2007 period. The pre-CICIG weights are produced with Jens Hainmueller’s entropy balancing algorithm, implemented in Stata using the “ebalance” package. The goal is a hypothetical control case where the pre-treatment trends are in parallel – a critical assumption of the difference-in-differences strategy. Secondly, there is the exchangeability assumption of the difference-in-differences approach. The argument is that an institution like the CICIG could have been established in any number of Latin American countries, eg, El Salvador, Colombia, Nicaragua or Mexico, that have experienced intra-state conflict, paramilitary activity, high homicide rates and poor prosecutorial performance. It was largely chance that a body like the CICIG was formed in Guatemala rather than elsewhere.
The table below shows the countries and their relative weight in the synthetic control. Based on the data, pre-2007 Guatemala is approximately equivalent (in terms of trends in homicide, economic well-being and infant health) to a hypothetical country that is composed of 38 per cent El Salvador, 18 per cent Dominican Republic, 13 per cent Nicaragua, 8 per cent Costa Rica, 8 per cent Honduras, 5 per cent Panama, 4 per cent Venezuela, 3 per cent Colombia and 2 per cent Mexico (the percentages do not add up to 100 due to rounding).
The following five charts illustrate the pre-2007 trends in homicides, household consumption, GDP per capita and under-five mortality rates in detail, as well as homicide rates broken down by country.
In order to compute the net effects of the CICIG on homicide, the standard difference-in-differences approach is used, operationalised in Stata using a two-way fixed-effects estimator on the weighted data. This process yields an average treatment effect of -2.77 homicides per 100,000 per year. That means that during the post-CICIG period, Guatemala had, on average, 2.77 fewer homicides per 100,000 per year than it would have been expected to.
Multiplying this average effect by the population of Guatemala in each year that the CICIG has been operational can approximate the number of avoided homicides. During the period 2007-2014, Guatemala’s growing population averaged 14.8 million people. The multiplication yields an estimate of 3,279 avoided homicides over eight years. Extrapolating these effects through 2017, the approximate number of avoided homicides is 4,658 (during the period 2007-2017, Guatemala’s population averaged 15.3 million).
The replication files for this analysis are hosted by Princeton’s Empirical Studies of Conflict program, a co-sponsor of the Economics of Conflict program. Replication files are available at: https://esoc.princeton.edu/subfiles/replication-material-guatemala-briefing
Source link : https://www.crisisgroup.org/latin-america-caribbean/central-america/guatemala/70-saving-guatemalas-fight-against-crime-and-impunity
Author :
Publish date : 2018-10-24 03:00:00
Copyright for syndicated content belongs to the linked Source.
