Statistical Analysis and Visualization of the Drug War in Mexico

On December 11, 2006 Felipe Calderon, as the first significant act of his presidency, sent the army to his home state of Michoacan. He claimed that it was to regain control of territories lost to the drug cartels, and indeed, a new cartel had started operating in Michocan. But the fact that he won the election by the slim margin of 0.6% and his main rival declared himself the legitimate president of Mexico surely influenced him.

In the beginning, the war proved a success by all objective measures: in 2007 the homicide rate decreased to its lowest level in recorded history and murders in Michoacan went down by more than 40%. Not that it mattered much, all the while the government was losing the psychological war—the use of torture and beheadings became common in executions carried out by drug cartels as they sought to protect their turfs and intimidate the population.

And then 2008 rolled around and the Sinaloa Cartel decided to take advantage of the weakening of the other cartels and the corruption that is endemic in Mexico to gain control of the drug trade…

Sources: Homicide data from INEGI, population data from CONAPO. 2009 estimate based on execution rates

In Mexico about 85% percent of crimes go unreported, so it’s difficult to obtain exact estimates other than general trends. However, murder is generally considered to be reliably reported, and thus it will be the main data source I’ll use to measure the ebbs and flows of the drug war.

There are two ways to measure the number of homicides:

1. Police records. This data is collected by law enforcement agencies. It is available from two sources:
   
   • The National System of Public Security (“Sistema Nacional de Seguridad Pública”) which collects the data from something called CIEISP forms. Each state law enforcement agency has to fill them out each month, and the SNSP tallies them. You can download the SNSP data from the ICESI, which is a civic institution not affiliated with the government.
   • The Statistical Yearbooks the statistical agency of the Mexican Government publishes each year on its website. The source for the data are the state law enforcement agencies (“Procuraduría General de Justicia del Estado. Dirección General de Información, Estadística e Identificación Criminal”)
   Even tough they are supposedly collected from the same source, there are important difference between them.
2. Vital Statistics from the INEGI. Produced from death certificates, they include data at the municipality level and the month of death. In addition they include sex, age, marital status, occupation, education, etc.

There are also data from the UN Crime Survey and the Panamerican Health Organization, but these are only available at the national level.

Sources

As you can see from the chart, the trend from all sources is roughly the same, even thought the homicide rate according to the SNSP is always higher than that from the INEGI—except for 2008. It turns out there is a simple reason for that.

Since the data is available at the state level for both police records and the vital statistics we can compare them:

Sources

Ciudad Juarez come forth!

The police records from the SNSP are missing 1,153 murders in Chihuahua! I am not able to rightly apprehend the kind of confusion that would provoke such a discrepancy. Just to give you some perspective, in Ciudad Juarez, Chihuahua’s biggest city, there were 1,653 homicides, more than the 1,414 reported in the whole state according to police records.

How can it be that there was such a big difference?

Well, first of all, the police records and the vital statistics records are collected using different methodologies: vital statistics from the INEGI are collected from death certificates and the police records from the SNSP are the number of police reports (“averiguaciones previas”) for the crime of murder—not the number of victims. For example, if there happened to occur a particular heinous crime in which 15 teens were massacred, but only one police report were filed, all the murders would be recorded in the database as one. But even taking this into account, the difference is too high.

You could also argue that the data are provisional—at least for 2008—but missing over a thousand murders in Chihuahua makes the data useless at the state level. I could understand it if it was an undercount by 10%–15%, or if they had added a disclaimer saying the data for Chihuahua was from July, but none of that happened and it just looks like a clumsy way to lie. It’s a pity several media outlets and the UN homicide statistics used this data to report the homicide rate in Mexico is lower than it really is.

For most states the differences in homicides rates seem stable

Police Records from the Statistical Yearbooks

The other source of homicides based on police records are the Statistical Yearbooks. They contain what is claimed to be the final number with at least 90% of all homicide reports entered into the database. Chihuahua is now “only” missing 466 murders compared to the vital statistics.

Sources

You might think that the Statistical Yearbooks are better, but in some of the years I looked up, the statistical agency of the Mexican government made the all too common mistake of confusing manslaugthers with homicides. You see, in Mexico the technical term for homicide is “homicidio doloso” and the term for manslaugther is “homicidio culposo” they look alike and there is no obvious way to tell which one is which (I had to look them up in a dictionary). I had to correct the 2008 data for Querétaro since the values were obviously transposed in the Statistical Yearbook.

The Statistical Yearbooks also report the number of victims instead of police reports, but only combined with the number of manslaughters, so it is impossible to know the true number of homicide victims from police records.

Questions remain for all the discrepancies. Perhaps the differences are due to data entry errors, improper use of preliminary data or plain lack of technical knowledge, and the local or federal governments are innocent of any wrongdoing. But it does seem awfully convinient to try and hide the fact that Chihuahua has turned into a war zone.
The INEGI, the statistical agency of the Mexican government, is widely regarded as more trustworthy that the police. Though that’s a bit like saying Episode III is widely regarded as better than The Phantom Menace. So we shouldn’t automatically trust the statistics just because the incredible rise in homicides in Chihuahua shows up in the data. To test its authenticity we can use Benford’s Law.

Benford’s law has got to be the coolest statistical “trick” after bootstraping. The basic idea is that the first digit of certain lists of numbers—one of the examples in Benford’s orginal paper was the death rate—are distributed in a non-uniform way tilting towards the lower digits, i.e, there are more numbers starting with 1s than with 9s.

This counter intuitive phenomenon makes it easy to detect if numbers have been altered since you would expect normal humans beings making up numbers whole cloth to produce the same amount of numbers starting with 1 than with 9.

Failure to follow Benford’s law is often interpreted as possible evidence of fraud and a prescription to investigate further.

Sources

The INEGI homicide data fails the chi-squared goodness-of-fit test and does not follow Benford’s law. This means we’ll have to investigate further to see if any fraud was commited.

Since the data are only available monthly, I decided to check if the big massacres that have occurred since 1994 were recorded in the database. For example, if Ensenada on average each month has 5 homicides, but suddenly during September there are 20, we can be pretty sure that the massacre that occurred was recorded. This is what I found:

• Aguas Blancas Massacre: (Warning: Graphic Video) 17 dead, June 28, 1995. In the INEGI homicide database
• Acteal Massacre: 45 dead, December 22, 1997. Not in the INEGI homicide database
• Ensenada: September 19, 1998, 18 dead. In the INEGI homidice database
• Decapitated Bodies in Yucatán: August 28, 2008, 12 dead. In the INEGI homicide database
• Tijuana prison riots: September 2008, 25 dead. In the INEGI homicide database
• 24 dead outside Mexico City: September 13, 2008, 24 dead. In the INEGI homicide database
• Reynosa prison riot: October 2008, 21 dead. In the INEGI homicide database

The Acteal massacre committed by paramilitary units with government backing against 45 Tzotzil Indians is missing from the vital statistics database. According to the INEGI there were only 2 deaths during December 1997 in the municipality of Chenalho, where the massacre occurred. What a silly way to avoid recording homicides! Now it is just a question of which data is less corrupt.

Since the recent vital statistics from the INEGI don’t look crazy wrong like the SNSP data, they are the ones I will use. Oh, I don’t doubt that here and there some homicides were not reported or even deleted on purpose from the database, but the INEGI homicide data for Chihuahua looks ok and at least includes all the recent massacres. For our purposes that will be enough, as long as we keep in mind its shortcomings.

Sources: INEGI and CIEISP forms from the SNSP (2005, 2006, 2007)

Overview of the Cartels
The Sinaloa Cartel
This is the most powerful cartel operating in Mexico. The leaders like to cultivate an image of being simple farmers that only grow pot to escape poverty. They claim to be adamantly against kidnappings, which makes them somewhat more popular than the other cartels. They are led by Joaquin “El Chapo” Guzman, who according to Forbes has a net worth of 1 billion dollars.

La Familia
The new kids on the block that managed to wrest control of the port of Lazaro Cardenas from the Zetas. They’re into Focus on the Family type Muscular Christianity and chopping heads off with machetes. As their official introduction they dropped 5 severed heads inside a nightclub in Michoacan. They are currently allied with the Sinaloa cartel.

The Juarez Cartel
Also known as the Vicente Carrillo Organization, they used to be allied with the Sinaloa Cartel till “El Chapo” ordered the killing of Vicente Carrillo’s brother in 2004. They control Ciudad Juarez, though, after more than 5,000 deaths, the Sinaloa cartel is pretty well entrenched there and it looks like it will be a fight to the death.

The Tijuana Cartel
This cartel has seen better days as all of its major leaders have been either arrested or killed. Their rivalry with the Sinaloa Cartel dates back to when they tried to kill “El Chapo” Guzman at the Guadalajara airport in 1993. That shootout ended with six people dead, among them a Mexican cardinal.

The Gulf Cartel
They used to be the most powerful cartel in Mexico till their leader got arrested in 2003, visibly weakened, “El Chapo” Guzman saw an opportunity and put the Beltran Leyva brothers in charge of taking over their central stronghold of Nuevo Laredo. They weren’t successful, but the Beltran Leyvas rose in the ranks of the Sinaloa Cartel, along with their chief gunman “La Barbie.”

Beltran Leyva Cartel
Formerly allied with the Sinaloa cartel, they broke off after one of the Beltran Leyva brothers was captured by the government and the rest of the clan blamed the leader of the Sinaloa cartel of tipping off the government. As revenge they killed “El Chapo” Guzman’s son. As of 2010 only one of the brothers remains free. He is currently battling his former ally “La Barbie” for control of the organization. They are currently allied with the Zetas

Zetas
They started as a group of army deserters who offered their services to the Gulf cartel. After the capture of Osiel Cárdenas in 2003 and succesfully defending Nuevo Laredo from La Barbie they started operating independently of the Gulf cartel, though they remained allies until 2010. They like to collect “taxes” and kidnap people in addition to trafficking drugs.



To produce the plots I discarded all data before 1994 since there were a lot homicides registered in the INEGI database with no year of occurrence recorded. I also ordered the data by year of occurrence instead of year of registration as is done by default by the INEGI.

Sources

If Chihuahua were a country it would have had the highest homicide rate in the world. The data for 2009 won’t be available until November, but from press reports things got much, much worse there.

Sources

As you can see the big increases in violence have been concentrated in a few states. It’s interesting that Tamaulipas, home to the Gulf Cartel and the Zetas, experienced a decrease in homicides from 2006 to 2008; however, things have changed since then. As of 2010 the states which have experienced big increases in homicides would also include Morelos, Guerrero, Coahuila, Nayarit, Nuevo León and Tamaulipas.

*I got the idea for this chart from the Learning R blog

Homicide Rates by State*

Sources

I used k-means to classify the different states into clusters, within each cluster the states are ordered by their homicide rate in 2008. I liked how they were classified, though I would switch Nayarit with Sonora.

• The top row consists of states home to drug cartels or in dispute. Notice the big increases at the end of the series
• Violence has decreased a lot since 1994 in Oaxaca, Morelos (this will change in 2010), State of Mexico and Michocán. 
• There was a big increase in the homicide rate in Baja California in 1998 due to the power vacuum left after the death of Amado Carrillo Fuentes.
• Notice how the homicide rate was higher in Michoacan in 1994 than in 2006 when the President of Mexico sent in the army.
• Notice also the rise and fall of homicide rates in Tamaulipas when the Sinaloa Cartel tried to take over Nuevo Laredo around 2005-2006.
• Chiapas wasn’t all that violent back when the Zapatista were in vogue—at least according to the Mexican government

*I got the idea for this chart from the Junkcharts blog


To display the murder rate at the municipality level I’ve switched to choropleths. Keep in mind that the big municipalities tend to draw more attention because of their size, but they also tend to have low population densities. For example, Tijuana, which is an important drug corridor and home to the Tijuana Cartel is barely visible.

1995	2000
2005	2006
2007	2008

It’s interesting that the most violent municipality in all of Mexico during the 1990-2008 period was probably Badiraguato, it’s served as cradle to some of Mexico’s most important cartel leaders: Caro Quintero, Fonseca Carrillo, “El Chapo” Guzmán, the Beltran Leyva brothers, Félix Gallardo, and lots of others.


Since the data for 2009 isn’t out yet, I can’t do any charts :(. But here is what happened:

• The war the Sinaloa cartel started to defeat the Juarez cartel kept going in 2009, leaving a total of 2,700 dead in Ciudad Juarez, compared to 1,600 in 2008. In 2010 the number of homicides will probably be similar to 2009 (perhaps a little lower).
• At the end of 2009 one of the Beltrán Leyva brothers was killed at a luxury apartment in Cuernava, capital of the state of Morelos. A week later another of the Beltran Leyvas was captured in Sinaloa. This sparked another inter-cartel war as the only remaining Beltran Leyva fought for the leadership of the organization with “La Barbie”
• At the beginning of 2010 a hit man allied with the Sinaloa Cartel, “El Teo,” was captured. He was sent to defeat the Tijuana cartel, and is believed to be responsible for over 300 murders. Interestingly, there were no massacres or increases in violence after his capture. This probably means the Sinaloa Cartel overextended itself and is waiting to defeat the Zetas, the Juarez Cartel and the Beltran Leyva organization before resuming its quest for Tijuana. There won’t be a quick ending to the cartel wars.
• At least at the beginning of 2010 the murder rates in Mazatlan and Nogales grew to be similar to those of Ciudad Juarez in 2008. The state of Sinaloa in 2010 will probably be about as violent as Chihuahua in 2008
• Like Finland in 1944, the Gulf cartel in 2010 saw the writing on the wall and decided to switch sides and ally itself with its former mortal enemy, the Sinaloa Cartel, the guys they stopped in Nuevo Laredo. To prove their worthiness they are now tasked with destroying their former paramilitary unit, the Zetas, with the help of La Familia.

One cartel to rule them all?

I think I went a little too far with the WWII analogies, Mexico is a pretty violent place, but it’s not going to turn into Iraq or Afghanistan. It is also not going to turn into a failed state like Somalia. Even with its high homicide rate Brazil’s economy has had higher growth rates than Mexico and is a pretty swell place to visit.

As recently as 1994 the man destined to win the presidential election was murdered in Mexico, stuff like what’s happening now isn’t exactly new, even if it is worse. Plus, the army is firmly under civilian control, there are free elections, there’s an independent judiciary, and important business groups. Even right now in 2010 I very much doubt the murder rate in Mexico has reached the peak murder years in Brazil (you’ll have to wait another year!). And just ask any middle class Mexican if he’s been able to stop paying taxes if you believe the failed state theory.

On the other hand, I don’t want to come off as glib, dismissing all the extra deaths the drug war has caused, which should total in the tens of thousands by time it is finished. This will be the most violent conflict in Mexico since the Cristero War. But no, Mexico is not in any danger of turning into a failed state, or having the drug cartels team up with Marxist revolutionaries to topple the Mexican Government and have Blackwater run PEMEX or whatever nonsense people like to believe. There will not be a phase transition if you will.

Drug use has increased a lot in Mexico. According to the National Survey of Addictions (“Encuesta Nacional de Adicciones 2008”), the percentage of the population who had consumed cocaine doubled from 1.2% in 2002 to 2.4% in 2008. The percentage of people who consumed cocaine in Mexico during the last year was .4%, what’s more, in Mexico one fourth of cocaine is consumed as crack compared to 14% in the US.

Since drug addicts are more likely to be gang members and serve as cannon fodder for the drug cartels, I thought of running some regressions comparing the homicide rate in 2008 with the percentage of the population that consumes different drugs. The results show a significant the relationship between cocaine and amphetamine consumption and homicide rates, the regression for marijuana was not significant, and the regression for consumption of all illegal drugs was barely insignificant (p = 0.7). Still, there were some big outliers in the states where violence has increased the most.

Sources: National Survey of Addictions and INEGI

Obviously, this is just a regression and in no way proves causation. One way to mitigate the problem would be to analyze the data from the previous survey in 2002 and see if the states with the biggest increases in drug use suffered the biggest increases in violence. Still, it may be due to the fact that people in states that are home to the drug cartels are simply more familiar with drugs. But it does show that even if the drug war stopped tomorrow things aren’t suddenly going to get better.

According to government statistics drug eradication efforts have faltered in spite of the drug war. Since there were lots of army operations designed to destroy marijuana crops (specifically Operation Triangulo Dorado) you’d think that at least the amount of marijuana eradicated in Mexico would be higher. But as far as marijuana and poppy are concerned, the drug war has been a failure, with less hectares eradicated, and an increase in area under cultivation.

Source: 2010 International Narcotics Control Strategy Report (INCSR)

To top it off the US government has been delaying the release of a report describing how easy it is for the cartels to produce methamphetamines.

However, Mexico has seized more cocaine since the drug war started, and this has been reflected in an increase in its street price. Considering that cocaine consumption has stayed constant in the US, this just means the drug cartels increased the rents they obtain. It would be interesting to know if the navy has been in charge of most cocaine seizures.

Sources: The execution rate after 2006 is based on the average of reported executions by Milenio and Reforma

Well, at least the President of Mexico got it right and the homicide rate did start rising in 2004 (though not according to the police records). But we can do better than that and look at the homicide and the homicide with firearm rates in the states that have had the biggest increases in murders.

Sources

Homicides involving a firearm pretty much follow the homicide rate. If it were true that the assault weapons ban expiration were the cause of the drug war, we would expect homicides with firearm as a proportion of all homicides to have increased since the ban expired. And indeed, the proportion of homicides with a firearm increased in all of Mexico by more than 10% since 2004! Keep in mind that while the assault weapons ban might have had an effect in the rest of Mexico (I’ll talk about that in another post), what I’m interested in is whether it caused the drug war. Again we can look at the states with the big increases in homicides or which are home to drug cartels:

Sources

As far as I can see only Sonora has an upward trend that predates 2008, and it started before the assault weapon ban expired. If the ban were to blame for the drug war we’d see big increases in the proportion of homicides with firearms since 2004-2005 in the states where most of the drug violence has occurred, and that’s simply not the case.