While US Hispanics share many cultural similarities they are in reality an amalgamation of many different groups that have their origins in many different countries and are not monolithic.
Examining Hispanic homicide mortality as a single group as is usually done in the so-called social sciences would leave us with an incomplete picture of violence in their vibrant and diverse communities since we can’t assume that their attitudes, cultural practices, or genetic origins are similar.
Heterogeneity in percentage foreign born, Native American and Black admixture and educational selectivity in who emigrates, all make it worthwhile to break down Hispanics into subgroups. Fortunately, the CDC records the Hispanic origin of the deceased in it’s death certificate data.
To analyze trends by Hispanic subgroup I downloaded homicide data from the multiple cause of death files for the years 2003-2016 from the NBER and used the 2001-2016 population from the American Community Survey, 2001-2017:
Average of 2003-2016 data
| race/ethnicity | total homicides | average population (2003-2016) | average rate |
| Black | 128445 | 37445852 | 22.904184 |
| Puerto Rican | 4859 | 4636070 | 7.101312 |
| Mexican | 32349 | 31797661 | 6.975989 |
| All Hispanics | 48414 | 49600521 | 6.669951 |
| Central American | 3589 | 4220482 | 6.034352 |
| Cuban | 1213 | 1807627 | 4.572700 |
| Dominican | 789 | 1495811 | 3.504267 |
| White | 83487 | 196678225 | 2.830295 |
| South American | 583 | 2897989 | 1.340240 |
| Central American and South American | 933 | NA | NA |
| Other | 4099 | NA | NA |
| Other Races | 9024 | NA | NA |
| Unknown | 1636 | NA | NA |
I’m actually under-counting rates by a tiny bit because of the unknowns. In particular, homicide rates for Central And South Americans could be vastly underestimated because of the Central and South American category. (The ACS has data by country and the CDC has the categories Central American; South American; and Central and South American.)
Here’s the information in chart form:
I’m kind of surprised that Dominicans have such a low homicide rate. Dominicans had very low homicide rate in 2003, so maybe there are some problems with how they’re coded. The obvious explanation being that some Dominicans are coded as simply Black (but I’ve never met a Dominican who would describe himself as Black)
Here’s the data visualized as a time series. (lol, the Ferguson effect is really obvious)
The chart shows that as of 2016 Mexicans are a little less violent than Puerto Ricans, but it is a recent phenomenon, they used to be more violent back in 2007.
And a chart showing only showing Hispanics and Whites:
It’s also worthwhile to remember from my other post that in 2003 the Hispanic homicide rate was well above the national average, while in 2016 it was a little bit lower (i.e. Hispanic homicides rates are decreasing faster that those of other races)
To download the NBER mortality files you can use this bash script (you’ll need to install [csvkit] (https://csvkit.readthedocs.io/en/1.0.3/), and R with dplyr)
1#!/bin/bash23OUT=mort1999-2016.csv4rm -f $OUT5DATADIR=data6mkdir -p $DATADIR7for i in {2003..2016}8do9 if [ ! -f "$DATADIR"/mort$i-homicide.csv ]; then10 TEMPFILE=$(tempfile).zip11 wget -O "$TEMPFILE" http://www.nber.org/mortality/$i/mort$i.csv.zip12 unzip -p "$TEMPFILE" > "$TEMPFILE".csv13 # the files have a column named ucod with the ICD-10 codes,14 # we need to find it's number to be able to filter15 col_number=$(awk -F',' ' { for (i = 1; i <= NF; ++i) print i, $i; exit } ' "$TEMPFILE".csv | grep ucod | sed 's|"ucod"||g')16 # filter to include only homicides and non-terrorism17 # operations of war/legal interventions18 csvgrep -c "$col_number" -f icd-homicide-codes.txt "$TEMPFILE".csv >> "$DATADIR"/mort$i-homicide.csv19 rm -f "$TEMPFILE" "$TEMPFILE".csv20 fi21 # join all the files22 if [ -f $OUT ]; then23 # use R to concat the csv files (but only include some columns24 # like ucod, sex, etc)25 Rscript -e 'options(stringsAsFactors = FALSE);df=dplyr::bind_rows(read.csv("mort1999-2016.csv"), read.csv(commandArgs(TRUE)[1]));write.csv(df[,-grep("recor|econ|enico|rnifla", names(df))], "mort1999-2016.csv", row.names=FALSE)' "$DATADIR"/mort$i-homicide.csv26 else27 cp "$DATADIR"/mort$i-homicide.csv $OUT2829 fi3031done
Because the CDC’s estimates of legal intervention deaths are unreliable I include them in the homicide totals. These are the codes I used to filter homicides (the ones starting with ’#’ were excluded):
icd-homicide-codes.txt
1#U011 (Terrorism involving destruction of aircraft)2#U012 (Terrorism involving other explosions and fragments)3#U014 (Terrorism involving firearms)4X855X866X877X888X899X9010X9111X9212X9313X9414X9515X9616X9717X9818X9919Y0020Y0121Y0222Y0323Y0424Y0525Y06026Y06127Y06228Y06829Y06930Y07031Y07132Y07233Y07334Y07835Y07936Y0837Y0938Y35039Y35140Y35241Y35342Y35543Y35644Y35745Y36246Y36447Y36648Y36749Y36850Y36951Y87152Y89053#Y891 (Sequelae of war operations)
