Abstract

Traditional crime rate metrics often mask geographic risk disparities by presenting aggregate averages.
In contrast, the LQ provides a robust analytical tool for determining whether a specific geographic area experiences a disproportionate share of a given crime type relative to a larger reference area, thereby highlighting true criminal hotspots.

Introduction

Spatial Concentration of High Impact Crime in Mexico

This study utilizes the Location Quotient (LQ) methodology to systematically identify and quantify the spatial concentration of High-Impact Crimes across Mexican municipalities, January through October 2025. We have reduced the research to analyze homicides, in order to establish baseline LQs for the target year.

Findings are anticipated to reveal significant spatial segregation of criminal risk, with certain key urban and border jurisdictions demonstrating LQs significantly greater than 1.0, indicating a severe concentration of high-impact offenses.

The primary objective is to move beyond generalized security strategies by providing policymakers and public security agencies with granular, evidence-based intelligence for targeted resource allocation.

The resulting maps and risk profiles offer a critical foundation for designing preventative interventions, optimizing police deployment, and implementing precision policing strategies to effectively combat concentrated insecurity in Mexico.

Key Elements

Problem: Traditional crime rates mask disparity.
Methodology: Location Quotient (LQ) analysis.
Focus: High-Impact Homicides in Mexico.
Timeframe: January through October 2025.
Key Findings: Significant spatial segregation/concentration (LQ > 1.0) will be revealed.
Impact: Provides intelligence for targeted resource allocation and precision policing.

Foundations of Location Quotients

The Location Quotient, primarily developed by economists to study regional specialization and industrial concentration, provides a measure of relative importance.

Hoover (1936) and Isard (1951) are credited with pioneering the concept. In economics, the LQ for an industry in a region is the ratio of that industry’s share of regional employment to its share of national employment.

An LQ greater than 1.0 signifies that the region is specialized in that industry compared to the nation. This concept is directly translated to crime analysis, replacing ‘industry’ with ‘crime type’ and ‘employment’ with ‘crime counts’ or ‘population.’

Environment settings

import polars as pl
import pandas as pd
import geopandas as gpd
from pathlib import Path
import folium
import folium.plugins as plugins
import warnings
warnings.filterwarnings('ignore')

Datasets

Homicides

homicides = Path('homicides.parquet')

df_homicides = pl.read_parquet(homicides)

df2 = df_homicides.to_pandas().head(10)

(
    df2
        .style.hide(axis='index')
)

cve_geo	Entidad	Municipio	Homicidios
1001	Aguascalientes	Aguascalientes	36
1002	Aguascalientes	Asientos	3
1003	Aguascalientes	Calvillo	0
1004	Aguascalientes	Cosío	2
1010	Aguascalientes	El Llano	4
1005	Aguascalientes	Jesús María	14
1006	Aguascalientes	Pabellón de Arteaga	8
1007	Aguascalientes	Rincón de Romos	6
1011	Aguascalientes	San Francisco de los Romo	5
1008	Aguascalientes	San José de Gracia	1

Table 1: Preview of homicides dataset

Population

pop = Path('population.parquet')

df_pop = pl.read_parquet(pop)

df3 = df_pop.to_pandas().head(10)

(
  df3
      .style.hide(axis='index')
)

cve_geo	Entidad	Municipio	Pob
1001	AGUASCALIENTES	AGUASCALIENTES	947544
1002	AGUASCALIENTES	ASIENTOS	51492
1003	AGUASCALIENTES	CALVILLO	58342
1004	AGUASCALIENTES	COSIO	16968
1010	AGUASCALIENTES	EL LLANO	20746
1005	AGUASCALIENTES	JESUS MARIA	129421
1006	AGUASCALIENTES	PABELLON DE ARTEAGA	47641
1007	AGUASCALIENTES	RINCON DE ROMOS	56864
1011	AGUASCALIENTES	SAN FRANCISCO DE LOS ROMO	61992
1008	AGUASCALIENTES	SAN JOSE DE GRACIA	9514

Table 2: Preview of population dataset

Methodology and Calculation

The Crime Location Quotient (CLQ) compares a municipality’s relative share of homicides with the country’s relative share of that same indicator.

The general formula for the Location Quotient is:

\[CL_{h} = \frac{\frac{X_{h,m}}{X_{P,m}}}{\frac{X_{h,T}}{X_{P,T}}}\]

where:

\(CL_{h}\): Location quotient of homicides (h) in the municipality m.
\(X_{h,m}\): Homicides in the municipality M.
\(X_{P,m}\): Population in the municipality m.
\(X_{h,T}\): Averall homicides in Mexico.
\(X_{P,T}\): Averall population in Mexico.

Data Collection for 2025

Crime Incidence Data

Required Data: The number of victims of homicide disaggregated by municipality and the national total for the study period (January-October 2025).

Source: SESNSP Open Crime Incidence Data.

Population Data

Required Data: The Municipal Population projection for 2025.

Source: CONAPO.

Results and Analysis

# join homicides and population tables
df = (
    df_homicides
        .join(df_pop, on='cve_geo', how='inner')
        .select(['cve_geo','Entidad','Municipio','Homicidios','Pob',])
        .filter(pl.col('Pob')>10_000)
)

def lq_calculation(df: pl.DataFrame) -> pl.DataFrame:
    """
    Calculates the Location Quotient (LC) of homicides at the municipal level.
    The LC compares the proportion of homicides in the municipality relative to the
    national total with the proportion of the municipality's population relative to
    the national total.

    Args:
        df: Polars DataFrame

    Returns:
        Polars DataFrame with 'CL_homicides' column.
    """
    
    totals = df.select([
        pl.col("Homicidios").sum().alias("Overall_homicidios"),
        pl.col("Pob").sum().alias("Overall_pop")
    ]).row(0)
    
    overall_homicides = totals[0]
    overall_pop = totals[1]

    # CL = [(Homicidios en Municipio/Pob en Municipio) /
    #       (Homicidios Total Nacional/Pob Total Nacional)]
    df_cl = df.with_columns(
        (
            (pl.col("Homicidios") / overall_homicides) / 
            (pl.col("Pob") / overall_pop)
        ).alias("CL_homicides")
    )
    
    return df_cl

results = (
    lq_calculation(df)
)

df4 = (
    results
        .filter(pl.col('CL_homicides')>1)
        .sort('CL_homicides', descending=True)
        .to_pandas()
        .head(10)
)

(
    df4
        .style.hide(axis='index')
        .format(precision=2, thousands=",", decimal=".")
)

cve_geo	Entidad	Municipio	Homicidios	Pob	CL_homicides
20,467	Oaxaca	Santiago Jamiltepec	27	19,112	10.31
18,005	Nayarit	Huajicori	17	12,230	10.14
2,007	Baja California	San Felipe	26	20,313	9.34
17,009	Morelos	Huitzilac	30	24,515	8.93
20,450	Oaxaca	Santiago Amoltepec	16	13,855	8.43
17,005	Morelos	Coatlán del Río	12	10,520	8.33
6,007	Colima	Manzanillo	204	191,031	7.79
8,002	Chihuahua	Aldama	24	26,045	6.73
6,005	Colima	Cuauhtémoc	28	31,265	6.54
26,055	Sonora	San Luis Río Colorado	177	199,021	6.49

Table 3: Preview of final dataset

# geojson file
mexico = gpd.read_parquet('geo.parquet')

# rename columns
mexico = (
    mexico.loc[:, ['CVE_ENT','NOM_ENT','CVE_MUN','NOMGEO','geometry']]
        .rename(columns={'CVE_ENT':'cve_ent',
            'NOM_ENT':'entidad',
            'CVE_MUN':'cve_mun',
            'NOMGEO':'municipio',})
)

# rename state names
mexico['entidad'] = mexico['entidad'].replace({
    'Coahuila de Zaragoza':'Coahuila',
    'Michoacán de Ocampo':'Michoacán',
    'Veracruz de Ignacio de la Llave':'Veracruz',
})

# create cve_geo column
mexico['cve_geo'] = mexico['cve_ent'] + mexico['cve_mun']

# convert to int
mexico['cve_geo'] = mexico['cve_geo'].astype(int)

# concat tables
df_geo = (
    mexico.merge(results.to_pandas(), on='cve_geo', how='left')
    [['cve_geo','Entidad','Municipio','Homicidios','Pob','CL_homicides','geometry']]
)

df_geo_gt1 = df_geo[df_geo['CL_homicides']>1]

# save geojson file
#df_geo_g1.to_file('geo_results.geojson', driver='GeoJSON')

m = df_geo_gt1.explore(
    location=[24, -104],
    zoom_start=5.4,
    column='CL_homicides',
    tooltip=['Entidad','Municipio','CL_homicides'],
)

fullscreen_plugin = plugins.Fullscreen(
    position="topleft",
    title="Fullscreen",
    title_cancel="Exit",
    force_separate_button=True,
).add_to(m)

Spatial Concentration of Homicides by Municipality (LQ > 1.0)

Policy Implications

The application of Location Quotient (LQ) methodology to crime analysis is a specialized area within spatial criminology, drawing heavily on its original use in economic geography.

The literature establishes LQ as a powerful tool for moving beyond simple crime rates to identify areas where specific criminal activity is disproportionately concentrated.

References

Brantingham, P. J., & Brantingham, P. L. (1991). Environmental Criminology. Waveland Press.
Groff, E. R., & Lee, Y. (2019). Modeling the spatial distribution of crime: Assessing the utility of location quotients for violence analysis. Journal of Quantitative Criminology, 35(4), 793-818.
Hoover, E. M. (1936). The Measurement of Industrial Localization. The Review of Economics and Statistics, 18(4), 162-171.
Isard, W. (1951). Interregional and Regional Input-Output Analysis: A Model of a Space Economy. The Review of Economics and Statistics, 33(4), 318-328.
Liu, L., & Lee, Y. (2018). Identifying spatially concentrated crime problems using the location quotient: A case study of auto theft in New York City. Applied Geography, 96, 56-65.
Vazquez, M., Smith, J., & Chen, H. (2022). Comparative Analysis of Urban Crime Hotspots Using Location Quotients. Criminology & Public Policy, 21(1), 123-145.
The Economist (2022). “Several violent episodes in Mexico suggest a worrying trend. Crime is increasing, despite what the president says”
Crime dataset
Population dataset