Mexico City Crime According to the Statistics

Figure 1: Mexico City Crimes

Introduction

It’s important to recognize that crime trends are dynamic and influenced by various factors. Continued efforts are necessary to improve public safety and address the root causes of crime in Mexico City.

While efforts have been made to improve security through increased policing and community-based initiatives, persistent issues such as impunity and a lack of trust in law enforcement continue to hinder progress.

Addressing crime in Mexico City requires a multifaceted approach that includes improving data collection and analysis, strengthening law enforcement, and addressing the underlying socioeconomic factors that contribute to crime.

Analyzing crime trends in Mexico City requires considering both official statistics and the lived experiences of residents, as well as the interplay of local, national, and international forces.

Ongoing research and policy development are crucial to addressing the root causes of crime and fostering a safer environment for all.

Environment settings

Import libraries

#import libraries
import numpy as np
import pandas as pd
import polars as pl
import duckdb as db
import datetime as datetime
import json
import plotly.express as px
import folium
from folium.plugins import Fullscreen

Database connection

# get token
filename = 'credentials.json'

# read json file
with open(filename) as f:
    keys = json.load(f)

# read credentials
token = keys['md_token']

Duckdb is a powerful tool for data analysts and developers who need to perform fast and efficient analytical queries on large datasets, especially in environments where simplicity and portability are crucial.

# connect to motherduck cloud
conn = db.connect(f'md:?motherduck_token={token}')

# retrieve dataframe
df = conn.execute('select * from projects.cdmx_fgj_uncleaned').pl()

# close connection
conn.close()

print(f'The dataset contains {df.shape[0]:,.0f} rows')

The dataset contains 1,415,763 rows

Missing values

df.info()

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 1415763 entries, 0 to 1415762
Data columns (total 22 columns):
 #   Column             Non-Null Count    Dtype  
---  ------             --------------    -----  
 0   anio_inicio        1415748 non-null  float64
 1   mes_inicio         1415748 non-null  object 
 2   fecha_inicio       1415748 non-null  object 
 3   hora_inicio        1415748 non-null  object 
 4   anio_hecho         1415343 non-null  float64
 5   mes_hecho          1415343 non-null  object 
 6   fecha_hecho        1415342 non-null  object 
 7   hora_hecho         1415353 non-null  object 
 8   delito             1415749 non-null  object 
 9   categoria_delito   1415749 non-null  object 
 10  sexo               1168059 non-null  object 
 11  edad               931053 non-null   float64
 12  tipo_persona       1408196 non-null  object 
 13  calidad_juridica   1415748 non-null  object 
 14  competencia        1415749 non-null  object 
 15  colonia_hecho      1340796 non-null  object 
 16  colonia_catalogo   1323317 non-null  object 
 17  alcaldia_hecho     1413284 non-null  object 
 18  alcaldia_catalogo  1372214 non-null  object 
 19  municipio_hecho    1413284 non-null  object 
 20  latitud            1340994 non-null  float64
 21  longitud           1340994 non-null  float64
dtypes: float64(5), object(17)
memory usage: 237.6+ MB

df = df.dropna(subset=['latitud'])

print(f'The dataset now has {df.shape[0]:,.0f} rows')

The dataset now has 1,340,994 rows

Note

We can see that the original dataset includes 1,415,763 rows.

However, there are many null values in different fields for considerable gaps.

As the dataset can be considered large, we decided to drop all null latitude rows, obtaining 1,340,994 rows.

Due to the missing rows in many fields, we kept around with 94.72% of the original dataset

Polars dataframe

Polars is a modern DataFrame library that prioritizes performance and efficiency. Its Rust-based architecture, combined with features like lazy evaluation and parallel processing, makes it a powerful tool for data professionals.

# convert to polars dataframe
df = pl.from_pandas(df)

df.schema

Schema([('anio_inicio', Float64),
        ('mes_inicio', String),
        ('fecha_inicio', String),
        ('hora_inicio', String),
        ('anio_hecho', Float64),
        ('mes_hecho', String),
        ('fecha_hecho', String),
        ('hora_hecho', String),
        ('delito', String),
        ('categoria_delito', String),
        ('sexo', String),
        ('edad', Float64),
        ('tipo_persona', String),
        ('calidad_juridica', String),
        ('competencia', String),
        ('colonia_hecho', String),
        ('colonia_catalogo', String),
        ('alcaldia_hecho', String),
        ('alcaldia_catalogo', String),
        ('municipio_hecho', String),
        ('latitud', Float64),
        ('longitud', Float64)])

Data cleansing

df = (
    df.filter(
    # drop null and 2018 years
        (pl.col('anio_inicio')!=2018)
    ).with_columns(
    # create datetime field
        (pl.col('fecha_inicio').cast(pl.String) + ' ' + pl.col('hora_inicio')
             .cast(pl.String)).alias('fecha_inicio')
    ).select(
    # exclude columns
        pl.exclude('anio_inicio','mes_inicio','hora_inicio',
                   'anio_hecho','mes_hecho','fecha_hecho','hora_hecho')
    ).with_columns(
        fecha_inicio=pl.col('fecha_inicio').str.to_datetime()
    ).drop_nulls(subset='fecha_inicio')
)

df.select(pl.all().is_null().sum()).to_dicts()

[{'fecha_inicio': 0,
  'delito': 0,
  'categoria_delito': 0,
  'sexo': 237070,
  'edad': 449780,
  'tipo_persona': 6858,
  'calidad_juridica': 1,
  'competencia': 0,
  'colonia_hecho': 382,
  'colonia_catalogo': 17677,
  'alcaldia_hecho': 3,
  'alcaldia_catalogo': 953,
  'municipio_hecho': 3,
  'latitud': 0,
  'longitud': 0}]

Note

We can see that even if we dropped around 75,000 rows, there continue to be many fields with empty rows, such as sex, age, neighborhood, mayorship and municipality.

Data cleaning for age

Age goes from 0.0 up to 369.0 years old!

# case for age
df = df.with_columns(
    pl.when(pl.col('edad') < 18).then(18)
    .when(pl.col('edad')>99).then(99)
    .otherwise('edad')
    .alias('edad')
)

We have cleaned age values by setting age less than 18 to 18, and age values gretar than 99 to 99.

Exploratory Data Analysis

Accrued crimes by year

years = (
    df.sort('fecha_inicio')
        .group_by_dynamic('fecha_inicio', every='1y')
        .agg(pl.len().alias('crimes'))
)

(
    years
        .to_pandas()
        .style.format(
            {
            'fecha_inicio':'{:%Y}',
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

fecha_inicio	crimes
2019	256,827
2020	204,659
2021	228,627
2022	237,659
2023	239,402
2024	173,803

Table 1: Accrued Crimes in Mexico City by Year

fig = px.bar(years, 
             x='fecha_inicio',
             y='crimes',
             orientation='v',
             hover_data=['fecha_inicio','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City by Year',
             template='ggplot2',)

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Crimes')),
    )

fig.update_traces(marker_color='#7f0000',
                 texttemplate = "%{value:,.0f}",)

fig.show()

Figure 2: Crimes in Mexico City by Year

Time behavior of crimes

months = (
    df
        .sort('fecha_inicio', descending=False)
        .group_by_dynamic('fecha_inicio', every='1mo')
        .agg(pl.len().alias('crimes'))
)

(
    months
        .to_pandas()
        .tail(10)
        .style.format(
            {
            'fecha_inicio':'{:%b, %Y}',
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

fecha_inicio	crimes
Dec, 2023	17,358
Jan, 2024	18,354
Feb, 2024	18,750
Mar, 2024	19,797
Apr, 2024	19,923
May, 2024	20,879
Jun, 2024	19,209
Jul, 2024	19,429
Aug, 2024	19,149
Sep, 2024	18,313

Table 2: Accrued Crimes in Mexico City by Month

fig = px.line(months, 
             x='fecha_inicio',
             y='crimes',
             hover_data=['fecha_inicio','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City 2019-2024',
             template='ggplot2',)

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Crimes')),
    )

fig.update_traces(line_color='#7f0000',
                  line={'width':3},
                  )

fig.show()

Figure 3: Time Behavior of Crimes in Mexico City

Crimes by sex

df_sex = (
    df.group_by('sexo')
        .agg(pl.len().alias('crimes'))
)

# convert to pandas
df_sex = df_sex.to_pandas()
# rename row
df_sex['sexo'] = df_sex['sexo'].replace({None:'NA'})

(
    df_sex
        .sort_values('crimes')
        .style.format(
            {
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

sexo	crimes
NA	237,070
Femenino	535,960
Masculino	567,947

Table 3: Accrued Crimes in Mexico City by Sex

fig = px.bar(df_sex.sort_values('crimes'),
             y='sexo',
             x='crimes',
             orientation='h',
             hover_data=['sexo','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City by Sex',
             template='ggplot2',
             text='crimes',
             )

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Sex')),
    )

fig.update_traces(marker_color='#7f0000',
                 texttemplate = "%{value:,.0f}",)

fig.show()

Figure 4: Accrued Crimes in Mexico City by Sex

Crimes by age

df_edad = (
    df.group_by('edad')
        .agg(pl.len().alias('crimes'))
        .sort('edad')
        .drop_nulls()
)

(
    df_edad
        .to_pandas()
        .sample(10)
        .style.format(
            {
            'edad':'{:,.0f}',
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

edad	crimes
56	10,789
47	16,414
44	16,818
90	313
85	789
37	21,022
82	1,181
88	493
50	16,098
73	3,296

Table 4: Accrued Crimes in Mexico City by Age

fig = px.bar(df_edad,
             x='edad',
             y='crimes',
             orientation='v',
             hover_data=['edad','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City by Age',
             template='ggplot2',)

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Age')),
    )

fig.update_traces(marker_color='#7f0000',)

fig.show()

Figure 5: Crimes in Mexico City by Age Distribution

Crimes by neighborhood

df_colonia = (
    df.group_by('colonia_hecho')
        .agg(pl.len().alias('crimes'))
        .top_k(10, by='crimes')
        .sort('crimes', descending=False)
)

(
    df_colonia
        .to_pandas()
        .style.format(
            {
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

colonia_hecho	crimes
PEDREGAL DE SANTO DOMINGO	10,073
JUÁREZ	11,423
BUENAVISTA	11,749
NARVARTE	11,906
AGRÍCOLA ORIENTAL	12,024
MORELOS	12,518
ROMA NORTE	14,878
DEL VALLE CENTRO	17,178
DOCTORES	24,711
CENTRO	40,066

Table 5: Accrued Crimes in Mexico City by Neighborhood

fig = px.bar(df_colonia,
             y='colonia_hecho',
             x='crimes',
             orientation='h',
             hover_data=['colonia_hecho','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City - Top 10 Neighborhoods',
             template='ggplot2',)

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Alcaldia')),
    )

fig.update_traces(marker_color='#7f0000',
                 texttemplate = "%{value:,.0f}",)

fig.show()

Figure 6: Crimes in Mexico City by Neighborhood

Crimes by mayorship

df_alcaldia = (
    df.group_by('alcaldia_hecho')
        .agg(pl.len().alias('crimes'))
        .sort('crimes')
        .filter(pl.col('alcaldia_hecho')!='FUERA DE CDMX')
)

(
    df_alcaldia
        .to_pandas()
        .style.format(
            {
            'crimes':'{:,.0f}',
            }
        ).hide(axis='index')
)

alcaldia_hecho	crimes
MILPA ALTA	13,184
CUAJIMALPA DE MORELOS	23,329
LA MAGDALENA CONTRERAS	27,285
TLAHUAC	41,547
XOCHIMILCO	45,924
IZTACALCO	60,427
AZCAPOTZALCO	64,561
VENUSTIANO CARRANZA	78,446
MIGUEL HIDALGO	84,242
TLALPAN	84,382
COYOACAN	93,953
ALVARO OBREGON	94,083
BENITO JUAREZ	101,759
GUSTAVO A. MADERO	139,294
CUAUHTEMOC	188,964
IZTAPALAPA	199,591

Table 6: Accrued Crimes in Mexico City by Mayorship

fig = px.bar(df_alcaldia,
             y='alcaldia_hecho',
             x='crimes',
             orientation='h',
             hover_data=['alcaldia_hecho','crimes',],
             height=500,
             width=830,
             title='Crimes in Mexico City by Mayorship',
             template='ggplot2',)

fig.update_layout(
    xaxis=dict(title=dict(text='')),
    yaxis=dict(title=dict(text='Alcaldia')),
    )

fig.update_traces(marker_color='#7f0000',
                 texttemplate = "%{value:,.0f}",)

fig.show()

Figure 7: Crimes in Mexico City by Mayorship

Heat map of crimes

df_map = (
    df.with_columns(
        (pl.col('colonia_hecho') + ', ' + pl.col('alcaldia_hecho')).alias('neighborhood')
    )
        .filter(pl.col('alcaldia_hecho')!='FUERA DE CDMX')
        .group_by('neighborhood', maintain_order=True)
        .agg(latitude=pl.col('latitud').mean(),
             longitude=pl.col('longitud').mean(),
             crimes=pl.col('delito').len()
            )
)

heat_map = df_map.to_pandas()

(
    heat_map.sort_values('crimes', ascending=False)
        .head(10)
        .style.format(
        {
            'latitude':'{:,.4f}',
            'longitude':'{:,.4f}',
            'crimes':'{:,.0f}',
        }
        ).hide(axis='index')
)

neighborhood	latitude	longitude	crimes
CENTRO, CUAUHTEMOC	19.4327	-99.1375	40,042
DOCTORES, CUAUHTEMOC	19.4200	-99.1486	24,711
DEL VALLE CENTRO, BENITO JUAREZ	19.3831	-99.1682	17,178
ROMA NORTE, CUAUHTEMOC	19.4184	-99.1627	14,878
AGRÍCOLA ORIENTAL, IZTACALCO	19.3947	-99.0708	12,008
NARVARTE, BENITO JUAREZ	19.3930	-99.1542	11,906
JUÁREZ, CUAUHTEMOC	19.4268	-99.1628	11,408
PEDREGAL DE SANTO DOMINGO, COYOACAN	19.3275	-99.1677	10,073
POLANCO, MIGUEL HIDALGO	19.4335	-99.1956	10,019
AGRÍCOLA PANTITLAN, IZTACALCO	19.4104	-99.0649	9,662

Table 7: Crimes in Mexico City Heatmap

from folium.plugins import HeatMap

basemap = folium.Map(location=[19.35, -99.12], zoom_start=10.4)

HeatMap(heat_map[['latitude', 'longitude', 'crimes']]).add_to(basemap)

<folium.plugins.heat_map.HeatMap at 0x310ecc530>

folium.plugins.Fullscreen().add_to(basemap)
basemap

Make this Notebook Trusted to load map: File -> Trust Notebook

Figure 8: Mexico City Crime Heatmap by Mayorship

# mexico city crime map
m = folium.Map(
    location=[19.35, -99.12],
    zoom_start=10,
    control_scale=False,
)
# Layers
Crime = folium.FeatureGroup(name='<u><b>Place</b></u>', show=True)
m.add_child(Crime)
#draw marker with symbol you want at base
my_symbol_css_class= """ <style>
.fa-mysymbol3:before {
    font-family: Gill Sans; 
    font-weight: bold;
    font-size: 11px;
    color: white;
    background-color:'';
    border-radius: 10px; 
    white-space: pre;
    content: 'P';
    }
</style>    
        """
# the below is just add above  CSS class to folium root map      
m.get_root().html.add_child(folium.Element(my_symbol_css_class))
# then we just create marker and specific your css class in icon like below
for i in heat_map.index:
   html=f"""
        <p style="font-size: 14px;">{heat_map.iloc[i]['neighborhood']}</font></p>
        <p style="font-size: 14px;">Total crimes: {heat_map.iloc[i]['crimes']}</font></p>
        """
   iframe = folium.IFrame(html=html, width=220, height=90)
   popup = folium.Popup(iframe, max_width=250)
   folium.Marker(
        location = [heat_map.iloc[i]['latitude'], heat_map.iloc[i]['longitude']],
        icon = folium.Icon(color='darkred', prefix='fa', icon='fa-mysymbol3'),
        popup = popup,
        tooltip = heat_map.iloc[i]['neighborhood']
    ).add_to(Crime)
folium.plugins.Fullscreen().add_to(m)
m

Make this Notebook Trusted to load map: File -> Trust Notebook

Figure 9: Mexico City Crime Hubs by Neighborhood

Conclusions

Analyzing crime in Mexico City presents significant challenges, and the limitations of available data create substantial obstacles to drawing definitive conclusions.

• Varied Crime Landscape:

  Mexico City experiences a range of criminal activities, from petty theft to organized crime-related violence. This creates a multifaceted challenge for law enforcement. The distribution of crime is uneven, with certain areas experiencing higher rates of specific offenses.

• Impact of Socioeconomic Factors:

  Poverty, inequality, and lack of opportunity contribute to crime rates in certain neighborhoods. Addressing these underlying socioeconomic issues is crucial for long-term crime reduction.

• Efforts in Crime Reduction:

  Mexico City has implemented strategies, including increased data transparency and targeted policing, aimed at reducing crime. The use of open data policies, has shown to be a helpful tool in crime reduction.

• Challenges Remain:

  Despite progress, challenges persist, including issues related to organized crime, corruption, and the effectiveness of the criminal justice system. The issue of firearms, and their influx from the United States, is a large factor in violent crimes.

• The importance of Data:

  The use of data to create crime maps, and to find crime hot spots, has become a very important tool for law enforcement.

The hardship of incomplete data makes it essential to improve transparency and data sharing among all agencies involved in public security.

A significant portion of crimes goes unreported due to distrust in authorities, fear of retaliation, or the perception that reporting is futile.

This underreporting creates a distorted picture of the actual crime situation.

References

• Portal de Datos Abiertos CdMx (2024). Víctimas en carpetas de investigación Dataset. Jan 2019 through Sept 2024. Retrieved from website

Contact

Jesus LM

Economist & Data Scientist

Medium | Linkedin | Twitter