openpolicedata 0.5.4

A Python library providing easy access to 276+ incident-level open datasets released by police departments including traffic stops, use of force, officer-involved shootings, and complaints data

OpenPoliceData

OpenPoliceData is a Python package that provides easy access to 365 (and growing) incident-level open datasets from police departments around the United States. Datasets include traffic stops, use of force, officer-involved shootings, complaints, and other types of police interactions.

Users request data by department name and type of data, and the data is returned as a pandas DataFrame. There is no need to manually find the data online or to know how to work with open data APIs (ArcGIS, Socrata, etc.).

Installation

Examples

Contributing

Querying Available Data

Loading and Working with Datasets

Installation

The source code is available at https://github.com/openpolicedata/openpolicedata.

OpenPoliceData can be installed from the Python Package Index (PyPI):

pip install openpolicedata

Additionally, geopandas can be installed to enable downloaded data tables to be returned as geopandas DataFrames instead of pandas DataFrames when there is geographic data. It is recommended to use conda to install geopandas.

Examples

You can try out OpenPoliceData and run examples online on Binder.

Basic usage of OpenPoliceData simply involves:

1. Finding datasets
2. Loading datasets

The query function is used to find data. To get all available datasets, query can be used with no inputs. To get all police stops datasets in Virginia, try the following:

> import openpolicedata as opd
> datasets = opd.datasets.query(state="Virginia", table_type="STOPS")
> datasets

State	SourceName	Agency	TableType	coverage_start	coverage_end
Virginia	Dumfries	Dumfries	STOPS	2021-07-01	2023-03-31
Virginia	Virginia	MULTIPLE	STOPS	2021-07-01	2023-03-31

(only 1st 6 columns shown above)

There are 2 stops (containing pedestrian and traffic stops) datasets in Virginia: 1 for the town of Dumfries and 1 for every police department in Virginia (indicated by Agency being MULTIPLE). Let's load in data from the state data with load_data_from_url. To do this, we will first need to create a Source:

> src = opd.Source("Virginia")

If we are only interested in data from a single police department, we will need to set the optional agency input for load_data_from_url if we do not want to load data from all police departments in the state. get_agencies can be used to find the exact department name (if it is not known) by searching for agencies containing the partial_name input ("Arlington" in this case).

> agencies = src.get_agencies(table_type="STOPS", partial_name="Arlington")
> agencies
["Arlington County Sheriff's Office", 'Arlington County Police Department']

Now, we are ready to load the data.

> tbl = src.load_from_url(year=2021, table_type="STOPS", agency="Arlington County Police Department")
> tbl.table.head(n=3)

incident_date	agency_name	agency	reason_for_stop	race	ethnicity
2021-01-01	Arlington County Police Department	ARLINGTON CO	OTHER	WHITE	HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	EQUIPMENT VIOLATION	WHITE	NON-HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	TRAFFIC VIOLATION	BLACK OR AFRICAN AMERICAN	NON-HISPANIC

(only 1st 6 columns shown above)

tbl.table is a pandas DataFrame and therefore, can be analyzed directly in Python using the powerful pandas analysis library. Alternatively, the table can be exported to a CSV file to analyze with your favorite tool:

> tbl.to_csv()

More examples can be found in the opd-examples repo.

Contributing

If you're interesting in helping out, see our Contributing Guide

Import

> import openpolicedata as opd

Querying Available Data

opd.datasets.query(source_name=None, state=None, agency=None, table_type=None)

Query the available datasets to see what is available. Various filters can be applied. By default, all datasets are returned.

> datasets = opd.datasets.query(state="California")
> datasets.head()

State	SourceName	Agency	TableType	Year
California	Anaheim	Anaheim	TRAFFIC STOPS	MULTI
California	Bakersfield	Bakersfield	TRAFFIC STOPS	MULTI
California	California	MULTI	STOPS	2018
California	California	MULTI	STOPS	2019
California	California	MULTI	STOPS	2020

(only 1st 5 columns shown above)

datasets is a pandas DataFrame. The first 5 datasets available from California include traffic stops data from multiples years from Anaheim and Bakersfield and data from every agency in California for all types of police stops for years 2018, 2019, and 2020.

opd.datasets.num_unique()

Returns the number of unique datasets in OpenPoliceData. This counts the number of datasets from distinct sources AND table types (stops, use of force, etc.).

opd.datasets.num_sources(full_states_only=False)

Returns the number of sources (police departments and states) that provide the data available in OpenPoliceData. Setting full_states_only to True returns only the number of states that share data for all agencies in the state.

opd.datasets.summary_by_state(by=None)

Returns a pandas DataFrame with the number of datasets available for each state. The optional input by can be used to further breakdown by "year" or "table".

opd.datasets.summary_by_table_type(by_year=False)

Returns a pandas DataFrame with the number of datasets available for each type of table (stops, use of force, etc.). Setting by_year to True also returns a breakdown of table types by year.

Loading and Working with Datasets

opd.Source(source_name, state=None)

Create a data source. A data source allows the user to easily import or export police data. It provides access to all datasets available from a source. source_name should match a value of SourceName for an available dataset. An optional state parameter is used to resolve ambiguities when the same source name is used in multiple states (such as multiple states have State Police).

> src = opd.Source(source_name="Virginia")
> src.datasets

State	SourceName	Agency	TableType	Year
Virginia	Virginia	MULTI	STOPS	MULTI

(only 1st 5 columns shown above)

There is 1 dataset available from the state of Virginia that contains data from every agency in Virginia for all types of police stops for multiple years.

get_tables_types()

Show all types of data available from a source.

> src.get_tables_types()
['STOPS']

get_years(table_type=None, force=False)

Show years available for one or more datasets. Results can be filtered to only show years for a specific table type. For CSV and Excel data types, get_years will not run unless force is set to True due to the necessity of reading in the entire file. It may be more efficient to run load_from_url and manually get the years.

> src.get_years(table_type="STOPS")
[2020, 2021, 2022]

get_agencies(table_type=None, year=None, partial_name=None)

Show agencies (police departments) that have data available. This is typically a single agency unless the data is from a state. Results can be filtered to only show agencies for a specific type of data and/or year. partial_name can be used to find only agencies containing a substring. This is useful for finding the exact name of a police department.

> agencies = src.get_agencies(partial_name="Arlington")
> print(agencies)
['Arlington County Police Department', "Arlington County Sheriff's Office"]

get_count( year=None, table_type=None, agency=None, force=False)

Get the number of records that would be returned for a table. table_type and year can be used to filter for a specific table. For datasets that allow for filtering by agency, the number of records can be requested for a specific agency by setting agency. For the Excel data type, get_count will not run unless force is set to True due to the necessity of reading in the entire file. It may be more efficient to run load_from_url and then to find the number of rows of the returned table.

load_from_url(year, table_type=None, agency=None, pbar=True, nrows=None, offset=0)

load_from_url_gen(year, table_type=None, agency=None, pbar=False, nbatch=10000, offset=0, force=False)

Load data from the source.

• load_from_url returns the entire data request at once as a DataFrame
• load_from_url_gen returns a generator for batch processing

Data for a year (i.e. 2020) or a range of years (i.e. [2020, 2022]) can be requested. If more than one data type is available, table_type must be specified. Optionally, for datasets containing multiple agencies (police departments) data, agency can be used to request data for a single agency. pbar can be set to false to not show a progress bar while loading. offset indicates the starting record in the data request. nrows is the total number rows to request. For example, to request records 10 to 30, offset would be 10 and nrow would be 20. offset and nrows allow requesting of data in batches. For load_from_url_gen, nbatch is the size of the batch read returned each iteration.

> agency = "Arlington County Police Department"
> tbl = src.load_from_url(year=2021, table_type="STOPS", agency=agency)
> tbl.table.head(n=3)

incident_date	agency_name	agency	reason_for_stop	race	ethnicity
2021-01-01	Arlington County Police Department	ARLINGTON CO	OTHER	WHITE	HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	EQUIPMENT VIOLATION	WHITE	NON-HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	TRAFFIC VIOLATION	BLACK OR AFRICAN AMERICAN	NON-HISPANIC

(only 1st 6 columns shown above)

> for tbl in src.load_from_url_gen(year=2021, table_type="STOPS", agency=agency, nbatch=1000):
      df = tbl.table.copy()
      # Do something with the 1000 row DataFrame that was loaded

The result of load_from_url is a Table object. The table contained in the Table object is either a geopandas or pandas DataFrame depending on whether the returned data contains geographic data or not.

to_csv(output_dir=None, filename=None)

Export table to CSV. The default output directory is the current directory. The default filename is automatically generated which enables the user to easily re-import the table to a new Table object.

> tbl.to_csv()

load_from_csv(year, output_dir=None, table_type=None, agency=None)

Import table from previously exported CSV. The directory to look in defaults to the current directory. The CSV file must have been automatically generated (see to_csv). year, table_type, and agency are defined the same as for load_from_url.

> new_src = opd.Source(source_name="Virginia")
new_t = new_src.load_from_csv(year=2021, agency=agency)


> tbl.table.head(n=3)

incident_date	agency_name	agency	reason_for_stop	race	ethnicity
2021-01-01	Arlington County Police Department	ARLINGTON CO	OTHER	WHITE	HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	EQUIPMENT VIOLATION	WHITE	NON-HISPANIC
2021-01-01	Arlington County Police Department	ARLINGTON CO	TRAFFIC VIOLATION	BLACK OR AFRICAN AMERICAN	NON-HISPANIC

(only 1st 6 columns shown above)

See the OpenPoliceData wiki for further documentation