In davben/arvig: Data on anti-refugee violence in Germany since 2014

We originally published the ARVIG dataset in 2016 and documented it in [@bencek_2016]. This vignette replicates all figures and tables of the paper.

Installation

As we have continually updated the ARVIG dataset and information was also added and corrected retroactively, we first need to install the version of the dataset used in the original publication.

#devtools::install_github("davben/arvig", ref = "v16.1.0")
library(tidyverse)
library(arvig)
library(sf)

Overview

First we can create an overview table distinguishing the different types of events present in the data and listing their frequencies.

arvig %>%
  count(category_en) %>%
  arrange(desc(n))

A few observations are mapped to more than one event type. In order to break those down into the four main categories, arvig offers the function split_events() that creates additional observations for each category.

arvig_split <- arvig %>%
  split_events()

Now we can take a look at the spatial distribution of events across Germany. To do this, we can use a district-level shapefile of Germany that is included in the arvig package and plot all events based on their geographical coordinates. Since the geo-coding is only precise at the level of municipalities, multiple events in the same town or city are plotted on top of one another. Adjusting the level of transparency can help visualize repeated event locations (adding a small amount of random noise to the coordinates would be an alternative solution, though one needs to be careful not to misrepresent the data).

load(system.file("extdata", "german_districts.rda", package = "arvig"))

arvig_split %>%
  ggplot() +
  geom_sf(data = german_districts, fill = "#cccccc") +
  geom_point(aes(longitude, latitude, colour = category_en), alpha = 0.4) +
  scale_color_colorblind() +
  coord_sf(crs = 4326) +
  theme_map() +
  theme(panel.grid.major = element_line(colour = "transparent")) +
  theme(legend.position = "bottom") +
  guides(colour = guide_legend(ncol = 2, direction = "horizontal", override.aes = list(alpha = 1),
                               title = "Event type", title.position = "top", title.hjust = 0.5))

Beside the spatial variance temporal clusters of events may also be of interest. A histogram of event counts by day helps to get a first impression.

arvig_split %>%
  count(date) %>%
  complete(date = full_seq(date, period = 1), fill = list(n = 0)) %>% # include days with zero observations
  ggplot(aes(n)) +
  geom_histogram(binwidth = 1, boundary = -0.5, fill="#E69F00", colour = "black") +
  labs(x = "Events per day", y = "Count") +
  theme_bw()