Protest Dynamics in Serbia Following the Train Station Roof Collapse
Setting the Stage
On November 11, 2024, the roof of a train station in Novi Sad, Serbia collapsed, resulting in the deaths of 15 individuals. In the aftermath, protests erupted, initially focused on demanding justice for the victims. Over time, these protests grew in scale and scope, with student groups and civil society organizations leading the call for broader political reform. This project analyzes changes in protest activity in Serbia following the collapse, examining patterns by protest group type, geographic distribution, and the temporal evolution of events. The aim is to provide a data-driven understanding of how the political crisis has influenced civic engagement and activism in Serbia.
Data Source
This data comes from ACLED which is an event-based data project designed for disaggregated conflict analysis and crisis mapping. The data is updated weekly. The two data sources I use contain information from January 2018 up until the first week of April 2025. I used these values to create visualizations that shed light on the civil response to the train station roof collapse.
Style Choices
My focus of this project was to highlight protest events and changing protest group dynamic by location and time: before and after the train station roof collapse. Given that I had many categorical variables to use in my visualizations, I chose an overall color scheme of hues that were less vibrant and on the duller side. I tried to choose colors that provided enough contrast to see the differences between protest events and protest groups. For variables that were portraying anything violent, I used different saturations of red. In contrast, for variables that were specifically peaceful, like “peaceful protest”, I chose different saturations of green. All of my visualizations have an offset white hue for the background and a black border around each plot. In my opinion, this made viewing the visualizations easier and cleaner as compared to the default white. For the map of Serbia, highlighting the top 10 cities by protest frequency following the train station roof collapse, I chose to fill the background as the Serbian flag.
Load Necessary Libraries and Data Upload/ Manipulation
library(readr)
library(dplyr)
library(ggplot2)
library(dplyr)
library(sf)
library(rnaturalearth)
library(rnaturalearthdata)
library(lubridate)
library(tidyr)
library(stringr)
library(geodata)
library(patchwork)
library(forcats)
library(scales)
Europe_Central_Asia_2018_2025_Apr04 <- read_csv("C:/Users/katam/OneDrive/Documents/Data Viz/Final Project Serbia/Europe-Central-Asia_2018-2025_Apr04.csv")
#filtering to only contain Serbia as the country
serb <- Europe_Central_Asia_2018_2025_Apr04[grepl("serbia",
Europe_Central_Asia_2018_2025_Apr04$country,
ignore.case = TRUE), ]
#making copy of serb to manipulate
serb1 <- serb
# Convert "4/4/2025" style dates into Date objects
serb1$event_date <- mdy(serb$event_date)
#filter the dataset and create train_roof_collapse variable
serb_clean <- serb1 %>%
filter(!is.na(latitude), !is.na(longitude), !is.na(event_date)) %>%
mutate(
train_roof_collapse = ifelse(event_date >= as.Date("2024-11-01"),
"after collapse", "before collapse")
)
#create new variable to remove city name followed by neighborhood to only contain city name
# new variable called location_clean
serb_exploded <- serb_clean %>%
mutate(location_clean = str_remove(location, " - .*")) # Removes anything after ' - '
# creating clean protest_group variable
serb_exploded <- serb_exploded %>%
separate_rows(assoc_actor_1, sep = ";") %>%
mutate(assoc_actor_1 = str_trim(assoc_actor_1)) %>%
mutate(
protest_group = case_when(
str_detect(assoc_actor_1, regex("Student", ignore_case = TRUE)) ~ "Students",
str_detect(assoc_actor_1, regex("Teacher", ignore_case = TRUE)) ~ "Teachers",
str_detect(assoc_actor_1, regex("Health|Medical", ignore_case = TRUE)) ~ "Healthcare Workers",
str_detect(assoc_actor_1, regex("Labor|Worker", ignore_case = TRUE)) ~ "Labor Group",
str_detect(assoc_actor_1, regex("Farmer", ignore_case = TRUE)) ~ "Farmers",
str_detect(assoc_actor_1, regex("Journalist", ignore_case = TRUE)) ~ "Journalists",
str_detect(assoc_actor_1, regex("Lawyer", ignore_case = TRUE)) ~ "Lawyers",
str_detect(assoc_actor_1, regex("Women", ignore_case = TRUE)) ~ "Women",
str_detect(assoc_actor_1, regex("Ecological|Environment|Green", ignore_case = TRUE)) ~ "Environmental Groups",
str_detect(assoc_actor_1, regex("1 of 5 Million", ignore_case = TRUE)) ~ "1 of 5 Million Movement",
TRUE ~ NA_character_
)
) %>%
filter(!is.na(protest_group))
#cleaning the disorder_type variable - now variable called disorder_category
serb_exploded <- serb_exploded %>%
separate_rows(disorder_type, sep = ";") %>%
mutate(disorder_type = str_trim(disorder_type)) %>%
mutate(
disorder_category = case_when(
str_detect(disorder_type, regex("Demonstration", ignore_case = TRUE)) ~ "Demonstration",
str_detect(disorder_type, regex("Political violence", ignore_case = TRUE)) ~ "Political Violence",
str_detect(disorder_type, regex("Strategic development", ignore_case = TRUE)) ~ "Strategic Development",
TRUE ~ NA_character_
)
) %>%
filter(!is.na(disorder_category))
#reading in second data set
number_of_political_violence_events_by_country_month_year_as_of_11Apr2025 <-
read_csv("C:/Users/katam/OneDrive/Documents/Data Viz/Final Project Serbia/number_of_political_violence_events_by_country-month-year_as-of-11Apr2025.csv")
#renaming the df and filtering to only contain Serbia
pol_viol_cmy <-
number_of_political_violence_events_by_country_month_year_as_of_11Apr2025[grepl("serbia",number_of_political_violence_events_by_country_month_year_as_of_11Apr2025$Country,
ignore.case = TRUE), ]
#converting months to numeric characters
pol_viol_cmy <- pol_viol_cmy %>%
mutate(
Month_num = match(Month, month.name), # Converts "January" to 1, etc.
Date = as.Date(paste(Year, Month_num, "01", sep = "-"))
)
#Define the collapse date
collapse_date <- as.Date("2024-11-01")Data Description
serb_exploded: a cleaned and processed data set that contains all protest events, including location, protest groups and event dates
pol_viol_cmy: a cleaned and processed data set that contains country-level monthly data on political violence events in Serbia
The geographic data I used to create maps was supplemented with spatial shape files from the {geodata} package
Political Violence in Serbia
In order to better understand the broader context of civil unrest in Serbia, it is essential to examine trends in political violence over time. This line graph analyzes the frequency of political violence events in Serbia from February 2018 to April 2025, providing a year-by-year and month-by-month breakdown. By identifying key spikes and periods of stability, we can gain insight into the political climate and shifts in public sentiment, particularly in relation to significant events such as the train station roof collapse in November 2024. This visualization will highlight how political violence has evolved and how it may be linked to broader socio-political developments in the country.
collapse_date <- as.Date("2024-11-01")
ggplot(pol_viol_cmy, aes(x = Date, y = Events)) +
geom_line(color = "darkred", size = 1) +
geom_vline(xintercept = collapse_date, linetype = "dashed", color = "black", size = 1) +
annotate("text", x = collapse_date, y = max(pol_viol_cmy$Events, na.rm = TRUE) * 0.87,
label = "Train Roof Collapse", angle = 90, vjust = -0.7, color = "black", size = 4) +
scale_x_date(breaks = seq(from = min(pol_viol_cmy$Date), to = as.Date("2025-04-01"), by = "3 months"),
date_labels = "%b %Y",
limits = c(min(pol_viol_cmy$Date), as.Date("2025-04-01"))) +
labs(
title = "A Turning Point? Political Violence Trends Around the 2024 Collapse",
x = NULL,
y = "Number of Political Violence Events"
) +
theme_minimal() +
theme(
plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
axis.text.x = element_text(angle = 45, hjust = 1, size = 12),
axis.text.y = element_text(size = 12),
axis.title.y = element_text(size = 12),
plot.title = element_text(size = 16, face = "bold", hjust = 0.5)
)
The line plot visualizes the frequency of political violence events in Serbia over time, with a distinct dashed line marking the date of the train station roof collapse in November 2024. Prior to this event, political violence incidents remained relatively steady, fluctuating modestly over the years. However, immediately following the collapse, there is a marked increase in the number of political violence events, signaling a potential shift in the political climate and public response. This spike suggests that the collapse may have acted as a catalyst for heightened unrest or more aggressive forms of dissent, possibly due to growing public dissatisfaction or perceived government negligence.
While the rise in political violence events provides a focused view of intensifying unrest, it is equally important to understand how different types of disorder evolved during this period. To offer a broader perspective, the following area chart disaggregates all disorder events by category and tracks their occurrence over time. This visualization allows for a more nuanced analysis of whether specific types of disorder—such as demonstrations, strategic developments, or political violence—have shifted in volume or prominence in the months surrounding the collapse
monthly_disorder <- serb_exploded %>%
filter(!is.na(disorder_category)) %>%
mutate(
month = floor_date(event_date, "month"),
train_roof_collapse = ifelse(event_date >= as.Date("2024-11-01"),
"after collapse", "before collapse")
) %>%
group_by(month, disorder_category) %>%
summarise(events = n(), .groups = "drop")
# Normalize events per month
last_date <- as.Date("2025-04-01")
ggplot(monthly_disorder, aes(x = month, y = events, fill = disorder_category)) +
geom_area(alpha = 0.8, position = "stack") +
geom_vline(xintercept = as.Date("2024-11-01"), linetype = "dashed", color = "black", size =1) +
annotate(
"text", x = last_date, y = 0, label = "April 2025", angle = 90,
hjust = 0, vjust = 1, size = 4, fontface = "italic"
) +
scale_x_date(
expand = expansion(mult = c(0, 0.05))
) +
scale_fill_manual(values = c(
"Demonstration" = "#1b9e77",
"Political Violence" = "darkred",
"Strategic Development" = "#00468b"
)) +
labs(
title = "Tracking Disorder in Serbia: Monthly Trends",
subtitle = "Dashed line = Train station roof collapse (Nov 2024)",
x = "", y = "Events per Month", fill = "Disorder Category"
) +
theme_minimal() +
theme(
plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
plot.subtitle = element_text(size = 12),
axis.title.y = element_text(size = 14, margin = margin(r = 10)),
axis.text.x = element_text(size = 12),
axis.text.y = element_text(size = 12),
plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
legend.text = element_text(size = 12),
legend.title = element_text(size = 13),
legend.position = "top"
)
The chart makes clear that demonstrations have consistently been the dominant form of unrest in Serbia throughout the observed period. Notably, there is a significant spike in demonstrations in 2019, largely driven by the 1 of 5 Million movement, a widespread protest campaign against government corruption and media suppression. A similar surge appears again following the roof collapse, reinforcing the trend that public dissent in Serbia has often manifested through mass, largely peaceful, demonstrations rather than violent conflict. This context provides an essential foundation for understanding the nature of subsequent disorder events and the groups involved.
To further unpack the nature of the rise in demonstrations following the train roof collapse, the next area chart disaggregates events by sub-event type, providing insight into how these demonstrations manifested on the ground. While Serbian leadership, including their current president Aleksandar Vučić, characterized the protests as violent, the data tells a different story. This visualization reveals that the vast majority of post-collapse protest activity was comprised of peaceful demonstrations, rather than violent or interventionist actions. By highlighting the composition of disorder events over time, this chart challenges political narratives and emphasizes the largely nonviolent nature of civic mobilization in the aftermath of the collapse.
monthly_sub_event <- serb_exploded %>%
filter(!is.na(sub_event_type)) %>%
mutate(
# Clean sub_event_type by grouping rare ones
sub_event_type_grouped = case_when(
sub_event_type %in% c("Attack", "Mob violence") ~ "Other",
TRUE ~ sub_event_type
),
# Date transformation
month = floor_date(event_date, "month"),
# Flag for collapse timing
train_roof_collapse = ifelse(event_date >= as.Date("2024-11-01"),
"after collapse", "before collapse")
) %>%
group_by(month, sub_event_type_grouped) %>%
summarise(events = n(), .groups = "drop")
last_date <- as.Date("2025-04-01")
ggplot(monthly_sub_event, aes(x = month, y = events, fill = sub_event_type_grouped)) +
geom_area(alpha = 0.8, position = "stack") +
geom_vline(xintercept = as.Date("2024-11-01"), linetype = "dashed", color = "black", size = 1) +
annotate("text",
x = last_date,
y = 0,
label = "April 2025",
angle = 90,
hjust = 0,
vjust = 1,
size = 4,
fontface = "italic") +
scale_x_date(
breaks = seq(from = as.Date("2018-01-01"), to = last_date, by = "2 year"),
labels = scales::date_format("%Y"),
expand = expansion(mult = c(0, 0.05))
) +
scale_fill_manual(values = c(
"Excessive force against protesters" = "#00468b",
"Other" = "#8b4600",
"Peaceful protest" = "#008b46",
"Protest with intervention" = "#8b008b",
"Violent demonstration" = "darkred"
)) +
labs(
title = "Sub-Events of Unrest in Serbia: A Stacked Timeline of Tension",
subtitle = "Dashed line = Train station roof collapse (Nov 2024)",
x = "", y = "Events per Month", fill = "Sub-Event Type"
) +
theme_minimal() +
theme(
plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
plot.subtitle = element_text(size = 12),
axis.title.y = element_text(size = 14, margin = margin(r = 10)),
axis.text.x = element_text(angle = 45, hjust = 0.5, vjust = 1, size = 12, margin = margin(t = 8)),
axis.text.y = element_text(size = 12),
plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
legend.position = "top",
legend.text = element_text(size = 12),
legend.title = element_text(size = 13),
)
While the area chart clearly demonstrates that the majority of the sub-event types in the post-roof collapse period were peaceful protests, it is crucial to understand who was participating in these demonstrations and how different groups engaged in the protests. To explore this further, the next bar graph examines protest group demographics across all events in the data set, not just those occurring after the train station roof collapse. This chart provides a broader perspective on the composition of protest groups throughout the entire period, offering insights into the diversity and scale of participation across various social and professional sectors in Serbia. By comparing this data to the previous analysis, we can better understand the dynamics of protest participation and the shifts in involvement across different groups over time.
protest_colors <- c(
"Students" = "#1b9e77",
"1 of 5 Million Movement" = "#d95f02",
"Labor Group" = "#8b008b",
"Teachers" = "#1f78b4",
"Environmental Groups" = "#66a61e",
"Farmers" = "#e6ab02",
"Women" = "#e7298a",
"Healthcare Workers" = "#8b0000",
"Journalists" = "#a6761d",
"Lawyers" = "#b2df8a"
)
serb_exploded %>%
filter(!is.na(protest_group)) %>%
count(protest_group, sort = TRUE) %>%
slice_max(n = 10, order_by = n) %>%
ggplot(aes(x = fct_reorder(protest_group, n), y = n, fill = protest_group)) +
geom_col(show.legend = FALSE) +
coord_flip() +
scale_fill_manual(values = protest_colors) +
labs(title = "The History of Faces Behind the Protests: Serbia’s Most Prominent Groups",
x = "",
y = "Protest Appearances by Group") +
theme_minimal() +
theme (
plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
axis.title.x = element_text(size = 14),
axis.text.x = element_text(size = 12, margin = margin(t = 8)),
axis.text.y = element_text(size = 12),
plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
)
The bar chart displaying the top 10 protest groups in Serbia by number of events highlights the central role of student-led activism in the country’s recent protest landscape. Students emerge as the most active group, followed closely by the 1 of 5 Million movement, labor groups, and teachers. The 1 of 5 Million movement, which gained momentum in late 2018 and peaked in 2019, was a broad-based citizen initiative sparked by public outrage over political violence and deteriorating democratic institutions. The anti-government movement began after an assault on opposition politician Borko Stefanović in November 2018 and was fueled by widespread concerns over political violence, media censorship, corruption, and the authoritarian tendencies of President Aleksandar Vučić’s government. The movement’s name originated from Vučić’s dismissive remark that he would not yield to even five million protesters, prompting the slogan “One in Five Million” to symbolize the collective power of the people.
While this chart offers a broad view of protest group activity across Serbia over several years, it doesn’t reflect how these dynamics may have shifted in response to a major national event — the train station roof collapse in November 2024 that resulted in the loss of 15 innocent lives. To better understand how the protest landscape evolved, we now narrow our focus to events that occurred after the collapse.
The following visualization highlights the most active protest groups in the top 10 cities that experienced the highest number of protest events following the incident. By examining protest group activity in these key urban centers, we can observe whether new groups emerged, existing ones intensified their efforts, or if certain types of protests became more geographically concentrated in the aftermath of the collapse.
serb_exploded <- serb_exploded %>%
filter(!is.na(latitude), !is.na(longitude), !is.na(event_date)) %>%
mutate(after_collapse = event_date >= as.Date("2024-11-01"))
top_10_cities <- serb_exploded %>%
filter(after_collapse == TRUE, !is.na(protest_group)) %>%
count(location_clean, sort = TRUE) %>%
slice_max(n = 10, order_by = n) %>%
pull(location_clean)
protest_color <- c(
"Students" = "#1b9e77",
"Labor Group" = "#8b008b",
"Teachers" = "#1f78b4",
"Farmers" = "#e6ab02",
"Healthcare Workers" = "#8b0000",
"Lawyers" = "#b2df8a"
)
# Filter and plot
serb_exploded %>%
filter(after_collapse == TRUE, !is.na(protest_group), location_clean %in% top_10_cities) %>%
count(location_clean, protest_group, sort = TRUE) %>%
group_by(location_clean) %>%
slice_max(n = 4, order_by = n) %>%
ggplot(aes(x = fct_reorder(location_clean, n), y = n, fill = protest_group)) +
geom_col(position = "stack") +
coord_flip() +
scale_fill_manual(values = protest_color) +
labs(title = "Post-Collapse Protest Dynamics: Top Groups in Serbia’s Leading Cities",
x = "", y = "Protest Appearances by Group", fill = "Protest Group") +
theme_minimal() +
theme(
plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
axis.title.x = element_text(size = 14),
axis.text.x = element_text(size = 12, margin = margin(t = 8)),
axis.text.y = element_text(size = 12),
plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
legend.text = element_text(size = 12),
legend.title = element_text(size = 13),
)
This visualization presents the distribution of protest group involvement across the ten Serbian cities with the highest number of protest events after the Novi Sad train station roof collapse. The chart reveals distinct patterns of mobilization, with student and teacher groups emerging as prominent actors in several urban centers. The variation in group presence across cities suggests that while the roof collapse served as a unifying catalyst, localized social and institutional dynamics influenced the nature and intensity of protest activity. Protest dynamics varied across locations — with students, teachers, and labor groups being especially active in cities like Belgrade, Novi Sad, and Nis. This suggests that while the collapse acted as a national trigger, local context shaped who mobilized and how strongly in different areas.
While the previous chart highlights the most active protest groups in the cities most affected after the train roof collapse, the following map provides a geographic view of where these protest events were most concentrated across Serbia. By visualizing the top 10 cities on the map, we can better understand the spatial dimension of post-collapse protest activity. Notably, Belgrade—the capital—recorded the highest number of protest events, followed by Novi Sad, where the roof collapse occurred. The distribution of these cities also shows that protests were not confined to one region but were spread throughout the country.
# Download Serbia administrative boundaries (level 1) using terra
serbia_spat <- geodata::gadm("SRB", level = 1, path = tempdir())
# Convert to sf object for plotting
serbia_sf <- sf::st_as_sf(serbia_spat)
# Assign a CRS
if (is.null(st_crs(serbia_sf))) {
st_crs(serbia_sf) <- 4326 # WGS 84
}
serb_exploded_top_10 <- serb_exploded %>%
filter(!is.na(latitude), !is.na(longitude), !is.na(event_date)) %>%
mutate(after_collapse = event_date >= as.Date("2024-11-01"))
top_10_cities <- serb_exploded_top_10 %>%
filter(after_collapse == TRUE, !is.na(protest_group)) %>%
count(location_clean, sort = TRUE) %>%
slice_max(n = 10, order_by = n) %>%
pull(location_clean)
cities <- tibble::tibble(
name = c("Belgrade", "Novi Sad", "Niš", "Kragujevac", "Subotica",
"Cacak", "Zrenjanin", "Kraljevo", "Valjevo", "Leskovac"),
lon = c(20.46, 19.84, 21.90, 20.92, 19.66,
20.35, 20.39, 20.69, 19.88, 21.95),
lat = c(44.82, 45.26, 43.32, 44.02, 46.10,
43.89, 45.38, 43.72, 44.27, 42.99),
color = c("white", "white", "black", "white","white",
"white", "white", "black", "white", "black"),
)
# Add a new column for shape_type to categorize Belgrade and Novi Sad
cities <- cities %>%
mutate(
shape_type = case_when(
name == "Belgrade" ~ "Capital",
name == "Novi Sad" ~ " Train Roof Collapse",
TRUE ~ "Regular"
)
)
# Get Serbia's bounding box
bbox <- st_bbox(serbia_sf)
# Calculate Y-ranges for 3 equal horizontal bands
y_top <- bbox["ymax"]
y_1third <- bbox["ymax"] - (1/3) * (bbox["ymax"] - bbox["ymin"])
y_2third <- bbox["ymax"] - (2/3) * (bbox["ymax"] - bbox["ymin"])
y_bottom <- bbox["ymin"]
# Create three rectangular bands as sf polygons
red_band <- st_as_sf(st_sfc(st_polygon(list(rbind(
c(bbox["xmin"], y_top),
c(bbox["xmax"], y_top),
c(bbox["xmax"], y_1third),
c(bbox["xmin"], y_1third),
c(bbox["xmin"], y_top)
))), crs = st_crs(serbia_sf)))
blue_band <- st_as_sf(st_sfc(st_polygon(list(rbind(
c(bbox["xmin"], y_1third),
c(bbox["xmax"], y_1third),
c(bbox["xmax"], y_2third),
c(bbox["xmin"], y_2third),
c(bbox["xmin"], y_1third)
))), crs = st_crs(serbia_sf)))
white_band <- st_as_sf(st_sfc(st_polygon(list(rbind(
c(bbox["xmin"], y_2third),
c(bbox["xmax"], y_2third),
c(bbox["xmax"], y_bottom),
c(bbox["xmin"], y_bottom),
c(bbox["xmin"], y_2third)
))), crs = st_crs(serbia_sf)))
#clip each band to serbias shape
red_band_clipped <- st_intersection(serbia_sf, red_band)
blue_band_clipped <- st_intersection(serbia_sf, blue_band)
white_band_clipped <- st_intersection(serbia_sf, white_band)
# Plot
ggplot() +
geom_sf(data = white_band_clipped, fill = "white", color = NA) +
geom_sf(data = blue_band_clipped, fill = "blue", color = NA) +
geom_sf(data = red_band_clipped, fill = "red", color = NA) +
geom_sf(data = serbia_sf, fill = NA, color = "black", size = 0.5) +
geom_text(
data = cities %>%
mutate(
lon = ifelse(name == "Belgrade", lon - 0.05, lon),
lat = ifelse(name == "Belgrade", lat - 0.1, lat),
lon = ifelse(name == "Novi Sad", lon - 0.1, lon),
lat = ifelse(name == "Novi Sad", lat + 0.0, lat),
lon = ifelse(name == "Niš", lon + 0.1, lon),
lat = ifelse(name == "Niš", lat + 0, lat),
lon = ifelse(name == "Kragujevac", lon - 0.08, lon),
lat = ifelse(name == "Kragujevac", lat + 0.15, lat),
lon = ifelse(name == "Subotica", lon - 0.1, lon),
lat = ifelse(name == "Subotica", lat - 0.1, lat),
lon = ifelse(name == "Cacak", lon + 0, lon),
lat = ifelse(name == "Cacak", lat + 0, lat),
lon = ifelse(name == "Zrenjanin", lon - 0.0, lon),
lat = ifelse(name == "Zrenjanin", lat + 0.15, lat),
lon = ifelse(name == "Kraljevo", lon - 0.1, lon),
lat = ifelse(name == "Kraljevo", lat - 0.2, lat),
lon = ifelse(name == "Valjevo", lon + 0.1, lon),
lat = ifelse(name == "Valjevo", lat + 0, lat),
lon = ifelse(name == "Leskovac", lon - 0.1, lon),
lat = ifelse(name == "Leskovac", lat + 0.0, lat)
),
aes(x = lon, y = lat, label = name, color = color),
size = 3.1, fontface = "bold", hjust = 0.5, vjust = 0.5, show.legend = FALSE
) +
scale_color_identity() +
labs(title = "Post-Collapse Unrest: Top 10 Serbian Cities by Protest Frequency",
subtitle = "Novi Sad: Train Station Roof Collapse Site | Belgrade: Capital of Serbia") +
theme_minimal() +
theme(
axis.title = element_blank(),
axis.text = element_blank(),
axis.ticks = element_blank(),
panel.grid = element_blank(),
plot.title = element_text(hjust = 0.5, face = "bold", size = 16),
plot.subtitle = element_text(hjust = 0.5, size = 14),
plot.caption = element_text(hjust = 0.5, size = 10)
)
While the geographic map reveals the spatial distribution of protest activity following the train roof collapse—highlighting Belgrade as the epicenter and Novi Sad as the site of the incident—it is equally important to examine how patterns of participation evolved over time. The following visualization shifts from a geographic lens to a temporal one, comparing protest group involvement before and after the collapse. This temporal comparison offers insight into which groups intensified their mobilization in the aftermath of the crisis, shedding light on broader shifts in civic engagement and collective action across Serbia.
serb_exploded %>%
group_by(after_collapse, protest_group) %>%
summarise(count = n(), .groups = "drop") %>%
mutate(period = ifelse(after_collapse, "After Collapse", "Before Collapse")) %>%
filter(protest_group != "1 of 5 Million Movement") %>% # Exclude this specific group
ggplot(aes(x = reorder(protest_group, count), y = count, fill = period)) +
geom_col(position = "dodge") +
coord_flip() +
labs(
title = "Shift in Protest Group Activity Before and After the Novi Sad Roof Collapse",
subtitle = "Students and teachers became the most active protest groups post-collapse",
x = "",
y = "Frequency of Protest Group Involvement",
fill = "Time Period"
) +
scale_fill_manual(
values = c("Before Collapse" = "#1f78b4", "After Collapse" = "#1b9e77")
) +
theme_minimal() +
theme( plot.background = element_rect(fill = "#f4f2e7", color = NA),
panel.background = element_rect(fill = "#f4f2e7", color = NA),
panel.border = element_rect(color = "black", fill = NA, size = 1),
panel.grid.major.x = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.x = element_blank(),
panel.grid.major.y = element_line(color = "lightgray", size = 0.5),
panel.grid.minor.y = element_line(color = "gray80", size = 0.25),
axis.title.x = element_text(size = 12),
axis.text.x = element_text(size = 12, margin = margin(t = 8)),
axis.text.y = element_text(size = 12),
plot.title = element_text(hjust = 0.5, size = 16, face = "bold"),
plot.subtitle = element_text(size = 14),
legend.text = element_text(size = 12),
legend.title = element_text(size = 13)
)
The comparison of protest group activity before and after the train station roof collapse reveals significant shifts in patterns of mobilization across Serbia. Notably, students and teachers emerged as the most active groups following the incident, with a marked increase in protest events compared to the period prior. This surge suggests that the collapse acted as a catalyst for heightened engagement, particularly among younger populations and educational professionals. The increased visibility of infrastructural neglect may have galvanized students to organize, while teachers likely joined in solidarity and to express broader concerns about the safety and conditions of public institutions. Healthcare workers also demonstrated a rise in mobilization after the collapse, potentially driven by heightened awareness of public safety failures and the strain such emergencies place on the medical system. These shifts indicate that the roof collapse was not merely an isolated event, but one that triggered deeper societal concerns and prompted targeted collective action among key civil sectors.
Key Insights and Implications
This project provided a comprehensive analysis of protest activity in Serbia from 2018 through early 2025, revealing important patterns in civic mobilization, protest group dynamics, and geographic concentrations of unrest. The data demonstrated that while various forms of disorder were present throughout the period, demonstrations—particularly peaceful protests—were by far the most common. A major turning point occurred in November 2024, following the tragic collapse of a train station roof in Novi Sad. In the aftermath, there was a sharp rise in protest events, especially among students, teachers, and healthcare workers, reflecting deep public dissatisfaction with government accountability and infrastructure neglect. Geographically, protest activity was not limited to urban centers, but spread across the country, with Belgrade and Novi Sad leading in event counts. The emergence and sustained presence of movements such as the “1 of 5 Million” further underscores the population’s ongoing struggle for transparency, reform, and democratic rights. Ultimately, this analysis highlights that protest in Serbia is not episodic but persistent—driven by an engaged citizenry that continues to demand accountability from those in power.
References
ACLED Data Export Tool – Official political violence & protest dataset used for analysis: https://acleddata.com/data/#download
Protests Leave Serbia’s Politics Finely Balanced – GMFUS summary of evolving civic unrest in Serbia: https://gmfus.org/news/protests-leave-serbias-politics-finely-balanced
Tensions Mount Ahead of a Pro‑government Rally in Serbia to Counter Massive Student Protests – ABC News reporting on April 2025 events: https://abcnews.go.com/International/wireStory/tensions-mount-ahead-pro-government-rally-serbia-counter-120741785
Serbia Train Station Roof Collapse, Arrests – AP News on arrests linked to the Novi Sad disaster: https://apnews.com/article/serbia-train-station-roof-collapse-arrests-75ce32518db2b255229655bce39cb4f8
Serbia Students Strasbourg EU Protest – AP News covering the student cycling protest to Strasbourg: https://apnews.com/article/serbia-students-strasbourg-eu-protest-398094de8251ffb3579231e805fc28a7
Serbia Protest: Students, Roof Collapse, Prosecutor, Vučić – AP News on student demonstrations and state response: https://apnews.com/article/serbia-protest-students-roof-collapse-prosecutor-vucic-b2aa7a208e53006dab4e086588b7355d