Popular culture has evolved beyond mere entertainment; it has become a significant medium for self-expression and social connection. In Indonesia, a surprising symbol of rebellion, adventure, and resistance against oppression has emerged among protesters. But how strong is the symbolic significance of this emblem? In this post, we will explore how data from the entertainment industry can illuminate the role of One Piece in Indonesia’s protests. By employing causal inference methods and data from Parrot Analytics, we aim to determine whether the popularity of this fictional world reflects real-life actions and what this reveals about the relationship between pop culture and social change.
Recently, Indonesia has seen a wave of protests fueled by economic and political frustrations, along with rising uncertainty about the future. Since late July, truck drivers in the country have started waving the Jolly Roger flag as a way to protest new transport regulations. What began as a viral trend on social media quickly turned into a collective movement, with demonstrators nationwide raising the flag to symbolize their discontent during Indonesia’s 80th National Day.
The Jolly Roger flag of the Straw Hat Pirates from One Piece, a hugely popular Japanese manga and anime series with over 1,100 episodes aired in more than 80 countries, symbolizes the adventurous and rebellious spirit of Monkey D. Luffy and his crew, who stand against a globally oppressive and corrupt government. Interestingly, protesters have adopted this flag as a symbol of collective identity; it represents not only the desire to end corruption but also embodies the Mugiwaras’ promise of freedom and justice. Émile Durkheim believed that symbols, along with shared beliefs and values, create a collective consciousness that holds society together. This illustrates that popular culture does more than entertain us; it shapes our daily lives and provides us with meaningful symbols that can inspire action. Such symbols transcend the fictional realm and become part of real-world struggles.
Stories like One Piece play a significant role in shaping the social landscape by helping people envision and embody their hopes for change. The symbols within these narratives serve as rallying cries, representing a form of cultural glocalization that blends global media with local struggles. These stories provide social movements with both a narrative and imagery that they can draw strength from, making the fight for a better Indonesia not just a political issue but also a collective experience of resistance and resilience rooted in global popular culture.
The first step is to start our R session and load the tidyverse toolkit for our analysis. I recommend using extensions for ggplot2, such as ggalt and ggtext, as they offer greater control over the customization of our visuals. Finally, I utilize stargazer to create well-formatted regression tables for presenting my results.
library(tidyverse) # For data wrangling
library(ggplot2) # For nice data visualization
library(ggalt) # For nice smooth splines
library(ggtext) # For modifying titles and labels text
library(stargazer) # For nice regression tablesThe next step is to plot the average weekly per-capita demand for One Piece over time in Indonesia, East Asia, and Southeast Asia (excluding Indonesia), while also comparing it to the overall regional average. I included East Asia in the analysis for two main reasons. First, to account for variations in the popularity of the show within the Asian market, such as new episode releases or events like One Piece Day. Second, to control for potential spillover effects that the protests in Indonesia may have generated in other Southeast Asian markets, such as Nepal. The graph below illustrates that after the Jolly Roger flag protest began, the demand for the show in Indonesia peaked at 1.88 times higher than the average demand in the region.
### Labels and data points for visuals ###
# Create a tibble with the coordinates of market labels
labels = data %>%
# Group by market
group_by(market) %>%
# Keep the last observation
filter(date == max(date)) %>%
# This trick is just to have a nice label
mutate(
market = ifelse(
# For the 'Southeast Asia' label
test = market == 'Southeast Asia',
# Add a nice line break and centered text
yes = 'Southeast east\n Asia',
# The rest stay the same
no = market
)
)
# Get min date limit
min_date = data %>%
filter(date == min(date)) %>%
pull(date) %>%
unique()
# Get max date limit
max_date = data %>%
filter(date == max(date)) %>%
pull(date) %>%
unique()
## Save relevant dates and demand ##
# Cutoff date ('2025-07-20')
cutoff = ymd('2025-07-20')
# Truck protest date
truck_date = data %>%
filter(market == 'Indonesia', date == ymd('2025-07-20')) %>%
pull(date)
# Truck protest demand
truck_demand = data %>%
filter(market == 'Indonesia', date == ymd('2025-07-20')) %>%
pull(region_multiplier)
# Indonesia's National day date
national_day_date = data %>%
filter(market == 'Indonesia', date == ymd('2025-08-17')) %>%
pull(date)
# Indonesia's National day demand
national_day_demand = data %>%
filter(market == 'Indonesia', date == ymd('2025-08-17')) %>%
pull(region_multiplier)
# Peak demand's date
peak_date = data %>%
filter(market == 'Indonesia') %>%
filter(region_multiplier == max(region_multiplier)) %>%
pull(date)
# Peak demand
peak_demand = data%>%
filter(market == 'Indonesia') %>%
filter(region_multiplier == max(region_multiplier)) %>%
pull(region_multiplier)
### Overal Time Series ###
# Create a ggplot canvas
ggplot(
# Data to plot
data = data,
# Data-related aesthetics
mapping = aes(x = date, y = region_multiplier, col = market)
) +
# ggalt geom_xpline give us soft lines using splines
geom_xspline(spline_shape = -0.5) +
# Add vertical line
geom_vline(
# This is the week when the first flag was shown (at leas in the news)
xintercept = c(cutoff),
# Modify the aesthetics
linetype = 'dashed', col = 'gray60'
) +
# Add points
geom_point() +
# Add text
geom_text(
# Use our labeled data
data = labels, aes(label = market),
# Modify justification and lineheight
vjust = 0.25, hjust = -0.15, lineheight = 0.9
) +
# Add segemnt from point to label
geom_segment(
# X coords
x = truck_date, xend = truck_date,
# Y coords
y = truck_demand, yend = truck_demand + 0.2,
# Set aesthetics
col = 'red2'
) +
# Annotate the relevant text
annotate(
# Use GeomRichtext as geom
geom = ggtext::GeomRichtext,
# Set coordinates
x = truck_date, y = truck_demand + 0.22,
# Set label
label = sprintf(
# Html and linked to data format
fmt = "<span style='color:black'>During **Truck drivers protests**<br>Demand was **%s times** higher<br>than in the overall region</span>",
round(x = truck_demand, digits = 2)
),
# Set aesthetics
size = 2.5, hjust = 1, vjust = 0.5, color = 'red2',
) +
# Add segemnt from point to label
geom_segment(
# X coords
x = national_day_date, xend = national_day_date,
# Y coords
y = national_day_demand, yend = national_day_demand + 0.2,
# Set aesthetics
col = 'red2'
) +
# Annotate the relevant text
annotate(
# Use GeomRichtext as geom
geom = ggtext::GeomRichtext,
# Set coordinates
x = national_day_date, y = national_day_demand + 0.22,
# Set label
label = sprintf(
# Html and linked to data format
fmt = "<span style='color:black'>During **Indonesia's National Day**<br>Demand was **%s times** higher<br>than in the overall region</span>",
round(x = national_day_demand,digits = 2)
),
# Set aesthetics
size = 2.5, hjust = 0, vjust = 0.5, color = 'red2',
) +
# Add segemnt from point to label
geom_segment(
# X coords
x = peak_date, xend = peak_date,
# Y coords
y = peak_demand, yend = peak_demand + 0.2,
# Set color
col = 'red2'
) +
# Annotate the relevant text
annotate(
# Use GeomRichtext as geom
geom = ggtext::GeomRichtext,
# Set coordinates
x = peak_date, y = peak_demand + 0.2,
# Html and linked to data format
label = sprintf(
fmt = "<span style='color:black'>**Peak of Demand** was **%s times**<br>higher than in the overall region</span>",
round(x = peak_demand, digits = 2)
),
# Set aesthetics
size = 2.5, hjust = 0, vjust = 0.5, color = 'red2',
) +
# Add labels
labs(
title = 'Weekly Demand Performance for One Piece in East and Southeast Asia',
subtitle = 'Overall region demand multiplier (x Times average demand for One Piece)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo'
) +
# Modify color scale (I repeated blue to match our modified label of Southeast East Asia)
scale_color_manual(values = c('green4','red2','blue2','blue')) +
# Modify x axis limits
scale_x_date(limits = c(min_date, max_date + 10), date_labels = "%b-%Y") +
# Modify y axis limits and label units
scale_y_continuous(limits = c(0.5,NA), labels = scales::unit_format(suffix = 'x')) +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
# Remove legend (recommended to avoid writing to many `show.legend = F`)
legend.position = 'none'
)
The central idea of DiD regression is to estimate the causal effect of an intervention, such as the Jolly Roger flag protest. This method involves comparing the average changes over time between a group that experiences the intervention (Indonesia) and a control group that does not (the other Southeast and East Asian markets). The plot below illustrates, with a dashed line, the average demand multiplier for each market before and after the initiation of the Jolly Roger flag protest in Indonesia.
### Average Time Series ###
# Create a variable to differentiate between the before and after
data = data %>%
mutate(intervention = ifelse(test = date < cutoff, yes = 0, no = 1)) %>%
# Get the market average for the before and after
with_groups(
.groups = c(market, intervention),
mutate,
average_multiplier = mean(region_multiplier)
)
# Create a ggplot canvas
ggplot(
# Data to plot
data = data,
# Data-related aesthetics
mapping = aes(x = date, y = region_multiplier, col = market)
) +
# Add vertical line
geom_vline(
# This is the week when the first flag was shown (at least in the news)
xintercept = c(cutoff),
# Modify the aesthetics
linetype = 'dashed', col = 'gray60'
) +
# Add point
geom_point() +
# Add a dashed line with the average multiplier for the Before
stat_smooth(
data = filter(data, date < cutoff),
mapping = aes(x = date, y = average_multiplier, col = market),
method = 'lm', linetype = 'dashed', linewidth = 0.75
) +
# Add a dashed line with the average multiplier for the After
stat_smooth(
data = filter(data, date >= cutoff),
mapping = aes(x = date, y = average_multiplier, col = market),
method = 'lm', linetype = 'dashed', linewidth = 0.75
) +
# Add the Before text
annotate(
geom = 'text',label = 'Before',
# Set coordinates
x = c(ymd('2025-06-22')), y = 1.95,
# Set aesthetics
size = 3.5, color = 'gray30',
) +
# Add the After text
annotate(
geom = 'text', label = 'After',
# Set coordinates
x = c(ymd('2025-08-21')), y = 1.95,
# Set aesthetics
size = 3.5, color = 'gray30',
) +
# Facet by market
facet_wrap(~market) +
# Add labels
labs(
title = 'Effect of the Jolly Roger flag protest in Weekly Demand Performance\nfor One Piece in Southeast and East Asia',
subtitle = 'Average demand multiplier (x Times average demand for One Piece in the region)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo'
) +
# Modify colors
scale_color_manual(values = c('green4','red2','blue2')) +
# Modify Y axis limits and units
scale_y_continuous(limits = c(0.5,2), labels = scales::unit_format(suffix = 'x')) +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
# Remove legend (recommended to avoid writting to many `show.legend = F`)
legend.position = 'none'
)
We can observe an increase in the average multiplier value for Indonesia after the protest began. Our objective is to determine whether the differences in demand before and after the protest are statistically significant when compared to the behavior of the control markets.
### Average before and after effects ###
# Create a ggplot canvas
ggplot(
data = data,
mapping = aes(x = intervention, y = region_multiplier, color = market)
) +
# Add points with 1.96 se
stat_summary(geom = 'pointrange', fun.data = 'mean_se', fun.args = list(mult = 1.96)) +
# Add labels
labs(
title = 'Effect of the Jolly Roger flag protest in Weekly Demand Performance\nfor One Piece in East and Southeast Asia (June 1st to September 13th)',
subtitle = 'Average demand multiplier (x Times average demand for One Piece in the region)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo'
) +
scale_x_continuous(breaks = c(0,1), labels = c('Before', 'After'), limits = c(-0.25,1.25)) +
facet_wrap(~market) +
scale_color_manual(values = c('green4','red2','blue2')) +
# Modify Y axis limits and units
scale_y_continuous(limits = c(0.5,2), labels = scales::unit_format(suffix = 'x')) +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
# Remove legend (recommended to avoid writting to many `show.legend = F`)
legend.position = 'none'
)
To gain a better understanding of the DiD procedure, we can create a table that displays the average values for both Indonesia and the control markets before and after the intervention. This table will show the average values prior to the Jolly Roger Flag’s protest, as well as the averages after the protest.
Next, we will calculate the changes over time for each group by subtracting the pre-protest average from the post-protest average. The key row of the table will highlight the difference between these two changes; this is known as the “difference-in-differences” estimate. This estimate represents the intended impact of the intervention. Essentially, this method isolates the portion of the change in the treated group (Indonesia) that exceeds or falls short of the change observed in the control group (the other Southeast and East Asian markets), providing a clear measure of the intervention’s effect.
| Jolly Roger flag protest | Before | After | Average Difference |
|---|---|---|---|
| Indonesia | A | B | B - A |
| Control Market | C | D | D - C |
| Average Difference | A - C | B - D | (B - A) - (D - C) |
We can easily fill in the first row for our treated market using dplyr syntax. After the intervention, we observed that the average weekly demand increased by 0.15 units in the multiplier.
data = data %>%
mutate(
# Create a treatment variable
treatment = ifelse(market == 'Indonesia', 1, 0),
# Create a DiD variable
DiD = treatment * intervention
)
# A cell
before_treatment = data %>%
# Treated market before intervention
filter(treatment == 1, intervention == 0) %>%
# Average multiplier
summarise(A = mean(region_multiplier)) %>%
# Pull the value
pull(A)
# B cell
after_treatment = data %>%
# Treated market after intervention
filter(treatment == 1, intervention == 1)%>%
# Average multiplier
summarise(B = mean(region_multiplier)) %>%
# Pul the value
pull(B)| Jolly Roger flag protest | Before | After | Average Difference |
|---|---|---|---|
| Indonesia | 1.554 | 1.703 | 0.149 |
| Control Market | C | D | D - C |
| Average Difference | A - C | B - D | (B - A) - (D - C) |
Let’s calculate the average for the Southeast Asian market, which will act as our control group. We can follow the same steps as in our previous analysis, making just a few modifications to our code to complete the rest of the table. After our intervention, we observed that the change in the demand multiplier for the control group was -0.01. By subtracting the average difference of the control group from the average difference of the treated group, we can determine our DiD estimation. Our calculations show that the average impact of the Jolly Roger Flag protest on the demand performance of One Piece resulted in an increase of approximately 0.16 units in the multiplier compared to the average market in Southeast Asia.
# Filter our interest market
data_se = data %>%
filter(market %in% c('Indonesia','Southeast Asia'))
# C cell
before_control = data_se %>%
# Control market before intervention
filter(treatment == 0, intervention == 0) %>%
# Average multiplier
summarise(C = mean(region_multiplier)) %>%
# Pull the value
pull(C)
# D cell
after_control = data_se %>%
# Control market after intervention
filter(treatment == 0, intervention == 1) %>%
# Average multiplier
summarise(D = mean(region_multiplier)) %>%
# Pull the value
pull(D)
# Diff in diff estimator
diff_in_diff = (after_treatment - before_treatment) - (after_control - before_control)| Jolly Roger flag protest | Before | After | Average Difference |
|---|---|---|---|
| Indonesia | 1.554 | 1.703 | 0.149 |
| Control Market | 0.756 | 0.743 | -0.013 |
| Average Difference | 0.798 | 0.96 | 0.162 |
If we plot our difference-in-differences calculations alongside our previously calculated average, we can confirm that they align with the average observed values we plotted earlier; in fact, they are the same.
### diff in diff graph ###
did_df = tibble(
# Market name
market = c('Indonesia', 'Indonesia','Southeast Asia','Southeast Asia'),
# Intervention column
intervention = c(0 , 1, 0, 1),
# Treatment column
treatment = c(1, 1, 0, 0),
# Calculated valures
values = c(before_treatment, after_treatment, before_control, after_control)
)
# Create a ggplot canvas
ggplot(data = data_se, mapping = aes(x = intervention, y = region_multiplier)) +
# Add points with 1.96 se
stat_summary(
# Add colour
mapping = aes(color = 'Observed'),
# Additional aesthetics
geom = 'pointrange', fun.data = 'mean_se', fun.args = list(mult = 1.96)
) +
# Add our calculated averages
geom_point(
data = did_df,
mapping = aes(x = intervention, y = values, col = 'Calculated'),
# Modify aesthetics
shape = 13, size = 4
) +
# Add labels
labs(
title = bquote('Effect of the Jolly Roger flag protest in Weekly Demand Performance\nfor One Piece in Southeast Asia (June 1st to September 13th)'),
subtitle = 'Average demand multiplier (x Times average demand for One Piece in the region)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo',
color = "Mean" # Legend title
) +
# Split by market
facet_wrap(~market) +
# Modify scales
scale_x_continuous(breaks = c(0,1), labels = c('Before', 'After'), limits = c(-0.25,1.25)) +
# Modify Y axis limits and units
scale_y_continuous(limits = c(0.5,2), labels = scales::unit_format(suffix = 'x')) +
# Scale color
scale_color_manual(values = c('gray50','red2')) +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
legend.position = 'bottom',
legend.title = element_blank()
)
A key requirement for the validity of the difference-in-differences method is the parallel trend assumption. This assumption posits that, in the absence of the intervention, the treated group and the control group would have exhibited similar trends or patterns in their outcomes over time. In simpler terms, had there been no changes related to the protest, the demand performance of One Piece in both markets would likely have followed a similar trajectory.
This assumption allows us to attribute any differences in trends observed after the intervention to the effect of the Jolly Roger protest, rather than to other factors or inherent differences between the groups. Essentially, it ensures that the only systematic difference noted after the intervention is due to the treatment itself. This validation supports the causal interpretation of the difference-in-differences estimate. The following plot provides a graphic representation of our parallel trend assumption, illustrated with a blue dashed line. The potential impact of the Jolly Roger flag protest in Indonesia is indicated by a vertical red dashed line.
# Min x for segment annotation
min_x = data_se %>%
filter(date >= cutoff) %>%
pull(date) %>% min()
# Max x for segment annotation
max_x = data_se %>%
filter(date >= cutoff) %>%
pull(date) %>% max()
# Create a ggplot canvas
ggplot(
# Data to plot
data = data_se,
# Data-related aesthetics
mapping = aes(x = date, y = region_multiplier, col = market)
) +
# Add vertical line
geom_vline(
# This is the week when the first flag was shown (at least in the news)
xintercept = c(cutoff) - 3,
# Modify the aesthetics
linetype = 'dashed', col = 'gray60'
) +
# Add point
geom_point(size = 1, alpha = 0.25) +
# Add a dashed line with the average multiplier for the Before
stat_smooth(
data = filter(data_se, date < cutoff),
mapping = aes(x = date, y = average_multiplier, col = market),
method = 'lm', linewidth = 0.25
) +
# Add a dashed line with the average multiplier for the After
stat_smooth(
data = filter(data_se, date >= cutoff),
mapping = aes(x = date, y = average_multiplier, col = market),
method = 'lm', linewidth = 0.25
) +# Draw the parallel trend
annotate(
geom = 'segment',
# Set coordinates
x = min_x, xend = max_x, y = before_treatment, yend = after_treatment - diff_in_diff,
# Set aesthetics
linetype = 'dashed', col = 'blue2', linewidth = 0.75
) +
# Add vertical gray projection lines
annotate(
geom = 'segment',
# Set coordinates
x = c(min_x, max_x), xend = c(min_x, max_x), y = before_treatment, yend = before_control,
# Set aesthetics
linetype = 'dotted', col = 'gray50'
) +
# Add the potential effect
annotate(
geom = 'segment',
# Set coordinates
x = c(min_x, max_x), xend = c(min_x, max_x), y = before_treatment, yend = after_treatment,
# Set aesthetics
linetype = 'dashed', col = 'red2', linewidth = 0.75
) +
# Add text
annotate(
geom = 'text', label = 'Indonesia before protest',
# Set coordinates
x = c(ymd('2025-06-02')), y = before_treatment + 0.1,
# Set aesthetics
size = 3.5, color = 'red', hjust = 0
) +
# Add text
annotate(
geom = 'text', label = 'Southeast Asia before protest',
# Set coordinates
x = c(ymd('2025-06-02')), y = before_control + 0.1,
# Set aesthetics
size = 3.5, color = 'blue2', hjust = 0
) +
# Add text
annotate(
geom = 'text', label = 'Indonesia after protest',
# Set coordinates
x = min_x + 1, y = after_treatment + 0.1,
# Set aesthetics
size = 3.5, color = 'red', hjust = 0
) +
# Add text
annotate(
geom = 'text', label = 'Southeast Asia after protest',
# Set coordinates
x = min_x + 1, y = after_control + 0.1,
# Set aesthetics
size = 3.5, color = 'blue2', hjust = 0
) +
# Add text
annotate(
geom = 'text', label = 'Indonesia "without" intervention\n(parallel trend assumption)',
# Set coordinates
x = min_x + 1, y = after_treatment - diff_in_diff - 0.1,
# Set aesthetics
size = 3.5, color = 'blue2', hjust = 0, fontface = "bold"
) +
# Add labels
labs(
title = 'Difference-in-differences parallel trend assumption',
subtitle = 'Average demand multiplier (x Times average demand for One Piece in the region)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo'
) +
# Modify colors
scale_color_manual(values = c('red2','blue2')) +
# Modify Y axis limits and units
scale_y_continuous(limits = c(0.5,2), labels = scales::unit_format(suffix = 'x')) +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
# Remove legend (recommended to avoid writting to many `show.legend = F`)
legend.position = 'none'
)
We want to determine whether the Difference-in-Differences (DiD) estimate of 0.16 is statistically significant. In other words, we aim to confirm that this estimate is genuinely different from zero and not merely a result of random variation.
Linear regression is an effective method for simultaneously analyzing the combined effects of multiple factors. When applied to a DiD analysis, it considers important information, such as whether a market received the intervention, whether the time period is before or after the intervention, and crucially, whether the market was treated after the change.
The fundamental idea is straightforward: we want to mathematically describe our red and blue dashed lines in Figure 2 using the following equation:
\[Demand Multiplier = \beta_0 + \beta_1intervention + \beta_2treatment + \beta_3DiD+\epsilon\]
The following table displays the results of our linear regression, which uses the average performance of Southeast Asian markets as a control. Notably, the estimates for each variable correspond with those from Table 1. This correlation occurs because linear regression uses matrix operations to estimate the coefficients, ( _i ), centered around the mean.
The advantage of conducting this statistical analysis is that it allows us to assess the statistical significance of our findings. As we can observe, the estimate of -0.01 for the intervention variable, which we calculated earlier, is not statistically significant. This means it is not significantly different from zero, indicating that our control market did not experience any change in average performance following the onset of the Jolly Roger flag protest.
In contrast, estimates that are statistically significant are considered different from zero. Specifically, our Difference-in-Differences (DiD) estimator indicates an increase of 0.16 units in the multiplier when compared to the control group, which can be attributed to the Jolly Roger flag protest associated with One Piece.
model_1 = lm(
# This is how we input the formula for the regresion model
formula = region_multiplier ~ intervention + treatment + `DiD`,
# We are using our previously filtered data
data = data_se
)| Dependent variable: | |
| Demand Multiplier | |
| intervention | -0.013 |
| (0.038) | |
| treatment | 0.798*** |
| (0.039) | |
| DiD | 0.162*** |
| (0.053) | |
| Constant | 0.756*** |
| (0.028) | |
| Observations | 30 |
| R2 | 0.977 |
| Adjusted R2 | 0.975 |
| Residual Std. Error | 0.073 (df = 26) |
| F Statistic | 372.429*** (df = 3; 26) |
| Note: | ⋆p<0.1;⋆⋆p<0.05; ⋆⋆⋆p<0.01 |
A key component of this analysis is a specific term that captures the combined effect of being in the treated group after the intervention. This DiD term indicates how much of the change in demand can be directly attributed to the intervention, helping to isolate its impact from other influencing factors and providing a clearer understanding of its true effects.
To strengthen the validity of the Southeast Asia model, we will estimate a new DiD regression using the average market data from East Asia as our control group. This approach allows us to account for variations in the show’s popularity within the Asian market, such as the release of new episodes or other relevant events. It also helps prevent underestimating the effects due to the influence Indonesia may have had on other Southeast Asian markets, such as Nepal.
Consistent with the previous model, the East Asia model shows a significant increase of 0.16 units in the multiplier compared to the average East Asian market. Additionally, we do not observe any significant impact from the Jolly Roger flag protest in the control market.
model_2 = data %>%
# Filter to remove the Southeast market
filter(market != 'Southeast Asia') %>%
# Run the regresion
lm(formula = region_multiplier ~ intervention + treatment + `DiD`)| Dependent variable: | ||
| Demand Multiplier | ||
| Southeast Asia | East Asia | |
| (1) | (2) | |
| intervention | -0.013 | -0.014 |
| (0.038) | (0.040) | |
| treatment | 0.798*** | 0.372*** |
| (0.039) | (0.041) | |
| DiD | 0.162*** | 0.163*** |
| (0.053) | (0.056) | |
| Constant | 0.756*** | 1.182*** |
| (0.028) | (0.029) | |
| Observations | 30 | 30 |
| R2 | 0.977 | 0.915 |
| Adjusted R2 | 0.975 | 0.905 |
| Residual Std. Error (df = 26) | 0.073 | 0.077 |
| F Statistic (df = 3; 26) | 372.429*** | 92.982*** |
| Note: | ⋆p<0.1; ⋆⋆p<0.05; ⋆⋆⋆p<0.01 | |
Finally, we can examine potential short-term spillovers in the Southeast Asian markets by applying the same methodology. We will designate the average Southeast Asian market as our treatment group, while the average East Asian market will serve as our control group for the DiD regression. The table below shows that neither the intervention nor the DiD estimates are statistically significant. This indicates that we cannot find evidence of any differentiated behavior between these two markets following the intervention. At least for the analyzed period, there were no short-term spillover effects observed in the average Southeast Asian market.
model_3 = data %>%
# Filter to remove the Southeast market
filter(market != 'Indonesia') %>%
mutate(
treatment = ifelse(market == 'Southeast Asia', 1, 0),
DiD = treatment * intervention) %>%
# Run the regresion
lm(formula = region_multiplier ~ intervention + treatment + `DiD`)| Dependent variable: | |
| Demand Multiplier | |
| intervention | -0.014 |
| (0.022) | |
| treatment | -0.426*** |
| (0.022) | |
| DiD | 0.001 |
| (0.031) | |
| Constant | 1.182*** |
| (0.016) | |
| Observations | 30 |
| R2 | 0.967 |
| Adjusted R2 | 0.964 |
| Residual Std. Error | 0.042 (df = 26) |
| F Statistic | 256.540*** (df = 3; 26) |
| Note: | ⋆p<0.1; ⋆⋆p<0.05; ⋆⋆⋆p<0.01 |
We have seen the profound power of One Piece and its iconic Jolly Roger flag,which extend beyond mere entertainment and resonate deeply within Indonesia’s protest movement. Statistical methods like difference-in-differences support this connection, reminding us that symbols like One Piece embody collective hopes and desires, serving as cultural anchors during times of social upheaval.
What is the symbolic value of One Piece for the Jolly Roger protests in Indonesia? Our analysis reveals that it is at least enough to raise the demand of the series, which manga has sold over 516 million copies and anime has amassed millions of viewing hours on platforms like Netflix, surged by nearly 9.5% in response to the protests. This increase underscores how popular culture can unite communities in their shared aspirations for change and justice.
data %>%
filter(market == 'Indonesia') %>%
# Create a ggplot canvas
ggplot(
# Data-related aesthetics
mapping = aes(x = date, y = region_multiplier, col = market)
) +
# Add vertical line
geom_vline(
# This is the week when the first flag was shown (at leas in the news)
xintercept = c(cutoff),
# Modify the aesthetics
linetype = 'dashed', col = 'gray60'
) +
# Add point
geom_point(size = 1, alpha = 0.25) +
# Add a dashed line with the average multiplier for the Before
geom_xspline(
data = filter(data, date < cutoff, market == 'Indonesia') %>%
mutate(date = ifelse(date == max(date), cutoff, date)),
mapping = aes(x = date, y = region_multiplier, col = market),
linewidth = 0.5, alpha = 0.25
) +
# Add a dashed line with the average multiplier for the After
geom_xspline(
data = filter(data, date >= cutoff, market == 'Indonesia'),
mapping = aes(x = date, y = region_multiplier, col = market),
spline_shape = -0.5, alpha = 0.25
) +
# Add a dashed line with the average multiplier for the Before
stat_smooth(
data = filter(data, date < cutoff, market == 'Indonesia')%>%
mutate(date = ifelse(date == max(date), cutoff, date)),
mapping = aes(x = date, y = average_multiplier),
method = 'lm', linetype = 'dashed', linewidth = 0.75, col = 'red2'
) +
# Add a dashed line with the average multiplier for the After
stat_smooth(
data = filter(data, date >= cutoff, market == 'Indonesia') ,
mapping = aes(x = date, y = average_multiplier),
method = 'lm', linetype = 'dashed', linewidth = 0.75, col = 'red2'
) +
annotate(
geom = 'segment',
x = cutoff, xend = cutoff,
y = before_treatment, yend = after_treatment,
col = 'red2', linewidth = 1
) +
annotate(
geom = 'point',
x = cutoff, y = c(before_treatment, after_treatment),
shape = 18, size = 4, col = 'red2'
) +
# Annotate the relevant text
annotate(
# Use GeomRichtext as geom
geom = ggtext::GeomRichtext,
# Set coordinates
x = cutoff - 1, y = after_treatment,
# Html and linked to data format
label = sprintf(
fmt =
"A <span style='color:black'>**%s%% increase** in One Piece performance<br>can be attributed to the Jolly Roger flag protest</span>",
round(x = 100 * diff_in_diff/after_treatment, digits = 2)
),
# Set aesthetics
size = 2.5, hjust = 1, vjust = 0.5, color = 'red2',
) +
# Add labels
labs(
title = 'Weekly Demand for One Piece in Indonesia',
subtitle = 'Overall Southeast and East Asia demand multiplier (x Times average demand for One Piece)',
caption = 'Source: own elaboration with data from Parrot Analytics\nAuthor:@renorosgo'
) +
# Modify colors
scale_color_manual(values = c('red2')) +
# Modify Y axis limits and units
scale_y_continuous(limits = c(1.45,1.85), labels = scales::unit_format(suffix = 'x')) +
scale_x_date(date_labels = "%d-%b-%y", date_breaks = "2 weeks") +
# Set the base theme
theme_bw(paper = 'gray99', base_family = 'Avenir Next') +
# Additional tweaks to the theme
theme(
# Remove titles from de axis
axis.title = element_blank(),
# Delete some grids
panel.grid.minor = element_blank(),
panel.grid.major.x = element_blank(),
# Remove legend (recommended to avoid writting to many `show.legend = F`)
legend.position = 'none'
)
Ultimately, this case illustrates that entertainment and storytelling are not merely forms of escapism; they are vital forces that shape social identities and inspire collective action. As societies continue to evolve, embracing these cultural dynamics becomes essential for engaging with and interpreting contemporary social movements.