Rationale

Agenda-setting theory says that the issues the news media cover the most tend to become the issues people see as most important. McCombs and Shaw (1972) first tested this idea during the 1968 presidential election in Chapel Hill, North Carolina. Their work built on an earlier point from Cohen (1963): the media may not always tell people what to think, but they are very good at telling people what to think about.

An important part of this theory is that the media agenda has limited space. News outlets only have so many reporters, so much airtime, and so much room on a homepage. Because of this, issues compete with each other for attention. Zhu (1992) called this the “zero-sum” side of agenda-setting. When one issue gets more coverage, another issue may get less. At other times, two issues may trade places, or they may stay at about the same level.

Using first-level agenda-setting theory, my project compares how much media attention two criminal justice issues received: police reform and violent crime. Both are about public safety, and both come up often in national politics, especially in 2026, a midterm election year. But they come at public safety from opposite sides. Police reform coverage is about how police behave, such as misconduct, use of force, and oversight. Violent crime coverage is about crime itself, such as shootings, homicides, and assaults. One issue points the public toward holding police accountable. The other points toward stronger policing. That makes them a good test of how competing issues share the media agenda.

The analysis compares the number of stories per week that MediaCloud.org’s United States - National collection gathered about each issue, from the week of Jan. 4, 2026, through the week ending Oct. 24, 2026. The results show whether attention to one side of the public safety debate comes at the expense of the other.

Hypothesis

Average weekly coverage volume of police reform and violent crime differed during the first 10 months of 2026.

Variables & method

The dependent variable is weekly coverage volume, a continuous variable. It is the number of stories published each week that match each issue’s MediaCloud.org query.

The independent variable is issue, a categorical variable with two categories: “Police Reform” and “Violent Crime.” Each category is defined by keywords likely to appear in stories about that issue.

Police reform stories were defined as stories matching this MediaCloud query:

reform_query <-
  "((police OR policing OR officer*) AND
     (reform OR accountability OR misconduct OR
      brutality OR \"use of force\" OR oversight OR
      \"body camera\" OR \"police violence\"))"

Violent crime stories were defined as stories matching this MediaCloud query:

crime_query <-
  "((homicide* OR murder* OR shooting* OR
     \"violent crime\" OR assault*) AND
     (suspect* OR victim* OR killed OR
      wounded OR charged))"

Some stories, such as those about police shootings, may match both queries and count toward both issues.

A paired-samples t-test checked whether the difference in weekly coverage between the two issues was statistically significant. A paired test fits because each week has one count for each issue.

Results & discussion

The graph and table below show each issue’s weekly coverage volume.

Media Cloud Topic Summary Statistics
topic Total_Stories Weekly_Minimum Weekly_Maximum Weekly_Median Weekly_Mean Weekly_SD
Violent Crime 133,857 2,178 5,354 3,117.0 3,187.07 596.87
Police Reform 45,769 661 1,941 1,087.5 1,089.74 270.17

Across the 42 weeks studied, violent crime got more coverage than police reform in every single week. On average, violent crime got more coverage than police reform. Police reform averaged 1,090 stories per week (median = 1,088). Violent crime averaged 3,187 stories per week (median = 3,117). The median, which unusually busy weeks affect less, was higher for violent crime.

Police reform coverage peaked the week of January 25, 2026, with 1,941 stories. Violent crime coverage peaked the week of January 25, 2026, with 5,354 stories. That week followed the Jan. 24 killing of Alex Pretti by federal agents in Minneapolis, less than three weeks after an ICE officer fatally shot Renee Good in the same city on Jan. 7. Stories about these events involved both deadly shootings and calls for accountability, so many likely matched both queries. That may explain why both issues spiked at once. It also shows the two issues do not always compete: one event can raise attention to both.

The standard deviations in the table show how steady each issue’s coverage was. A standard deviation that is large compared with the mean means coverage jumped around certain events instead of staying steady.

One likely reason for the gap is the kind of news each issue produces. Violent crime produces stories almost every day, which keeps its coverage high. Police reform coverage depends more on specific events, like a viral video, a major court ruling, or a new law. Those events cause spikes followed by quiet periods. Agenda-setting theory also suggests that when one issue surges, the other may lose ground. A major misconduct case could briefly push crime coverage down, and a wave of shootings could push reform coverage aside. Because 2026 is a midterm election year, candidates’ messages about public safety may also have shaped coverage, especially before the November election.

Paired-samples t-test

The paired-samples t-test checked whether the average weekly difference between the two issues was reliably above or below zero. This box plot compares the weekly story counts for police reform (V1) and violent crime (V2):

The pair differences

Stories about violent crime outnumbered stories about police reform by an average of 2,097 per week, with a standard deviation of 461. The average gap was larger than the standard deviation, so the difference was both large and fairly steady from week to week.

Descriptive Statistics: Pair Differences
count mean sd min max
42.000 2,097.333 461.190 1,390.000 3,534.000

A Shapiro-Wilk test found the pair differences were not normally distributed (p = 0.002). But with 42 weeks of paired data, the t-test is still reliable. Normality matters most for small samples.

Normality Test (Shapiro-Wilk)
statistic p.value method
0.9036 0.0018 Shapiro-Wilk normality test
If the P.VALUE is 0.05 or less, the number of pairs is fewer than 40, and the distribution of pair differences shows obvious non-normality or outliers, consider using the Wilcoxon Signed Rank Test results instead of the Paired-Samples t-Test results.

Paired t-test results

The t-test found that the average weekly difference was significantly different from zero (t = 29.47, df = 41, p < 0.001). The hypothesis was supported. Average weekly coverage of police reform and violent crime differed during the first 10 months of 2026, and violent crime got more coverage.

Paired-Samples t-Test
statistic parameter p.value conf.low conf.high method
29.4722 41 0.0000 1,953.6165 2,241.0502 Paired t-test
Group Means and SDs (t-Test)
V1_Mean V2_Mean V1_SD V2_SD
1,089.738 3,187.071 270.167 596.868

Discussion

The results fit the zero-sum view of agenda-setting. Violent crime held a much larger share of the media agenda than police reform. Agenda-setting theory suggests news audiences were prompted to think about violent crime more often than police reform when they thought about public safety. The shared peak in late January adds a twist: the issues did not always compete, since one event sometimes raised attention to both.

These results have limits. A t-test can rule out random chance, but not every other explanation. The queries matter. The violent crime query catches any story about a homicide, shooting or assault, and those appear almost daily. The police reform query needs both a policing term and an accountability term, which narrows it. Different keywords, news sources or dates could give a different result. A next step would be to read stories from the peak weeks to see what drove the spikes and whether the queries caught what they were meant to.

References

Cohen, B. C. (1963). The press and foreign policy. Princeton University Press.

McCombs, M. E., & Shaw, D. L. (1972). The agenda-setting function of mass media. Public Opinion Quarterly, 36(2), 176–187.

Zhu, J.-H. (1992). Issue competition and attention distraction: A zero-sum theory of agenda-setting. Journalism Quarterly, 69(4), 825–836.

Code

Here is the code that produced the analysis.

############################################################
# STEP 1: Install and Load Required Packages
############################################################

required_packages <- c(
  "tidyverse",
  "httr2",
  "lubridate",
  "plotly",
  "knitr",
  "kableExtra"
)

for (pkg in required_packages) {
  if (!requireNamespace(pkg, quietly = TRUE)) {
    install.packages(pkg)
  }
}

library(tidyverse)
library(httr2)
library(lubridate)
library(plotly)
library(knitr)
library(kableExtra)
############################################################
# STEP 2: Verify Media Cloud API Key
############################################################

if (!nzchar(Sys.getenv("MEDIACLOUD_KEY"))) {
  stop(
    "No Media Cloud API key was found.\n",
    "Run the setup section once,\n",
    "then restart R and try again."
  )
}

mc_key <- Sys.getenv("MEDIACLOUD_KEY")
############################################################
# STEP 3: Define the Date Range
############################################################

start_date <- "2026-01-04"
end_date   <- "2026-10-24"
############################################################
# STEP 4: Define Search Queries
############################################################

reform_query <-
  "((police OR policing OR officer*) AND
     (reform OR accountability OR misconduct OR
      brutality OR \"use of force\" OR oversight OR
      \"body camera\" OR \"police violence\"))"

crime_query <-
  "((homicide* OR murder* OR shooting* OR
     \"violent crime\" OR assault*) AND
     (suspect* OR victim* OR killed OR
      wounded OR charged))"
############################################################
# STEP 5: Create Topic Definitions
############################################################

topics <- c(
  "Police Reform" = reform_query,
  "Violent Crime" = crime_query
)
############################################################
# STEP 6: Create a Function to Retrieve Daily
# Story Counts
############################################################

mc_counts_over_time <- function(
    query,
    start_date,
    end_date,
    collection_id = 34412234
) {

  resp <-
    request(
      "https://search.mediacloud.org/api/search/count-over-time"
    ) |>
    req_headers(
      Authorization = paste("Token", mc_key)
    ) |>
    req_url_query(
      q = query,
      start = start_date,
      end = end_date,
      platform = "onlinenews-mediacloud",
      cs = collection_id
    ) |>
    req_perform()

  results <- resp |>
    resp_body_json()

  map_dfr(
    results$count_over_time$counts,
    as_tibble
  ) |>
    mutate(
      date = as.Date(date)
    )
}
############################################################
# STEP 7: Download Daily Counts for All Topics
############################################################

topic_data <-
  imap_dfr(
    topics,
    function(search_string, topic_name) {

      message(
        "Downloading: ",
        topic_name
      )

      results <-
        mc_counts_over_time(
          query = search_string,
          start_date = start_date,
          end_date = end_date
        )

      Sys.sleep(35)

      results |>
        mutate(
          topic = topic_name
        )
    }
  )
############################################################
# STEP 8: Examine the Downloaded Daily Data
############################################################

head(topic_data)

glimpse(topic_data)
############################################################
# STEP 9: Convert Daily Counts to Weekly Counts
############################################################

weekly_counts <-
  topic_data |>
  mutate(
    week = floor_date(
      date,
      unit = "week"
    )
  ) |>
  group_by(
    topic,
    week
  ) |>
  summarise(
    stories = sum(count),
    .groups = "drop"
  )
############################################################
# STEP 10: Examine Weekly Counts
############################################################

weekly_counts
############################################################
# STEP 11: Summarize Coverage by Topic
############################################################

topic_summary <-
  weekly_counts |>
  group_by(topic) |>
  summarise(
    Total_Stories = sum(stories),
    Weekly_Minimum = min(stories),
    Weekly_Maximum = max(stories),
    Weekly_Median = median(stories),
    Weekly_Mean = round(
      mean(stories),
      2
    ),
    Weekly_SD = round(
      sd(stories),
      2
    ),
    .groups = "drop"
  ) |>
  arrange(
    desc(Total_Stories)
  )

Table <-
  topic_summary |>
  kbl(
    caption =
      "Media Cloud Topic Summary Statistics",
    align = "lrrrrrr",
    format.args = list(big.mark = ",")
  ) |>
  kable_styling(
    full_width = FALSE,
    bootstrap_options = c(
      "striped",
      "hover",
      "condensed"
    )
  )
############################################################
# STEP 12: Create an Interactive Plotly Graph
############################################################

weekly_counts <- weekly_counts |>
  mutate(
    stories_formatted = formatC(
      stories,
      format = "d",
      big.mark = ","
    )
  )

Graph <- plot_ly(
  data = weekly_counts,

  x = ~week,
  y = ~stories,

  color = ~topic,
  colors = "Set1",

  type = "scatter",
  mode = "lines+markers",

  customdata = ~topic,

  text = ~stories_formatted,

  hovertemplate =
    paste(
      "%{customdata}<br>",
      "Week: %{x|%Y-%m-%d}<br>",
      "Stories: %{text}",
      "<extra></extra>"
    )
) |>
  layout(
    title =
      "Weekly U.S. Coverage Volume",

    hovermode = "x unified",

    xaxis = list(
      title = "Week"
    ),

    yaxis = list(
      title = "Weekly Story Count",
      separatethousands = TRUE
    )
  )
############################################################
# STEP 13: Calculate Values Reported in the Text
############################################################

ref <- topic_summary |> filter(topic == "Police Reform")
cri <- topic_summary |> filter(topic == "Violent Crime")

fmt <- function(x) formatC(x, format = "f", digits = 0, big.mark = ",")

higher_mean   <- if (ref$Weekly_Mean > cri$Weekly_Mean) "police reform" else "violent crime"
lower_mean    <- if (ref$Weekly_Mean > cri$Weekly_Mean) "violent crime" else "police reform"
higher_median <- if (ref$Weekly_Median > cri$Weekly_Median) "police reform" else "violent crime"

weeks_ref_ahead <-
  weekly_counts |>
  select(topic, week, stories) |>
  pivot_wider(names_from = topic, values_from = stories) |>
  summarise(n = sum(`Police Reform` > `Violent Crime`, na.rm = TRUE)) |>
  pull(n)

total_weeks <- n_distinct(weekly_counts$week)

weeks_text <-
  if (weeks_ref_ahead == 0) {
    "violent crime got more coverage than police reform in every single week."
  } else if (weeks_ref_ahead == total_weeks) {
    "police reform got more coverage than violent crime in every single week."
  } else {
    paste0(
      "police reform got more coverage than violent crime in ", weeks_ref_ahead,
      " weeks, and violent crime got more coverage in the other ",
      total_weeks - weeks_ref_ahead, "."
    )
  }

peak_ref <- weekly_counts |>
  filter(topic == "Police Reform") |>
  slice_max(stories, n = 1, with_ties = FALSE)

peak_cri <- weekly_counts |>
  filter(topic == "Violent Crime") |>
  slice_max(stories, n = 1, with_ties = FALSE)
############################################################
# STEP 14: Reshape the Data
############################################################

weekly_counts_wide <-
  weekly_counts %>%
  mutate(
    topic = make.names(topic)
  ) %>%
  select(week, topic, stories) %>%
  pivot_wider(
    names_from  = topic,
    values_from = stories
  )
# ============================================================
#  STEP 15: Paired-Samples t-Test
#
#  Setup: Install and Load Required Packages
# ============================================================
if (!require("tidyverse")) install.packages("tidyverse")
if (!require("plotly")) install.packages("plotly")
if (!require("gt")) install.packages("gt")
if (!require("gtExtras")) install.packages("gtExtras")
if (!require("broom")) install.packages("broom")

library(tidyverse)
library(plotly)
library(gt)
library(gtExtras)
library(broom)

options(scipen = 999)

# ============================================================
#  Data Import
# ============================================================
mydata <- weekly_counts_wide

# Specify the two variables involved
mydata$V1 <- mydata$Police.Reform
mydata$V2 <- mydata$Violent.Crime

# ============================================================
#  Compute Pair Differences
# ============================================================
mydata$PairDifferences <- mydata$V2 - mydata$V1

# ============================================================
#  Interactive Histogram of Pair Differences
# ============================================================
hist_plot <- plot_ly(
  data = mydata,
  x = ~PairDifferences,
  type = "histogram",
  marker = list(color = "#1f78b4", line = list(color = "black", width = 1))
) %>%
  layout(
    title = "Distribution of Pair Differences",
    xaxis = list(title = "Pair Differences"),
    yaxis = list(title = "Count"),
    shapes = list(
      list(
        type = "line",
        x0 = mean(mydata$PairDifferences, na.rm = TRUE),
        x1 = mean(mydata$PairDifferences, na.rm = TRUE),
        y0 = 0,
        y1 = max(table(mydata$PairDifferences)),
        line = list(color = "red", dash = "dash")
      )
    )
  )

# ============================================================
#  Descriptive Statistics
# ============================================================
desc_stats <- mydata %>%
  summarise(
    count = n(),
    mean = mean(PairDifferences, na.rm = TRUE),
    sd = sd(PairDifferences, na.rm = TRUE),
    min = min(PairDifferences, na.rm = TRUE),
    max = max(PairDifferences, na.rm = TRUE)
  )

desc_table <- desc_stats %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Descriptive Statistics: Pair Differences") %>%
  fmt_number(columns = where(is.numeric), decimals = 3)

# ============================================================
#  Normality Test (Shapiro-Wilk)
# ============================================================
shapiro_res <- shapiro.test(mydata$PairDifferences)
shapiro_table <- tidy(shapiro_res) %>%
  select(statistic, p.value, method) %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Normality Test (Shapiro-Wilk)") %>%
  fmt_number(columns = c(statistic, p.value), decimals = 4) %>%
  tab_source_note(
    source_note = "If the P.VALUE is 0.05 or less, the number of pairs is fewer than 40, and the distribution of pair differences shows obvious non-normality or outliers, consider using the Wilcoxon Signed Rank Test results instead of the Paired-Samples t-Test results."
  )

# ============================================================
#  Reshape Data for Repeated-Measures Plot
# ============================================================
df_long <- mydata %>%
  pivot_longer(cols = c(V1, V2),
               names_to = "Measure",
               values_to = "Value")

# ============================================================
#  Repeated-Measures Boxplot (Interactive, with Means)
# ============================================================
group_means <- df_long %>%
  group_by(Measure) %>%
  summarise(mean_value = mean(Value), .groups = "drop")

boxplot_measures <- plot_ly() %>%
  add_trace(
    data = df_long,
    x = ~Measure, y = ~Value,
    type = "box",
    boxpoints = "outliers",
    marker = list(color = "red", size = 4),
    line = list(color = "black"),
    fillcolor = "royalblue",
    name = ""
  ) %>%
  add_trace(
    data = group_means,
    x = ~Measure, y = ~mean_value,
    type = "scatter", mode = "markers",
    marker = list(
      symbol = "diamond", size = 9,
      color = "black", line = list(color = "white", width = 1)
    ),
    text = ~paste0("Mean = ", round(mean_value, 2)),
    hoverinfo = "text",
    name = "Group Mean"
  ) %>%
  layout(
    title = "Boxplot of Repeated Measures (V1 vs V2) with Means",
    xaxis = list(title = "Measure"),
    yaxis = list(title = "Value"),
    showlegend = FALSE
  )

# ============================================================
#  Parametric Test (Paired-Samples t-Test)
# ============================================================
t_res <- t.test(mydata$V2, mydata$V1, paired = TRUE)
t_table <- tidy(t_res) %>%
  select(statistic, parameter, p.value, conf.low, conf.high, method) %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Paired-Samples t-Test") %>%
  fmt_number(columns = c(statistic, p.value, conf.low, conf.high), decimals = 4)

t_summary <- mydata %>%
  select(V1, V2) %>%
  summarise_all(list(Mean = mean, SD = sd)) %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Group Means and SDs (t-Test)") %>%
  fmt_number(columns = everything(), decimals = 3)

# ============================================================
#  Nonparametric Test (Wilcoxon Signed Rank)
# ============================================================
wilcox_res <- wilcox.test(mydata$V1, mydata$V2, paired = TRUE,
                          exact = FALSE)
wilcox_table <- tidy(wilcox_res) %>%
  select(statistic, p.value, method) %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Wilcoxon Signed Rank Test") %>%
  fmt_number(columns = c(statistic, p.value), decimals = 4)

wilcox_summary <- mydata %>%
  select(V1, V2) %>%
  summarise_all(list(Mean = mean, SD = sd)) %>%
  gt() %>%
  gt_theme_538() %>%
  tab_header(title = "Group Means and SDs (Wilcoxon)") %>%
  fmt_number(columns = everything(), decimals = 3)
############################################################
# STEP 16: Calculate t-Test Values Reported in the Text
############################################################

n_pairs   <- nrow(mydata)
diff_mean <- mean(mydata$PairDifferences, na.rm = TRUE)
diff_sd   <- sd(mydata$PairDifferences, na.rm = TRUE)

shapiro_p      <- shapiro_res$p.value
shapiro_normal <- shapiro_p > 0.05

t_value <- formatC(unname(t_res$statistic), format = "f", digits = 2)
t_df    <- unname(t_res$parameter)
t_p     <- t_res$p.value
t_sig   <- t_p <= 0.05

p_text <- function(p) {
  if (p < 0.001) "p < 0.001" else paste0("p = ", formatC(p, format = "f", digits = 3))
}

more_issue <- if (diff_mean > 0) "violent crime" else "police reform"
less_issue <- if (diff_mean > 0) "police reform" else "violent crime"
Graph
Table
boxplot_measures
desc_table
shapiro_table
t_table
t_summary