Replication of “Anti-Immigrant Rhetoric and ICE Reporting Interest: Evidence from a Large-Scale Study of Web Search Data”
Suvedei Soyol-Erdene (Baruch College) and Eric Guan (Hunter College)
7/6/2024
# Loading in all datasets used for replication
# Dataset downloaded from Google Trends 2024
google_trends_2024 <- read.csv("data/from_google_trends_2024/google_trend_search.csv", header = FALSE, col.names = c("ym", "report", "crime", "welfare"))
google_trends_2024 <- google_trends_2024[-c(1, 2), ] %>% mutate(ym = ym(ym), report = as.numeric(report), crime = as.numeric(crime), welfare = as.numeric(welfare)) %>% filter(ym < ym("2020-01"))
google_trends_2024 <- google_trends_2024 %>% mutate( president = case_when(
ym < ym("2009-01") ~ "Bush" ,
ym < ym("2017-01") ~ "Obama" ,
.default = "Trump"))
# Original dataset used in the paper
original_crime <- read.csv("data/from_replication_files/google_trends_crime.csv", col.names = c("year", "month", "crime")) %>%
unite(date,c(year, month), sep = "-")
original_report <- read.csv("data/from_replication_files/google_trends_report.csv" , col.names = c("year", "month", "report")) %>%
unite(date,c(year, month), sep = "-")
original_welfare <- read.csv("data/from_replication_files/google_trends_welfare.csv", col.names = c("year", "month", "welfare")) %>%
unite(date,c(year, month), sep = "-")
trends_original <- left_join(original_crime, original_report, by = join_by(date)) %>%
left_join(original_welfare, by = join_by(date)) %>%
mutate(date = ym(date))
trends_original <- trends_original %>% mutate( president = case_when(
date < ym("2009-01") ~ "Bush" ,
date < ym("2017-01") ~ "Obama" ,
.default = "Trump"))
# Import the topic model
library(stm)
load("data/topic_model_lite.RData")
# Import Daily Google Trends dataset
daily_report <- read.csv(("data/from_replication_files/gt_report_daily.csv")) %>%
mutate(ymd = ymd(date)) %>%
select(ymd, search, search_adj)Figure 4 Replication: Google Trend data split by Presidential Term
Replication of Figure 4 using 2024 Google Trends Data
#plotting reporting
google_trends_2024 %>%
ggplot(aes(x = ym, y = report, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") +
xlab("") +
ggtitle("Reporting Trends")
#plotting crime
google_trends_2024 %>%
ggplot(aes(x = ym, y = crime, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") + xlab("") +
scale_y_continuous(limits = c(0, 100)) +
ggtitle("Crime Trends")
#plotting welfare
google_trends_2024 %>%
ggplot(aes(x = ym, y = welfare, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") + xlab("") + scale_y_continuous(limits = c(0, 100)) +
ggtitle("Welfare Trends")
Replication of Figure 4 using the Original Dataset
#plotting reporting
trends_original %>%
ggplot(aes(x = date, y = report, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") +
xlab("") +
ggtitle("Reporting Trends")
#plotting crime
trends_original %>%
ggplot(aes(x = date, y = crime, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") + xlab("") +
scale_y_continuous(limits = c(0, 100)) +
ggtitle("Crime Trends")
#plotting welfare
trends_original %>%
ggplot(aes(x = date, y = welfare, color = president), group_by = president) +
geom_point() + geom_smooth(method = "lm", se = FALSE) +
ylab("Google Trends") + xlab("") + scale_y_continuous(limits = c(0, 100)) +
ggtitle("Welfare Trends")
Figure 4 Report:
When replicating Figure 4 using the orignal dataset, we got near identical results. However, when replicating with the 2024 Google Trends data, we noticed a lot of zero values that weren’t present in the original dataset for the paper. This resulted in slightly skewed regression lines for the Bush administration. We suspect that maybe Google Trends dropped the data for those specific days.
Table 3 Replication: Regression Model using Google Trends data
Replication of Table 3 using 2024 Google Trends Data
#Crime
google_trends_2024$president <- relevel(factor(google_trends_2024$president), ref = "Obama")
crime_regression <- lm(crime ~ ym + president , data = google_trends_2024)
print("Crime_Regression")## [1] "Crime_Regression"summary(crime_regression)##
## Call:
## lm(formula = crime ~ ym + president, data = google_trends_2024)
##
## Residuals:
## Min 1Q Median 3Q Max
## -13.508 -4.959 -2.184 3.035 37.960
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -3.160e+01 1.367e+01 -2.312 0.02188 *
## ym 2.584e-03 8.696e-04 2.971 0.00336 **
## presidentBush 3.179e+00 2.465e+00 1.290 0.19867
## presidentTrump 7.206e+00 2.368e+00 3.043 0.00268 **
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 8.183 on 188 degrees of freedom
## Multiple R-squared: 0.3324, Adjusted R-squared: 0.3218
## F-statistic: 31.21 on 3 and 188 DF, p-value: < 2.2e-16report_regression <- lm(report ~ ym + president , data = google_trends_2024)
print("Report_Regression")## [1] "Report_Regression"summary(report_regression)##
## Call:
## lm(formula = report ~ ym + president, data = google_trends_2024)
##
## Residuals:
## Min 1Q Median 3Q Max
## -47.959 -5.455 -0.649 4.599 42.899
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 70.042554 17.008321 4.118 5.72e-05 ***
## ym -0.001700 0.001082 -1.571 0.118
## presidentBush -0.976845 3.067040 -0.318 0.750
## presidentTrump 16.289312 2.946926 5.528 1.07e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 10.18 on 188 degrees of freedom
## Multiple R-squared: 0.1911, Adjusted R-squared: 0.1782
## F-statistic: 14.8 on 3 and 188 DF, p-value: 1.081e-08welfare_regression <- lm(welfare ~ ym + president, data = google_trends_2024)
print("Welfare_Regression")## [1] "Welfare_Regression"summary(welfare_regression)##
## Call:
## lm(formula = welfare ~ ym + president, data = google_trends_2024)
##
## Residuals:
## Min 1Q Median 3Q Max
## -11.0316 -3.7726 -0.3853 2.6861 27.3412
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -1.067e+01 1.171e+01 -0.910 0.3638
## ym 1.408e-03 7.452e-04 1.889 0.0604 .
## presidentBush 1.682e+00 2.112e+00 0.796 0.4268
## presidentTrump 4.744e+00 2.030e+00 2.337 0.0205 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 7.013 on 188 degrees of freedom
## Multiple R-squared: 0.1957, Adjusted R-squared: 0.1829
## F-statistic: 15.25 on 3 and 188 DF, p-value: 6.391e-09Replication of Table 3 using the Original Dataset
#Crime
trends_original$president <- relevel(factor(trends_original$president), ref = "Obama")
crime_regression <- lm(crime ~ date + president , data = trends_original)
print("Crime_Regression")## [1] "Crime_Regression"summary(crime_regression)##
## Call:
## lm(formula = crime ~ date + president, data = trends_original)
##
## Residuals:
## Min 1Q Median 3Q Max
## -21.480 -6.927 -3.641 3.294 62.821
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) -4.163158 23.241077 -0.179 0.8580
## date 0.001247 0.001479 0.843 0.4002
## presidentBush 8.280386 4.187005 1.978 0.0494 *
## presidentTrump 19.902636 4.027082 4.942 1.72e-06 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 13.87 on 186 degrees of freedom
## Multiple R-squared: 0.2621, Adjusted R-squared: 0.2502
## F-statistic: 22.03 on 3 and 186 DF, p-value: 2.988e-12report_regression <- lm(report ~ date + president , data = trends_original)
print("Report_Regression")## [1] "Report_Regression"summary(report_regression)##
## Call:
## lm(formula = report ~ date + president, data = trends_original)
##
## Residuals:
## Min 1Q Median 3Q Max
## -22.050 -5.662 -1.293 4.796 38.482
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 72.573674 16.089215 4.511 1.14e-05 ***
## date -0.001789 0.001024 -1.748 0.0821 .
## presidentBush 0.785389 2.898559 0.271 0.7867
## presidentTrump 19.716343 2.787848 7.072 2.99e-11 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 9.601 on 186 degrees of freedom
## Multiple R-squared: 0.2863, Adjusted R-squared: 0.2748
## F-statistic: 24.87 on 3 and 186 DF, p-value: 1.411e-13welfare_regression <- lm(welfare ~ date + president, data = trends_original)
print("Welfare_Regression")## [1] "Welfare_Regression"summary(welfare_regression)##
## Call:
## lm(formula = welfare ~ date + president, data = trends_original)
##
## Residuals:
## Min 1Q Median 3Q Max
## -19.966 -7.986 -1.646 4.774 56.953
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 28.584544 20.709656 1.380 0.1692
## date -0.000178 0.001318 -0.135 0.8927
## presidentBush 7.819065 3.730956 2.096 0.0375 *
## presidentTrump 18.560493 3.588452 5.172 5.95e-07 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 12.36 on 186 degrees of freedom
## Multiple R-squared: 0.237, Adjusted R-squared: 0.2247
## F-statistic: 19.26 on 3 and 186 DF, p-value: 6.428e-11Table 3 Report:
When we replicated Table 3 using the original data set, we got near identical results. However, when we tried to replicate Table 3 using the 2024 Google Trends data, only the “report trends” table had close results to the paper. The other two (welfare and crime) had drastically different coefficients, although the p-values are similar. This is most likely due to the effects of the extra zero values.
Figure 2-3 Replication: News Coverage data by News Channel, split by Trump’s Campaign Period
Replication of Figure 2 using the Original Topic Model
document_topics %>%
mutate( ym = ym(format(ymd(date), "%Y-%m"))) %>%
group_by(channel, ym, post_election, post_trump, time ) %>%
summarize( total_duration = sum(duration)) %>%
ggplot(aes(x = ym, y = total_duration, color = channel, group = interaction(channel, time))) +
geom_point() +
geom_smooth( se = F) +
geom_vline(xintercept = c(ym("2015-06"), ym("2017-01")), linetype = "dashed") +
ggtitle("Figure 2") +
ylab("Num Monthly Immigration Segs") +
xlab("")
Replication of Figure 3 using the Original Topic Model
document_topics %>%
select(Topic1, Topic3, channel, date, duration, post_election, post_trump, time) %>%
mutate(Topic = Topic1 + Topic3)%>%
mutate( ym = ym(format(ymd(date), "%Y-%m"))) %>%
group_by(channel, ym, post_election, post_trump, time ) %>%
summarize( total_duration = sum(Topic)) %>%
ggplot(aes(x = ym, y = total_duration, color = channel, group = interaction(channel, time))) +
geom_point() +
geom_smooth( se = F) +
geom_vline(xintercept = c(ym("2015-06"), ym("2017-01")), linetype = "dashed") +
ylab("Immigr + Crime \n News Coverage") +
xlab("") +
ggtitle("Figure 3")
document_topics %>%
select(Topic13, channel, date, duration, post_election, post_trump, time) %>%
mutate( ym = ym(format(ymd(date), "%Y-%m"))) %>%
group_by(channel, ym, post_election, post_trump, time ) %>%
summarize( total_duration = sum(Topic13)) %>%
ggplot(aes(x = ym, y = total_duration, color = channel, group = interaction(channel, time))) +
geom_point() +
geom_smooth( se = F) +
geom_vline(xintercept = c(ym("2015-06"), ym("2017-01")), linetype = "dashed")+
ylab("Immigr + Welfare \n News Coverage") +
xlab("") +
ggtitle("Figure 3")
Figure 2-3 Report:
Using the topic model data used by the research paper, we were able to make near identical graphs.
Table 4 Replication: Regression Model using News Coverage and Google Trends Data
Replication of Table 4 using Daily Topic Model and Google Trends data
immig_segment <- document_topics %>%
mutate( ymd = ymd(format(ymd(date), "%Y-%m-%d"))) %>%
group_by(ymd) %>%
summarize( segment = sum(duration))
immig_crime <- document_topics %>%
select(Topic1, Topic3, date) %>%
mutate(Topic = Topic1 + Topic3)%>%
mutate( ymd = ymd(format(ymd(date), "%Y-%m-%d"))) %>%
group_by(ymd) %>%
summarize( crime = sum(Topic))
immig_welfare <- document_topics %>%
select(Topic13, channel, date, duration, post_election, post_trump, time) %>%
mutate( ymd = ymd(format(ymd(date), "%Y-%m-%d"))) %>%
group_by(ymd) %>%
summarize( welfare = sum(Topic13))
immig_comb <- left_join(immig_crime, immig_segment, by = join_by(ymd)) %>%
left_join(immig_welfare, by = join_by(ymd)) %>%
mutate( Trump_admin = case_when(
ymd("2017-01-01") < ymd ~ TRUE,
.default = FALSE
), weekday = weekdays.Date(ymd), month = months.Date(ymd))# Merge all data frames together
table_4 <- inner_join(daily_report, immig_comb)
print("Search Data")## [1] "Search Data"regression <- lm(search ~ ymd + crime + segment + welfare + Trump_admin + weekday + month, table_4)
summary(regression)##
## Call:
## lm(formula = search ~ ymd + crime + segment + welfare + Trump_admin +
## weekday + month, data = table_4)
##
## Residuals:
## Min 1Q Median 3Q Max
## -47.254 -13.465 -1.511 11.179 62.069
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 77.908253 23.780954 3.276 0.001071 **
## ymd -0.002196 0.001432 -1.534 0.125254
## crime 0.064368 0.261010 0.247 0.805233
## segment 0.047044 0.018910 2.488 0.012932 *
## welfare 1.523201 0.666820 2.284 0.022459 *
## Trump_adminTRUE 6.251839 1.692285 3.694 0.000226 ***
## weekdayMonday 1.568848 1.543870 1.016 0.309667
## weekdaySaturday -3.043553 1.546315 -1.968 0.049175 *
## weekdaySunday -4.889389 1.540464 -3.174 0.001526 **
## weekdayThursday 3.177135 1.540152 2.063 0.039253 *
## weekdayTuesday -0.334169 1.538628 -0.217 0.828085
## weekdayWednesday 4.002338 1.537512 2.603 0.009306 **
## monthAugust -1.545502 1.975488 -0.782 0.434107
## monthDecember 4.896784 2.089877 2.343 0.019222 *
## monthFebruary 1.574983 2.006999 0.785 0.432696
## monthJanuary -1.247298 1.982518 -0.629 0.529324
## monthJuly 2.271054 1.977468 1.148 0.250913
## monthJune 3.280531 1.996264 1.643 0.100470
## monthMarch -6.166266 1.978310 -3.117 0.001853 **
## monthMay -0.207390 1.975072 -0.105 0.916383
## monthNovember 2.160816 2.080258 1.039 0.299058
## monthOctober -0.652621 2.071266 -0.315 0.752731
## monthSeptember -0.635930 2.041850 -0.311 0.755492
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.55 on 2007 degrees of freedom
## Multiple R-squared: 0.08522, Adjusted R-squared: 0.07519
## F-statistic: 8.499 on 22 and 2007 DF, p-value: < 2.2e-16print("Search Adjusted Data")## [1] "Search Adjusted Data"regression_adj <- lm(search_adj ~ ymd + crime + segment + welfare + Trump_admin + weekday + month, table_4)
summary(regression_adj)##
## Call:
## lm(formula = search_adj ~ ymd + crime + segment + welfare + Trump_admin +
## weekday + month, data = table_4)
##
## Residuals:
## Min 1Q Median 3Q Max
## -46.311 -11.892 -1.991 8.933 138.084
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 98.707429 24.124352 4.092 4.45e-05 ***
## ymd -0.003863 0.001453 -2.659 0.007901 **
## crime 0.728080 0.264779 2.750 0.006017 **
## segment 0.048404 0.019183 2.523 0.011701 *
## welfare 2.367886 0.676449 3.500 0.000475 ***
## Trump_adminTRUE 18.634297 1.716722 10.855 < 2e-16 ***
## weekdayMonday 0.605772 1.566164 0.387 0.698955
## weekdaySaturday -3.815175 1.568644 -2.432 0.015096 *
## weekdaySunday -5.312643 1.562708 -3.400 0.000688 ***
## weekdayThursday 2.510494 1.562391 1.607 0.108249
## weekdayTuesday -1.210984 1.560845 -0.776 0.437928
## weekdayWednesday 2.953994 1.559714 1.894 0.058378 .
## monthAugust 0.365126 2.004014 0.182 0.855446
## monthDecember 1.895632 2.120055 0.894 0.371353
## monthFebruary 11.327658 2.035980 5.564 2.99e-08 ***
## monthJanuary 10.715976 2.011146 5.328 1.10e-07 ***
## monthJuly 1.187310 2.006022 0.592 0.554002
## monthJune 3.963759 2.025091 1.957 0.050448 .
## monthMarch 2.839990 2.006877 1.415 0.157186
## monthMay 0.134669 2.003592 0.067 0.946418
## monthNovember 1.278942 2.110297 0.606 0.544551
## monthOctober -1.438794 2.101176 -0.685 0.493576
## monthSeptember -0.547946 2.071334 -0.265 0.791393
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.82 on 2007 degrees of freedom
## Multiple R-squared: 0.2591, Adjusted R-squared: 0.251
## F-statistic: 31.91 on 22 and 2007 DF, p-value: < 2.2e-16Table 4 Report:
We replicated a regression analysis on both for predicting the standard Google Trends “Reporting” data and the adjusted Google Trends “Reporting” data (adj = monthly search / mean weekly search). We found that by using the adjusted data, we got similar outcomes, particularly for the “Trump_admin” feature (18.634301), but the intercept seems to be different. When running on the standard data, the intercept is closer to the reported table, but the “Trump_admin” feature is drastically lower (6.251847). However, the p-values replicated are similar, so there does seem to be a correlation between the Trump administration and the search data.
Extension:
Question 1:
The paper uses “Trump’s Presidential Term” as an indicator for “government support for deportation.” The analysis from this decision supports Hypothesis 1 of this paper: “People will have more interest in reporting immigrants when they believe the government supports deportation.” However, perhaps a better metric to use would be the passage of specific immigration based policies. By identifying positive and negative key immigration policies passed and the exact day of passage, we can explore the reaction of the “reporting”, “welfare”, and “crime” search data.
Question 2:
Investigating further for Hypothesis 1: “H1: People will have more interest in reporting immigrants when they believe the government supports deportation.” Table 5 supported H1 by stating that “Regarding Trump’s campaign cues, the regressions did not show a significant positive effect of Trump’s immigration coverage on anti-immigrant searches during the campaign. However, the cues about Trump’s presidential immigration policy positively and significantly affected reporting searches. Both of these findings provide further support for Hypothesis 1.”
But we thought that not just because of government supports, there is an increase in interest in reporting immigrants, but also it is because there is more news activity during Trump presidential. Therefore we thought it is worth to explore further by adjusting report search by the total number of news streamed. (#time interval we chose “monthly adjustment”)
Replication of Table 5 using Original Topic Model and Original Dataset
#Importing Trump segments and monthly adjustment
trump_segs <- document_topics %>%
select(date, trump) %>%
mutate( ymd = ymd(format(ymd(date), "%Y-%m-%d"))) %>%
group_by(ymd) %>%
summarize( trump = sum(trump), count = n()) %>% mutate(ym = format(ymd, "%Y-%m")) %>% group_by(ym) %>% mutate( max = cumsum(count), max = max(max))
table_5 <- inner_join(table_4, trump_segs) %>% mutate(news_adjust = search_adj / max)## [1] "Replicating Table 5"##
## Call:
## lm(formula = search_adj ~ segment + trump + Trump_admin + Trump_admin *
## trump + crime + welfare + ymd + weekday + month, data = table_5)
##
## Residuals:
## Min 1Q Median 3Q Max
## -49.145 -11.762 -2.061 8.958 139.752
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 91.1603546 24.8416056 3.670 0.000249 ***
## segment 0.0008597 0.0322341 0.027 0.978725
## trump -0.0286036 0.0555613 -0.515 0.606742
## Trump_adminTRUE 15.9952046 1.8499136 8.646 < 2e-16 ***
## crime 0.8680465 0.2674561 3.246 0.001191 **
## welfare 2.0377912 0.6873738 2.965 0.003066 **
## ymd -0.0033704 0.0014955 -2.254 0.024324 *
## weekdayMonday 0.5685343 1.5607108 0.364 0.715688
## weekdaySaturday -3.9442121 1.5625361 -2.524 0.011672 *
## weekdaySunday -5.4298245 1.5569654 -3.487 0.000498 ***
## weekdayThursday 2.6148097 1.5561924 1.680 0.093062 .
## weekdayTuesday -1.3198180 1.5547484 -0.849 0.396041
## weekdayWednesday 3.0430072 1.5535500 1.959 0.050281 .
## monthAugust 1.4808775 2.0179881 0.734 0.463133
## monthDecember 2.1738161 2.1136517 1.028 0.303855
## monthFebruary 11.2646437 2.0277062 5.555 3.14e-08 ***
## monthJanuary 10.4765642 2.0037581 5.228 1.89e-07 ***
## monthJuly 1.9820603 2.0088748 0.987 0.323932
## monthJune 3.9985449 2.0168297 1.983 0.047550 *
## monthMarch 3.2650911 2.0011768 1.632 0.102924
## monthMay 0.6076047 1.9984211 0.304 0.761127
## monthNovember 2.1050647 2.1148171 0.995 0.319667
## monthOctober -1.1940365 2.0935285 -0.570 0.568507
## monthSeptember 0.2375459 2.0712677 0.115 0.908705
## trump:Trump_adminTRUE 0.2655726 0.0618434 4.294 1.84e-05 ***
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 18.74 on 2005 degrees of freedom
## Multiple R-squared: 0.2659, Adjusted R-squared: 0.2571
## F-statistic: 30.26 on 24 and 2005 DF, p-value: < 2.2e-16## [1] "Adjusted Report Data by total number of segmens in a month"##
## Call:
## lm(formula = news_adjust ~ segment + trump + Trump_admin + Trump_admin *
## trump + crime + welfare + ymd + weekday + month, data = table_5)
##
## Residuals:
## Min 1Q Median 3Q Max
## -0.13917 -0.03848 -0.00955 0.02350 0.36383
##
## Coefficients:
## Estimate Std. Error t value Pr(>|t|)
## (Intercept) 1.495e+00 7.882e-02 18.962 < 2e-16 ***
## segment -5.257e-04 1.023e-04 -5.140 3.01e-07 ***
## trump 2.116e-04 1.763e-04 1.201 0.23007
## Trump_adminTRUE 5.544e-02 5.870e-03 9.445 < 2e-16 ***
## crime 4.918e-04 8.486e-04 0.580 0.56226
## welfare 2.773e-03 2.181e-03 1.271 0.20379
## ymd -8.287e-05 4.745e-06 -17.464 < 2e-16 ***
## weekdayMonday 2.784e-03 4.952e-03 0.562 0.57406
## weekdaySaturday -8.386e-03 4.958e-03 -1.692 0.09088 .
## weekdaySunday -9.050e-03 4.940e-03 -1.832 0.06710 .
## weekdayThursday 5.514e-03 4.938e-03 1.117 0.26425
## weekdayTuesday 2.611e-03 4.933e-03 0.529 0.59664
## weekdayWednesday 8.408e-03 4.929e-03 1.706 0.08823 .
## monthAugust -3.671e-02 6.403e-03 -5.733 1.14e-08 ***
## monthDecember -2.001e-02 6.706e-03 -2.983 0.00289 **
## monthFebruary -3.008e-02 6.434e-03 -4.676 3.12e-06 ***
## monthJanuary -2.981e-02 6.358e-03 -4.688 2.94e-06 ***
## monthJuly -3.942e-02 6.374e-03 -6.184 7.55e-10 ***
## monthJune -2.574e-02 6.399e-03 -4.022 5.99e-05 ***
## monthMarch 1.257e-02 6.350e-03 1.980 0.04780 *
## monthMay 1.593e-02 6.341e-03 2.513 0.01206 *
## monthNovember -3.963e-02 6.710e-03 -5.905 4.12e-09 ***
## monthOctober 1.911e-03 6.643e-03 0.288 0.77362
## monthSeptember 9.896e-03 6.572e-03 1.506 0.13227
## trump:Trump_adminTRUE 3.864e-04 1.962e-04 1.969 0.04907 *
## ---
## Signif. codes: 0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
##
## Residual standard error: 0.05946 on 2005 degrees of freedom
## Multiple R-squared: 0.3066, Adjusted R-squared: 0.2983
## F-statistic: 36.93 on 24 and 2005 DF, p-value: < 2.2e-16Report:
After adjusted search of report by total numbers of news in a month, we observed that p value for interaction of Trump News and Trump Admin increased to 0.01. However, it shows different result that mentioned in the notes of Table 5 : ” The interaction between Trump News and Trump Admin, while significant,does not weaken the relationship between the crime/welfare news coverage and anti-immigrant searches.” After adjustment, the interaction between Trump News and Trump Admin that does weaken the relationship between the crime/welfare news coverage and anti-immigrant searches. (Comparing regression results for welfare and crime variable that represent crime and welfare news coverage)