Experiment 1: Morocco as European

Experiment 2: Mexico as North American

Experiment 3: Turkey as European

Experiment 4: Czechia

Morocco Location Pins Placed Further North

library(coin) # For permutation tests

## Loading required package: survival

load("Morocco_Data") # Experiment 1: Morocco as European

sum(!is.na(spdata3$Y.Pin)) # Only 193 of 288 responses drew maps. 

## [1] 193

tapply(spdata3$Y.Pin, spdata3$treatment, mean, na.rm = TRUE) # This is showing the mean aggregated difference on the location of the pin dropped (to locate Morocco) on the y axis between the control group (Treatment == 0) and the treatment group (Treatment == 1). It is measured on a 1000 point scale, with (counter-intuitively) 1000 on the bottom of the scale and 0 on the top. 

##         0         1 
## 0.3663314 0.3273980

independence_test(Y.Pin~factor(treatment), 
                  data = spdata3, distribution = "approximate") # From the coin package which does permutation testing. Default number of shuffles is 10,000.

## 
## 	Approximative General Independence Test
## 
## data:  Y.Pin by factor(treatment) (0, 1)
## Z = 3.7684, p-value < 1e-04
## alternative hypothesis: two.sided

Morocco Regions Drawn Further North

tapply(spdata3$Y_Point_Middle, spdata3$treatment, mean, na.rm = TRUE) # This is showing the difference in the mean calculated centroid on the y axis between the control group (Treatment == 0) and the treatment group (Treatment == 1). It is measured on a 1000 point scale, with (counter-intuitively) 1000 on the bottom of the scale and 0 on the top. 

##        0        1 
## 376.1782 340.4565

independence_test(Y_Point_Middle~factor(treatment), 
                  data = spdata3, distribution = "approximate") 

## 
## 	Approximative General Independence Test
## 
## data:  Y_Point_Middle by factor(treatment) (0, 1)
## Z = 3.1729, p-value = 0.0013
## alternative hypothesis: two.sided

Morocco Label Changes

# spdata2$region <- trimws(spdata2$Label) # Getting rid of spaces in participants open-ended responses about the region Morocco belongs to. 

# Recategorizing each open-ended response:

# spdata2$region[spdata2$region=="Africa???"|spdata2$region=="africa"] <- "Africa"

# spdata2$region[spdata2$region=="Eastern Africa"] <- "East Africa"
 
# spdata2$region[spdata2$region=="europe"|spdata2$region=="European"] <- "Europe"

# spdata2$region[spdata2$region=="The Maghreb"] <- "Maghreb"
 
# spdata2$region[spdata2$region=="Bordering the Mediterranean"|spdata2$region=="mideterranian"|spdata2$region=="mediterranean"|spdata2$region=="Meditteranean"|spdata2$region=="mediterranian"|spdata2$region=="Southern Europe/North Africa"] <- "Mediterranean"

# spdata2$region[spdata2$region=="NOrther Africa/Middle East"|spdata2$region=="Northern Africa (and the Middle East to be more broad)"|spdata2$region=="Northern Africa/Middle East"|spdata2$region=="Middle East/North Africa"|spdata2$region=="Middle East and North Africa"|spdata2$region=="North Africa / Arabian Peninsula"|spdata2$region=="Northern Africa/ Middle East"] <- "MENA"
 
# spdata2$region[spdata2$region=="middle east"] <- "Middle East"
 
# spdata2$region[spdata2$region=="Moracco"|spdata2$region=="moroccan region"|spdata2$region=="Morocco?"|spdata2$region=="Morrocco"|spdata2$region=="morocco"|spdata2$region=="Casablanca-Settat"|spdata2$region=="maracoo"|spdata2$region=="Marocco"|spdata2$region=="Maroon"] <- "Morocco"
 
# spdata2$region[spdata2$region=="north africa"|spdata2$region=="Sahara"|spdata2$region=="northern africa"|spdata2$region=="North African"|spdata2$region=="Northern Africa"|spdata2$region=="northeast africa"|spdata2$region=="NorthWest Africa"|spdata2$region=="North West Africa"|spdata2$region=="Northwestern Africa"|spdata2$region=="north west coast of africa"|spdata2$region=="Northeast Africa"|spdata2$region=="North African"|spdata2$region=="Northwest Africa"] <- "North Africa"

# spdata2$region[spdata2$region=="Central America"|spdata2$region=="Latin America"|spdata2$region=="Asia"|spdata2$region=="SE Asia"|spdata2$region=="Near Algeria"|spdata2$region=="S. Mexico"|spdata2$region=="Eurafrican Morocco"|spdata2$region=="south america"|spdata2$region=="Europe/Africa"|spdata2$region=="Morocco-Spain Border"|spdata2$region=="northern africa/europe/middle east"] <- "Other"
 
# spdata2$region[spdata2$region=="Spain"|spdata2$region=="Spain region"|spdata2$region=="Southwestern Europe"|spdata2$region=="Southeast Europe"] <- "Southern Europe"
 
# spdata2$region[spdata2$region=="idk"|spdata2$region=="i dont know"|spdata2$region=="not sure"|spdata2$region=="I'm not sure"|spdata2$region=="id"] <- "Unsure"

# spdata2$region[spdata2$region=="west africa"|spdata2$region=="Western Africa"] <- "West Africa"
 
# spdata2$region[spdata2$region==""|spdata2$region=="I don't understand the question"] <- NA

#spdata2$change_region <- ifelse(spdata2$region== "Europe"|spdata2$region== "Mediterranean"|spdata2$region== "Southern Europe", change_region1+1, change_region1+0)

#table(spdata2$change_region[spdata2$treatment==0]) # 0 put Europe in control group
#table(spdata2$change_region[spdata2$treatment==1]) # 23 put Europe in treatment group

#independence_test(change_region~factor(treatment), 
              #    data = spdata2, distribution = "approximate") # Based on 10k resamples, p = 0.0001

Turkey Regions Drawn

load("Turkey_Data") # Survey 3: Turkey as European

sum(!is.na(Coordinate_Data$`pin/0`)) # Only 331 of 587 responses drew maps. 

## [1] 331

# Checking the area of each polygon drawn (i.e. the regions):

# Define coordinate column names
x_cols_area <- paste0("points/", 0:51, "/x")
y_cols_area <- paste0("points/", 0:51, "/y")

# Shoelace area function
shoelace_area <- function(x, y) {
  x <- c(x, x[1])
  y <- c(y, y[1])
  0.5 * abs(sum(x[-1] * y[-length(y)] - x[-length(x)] * y[-1]))
}

# Calculate polygon area per row, handling NAs
calc_polygon_area <- function(row, x_cols, y_cols) {
  x <- as.numeric(row[x_cols])
  y <- as.numeric(row[y_cols])
  valid <- !(is.na(x) | is.na(y))
  x <- x[valid]
  y <- y[valid]
  if(length(x) >= 3) {
    shoelace_area(x, y)
  } else {
    NA_real_
  }
}

# Apply function to all rows
Coordinate_Data$Polygon_Area <- apply(
  Coordinate_Data,
  1,
  function(r) calc_polygon_area(r, x_cols_area, y_cols_area)
)

# Summary of results
summary(Coordinate_Data$Polygon_Area)

##      Min.   1st Qu.    Median      Mean   3rd Qu.      Max.      NA's 
##     67.24   1262.38   2374.70   9002.61  12996.32 169601.74       259

t.test(Polygon_Area~ Treatment, data = Coordinate_Data) # difference in size is 1300, p = 0.42 Polygons got larger in treatment group. That's about a 16 percent increase in average polygon size. 

## 
## 	Welch Two Sample t-test
## 
## data:  Polygon_Area by Treatment
## t = -0.80311, df = 320.79, p-value = 0.4225
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -4610.361  1937.465
## sample estimates:
## mean in group 0 mean in group 1 
##        8346.613        9683.061

Turkey Label Changes

library(dplyr)

## 
## Attaching package: 'dplyr'

## The following objects are masked from 'package:stats':
## 
##     filter, lag

## The following objects are masked from 'package:base':
## 
##     intersect, setdiff, setequal, union

# Checking Asian labels. Doing so by simply including all labels where they actually use the word "asia."  

Coordinate_Data <- Coordinate_Data %>%
  mutate(
    Asia = case_when(
      is.na(label) ~ NA_real_,
      grepl("asia", label, ignore.case = TRUE) ~ 1,
      TRUE ~ 0
    )
  )

# Which provides us with these Asian words:

Coordinate_Data %>%
  filter(Asia == 1) %>%
  select(label) %>%
  print()

## # A tibble: 33 × 1
##    label                                            
##    <chr>                                            
##  1 Eurasia                                          
##  2 Europe and Asia                                  
##  3 Asia Minor                                       
##  4 Asia                                             
##  5 Europe & Asia                                    
##  6 Far East Europe/Northern Middle East/Western Asia
##  7 Eurasia                                          
##  8 Anatolia and Caucasia                            
##  9 euroasia                                         
## 10 Asia                                             
## # ℹ 23 more rows

# And these non-Asian words:

Coordinate_Data %>%
  filter(Asia == 0) %>%
  select(label) %>%
  print()

## # A tibble: 298 × 1
##    label             
##    <chr>             
##  1 Turkey            
##  2 turkey            
##  3 Turkey or Anatolia
##  4 anatolia          
##  5 Western Europe    
##  6 Europe            
##  7 Middle East       
##  8 turkey            
##  9 ?                 
## 10 not sure          
## # ℹ 288 more rows

independence_test(Asia~factor(Treatment), data = Coordinate_Data, 
distribution = "approximate") # 0.03  

## 
## 	Approximative General Independence Test
## 
## data:  Asia by factor(Treatment) (0, 1)
## Z = 2.2904, p-value = 0.0238
## alternative hypothesis: two.sided

Turkey Regional Word Association Change

# Open response words coded as having to do with Asia or the Middle East:

Coordinate_Data$Middle_East_or_Asia_Words <- ifelse(
  Coordinate_Data$oneword %in% c(
    "oil", "Kurds", "muslim", "mena", "Middle", "middle eastern",
    "Muslim", "Arabs", "Islam", "Arab", "arab", "desert", "Asia",
    "MENA", "east", "East", "Desert", "Religious", "MiddleEast",
    "Desert/Different", "middle", "Syria", "middle-eastern",
    "islam", "middle east", "Middle Eastern", "Eastern World",
    "arabs", "Middle East", "Arabic", "asia", "Oil"
  ),
  1,
  0
)

independence_test(
  Middle_East_or_Asia_Words ~ factor(Treatment),
  data = Coordinate_Data,
  distribution = approximate(nresample = 10000)
) # p = 0.07

## 
## 	Approximative General Independence Test
## 
## data:  Middle_East_or_Asia_Words by factor(Treatment) (0, 1)
## Z = 1.8516, p-value = 0.0748
## alternative hypothesis: two.sided

The Importance of Regional Identity

load("Mexico_Data") # Survey 2: Mexico as North American

# How important is a North American identity to you? 

# spdata$North_Am_ID_Imp_Num <- as.numeric(as.factor(spdata$North_Am_ID_Imp))
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==1] <- NA  # Did not answer
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==2] <- 0.75 # Important
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==3] <- 0.5 # Neither important nor unimportant
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==4] <- 0.25 # Not important
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==5] <- 0 # Totally meaningless
# spdata$North_Am_ID_Imp_Num[spdata$North_Am_ID_Imp_Num==6] <- 1  # Very important

tapply(spdata$North_Am_ID_Imp_Num, spdata$Treatment, mean, na.rm = TRUE) # The difference between the treatment and control groups is 10% on the assessed importance of North American identity. 

##         0         1 
## 0.4355469 0.5354167

independence_test(
  North_Am_ID_Imp_Num ~ factor(Treatment),
  data = spdata,
  distribution = approximate(nresample = 10000)
) # p = 0.006

## 
## 	Approximative General Independence Test
## 
## data:  North_Am_ID_Imp_Num by factor(Treatment) (0, 1)
## Z = -2.7409, p-value = 0.0053
## alternative hypothesis: two.sided

North America Identity Meaning

library(knitr) # For kable table to display
kable(spdata[, c("North_ID_Mean", "Meaning_Coding")],
      row.names = FALSE,
      col.names = NULL) # This shows the open ended responses to what a North American identity means to respondents and how I coded them.

Difference in Region Meaning by Treatment Exposure

## 
## ----------------------------------
## Variable: Liberal_ID 
## Mean (Treatment = 0): 0.05 
## Mean (Treatment = 1): 0.1068702 
## 
## 	Welch Two Sample t-test
## 
## data:  Liberal_ID by Treatment
## t = -1.7338, df = 232.14, p-value = 0.08429
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.121497250  0.007756791
## sample estimates:
## mean in group 0 mean in group 1 
##       0.0500000       0.1068702 
## 
## 
## ----------------------------------
## Variable: Western_ID 
## Mean (Treatment = 0): 0.007142857 
## Mean (Treatment = 1): 0.03816794 
## 
## 	Welch Two Sample t-test
## 
## data:  Western_ID by Treatment
## t = -1.6991, df = 175.85, p-value = 0.09107
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.067061290  0.005011126
## sample estimates:
## mean in group 0 mean in group 1 
##     0.007142857     0.038167939 
## 
## 
## ----------------------------------
## Variable: Other_Positive_ID 
## Mean (Treatment = 0): 0.01428571 
## Mean (Treatment = 1): 0.03816794 
## 
## 	Welch Two Sample t-test
## 
## data:  Other_Positive_ID by Treatment
## t = -1.2192, df = 214.22, p-value = 0.2241
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.06249265  0.01472820
## sample estimates:
## mean in group 0 mean in group 1 
##      0.01428571      0.03816794 
## 
## 
## ----------------------------------
## Variable: Wealth_and_Opportunity_ID 
## Mean (Treatment = 0): 0.07857143 
## Mean (Treatment = 1): 0.07633588 
## 
## 	Welch Two Sample t-test
## 
## data:  Wealth_and_Opportunity_ID by Treatment
## t = 0.06856, df = 268.23, p-value = 0.9454
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.06196287  0.06643397
## sample estimates:
## mean in group 0 mean in group 1 
##      0.07857143      0.07633588 
## 
## 
## ----------------------------------
## Variable: Geographic_ID 
## Mean (Treatment = 0): 0.5714286 
## Mean (Treatment = 1): 0.5725191 
## 
## 	Welch Two Sample t-test
## 
## data:  Geographic_ID by Treatment
## t = -0.018064, df = 267.81, p-value = 0.9856
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.1199497  0.1177687
## sample estimates:
## mean in group 0 mean in group 1 
##       0.5714286       0.5725191 
## 
## 
## ----------------------------------
## Variable: Colonial_North_American_ID 
## Mean (Treatment = 0): 0.04285714 
## Mean (Treatment = 1): 0.02290076 
## 
## 	Welch Two Sample t-test
## 
## data:  Colonial_North_American_ID by Treatment
## t = 0.92324, df = 255.5, p-value = 0.3568
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.02261116  0.06252391
## sample estimates:
## mean in group 0 mean in group 1 
##      0.04285714      0.02290076 
## 
## 
## ----------------------------------
## Variable: Negative_ID 
## Mean (Treatment = 0): 0.03571429 
## Mean (Treatment = 1): 0.02290076 
## 
## 	Welch Two Sample t-test
## 
## data:  Negative_ID by Treatment
## t = 0.62532, df = 263.33, p-value = 0.5323
## alternative hypothesis: true difference in means between group 0 and group 1 is not equal to 0
## 95 percent confidence interval:
##  -0.02753380  0.05316084
## sample estimates:
## mean in group 0 mean in group 1 
##      0.03571429      0.02290076

The Role of Geographic Knowledge

# The interaction of Treatment and Geographic Knowledge (called Familiarity) on the central point of regions drawn for Morocco's region (measured on a 0 to 1000 scale)

summary(lm(Y_Point_Middle ~ treatment * Familiarity, data = spdata3))

## 
## Call:
## lm(formula = Y_Point_Middle ~ treatment * Familiarity, data = spdata3)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -223.36  -36.36   -8.60   22.55  310.72 
## 
## Coefficients:
##                       Estimate Std. Error t value Pr(>|t|)    
## (Intercept)             397.28      12.60  31.532  < 2e-16 ***
## treatment               -66.68      17.63  -3.782 0.000208 ***
## Familiarity             -22.92      10.98  -2.087 0.038192 *  
## treatment:Familiarity    33.34      14.87   2.242 0.026139 *  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 75.56 on 189 degrees of freedom
##   (95 observations deleted due to missingness)
## Multiple R-squared:  0.07893,	Adjusted R-squared:  0.06431 
## F-statistic: 5.399 on 3 and 189 DF,  p-value: 0.001383

# Whether respondents changed the label of their regions for Morocco regions. The coefficients can be interpreted as differences in probabilities.

# summary(lm(change_region ~ treatment * Familiarity, data = spdata2))
# 
# Coefficients:
#                        Estimate Std. Error t value Pr(>|t|)    
# (Intercept)            9.29e-16   4.94e-02    0.00    1.000    
# treatment              3.63e-01   6.91e-02    5.25  4.1e-07 ***
# Familiarity           -4.35e-17   4.30e-02    0.00    1.000    
# treatment:Familiarity -1.19e-01   5.82e-02   -2.04    0.042 *  
# ---
# Signif. codes:  0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1
# 
# Residual standard error: 0.296 on 188 degrees of freedom
#   (96 observations deleted due to missingness)
# Multiple R-squared:  0.188,	Adjusted R-squared:  0.175 
# F-statistic: 14.5 on 3 and 188 DF,  p-value: 1.6e-08

# Difference in Turkey median east-west value based on treatment and geographic knowledge. Measured on a 0 to 1000 scale. 

summary(lm(Median_X~Treatment * Geographic_Knowledge, data = Coordinate_Data))

## 
## Call:
## lm(formula = Median_X ~ Treatment * Geographic_Knowledge, data = Coordinate_Data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -130.921  -17.985   -4.496    6.726  235.248 
## 
## Coefficients:
##                                Estimate Std. Error t value Pr(>|t|)    
## (Intercept)                      554.82      13.28  41.774  < 2e-16 ***
## Treatment                        -34.40      19.76  -1.741  0.08269 .  
## Geographic_Knowledge             -49.76      16.21  -3.071  0.00233 ** 
## Treatment:Geographic_Knowledge    43.99      24.11   1.825  0.06902 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 47.69 on 300 degrees of freedom
##   (283 observations deleted due to missingness)
## Multiple R-squared:  0.0308,	Adjusted R-squared:  0.02111 
## F-statistic: 3.178 on 3 and 300 DF,  p-value: 0.02442

summary(lm(Middle_East_or_Asia_Words~Treatment * Geographic_Knowledge, data = Coordinate_Data))

## 
## Call:
## lm(formula = Middle_East_or_Asia_Words ~ Treatment * Geographic_Knowledge, 
##     data = Coordinate_Data)
## 
## Residuals:
##      Min       1Q   Median       3Q      Max 
## -0.20129 -0.15173 -0.12695 -0.09597  0.94550 
## 
## Coefficients:
##                                Estimate Std. Error t value Pr(>|t|)  
## (Intercept)                     0.05261    0.06524   0.806   0.4204  
## Treatment                      -0.03958    0.10223  -0.387   0.6988  
## Geographic_Knowledge            0.14868    0.08227   1.807   0.0713 .
## Treatment:Geographic_Knowledge -0.02428    0.12687  -0.191   0.8483  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 0.3429 on 520 degrees of freedom
##   (63 observations deleted due to missingness)
## Multiple R-squared:  0.01535,	Adjusted R-squared:  0.009669 
## F-statistic: 2.702 on 3 and 520 DF,  p-value: 0.04497

Survey 1: Morocco as European

Survey 2: Mexico as North American

Survey 3: Turkey as European

Survey 4: Czechia