Granger Causalities

Beginning with ADF tests, we test for stationary and, in the event of non-stationary, we difference the variables and retest for stationary. Once stationary is achieved, we test for granger causality both ways. We examine the hashtags against ACLED, AMCHA, and ADL data.

FBI

We will first look at FBI-reported antisemitic incidents. which are possibly caused by “NWO” and likely caused by variants of “NWO/new world order/newworldorder.”

adf.test(polarization_before$FBI)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(polarization_before$nwo_comb)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$nwo_comb
## Dickey-Fuller = -2.9386, Lag order = 7, p-value = 0.1808
## alternative hypothesis: stationary
VARselect(polarization_after2$FBI + polarization_after2$nwo_comb)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      6      2      2      6 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     12.84414     12.83058     12.82795     12.82351     12.82487
## HQ(n)      12.85293     12.84377     12.84553     12.84550     12.85125
## SC(n)      12.86623     12.86372     12.87213     12.87874     12.89115
## FPE(n) 378562.29043 373464.81214 372482.90052 370835.51305 371340.69959
##                   6            7            8            9           10
## AIC(n)     12.82223     12.82609     12.83003     12.83291     12.83809
## HQ(n)      12.85301     12.86127     12.86961     12.87688     12.88646
## SC(n)      12.89956     12.91446     12.92945     12.94337     12.95960
## FPE(n) 370363.02988 371794.55999 373264.30907 374343.12504 376289.29740
grangertest(polarization_before$FBI , polarization_before$nwo_comb, order=6)
## Granger causality test
## 
## Model 1: polarization_before$nwo_comb ~ Lags(polarization_before$nwo_comb, 1:6) + Lags(polarization_before$FBI, 1:6)
## Model 2: polarization_before$nwo_comb ~ Lags(polarization_before$nwo_comb, 1:6)
##   Res.Df Df      F Pr(>F)
## 1    353                 
## 2    359 -6 1.2902 0.2609
grangertest(polarization_before$FBI ~ polarization_before$nwo_comb, order=6)
## Granger causality test
## 
## Model 1: polarization_before$FBI ~ Lags(polarization_before$FBI, 1:6) + Lags(polarization_before$nwo_comb, 1:6)
## Model 2: polarization_before$FBI ~ Lags(polarization_before$FBI, 1:6)
##   Res.Df Df      F    Pr(>F)    
## 1    353                        
## 2    359 -6 3.9327 0.0008068 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
adf.test(polarization_before$FBI + polarization_before$sum_antizionist)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI + polarization_before$sum_antizionist
## Dickey-Fuller = -6.1057, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(polarization_before$FBI + polarization_before$sum_antizionist)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      4      4      3      4 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n) 1.733120e+01 1.726842e+01 1.724976e+01 1.724543e+01 1.724764e+01
## HQ(n)  1.733975e+01 1.728124e+01 1.726686e+01 1.726680e+01 1.727329e+01
## SC(n)  1.735270e+01 1.730067e+01 1.729276e+01 1.729918e+01 1.731215e+01
## FPE(n) 3.363918e+07 3.159203e+07 3.100823e+07 3.087419e+07 3.094271e+07
##                   6            7            8            9           10
## AIC(n) 1.725316e+01 1.724725e+01 1.724797e+01 1.725268e+01 1.725807e+01
## HQ(n)  1.728307e+01 1.728144e+01 1.728643e+01 1.729542e+01 1.730508e+01
## SC(n)  1.732841e+01 1.733326e+01 1.734472e+01 1.736018e+01 1.737633e+01
## FPE(n) 3.111384e+07 3.093068e+07 3.095293e+07 3.109926e+07 3.126761e+07

NWO

FBI incidents are also possibly caused by “New World Order” (p=0.09).

adf.test(polarization_before$FBI )
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(polarization_before$new_world)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$new_world
## Dickey-Fuller = -2.6029, Lag order = 7, p-value = 0.3225
## alternative hypothesis: stationary
dif_nwo<- diff(polarization_before$nwo)
dif_new_world<- diff(polarization_before$new_world)
dif_fbi<- diff(polarization_before$FBI)

VARselect(polarization_before$FBI + polarization_before$new_world)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##     10      3      3     10 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    10.15393    10.14123    10.12349    10.12412    10.12631    10.13100
## HQ(n)     10.16248    10.15405    10.14058    10.14549    10.15196    10.16092
## SC(n)     10.17543    10.17348    10.16649    10.17788    10.19082    10.20626
## FPE(n) 25691.83069 25367.59490 24921.62369 24937.42357 24992.17431 25109.65306
##                  7           8           9          10
## AIC(n)    10.11057    10.11529    10.11906    10.10631
## HQ(n)     10.14476    10.15376    10.16180    10.15332
## SC(n)     10.19657    10.21205    10.22657    10.22456
## FPE(n) 24601.87280 24718.42209 24811.81430 24497.53597
grangertest(dif_new_world , dif_fbi, order=10)
## Granger causality test
## 
## Model 1: dif_fbi ~ Lags(dif_fbi, 1:10) + Lags(dif_new_world, 1:10)
## Model 2: dif_fbi ~ Lags(dif_fbi, 1:10)
##   Res.Df  Df      F Pr(>F)
## 1    340                  
## 2    350 -10 1.3865 0.1848
grangertest(dif_new_world ~ dif_fbi, order=10)
## Granger causality test
## 
## Model 1: dif_new_world ~ Lags(dif_new_world, 1:10) + Lags(dif_fbi, 1:10)
## Model 2: dif_new_world ~ Lags(dif_new_world, 1:10)
##   Res.Df  Df      F Pr(>F)
## 1    340                  
## 2    350 -10 0.6328 0.7857
adf.test(dif_fbi)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_fbi
## Dickey-Fuller = -11.038, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(dif_nwo)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_nwo
## Dickey-Fuller = -9.261, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(dif_fbi + dif_nwo)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_fbi + dif_nwo
## Dickey-Fuller = -9.2597, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_fbi + dif_nwo)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##     10      2      2     10 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     12.57494     12.44291     12.44117     12.44098     12.43972
## HQ(n)      12.58351     12.45576     12.45831     12.46239     12.46542
## SC(n)      12.59649     12.47523     12.48426     12.49484     12.50436
## FPE(n) 289220.21758 253448.00040 253007.59927 252958.07040 252641.80995
##                   6            7            8            9           10
## AIC(n)     12.44062     12.44219     12.44333     12.44533     12.43706
## HQ(n)      12.47060     12.47645     12.48187     12.48816     12.48418
## SC(n)      12.51603     12.52837     12.54028     12.55305     12.55556
## FPE(n) 252869.61027 253266.80371 253555.12206 254064.32532 251974.25522
grangertest(dif_fbi , dif_nwo, order=10)
## Granger causality test
## 
## Model 1: dif_nwo ~ Lags(dif_nwo, 1:10) + Lags(dif_fbi, 1:10)
## Model 2: dif_nwo ~ Lags(dif_nwo, 1:10)
##   Res.Df  Df      F Pr(>F)
## 1    340                  
## 2    350 -10 1.0652 0.3886
grangertest(dif_fbi ~ dif_nwo, order=10)
## Granger causality test
## 
## Model 1: dif_fbi ~ Lags(dif_fbi, 1:10) + Lags(dif_nwo, 1:10)
## Model 2: dif_fbi ~ Lags(dif_fbi, 1:10)
##   Res.Df  Df      F  Pr(>F)  
## 1    340                     
## 2    350 -10 1.6502 0.09131 .
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
adf.test(polarization_before$FBI + polarization_before$newworld...2)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI + polarization_before$newworld...2
## Dickey-Fuller = -2.4162, Lag order = 7, p-value = 0.4013
## alternative hypothesis: stationary
adf.test(polarization_before$FBI)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(polarization_before$newworld...2)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$newworld...2
## Dickey-Fuller = -2.3798, Lag order = 7, p-value = 0.4167
## alternative hypothesis: stationary
dif_new_w<- diff(polarization_before$newworld...2)
VARselect(dif_new_w + dif_fbi)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      5      5      5 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    7.684349    7.528097    7.500826    7.503625    7.472962    7.477817
## HQ(n)     7.692914    7.540946    7.517957    7.525040    7.498659    7.507797
## SC(n)     7.705894    7.560415    7.543916    7.557488    7.537597    7.553224
## FPE(n) 2174.053453 1859.564914 1809.537744 1814.612030 1759.816921 1768.384000
##                  7           8           9          10
## AIC(n)    7.482960    7.487873    7.493060    7.497958
## HQ(n)     7.517224    7.526419    7.535889    7.545070
## SC(n)     7.569140    7.584826    7.600785    7.616456
## FPE(n) 1777.507676 1786.267154 1795.562816 1804.388254
grangertest(dif_new_w , dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_fbi ~ Lags(dif_fbi, 1:5) + Lags(dif_new_w, 1:5)
## Model 2: dif_fbi ~ Lags(dif_fbi, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.1593 0.9771
grangertest(dif_new_w ~ dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_new_w ~ Lags(dif_new_w, 1:5) + Lags(dif_fbi, 1:5)
## Model 2: dif_new_w ~ Lags(dif_new_w, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.5372 0.7481

Soros

FBI incidents unlikely to be caused by “Soros.”

adf.test(polarization_before$FBI)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(polarization_before$soros)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$soros
## Dickey-Fuller = -4.7583, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(polarization_before$FBI + polarization_before$soros)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      3      3      3      3 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n) 1.577396e+01 1.570019e+01 1.568729e+01 1.569228e+01 1.569781e+01
## HQ(n)  1.578251e+01 1.571301e+01 1.570439e+01 1.571365e+01 1.572345e+01
## SC(n)  1.579546e+01 1.573244e+01 1.573030e+01 1.574604e+01 1.576231e+01
## FPE(n) 7.088325e+06 6.584257e+06 6.499881e+06 6.532409e+06 6.568602e+06
##                   6            7            8            9           10
## AIC(n) 1.569820e+01 1.570274e+01 1.570619e+01 1.571171e+01 1.571705e+01
## HQ(n)  1.572811e+01 1.573693e+01 1.574465e+01 1.575445e+01 1.576406e+01
## SC(n)  1.577345e+01 1.578874e+01 1.580294e+01 1.581921e+01 1.583531e+01
## FPE(n) 6.571177e+06 6.601111e+06 6.623922e+06 6.660625e+06 6.696347e+06
grangertest(polarization_before$soros , polarization_before$FBI, order=3)
## Granger causality test
## 
## Model 1: polarization_before$FBI ~ Lags(polarization_before$FBI, 1:3) + Lags(polarization_before$soros, 1:3)
## Model 2: polarization_before$FBI ~ Lags(polarization_before$FBI, 1:3)
##   Res.Df Df      F Pr(>F)
## 1    362                 
## 2    365 -3 1.8233 0.1425
grangertest(polarization_before$soros ~ polarization_before$FBI, order=3)
## Granger causality test
## 
## Model 1: polarization_before$soros ~ Lags(polarization_before$soros, 1:3) + Lags(polarization_before$FBI, 1:3)
## Model 2: polarization_before$soros ~ Lags(polarization_before$soros, 1:3)
##   Res.Df Df      F Pr(>F)
## 1    362                 
## 2    365 -3 0.4474 0.7193

Globalist

No correlations with “Globalist”

adf.test(polarization_before$globalist)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$globalist
## Dickey-Fuller = -2.8087, Lag order = 7, p-value = 0.2356
## alternative hypothesis: stationary
adf.test(polarization_before$FBI)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
dif_wwg<- diff(polarization_before$wwg1wga)
dif_glob<- diff(polarization_before$globalist)

VARselect(dif_glob + dif_fbi)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      5      2      5 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    10.80379    10.75571    10.75771    10.74928    10.73941    10.74311
## HQ(n)     10.81236    10.76856    10.77484    10.77069    10.76510    10.77309
## SC(n)     10.82534    10.78803    10.80080    10.80314    10.80404    10.81852
## FPE(n) 49207.06045 46897.11072 46991.12935 46596.43977 46138.81570 46310.07473
##                  7           8           9          10
## AIC(n)    10.74850    10.75256    10.75759    10.75522
## HQ(n)     10.78276    10.79111    10.80042    10.80234
## SC(n)     10.83468    10.84951    10.86532    10.87372
## FPE(n) 46560.39608 46750.10411 46986.08362 46875.18834
grangertest(dif_glob , dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_fbi ~ Lags(dif_fbi, 1:5) + Lags(dif_glob, 1:5)
## Model 2: dif_fbi ~ Lags(dif_fbi, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.9934 0.4216
grangertest(dif_glob ~ dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_glob ~ Lags(dif_glob, 1:5) + Lags(dif_fbi, 1:5)
## Model 2: dif_glob ~ Lags(dif_glob, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.4278 0.8292

WWG1WGA

“WWG1WGA” is possibly caused by FBI-reported incidents (p = 0.052).

adf.test(polarization_before$FBI)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$FBI
## Dickey-Fuller = -5.303, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test( polarization_before$wwg1wga)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$wwg1wga
## Dickey-Fuller = -2.273, Lag order = 7, p-value = 0.4618
## alternative hypothesis: stationary
dif_wwg<- diff(polarization_before$wwg1wga)

VARselect(dif_wwg + dif_fbi)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      3      3      5 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     13.44666     13.41585     13.39098     13.38966     13.38530
## HQ(n)      13.45522     13.42870     13.40811     13.41107     13.41100
## SC(n)      13.46820     13.44817     13.43407     13.44352     13.44994
## FPE(n) 691527.15941 670549.53396 654077.32829 653214.77203 650375.96121
##                   6            7            8            9           10
## AIC(n)     13.38792     13.39287     13.39033     13.39368     13.39609
## HQ(n)      13.41790     13.42713     13.42887     13.43651     13.44320
## SC(n)      13.46333     13.47905     13.48728     13.50140     13.51458
## FPE(n) 652082.58675 655317.48086 653657.43980 655852.21730 657436.88143
grangertest(dif_wwg , dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_fbi ~ Lags(dif_fbi, 1:5) + Lags(dif_wwg, 1:5)
## Model 2: dif_fbi ~ Lags(dif_fbi, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.0138 0.9999
grangertest(dif_wwg ~ dif_fbi, order=5)
## Granger causality test
## 
## Model 1: dif_wwg ~ Lags(dif_wwg, 1:5) + Lags(dif_fbi, 1:5)
## Model 2: dif_wwg ~ Lags(dif_wwg, 1:5)
##   Res.Df Df      F  Pr(>F)  
## 1    355                    
## 2    360 -5 2.2216 0.05171 .
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Elections (“Stop the Steal”)

Now we will assess “Stop the Steal” relationships with all the hashtags.

NWO

“NWO” likely caused Election incidents.

adf.test(polarization_before$newworld...2)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$newworld...2
## Dickey-Fuller = -2.3798, Lag order = 7, p-value = 0.4167
## alternative hypothesis: stationary
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
dif_newwo <- diff(polarization_before$newworld...2)
dif_elec <- diff(polarization_before$Elections)
adf.test(dif_elec + dif_newwo)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_elec + dif_newwo
## Dickey-Fuller = -6.9007, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_elec + dif_newwo)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      5      5      5 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    7.661250    7.514729    7.489408    7.492542    7.465271    7.470252
## HQ(n)     7.669815    7.527578    7.506540    7.513956    7.490969    7.500232
## SC(n)     7.682795    7.547047    7.532498    7.546404    7.529907    7.545659
## FPE(n) 2124.410785 1834.871010 1788.994096 1794.611036 1746.334801 1755.057072
##                  7           8           9          10
## AIC(n)    7.475220    7.479881    7.485071    7.490330
## HQ(n)     7.509484    7.518427    7.527900    7.537442
## SC(n)     7.561400    7.576833    7.592796    7.608828
## FPE(n) 1763.802944 1772.047589 1781.275902 1790.676800
grangertest(dif_elec , dif_newwo, order=5)
## Granger causality test
## 
## Model 1: dif_newwo ~ Lags(dif_newwo, 1:5) + Lags(dif_elec, 1:5)
## Model 2: dif_newwo ~ Lags(dif_newwo, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 1.7096 0.1316
grangertest(dif_elec ~ dif_newwo, order=5)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:5) + Lags(dif_newwo, 1:5)
## Model 2: dif_elec ~ Lags(dif_elec, 1:5)
##   Res.Df Df      F    Pr(>F)    
## 1    355                        
## 2    360 -5 15.058 2.038e-13 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
adf.test(polarization_before$nwo)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$nwo
## Dickey-Fuller = -3.3771, Lag order = 7, p-value = 0.05815
## alternative hypothesis: stationary
dif_new_world <- diff(polarization_before$nwo)
adf.test(dif_elec + dif_new_world)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_elec + dif_new_world
## Dickey-Fuller = -9.2603, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_elec + dif_new_world)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##     10      2      2     10 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     12.57482     12.44302     12.44124     12.44112     12.43986
## HQ(n)      12.58338     12.45587     12.45837     12.46253     12.46556
## SC(n)      12.59636     12.47534     12.48433     12.49498     12.50450
## FPE(n) 289183.98520 253474.38541 253024.23637 252994.28947 252676.72311
##                   6            7            8            9           10
## AIC(n)     12.44077     12.44228     12.44340     12.44541     12.43734
## HQ(n)      12.47075     12.47654     12.48194     12.48824     12.48445
## SC(n)      12.51618     12.52846     12.54035     12.55314     12.55584
## FPE(n) 252905.94577 253288.93673 253572.71518 254086.27448 252043.80871
grangertest(dif_elec , dif_new_world, order=9)
## Granger causality test
## 
## Model 1: dif_new_world ~ Lags(dif_new_world, 1:9) + Lags(dif_elec, 1:9)
## Model 2: dif_new_world ~ Lags(dif_new_world, 1:9)
##   Res.Df Df      F    Pr(>F)    
## 1    343                        
## 2    352 -9 3.5216 0.0003438 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
grangertest(dif_elec ~ dif_new_world, order=9)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:9) + Lags(dif_new_world, 1:9)
## Model 2: dif_elec ~ Lags(dif_elec, 1:9)
##   Res.Df Df      F Pr(>F)
## 1    343                 
## 2    352 -9 0.5051 0.8707
adf.test(polarization_before$Elections + polarization_before$new_world)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections + polarization_before$new_world
## Dickey-Fuller = -2.5998, Lag order = 7, p-value = 0.3238
## alternative hypothesis: stationary
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
adf.test(polarization_before$new_world)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$new_world
## Dickey-Fuller = -2.6029, Lag order = 7, p-value = 0.3225
## alternative hypothesis: stationary
dif_new_w <- diff(polarization_before$new_world)
adf.test(dif_elec + dif_new_w)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_elec + dif_new_w
## Dickey-Fuller = -8.6453, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_elec + dif_new_w)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      9      6      2      9 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    10.17703    10.14961    10.15325    10.15201    10.15490    10.12733
## HQ(n)     10.18559    10.16245    10.17038    10.17342    10.18060    10.15731
## SC(n)     10.19857    10.18192    10.19634    10.20587    10.21954    10.20273
## FPE(n) 26292.18418 25581.01634 25674.43928 25642.55142 25716.97228 25017.47671
##                  7           8           9          10
## AIC(n)    10.13272    10.13506    10.12633    10.12708
## HQ(n)     10.16698    10.17361    10.16916    10.17419
## SC(n)     10.21890    10.23201    10.23406    10.24558
## FPE(n) 25152.82015 25211.88771 24992.84861 25011.76842
grangertest(dif_elec , dif_new_w, order=5)
## Granger causality test
## 
## Model 1: dif_new_w ~ Lags(dif_new_w, 1:5) + Lags(dif_elec, 1:5)
## Model 2: dif_new_w ~ Lags(dif_new_w, 1:5)
##   Res.Df Df      F  Pr(>F)  
## 1    355                    
## 2    360 -5 2.9456 0.01277 *
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
grangertest(dif_elec ~ dif_new_w, order=5)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:5) + Lags(dif_new_w, 1:5)
## Model 2: dif_elec ~ Lags(dif_elec, 1:5)
##   Res.Df Df     F Pr(>F)
## 1    355                
## 2    360 -5 1.641 0.1484
dif_new_world_com <- diff(polarization_before$nwo_comb)

VARselect(dif_elec + dif_new_world_com)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      2      2      2      2 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     12.84002     12.75536     12.76040     12.76285     12.75810
## HQ(n)      12.84859     12.76821     12.77754     12.78427     12.78380
## SC(n)      12.86156     12.78768     12.80349     12.81671     12.82274
## FPE(n) 377007.14960 346403.99811 348155.43011 349009.16763 347356.28749
##                   6            7            8            9           10
## AIC(n)     12.75889     12.76328     12.76626     12.77135     12.76710
## HQ(n)      12.78887     12.79754     12.80481     12.81418     12.81422
## SC(n)      12.83430     12.84946     12.86321     12.87908     12.88560
## FPE(n) 347630.90663 349159.54898 350204.03934 351992.70214 350502.62301
grangertest(dif_elec , dif_new_world_com, order=7)
## Granger causality test
## 
## Model 1: dif_new_world_com ~ Lags(dif_new_world_com, 1:7) + Lags(dif_elec, 1:7)
## Model 2: dif_new_world_com ~ Lags(dif_new_world_com, 1:7)
##   Res.Df Df      F    Pr(>F)    
## 1    349                        
## 2    356 -7 5.3425 7.982e-06 ***
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
grangertest(dif_elec ~ dif_new_world_com, order=7)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:7) + Lags(dif_new_world_com, 1:7)
## Model 2: dif_elec ~ Lags(dif_elec, 1:7)
##   Res.Df Df     F Pr(>F)
## 1    349                
## 2    356 -7 0.409 0.8966

Soros

There is some evidence to suggest election incidents were caused by and (to a lesser degree) caused the term “Soros.”

adf.test(polarization_before$soros)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$soros
## Dickey-Fuller = -4.7583, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
VARselect(polarization_before$Elections + polarization_before$soros)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      3      3      3      3 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n) 1.577403e+01 1.570025e+01 1.568733e+01 1.569233e+01 1.569785e+01
## HQ(n)  1.578258e+01 1.571307e+01 1.570442e+01 1.571370e+01 1.572349e+01
## SC(n)  1.579554e+01 1.573250e+01 1.573033e+01 1.574608e+01 1.576235e+01
## FPE(n) 7.088858e+06 6.584623e+06 6.500111e+06 6.532693e+06 6.568885e+06
##                   6            7            8            9           10
## AIC(n) 1.569826e+01 1.570280e+01 1.570626e+01 1.571178e+01 1.571712e+01
## HQ(n)  1.572818e+01 1.573699e+01 1.574472e+01 1.575452e+01 1.576413e+01
## SC(n)  1.577351e+01 1.578881e+01 1.580301e+01 1.581928e+01 1.583538e+01
## FPE(n) 6.571589e+06 6.601527e+06 6.624388e+06 6.661097e+06 6.696818e+06
grangertest(polarization_before$Elections , polarization_before$soros, order=3)
## Granger causality test
## 
## Model 1: polarization_before$soros ~ Lags(polarization_before$soros, 1:3) + Lags(polarization_before$Elections, 1:3)
## Model 2: polarization_before$soros ~ Lags(polarization_before$soros, 1:3)
##   Res.Df Df      F Pr(>F)
## 1    362                 
## 2    365 -3 0.5199 0.6688
grangertest(polarization_before$Elections ~ polarization_before$soros, order=3)
## Granger causality test
## 
## Model 1: polarization_before$Elections ~ Lags(polarization_before$Elections, 1:3) + Lags(polarization_before$soros, 1:3)
## Model 2: polarization_before$Elections ~ Lags(polarization_before$Elections, 1:3)
##   Res.Df Df      F Pr(>F)
## 1    362                 
## 2    365 -3 0.2542 0.8583

WWG1WGA

No causality for WWG1WGA

adf.test(polarization_before$Elections + polarization_before$wwg1wga)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections + polarization_before$wwg1wga
## Dickey-Fuller = -2.2738, Lag order = 7, p-value = 0.4614
## alternative hypothesis: stationary
adf.test(polarization_before$wwg1wga)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$wwg1wga
## Dickey-Fuller = -2.273, Lag order = 7, p-value = 0.4618
## alternative hypothesis: stationary
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
dif_wwg <- diff(polarization_before$wwg1wga)
adf.test(dif_elec + dif_wwg)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_elec + dif_wwg
## Dickey-Fuller = -8.6678, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_elec + dif_wwg)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      3      3      5 
## 
## $criteria
##                   1            2            3            4            5
## AIC(n)     13.44622     13.41548     13.39069     13.38939     13.38502
## HQ(n)      13.45479     13.42833     13.40783     13.41080     13.41072
## SC(n)      13.46777     13.44780     13.43378     13.44325     13.44965
## FPE(n) 691228.01636 670299.21683 653890.06422 653038.64842 650191.69225
##                   6            7            8            9           10
## AIC(n)     13.38765     13.39259     13.39007     13.39341     13.39579
## HQ(n)      13.41763     13.42685     13.42862     13.43624     13.44290
## SC(n)      13.46305     13.47877     13.48703     13.50114     13.51429
## FPE(n) 651903.45676 655136.40973 653490.84122 655678.07832 657244.10800
grangertest(dif_elec , dif_wwg, order=5)
## Granger causality test
## 
## Model 1: dif_wwg ~ Lags(dif_wwg, 1:5) + Lags(dif_elec, 1:5)
## Model 2: dif_wwg ~ Lags(dif_wwg, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.0356 0.9993
grangertest(dif_elec ~ dif_wwg, order=5)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:5) + Lags(dif_wwg, 1:5)
## Model 2: dif_elec ~ Lags(dif_elec, 1:5)
##   Res.Df Df      F Pr(>F)
## 1    355                 
## 2    360 -5 0.0274 0.9996

Globalist

After differencing, we see election incidents are likely to cause but not be caused by “globalist.”

adf.test(polarization_before$Elections + polarization_before$globalist)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections + polarization_before$globalist
## Dickey-Fuller = -2.8047, Lag order = 7, p-value = 0.2373
## alternative hypothesis: stationary
adf.test(polarization_before$Elections)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$Elections
## Dickey-Fuller = -3.2459, Lag order = 7, p-value = 0.08036
## alternative hypothesis: stationary
adf.test(polarization_before$globalist)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  polarization_before$globalist
## Dickey-Fuller = -2.8087, Lag order = 7, p-value = 0.2356
## alternative hypothesis: stationary
dif_glob <- diff(polarization_before$globalist)
adf.test(dif_elec)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_elec
## Dickey-Fuller = -6.6195, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
adf.test(dif_glob)
## 
##  Augmented Dickey-Fuller Test
## 
## data:  dif_glob
## Dickey-Fuller = -7.6349, Lag order = 7, p-value = 0.01
## alternative hypothesis: stationary
VARselect(dif_elec + dif_glob)
## $selection
## AIC(n)  HQ(n)  SC(n) FPE(n) 
##      5      5      2      5 
## 
## $criteria
##                  1           2           3           4           5           6
## AIC(n)    10.80422    10.75546    10.75767    10.74921    10.73950    10.74323
## HQ(n)     10.81279    10.76830    10.77480    10.77062    10.76520    10.77321
## SC(n)     10.82577    10.78777    10.80076    10.80307    10.80414    10.81863
## FPE(n) 49228.11427 46885.15784 46989.04279 46593.07215 46143.19400 46315.51053
##                  7           8           9          10
## AIC(n)    10.74864    10.75268    10.75775    10.75519
## HQ(n)     10.78290    10.79122    10.80057    10.80230
## SC(n)     10.83482    10.84963    10.86547    10.87368
## FPE(n) 46566.76322 46755.45635 46993.26749 46873.31809
grangertest(dif_elec ~ dif_glob, order=5)
## Granger causality test
## 
## Model 1: dif_elec ~ Lags(dif_elec, 1:5) + Lags(dif_glob, 1:5)
## Model 2: dif_elec ~ Lags(dif_elec, 1:5)
##   Res.Df Df      F  Pr(>F)  
## 1    355                    
## 2    360 -5 3.0178 0.01107 *
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
grangertest(dif_elec , dif_glob, order=5)
## Granger causality test
## 
## Model 1: dif_glob ~ Lags(dif_glob, 1:5) + Lags(dif_elec, 1:5)
## Model 2: dif_glob ~ Lags(dif_glob, 1:5)
##   Res.Df Df      F  Pr(>F)  
## 1    355                    
## 2    360 -5 2.4101 0.03618 *
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1

Elections Correlations Before J6

Correlations prior to J6 show the strongest correlation for “Globalist” and election incidents

cor.test(polarization_before$soros , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$soros and polarization_before$Elections
## t = -0.090606, df = 370, p-value = 0.9279
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.10633846  0.09701518
## sample estimates:
##          cor 
## -0.004710338
cor.test(polarization_before$globalist , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$globalist and polarization_before$Elections
## t = 5.0487, df = 370, p-value = 6.996e-07
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.1562216 0.3466003
## sample estimates:
##       cor 
## 0.2538681
cor.test(polarization_before$wwg1wga , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$wwg1wga and polarization_before$Elections
## t = -3.8307, df = 370, p-value = 0.0001501
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.29120924 -0.09553153
## sample estimates:
##        cor 
## -0.1953134
cor.test(polarization_before$nwo , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$nwo and polarization_before$Elections
## t = 2.265, df = 370, p-value = 0.02409
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.01544617 0.21605165
## sample estimates:
##       cor 
## 0.1169416
cor.test(polarization_before$newworld...2 , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$newworld...2 and polarization_before$Elections
## t = 0.70881, df = 370, p-value = 0.4789
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.06509847  0.13798672
## sample estimates:
##        cor 
## 0.03682433
cor.test(polarization_before$new_world , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$new_world and polarization_before$Elections
## t = 1.7033, df = 370, p-value = 0.08935
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.01359593  0.18819625
## sample estimates:
##        cor 
## 0.08820506
cor.test(polarization_before$sum_right , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$sum_right and polarization_before$Elections
## t = -0.39446, df = 370, p-value = 0.6935
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  -0.12192770  0.08134581
## sample estimates:
##         cor 
## -0.02050283
cor.test(polarization_before$nwo_comb , polarization_before$Elections)
## 
##  Pearson's product-moment correlation
## 
## data:  polarization_before$nwo_comb and polarization_before$Elections
## t = 2.1947, df = 370, p-value = 0.02881
## alternative hypothesis: true correlation is not equal to 0
## 95 percent confidence interval:
##  0.01181776 0.21258926
## sample estimates:
##       cor 
## 0.1133606