Kcore Analytics WebsiteView Now

Live Daily Prediction Using Artificial Intelligence

Through collecting a massively large number of tweets and building machine learning models, we investigate the dynamics of the Twitter social network formed by the interactions among millions of Twitter supporters. We then infer the opinion of each user with Artificial Intelligence on the candidates of the 2019 Argentina presidential election. Our predictions are updated daily at 2AM.

Final Predictions as of Oct 27, 2019:
Raw data from Twitter: FF = 45.9%, MP = 32.5%, Third party = 21.6%.
Rescaled data according to gender/age and homophily: FF = 48.9%, MP = 39.6%, Third party = 11.4%

Frente de Todos

Fernández - Fernández

48.9%

Model 3 Final Prediction Updated

October 27, 2019

Third party (11.4%)

Juntos por el Cambio

Macri - Pichetto

39.6%

Third party (11.5%)

Macri - Pichetto

40.26%

A subset of traditional polls from Wikipedia.

UPDATES:

Update November 17, 2019. All the articles in journals documenting the Argentina election are here.
Update October 27, 2019. We have obtained our final estimates for the election. The final Model 3 consists on first obtaining the cumulative classification of the tweets and users (Model 1), then predicting the undecided voters by homophily in the network of contacts (Model 2), and finally applying a re-weighting of the results by age and gender (Model 3).

The final results of the Model 3 are:

FF = 48.9%, MP = 39.6%, Third Party = 11.4%, Difference = 9.3%

The official results of the election with 97.13% of the votes computed are:

FF = 48.10%, MP = 40.37%, Third Party = 11.53%, Difference = 7.73%,

Update September 5, 2019. Hedge fund Autonomy Capital lost about $1 billion last month largely on investments tied to Argentina, making it one of the most prominent investors caught on the wrong side of market turmoil in that country. [WSJ]
Update August 11, 2019. The PASO elections have shown a 15.57% difference for FF over the MP. No pollster anticipated such a large diffence. [Clarin] Our cumulative results in chart 39 show very good agreement with PASO results. Methods are explained in chart 40.
Update August 1, 2019. We updated our model with new set of curated hashtags.
Update July 31, 2019. We have obtained the gender, age and location of a number of users and use this information to renormalize the data according to the demographic of the census population. See Charts 37 and 38.
Update July 25, 2019. We have estimated the intention of vote of the undecided population by homophily.
Update July 8, 2019. Macri-Pichetto has improved its vote intention but Fernandez-Fernandez is still above Macri in the balotage and in first round taking into account undecided. We observe an extreme polarization. As of today, no third candidate is viable, including Lavagna and Espert. We are closely following the undecided to predict their vote intention using graph theory and other analytical methods.
Update June 25, 2019. We have trained a new model incorporating the new hashtags after the closing of the PASO lists of candidates.
Update June 7, 2019. A new survey from Synopsis reported in Clarin gives an improvement of Macri against Cristina of 50.7% over 49.3%. Clarin reports that the majority of the consulting firms now give Macri a “repunte” (improvement) over the last couple of weeks. Kcore anticipated this result already on May 13, when Cristina reached its peak of 57% and then went down steadly after the announcement of the formula Fernandez-Fernandez, until reaching a technical tie with Macri on May 31. As of today, June 7, Kcore is predicting a new change of trend with an improvement of Cristina over Macri: 53% to 47%. This result has not been predicted yet by any traditional polling firm.
Update May 28, 2019. Our predictions show Cristina going down in preference after her announcement of the formula Fernandez-Fernandez. This prediction was confirmed today by Opinaia in a note in Clarin. See Mauricio Macri pasa al frente en el balotaje: 51 a 49
Update May 7, 2019. New poll from Insomnia reported by Clarin reveals that, in an eventual ballot, the advantage on the intention of vote of Cristina over Macri has decreased in relation to the first half of April.
Update May 3, 2019. Trusted sources according to Clarin: Isonomia, Gustavo Cordoba, Ricardo Rouvier, Opinaia, Imagen y Gestión Política, Demos Consulting, Gustavo Marangoni, Giacobbe, M&F. Trusted sources from the 2015 election: Management & Fit, Poliarquia, OPSM, Giacobbe, IPA.
Update April 29, 2019. A recent poll featured in Clarin from Opinaia ("tuvo buenos pronósticos en 2015 y 2017") brings some "alivio" calm to the Macri camp showing a 51% for Cristina and 49% for Macri in ballotage. AI predicted similar results one week before and confirmed the result this week.
Update April 24, 2019. Kcore-Analytics Artificial Intelligence predicts in advance of the traditional polls the results of the 2019 Argentine Presidential election. We use artificial intelligence to predict the opinion of 3.8 million users from Twitter. The Twitter data is selected both by content and by frequency eliminating bots so it is quite robust to predict the results. It is measured in real time and we can identify influencers, hashtags and the results at the same time.
Update April 18, 2019. We are currently following the Argentinian Election Campaign. Our results show an advantage of Cristina Kirchner over Mauricio Macri of 10 points since February 2019. This prediction was confirmed by Isonomia on April 18, 2019, a pollster consulting firm working for the Macri administration, creating a large concern among politicians in Cambiemos Party, increasing the 'riesgo país' index to 854 points and bringing down stocks of Argentinian firms in Wall Street by 8%. See [clarin.com] [infobae.com]. Our AI anticipated this result by one month.
trusted sources: M&F, Elypsis


#Hashtags

(15) Top 50 cumulative #hashtags (2019-08-11 updated)
# Hashtag Count Camp
1 cambiemos 144195 M
2 macri 119081 M
3 sinceramente 92838 K
4 argentina 87724
5 habraconsecuencias 84675 K
6 elecciones2019 84138
7 cristina 71395 K
8 cfk 68295 K
9 elecciones 60086
10 urgente 50908
11 novuelvenmas 49752 M
12 encuesta 48483
13 juntosporelcambio 45985 M
14 cronicaanunciada 44841 K
15 politica 41610
16 yovotomm 37397 M
17 axelgobernador 37222 K
18 navarro2019 35916 K
19 animalessueltos 34816
20 buenlunes 34356
21 fuerzacristina 33178 K
22 macritevuelvoaelegir 32530 M
23 buenmartes 30750
24 defensoresdelcambio 28739 M
25 lacornisa 27464
26 debodecir 25126 M
27 buenmiercoles 24194
28 sobredosisdetv 23966
29 buendomingo 23825
30 fernandezfernandez 23213 K
31 alerta 22572
32 operacionlibertad 22516
33 vidal 22445 M
34 cordoba 22350
35 ultimahora 22034
36 buenjueves 21761
37 massa 20590 K
38 escontodos 20302 K
39 chaumacri 19843 K
40 juntossomosimparables 19772 M
41 hayotrocamino 19226 K
42 andatemacri 18712 K
43 humor 17977
44 mm2019 17649 M
45 frentedetodos 17401 K
46 buensabado 17323
47 lanochedeml 17129
48 sinluz 16771
49 fernandezfernandez2019 16238 K
50 sracristinalecuentoque 16113 M
(16) #hashtags from camps (2019-08-11 updated)
# Hashtag Camp Count
1 cambiemos M 144195
2 macri M 119081
3 sinceramente K 92838
4 habraconsecuencias K 84675
5 cristina K 71395
6 cfk K 68295
7 novuelvenmas M 49752
8 juntosporelcambio M 45985
9 cronicaanunciada K 44841
10 yovotomm M 37397
11 axelgobernador K 37222
12 navarro2019 K 35916
13 fuerzacristina K 33178
14 macritevuelvoaelegir M 32530
15 defensoresdelcambio M 28739
16 debodecir M 25126
17 fernandezfernandez K 23213
18 vidal M 22445
19 massa K 20590
20 escontodos K 20302
21 chaumacri K 19843
22 juntossomosimparables M 19772
23 hayotrocamino K 19226
24 andatemacri K 18712
25 mm2019 M 17649
26 frentedetodos K 17401
27 fernandezfernandez2019 K 16238
28 sracristinalecuentoque M 16113
29 vidalespeorquemacri K 15817
30 ucr M 14647
31 mesaza M 14057
32 macripichetto2019 M 13902
33 macrisis K 13286
34 albertoycristina K 13239
35 chautelefe K 13209
36 ellalegana K 12973
37 cristina2019 K 12959
38 albertopresidente K 12818
39 mm2019listacompleta M 12683
40 haganalgo K 12086
41 pichetto M 11798
42 todosarosario K 11626
43 pj K 11135
44 podemoshablar K 11133
45 kirchner K 11127
46 hayfuturoparavos K 10529
47 nvm M 10418
48 macrimiente K 10131
49 cambiemosnosfundio K 10103
50 cristinateesperamos K 9582
51 cfkenlaferia K 9393
52 macri2019 M 9155
53 labandaajuicio M 9047
54 elfuturoescontodos K 9034
55 pongamosunafotodecfk K 9016
56 mananasylvestre10anos K 8989
57 sinjusticianohayrepublica M 8945
58 sevan K 8898
59 paronacional K 8837
60 nimacrinicristina L 8649
61 2deagosto M 8193
62 novuelvenmasensudelictivavida M 8116
63 fuerzapresidente M 8097
64 lavagna L 8056
65 cristinaenelpj K 7904
66 sevamacri K 7836
67 futurocontodos K 7813
68 mauroenamerica M 7760
69 nuevamayoria K 7658
70 chaucambiemos K 7524
71 macriteodia K 7427
72 gutierrezesmacri M 7420
73 nestorcumple K 7358
74 macriespobreza K 7219
75 albertofernandez K 7179
76 nuncamasmacri K 6975
77 yovotoff K 6635
78 boletacompleta M 6597
79 concfkviviasmejor K 6463
80 hayalternativa K 6385
81 hartosdelosk M 6382
82 juiciopoliticoamacri K 6364
83 macrigato K 6200
84 consensofederal L 6101
85 macrichau K 5848
86 sisepuede M 5723
87 macriestoyconvos M 5722
88 vidallapeor K 5666
89 macrilandia K 5352
90 lasmentirasdevidal K 5312
91 vidalespeor K 5281
92 evita100anos K 5227
93 argentinaexplota K 5130
94 blindadaporpichetto M 5093
95 fuerzaflorencia K 5032
96 yovotoalbertopresidente K 4988
97 macrinuncamas K 4985
98 ellavuelve K 4907
99 confirmenaramospadilla K 4877
100 yovotoavidal M 4877
101 majulmercenario K 4787
102 elijoalosfernandez K 4703
103 nestorpudo K 4628
104 derrumbepro K 4596
105 dalessiogate K 4454
106 yovotoalgato M 4417
107 macrivendepatria K 4345
108 macripichetto2019listacompleta M 4319
109 lasobrasnomienten M 4301
110 ruidazo K 4297
111 kicillofmagario K 4253
112 juntosporelcambioboletacompleta M 4236
113 kicillof K 4228
114 ramospadillasomostodos K 4192
115 yovotoalfrentedetodos K 4172
116 macriarruga K 3997
117 peponilaconmacri M 3945
118 cristinaenmardelplata K 3928
119 manotazopro K 3923
120 chautimba K 3923
121 dejencompetiraespert L 3850
122 cristinaenmendoza K 3846
123 vidal2019 M 3840
124 conalbertohayfuturo K 3792
125 macriescaos K 3708
126 elplandecristina M 3686
127 ruidazonacional K 3648
128 peronvive K 3620
129 macrichetto M 3596
130 macrijefedelabanda K 3581
131 cristinasomostodos K 3573
132 macriaprietaalaprensa K 3558
133 desafueroastornelliya K 3501
134 porunaargentinamejor M 3485
135 bonadiotrucho K 3364
136 sefugarontodo M 3271
137 lapropuestadetodos K 3249
138 cambiandojuntos M 3231
139 1mruidazonacional K 3221
140 sevanenprimeravuelta K 3213
141 yosigoconfiando M 3181
142 novuelvennuncamas M 3180
143 macriconmajul M 3169
144 chautn K 3055
145 todosconalberto K 3050
146 macriesviolencia K 3021
147 cfkenlarural K 2964
148 cordobacambia M 2937
149 hambrepro K 2929
150 elversodelascloacas K 2904
151 macrienlarural M 2857
152 camaradaaxel M 2854
153 cristinaenmalvinas K 2849
154 cambiemosacambiemos K 2824
155 nonosdejemospsicopatear M 2720
156 cacerolazo M 2682
157 lilitaconmajul M 2579
158 colapsopro K 2576
159 abismopro K 2523
160 loquequeremostodos K 2496
161 chautelefenoticias K 2486
162 lamafiademacri K 2434
163 stornelliaindagatoria K 2433
164 kirchnerista K 2402
165 nestorvive K 2402
166 cristinaensantiago K 2368
167 arrugocristina K 2358
168 conramospadillano K 2343
169 serobaronlasjubilaciones K 2261
170 proeshambre K 2214
171 ramospadillanosetoca K 2146
172 patriaproductiva K 2050
173 conffjubilacionesdignas K 1995
174 mananacordobacambia M 1958
175 21mporlaverdad K 1837
176 riesgomacri K 1768
177 caecarrio K 1664
178 mafiajudicialpro K 1657
179 votoaxelgobernador K 1627
180 elriesgoescristina M 1547
181 fuerzamauricio M 1531
182 nosunimos K 1504
183 rompierontodo K 1389
184 vienenairesbuenos K 1330
185 elfuturoesmacri M 1284
186 duhalde L 1278
187 derrotemosalfmi K 1132
188 recuperarlorobado M 1109
189 todomal K 1087
190 mariugeniadelmal K 950
191 macrilohizo K 941
192 yotevotoalberto K 910
193 massa2019 K 529
194 ponelelavagnaalprograma L 215
195 noaflojamos M 207
196 aumentatodo K 48
197 24a M 8
198 macrichauchauchau K 8
199 macrieselfuturo M 6
200 lodamosvuelta M 6
201 ladamosvuelta M 6
202 apagonamajul M 6
203 macrihacetecargo K 5
204 siguenmintiendo K 4
205 macrivsmacri K 2
(17) Hashtag network updated on 2019-08-08 (p-value<10-7)
ht_network
(18) Hashtag Cloud of FF camp
K_cloud
(19) Hashtag Cloud of MP camp
M_cloud

Bots dynamics



Fitting with polls

Weekly opinion flowing
(41) Cumulative users from March 1st
AugustSeptemberOctober
Ultra loyal FF505875578168612402
Ultra loyal MP311410365065384819
Undecided - FF72712161319
Undecided - MP132918662033
Undecided - Undecided108955140180148655
loyal FF - FF202050260255278230
loyal FF - MP535671518269
loyal FF - Undecided93141353815437
loyal MP - FF184635653704
loyal MP - MP132805180513196932
loyal MP - Undecided514085538637
Unclassified215520251796263739
150032718118661924176
homophily (undecided)
FF (from undecided)537046636066736
MP (from undecided)192822912932465
Undecided (from undecided)380254777352806
111011143262152007
homophily (unclassfied)
FF (from unclassified)722188677483264
MP (from unclassified)218932945533653
Unclassified (from unclassified)121409135567146822
215520251796263739
(42) Prediction results
MODEL 1AugustSeptemberOctoberNote
FF722595859112914338Ultra loyal FF + loyal FF-FF + loyal FF-MP + loyal FF-Undecided
MP451201557696594092Ultra loyal MP + loyal MP-MP + loyal MP-FF + loyal MP-Undecided
third party326531395058415746Undecided-MP + Undecided-FF + Undecided-Undecided + Unclassified
FF (%)48.2%47.4%47.5%
MP (%)30.1%30.8%30.9%
third party (%)21.8%21.8%21.6%
MODEL 2AugustSeptemberOctoberNote
FF776299925472981074Ultra loyal FF + loyal FF-FF + loyal FF-MP + loyal FF-Undecided + FF(undeicded)
MP470483586825626557Ultra loyal MP + loyal MP-MP + loyal MP-FF + loyal MP-Undecided + MP(undecided)
third party253545299569316545Undecided(undecided) + Unclassfied
FF (%)51.7%51.1%51.0%
MP (%)31.4%32.4%32.6%
third party (%)16.9%16.5%16.5%
MODEL 3AugustSeptemberOctoberNote
FF (%)50.2%50.2%50.1%rescaling the results from MODEL 2
MP (%)36.9%38.0%38.3%
third party (%)12.9%11.8%11.6%
MODEL 4AugustSeptemberOctoberNote
FF84851710122461064338Ultra loyal FF + loyal FF-FF + loyal FF-MP + loyal FF-Undecided + FF(undeicded) + FF(unclassified)
MP492376616280660210Ultra loyal MP + loyal MP-MP + loyal MP-FF + loyal MP-Undecided + MP(undecided) + MP(unclassified)
third party159434183340199628Undecided(undecided) + Unclassfied(unclassified)
FF (%)56.6%55.9%55.3%
MP (%)32.8%34.0%34.3%
third party (%)10.6%10.1%10.4%
MODEL 5AugustSeptemberOctoberNote
FF (%)53.2%52.9%52.5%rescaling the results from MODEL 4
MP (%)38.0%39.1%39.4%
third party (%)8.8%8.1%8.0%
(43) Distribution of population in Argentina from Twitter electorate and Census

Census
Twitter
(before PASO)
(after PASO)

2016 US census
2016 US voters

(45) Distribution of voters according to ages and genders

270 thousand cumulative users from Twitter
100 thousand users from Twitter in one week
3.5 thousand respondents (before PASO)
3.1 thousand respondents (after PASO)
(46) Twitter supporter dynamics from AI compared with PASO (cumulative, p=0.68)

(50) Population distribution in Argentina according to areas

(51) Population distribution in US according to states (2.59 million users)

(52) Daily new users

ht_network
New users (w=14 days)

ht_network
(53) Swing and loyal voters in FF and MP camp

ht_network
(54) Distribution of FF (MP) tweets of FF (MP) supporters

ht_network
(54) CCDF of FF (MP) tweets of FF (MP) supporters

ht_network