Brazil in the General Assembly: from redemocratization to date

Amena Yassine - Statistics for International Relations Research II

The views and opinions expressed in this article are those of the author and do not reflect the official policy or the position of the institutions to which the author is affiliated.

This project was submitted in partial fulfillment of the requirements of the course RI-SP062 and not for public disclosure

Cleaning environment

Loading packages

1. Introduction

This project will attempt to answer the following research questions:

  • have the speeches of Brazil in the General-Debates of the United Nations General Assembly remained stable since the country’s redemocratization in 1985?

  • have peace and development remained constant priorities of Brazilian Foreign Policy in these speeches?

Brazil has the oldest professional diplomacy in the Americas, as the country became the site of colonial power when the Portuguese court fled the Napoleonic wars. Considered a precursor of Brazilian diplomacy, Alexandre de Gusmão, born in the Portuguese colony of Brazil was one of the main negotiators of the Treaty of Madrid in 1761 between Portugal and Spain, roughly defining the borders of future independent Brazil. In fact, diplomacy played a crucial role in the very constitution of Brazil as an independent country, since the Brazilian borders were all defined by diplomatic negotiations rather than wars with its numerous neighbors (Pinheiro, 2010; Cervo & Bueno, 2002; Mendonça, 2013).

Due to its long professional tradition and the influence that the Ministry of Foreign Affairs of Brazil still enjoys to date, the tenants of the Brazilian foreign policy are thought to have remained stable over time. Sharing boarders with 10 out of 12 South American countries, Brazil has not been involved in armed conflicts with any of its neighbors for the past 150 years. Therefore, “peace” has been pointed as a priority for the Brazilian foreign policy (Cervo & Bueno, 2002). That the search for “development” has remained a main priority is however a disputed argument among historians (Pinheiro, 2010; Cervo & Bueno, 2002).

Since redemocratization in 1985, the Brazilian history has been marked by a few ruptures, with the impeachment of President Fernando Collor de Mello (1992) and of President Dilma Rousseff (2016). There have also been a variety of administrations, with different political affiliations, from liberal to leftist to ultraconservative. I want to verify whether the speeches of Brazil in the UNGA remained stable across these administrations and with a focus on peace and development, even if the country faced political ruptures over time.

Therefore, I want to test the hypotheses that (i) the policy positions of Brazil in UNGA have remained constant since the redemocratization period and that (ii) the policy positions of Brazil in UNGA focused on peace and development during this time.

2. Methodology

The speeches of Brazil in the UNGA have a singular relevance, as by tradition the country opens the General Debate of the UNGA and firstly speaks about its perception of the world affairs and about its foreign policy priorities for the entire UN membership (Vargas, 2012).

By employing the technique of Natural Language Processing (NLP), I will assess Brazil’s 37 speeches delivered in the General Debates of the UN General Assembly (UNGA), between 1985 to 2020.

NLP uses machine learning techniques to assess a large number of natural language data. A subfield of artificial intelligence, NLP provides the most appropriate methods (tokenization, stemming, parsing, among others) for content analysis that allows for extracting keywords, identifying relations between words, establishing the structure of the text, measuring the degree of lexical homogeneity of the corpora, etc.

In this light, NLP will assist us in measuring:

  • how much the policy positions of Brazil in the GD of the UNGAs have changed from one speech to another since 1985 to date; and,

  • whether development and peace have remained the main features of such corpora.

3. Data

Describing the Data

BRUNGA is a compilation of the 37 speeches of Brazil since 1985 to date, which I built by drawing on a corpus of data that is publicly available in the official websites of the United Nations General Assembly and in the book “Brazil in the United Nations: 1945 - 2011”, organized by Seixas Corrêa and the publisher house of the MFA of Brazil (FUNAG) (Corrêa and Fundação Alexandre de Gusmão, 2013). The data comprise the speeches delivered from UNGA40 to UNGA75. The speeches were originally delivered in Portuguese; however, an English version of the speeches was provided by the Brazilian delegation in each occasion. The speeches used here are, therefore, in English language.

I also built eight separated data for each different administration, as follows:

Sarney (term 1985-1990): a compilation of the speeches delivered in UNGA40 in 1985, UNGA41 in 1986, UNGA42 in 1987, UNGA43 in 1988 and UNGA44 in 1989. The UNGA40 and UNGA44 speeches were delivered by President José Sarney, and the others by the Minister of Foreign Affairs Abreu Sodré.

Collor (term 1990-1992): a compilation of the speeches delivered in UNGA45 in 1990, UNGA46 in 1991 and UNGA47 in 1992. The UNGA45 and UNGA46 speeches were delivered by President Fernando Collor de Mello (also known as Collor), and the other by the Minister of Foreign Affairs Celso Lafer. Note that Minister Celso Lafer was also MFA of Brazil in the last year of the adminstration of President Fernando Henrique Cardoso (also known as FHC).

Franco (term 1992-1995): a compilation of the speeches delivered in UNGA48 in 1993 and UNGA49 in 1994. The UNGA48 and UNGA49 speeches were delivered by the Minister of Foreign Affairs Celso Amorim. President Itamar Franco did not deliver speeches in the GA. Note that Minister Celso Amorim was also MFA of Brazil under President Lula.

FHC (term 1995-2003): a compilation of the speeches delivered in UNGA50 in 1995, UNGA51 in 1996, UNGA52 in 1997, UNGA53 in 1998, UNGA54 in 1999, UNGA55 in 2000, UNGA56 in 2001 and UNGA57 in 2002. The Minister of Foreign Affairs Luiz Felipe Lampreia delivered the speeches between UNGA50 and UNGA55. President Fernando Henrique Cardoso delivered the UNGA56 speech. And the Minister of Foreign Affairs Celso Lafer delivered the last speech of the adminstration FHC in UNGA 57.

Lula (term 2003-2011): a compilation of the speeches delivered in UNGA58 in 2003, UNGA59 in 2004, UNGA60 in 2005, UNGA61 in 2006, UNGA62 in 2007, UNGA63 in 2008, UNGA64 in 2009 and UNGA65 in 2010. Except for UNGA60 and UNGA65, where the Minister of Foreign Affairs Celso Amorim delivered the speeches, all others were delivered by President Luiz Inácio Lula da Silva (also known as Lula).

Rousseff (term 2011-2016): a compilation of the speeches delivered in UNGA66 in 2011, UNGA67 in 2012, UNGA68 in 2013, UNGA69 in 2014 and UNGA70 in 2015. All the speeches were delivered by President Dilma Rousseff.

Temer (term 2016-2019): a compilation of the speeches delivered in UNGA71 in 2016, UNGA72 in 2017 and UNGA73 in 2018. All the speeches were delivered by President Michel Temer.

Bolsonaro (term 2019-present): a compilation of the speeches delivered in UNGA74 in 2019 and UNGA75 in 2020. All the speeches were delivered by President Jair Bolsonaro.

Presidents Sarney, Franco and Temer belonged to the same center-right, liberal party: Movimento Democrático Brasileiro (PMDB or, currently, MDB). President Collor was from a right-wing, liberal-conservative party: Partido da Reconstrução Nacional (PRN or, currently, PTC). President Lula and Rousseff belong to the same left-wing party: Partido dos Trabalhadores (PT). President Bolsonaro, currently independent, was elected on a far-right platform and belonged to Partido Social Liberal (PSL).

Access to the speeches

The speeches are on the google drive that I shared with Prof Hollway and Juliette’s email address and can be assessed here:

BRUNGA https://drive.google.com/drive/folders/1fRDlvEufY3R8QCK5jlqzIL5oq7cmfAC2?usp=sharing

BRUNGAADM/by administration dataset https://drive.google.com/drive/folders/1mUSI_YgEQBhtVa3tMPeuJZ3Q6-wxWp3Q?usp=sharing

Treating the data

Importing the data

Creating a corpus

corp_Sarney <- corpus(Sarney, 
                     docvars = data.frame(party = names(Sarney)))
corp_Collor <- corpus(Collor, 
                     docvars = data.frame(party = names(Collor)))
corp_Franco <- corpus(Franco, 
                     docvars = data.frame(party = names(Franco)))
corp_FHC <- corpus(FHC, 
                     docvars = data.frame(party = names(FHC)))
corp_Lula <- corpus(Lula, 
                     docvars = data.frame(party = names(Lula)))
corp_Rousseff <- corpus(Rousseff, 
                     docvars = data.frame(party = names(Rousseff)))
corp_Temer <- corpus(Temer, 
                     docvars = data.frame(party = names(Temer)))
corp_Bolsonaro <- corpus(Bolsonaro, 
                     docvars = data.frame(party = names(Bolsonaro)))

Adding a column for Administration

## Corpus consisting of 36 documents, showing 36 documents:
## 
##                 Text Types Tokens Sentences Administration
##     UNGA40Sarney.pdf  1839   6306       254         Sarney
##      UNGA41Sodre.pdf  1664   5564       193         Sarney
##      UNGA42Sodre.pdf  1098   3391       106         Sarney
##      UNGA43Sodre.pdf  1222   3564       119         Sarney
##     UNGA44Sarney.pdf  1490   4713       210         Sarney
##     UNGA45Collor.pdf  1167   3721       147         Collor
##     UNGA46Collor.pdf  1130   3624       141         Collor
##      UNGA47Lafer.pdf  1283   4196       165         Collor
##     UNGA48Amorim.pdf  1428   4567       133         Franco
##     UNGA49Amorim.pdf  1243   3764       144         Franco
##   UNGA50Lampreia.pdf  1196   3824       137            FHC
##   UNGA51Lampreia.pdf   983   3180       104            FHC
##   UNGA52Lampreia.pdf   918   2821       127            FHC
##   UNGA53Lampreia.pdf  1057   3212       117            FHC
##   UNGA54Lampreia.pdf   965   2672        88            FHC
##   UNGA55Lampreia.pdf   962   2630       103            FHC
##    UNGA56Cardoso.pdf   882   2409        97            FHC
##      UNGA57Lafer.pdf   654   1579        63            FHC
##       UNGA58Lula.pdf  1058   2853       128           Lula
##       UNGA59Lula.pdf   907   2291       102           Lula
##     UNGA60Amorim.pdf   865   2284       100           Lula
##       UNGA61Lula.pdf   850   2282       113           Lula
##       UNGA62Lula.pdf   760   1816       100           Lula
##       UNGA63Lula.pdf   790   1872        79           Lula
##       UNGA64Lula.pdf   832   2174        94           Lula
##     UNGA65Amorim.pdf   991   2537       118           Lula
##   UNGA66Rousseff.pdf   930   2668       123       Rousseff
##   UNGA67Rousseff.pdf   977   2608       113       Rousseff
##   UNGA68Rousseff.pdf   970   2694       120       Rousseff
##   UNGA69Rousseff.pdf  1087   2935       131       Rousseff
##   UNGA70Rousseff.pdf   968   2611       108       Rousseff
##      UNGA71Temer.pdf   955   2455       149          Temer
##      UNGA72Temer.pdf   851   2233       127          Temer
##      UNGA73Temer.pdf   870   2426       135          Temer
##  UNGA74Bolsonaro.pdf  1290   3856       159      Bolsonaro
##  UNGA75Bolsonaro.pdf   854   2078        86      Bolsonaro

Adding a column for Political Affiliation

## Corpus consisting of 36 documents, showing 36 documents:
## 
##                 Text Types Tokens Sentences Administration  Affiliation
##     UNGA40Sarney.pdf  1839   6306       254         Sarney Center Right
##      UNGA41Sodre.pdf  1664   5564       193         Sarney Center Right
##      UNGA42Sodre.pdf  1098   3391       106         Sarney Center Right
##      UNGA43Sodre.pdf  1222   3564       119         Sarney Center Right
##     UNGA44Sarney.pdf  1490   4713       210         Sarney Center Right
##     UNGA45Collor.pdf  1167   3721       147         Collor        Right
##     UNGA46Collor.pdf  1130   3624       141         Collor        Right
##      UNGA47Lafer.pdf  1283   4196       165         Collor        Right
##     UNGA48Amorim.pdf  1428   4567       133         Franco Center Right
##     UNGA49Amorim.pdf  1243   3764       144         Franco Center Right
##   UNGA50Lampreia.pdf  1196   3824       137            FHC Center Right
##   UNGA51Lampreia.pdf   983   3180       104            FHC Center Right
##   UNGA52Lampreia.pdf   918   2821       127            FHC Center Right
##   UNGA53Lampreia.pdf  1057   3212       117            FHC Center Right
##   UNGA54Lampreia.pdf   965   2672        88            FHC Center Right
##   UNGA55Lampreia.pdf   962   2630       103            FHC Center Right
##    UNGA56Cardoso.pdf   882   2409        97            FHC Center Right
##      UNGA57Lafer.pdf   654   1579        63            FHC Center Right
##       UNGA58Lula.pdf  1058   2853       128           Lula         Left
##       UNGA59Lula.pdf   907   2291       102           Lula         Left
##     UNGA60Amorim.pdf   865   2284       100           Lula         Left
##       UNGA61Lula.pdf   850   2282       113           Lula         Left
##       UNGA62Lula.pdf   760   1816       100           Lula         Left
##       UNGA63Lula.pdf   790   1872        79           Lula         Left
##       UNGA64Lula.pdf   832   2174        94           Lula         Left
##     UNGA65Amorim.pdf   991   2537       118           Lula         Left
##   UNGA66Rousseff.pdf   930   2668       123       Rousseff         Left
##   UNGA67Rousseff.pdf   977   2608       113       Rousseff         Left
##   UNGA68Rousseff.pdf   970   2694       120       Rousseff         Left
##   UNGA69Rousseff.pdf  1087   2935       131       Rousseff         Left
##   UNGA70Rousseff.pdf   968   2611       108       Rousseff         Left
##      UNGA71Temer.pdf   955   2455       149          Temer Center Right
##      UNGA72Temer.pdf   851   2233       127          Temer Center Right
##      UNGA73Temer.pdf   870   2426       135          Temer Center Right
##  UNGA74Bolsonaro.pdf  1290   3856       159      Bolsonaro    Far Right
##  UNGA75Bolsonaro.pdf   854   2078        86      Bolsonaro    Far Right

Adding a column for year of the statement and visualizing the table

docvars(corp_BRUNGA, "Year") <- 1984
docvars(corp_BRUNGA, field = "Year") <- floor(docvars(corp_BRUNGA, field = "Year")) + 1:36
summary(corp_BRUNGA)

## Corpus consisting of 36 documents, showing 36 documents:
## 
##                 Text Types Tokens Sentences Administration  Affiliation Year
##     UNGA40Sarney.pdf  1839   6306       254         Sarney Center Right 1985
##      UNGA41Sodre.pdf  1664   5564       193         Sarney Center Right 1986
##      UNGA42Sodre.pdf  1098   3391       106         Sarney Center Right 1987
##      UNGA43Sodre.pdf  1222   3564       119         Sarney Center Right 1988
##     UNGA44Sarney.pdf  1490   4713       210         Sarney Center Right 1989
##     UNGA45Collor.pdf  1167   3721       147         Collor        Right 1990
##     UNGA46Collor.pdf  1130   3624       141         Collor        Right 1991
##      UNGA47Lafer.pdf  1283   4196       165         Collor        Right 1992
##     UNGA48Amorim.pdf  1428   4567       133         Franco Center Right 1993
##     UNGA49Amorim.pdf  1243   3764       144         Franco Center Right 1994
##   UNGA50Lampreia.pdf  1196   3824       137            FHC Center Right 1995
##   UNGA51Lampreia.pdf   983   3180       104            FHC Center Right 1996
##   UNGA52Lampreia.pdf   918   2821       127            FHC Center Right 1997
##   UNGA53Lampreia.pdf  1057   3212       117            FHC Center Right 1998
##   UNGA54Lampreia.pdf   965   2672        88            FHC Center Right 1999
##   UNGA55Lampreia.pdf   962   2630       103            FHC Center Right 2000
##    UNGA56Cardoso.pdf   882   2409        97            FHC Center Right 2001
##      UNGA57Lafer.pdf   654   1579        63            FHC Center Right 2002
##       UNGA58Lula.pdf  1058   2853       128           Lula         Left 2003
##       UNGA59Lula.pdf   907   2291       102           Lula         Left 2004
##     UNGA60Amorim.pdf   865   2284       100           Lula         Left 2005
##       UNGA61Lula.pdf   850   2282       113           Lula         Left 2006
##       UNGA62Lula.pdf   760   1816       100           Lula         Left 2007
##       UNGA63Lula.pdf   790   1872        79           Lula         Left 2008
##       UNGA64Lula.pdf   832   2174        94           Lula         Left 2009
##     UNGA65Amorim.pdf   991   2537       118           Lula         Left 2010
##   UNGA66Rousseff.pdf   930   2668       123       Rousseff         Left 2011
##   UNGA67Rousseff.pdf   977   2608       113       Rousseff         Left 2012
##   UNGA68Rousseff.pdf   970   2694       120       Rousseff         Left 2013
##   UNGA69Rousseff.pdf  1087   2935       131       Rousseff         Left 2014
##   UNGA70Rousseff.pdf   968   2611       108       Rousseff         Left 2015
##      UNGA71Temer.pdf   955   2455       149          Temer Center Right 2016
##      UNGA72Temer.pdf   851   2233       127          Temer Center Right 2017
##      UNGA73Temer.pdf   870   2426       135          Temer Center Right 2018
##  UNGA74Bolsonaro.pdf  1290   3856       159      Bolsonaro    Far Right 2019
##  UNGA75Bolsonaro.pdf   854   2078        86      Bolsonaro    Far Right 2020

Transforming corpus into tokens

toks_Sarney <- tokens(corp_Sarney)
toks_Collor <- tokens(corp_Collor)
toks_Franco <- tokens(corp_Franco)
toks_FHC <- tokens(corp_FHC)
toks_Lula <- tokens(corp_Lula)
toks_Rousseff <- tokens(corp_Rousseff)
toks_Temer <- tokens(corp_Temer)
toks_Bolsonaro <- tokens(corp_Bolsonaro)
toks_BRUNGA <- tokens(corp_BRUNGA)

Removing punctuations

nopunct_Sarney <- tokens(toks_Sarney, remove_punct = TRUE)
nopunct_Collor <- tokens(toks_Collor, remove_punct = TRUE)
nopunct_Franco <- tokens(toks_Franco, remove_punct = TRUE)
nopunct_FHC <- tokens(toks_FHC, remove_punct = TRUE)
nopunct_Lula <- tokens(toks_Lula, remove_punct = TRUE)
nopunct_Rousseff <- tokens(toks_Rousseff, remove_punct = TRUE)
nopunct_Temer <- tokens(toks_Temer, remove_punct = TRUE)
nopunct_Bolsonaro <- tokens(toks_Bolsonaro, remove_punct = TRUE)
nopunct_BRUNGA <- tokens(toks_BRUNGA, remove_punct = TRUE)

Removing stop words

cleantoks_Sarney <- tokens_select(nopunct_Sarney, pattern = stopwords("en"), selection = "remove")
cleantoks_Collor <- tokens_select(nopunct_Collor, pattern = stopwords("en"), selection = "remove")
cleantoks_Franco <- tokens_select(nopunct_Franco, pattern = stopwords("en"), selection = "remove")
cleantoks_FHC <- tokens_select(nopunct_FHC, pattern = stopwords("en"), selection = "remove")
cleantoks_Lula <- tokens_select(nopunct_Lula, pattern = stopwords("en"), selection = "remove")
cleantoks_Rousseff <- tokens_select(nopunct_Rousseff, pattern = stopwords("en"), selection = "remove")
cleantoks_Temer <- tokens_select(nopunct_Temer, pattern = stopwords("en"), selection = "remove")
cleantoks_Bolsonaro <- tokens_select(nopunct_Bolsonaro, pattern = stopwords("en"), selection = "remove")
cleantoks_BRUNGA <- tokens_select(nopunct_BRUNGA, pattern = stopwords("en"), selection = "remove")

Creating DFMs

dfmtoks_Sarney <- tokens(cleantoks_Sarney)
dfmat_Sarney <- dfm(dfmtoks_Sarney)
dfmtoks_Collor <- tokens(cleantoks_Collor)
dfmat_Collor <- dfm(dfmtoks_Collor)
dfmtoks_Franco <- tokens(cleantoks_Franco)
dfmat_Franco <- dfm(dfmtoks_Franco)
dfmtoks_FHC <- tokens(cleantoks_FHC)
dfmat_FHC <- dfm(dfmtoks_FHC)
dfmtoks_FHC <- tokens(cleantoks_FHC)
dfmat_FHC <- dfm(dfmtoks_FHC)
dfmtoks_Lula <- tokens(cleantoks_Lula)
dfmat_Lula <- dfm(dfmtoks_Lula)
dfmtoks_Rousseff <- tokens(cleantoks_Rousseff)
dfmat_Rousseff <- dfm(dfmtoks_Rousseff)
dfmtoks_Temer <- tokens(cleantoks_Temer)
dfmat_Temer <- dfm(dfmtoks_Temer)
dfmtoks_Bolsonaro <- tokens(cleantoks_Bolsonaro)
dfmat_Bolsonaro <- dfm(dfmtoks_Bolsonaro)
dfmtoks_BRUNGA <- tokens(cleantoks_BRUNGA)
dfmat_BRUNGA <- dfm(dfmtoks_BRUNGA)

Creating clean DFMs

BRUNGA

Sarney (term 1985-1990)

Collor (term 1990-1992)

Franco (term 1992-1995)

FHC (term 1995-2003)

Lula (term 2003-2011)

Rousseff (term 2011-2016)

Temer (term 2016-2019)

Bolsonaro (term 2019-present)

Creating feature co-occurrence matrix (FCM)

fcmat_Sarney <- fcm(cleandfmat_Sarney)
fcmat_Collor <- fcm(cleandfmat_Collor)
fcmat_Franco <- fcm(cleandfmat_Franco)
fcmat_FHC <- fcm(cleandfmat_FHC)
fcmat_Lula <- fcm(cleandfmat_Lula)
fcmat_Rousseff <- fcm(cleandfmat_Rousseff)
fcmat_Temer <- fcm(cleandfmat_Temer)
fcmat_Bolsonaro <- fcm(cleandfmat_Bolsonaro)
fcmat_BRUNGA <- fcm(cleandfmat_BRUNGA)

4. Empirical findings about the stability and the thematic axes of the speeches of Brazil in UNGA between 1985 and 2020

Brazil’s policy positions in the UNGA (1985-2020):

Table 1

Top features for all 37 speeches

topfeatures(cleandfmat_BRUNGA)

##       peace development    economic    security      rights      social 
##         342         314         281         197         184         181 
##       human   democracy cooperation       trade 
##         179         151         147         135

Scaling all 37 speeches

## 
## Call:
## textmodel_wordfish.dfm(x = cleandfmat_BRUNGA)
## 
## Estimated Document Positions:
##                        theta      se
## UNGA40Sarney.pdf     1.49378 0.02416
## UNGA41Sodre.pdf      1.54413 0.02486
## UNGA42Sodre.pdf      1.37687 0.03591
## UNGA43Sodre.pdf      1.30132 0.03642
## UNGA44Sarney.pdf     1.25324 0.03291
## UNGA45Collor.pdf     1.79450 0.02515
## UNGA46Collor.pdf     1.74891 0.02676
## UNGA47Lafer.pdf      1.09047 0.03794
## UNGA48Amorim.pdf     0.71158 0.04149
## UNGA49Amorim.pdf     0.62423 0.04603
## UNGA50Lampreia.pdf   0.43760 0.04680
## UNGA51Lampreia.pdf   0.66355 0.05224
## UNGA52Lampreia.pdf   0.51599 0.05509
## UNGA53Lampreia.pdf   0.48512 0.05208
## UNGA54Lampreia.pdf   0.57424 0.05488
## UNGA55Lampreia.pdf   0.04149 0.05598
## UNGA56Cardoso.pdf   -0.08729 0.05808
## UNGA57Lafer.pdf      0.13048 0.07300
## UNGA58Lula.pdf      -0.66084 0.04157
## UNGA59Lula.pdf      -0.65221 0.04703
## UNGA60Amorim.pdf    -0.43531 0.05061
## UNGA61Lula.pdf      -0.76259 0.04597
## UNGA62Lula.pdf      -1.05695 0.04092
## UNGA63Lula.pdf      -1.00936 0.04197
## UNGA64Lula.pdf      -1.22956 0.03445
## UNGA65Amorim.pdf    -0.65427 0.04423
## UNGA66Rousseff.pdf  -0.66668 0.04597
## UNGA67Rousseff.pdf  -0.67697 0.04449
## UNGA68Rousseff.pdf  -0.74938 0.04242
## UNGA69Rousseff.pdf  -0.96306 0.03497
## UNGA70Rousseff.pdf  -1.20361 0.03212
## UNGA71Temer.pdf     -0.58541 0.04785
## UNGA72Temer.pdf     -0.65415 0.05005
## UNGA73Temer.pdf     -0.74033 0.04575
## UNGA74Bolsonaro.pdf -1.54536 0.02075
## UNGA75Bolsonaro.pdf -1.45417 0.02991
## 
## Estimated Feature Scores:
##      regular   1985    eyes suffering mexico   just undergone landed     see
## beta  0.4379  2.531  0.7843    0.2867  1.207 0.4589     1.180  1.944  0.5135
## psi   0.2841 -3.063 -1.3365   -0.3198 -2.316 0.8400    -3.542 -5.107 -0.2426
##      tragedy happened  affirm  people solidarity   dare   hope conveyed
## beta  0.2374    1.945  0.7871 0.05543     0.1126  1.883 0.2772    1.944
## psi  -1.4421   -4.010 -2.5170 1.24015     0.5259 -3.644 0.4777   -5.107
##      feelings   rest   begin statement assuring   entire tribune instills
## beta   0.9241  0.183  0.8343   -0.1039   0.1175  0.08186   1.135    2.428
## psi   -2.3981 -1.431 -1.8582   -0.9047  -2.9246 -0.61823  -2.806   -5.059
##      respect dignity loftiest  nation mighty
## beta  0.1617  0.1446    1.843  0.5088  1.390
## psi   0.6799 -0.3634   -3.594 -0.4046 -3.065

Graph 1

textplot_scale1d(tmod_BRUNGA) + 
  theme_economist() +
  geom_hline(aes(yintercept = 0), col = "red", linetype = "longdash")

Indexing Sarney UNGA 40 and Bolsonaro UNGA 74

Establishing reference scores for Sarney UNGA 40 and Bolsonaro UNGA 74

Modeling Wordscores for Sarney UNGA 40 and Bolsonaro UNGA 74

Retrieving Wordscores

pred_ws_Sarney40Bolsonaro74 <- predict(ws_Sarney40Bolsonaro74, rescaling = "mv")
pred_ws_Sarney40Bolsonaro74

##    UNGA40Sarney.pdf     UNGA41Sodre.pdf     UNGA42Sodre.pdf     UNGA43Sodre.pdf 
##        -1.000000000        -0.258146089        -0.233624688        -0.148348564 
##    UNGA44Sarney.pdf    UNGA45Collor.pdf    UNGA46Collor.pdf     UNGA47Lafer.pdf 
##        -0.220405744        -0.165319750        -0.165493181        -0.132578947 
##    UNGA48Amorim.pdf    UNGA49Amorim.pdf  UNGA50Lampreia.pdf  UNGA51Lampreia.pdf 
##        -0.087136904        -0.137537304        -0.144360123        -0.101553756 
##  UNGA52Lampreia.pdf  UNGA53Lampreia.pdf  UNGA54Lampreia.pdf  UNGA55Lampreia.pdf 
##        -0.127297894        -0.051779930        -0.119934407        -0.042397014 
##   UNGA56Cardoso.pdf     UNGA57Lafer.pdf      UNGA58Lula.pdf      UNGA59Lula.pdf 
##        -0.113909826        -0.124694215        -0.118447369        -0.120183023 
##    UNGA60Amorim.pdf      UNGA61Lula.pdf      UNGA62Lula.pdf      UNGA63Lula.pdf 
##        -0.051724896        -0.075265738         0.014892868        -0.068437664 
##      UNGA64Lula.pdf    UNGA65Amorim.pdf  UNGA66Rousseff.pdf  UNGA67Rousseff.pdf 
##        -0.024335844        -0.043858479        -0.120172500        -0.103245226 
##  UNGA68Rousseff.pdf  UNGA69Rousseff.pdf  UNGA70Rousseff.pdf     UNGA71Temer.pdf 
##        -0.045825247        -0.023278044         0.007985528        -0.051577371 
##     UNGA72Temer.pdf     UNGA73Temer.pdf UNGA74Bolsonaro.pdf UNGA75Bolsonaro.pdf 
##        -0.042275574        -0.038585296         1.000000000         0.194818212

Graph 2

Plotting Wordscores

BRUNGA$ws_Sarney40Bolsonaro74 = pred_ws_Sarney40Bolsonaro74
BRUNGA %>% ggplot (aes(x=ws_Sarney40Bolsonaro74, y=reorder(doc_id, -ws_Sarney40Bolsonaro74))) + geom_point()

Graph 3

Comparing all speeches from 1985 to 2020 for frequency of words

textplot_scale1d(tmod_BRUNGA, margin = "features", highlighted = c("development", "nuclear", "peace", "social", "democracy", "god", "gender", "women", "indebtedness", "religion", "liberalism", "socialism", "indigenous"), alpha = 1) + labs(title = "Estimated word positions for speeches by the Brazil \nat UNGA sessions") + theme_economist()

Graph 4

set.seed(132)
textplot_wordcloud(cleandfmat_BRUNGA, max_words = 100, colors = RColorBrewer::brewer.pal(8,"Dark2"), random_order = FALSE,
                   rotation = .25, )

## Warning: colors is deprecated; use color instead

Graph 5

size <- log(colSums(dfm_select(cleandfmat_BRUNGA, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_BRUNGA, 30))
fcm_select(fcmat_BRUNGA, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

Graph 6

tstat_dist <- as.dist(textstat_dist(cleandfmat_BRUNGA))
clust <- hclust(tstat_dist)
plot(clust, xlab = "Distance", ylab = NULL)

The unsupervised method (Graph 1) shows that the majority of speeches fall between theta -1 and 1, indicating a relatively stable policy position across the different administrations since redemocratization of Brazil. There are, however, some notable variations in policy positions, as follows:

  • in the early years of the redemocratization (during the administrations of Presidents Sarney and Collor, from UNGAs 40 to 47), where theta >= 1;
  • in the year before last of administration Lula (UNGAs 64) and in the last year of the administration Rousseff (UNGA 70), where theta <= -1;
  • in the two years of Bolsonaro’s administration (UNGAs 74 and 75), where theta <= -1.

It is possible to notice a variation within Collor’s administration. In his first two years, President Collor’s speeches (UNGAs 45 and 46) scored a theta >= 1; however, in the last year of his mandate, the UNGA47 speech fell within 0 <= theta >= 1. The impeachment of President Collor did not seem to impact the policy position of his successor, President Franco in UNGAs 48 and 49. The administration of Franco registered a 0 <= theta >= 1, as in the last year of Collor’s mandate. This is an important observation, because Collor was impeached and Franco was his successor.

In the same line, FHC’s administration (UNGAs 50 to 54) registered a 0 <= theta >= 1 for most part, except for the last two years (UNGAs 55 and 56), where theta <= 0. From 1992 to 2000 (a period that corresponds to the last year President Collor’ term to the year before last of President FHC’s term), the policy position of Brazil ranged between 0 <= theta >= 1. The outcome was expected, since these administrations (Collor, Franco and FHC) leaned towards a right or center-right position.

The administrations of Lula (UNGAs 58 to 65), Rousseff (UNGAs 66 to 70) and Temer (UNGAs 71 to 73) continued the tendency showed during the last years of President FHC (UNGAs 55 and 56), where -1 <= theta >= 0. Note, however, that Presidents Lula and Rousseff lean towards a left position, while President Temer lean towards a center-right position. Although the three speeches of administration Temer fell within the -1 <= theta >= 0, it is possible to note that the year after the impeachment of Rousseff, President Temer’s speech in UNGA 71 was closer to zero. President Temer’s speech in UNGA 71, however, was not that different from that of left-wing Lula (UNGA58 and 60). This is an important observation, because Rousseff was impeached and Temer was her successor. In subsequent years, the speeches of President Temer (UNGA 72 and 73) approached those of left-wing Rousseff (UNGA 66, 67 and 68).

The supervised method (Graph 2), which has Bolsonaro’s first speech in the GA (UNGA74) and the first speech of the redemocratization period (Sarney, during the 40th session of UNGA) as indexes, confirms the outcomes of the unsupervised method (Graph 1), while indicating that majority of speeches of Brazil is clustered between a wordscore of -0.25 and 0.25. Once again, the indication is that there is a relative stable policy position across the different administrations since redemocratization of Brazil. The supervised method (Graph 2) also confirms that the highest variations are during the redemocratization period (UNGAs 41, 42 and 43, with the exception of UNGA 43) and the second intervention of President Bolsonaro (during the 75th session of UNGA). The estimated word positions (Graph 3) give some hints regarding the variations in the speeches during the redemocratization and the Bolsonaro periods. The right side of Graph 3 indicates that, in the early years redemocratization, there were concerns with the Latin America debt crises of the 1980s and the reopening of the Brazilian economy after years of military rule. The left side of Graph 3 indicates that the Bolsonaro administration seemed rather concerned with matters related to indigenous groups, gender issues, religion and socialism. Overall, the top center of Graph 3 confirms that peace and development were the main thematic axes of the speeches of Brazil since 1984, as wel as concerns with social, democracy and nuclear matters.

The cluster dendrogram (Graph 6) also shows, on the basis of the tree, that the vast majority of speeches are similar among them (UNGA 42, 43, 44, 49, 51, 52, 53, 54, 55, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 72, 74 and 75), regardless of the political affiliation. Clustered together on the basis of the tree are left-wing administrations, such as Lula’s and Rousseff’s, and center and center-right administrations, such as Sarney’s, Franco’s, FHC’s, Temer’s. The speeches of the first 2 years of the redemocratization period (UNGAs 40 and 41), on the top left of the tree, are the most distant from the rest of the speeches of Brazil in UNGA. The first speech of Bolsonaro in the UNGA is also one of the most distant from the rest of the speeches. The third branch of the tree (UNGAs 45, 46, 47, 48 and 50) refers to the early years of the redemocratization and is the third most distant cluster from the rest of the speeches. One important observation is that Bolsonaro’s second speech (which was indexed in the supervised method - Graph 2) is not as distant from the rest of majority of the speeches.

The top features method (Table 1) once again confirms that peace and development are the two thematic axes of the 37 speeches of Brazil in the UNGA since redemocratization. From the plot of the frequency of words in the Brazilian speeches in the General Assembly since the redemocratization (Table 1), the words peace (342 observations) and development (314 observations) are the most frequently mentioned. The wordcloud graph (Graph 4) confirms the tendency and adds economic, political, social, security, rights and human as other frequent words.

The outcome of the network graph (Graph 5) shows that development, developing and developed are nodal points. Development is related to security, rights, trade, south, Latin America and sustainable. It is noticeable that peace does not appear as a nodal point. That is the case, because the network of co-occurrences graph is based on semantic distance between words. In this sense, the networks of co-occurrences shows the most frequent chunk of words after removing stopwords and applying stemming. Note that security emerges as a nodal point and is related to council and reform. The reform of the Security Council has been another priority of the Brazilian diplomacy (Pinheiro, 2004; Cervo & Bueno, 2002; Mendonça, 2013).

5. Robustness tests

As robustness tests, I will compare all 37 the speeches by administration’s political affiliation and will assess the frequency of words and co-occurrence networks for each individual administration. I will also compare the relative frequency of words by administration and lexical dispersion of the words development and peace in the corpus of the 37 speeches year over year.

Comparing all speeches from 1985 to 2020 by indexing Lula (left) and Bolsonaro (far-right):

Indexing Lula and Bolsonaro

Establishing reference scores for Lula and Bolsonaro

Modeling Wordscores for Bolsonaro 74 UNGA and Bolsonaro 74 UNGA

Retrieving Wordscores

pred_ws_LulaBolsonaro <- predict(ws_LulaBolsonaro, rescaling = "mv")
pred_ws_LulaBolsonaro

##    UNGA40Sarney.pdf     UNGA41Sodre.pdf     UNGA42Sodre.pdf     UNGA43Sodre.pdf 
##          -0.5912376          -0.5904842          -0.6914821          -0.5241864 
##    UNGA44Sarney.pdf    UNGA45Collor.pdf    UNGA46Collor.pdf     UNGA47Lafer.pdf 
##          -0.5662680          -0.5413491          -0.6229430          -0.5746850 
##    UNGA48Amorim.pdf    UNGA49Amorim.pdf  UNGA50Lampreia.pdf  UNGA51Lampreia.pdf 
##          -0.5457995          -0.5737767          -0.6157445          -0.5155816 
##  UNGA52Lampreia.pdf  UNGA53Lampreia.pdf  UNGA54Lampreia.pdf  UNGA55Lampreia.pdf 
##          -0.6279497          -0.5537841          -0.6395312          -0.5744135 
##   UNGA56Cardoso.pdf     UNGA57Lafer.pdf      UNGA58Lula.pdf      UNGA59Lula.pdf 
##          -0.6396015          -0.6447301          -1.0173184          -0.9681899 
##    UNGA60Amorim.pdf      UNGA61Lula.pdf      UNGA62Lula.pdf      UNGA63Lula.pdf 
##          -0.9448528          -0.9490014          -1.0203013          -1.0990778 
##      UNGA64Lula.pdf    UNGA65Amorim.pdf  UNGA66Rousseff.pdf  UNGA67Rousseff.pdf 
##          -1.0185647          -0.9870661          -0.6949221          -0.5567390 
##  UNGA68Rousseff.pdf  UNGA69Rousseff.pdf  UNGA70Rousseff.pdf     UNGA71Temer.pdf 
##          -0.5791896          -0.4818133          -0.5903937          -0.5356510 
##     UNGA72Temer.pdf     UNGA73Temer.pdf UNGA74Bolsonaro.pdf UNGA75Bolsonaro.pdf 
##          -0.5499090          -0.4979098           1.0000000           1.0000000

Graph 7

Plotting Wordscores

BRUNGA$ws_LulaBolsonaro = pred_ws_LulaBolsonaro
BRUNGA %>% ggplot (aes(x=ws_LulaBolsonaro, y=reorder(doc_id, -ws_LulaBolsonaro))) + geom_point()

The robustness test (Graph 7) confirms that the majority of the speeches are clustered and shows little variation. While indexing Lula’s (left) and Bolsonaro’s (far-right) administrations for political affiliation, the speeches of all other periods (Collor, Franco, FHC, Rousseff and Temer) fall between wordscores -0.75 and -0.5. Graph 7 also shows variations in the speeches in UNGAs 59, 60, 61, 63, 65, during Lula’s administration. It is also noticeable that the majority of speeches clustered around wordscores -0.75 and -0.5 is closer to the speeches of Lula’s administration than to those of Bolsonaro’s.

Sarney (1985-1990):

Table 2

topfeatures(cleandfmat_Sarney)

##       peace    economic       latin     america development cooperation 
##          88          61          52          48          42          37 
##      people   democracy  developing     peoples 
##          34          33          29          28

Graph 8

size <- log(colSums(dfm_select(cleandfmat_Sarney, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Sarney, 30))
fcm_select(fcmat_Sarney, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

The top features function (Table 2) show that the most frequent words in the speeches of Brazil under Sarney’s administration was peace, followed by economic, Latin America, development, cooperation, people, political, democracy and developing. If taken together, development and developing appear 71 times. Therefore, development and peace were two most spoken words in Brazil’s speeches in the UNGA 40, 41, 42, 43 and 44.

The words development, developing, developed are also central nodal points in the network of feature co-occurrences (Graph 8) under the Sarney’s administration and the early years of the redemocratization of Brazil. Here, I can note that development is associated with Latin America and the South, economy, security and peoples.

The outcome of the network graph (Graph 8) also show that Brazil has prioritized in international fora the economic development and the security of Latin America and the South. Other co-occurrences with Latin America are integration and trade. In fact, the economic integration with Latin America during Sarney’s administration was one vector of development, as historians point out (Cervo and Bueno, 2002). Although peace is one of the most frequently spoken word along with development (and variations), peace does not appear as a nodal point in Graph 8, as it is not captured as a frequent chunk of words in these speeches.

It is also worth mentioning that the debt crisis that Latin America has faced in the 1980s is reflected in the speeches of Brazil, as well as concerns with wars, bomb and nuclear (weapons/threat). The Latin America debt crisis and the ever present nuclear threat during the Cold War years are a reflection of the time in the Brazilian speeches in the UN.

Collor (1990-1992):

Table 3

topfeatures(cleandfmat_Collor)

##       peace development    economic   democracy cooperation      rights 
##          49          49          36          28          26          25 
##      social   universal  democratic    peaceful 
##          25          23          18          18

Graph 10

size <- log(colSums(dfm_select(cleandfmat_Collor, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Collor, 30))
fcm_select(fcmat_Collor, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

Once again, the top features function (Table 3) has peace and development as the most frequent words in Brazil’s speeches in UNGAs 45, 46 and 47, under Collor’s administration. It is worth noting, however, that economic becomes the third most relevant word in President Collor’s speeches, reflecting the priority of his neoliberal agenda.

The feature co-occurrence network graph (Graph 10) shows that development however ceased to be associated with Latin America and the South, pointing to a change of perception regarding development. In other words, development is no longer a project for the Southern Hemisphere, in an attempt to bridge the Cold War divide between the developed North and the underdeveloped South.

In President Collor’s years, development is rather linked to economic, social, human, democratic, technology, disarmament, weapons, security, peaceful, disarmament, progress, trade, environmentally and sustainable, among other connections. The links associated with the nodal point development corroborates the argument that the new model of development adopted by President Collor prioritized access to markets and technology, as well as the abandonment of any aspiration for Brazil to become a nuclear power. Although peace is not a nodal point, peaceful is and is associated with nuclear, Argentina, democratic, social, economic, development. The lexicon reflects Collor’s turn to the peaceful uses of the nuclear technology and the agreement with Argentina on the matter, one of the main legacies of his administration.

Franco (1992-1995):

Table 4

topfeatures(cleandfmat_Franco)

## development       peace      rights       human       south cooperation 
##          33          27          25          23          21          16 
##  democratic   democracy     society    economic 
##          14          13          13          13

Graph 12

size <- log(colSums(dfm_select(cleandfmat_Franco, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Franco, 30))
fcm_select(fcmat_Franco, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

Despite the political rupture provoked by the impeachment of President Collor, peace and development remain the two most important features in the speeches of President Franco in UNGAs 48 and 49 (Table 4). A novelty is the shift of emphasis from economic to human rights in the outcome of the top features function (Table 4).

The co-occurrence network graph (Graph 12) shows that development is still associated with economic and, as in Sarney’s years, Latin America, Africa and national. However, development gains new connotations, being also associated with human, rights and conference. Under President Franco, Brazil hosted the United Nations Conference on Environment and Development (emphasis added), also known as the Rio de Janeiro Earth Summit. In 1992, Brazil also signed the International Covenant on Economic, Social and Cultural Rights and the International Covenant on Civil and Political Rights. The country has also signed the Organization of American States’ Convention on Human Rights, also known as Pact of San José da Costa Rica.

Although a frequent word in the wordcloud, peace is not a nodal point in the network graph (Graph 12), as it is not captured as a frequent chunk of words in these speeches.

FHC (1995-2003):

Table 5

topfeatures(cleandfmat_FHC)

##       peace    economic development    security       human       trade 
##          68          65          64          55          39          38 
##     council      reform cooperation   democracy 
##          35          33          32          30

Graph 14

size <- log(colSums(dfm_select(cleandfmat_FHC, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_FHC, 30))
fcm_select(fcmat_FHC, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

The triad peace, economic and development becomes more salient in the speeches of Brazil in UNGAS 50 to 57 (Table 5). Under President FHC, economic has even more prominence than during the Collor’s years, as the top features function shows (Table 5). The co-occurrence network graph (Graph 14) shows that economic is tied to free and trade to freedom, reflecting the neoliberal agenda of the FHC years. In the same vein, developing is linked to trade, environment, freedom and challenges. At this point in time, there were various attempts to negotiate free trade zones with the Americas (ALCA) and the EU (Mercorsur and EU). It is also noticeable that the reform of the Security Council gained prominence in the speeches of the FHC administration, reflecting an attempt to project the country globally. Although a frequent word in the wordcloud, peace is not a nodal point in the network graph, as it is not captured as a frequent chunk of words in these speeches.

Lula (2003-2011):

Table 6

topfeatures(cleandfmat_Lula)

##       peace      social    economic development      hunger    security 
##          57          57          55          53          45          40 
##       south     poverty     council      people 
##          38          36          34          31

Graph 16

size <- log(colSums(dfm_select(cleandfmat_Lula, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Lula, 30))
fcm_select(fcmat_Lula, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

The left-wing administration of President Lula added the word social to the triad development, peace and economic, which were present in Collor’s and FHC’s speeches in the UN (Table 6). The concern with social issues is also reflected by the frequency of the words hunger and poverty (Table 6). The top features function shows that the word social comes in second place in frequency, along with peace.

The novelty of the concern with social issues in the speeches of Brazil in the UNGAs 58 to 65 can explain that the theta of the Lula administration was mostly between zero and -1, in the comparison of all 37 speeches of Brazil after the redemocratization (Graph 1). The fact that the administration Lula maintained the triad development, peace and economic can also explain that some speeches of the neoliberal administration FHC (UNGAs 55 and 56) are close to those of Lula’s left-wing period (Graph 1).

The network graph (Graph 16) also reveals that development and developing are linked to the Doha Development Round (emphasis added) of the World Trade Organization (WTO). The graph also reveals other important aspects of the Brazilian diplomacy under Lula, such as the Brazilian command of the military component of the UN peacekeeping mission in Haiti and the Brazilian engagement with the establishment of the G20. Although a frequent word in the wordcloud, peace is not a nodal point in the network graph, as it is not captured as a frequent chunk of words in these speeches.

Rousseff (2011-2016):

Table 7

topfeatures(cleandfmat_Rousseff)

## development    economic      social     poverty      crisis       peace 
##          38          28          28          28          27          27 
##      rights    security       women     council 
##          26          25          23          21

Graph 18

size <- log(colSums(dfm_select(cleandfmat_Rousseff, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Rousseff, 30))
fcm_select(fcmat_Rousseff, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

Development was the flagship of Rousseff’s administration, as her two National Development Plans evidence. Her concern with development is clearly reflected in her speeches from UNGAs 66 to 70 (Table 7). Like her predecessor, Rousseff’s administration also evidences a concern with social issues in her interventions in the United Nations, with the words social and poverty appearing frequently.

One very distinct feature of the speeches of President Rousseff is the frequency of the word women, which gain prominence for the first time in Brazil’s interventions in the United Nations (Table 7). It is noticeable that President Rousseff was the first female Head of State and Government of Brazil and, as such, the first woman to ever open a General Debate of the United Nations.

Development and its variations is linked with education, poverty, rights, peace, security, public, climate, sustainable and Rio (Graph 18). Under President Rousseff’s administration, Brazil hosted the UN Conference on Sustainable Development in Rio de Janeiro (emphasis added). The network graph also reveals other important aspects of the Brazilian diplomacy under Rousseff, such as the Olympic Games in Rio. Peace appears as a nodal point in the network graph and is linked to Syrian, information, security, council, development and rights, indicating that Brazil’s speeches are now addressing peace from a more global than local or regional perspective.

Temer (2016-2019):

Table 8

topfeatures(cleandfmat_Temer)

## development     nuclear    security       peace      rights       human 
##          26          20          18          17          16          14 
##       trade   diplomacy    dialogue    economic 
##          13          12          11          11

Graph 20

size <- log(colSums(dfm_select(cleandfmat_Temer, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Temer, 30))
fcm_select(fcmat_Temer, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

Despite the political rupture provoked by the impeachment of President Rousseff, development remained the most important feature in the speeches of President Temer in UNGAs 71 to 73 (Table 8). However, peace lost some of its prominence, appearing in fourth place after nuclear and security (Table 8); however, peace is still one of the most frequent words. Additionally, the social concern has also disappeared, which evidences the shift back to a center-right, liberal politics. Peace, security and nuclear gain prominence due to a global nuclear threat that started to loom in the horizon with North Korea and Iran and to the conclusion of the Treaty on the Prohibition of Nuclear Weapons (TPNW). Economic and trade regain importance, as per top features function (Table 8). Development is again associated to economic (Graph 20). It is noticeable that women loses relevance in Temer’s speeches. Pacific and peace are nodal points and linked to human, agriculture, democracy, economic, integration, nuclear, development, security. The links with peace (Graph 20) seem very general and connected to global politics, except for the reference to agriculture (one of the main economic activities in Brazil).

Bolsonaro (2019-present):

Table 9

topfeatures(cleandfmat_Bolsonaro)

##    indigenous       freedom      economic         human        rights 
##            31            14            12            12            11 
## environmental       peoples        amazon        people   development 
##            11            11            10             9             9

Graph 22

size <- log(colSums(dfm_select(cleandfmat_Bolsonaro, selection = "keep")))
set.seed(144)
feat <- names(topfeatures(fcmat_Bolsonaro, 30))
fcm_select(fcmat_Bolsonaro, pattern = feat) %>%
    textplot_network(min_freq = 0.5)

As Graphs 1 and 6 showed, the speech of President Bolsonaro in UNGAs 74 is the most distant from all the others during the redemocratization period. The robustness tests confirm the trend. For the first time since 1985, development and peace cease to have prominence in Brazil’s speeches in the UNGA (Table 9). Development is the 10th most spoken word and peace does not appear in the top features table (Table 9). Indigenous appears as the most spoken word for the first time in the top features and wordcloud and is associated with development, the Amazon, environmental, freedom, progress, peoples, invaded, territory, sovereignty and media. Peace is a nodal point and is linked to indigenous, environmental, security, Israel and peacekeeping (Graph 22). In this light, peace seems connected to indigenous or territorial issues (possibly sovereignty in the Amazon region), but also to global politics (Israel, as the main allied of Bolsonaro’s administration).

Peace and Development

Graph 23

dfm_weight <- dfmatclean_BRUNGA

# Calculate relative frequency by administration
freq_weight <- textstat_frequency(dfm_weight, n = 5, 
                                  groups = dfm_weight$Administration)

ggplot(data = freq_weight, aes(x = nrow(freq_weight):1, y = frequency)) +
     geom_point() +
     facet_wrap(~ group, scales = "free") +
     coord_flip() +
     scale_x_continuous(breaks = nrow(freq_weight):1,
                        labels = freq_weight$feature) +
     labs(x = NULL, y = "Relative frequency")

The relative frequency graph (Graph 23) shows that, in relative terms, the left-wing governments of Lula and Rousseff did not abandon the triad peace, development and economic of the Collor’s and FHC’s administration, adding to it a social concern. During the early years of redemocratization (Sarney’s years), development was less prominent in relative terms. In the same vein, peace and development were less prominent under Bolsonaro in relative terms.

Graph 24

data_corpus_inaugural_subset <- 
    corpus_subset(corp_BRUNGA, Year > 1984)

    
toks <- tokens(data_corpus_inaugural_subset)
textplot_xray(
    kwic(toks, pattern = "development"),
    kwic(toks, pattern = "peace"),
    scale = "absolute"
)

## Warning: Use of `x$ntokens` is discouraged. Use `ntokens` instead.

The dispersion plot (Graph 24) shows the lexical dispersion of the words development and peace in the corpus of the 37 speeches year over year. Graph 24 is useful to show how homogeneous the employment of these two words have been over time across the parts of the corpus. The output of graph 24 indicates that both development and peace are widely dispersed across the texts of the speeches of Brazil in the UNGA over time.

6. Conclusion

The NLP method corroborates the hypotheses that the speeches of Brazil from UNGAs 40 to 75 have remained relatively stable and centered around the two main thematic axes: peace and development.

Graph 1 (unsupervised method) shows that the majority of speeches falls between theta -1 and 1, indicating a relatively stable policy position across the eight administrations. The most noticeable variations are observed during the early years of the redemocratization period and of the Bolsonaro administration. The two moments of political rupture (after Collor’s and Rousseff’s impeachment) did not show any radical transformation, although the Temer’s speech suffered some small adjustments becoming closer to Lula’s period (UNGA58 and 60) immediately after President Rousseff was toppled but approached Rousseff’s speeches in subsequent years.

Additionally, Graph 2 (supervised method) indicates that majority of speeches of Brazil is clustered between a wordscore of -0.25 and 0.25, when indexing Bolsonaro’s first speech in the GA (UNGA74) and the first speech of the redemocratization period (Sarney, during the 40th session of UNGA). Graph 2 confirms that the highest variations are during the redemocratization period (UNGAs 41, 42 and 43, with the exception of UNGA 43) and the Bolsonaro’s administration (during the 75th session of UNGA).

Finally, Graph 6 (Cluster Dendrogram) shows that the vast majority of speeches are similar among them (UNGA 42, 43, 44, 49, 51, 52, 53, 54, 55, 59, 60, 61, 62, 63, 64, 65, 66, 67, 68, 69, 70, 71, 72, 72, 74 and 75), regardless of the political affiliation. The speeches of the first 2 years of the redemocratization period (UNGAs 40 and 41) and the first speech of Bolsonaro in the UNGA were the most distant from the rest of the speeches of Brazil in UNGA.

As Graph 3 (Estimated word positions) suggests, in the early years redemocratization, there was a concern with the Latin America debt crises of the 1980s and the reopening of the Brazilian economy after years of military rule, whereas, in the Bolsonaro years, there were concerns with matters related to indigenous groups, gender issues, religion and socialism.

The top features method (Table 1) once again confirms that peace and development are the two thematic axes of the 37 speeches of Brazil in the UNGA since redemocratization: the words peace (342 observations) and development (314 observations) are the most frequently mentioned in all 37 speeches.

The outcome of the network graph (Graph 5) shows that development, developing and developed are nodal points and related to security, rights, trade, south, Latin America and sustainable. Peace however does not appear as a nodal point, because chunk of words containing peace are not frequent in the 37 speeches, after removing stopwords and applying stemming.

As robustness tests, I compared all 37 the speeches by administration’s political affiliation (Graph 7) and assessed the frequency of words (Tables 2 to 9) and co-occurrence networks (Graphs 8, 10, 12, 14, 16, 18, 20 and 22) for each individual administration. I also compared the relative frequency of words by administration (Graph 23) and lexical dispersion of the words development and peace in the corpus of the 37 speeches year over year (Graph 24).

Graph 7 (supervised method by political affiliation for all 37 speeches) confirmed that the majority of the speeches is clustered and showed little variation, after indexing Lula’s left wing administration and Bolsonaro’s far-right administration. It is also noticeable that the majority of speeches clustered around wordscores -0.75 and -0.5 is closer to the speeches of Lula’s administration than to those of Bolsonaro’s.

Although development remained a constant feature in the speeches from one administration to another with the exception of Bolsonaro’s period, there have been some variations as to which word development was associated with. In the early years of the redemocratization process, development was linked with Latin America and the South, economy, security and peoples (Graph 8). Indeed, development, under Sarney’s administration, was a project for the Southern Hemisphere, in an attempt to bridge the Cold War divide between the developed North and the underdeveloped South. Afterwards, development became associated with economy and trade in the neoliberal administrations of Collor and FHC (Graphs 10 and 14). The economic integration with Latin America, which was possible after the nuclear agreement between Argentina and Brazil, was the main legacy of Collor’s years. As for FHC, the priority was the negotiation of free trade areas, as the ALCA and the Mercosur-EU agreement. Franco’s speeches recovers the link between development and Latin America and introduces a new link with human rights and cooperation (Graph 12). During Franco’s years, Brazil signed the UN Covenants on human rights and the OAS Convention on the same matter. Another variation was introduced by the left-wing administrations of Presidents Lula and Rousseff, which added a link between development and social (Graphs 16 and 18). It was the first time that social issues were linked to development and that reflects the leftist inclination of Lula’s and Rousseff’s administrations.

Peace has also remained a constant feature, appearing frequently in the speeches of Brazil in the UNGA over time. Peace and variations (peaceful, pacific) appeared as nodal points only in Collor’s, Rousseff’s, Temer’s and Bolsonaro’s administrations (Graphs 10, 18, 20 and 22). Peace was linked with nuclear, Argentina, democratic, social, economic and development in Collor’s administration and reflects his foreign policy priority of signing an agreement on the peaceful uses of the nuclear technology with Argentina and of abandoning any ambitions of a nuclear program other than for peaceful purposes for Brazil (Graph 10). Under Rousseff’s administration, peace is associated with Syrian, information, security, council, development and rights, indicating that Brazil’s speeches were addressing peace from a more global than local or regional perspective (Graph 18). The tendency is maintained in Temer’s administration (Graph 20), except for the connection between peace and agriculture (one of the main economic activities in Brazil). Under Bolsonaro’s administration, peace seems to be connected to both global (Israel) and local (indigenous) politics (Graph 22).

Peace and development clearly lose relevance in President Bolsonaro’s speeches in the UNGAs 74 and 75 (Table 9 and Graph 22). While emphasizing aspects related to the Amazon, indigenous people and territory, Bolsonaro’s speeches represent the highest variation when compared to all other interventions since the redemocratization of Brazil.

The relative frequency graph (Graph 23) shows that the left-wing governments of Lula and Rousseff did not abandon the triad peace, development and economic of the Collor’s and FHC’s administration, adding to it a social concern. The novelty of the concern with social issues in the speeches of Brazil in the UNGAs 58 to 65 can explain that the theta of the Lula administration was mostly between zero and -1, in the comparison of all 37 speeches of Brazil after the redemocratization (Graph 1). The fact that the administration Lula maintained the triad development, peace and economic can also explain that some speeches of the neoliberal administration FHC (UNGAs 55 and 56) are close to those of Lula’s left-wing period (Graph 1). Such a trend explains the little variation from one administration to another, regardless of political affiliation. During the early years of redemocratization (Sarney’s years), development was less prominent in relative terms; similarly, peace and development were less prominent under Bolsonaro in relative terms (Graph 23).

Overall, the output of graph 24 (lexical dispersion) indicates that both development and peace are widely dispersed across the speeches of Brazil in the UNGAs.

7. References

Benoit, K. et al. (2018) quanteda: An R package for the quantitative analysis of textual data. Journal of Open Source Software. [Online] 3 (30), 774.

Cervo, A. L. & Bueno, C. (2002) História da política exterior do Brasil. 4a edição revista e ampliada. Brasília, DF: Editora UnB.

Corrêa, L. F. de S. & Fundação Alexandre de Gusmão (eds.) (2013) Brazil in the United Nations, 1946 - 2011. Third edition. Brasília: Fundação Alexandre de Gusmão.

Dag Hammarskjold Library (n.d.) Ask DAG! [online]. Available from: https://ask.un.org/faq/70473 (Accessed 23 May 2021).

Garcia, E. V. (2012) O sexto membro permanente: o Brasil e a criação da ONU. 1a ed. Rio de Janeiro, RJ: Contraponto.

Mendonça, R. (2013) História da política exterior do Brasil: do período colonial ao reconhecimento do império (1500-1825). Brasília, DF: Fundação Alexandre de Gusmão.

Pinheiro, L. (2004) Política externa brasileira (1889-2002). Rio de Janeiro, RJ: Jorge Zahar Editor.

UCLSPP (n.d.) Advanced Quantitative Methods. [online]. Available from: https://uclspp.github.io/PUBLG088/seminar7.html (Accessed 23 May 2021).

Associate Professor of International Relations and Political Science

International institutions and political networks.

comments powered by Disqus