[pt] No início de 1995, dando continuidade ao processo de combate inflacionário, o regime cambial brasileiro passou a ser um Crawling Peg. Assim, a política monetária deveria ser fixada de forma passiva, e caberia a política fiscal o controle da demanda doméstica e, por conseguinte, a manutenção da estabilidade do nível de preços. Entretanto, Salgado, Garcia e Medeiros (2005) encontram fortes evidências de que a política monetária foi orientada para o controle da inflação compensando a política fiscal frouxa praticada pelo governo, aplicando um modelo TAR (Threshold Autoregressive Model). A especificação proposta pelos autores consiste na estimação de modelos lineares distintos dependendo do estado da economia (em crise ou fora de crise). Períodos de crise são determinados por uma variável observada (variação acumulada de reservas). Neste trabalho, utilizamos um modelo com chaveamento Markoviano para dinâmica da taxa de juros nominal onde a determinação de períodos de crise é feita por uma variável não-observada. Além disso, procuramos resolver possíveis problemas de endogeneidade. Finalmente, motivamos a estimação através de modelos econômicos teóricos cujas dinâmicas são compatíveis com uma regra de fixação de juros não-linear. Os resultados encontrados no trabalho foram bastante satisfatórios. / [en] In the beginning of 1995, continuing the process of inflation combat, the exchange rate regime adopted was a Crawling Peg. Thus, the monetary policy should have been passive and the fiscal policy would have been directed to control the aggregate demand, affecting the price level stability. However, Salgado, Garcia e Medeiros (2005) find strong evidences in Brazilian data that the monetary policy was used to control inflation, compensating the loose fiscal policy implemented by the Central Government, using a TAR Model (Threshold Autoregressive Model). The specification proposed by the authors consists in an estimation of two independent linear models for different states of the nature (crises or not crises). An observable variable (the accumulated changes in the International Reserve) determines the crises periods. Here we estimate a model where the dynamic of the nominal interest rate follows a Markov Switching process and the regimes are unobservable variables. Furthermore, we try to handle the possible problem of endogenity. Finally, we motivate the estimations analyzing models where the theoretical dynamics of the economy are compatible with a nonlinear interest rate rule. Our results are quiet satisfactory.

[pt] Estimamos um modelo DSGE para a economia brasileira abrangendo a transição do regime de bandas cambiais para o regime de metas para a inflação ocorrida em 1999. Utilizamos um modelo novo keynesiano de pequena economia aberta no qual o Banco Central segue uma regra de política monetária estruturalmente distinta em cada regime. Encontramos diferenças significativas na dinâmica macroeconômica e nos mecanismos de transmissão dos choques estruturais, com destaque àqueles relacionados ao setor externo da economia. Realizamos experimentos contrafactuais onde simulamos o impacto de cenários alternativos para a transição de regime na trajetória das variáveis econômicas brasileiras. Entre outros resultados, as simulações sugerem que a manutenção do sistema de bandas cambiais seria insustentável enquanto a antecipação da implementação do regime de metas para a inflação para antes da crise da Rússia de 1998 poderia deixar a economia brasileira em situação mais favorável. Por fim, mostramos que um teste de quebra estrutural aplicado sobre todo o período amostral detecta com precisão a ocorrência da mudança de regime em 1999. / [en] We estimate a DSGE model of the brazilian economy taking into account the transition from the exchange rate band system to inflation targeting occurred in 1999. We use a new Keynesian small open economy model where the Central Bank follows structurally different monetary policy rules in each regime. By comparing the transmission channels of exogenous shocks we find significant differences across the regimes, specially on those shocks related to the foreign sector of the economy.We then perform counterfactual experiments where we simulate the response of key macro variables under alternative scenarios for the regime transition. Among other results our simulations suggest that the continuation of the exchange rate band system could have been unsustainable while anticipating the transition to inflation targeting before the Russian crisis of 1998 could have benefited the economy. Additionally, we show that a structural break test applied to the whole data sample correctly identifies the regime change in 1999.

Τα πρωτότυπα αποτελέσματα της διατριβής περιέχονται στα κεφ. 2, 3 και 4. Στο κεφ. 2 μελετούμε με αριθμητικές μεθόδους ένα 3-διάστατο διακριτό μοντέλο μακροοικονομίας με σταθερές ισοτιμίες. Χρησιμοποιώντας μια μέθοδο πλέγματος, βρίσκουμε την περιοχή ευστάθειας στον παραμετρικό χώρο, προσδιορίζουμε την καμπύλη διακλαδώσεων Hopf-Neimark και θεωρούμε σύντομα την εμφάνιση “γλωσσών” Arnold. Υπολογίζονται διαγράμματα διακλαδώσεων και εκθέτη Lyapunov που δίνουν πληροφορίες για τους επιχειρηματικούς κύκλους και την πολύπλοκη δυναμική του μοντέλου και. παρουσιάζουμε παραδείγματα κυκλικών και χαοτικών ελκυστών. Στο κεφ. 3 μελετούμε με τις ίδιες μεθόδους ένα διακριτό μοντέλο αλληλεπίδρασης περιοχών με σταθερές ισοτιμίες, επέκταση του προηγούμενου μοντέλου σε 5 διαστάσεις. Στόχος ήταν να δείξουμε πόσο εφικτή και αποτελεσματική είναι μία αριθμητική μελέτη για ηπίως πολυδιάστατα διακριτά δυναμικά συστήματα με πολλές παραμέτρους. Βρήκαμε ότι η κίνηση κεφαλαίων δεν αρκεί για τη δημιουργία κύκλων όταν είναι χαμηλή η εμπορική αλληλεπίδραση. Το κατώφλι εμπορίου προβλέπεται περίπου στο 15% των εμπορικών συναλλαγών. Αντίθετα, το μοντέλο δεν προβλέπει αναγκαίο ελάχιστο επίπεδο κίνησης κεφαλαίων για την εμφάνιση των κύκλων. Δίνουμε παραδείγματα διαγραμμάτων διακλάδωσης και εκθέτη Lyapunov που δείχνουν την εμφάνιση κύκλων ή ακολουθίας διπλασιασμού περιόδου, και παραδείγματα της ανάπτυξής τους. Το κεφ. 4 περιέχει σύντομη περιγραφή συμπληρωματικών αποτελεσμάτων στα παραπάνω μοντέλα, και στα αντίστοιχα μοντέλα μεταβλητής ισοτιμίας συναλλάγματος, καθώς και κατευθύνσεις μελλοντικής έρευνας. Στο κεφ. 5 περιγράφονται σύντομα οι υπολογιστικές τεχνικές που χρησιμοποιήσαμε. Η διατριβή δείχνει την αποτελεσματικότητα της αριθμητικής προσέγγισης για πολυδιάστατα διακριτά μοντέλα. / The original results of the dissertation are contained in ch. 2, 3 and 4, and concern mainly the problem of business cycles. In ch. 2 we explore numerically a 3D discrete Kaldorian macrodynamic model of open economy with fixed exchange rates. Using a grid search method we determine the stability region in parameter space, and the Hopf-Neimark bifurcation curve, and discuss briefly the occurrence of Arnold tongues. Bifurcation and Lyapunov exponent diagrams are computed providing information on the business cycles and illustrating the complex dynamics involved. Examples of cycles and chaotic attractors are presented. In ch. 3 we explore a 5D extension of the previous model using the same methods. The aim was to demonstrate the feasibility and effectiveness of the numerical approach for discrete dynamical systems of moderately high dimensionality and several parameters. We found that capital movement is not sufficient to generate interregional business cycles when trade interaction is low. The trade threshold is predicted at about 15% of trade transactions. By contrast, no minimum level of capital mobility exists as a requirement for the emergence of business cycles. Examples of bifurcation and Lyapunov exponent diagrams illustrating the occurrence of cycles or period doubling, and examples of their development, are given. Ch. 4 contains a short description of complementary results on the above models, and on two other models which extend the previous models to the case of flexible exchange rates, as well as some lines of future research. In ch. 5, the computational techniques employed in the present study are briefly described. The dissertation indicates the effectiveness of the numerical approach for high dimensional discrete models.