Model selection, estimation and forecasting in VAR models with short-run and long-run restrictions

We study the joint determination of the lag length, the dimension of the cointegrating space and the rank of the matrix of short-run parameters of a vector autoregressive (VAR) model using model selection criteria. We consider model selection criteria which have data-dependent penalties for a lack of parsimony, as well as the traditional ones. We suggest a new procedure which is a hybrid of traditional criteria and criteria with data-dependant penalties. In order to compute the fit of each model, we propose an iterative procedure to compute the maximum likelihood estimates of parameters of a VAR model with short-run and long-run restrictions. Our Monte Carlo simulations measure the improvements in forecasting accuracy that can arise from the joint determination of lag-length and rank, relative to the commonly used procedure of selecting the lag-length only and then testing for cointegration.

Three essays on local demand for public services

Local provision of public services has the positive effect of increasing the efficiency because each locality has its idiosyncrasies that determine a particular demand for public services. This dissertation addresses different aspects of the local demand for public goods and services and their relationship with political incentives. The text is divided in three essays. The first essay aims to test the existence of yardstick competition in education spending using panel data from Brazilian municipalities. The essay estimates two-regime spatial Durbin models with time and spatial fixed effects using maximum likelihood, where the regimes represent different electoral and educational accountability institutional settings. First, it is investigated whether the lame duck incumbents tend to engage in less strategic interaction as a result of the impossibility of reelection, which lowers the incentives for them to signal their type (good or bad) to the voters by mimicking their neighbors’ expenditures. Additionally, it is evaluated whether the lack of electorate support faced by the minority governments causes the incumbents to mimic the neighbors’ spending to a greater extent to increase their odds of reelection. Next, the essay estimates the effects of the institutional change introduced by the disclosure on April 2007 of the Basic Education Development Index (known as IDEB) and its goals on the strategic interaction at the municipality level. This institutional change potentially increased the incentives for incumbents to follow the national best practices in an attempt to signal their type to voters, thus reducing the importance of local information spillover. The same model is also tested using school inputs that are believed to improve students’ performance in place of education spending. The results show evidence for yardstick competition in education spending. Spatial auto-correlation is lower among the lame ducks and higher among the incumbents with minority support (a smaller vote margin). In addition, the institutional change introduced by the IDEB reduced the spatial interaction in education spending and input-setting, thus diminishing the importance of local information spillover. The second essay investigates the role played by the geographic distance between the poor and non-poor in the local demand for income redistribution. In particular, the study provides an empirical test of the geographically limited altruism model proposed in Pauly (1973), incorporating the possibility of participation costs associated with the provision of transfers (Van de Wale, 1998). First, the discussion is motivated by allowing for an “iceberg cost” of participation in the programs for the poor individuals in Pauly’s original model. Next, using data from the 2000 Brazilian Census and a panel of municipalities based on the National Household Sample Survey (PNAD) from 2001 to 2007, all the distance-related explanatory variables indicate that an increased proximity between poor and non-poor is associated with better targeting of the programs (demand for redistribution). For instance, a 1-hour increase in the time spent commuting by the poor reduces the targeting by 3.158 percentage points. This result is similar to that of Ashworth, Heyndels and Smolders (2002) but is definitely not due to the program leakages. To empirically disentangle participation costs and spatially restricted altruism effects, an additional test is conducted using unique panel data based on the 2004 and 2006 PNAD, which assess the number of benefits and the average benefit value received by beneficiaries. The estimates suggest that both cost and altruism play important roles in targeting determination in Brazil, and thus, in the determination of the demand for redistribution. Lastly, the results indicate that ‘size matters’; i.e., the budget for redistribution has a positive impact on targeting. The third essay aims to empirically test the validity of the median voter model for the Brazilian case. Information on municipalities are obtained from the Population Census and the Brazilian Supreme Electoral Court for the year 2000. First, the median voter demand for local public services is estimated. The bundles of services offered by reelection candidates are identified as the expenditures realized during incumbents’ first term in office. The assumption of perfect information of candidates concerning the median demand is relaxed and a weaker hypothesis, of rational expectation, is imposed. Thus, incumbents make mistakes about the median demand that are referred to as misperception errors. Thus, at a given point in time, incumbents can provide a bundle (given by the amount of expenditures per capita) that differs from median voter’s demand for public services by a multiplicative error term, which is included in the residuals of the demand equation. Next, it is estimated the impact of the module of this misperception error on the electoral performance of incumbents using a selection models. The result suggests that the median voter model is valid for the case of Brazilian municipalities.

A (semi-)parametric functional coefficient autoregressive conditional duration model

In this paper, we propose a class of ACD-type models that accommodates overdispersion, intermittent dynamics, multiple regimes, and sign and size asymmetries in financial durations. In particular, our functional coefficient autoregressive conditional duration (FC-ACD) model relies on a smooth-transition autoregressive specification. The motivation lies on the fact that the latter yields a universal approximation if one lets the number of regimes grows without bound. After establishing that the sufficient conditions for strict stationarity do not exclude explosive regimes, we address model identifiability as well as the existence, consistency, and asymptotic normality of the quasi-maximum likelihood (QML) estimator for the FC-ACD model with a fixed number of regimes. In addition, we also discuss how to consistently estimate using a sieve approach a semiparametric variant of the FC-ACD model that takes the number of regimes to infinity. An empirical illustration indicates that our functional coefficient model is flexible enough to model IBM price durations.

Time-dependent or state-dependent pricing? : evidence from a large currency devaluation episode

Duas classes de modelos buscam explicar o padrão de ajustamento de preço das firmas: modelos tempo-dependente e estado-dependente. O objetivo deste trabalho é levantar algumas evidencias empíricas de modo a distinguir os modelos, ou seja, identificar de que maneira as firmas realmente precificam. Para isso, escolheu-se a grande desvalorização cambial de 1999 como principal ferramenta e ambiente de análise. A hipótese fundamental é que o choque cambial impacta significativamente o custo de algumas indústrias, em alguns casos induzindo-as a alterarem seus preço após o choque. A partir de uma imensa base de micro dados formada por preços que compõem o CPI, algumas estimações importantes como a probabilidade e a magnitude média das trocas foram levantadas. A magnitude é dada por uma média simples, enquanto a probabilidade é estimada pelo método da máxima verossimilhança. Os resultados indicam um comportamento de precificação similar ao proposto por modelos estado-dependente.

Asymptotic efficiency in an instrumental variable model

Esta dissertação se propõe ao estudo de inferência usando estimação por método generalizado dos momentos (GMM) baseado no uso de instrumentos. A motivação para o estudo está no fato de que sob identificação fraca dos parâmetros, a inferência tradicional pode levar a resultados enganosos. Dessa forma, é feita uma revisão dos mais usuais testes para superar tal problema e uma apresentação dos arcabouços propostos por Moreira (2002) e Moreira & Moreira (2013), e Kleibergen (2005). Com isso, o trabalho concilia as estatísticas utilizadas por eles para realizar inferência e reescreve o teste score proposto em Kleibergen (2005) utilizando as estatísticas de Moreira & Moreira (2013), e é obtido usando a teoria assintótica em Newey & McFadden (1984) a estatística do teste score ótimo. Além disso, mostra-se a equivalência entre a abordagem por GMM e a que usa sistema de equações e verossimilhança para abordar o problema de identificação fraca.

Estimates of genetic parameters for visual scores and their correlation with production and reproductive traits in Brahman cattle

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Parâmetros Genéticos para a Produção de Leite, Gordura e Proteína em Bubalinos

The objective of this study was to estimate the genetic parameters affecting milk production (MP), fat (%F) and protein (%P) contents of buffalo milk. Restricted Maximum Likelihood (REML) using MTDFREML program under animal model analyzed a total of 1744 lactations records from 1268 cows. The means were: MP = 1259.47 +/- 523.09 kg, %F = 6.87 +/- 0.88% and %P = 3.91 +/- 0.61%. The estimates of repeatability and heritability coefficients were: MP = 0.38 and 0.24, %F = 0.28 and 0.21 and %P = 0.30 and 0.26, respectively. The estimated genetic and phenotypic correlations were MP x %F = -0.18 and -0.62, MP x %P = -0.23 and -0.59 and %F x %P = 0.50 and 0.77, respectively. According to these results it is possible to conclude that selection is a proper way to increase milk yield, fat and protein percentage. Although negative low values of genetic correlations among traits, it should be take into account that simultaneous selection based on these traits could not be so efficient.

Índices de seleção para um rebanho Caracu de duplo propósito

Neste trabalho, objetivou-se determinar índices de seleção para um rebanho da raça Caracu de duplo propósito, cujo objetivo de seleção (H) incluiu a venda de bezerros desmamados e a produção de leite. As características que compuseram H foram: produção total de leite, idade ao primeiro parto, período de serviço, peso à desmama e duração da vida produtiva. Foram propostos dois índices de seleção para H. Os critérios de seleção adotados no índice 1 foram: produção de leite na primeira lactação, primeiro período de serviço, peso à desmama (PD) e perímetro escrotal. No índice 2, foram consideradas as mesmas características, sendo, entretanto, o ganho médio diário do nascimento à desmama empregado como critério de seleção para PD. As análises estatísticas para a obtenção dos componentes de (co)variâncias e das estimativas dos parâmetros genéticos e fenotípicos foram realizadas pelo método da máxima verossimilhança restrita livre de derivada, por um modelo animal (uni e bi-característica). As informações zootécnicas e genealógicas, os preços e as quantidades dos insumos e dos produtos foram obtidos no escritório da empresa estudada. Os custos de produção e as receitas da atividade pecuária foram determinados para o período de 1994 a 2000. A equação de lucro foi tomada em base anual, usando-se valores médios para número de animais por categoria, características biológicas e preços..¹ O valor econômico para cada característica foi obtido pela derivada parcial da função de lucro em relação à característica em questão. Os dois índices econômicos de seleção trariam considerável resposta para o objetivo proposto. Entretanto, o índice 1, que incluiu PD, seria um pouco mais eficiente em termos de resposta à seleção total.

Genetic associations between weight at maturity and maturation rate with ages and weights at first and second calving in Canchim beef cattle

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Genetic associations between age at first calving and heifer body weight and scrotal circumference in Nelore cattle

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Genetic associations between accumulated productivity, and reproductive and growth traits in Nelore cattle

The total meat yield in a beef cattle production cycle is economically very important and depends on the number of calves born per year or birth season, being directly related to reproductive potential. Accumulated Productivity (ACP) is an index that expresses a cow's capacity to give birth regularly at a young age and to wean animals of greater body weight. Using data from cattle participating in the "Program for Genetic Improvement of the Nelore Breed" (PMGRN - Nelore Brasil), bi-trait analyses were performed using the Restricted Maximum Likelihood method based on an ACP animal model and the following traits: age at first calving (AFC), female body weight adjusted for 365 (BW365) and 450 (BW450) days of age, and male scrotal circumference adjusted for 365 (SC365), 450 (SC450), 550 (SC550) and 730 (SC730) days of age. Median estimated ACP heritability was 0.19 and the genetic correlations with AFC, BW365, BW450, SC365, SC450, SC550 and SC730 were 0.33, 0.70, 0.65, 0.08, 0.07, 0.12 and 0.16, respectively. ACP increased and AFC decreased over time, revealing that the selection criteria genetically improved these traits. Selection based on ACP appears to favor the heaviest females at 365 and 450 days of age who showed better reproductive performance as regards AFC. Scrotal circumference was not genetically associated with ACP. (C) 2007 Elsevier B.V. All rights reserved.

Heritabilities and genetic correlations for reproductive traits in an F2 reciprocal cross chicken population

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Genetic parameters and environmental effects on temperament score and reproductive traits of Nellore cattle

Coordenação de Aperfeiçoamento de Pessoal de Nível Superior (CAPES)

Estimation of genetic parameters for partial egg production periods by means of random regression models

Conselho Nacional de Desenvolvimento Científico e Tecnológico (CNPq)

Herdabilidade e correlações quanto a peso, perímetro escrotal e escores visuais à desmama, em bovinos Canchim

O objetivo deste trabalho foi estabelecer o modelo mais adequado para avaliação genética de bovinos Canchim e estimar os parâmetros genéticos de características produtivas à desmama. Foram utilizados dados de: 12.103 animais, quanto ao peso (PD); 5.278, quanto ao perímetro escrotal (PE); 8.343, quanto ao escore visual da conformação frigorífica (CF); 9.111, quanto ao escore de umbigo (UM); e 7.986, quanto ao escore de pelame (PEL). Os modelos estatísticos incluíram os efeitos fixos e os efeitos aleatórios genéticos aditivos direto, materno e de ambiente permanente materno, em diferentes combinações. As análises foram feitas pelo método da máxima verossimilhança restrita livre de derivadas. O modelo completo foi o mais adequado para PD, PE, CF e UM, enquanto o modelo com uso apenas dos efeitos genéticos aditivos direto e materno foi o mais adequado para PEL. As estimativas de herdabilidade direta foram 0,17, 0,13, 0,20, 0,18, e 0,52 para PD, PE, CF, UM e PEL, respectivamente, o que indica a possibilidade de se obter progresso genético por meio da seleção para essas características, principalmente para PEL. As correlações genéticas aditivas diretas entre as características variaram de -0,16 a 0,61. As correlações entre PD e PE e entre PD e CF indicam que a seleção para PD deve proporcionar ganho genético em PE e CF