Home  /  Users Group meetings  /  2017 Brazil

The 2nd Brazilian Stata Users Group meeting was Friday, 8 December 2017 at the Faculty of Economics, Administration and Accounting of the University of São Paulo, but you can view the program and presentation slides below.

Spatial regression models: Identification strategy using Stata
Abstract: Spatial regression models are usually for estimating parameters related to the interaction of agents across space, e.g. social interactions, agglomeration externalities, technological spillovers, strategic interactions between governments, etc. In this practical class, I will explore estimation of different cross-sectional spatial models using Stata. I will discuss variants of spatial econometrics and social interactions, regression specifications, and challenges to estimation or "identification" of different types of interaction:

  • SAR/endogenous interactions
  • SLX/contextual interactions
  • SE/correlated interactions

I will focus on methods for getting consistent "causal" estimates of the effect. I will also discuss methods of estimation and problems of interpretation. As an example, I will use some data on Brazilian housing transactions and look at the effects of crime on neighbouring housing sales price.

Additional information:

Tatiane Menezes
Universidade Federal de Pernambuco

Latent class models applied to quality-of-life scores: A case study using gsem to optimize a latent profile analysis
Abstract: Latent class analysis (LCA) is one of the highlights available in Stata 15. This new feature allows identification of "unknown groups" (or classes) within a given population. When dealing with continuous observed variables, a latent class model is named "latent profile analysis" (LPA). To check how it works, we used quality-of-life (QOL) scores in an LPA to fit a generalized structural equation model (GSEM). In this case study of 600 individuals, 4 domains of the questionnaire WHOQOL-BREF are the observed variables, and scores I converted to a 0-100 scale. The goal of the modeling strategy was identifying the "most appropriate" number of classes. To achieve this task, I specified different numbers of classes in a sequence of models. After that, I estimated the marginal predicted means (with 95% confidence intervals) of each domain within each latent class. I also estimated the posterior probability of individuals being in a given class. The Akaike information criterion (AIC) and the Bayesian information criterion (BIC) were used as a measure to assess the relative quality of the model. Plots of the parameters of the "best fit" model and interpretation for the results concerning the identification of (so far) "unknown" groups are presented.

Additional information:

Marcos Almeida
Universidade de Tiradentes

Decomposing wage differential by sectors and occupations: What explains a glass ceiling effect in Brazil by gender?
Abstract: In Brazil, we have a large wage differential between women and men, particularly when we consider specific sectors and occupations. In this presentation, I attempt to understand the gender wage gap across the wage distribution. Using data from the Household National Survey from 2015 (PNAD/2015), I will show that the gender log wage gap in Brazil increases throughout the wage distribution and accelerates in the upper tail. This movement is different across sectors and specific occupations. I interpret this as a heterogeneous glass ceiling effect between sectors and occupations. I use quantile regression decompositions to examine whether this pattern can be ascribed primarily to gender differences in the labor market characteristics, the rewards to those characteristics depending the sector, or the occupation. Following the framework of counterfactual distributions functions proposed by Juhn, Murphy, and Pierce (1993) and the estimator proposed by Melly (2005) in the Stata language, I use a generalized version of the Oaxaca—Blinder decomposition, allowing me to separate three components: changes of the structural parameters, changes in the distribution of the covariates, and changes in the residuals.

Additional information:

Danielle Carusi Machado
Universidade Federal Fluminense
Jesus Óregon
Universidade Federal Fluminense

Does crime affect economic growth? A case study of the city of Manaus
Abstract: In recent years, Brazil has become the 7th largest economy in the world, according to the World Bank—making great economic and social developments. However, the crime rate has increased, and Brazil was the 11th most violent country in the world measured by homicide rates in 2012 according to the United Nations. In this presentation, I will carry out a case study for Manaus, where a criminal organization called "Família do Norte" was created in 2007. I analyze its effects on the GDP per capita and apply a synthetic control method to estimate their counterfactual economic performance in the absence of organized crime. The synthetic control is a weighted average of other regions unaffected by criminal organization activity that mimics the economic structure of Manaus a few years before 2007. The comparison of actual and counterfactual development shows that the presence of "Família do Norte" lowers GDP per capita by 20%.

Additional information:

Pedro Davi
Universidade de São Paulo
Now what do I do with this function?
Abstract: Nonparametric analysis has been traditionally descriptive. We fit the regression function that relates the outcome of interest and the covariates, and then we graph. However, we can go beyond the descriptive and use this function to compute marginal effects, counterfactuals, and other statistics of interest. In other words, we may use margins after npregress to conduct semiparametric analysis. I will show you how.

Additional information:

Enrique Pinzón

Auto evaluation of health and maternity
Abstract: Health self-assessment is a predictive measure of mortality, morbidity, and use of health services. Swedish studies have found that self-assessment of women's health may undergo changes depending on the condition of motherhood. This study aimed to verify if this relationship is also observed in Brazil, based on data from the 2013 PNS.

METHODS: The analysis was performed using Stata software version 12.0. The 2013 PNS data are the result of complex sampling. For this study, 22,621 observations were made by residents selected from the survey, representing 50,143,520 women under 50 years of age. The analyses were carried out with the application of corresponding sample weights.

RESULTS: Women with a recent reproductive history tended to better evaluate their health compared with women with no reproductive history, and women with a higher reproductive history rated their health worse.

CONCLUSION: The results are in line with those of the studies carried out in Sweden on the relation between motherhood and health self-evaluation, suggesting reproducibility and generalization potential for other populations. Stata software version 12.0 was adequate to perform the analysis and processing involved, considering the necessary resources to work with nationally-based study data with complex sampling.

Additional information:

Matheus Souza Ferreira
Universidade de São Paulo

Presentation of the research group: Brazilian Women in Economics (BWE)
Abstract: Brazilian Women in Economics is a research group that aims to study the various dimensions of gender gap in Brazil. We are specifically interested in the relative lack of women in science, technology, engineering, and mathematics (STEM) academic disciplines in the country. We also intend to encourage more women to study economics, and we promote the inclusion of women economists in the labor market, increasing the connections between those who work in the private and public sector and those who work in academia.

Additional information:

Paula Pereda
Universidade de São Paulo
Bruna Borges
Universidade de São Paulo
Multilevel mixed-effects generalized linear models in Stata
Abstract: Multilevel mixed-effects generalized linear models can be fit in Stata considering the existence of random effects in different levels and allowing outcome variables to follow distributions such as Gaussian, Bernoulli, binomial, gamma, negative binomial, ordinal, and Poisson. I use meglm to fit GLMs to multilevel datasets where outcome variables follow Bernoulli, Poisson, and negative binomial distributions while accounting for three link functions (logit, probit and log). I provide maximum likelihood estimation methods and likelihood-ratio tests comparing the proposed models with non-multilevel generalized linear models. meglm also allows flexible methods of modeling with general specifications of variance components—random intercepts and random slopes can be modeled as independent, correlated, or independent with equal variances, and so forth. These models are becoming increasingly popular in areas such as microecomics, finance, transportation, real estate, leisure, ecology, education, and health.

Additional information:

Luiz Paulo Fávero
Universidade de São Paulo
Matheus Albergaria
Universidade de São Paulo

Consequences of extending paid maternity leave: The Empresa Cidadã Program experience
Abstract: This presentation investigates the effects of paid maternity leave extension from 120 to 180 days on firms that have joined the Empresa Cidadã Program enacted by the Brazilian federal law 11.770 in 2008. The extension of the leave is not mandatory for the employees of participant firms, that is, women opt whether to take the extra 60 days. This study intends to understand the impact of this program on the entitled workers' careers using administrative information on all formal workers in Brazil gathered by the Ministry of Labor. Specifically, I explore effects on wage and employment of women in participant firms. Extensions on paid maternity leave produce a twofold effect; while they can attract women to the workforce, women become more onerous to employees. Although over 120 countries provide paid maternity leave, there is not a thorough comprehension of the effects of this sort of benefit, and the existing studies have shown mixed results. This research intends to broaden the debate on maternity leave extension; up to this moment, there is only one published article on this subject in Brazil.

Additional information:

Larissa Bueno Stolar
Universidade de São Paulo

Estimate the efficiency of Brazilian federal highway concessions: DEA and SFA—X Factor
Abstract: This presentation aims to provide an appropriate methodology for quallity discussion of the efficiency of Brazilian federal highway concessions and provide a way to implement yardstick competition with the X-Factor as recommended by international practice. I use data from the Brazilian federal highway concession contracts of the 1st and 2nd Stages of the Federal Government Concessions Program. To evaluate efficiency, I used a Stata implementation of data envelopment analysis (DEA) and of stochastic frontier analysis (SFA). The results from efficiency evaluation are used to calculate the X-Factor, which is part of a CPI-X regulation equation that forces a better performance from the concessionaries through time. Because the methodology is not defined in the current concession contracts, despite its predicted contractual clause, this work can enlighten the current discussions in the National Agency of Land Transportation (ANTT).
Carlos Eduardo Véras Neves
Agência Nacional de Transportes Terrestres—ANTT

Air pollution consequences in São Paulo: Evidence for health
Abstract: Pollution emissions cause negative externalities on human health, especially to vulnerable groups such as children and the elderly. I propose a model to estimate the impact of air pollution on respiratory hospitalizations in the São Paulo Metropolitan Region, which represents approximately 10% out of the Brazilian population. I deal with the endogeneity of air pollution exposure using wind speed as an instrument. The results show that air pollution positively affects hospitalizations in the short term. Sensitivity analysis corroborates the results and shows we have a strong instrument so far. Because of the data limitations of this preliminary study, I intend to improve the analysis by using satellite measurements and increasing the units of observation. My contribution to the literature is that I fit a model for a developing country correcting for potential endogeneity and run falsification and placebo tests that corroborate my results.

Additional information:

Bruna Guidetti
Universidade de São Paulo
Paula Pereda
Universidade de São Paulo
Dynamic document generation in Stata
Abstract: In the best of all possible worlds, any published paper, any monthly report, or any set of lecture notes should be easily reproducible. Even more so, they all should be simple to update based on new data or bug fixes. Such so-called dynamic documents are a part of the larger concept of reproducible research. In this presentation, I will present the new tools in Stata 15 for creating dynamic documents, including markdown and dyndoc for creation of web content and putdocx for creation of Open Office (docx) documents. I will also briefly look at some useful community-contributed tools and discuss how they differ or complement the official tools.

Additional information:

Bill Rising
Wishes and grumbles


Registration is closed.

The meeting is free, though registration is required. Seating is limited, so we strongly encourage you to register as soon as possible. Participants must travel at their own expense.


Scientific committee

Logistics organizer

The logistics organizer for the 2017 Brazilian Stata Users Group meeting is Timberlake Analytics Brasil, the distributor of Stata in Brazil.

View the proceedings of previous Stata Users Group meetings.