The Brazilian Stata Users Group was held on 7 December 2018 at the School of Economics and Administration of the University of São Paulo. You can view the program below.
overdisp: A direct command to detect overdispersion in count-data models using Stata (and Mata)
Abstract: Identifying overdispersion in count-data models is important for the definition of estimations such as Poisson or negative binomial. The new command overdisp reports directly the existence of overdispersion in the dependent variable, conditional on explanatory variables, and can be used without the necessity of previously estimating count-data models.
Luíz Paulo Fávero
Social media and customer service: The case of a large retail bank in Brazil
Abstract: In the age of social media, banks might have to rely more on technology-mediated interactions with clients. This presentation offers an insight on how information technology is used in interactions by bringing evidence on the implementation of social media as a channel for bank customer services. Evidence from a large commercial Brazilian bank was collected during five semesters after the institution introduced Facebook and Twitter in 2013.
Analyzing the profitability of providing customer services using social media versus the customer services with more traditional methods is assessed (Facebook and Twitter versus call centers). We estimate a linear regression model with cross-section data and a linear model with panel data and client fixed effects, where the outcomes of interest are explained by a client's characteristics and client-bank interactions. Regression shows a statistically significant and positive relation between client-bank interactions and profits or product consumption.
It also provides an estimate of the average treatment effect of high-technology interactions using a doubly robust estimator. For some groups of clients, Facebook and Twitter services can be considered profitable as traditional customer services. Results are robust to different specifications and indicate benefits from the interactions, even though some contacts are related to complaints and product cancellation.
Guilherme Gomes Cardim Mattos
Federal University of ABC
Natural gas vehicles in Brazil: Consequences to fuel markets
Abstract: Natural gas vehicles (NGVs) are very popular in many countries around the world, including Brazil. The Brazilian State of Rio de Janeiro has the largest NGV fleet in the country. Using a panel database extending for 15 years, we evaluate the impact of the NGV fleet growth in Rio de Janeiro on the retailer's prices and margins of gasoline and sugarcane ethanol. By correcting for endogeneity, we can identify a negative impact of the former variables on the last ones.
The result is generally robust to different specifications of our model and instrument as well as to data adjustment. We also calculate that the NGV fleet growth has benefited the environment through lower pollutant emissions. Hence, the increase in the NGV fleet is beneficial to society not only through less pollution but also through lower gasoline and ethanol prices.
Roberto Amaral de Castro Prado Santos
School of Economics of São Paulo
Population dynamics of the primary malaria vector Nyssorhynchus darlingi in a high-transmission setting dominated by fish farming in western Amazonian Brazil
Abstract: The Vale do Jurua (Acre) in the western Amazon reported the highest annual parasite incidence (API) of malaria in Brazil since 2005, a persistent hotspot. Fish farming was encouraged, resulting in breeding-sites for vector Nyssorhynchus darlingi that, combined with the high human density and mobility at urban zones, added to the local malaria burden.
This study reports an entomological profile of immature Ny. darlingi at three sites in Mâncio Lima municipality, Acre, during the rainy and dry seasons (February to September 2017). From 63 identified fishponds, Anophelinae larvae was collected 5,512 first-instar, and 4,927 second, third, and fourth-instar, of which 8.5% (n=420) were Ny. darlingi species. Seasonality analysis performed found no significant difference for Ny. darlingi abundance between February (baseline) and the remaining months in urban landscape sites, showing that larval population size remain stable in the transition of rainy to dry seasons.
For rural sites, significantly higher incidence ratio for the last two months was found: August [5.80 (1.11-30.41)] and September [6.62 (1.29- 33.89)], suggesting an important role of fish ponds in maintenance of vector population during seasonal transition. These data demonstrate spatiotemporal microgeographic features for primary malaria vector and can indicate larval control efforts as an essential part of integrated disease control for the Vale do Juruá region.
The labor market effects of maternity leave extension
Abstract: This presentation shows the effects of a sizeable maternity leave extension on labor market outcomes of women in Brazil, using detailed information on workers and firms in the formal labor market. Taking advantage of the exact leave-taking dates and the staggered implementation of the extended leave policy across firms, our analysis compares outcomes within firms before and after the eligibility cutoff. While eligible women could have extended their leave period by 50% (from 120 to 180 days), they take up increases only modestly. Moreover, employment effects are restricted to the maternity leave spell itself, with no permanent effects on employment in the long run. Taken together, our findings indicate that this policy is targeted to a select group of workers, while it is not able to retain them in the workforce.
Valdemar Pinho Neto
Individualistic culture and entrepreneurial opportunities
Abstract: This presentation evaluates the effect of living in an individualistic society on entrepreneurial opportunities using cross-country data from the Global Entrepreneurship Index (GEI) in 2017.
Individualism is a social trait that emphasizes freedom and rewards the personal achievements. We choose to combine the fractional probity regression model with an instrumental variable approach through the conditional mixed-process (CMP) framework implemented by David Roodman's cmp command in Stata.
One of the reasons to use this method is that we can jointly estimate two or more equations with linkages among their error processes. And the individual equations need not be classical regressions with a continuous dependent variable. So instead of two linear OLS regressions in the usual 2SLS estimation, our IV-FPRM model performs a two-step procedure with two fractional probity estimations.
Then, we find that the number of opportunity startups (as opposed to necessity startups) is higher in individualistic countries. Part of this effect occurs because individualistic people perceive opportunities in a optimistic way because they pretend to realize personal aims and because startups in individualistic countries are more innovative. These findings are robust to differences in institutions, religious affiliation, fertility, unemployment, and education.
Daisy Assmann Lima
Florângela Cunha Coelho
Catholic University of Brasília
Cigarette consumers' behavior: Effects of smoking bans in Brazil
Abstract: Approximately 135,000 Brazilians die from smoking-related diseases every year. Because of the relevance of smoking as a public health problem, some states and municipalities implemented restrictive smoke-free environments from 2008 onward. The national regulation became effective only in 2014. This presentation will explore the regional differences in the adoption of this policy to evaluate its impacts on cigarette consumption in Brazil. We propose a difference-in-differences approach to estimate the effects throughout the years on the smoking behavior of two age groups: adults and youths.
We also estimate the impacts according to different levels of enforcement among states. We have built our panel using microdata from the National Health Research, collected by the Brazilian Bureau of Statistics (IBGE) in 2013. Our results indicate that smoking bans reduced daily cigarette consumption among young individuals by 7% on average by reducing smoking initiation. This impact increases to 10% when considering higher enforced smoking bans and vanishes when the law is not enforced.
It represents 80 thousand fewer smokers among youths in the treated capitals. No impacts were found on smoking prevalence and initiation among adults, but high enforced smoking bans are related to a 1.2% smoking cessation rate. Robustness tests show that the results are consistent. This presentation contributes to the literature providing evidence to developing countries on the effects of tobacco control policies.
Paula Carvalho Pereda
Rally round the gun? Drug battles and law and order candidates in Rio de Janeiro
Abstract: This presentation studies the effects caused by the exposure of violence on electoral preferences in Rio de Janeiro's state representatives' elections. I test if higher exposure leads to more votes for "law and order" candidates, those who promise tougher responses to public security issues, identified by their registered profession and by running with commission on their name.
With a database of anonymous reports, the occurrence of shootings was mapped across town and is used as an explanatory variable in our model. I find negative and significant effects for this type of candidate. There is also a negative effect for left-wing parties and for incumbent candidates but no effect for turnout or for blank and null vote share.
Because these results could be influenced by police violence in certain regions, I verify if they differ according to official data for deaths in police operations (DPOs). In fact, my main estimates for candidates whose profession is in a security area seem solely driven by places with above median DPOs, suggesting that the public's negative view of police plays a central role for such candidate's image as well.
Felipe Campos Ronchini Lima
Decomposing differences in distributions
Abstract: This presentation will focus on the decomposition methods developed by Firpo, Fortin, and Lemieux (2011). These methods are used to decompose the difference in a distributional statistic between two groups, or its change over time, into various explanatory factors. I will present the Stata command related to this method.
Conditional correlation and volatility between spot and futures markets for soybean and corn
Abstract: This presentation investigates the dynamics of volatility and conditional correlations between corn and soybean prices in the spot and futures markets. Although most rural producers are not directly involved in the derivatives markets, they are, however, increasingly affected by them. Faced with price and production risks, farmers must use all information available in their risk management process, both in their product's spot and futures markets and in related products' markets, either domestic or foreign.
Constant conditional correlation and dynamic conditional correlation specifications with a bivariate VAR-GARCH model are employed, with data for Brazilian and US markets in the period 2004-2017, during which several structural changes and extreme climate phenomena have occurred. We find the highest correlations between the corn and soybean spot markets and evidence of contagion in both products' spot and futures markets. We also find that the financial crisis has significantly affected the relationship between the corn and soybean markets.
Julyerme Matheus Tonin
State University of Maringá
Carlos Manuel Rodrigues Vieira
Rui Manoel De Sousa Fragoso
CEFAGE/University of Évora
João Gomes Martines Filho
Internal migration, rainfall, and local labor markets in Brazil
Abstract: Individuals living in poorer areas usually perform worse on several socioeconomic dimensions. There is a growing concern that climate change could be more harmful to vulnerable populations in rural areas of developing nations. An important coping strategy that could help to attenuate the welfare impacts of climate change is internal migration. It can be a powerful tool to fight poverty and, hence, play a key role in reshaping the economic geography of lower-income countries. Estimating the impact of migration on the native population in destination regions is difficult.
First, from a purely empirical point of view, the decision of whether to migrate and where to move to are not random and may correlate with economic outcomes at the destination. Such feature may introduce confounding factors that will bias the estimates. Moreover, the arrival of migrants does not affect all local individuals in the same way. While some may benefit from an increase in demand for their skills, others may suffer because of increased competition in the labor market, potentially leading to higher wage inequality.
In this presentation, we study the impact of internal migrants escaping from droughts in the Brazilian Semiarid on labor markets at the destination. To circumvent the estimation challenges, we propose a two-step instrumental variable procedure. First, we instrument migration outflow from the Semiarid with rainfall deviations from historical averages to predict the number of internal migrants leaving their hometowns. Then, we assign a share of the predicted outflow according to the preexisting network in each destination based on the migrant's municipality of origin.
We find that for every 50 migrants, 3 native workers in the formal sector lose their job and 1 moves to the informal sector. The effect on the average wage of native workers is small and insignificant. However, this result masks considerable heterogeneity. The effect is significantly negative for workers below the 20th percentile of the wage distribution and positive for those higher up on the wage ladder. Our results help to explain the increase in income inequality observed in destination municipalities from 1970–1990.
Econometrics for experimental economics using Stata: Treatment testing
Abstract: The aim of this presentation is to discuss econometric design issues of experimental economics in the context of treatment testing, such as the choice between parametric and non-parametric methods. One of the key issues in this choice is the scale of measurement of the outcome variable (nominal, ordinal, cardinal).
Several parametric tests rely on distributional assumptions that can hold only if the variables under analysis are measured on a cardinal scale. The distributional assumption commonly required is that of normality in the outcome variable.
Because of nonnormal data from economic experiments, nonparametric tests are used, although this strategy is likely to be costly if data have a cardinal nature. As a result, I will demonstrate various parametric and nonparametric tests of treatment effects using a dataset from the experiment conducted by Ellingsen & Johannesson (2004) as well as discuss methods available that provide a means of conducting a parametric test without making any assumptions about the distribution of the data.
Victor Eduardo da Motta
Efficient dynamic documents using Stata
Abstract: Stata 15 includes three new commands for producing dynamic documents: dyndoc, putdocx, and putpdf. These commands have generated much interest in the user community; this has led to a large amount of community-contributed software. In this talk, I'll give some tips about how to use the commands efficiently both with official Stata software and with some of these community-contributed tools.
Wishes and grumbles
Abstract: Stata developers present will carefully and cautiously consider wishes and grumbles from Stata users in the audience. Questions, and possibly answers, may concern reports of present bugs and limitations or requests for new features in future releases of the software.
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.
The logistics organizer for the 2018 Brazilian Stata Users Group meeting is Timberlake Analytics Brazil, the distributor of Stata in Brazil.
View the proceedings of previous Stata Users Group meetings.