BETS - Brazilian Economic Time Series: Basic Usage

2018-06-10

Introduction

The BETS package (an abbreviation for Brazilian Economic Time Series) for R [@R] allows easy access to the most important Brazilian economic time series and a range of tools for analyzing them. It provides a much-needed single point of access to the many Brazilian series and a simple, flexible and robust interface.

The series in the BETS database are produced by three important and respected institutions: the Central Bank of Brazil (BACEN), the Brazilian Institute of Geography and Statistics (IBGE) and the Brazilian Institute of Economics at the Fundação Getúlio Vargas (FGV/IBRE and FGV Data). BETS was originally conceived to bring together as many series from these centers as possible in a single place because of the difficulties researchers faced obtaining this information. This objective was achieved and the package now contains more than 18,000 Brazilian economic series.

Because of the extremely large size of the databases, BETS was expanded to provide mechanisms that would help analysts to search for, extract and export series. The development of the package then took a new direction and BETS was to become an integrated analysis and learning environment. A range of functionality missing from the R universe was included to make modeling and interpretation of the series available in BETS even easier. In addition, some functions already include the option of generating explanatory outputs that encourage and help users to expand their knowledge.

This article seeks to describe the structure and some of the more important functionality of BETS. Typical examples of how the package can be used are described step by step. In the next section we discuss the database in detail. Then in Section 3 we describe the structure of the package and the different elements it is currently composed of. Section 4 introduces the basic ways in which BETS can be used and shows how users can search for, read and store series. The more advanced functions are also discussed in this section.

Database

Traditionally, the analyst could obtain the IBRE series in the FGV Data system1, which contains all the FGV statistical output and a set of selected indicators from other centers. With BETS it is no longer necessary, since we built a database to store this data and provide access to it. The figure below shows a schematic representation of the BETS database in its current form.

Series in the BETS database

Series in the BETS database

An important discussion during the conceptual design phase for BETS concerned the location of the database. The option of distributing the data together with the package was discarded as CRAN does not accept very large data files (.rdas). This left two alternatives: to collect the data dynamically using the external APIs in the Central Bank of Brazil, IBGE and IBRE systems or to maintain a database with a database management system (DBMS). In its current form, the package implements a hybrid solution. The IBGE and BACEN series are acquired on the fly using the respective APIs. This procedure is not necessary for the FGV/IBRE series as the data are provided directly by the institution, which is supporting the project. These series are kept in a database built using MySQL.

This division is invisible to the user, who does not need to know which database he should look in and will always have access to the most up-to-date data.2 This architecture facilitates implementation and maintenance of the structures and increases the efficiency of the package. The data are updated by a team of maintenance staff with the aid of servers, avoiding the need for the user to obtain new data manually. However, this requires that the user be connected to the Internet.

Implementation of fast, easy access would not have been possible without a table of the metadata for the available series. This table, which is maintained in the MySQL database, contains information such as description, periodicity, the unit in which the data are represented and the start and end dates for all the series that can be accessed using the package. Each series has a unique code, which is treated as an index by the DBMS. The auxiliary package RMySQL [@RMySQL] acts at the interface, allowing R to connect to the MySQL database.

Structure of the Package

In the previous section we looked at the different stages involved in data retrieval and organization. We now show how the package is structured and then explain how the end user accesses the databases.

BETS functionality groups

BETS functionality groups

BETS functionality can be divided into four groups:

The user interface with the database is quite intuitive. There is one function for searching, one for extracting the data and a class of functions for external storage (see table below), so that the data can be processed with popular software such as SaS, Stata or SPSS. In the next section we will look at how these functions should be used.

Name Description
BETSsearch Searches series for different characteristics.
BETSget Completely extracts the series from the database and loads it in the R environment
saveSpss Exports the time series in a file with an .spss extension
saveSas Exports the time series in a file with an .sas extension
saveStata Exports the time series in a file with a .dta extension

With the information provided so far, the promise that BETS holds becomes much more apparent. By going beyond the mere supply of data and providing a wide range of tools for studying time series, this pioneering package allows an inexperienced programmer anywhere in the world to analyze the Brazilian economic scenario. All that it takes to unleash the power of BETS is a simple install.packages("BETS"), the only command needed to install the package.

Using BETS

In this section we discuss some of the basic ways in which the package can be used.

Interface with the Database

BETSsearch

Because the database is large, it was necessary to develop a way of searching for series using the metadata, i.e., a search tool that used some of the information about the series as keywords.

The BETSsearch function performs searches in each field of the metadata table described in Section Database. It naturally allows combinations of these characteristics, making searches more flexible. Note that access to the BETS database is by means of the sqldf package, which makes the processing of searches sufficiently fast and ensures that the package performs well in any environment.

The BETSsearch prototype has the form:

BETSsearch(description, src, periodicity, unit, code, view = TRUE, lang = "en")

where the arguments are, respectively

To refine the search, there are syntax rules for the parameter description:

  1. To look for alternative words, separate them by blank spaces. Example: description = 'core ipca' means that the description of the series should contain “core” and “ipca”.

  2. To search for complete expressions, put them inside ‘’. Example: description = 'index and 'core ipca'' means that the description of the series should contain “core ipca” and “index”.

  3. To exclude words from the search, insert a ~ before each word. Example: description = 'ipca ~ core' means that the description of the series should contain “ipca” and should not contain “core”.

  4. To exclude all the expressions from a search, as in the previous item, place them inside ‘’ and insert a ~ before each of them. Example: description = '~ index 'core ipca'' means that the description of the series should contain “index” and should not contain “core ipca”.

  5. It is possible to search for or exclude certain words as long as these rules are obeyed.

  6. A blank space is not required after the exclusion sign (~), but is required after each expression or word.

Some examples of how this function is used are given below. We have not shown the results in some cases as the output can be a very long table. However, we guarantee that all the calls work and invite the reader to test them.

# Some examples
BETSsearch(description = "sales ~ retail",view = F)
BETSsearch(description = "'sales volume index' ~ vehicles",view = F)
BETSsearch(description = "'distrito federal'", periodicity = 'A', src = 'IBGE',view = F)
# Search for accumulated GDP series
BETSsearch(description = "gdp accumulated", unit = "US$", view = F)

For further information on BETSsearch, including valid values for each field, consult the reference manual by typing ?BETSsearch in the R console.

BETSget

BETSget only works with the reference code for the series, which is obtained using BETSsearch. The command for this function takes the form:

BETSget(code, data.frame = FALSE)

The parameter code is obviously mandatory. The optional argument data.frame represents the type of object that will be returned. Its default value is FALSE, indicating that the object returned by the function will be a ts ( time series ). If data.frame = TRUE, the series will be stored in an object of type data.frame.

We will extract two series that we looked up previously.

# Get the 12-month cumulative GDP series in dollars
gdp_accum <- BETSget(4192)
window(gdp_accum, start = c(2014,1))
#Get the series for the GDP of the Federal District at market prices
gdp_df <- BETSget(23992, data.frame = T)
head(gdp_df)

save

To allow greater flexibility in the way BETS series are stored, files containing the series can be created in proprietary formats, i.e., formats associated with proprietary software. Basically, save extracts the time series from the database in the package in the form of a data.frame and creates a file in the specified format. There is a table in the file in which the first column contains the dates and the second, the data.

There are three variations of this function, whose prototypes are shown below:

saveSas(code, data = NULL, file.name = "series")
saveSpss(code, data = NULL, file.name = "series")
saveStata(code, data = NULL, file.name = "series")

Again, the parameter code receives the code for the series. The user can provide his own series with the argument data, which can be a data.frame or a ts. There is no need to add the extension to the file name in the parameter file.name.

Some examples of how this function is used are:

# Save the series for the net public debt in the default Excel format
saveStata(code = 2078, file.name = "series_stata.dta")

# Save the series for the net public debt in the default Excel format
saveStata(code = 2078, file.name = "series_stata.dta")

# Save any series in SPSS format
my.series <- BETSget(4447)
saveSpss(data = my.series, file.name = "series_spss")

Some Additional Functions

ere we are going to discuss some of the most special BETS functions.

chart

chart was originally designed to be a private function for use with dashboard. However, we felt it would be very helpful for users to have a way to obtain the dashboard graphs separately so that they could include them in their own work.

The chart prototype is:

chart(ts, file = NULL, open = TRUE, lang = "en", params = NULL)

The parameter ts takes one of several predefined graph options or a user defined series. There is also the option of saving the output in the working directory by defining the name of the file with the parameter file. If the file should be opened after it has been created, open should be set to TRUE. The parameter params is reserved for graphs of the user’s own series, i.e., series that are not predefined. It is a list that can contain the field codace, which receives a Boolean and indicates whether shaded areas corresponding to recessions identified by CODACE (FGV/IBRE) should be drawn, and the field start, which specifies what the start date for the series should be. As the graph concerns the economic scenario, the end date cannot be changed and is always the last item of data available.

Let us look at two examples of how chart is used. The outputs are displayed in the figures below.

# Uncertainty Index chart
chart(ts = 'iie_br', file = "iie_br", open = TRUE)

# Leading and Coincident Labor Indicators charts
chart(ts = "lab_mrkt", file = "lab_mrkt.png", open = TRUE)
Graph of the IIE-Br (Uncertainty Index for the Brazilian Economy) produced with chart

Graph of the IIE-Br (Uncertainty Index for the Brazilian Economy) produced with chart

Graph of the leading and coincident indicators for the Brazilian labor market produced with chart. The shaded areas represent recessions identified by CODACE.

Graph of the leading and coincident indicators for the Brazilian labor market produced with chart. The shaded areas represent recessions identified by CODACE.

For a complete list of the available graphs, consult the chart reference manual.

dashboard

In a previous section, we said that BETS includes a powerful tool for scenario analysis, the dashboards. Currently, we provide two options of dashboards: a business cycle and a macroeconomic situation dashboard. We also plan to extend the dashboards to cover other items in addition to those that are implemented.

dashboard(type = "business_cycle", charts = "all", saveas = NA, parameters = NULL)

To create a dashboard, we call dashboard, setting the type parameter either to business_cycle or “macro_situation”. It generates a .pdf and, if type is set to “busines_cycle”, one of its five pages is going to look similar to that in the next figure. In the example, the user chooses to save the file with the name survey.pdf. The graphs shown can also be chosen by the user with the charts parameter, which is “all” by default. The reference manual contains a complete list of the available graphs.

dashboard(type = "business_cycle", saveas = "survey.pdf")
A page of the business cycle dashboard.

A page of the business cycle dashboard.

The macroeconomic situation dashboard allows the user to insert a custom explanatory text, as well as a personal logo, email and website. For instance, the call below generates a four page dashboard whose first page is shown below.

parameters = list(author = "FGV/IBRE", 
                  url = "http://portalibre.fgv.br/",
                  text = "text.txt",
                  logo = "logo_ibre.png")

dashboard(type = "macro_situation", parameters = parameters)

The main advantage of such dashboards is that they are a convenient instrument to quickly understand the subject they are about. Their values are always the most recent ones, they are organized in well defined sections and its desing is highly informative.