Brazilianbanks

<!-- badges: start -->

<!-- badges: end -->

brazilianbanks is an R package to download publicly-available data on Brazilian banks and other regulated financial institutions. This is a rich panel dataset that includes information ranging from total assets to number of physical branches to the growth in credit to specific economic sectors according to a maturity bucket.

The package should be functional but has not been officially released yet. Its alpha version will be released once the appropriate tests are included and the documentation is written more extensively. The plan is to eventually submit the package to CRAN.

Source of the data

The data is obtained from the Central Bank of Brazil's "IF.Data" page. The value added compared to accessing data from the IF.Data website is that brazilianbanks enables you to download the data across multiple periods as a panel dataset, and in a tidy format that you can readily use in your data science or econometric pipeline.

Install

In your R console, type:

devtools::install_github("dkgaraujo/brazilianbanks")

Usage

Bank-level information

The brazilianbanks package already includes off-the-shelf two datasets of banks and other financial institutions' accounting, risk and other information. One such dataset includes all information from 2017 Q1 up to the most recent quarter: brazilian_banks_201703_onwards. A shorter version of this dataset covers the period of 2017 Q1 to 2021 Q4: brazilian_banks_201703_202112.

library(brazilianbanks)

brazilian_banks_201703_onwards

Using these pre-loaded datasets is usually a quicker alternative to get_bank_stats(), the main function to retrieve data in a more customised way.

The function all_available_quarters() returns a vector with all the quarters for which there is available data. I suggest you run this function first, in particular to check what the most recent quarter is.

If you are interested in bank-level variables (eg, accounting or credit risk figures), the main function is get_bank_stats(). It returns a tibble data frame with the panel data:

library(brazilianbanks)

quarters <- all_available_quarters()
bank_df <- get_bank_stats(yyyymm_start = 201903, yyyymm_end = max(quarters))

The initial and final quarters (arguments yyyymm_start and yyyymm_end respectively) are defined by the user according to the YYYYMM format, ie March 2014 is 201403. Note that the code chunk above uses the most recent available quarter as the final quarter for data download.

Banking statistics at the municipality level

If your focus is on municipality-level banking statistics, the main function is get_municipal_stats(). It returns a tibble data frame with bank balance sheet and financial statement variables at the municipal level, at a monthly frequency:

library(brazilianbanks)

municipal_banking_data <- get_municipal_stats(yyyymm_start = 201111, yyyymm_end = 201808)

Definitions

The result of get_bank_stats() represents only data at the consolidated levels, equivalent to "bank holding company" level in other jurisdictions. Brazilian regulation differentiates between "prudential" and "financial" conglomerate consolidation perspectives; both are considered to be the same bank.

Speed

Note that the first download of the data from the BCB servers may be slow (due to server-side speed limits I believe). To mitigate that, you are encouraged to always use cache_json = TRUE, as set by default. This will ensure your downloaded files are stored locally, saving you time in future uses. Also, processing the information from the original JSON files is not optimised for speed. To bypass these speed issues, released versions will put processed files available as part of the package, which should be good enough for users that do not mind not using the data directly from the BCB source.

Comments, bugs, issues, feature requests and suggestions for improvement

Please write them here. Pull requests are welcome after launch of the alpha version and its posterior versions (before that, unit tests will not be available to check PRs for consistency). Contributors are encouraged but not required to describe the issue first in this previous link.

Thank you very much for your interest!