--- title: "wbstats" author: "Jesse Piburn and Mauricio Vargas Sepulveda" output: litedown::html_format: options: toc: true number_sections: true vignette: > %\VignetteIndexEntry{02 - Setup} %\VignetteEngine{litedown::vignette} %\VignetteEncoding{UTF-8} editor: markdown: wrap: sentence --- # Programmatic Access to Data and Statistics from the World Bank API You can install the latest release version from CRAN with ```r install.packages("wbstats") ``` or The latest development version from github with ```r remotes::install_github("pachadotdev/wbstats") ``` # Introduction The World Bank^[] is a tremendous source of global socio-economic data; spanning several decades and dozens of topics, it has the potential to shed light on numerous global issues. To help provide access to this rich source of information, The World Bank themselves, provide a well structured RESTful API. While this API is very useful for integration into web services and other high-level applications, it becomes quickly overwhelming for researchers who have neither the time nor the expertise to develop software to interface with the API. This leaves the researcher to rely on manual bulk downloads of spreadsheets of the data they are interested in. This too is can quickly become overwhelming, as the work is manual, time consuming, and not easily reproducible. The goal of the `wbstats` package is to provide a bridge between these alternatives and allow researchers to focus on their research questions and not the question of accessing the data. The `wbstats` package allows researchers to quickly search and download the data of their particular interest in a programmatic and reproducible fashion; this facilitates a seamless integration into their workflow and allows analysis to be quickly rerun on different areas of interest and with realtime access to the latest available data. ### Highlighted features of the `wbstats` package: - Uses version 2 of the World Bank API that provides access to more indicators and metadata than the previous API version - Access to all annual, quarterly, and monthly data available in the API - Support for searching and downloading data in multiple languages - Returns data in either wide (default) or long format - Support for Most Recent Value queries - Support for `grep` style searching for data descriptions and names - Ability to download data not only by country, but by aggregates as well, such as High Income or South Asia # Getting Started Unless you know the country and indicator codes that you want to download the first step would be searching for the data you are interested in. `wb_search()` provides `grep` style searching of all available indicators from the World Bank API and returns the indicator information that matches your query. To access what countries or regions are available you can use the `countries` data frame from either `wb_cachelist` or the saved return from `wb_cache()`. This data frame contains relevant information regarding each country or region. More information on how to use this for downloading data is covered later. ## Finding available data with `wb_cachelist` For performance and ease of use, a cached version of useful information is provided with the `wbstats` package. This data is called `wb_cachelist` and provides a snapshot of available countries, indicators, and other relevant information. `wb_cachelist` is by default the the source from which `wb_search()` and `wb_data()` uses to find matching information. The structure of `wb_cachelist` is as follows ```r library(wbstats) str(wb_cachelist, max.level = 1) # List of 8 # $ countries :Classes ‘data.table’ and 'data.frame': 295 obs. of 18 variables: # $ indicators :Classes ‘data.table’ and 'data.frame': 28517 obs. of 8 variables: # $ sources :Classes ‘data.table’ and 'data.frame': 71 obs. of 9 variables: # $ topics :Classes ‘data.table’ and 'data.frame': 21 obs. of 3 variables: # $ regions :Classes ‘data.table’ and 'data.frame': 43 obs. of 4 variables: # $ income_levels:Classes ‘data.table’ and 'data.frame': 7 obs. of 3 variables: # $ lending_types:Classes ‘data.table’ and 'data.frame': 4 obs. of 3 variables: # $ languages :Classes ‘data.table’ and 'data.frame': 23 obs. of 3 variables: ``` ## Accessing updated available data with `wb_cache()` For the most recent information on available data from the World Bank API `wb_cache()` downloads an updated version of the information stored in `wb_cachelist`. `wb_cachelist` is simply a saved return of `wb_cache(lang = "en")`. To use this updated information in `wb_search()` or `wb_data()`, set the `cache` parameter to the saved `list` returned from `wb_cache()`. It is always a good idea to use this updated information to insure that you have access to the latest available information, such as newly added indicators or data sources. There are also cases in which indicators that were previously available from the API have been removed or deprecated. ```r library(wbstats) # default language is english new_cache <- wb_cache() ``` ## Search available data with `wb_search()` `wb_search()` searches through the `indicators` data frame to find indicators that match a search pattern. An example of the structure of this data frame is below ```r str(wb_search("GDP per person employed")) # Classes ‘data.table’ and 'data.frame': 2 obs. of 3 variables: # $ indicator_id : chr "SL.GDP.PCAP.EM.KD" "SL.GDP.PCAP.EM.KD.ZG" # $ indicator : chr "GDP per person employed (constant 2021 PPP $)" "GDP per person employed (annual % growth)" # $ indicator_desc: chr "GDP per person employed is gross domestic product (GDP) divided by total employment in the economy. Purchasing "| __truncated__ "GDP per person employed is gross domestic product (GDP) divided by total employment in the economy." ``` By default the search is done over the `indicator_id`, `indicator`, and `indicator_desc` fields and returns the those 3 columns of the matching rows. The `indicator_id` values are inputs into `wb_data()`, the function for downloading the data. To return all columns for the `indicators` data table, you can set `extra = TRUE` as below ```r str(wb_search("GDP per person employed", extra = TRUE)) # Classes ‘data.table’ and 'data.frame': 2 obs. of 8 variables: # $ indicator_id : chr "SL.GDP.PCAP.EM.KD" "SL.GDP.PCAP.EM.KD.ZG" # $ indicator : chr "GDP per person employed (constant 2021 PPP $)" "GDP per person employed (annual % growth)" # $ unit : logi NA NA # $ indicator_desc: chr "GDP per person employed is gross domestic product (GDP) divided by total employment in the economy. Purchasing "| __truncated__ "GDP per person employed is gross domestic product (GDP) divided by total employment in the economy." # $ source_org : chr "Staff estimates, World Bank (WB), note: Estimates are based on employment, population, GDP, and PPP data obtain"| __truncated__ "International Labour Organization, Key Indicators of the Labour Market database." # $ topics :List of 2 # ..$ :'data.frame': 1 obs. of 2 variables: # .. ..$ id : chr "10" # .. ..$ value: chr "Social Protection & Labor" # ..$ :'data.frame': 0 obs. of 0 variables # $ source_id : int 2 11 # $ source : chr "World Development Indicators" "Africa Development Indicators" ``` Sometimes the tables are quite long because of a description. Use `data.table` options to prevent full printing. ```r options(datatable.prettyprint.char = 25L) unemploy_inds<- wb_search("unemployment") head(unemploy_inds) # indicator_id indicator indicator_desc # # 1: fin37 Received government trans... The percentage of respond... # 2: fin37.1 Received government trans... The percentage of respond... # 3: fin37.2 Received government trans... The percentage of respond... # 4: fin37.38 Received government trans... The percentage of respond... # 5: fin37.38.1 Received government trans... The percentage of respond... # 6: fin37.38.2 Received government trans... The percentage of respond... ``` Other fields can be searched by simply changing the `fields` parameter. For example ```r blmbrg_vars <- wb_search("Bloomberg", fields = "source_org") head(blmbrg_vars) # indicator_id indicator indicator_desc # # 1: GFDD.OM.02 Stock market return (%, y... Stock market return is th... # 2: GFDD.SM.01 Stock price volatility Stock price volatility is... ``` Regular expressions are also supported ```r # 'poverty' OR 'unemployment' OR 'employment' povemply_inds <- wb_search(pattern = "poverty|unemployment|employment") head(povemply_inds) # indicator_id indicator # # 1: 1.0.HCount.1.90usd Poverty Headcount ($1.90 ... # 2: 1.0.HCount.2.5usd Poverty Headcount ($2.50 ... # 3: 1.0.HCount.Mid10to50 Middle Class ($10-50 a da... # 4: 1.0.HCount.Ofcl Official Moderate Poverty... # 5: 1.0.HCount.Poor4uds Poverty Headcount ($4 a d... # 6: 1.0.HCount.Vul4to10 Vulnerable ($4-10 a day) ... # indicator_desc # # 1: The poverty headcount ind... # 2: The poverty headcount ind... # 3: The poverty headcount ind... # 4: The poverty headcount ind... # 5: The poverty headcount ind... # 6: The poverty headcount ind... ``` As well as any `grep` function argument ```r # contains "gdp" and NOT "trade" gdp_no_trade_inds <- wb_search("^(?=.*gdp)(?!.*trade).*", perl = TRUE) head(gdp_no_trade_inds) # indicator_id indicator # # 1: 6.0.GDP_current GDP (current $) # 2: 6.0.GDP_growth GDP growth (annual %) # 3: 6.0.GDP_usd GDP (constant 2005 $) # 4: 6.0.GDPpc_constant GDP per capita, PPP (cons... # 5: BI.WAG.TOTL.GD.ZS Wage bill as a percentage... # 6: BM.KLT.DINV.WD.GD.ZS Foreign direct investment... # indicator_desc # # 1: GDP is the sum of gross v... # 2: Annual percentage growth ... # 3: GDP is the sum of gross v... # 4: GDP per capita based on p... # 5: # 6: Foreign direct investment... ``` The default cached data in `wb_cachelist` is in English. To search indicators in a different language, you can download an updated copy of `wb_cachelist` using `wb_cache()`, with the `lang` parameter set to the language of interest and then set this as the `cache` parameter in `wb_search()`. Other languages are supported in so far as they are supported by the original data sources. Some sources provide full support for other languages, while some have very limited support. If the data source does not have a translation for a certain field or indicator then the result is `NA`, this may result in a varying number matches depending upon the language you select. To see a list of availabe languages call `wb_languages()` ```r wb_langs <- wb_languages() ``` ## Downloading data with `wb_data()` Once you have found the set of indicators that you would like to explore further, the next step is downloading the data with `wb_data()`. The following examples are meant to highlight the different ways in which `wb_data()` can be used and demonstrate the major optional parameters. The default value for the `country` parameter is a special value of `"countries_only"`, which as you might expect, returns data on the selected `indicator` for only countries. This is in contrast to `country = "all"` or `country = "regions_only"` which would return data for countries and regional aggregates together, or only regional aggregates, respectively ```r # Population, total pop_data <- wb_data("SP.POP.TOTL", start_date = 2000, end_date = 2002) head(pop_data) # iso2c iso3c country date SP.POP.TOTL unit obs_status footnote # # 1: AW ABW Aruba 2000 90588 # 2: AW ABW Aruba 2001 91439 # 3: AW ABW Aruba 2002 92074 # 4: AF AFG Afghanistan 2000 20130327 # 5: AF AFG Afghanistan 2001 20284307 # 6: AF AFG Afghanistan 2002 21378117 # last_updated # # 1: 2026-07-13 # 2: 2026-07-13 # 3: 2026-07-13 # 4: 2026-07-13 # 5: 2026-07-13 # 6: 2026-07-13 ``` If you are interested in only some subset of countries or regions you can pass along the specific codes to the `country` parameter. The country and region codes and names that can be passed to the `country` parameter as well, most prominently the coded values from the `iso2c` and `iso3c` from the `countries` data frame in `wb_cachelist` or the return of `wb_cache()`. Any values from the above columns can mixed together and passed to the same call. ```r # you can mix different ids and they are case insensitive # you can even use SpOnGeBoB CaSe if that's the kind of thing you're into # iso3c, iso2c, country, region_iso3c, admin_region_iso3c, admin_region, income_level example_geos <- c("ABW","AF", "albania", "SSF", "eca", "South Asia", "HiGh InCoMe") pop_data <- wb_data("SP.POP.TOTL", country = example_geos, start_date = 2012, end_date = 2012) pop_data # iso2c iso3c country date SP.POP.TOTL unit # # 1: XD High income 2012 1342428399 # 2: AW ABW Aruba 2012 104110 # 3: AF AFG Afghanistan 2012 30560034 # 4: AL ALB Albania 2012 2860708 # 5: 7E ECA Europe & Central Asia (ex... 2012 233875199 # 6: 8S SAS South Asia 2012 1483553073 # 7: ZG SSF Sub-Saharan Africa 2012 944523292 # obs_status footnote last_updated # # 1: 2026-07-13 # 2: 2026-07-13 # 3: 2026-07-13 # 4: WB estimate by interpolat... 2026-07-13 # 5: 2026-07-13 # 6: 2026-07-13 # 7: 2026-07-13 ``` As of `wbstats 1.0` queries are now returned in wide format. This was a request made by multiple users and is in line with the principles of [tidy data](https://www.jstatsoft.org/article/view/v059i10). If you would like to return the data in a long format, you can set `return_wide = FALSE`. Now that each indicator is it's own column, we can allow custom names for the indicators ```r my_indicators = c("pop" = "SP.POP.TOTL", "gdp" = "NY.GDP.MKTP.CD") pop_gdp <- wb_data(my_indicators, start_date = 2010, end_date = 2012) head(pop_gdp) # iso2c iso3c country date gdp pop # # 1: AW ABW Aruba 2010 2453597207 101838 # 2: AW ABW Aruba 2011 2637859218 102591 # 3: AW ABW Aruba 2012 2615208380 104110 # 4: AF AFG Afghanistan 2010 15856668556 28284089 # 5: AF AFG Afghanistan 2011 17805098206 29347708 # 6: AF AFG Afghanistan 2012 19907329778 30560034 ``` You'll notice that when you query only one indicator, as in the first two examples above, it returns the extra fields `unit`, `obs_status`, `footnote`, and `last_updated`, but when we queried multiple indicators at once, as in our last example, they are dropped. This is because those extra fields are tied to a specific observation of a single indicator and when we have multiple indicator values in a single row, they are no longer consistent with the tidy data format. If you would like that information for multiple indicators, you can use `return_wide = FALSE` ```r my_indicators = c("pop" = "SP.POP.TOTL", "gdp" = "NY.GDP.MKTP.CD") pop_gdp_long <- wb_data(my_indicators, start_date = 2010, end_date = 2012, return_wide = FALSE) head(pop_gdp_long) # indicator_id indicator iso2c iso3c country date value # # 1: SP.POP.TOTL Population, total AF AFG Afghanistan 2012 30560034 # 2: SP.POP.TOTL Population, total AF AFG Afghanistan 2011 29347708 # 3: SP.POP.TOTL Population, total AF AFG Afghanistan 2010 28284089 # 4: SP.POP.TOTL Population, total AL ALB Albania 2012 2860708 # 5: SP.POP.TOTL Population, total AL ALB Albania 2011 2905195 # 6: SP.POP.TOTL Population, total AL ALB Albania 2010 2913021 # unit obs_status last_updated # # 1: 2026-07-13 # 2: 2026-07-13 # 3: 2026-07-13 # 4: 2026-07-13 # 5: 2026-07-13 # 6: 2026-07-13 ``` ### Using `mrv` and `mrnev` If you do not know the latest date an indicator you are interested in is available for you country you can use the `mrv` instead of `start_date` and `end_date`. `mrv` stands for most recent value and takes a `integer` corresponding to the number of most recent values you wish to return ```r # most recent gdp per captia estimates gdp_capita <- wb_data("NY.GDP.PCAP.CD", mrv = 1) head(gdp_capita) # iso2c iso3c country date NY.GDP.PCAP.CD unit obs_status # # 1: AW ABW Aruba 2025 NA # 2: AF AFG Afghanistan 2025 NA # 3: AO AGO Angola 2025 3129.477 # 4: AL ALB Albania 2025 12998.148 # 5: AD AND Andorra 2025 54291.503 # 6: AE ARE United Arab Emirates 2025 NA # footnote last_updated # # 1: 2026-07-13 # 2: 2026-07-13 # 3: 2026-07-13 # 4: 2026-07-13 # 5: 2026-07-13 # 6: 2026-07-13 ``` Often it is the case that the latest available data is different from country to country. There may be 2020 estimates for one location, while another only has estimates up to 2019. This is especially true for survey data. When you would like to return the latest avialble data for each country regardless of its temporal misalignment, you can use the `mrnev` instead of `mrnev`. `mrnev` stands for most recent non empty value. ```r gdp_capita <- wb_data("NY.GDP.PCAP.CD", mrnev = 1) head(gdp_capita) # iso2c iso3c country date NY.GDP.PCAP.CD obs_status # # 1: AW ABW Aruba 2024 38590.5650 # 2: AF AFG Afghanistan 2024 416.8711 # 3: AO AGO Angola 2025 3129.4766 # 4: AL ALB Albania 2025 12998.1479 # 5: AD AND Andorra 2025 54291.5026 # 6: AE ARE United Arab Emirates 2024 50273.5126 # last_updated # # 1: 2026-07-13 # 2: 2026-07-13 # 3: 2026-07-13 # 4: 2026-07-13 # 5: 2026-07-13 # 6: 2026-07-13 ``` ### Dates Because the majority of data available from the World Bank is at the annual resolution, by default dates in `wbstats` are returned as `numeric`s. This default makes common tasks like filtering easier. If you would like the date field to be of class `Date` you can set `date_as_class_date = TRUE` # Some Sharp Corners There are a few behaviors of the World Bank API that being aware of could help explain some potentially unexpected results. These results are known but no special actions are taken to mitigate them as they are the result of the API itself and artifically limiting the inputs or results could potentially causes problems or create unnecessary rescrictions in the future. ## Searching in other languages Not all data sources support all languages. If an indicator does not have a translation for a particular language, the non-supported fields will return as `NA`. This could potentially result in a differing number of matching indicators from `wb_search()` ```r # english cache_en <- wb_cache() sum(is.na(cache_en$indicators$indicator)) #> [1] 0 # spanish cache_es <- wb_cache(lang = "es") sum(is.na(cache_es$indicators$indicator)) #> [1] 14791 ``` # Legal The World Bank Group, or any of its member instutions, do not support or endorse this software and are not libable for any findings or conclusions that come from the use of this software.