As part of the World Bank's Open Data initiative, the World Bank provides free access to their extensive database through their Data Catalog API. This Julia package makes use of the API to interactively access and download World Bank data in Julia.

Similar to the WDI R package, data can be downloaded using function wdi. This function accepts indicators and countries as String or Array{String, 1}. Thereby, countries must be given as World Bank's iso2c code. Two additional Int inputs allow specifying beginning and end of the data period, while input true for keyword argument extra attaches further country information like the capital, longitude, latitude, income range, etc. to each observation. Output will be of type TimeData by default, but can be set to DataFrame through keyword argument format.

Pkg.clone("https://github.com/JuliaFinMetriX/WorldBankDataTd.jl.git")

using WorldBankDataTd

## single indicator, single country
gnp = WorldBankDataTd.wdi("NY.GNP.PCAP.CD", "BR")

## single indicator, multiple countries
gnp = WorldBankDataTd.wdi("NY.GNP.PCAP.CD", ["BR", "US", "DE"])

## multiple indicators, single country
data = WorldBankDataTd.wdi(["NY.GNP.PCAP.CD", "SP.DYN.LE00.IN"],
                         "BR")

## multiple indicators, multiple countries
data = WorldBankDataTd.wdi(["NY.GNP.PCAP.CD", "SP.DYN.LE00.IN"],
                         ["BR", "US", "DE"])

## same, but as DataFrame
data = WorldBankDataTd.wdi(["NY.GNP.PCAP.CD", "SP.DYN.LE00.IN"],
                         ["BR", "US", "DE"],
                         format = DataFrame)

## multiple indicators, multiple countries, additional information
data = WorldBankDataTd.wdi(["NY.GNP.PCAP.CD", "SP.DYN.LE00.IN"],
                         ["BR", "US", "DE"], extra = true)

data[1:5, :]

In addition to the indicator itself, the World Bank also provides a lot of metadata for countries and indicators which can be downloaded with function getWBMeta.

countryData = getWBMeta("countries")
countryData[1:5, :]

Country metadata provides information for each country or aggregated entity with respect to the following characteristics:

names(countryData)

Overall, the country metadata is a DataFrame of the following size:

size(countryData)

Quite similarly, you can access the complete indicator metadata through input indicators.

indicatorData = getWBMeta("indicators")
indicatorData[1:5, :]

The metadata contains information about each indicator regarding the following characteristics:

names(indicatorData)

Just take a look at the size of this table in order to get an impression about the huge amount of indicators that are provided by the World Bank.

size(indicatorData)

You can use these two DataFrames of country and indicator metadata to easily search for information through function search_wdi. Before looking at that, however, let's first spend a few words on downloading and caching of country and indicator metadata.

Function getWBMeta is required to force a fresh download of metadata. In general, however, you do not need to download metadata each time that you want to conduct a search. Hence, you should rather use function loadWBMeta instead, which only downloads metadata in case that it can neither be accessed from disk nor from the workspace of the current Julia session. getWBMeta should be called only if you think that your local metadata was corrupted in some way and needs to be refreshed.

Calling loadWBMeta will first try to access metadata stored in global variables WorldBankDataTd.country_cache and WorldBankDataTd.indicator_cache of the current session. If these variables have not been assigned yet, it looks for a local version of the data in directory WorldBankDataTd/data. If there is no local version yet, it will automatically call getWBMeta, which will download the data and store it both on local disk and as caching variables in the current session.

Hence, as we already did load the data, we can access the cached versions in our current workspace.

WorldBankDataTd.country_cache[1:5, :]

WorldBankDataTd.indicator_cache[1:5, :]

The most convenient way to explore indicators probably still is the World Bank webpage, where you can easily use the search functionality to find what your are looking for. Once you found the indicator of interest, you can read off the indicator shortcut name (e.g. SP.DYN.LE00.IN) from the URL of the indicator webpage.

Alternatively, however, this package also contains functionality to interactively search the database from Julia itself. Thereby, function search_wdi makes use of cached country and indicator metadata to speed up search operations. As first argument, you need to choose the metadata to be searched. This can be either "countries" or "indicators". The second argument needs to specify the column to be searched, while the actual search term needs to be given as regex as third argument.

For example, searching the :name column of countries for a case insensitive occurrence of "united":

res = search_wdi("countries", :name, r"united"i)
res

Or, searching indicators by some given description:

res = search_wdi("indicators", :description, r"gross national expenditure"i)
res[:name]

Some further search examples:

search_wdi("countries", :iso2c, r"TZ"i)
search_wdi("countries", :income, r"upper middle"i)
search_wdi("countries", :region, r"Latin America"i)
search_wdi("countries", :capital, r"^Ka"i)
search_wdi("countries", :lending, r"IBRD"i)
search_wdi("indicators", :name, r"gross national expenditure"i)
search_wdi("indicators", :description, r"gross national expenditure"i)
search_wdi("indicators", :source_database, r"Sustainable"i)
search_wdi("indicators", :source_organization,
           r"Global Partnership"i)[1:5, :]

You can subset your data with respect to rows, columns or individual entries through the ordinary TimeData functions. For example, selecting entries of US only:

data = wdi("NY.GNP.PCAP.CD", ["US","BR"], 1980, 2012, extra = true)
usData = chkDates(x-> x[:iso2c] .== "US", eachdate(data)) |>
         x -> asArr(x, Bool, false) |>
         x -> data[x[:], :]
usData

Furthermore, data can be visualized through the TimeData plotting functions. Simply call function loadPlotting to load Winston and Gadfly packages, and directly plot the data as Timenum object through wstPlot or gdfPlot:

data = wdi("AG.LND.ARBL.HA.PC", "US", 1900, 2011)
arableLand = convert(Timematr, data[symbol("AG.LND.ARBL.HA.PC")])

loadPlotting()

## using Winston
wstPlot(arableLand)

## using Gadfly
gdfPlot(arableLand)

In case of missing values, wdi will return an empty TimeData object without warning.

dfAS = wdi("EN.ATM.CO2E.KT", "AS")

Most of the package originally has been developed by 4gh (Frank Herrmann) and can still be accessed through the official Julia package repository and on github.