World Bank Data in Julia

The World Bank provides free access to data about development at data.worldbank.org.

The primary collection of development indicators is called World Development Indicators (WDI).

This module provides two functions to access and download the data: search_wdi() and wdi().

It follows roughly the R WDI package.

Installation

using Pkg
Pkg.add("WorldBankData")

Basic Examples

Obtain a DataFrame with the gross national income per capita for the US and Brazil:

using WorldBankData

df = wdi("NY.GNP.PCAP.CD", ["US","BR"])

The WDI indicator NY.GNP.PCAP.CD becomes the symbol NY_GNP_PCAP_CD in the DataFrame, i.e. . gets replaced by _.

Multiple countries and indicators

using WorldBankData

df = wdi(["NY.GNP.PCAP.CD","AG.LND.ARBL.HA.PC"], ["US","BR"]; startyear = 1980, endyear = 2008, extra=true)

This returns the GNI per capita and the arable land (hectares per person) for the time range 1980-2008 for the US and Brazil. It also attaches extra country information (the extra = true argument) like the capital, longitude, latitude, income range, etc.

Arguments

The wdi function has the following arguments:

wdi(indicators::Union{String,Array{String,1}}, countries::Union{String,Array{String,1}}; startyear::Integer=1800, endyear::Integer=3000, extra::Bool=false, verbose::Bool=false)

It needs a minimum of two arguments: the indicators (from the WDI database) and the countries (ISO two letter country codes). The rest are optional arguments.

startyear: First year to include.

endyear: Last year to include.

extra: If extra=true, wdi() attaches extra country data (like the capital) to the returned DataFrame.

verbose: If verbose=true, wdi() will print URL download information. This can be used as a progress indicator if many countries and indicators are requested.

Searching

The most convenient way to explore the database is probably through a web browser at data.worldbank.org.

However, the module does provide a search function: search_wdi().

One can search for "countries" or "indicators" data.

Example for country search by name

julia> using WorldBankData
julia> res = search_wdi("countries", "name", r"united"i)
julia> res.name
3-element DataArray{UTF8String,1}:
 "United Arab Emirates"
 "United Kingdom"
 "United States"
julia> res.iso2c
3-element DataArray{ASCIIString,1}:
 "AE"
 "GB"
 "US"

Example for indicator search by description

julia> using WorldBankData
julia> res = search_wdi("indicators", "description", r"gross national expenditure"i)
6x5 DataFrame
|-------|---------------------|------------|---------|
| Col # | Name                | Type       | Missing |
| 1     | description         | UTF8String | 0       |
| 2     | indicator           | UTF8String | 0       |
| 3     | name                | UTF8String | 0       |
| 4     | source_database     | UTF8String | 0       |
| 5     | source_organization | UTF8String | 0       |

julia> res.name
6-element DataArray{UTF8String,1}:
 "Gross national expenditure deflator (base year varies by country)"
 "Gross national expenditure (current US\$)"
 "Gross national expenditure (current LCU)"
 "Gross national expenditure (constant 2005 US\$)"
 "Gross national expenditure (constant LCU)"
 "Gross national expenditure (% of GDP)"

julia> res.indicator
6-element DataArray{UTF8String,1}:
 "NE.DAB.DEFL.ZS"
 "NE.DAB.TOTL.CD"
 "NE.DAB.TOTL.CN"
 "NE.DAB.TOTL.KD"
 "NE.DAB.TOTL.KN"
 "NE.DAB.TOTL.ZS"

The search_wdi() function

The search_wdi() function has the following arguments

search_wdi(data::String, entry::String, regx::Regex)::DataFrame

data: Either countries or indicators.

entry: One of the attributes (like name).

regex: Regular expression to search for.

"countries" can be searched for "name", "region", "capital", "iso2c", "iso3c", "income", and "lending".

"indicators" can be searched for "name", "description", "topics", "source_database", and "source_organization".

The search function uses two DataFrames country_cache and indicator_cache and searches through these. On the first search it will download the data from the World Bank website. This takes much longer for the larger indicators data. This only happens once per session. After the first use the data is cached.

Note that the last argument to search_wdi() is a regular expression denoted by r"..." and an i at the end means that it is case insensitive.

Examples of country searches

julia> search_wdi("countries", "iso2c", r"TZ"i)
1x9 DataFrame
|-------|---------|------------|-------|-------|----------|---------|-----------|----------|----------------------------------------|
| Row # | capital | income     | iso2c | iso3c | latitude | lending | longitude | name     | region                                 |
| 1     | Dodoma  | Low income | TZ    | TZA   | -6.17486 | IDA     | 35.7382   | Tanzania | Sub-Saharan Africa (all income levels) |


julia> search_wdi("countries", "income", r"upper middle"i)
...

julia> search_wdi("countries", "region", r"Latin America"i)
...

julia> search_wdi("countries", "capital", r"^Ka"i)
3x9 DataFrame
|-------|-----------|------------|-------|-------|----------|---------|-----------|-------------|----------------------------------------|
| Row # | capital   | income     | iso2c | iso3c | latitude | lending | longitude | name        | region                                 |
| 1     | Kabul     | Low income | AF    | AFG   | 34.5228  | IDA     | 69.1761   | Afghanistan | South Asia                             |
| 2     | Kathmandu | Low income | NP    | NPL   | 27.6939  | IDA     | 85.3157   | Nepal       | South Asia                             |
| 3     | Kampala   | Low income | UG    | UGA   | 0.314269 | IDA     | 32.5729   | Uganda      | Sub-Saharan Africa (all income levels) |

julia> search_wdi("countries", "lending", r"IBRD"i)
...

Examples of indicator searches

julia> search_wdi("indicators", "name", r"gross national expenditure"i)
...
julia> search_wdi("indicators", "description", r"gross national expenditure"i)
...
julia> search_wdi("indicators", "source_database", r"Sustainable"i)
...
julia> search_wdi("indicators", "source_organization", r"Global Partnership"i)

Tips and Tricks

Extracting country data from results

df = wdi("NY.GNP.PCAP.CD", ["US","BR"]; startyear = 1980, endyear = 2012, extra = true)
us_gnp = df[df.iso2c .== "US", :]

Year format

For similarity with the R WDI package the :year column is in Float64 format. WDI data is yearly.

You can easily convert this to a Date series:

using WorldBankData
using Dates

df = wdi("AG.LND.ARBL.HA.PC", "US", startyear = 1900, endyear = 2011)
df.year = Date.(df.year)

Plotting

Install the Plots.jl package with Pkg.add("Plots").

using WorldBankData
using Plots

df = wdi("AG.LND.ARBL.HA.PC", "US"; startyear = 1980, endyear = 2010)

plot(df.year, df.AG_LND_ARBL_HA_PC)

Cache

The data in the World Bank database changes infrequently. Therefore it makes little sense to download it every time a script is run.

Metadata

The search_wdi() function internally caches the country and indicator metadata and therefore downloads the country and indicator data only once per session. Even that is usually not necessary. This data can easily be stored on local disk.

Download and store the country and indicator information in csv files:

using WorldBankData
using DataFrames
using CSV
CSV.write("country_cache.csv",WorldBankData.get_countries())
CSV.write("indicator_cache.csv", WorldBankData.get_indicators())

These can be used in the script to set the WorldBankData cache variables WorldBankData.country_cache and WorldBankData.indicator_cache (which are initialized to false) using the WorldBankData.set_country_cache() and WorldBankData.set_indicator_cache() functions:

using WorldBankData
using DataFrames
WorldBankData.set_country_cache(CSV.read("country_cache.csv"))
WorldBankData.set_indicator_cache(CSV.read("indicator_cache.csv"))

From then on the search_wdi() function will use the data read from disk.

The caches can be reset with WorldBankData.reset_country_cache() and WorldBankData.reset_indicator_cache().

Indicator data

In a similar way the indicator data itself can be cached.

using WorldBankData
using DataFrames
using CSV

function update_us_gnp_per_cap()
    df = wdi("NY.GNP.PCAP.CD", "US")
    CSV.write("us_gnp.csv",df)
end

df = CSV.read("us_gnp.csv")

one then runs the update_us_gnp_per_cap() function only when needed.

nilshg/WorldBankData.jl