Background

From Shaefer et al 2016:

Can Poverty in America Be Compared to Conditions in the World’s Poorest Countries?

World Development Indicators used in this example:

There are currently two R packages on CRAN with some differences, we compare them here:

1. Get data

First we download an updated list of inidicators. Both packages come with a downloaded list. That list appears to be still current for wbstats, as it is a newer package, for WDI however it is not up to date.

wbstats

library(wbstats)
new_wb_cache <- wbcache()

WDI

library(WDI)
new_wdi_cache <- WDIcache()

Next we find out the IDs for the indicators we are intersted in: GDP, life expectancy, and infant mortality.

Both packages support regular expressions for the search for indicator IDs.

wbstats

wbsearch("gdp.*capita.*US\\$", cache = new_wb_cache)
##         indicatorID                          indicator
## 7728 NY.GDP.PCAP.CD       GDP per capita (current US$)
## 7730 NY.GDP.PCAP.KD GDP per capita (constant 2010 US$)
wbsearch("life expectancy at birth.*total", cache = new_wb_cache)
##          indicatorID                               indicator
## 14389 SP.DYN.LE00.IN Life expectancy at birth, total (years)
wbsearch("^mortality.*rate.*infant", cache = new_wb_cache)
##             indicatorID
## 14380 SP.DYN.IMRT.FE.IN
## 14381    SP.DYN.IMRT.IN
## 14382 SP.DYN.IMRT.MA.IN
##                                                    indicator
## 14380 Mortality rate, infant, female (per 1,000 live births)
## 14381         Mortality rate, infant (per 1,000 live births)
## 14382   Mortality rate, infant, male (per 1,000 live births)

WDI

WDIsearch("gdp.*capita.*US\\$", cache = new_wdi_cache)
##      indicator        name                                
## [1,] "NY.GDP.PCAP.CD" "GDP per capita (current US$)"      
## [2,] "NY.GDP.PCAP.KD" "GDP per capita (constant 2010 US$)"
WDIsearch("life expectancy at birth.*total", cache = new_wdi_cache)
##                                 indicator 
##                          "SP.DYN.LE00.IN" 
##                                      name 
## "Life expectancy at birth, total (years)"
WDIsearch("^mortality.*rate.*infant", cache = new_wdi_cache)
##      indicator          
## [1,] "SP.DYN.IMRT.FE.IN"
## [2,] "SP.DYN.IMRT.IN"   
## [3,] "SP.DYN.IMRT.MA.IN"
##      name                                                    
## [1,] "Mortality rate, infant, female (per 1,000 live births)"
## [2,] "Mortality rate, infant (per 1,000 live births)"        
## [3,] "Mortality rate, infant, male (per 1,000 live births)"

Finally we download the data. The download functions provided in the two packages mainly differ in their arguments and in the format of the data they return.

The WDI()function takes the argument extra=.. which can be set to TRUE to download more data fields (e.g. region, iso3c code, income_level).
The the wb() function does not have these particular options. In order to get this additional information (for exampe, to link countries to regions) it is necessary to download a separate table. Note, though that wb() comes with other options (e.g. download of quarterly or monthly data, most recent values) that WDI() does not have.

wbstats

wb_dat <- wb(indicator = c("NY.GDP.PCAP.KD", "SP.DYN.LE00.IN", "SP.DYN.IMRT.IN")) 
names(wb_dat)
## [1] "value"       "date"        "indicatorID" "indicator"   "iso2c"      
## [6] "country"

WDI

wdi_dat <- WDI(indicator = c("NY.GDP.PCAP.KD", "SP.DYN.LE00.IN", "SP.DYN.IMRT.IN"), start = 1960, end = 2015, extra = TRUE) 
names(wdi_dat)
##  [1] "iso2c"          "country"        "year"           "NY.GDP.PCAP.KD"
##  [5] "SP.DYN.LE00.IN" "SP.DYN.IMRT.IN" "iso3c"          "region"        
##  [9] "capital"        "longitude"      "latitude"       "income"        
## [13] "lending"

WDI() returns a data frame in wide format. Rows are per year and country/region, columns are the different indicators.
wb() returns a data frame in long format. Each row is one observation.

2. Clean up

We remove all entries that are aggregated regional values.

Using the wbstats package we need to add two extra steps to download a separate table with the regions names and join it to our data table before eliminating the undesired rows.

wbstats

wb_countries <- wbcountries() 
names(wb_countries)
##  [1] "iso3c"     "iso2c"     "country"   "capital"   "long"     
##  [6] "lat"       "regionID"  "region"    "adminID"   "admin"    
## [11] "incomeID"  "income"    "lendingID" "lending"
wb_dat <- merge(wb_dat, y = wb_countries[c("iso2c", "region")], by = "iso2c", all.x = TRUE)

wb_dat <- subset(wb_dat, region != "Aggregates") # this also removes NAs

WDI

wdi_dat <- subset(wdi_dat, region != "Aggregates") # this also removes NAs

Now we rename the indicators to something we understand better.

With wbstats we need to add an extra step to reformat the table into wide format that we will use as input for the plotting.

wbstats

wb_dat$indicatorID[wb_dat$indicatorID == "NY.GDP.PCAP.KD"] <- "GDP"
wb_dat$indicatorID[wb_dat$indicatorID == "SP.DYN.LE00.IN"] <- "life_expectancy"
wb_dat$indicatorID[wb_dat$indicatorID == "SP.DYN.IMRT.IN"] <- "infant_mortality"

library(reshape2)
wb_dat <- dcast(wb_dat, iso2c + country + date + region ~ indicatorID,  value.var = 'value')

WDI

names(wdi_dat)[which(names(wdi_dat) == "NY.GDP.PCAP.KD")] <- "GDP"
names(wdi_dat)[which(names(wdi_dat) == "SP.DYN.LE00.IN")] <- "life_expectancy"
names(wdi_dat)[which(names(wdi_dat) == "SP.DYN.IMRT.IN")] <- "infant_mortality"

Now we have pretty much identical tables.

3. Plot a graph

wbstats

library(ggplot2)
ggplot(subset(wb_dat, date == "2008"), aes(x = GDP, y = infant_mortality)) + geom_point()

WDI

ggplot(subset(wdi_dat, year == 2008), aes(x = GDP, y = infant_mortality)) + geom_point()

4. Reproduce graphs from the paper

Life Expectancy

lifexp_countries <- subset(wb_dat, country %in% c("United States", "Rwanda", "Mongolia", "Pakistan", "Lao PDR", "Bhutan", "Malaysia", "Brazil", "Ireland", "Japan", "Sweden", "Netherlands"))

ggplot(subset(lifexp_countries, date == "2008"), aes(x = GDP, y = life_expectancy, color = country == "United States")) + 
  geom_point() +  
  geom_text(aes(label = country), size=3, nudge_y = 0.4) +
  scale_x_continuous(limits = c(0, 70000))

(From Shaefer et al 2016)

(From Shaefer et al 2016)

Infant Mortality

infmort_countries <- subset(wb_dat, country %in% c("United States", "Tonga", "Colombia", "Grenada", "Sri Lanka", "Malta", "Germany", "Japan", "Sweden", "Netherlands"))

ggplot(subset(infmort_countries, date == "2008"), aes(x = GDP, y = infant_mortality, color = country == "United States")) + 
  geom_point() +  
  geom_text(aes(label = country), size=3, nudge_y = 0.2) +
  scale_x_continuous(limits = c(0, 70000))

(From Shaefer et al 2016)

(From Shaefer et al 2016)

5. Simple interactivity

library(plotly)

g <- ggplot(subset(infmort_countries, date > "1999" & date <= "2015"), aes(x = GDP, y = infant_mortality, color = country == "United States", tooltip = country)) + 
  geom_point() + 
  facet_wrap(~ date) + 
  theme(legend.position="none")

ggplotly(g)

Reference:

H. Luke Shaefer, Pinghui Wu, Kathryn Edin, 2016: Can Poverty in America Be Compared to Conditions in the World’s Poorest Countries? Publication, Stanford Center on Poverty & Inequality. http://inequality.stanford.edu/publications/media/details/can-poverty-america-be-compared-conditions-worlds-poorest-countries