KPMG Technical Assesment - Statistical Analysis on GDP per capita using R
Janelle Tang
25/03/2020
INTRODUCTION
This is an analysis of data from different countries, focusing on GDP per capita and its predictors. The dataset was retrieved from Kaggle, and contains information on population, region, area size, infant mortality and more.
countries<-read.csv("/Users/janelletang/Documents/git/countries of the world.csv",na.strings=c("", "NA"),dec=",")
head(countries)## Country Region Population
## 1 Afghanistan ASIA (EX. NEAR EAST) 31056997
## 2 Albania EASTERN EUROPE 3581655
## 3 Algeria NORTHERN AFRICA 32930091
## 4 American Samoa OCEANIA 57794
## 5 Andorra WESTERN EUROPE 71201
## 6 Angola SUB-SAHARAN AFRICA 12127071
## Area..sq..mi.. Pop..Density..per.sq..mi.. Coastline..coast.area.ratio.
## 1 647500 48.0 0.00
## 2 28748 124.6 1.26
## 3 2381740 13.8 0.04
## 4 199 290.4 58.29
## 5 468 152.1 0.00
## 6 1246700 9.7 0.13
## Net.migration Infant.mortality..per.1000.births. GDP....per.capita.
## 1 23.06 163.07 700
## 2 -4.93 21.52 4500
## 3 -0.39 31.00 6000
## 4 -20.71 9.27 8000
## 5 6.60 4.05 19000
## 6 0.00 191.19 1900
## Literacy.... Phones..per.1000. Arable.... Crops.... Other.... Climate
## 1 36.0 3.2 12.13 0.22 87.65 1
## 2 86.5 71.2 21.09 4.42 74.49 3
## 3 70.0 78.1 3.22 0.25 96.53 1
## 4 97.0 259.5 10.00 15.00 75.00 2
## 5 100.0 497.2 2.22 0.00 97.78 3
## 6 42.0 7.8 2.41 0.24 97.35 NA
## Birthrate Deathrate Agriculture Industry Service
## 1 46.60 20.34 0.380 0.240 0.380
## 2 15.11 5.22 0.232 0.188 0.579
## 3 17.14 4.61 0.101 0.600 0.298
## 4 22.46 3.27 NA NA NA
## 5 8.71 6.25 NA NA NA
## 6 45.11 24.20 0.096 0.658 0.246#Renaming Columns
colnames(countries) <- c("Country", "Region", "Population", "Area_SqMiles", "PopDens",
"Coastline", "NetMigration", "InfantMortality", "GDPpc",
"LiteracyRate", "Phones", "Arable_percent", "Crops_percent",
"Other_percent", "ClimateRating","BirthRate","DeathRate",
"PrimarySector_percent","SecondarySector_percent","TertiarySector_percent")
summary(countries)## Country Region
## Afghanistan : 1 SUB-SAHARAN AFRICA :51
## Albania : 1 LATIN AMER. & CARIB :45
## Algeria : 1 ASIA (EX. NEAR EAST) :28
## American Samoa : 1 WESTERN EUROPE :28
## Andorra : 1 OCEANIA :21
## Angola : 1 NEAR EAST :16
## (Other) :221 (Other) :38
## Population Area_SqMiles PopDens
## Min. :7.026e+03 Min. : 2 Min. : 0.00
## 1st Qu.:4.376e+05 1st Qu.: 4648 1st Qu.: 29.15
## Median :4.787e+06 Median : 86600 Median : 78.80
## Mean :2.874e+07 Mean : 598227 Mean : 379.05
## 3rd Qu.:1.750e+07 3rd Qu.: 441811 3rd Qu.: 190.15
## Max. :1.314e+09 Max. :17075200 Max. :16271.50
##
## Coastline NetMigration InfantMortality GDPpc
## Min. : 0.00 Min. :-20.99000 Min. : 2.29 Min. : 500
## 1st Qu.: 0.10 1st Qu.: -0.92750 1st Qu.: 8.15 1st Qu.: 1900
## Median : 0.73 Median : 0.00000 Median : 21.00 Median : 5550
## Mean : 21.17 Mean : 0.03812 Mean : 35.51 Mean : 9690
## 3rd Qu.: 10.35 3rd Qu.: 0.99750 3rd Qu.: 55.70 3rd Qu.:15700
## Max. :870.66 Max. : 23.06000 Max. :191.19 Max. :55100
## NA's :3 NA's :3 NA's :1
## LiteracyRate Phones Arable_percent Crops_percent
## Min. : 17.60 Min. : 0.2 Min. : 0.00 Min. : 0.000
## 1st Qu.: 70.60 1st Qu.: 37.8 1st Qu.: 3.22 1st Qu.: 0.190
## Median : 92.50 Median : 176.2 Median :10.42 Median : 1.030
## Mean : 82.84 Mean : 236.1 Mean :13.80 Mean : 4.564
## 3rd Qu.: 98.00 3rd Qu.: 389.6 3rd Qu.:20.00 3rd Qu.: 4.440
## Max. :100.00 Max. :1035.6 Max. :62.11 Max. :50.680
## NA's :18 NA's :4 NA's :2 NA's :2
## Other_percent ClimateRating BirthRate DeathRate
## Min. : 33.33 Min. :1.000 Min. : 7.29 Min. : 2.290
## 1st Qu.: 71.65 1st Qu.:2.000 1st Qu.:12.67 1st Qu.: 5.910
## Median : 85.70 Median :2.000 Median :18.79 Median : 7.840
## Mean : 81.64 Mean :2.139 Mean :22.11 Mean : 9.241
## 3rd Qu.: 95.44 3rd Qu.:3.000 3rd Qu.:29.82 3rd Qu.:10.605
## Max. :100.00 Max. :4.000 Max. :50.73 Max. :29.740
## NA's :2 NA's :22 NA's :3 NA's :4
## PrimarySector_percent SecondarySector_percent TertiarySector_percent
## Min. :0.00000 Min. :0.0200 Min. :0.0620
## 1st Qu.:0.03775 1st Qu.:0.1930 1st Qu.:0.4293
## Median :0.09900 Median :0.2720 Median :0.5710
## Mean :0.15084 Mean :0.2827 Mean :0.5653
## 3rd Qu.:0.22100 3rd Qu.:0.3410 3rd Qu.:0.6785
## Max. :0.76900 Max. :0.9060 Max. :0.9540
## NA's :15 NA's :16 NA's :15dim(countries)## [1] 227 20Some obvservations appear to have missing values e.g. GDP per capita should not be NA.
full_countries<-countries
#removing observations with missing values
countries<-na.omit(countries)#extract legend of a ggplot graph
#https://github.com/hadley/ggplot2/wiki/Share-a-legend-between-two-ggplot2-graphs
get_legend<-function(a.gplot){
tmp <- ggplot_gtable(ggplot_build(a.gplot))
leg <- which(sapply(tmp$grobs, function(x) x$name) == "guide-box")
legend <- tmp$grobs[[leg]]
return(legend)}####Graphical Analysis
#Bar graph to compare the mean GDP per capita for each region
meanGDP_1<- ggplot(countries) +
geom_bar(aes(Region, GDPpc/1000, fill = as.factor(Region)),
position = "dodge", stat = "summary", fun.y = "mean") +
theme(axis.text.x=element_blank())+
labs(y="GDP per capita ('000s)",
title = "Average GDP per capita",fill="Regions")
#Boxplot to show the range for each region
meanGDP_2<- ggplot(countries, aes(x=Region, y=GDPpc, fill=as.factor(Region))) +
geom_boxplot(alpha=0.3)+
theme(axis.text.x=element_blank(),legend.position="none")+
labs(title = "Boxplot of GDP per capita",fill="Regions")
region_legend<-get_legend(meanGDP_1)
grid.arrange(arrangeGrob(meanGDP_1+theme(legend.position = "none"),
meanGDP_2,nrow=2),
region_legend,ncol=2,widths=c(2.2,1))
#Plot the density of GDP per capita
densityGDP<-plot (density (countries$GDPpc), main="GDP per capita")
rug (countries$GDPpc) #adding a "rug" to density estimate - 
#makes it possible to discern the individual data pointsIt appears that the North American region has the highest average GDP per capita, with a very small range, while the Sub-Saharan African region has the lowest, again with a very small range, but with more outliers.
Asia, both near east and elsewhere, have a large range.
#Total Population per region
pop_plot1<-ggplot(countries) +
geom_bar(aes(Region, Population/1000000, fill = as.factor(Region)),
position = "dodge", stat = "summary", fun.y = "sum") +
theme(axis.text.x=element_blank())+
labs(title = "Total Population",y="Total Population (millions)",
fill="Regions")
#Average Population Density per region
pop_plot2<-ggplot(countries) +
geom_bar(aes(Region, PopDens, fill = as.factor(Region)), position = "dodge",
stat = "summary", fun.y = "mean") +
theme(axis.text.x=element_blank())+
labs(title = "Average Population Density",y="Population Density (Person/sqmile)",
fill="Regions")
grid.arrange(arrangeGrob(pop_plot2+theme(legend.position="none"),
pop_plot1+theme(legend.position="none")),
region_legend,ncol=2,widths=c(2.2,1))
Comparing population data to mean GDP per capita for each region, Lets see if there are any variables that have a strong correlation to GDP per capita.
cor_matrix<-round(cor(countries[3:20]),1)
cor_GDPpc<-cor(countries$GDPpc,countries[4:20])
cor_GDPpc[1,]## Area_SqMiles PopDens Coastline
## 0.06835595 0.19012217 0.03581518
## NetMigration InfantMortality GDPpc
## 0.37879042 -0.63908984 1.00000000
## LiteracyRate Phones Arable_percent
## 0.52288013 0.88352011 0.04646475
## Crops_percent Other_percent ClimateRating
## -0.20784374 0.06644526 0.36056651
## BirthRate DeathRate PrimarySector_percent
## -0.65879545 -0.24756242 -0.61691880
## SecondarySector_percent TertiarySector_percent
## 0.03285465 0.53655075#Graphical visualisation of correlation between variables
corrplot(cor_matrix, method="color",type="upper")
corrplot(cor_GDPpc, method="color", cl.pos='n')
- GDP per capita has a relatively strong positive correlation with the number of phones, literacy rate, and the proportional size of the tertiary sector
- GDP per capital has a positive correlation with population density, net migration, and climate rating.
- GDP per capital has a weak positive correlation with area, coastline, and percentage of arable land.
- GDP per capita has a relatively strong negative correlation with infant mortality, the size of the proportional size of the primary sector and birth rate.
- GDP per capital has a negative correlation with death rate, percentage of crops.
- GDP per capital has a weak negative correlation with percentage of crops.
From an economic standpoint, many of these observations make sense. It seems obvious that a country with high GDP per capita would have a higher number of phone usage, since higher income means greater ability to afford phone (necessity vs. luxury). Also, higher GDP per capita economies have a larger tertiary sector (since GDP is calculated based on total spending). Similarly, with economic growth, the birth rate would drop due to a number of factors, for example less time for individuals to raise a child, and more years spent in higher education, as well as a deeper understanding of contraception. Additionally, infant mortality drops with economic growth as
Lets take a closer look at some of the highly correlated predictors with GDP per capita.
#create new data frame
countries_infant<-data.frame(full_countries[,c(1,8,9)])
countries_infant<-na.omit(countries_infant)
#plotting the infant mortality rate to show clear correlation with gdppc
infant_graph<-ggplot(countries_infant, aes(y=GDPpc,x=InfantMortality),fill=InfantMortality)+
geom_point(color=3)+
geom_smooth(color=4)+
labs(y="GDP per capita",title = "Infant Mortality against GDP per capita")
#create new data frame
countries_lit<-data.frame(full_countries[,c(1,9,10)])
countries_lit<-na.omit(countries_lit)
#plotting literacy rates against GDPpc
lit_graph<-ggplot(countries_lit, aes(LiteracyRate,GDPpc),fill=LiteracyRate)+
geom_point(color=3)+
geom_smooth(color=4)+
labs(y="GDP per capita",title = "Literacy rate against GDP per capita")
grid.arrange(infant_graph,lit_graph,ncol=2)## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
While real GDP remains the most used when measuring economic growth and living standards, in many cases, the human development index (HDI) is used to measure economic development, which is similar to economic growth, but encompasses many other factors, such as poverty rate, income equality, sustainability etc. The HDI is a composite measure comprising of:
- GDP per capita
- Life Expectancy from birth
- Literacy rate (mean years of schooling)
#create new data frame for analysing birth and death rates
countries_bd<-data.frame(full_countries[,c(1,9,16,17)])
countries_bd[countries_bd==""] <- NA
countries_bd<-na.omit(countries_bd)
head(countries_bd)## Country GDPpc BirthRate DeathRate
## 1 Afghanistan 700 46.60 20.34
## 2 Albania 4500 15.11 5.22
## 3 Algeria 6000 17.14 4.61
## 4 American Samoa 8000 22.46 3.27
## 5 Andorra 19000 8.71 6.25
## 6 Angola 1900 45.11 24.20#plotting the birth and death rates data to see an overview trend
birth_graph<-ggplot(countries_bd, aes(y=GDPpc,x=BirthRate),fill=BirthRate)+
geom_point(color=3)+
geom_smooth(color=4)+
labs(y="GDP per capita",title = "Birth rate against GDP per capita")
death_graph<-ggplot(countries_bd, aes(y=GDPpc,x=DeathRate),fill=DeathRate)+
geom_point(color=6)+
geom_smooth(color=4)+geom_smooth(method="lm",color=7)+
labs(y="GDP per capita",title = "Death rate against GDP per capita")
grid.arrange(birth_graph, death_graph,ncol=2)## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
## `geom_smooth()` using method = 'loess' and formula 'y ~ x'
We can see that death rates actually vary quite a bit, with a slight downward trend, but with great fluctuations. Initially, in low development countries, the death rate appears to be quite sparse, which can be reflective of the different types of living conditions and climates that vary across LEDCs. However, with industrialisation, the death rate should drop quickly, by removing high cause of death factors such as malnourishment, and curable diseases. However, at higher levesl of GDP per capita, the death rate begins to rise again. We could theorise that individuals would be working more, which results in a more unhealthy lifestyle, and less energy to care of the elderly. Additionally, with higher life expectancy comes more health risks such as cancer. However, there would need to be additional analysis on datasets focusing on death rates with lifestyle choices to make a clear judgement.
br/> We also see a similar effect to literacy rates and infant mortality - with higher GDP per capita, the variability of birth and death rates decreases.
Another significant factor is the composition of primary, secondary, and tertiary sector in an economy.
#create new data frame
countries_sec<-data.frame(full_countries[,c(1,9,18,19,20)])
countries_sec<-na.omit(countries_sec)
head(countries_sec)## Country GDPpc PrimarySector_percent SecondarySector_percent
## 1 Afghanistan 700 0.380 0.240
## 2 Albania 4500 0.232 0.188
## 3 Algeria 6000 0.101 0.600
## 6 Angola 1900 0.096 0.658
## 7 Anguilla 8600 0.040 0.180
## 8 Antigua & Barbuda 11000 0.038 0.220
## TertiarySector_percent
## 1 0.380
## 2 0.579
## 3 0.298
## 6 0.246
## 7 0.780
## 8 0.743#create data frame for top and bottom 25 countries ranked by GDP per capita
top25<-head(arrange(countries_sec,desc(GDPpc)), n = 25)
top25<-top25[,!(names(top25) == 'GDPpc')]
top25$Country <- factor(top25$Country, levels = top25$Country)
#arrange data vertical so that we can plot a stacked bar graph
top25.m <- melt(top25,id.vars = "Country")
bottom25<-head(arrange(countries_sec,desc(-GDPpc)), n = 25)
bottom25<-bottom25[,!(names(bottom25) == 'GDPpc')]
bottom25$Country <- factor(bottom25$Country, levels = bottom25$Country)
bottom25.m <- melt(bottom25,id.vars = "Country")
#Graphical visualisation of proportion of sectors
ggplot(top25.m, aes(x = Country, y = value,fill=variable)) +
geom_bar(stat='identity')+
theme(axis.text.x=element_text(angle=90,hjust=1,vjust=0.2),
axis.title.y=element_blank())+
scale_y_continuous(labels=percent)+
labs(title = "Proportion of different sectors in high GDP per capita countries",
fill="Sectors")
ggplot(bottom25.m, aes(x = Country, y = value,fill=variable)) +
geom_bar(stat='identity')+
theme(axis.text.x=element_text(angle=90,hjust=1,vjust=0.2),
axis.title.y=element_blank())+
scale_y_continuous(labels=percent)+
labs(title = "Proportion of different sectors in low GDP per capita countries",
fill="Sectors")
It is clear that for the higher GDP per capita countries, services take up a larger proportion of the economy, while agricultural/primary commodities take up a very insignificant percentage. The opposite is often true for low GDP per capita countries, where primary commodities take up a larger proportion - this is not always true, as in the case of East Timor, with the lowest GDP per capita, among others. It is likely that these are ust outliers. Additionally, most of these low income per capita countries appear to have a wider variety of composition.