The data set was taken from the website of the national institute of informatics. Downloaded in excel format, it contains info about annual exportation of products by sector, given in thousands of USD.
The excel file is separated in different sheets per state. Each sheet contains the same information: annual exportation of products by sector, given in thousands of USD, where each column contains information of the year, from 2007 to 2017.
Due to the format of the file, our first step will be to tidy up the data in a format that is useful to R and more manageable by us.
## Calling libraries
library(readxl)
library(tidyverse)
library(cowplot)
## Import from Excel
## The sheet names are the name of each state
estados <- excel_sheets('ignore/EAEF_Entidad_Subsector.xlsx')
## Import column names and avoiding useless rows
exportation <- read_xlsx('ignore/EAEF_Entidad_Subsector.xlsx',
range = "A5:N5", sheet = 1,
col_types = c("numeric", "text",
rep("numeric", 12)))
## Create state column
exportation <- mutate(exportation, state = "Aguascalientes")
## Iterate over every sheet
for (i in 1:length(estados)){
extracted <- read_xlsx('ignore/EAEF_Entidad_Subsector.xlsx',
range = "A7:N31", sheet = i,
col_names = F, na = "-",
col_types = c("numeric", "text",
rep("numeric", 12)))
colnames(extracted) <- colnames(exportation)[1:14]
extracted <- mutate(extracted,
state = estados[i])
exportation <- full_join(exportation, extracted)
}
## Fix couple of column names
colnames(exportation)[c(3, 14)] <- c("2007", "2018")
exportation$`Código` <- parse_factor(as.character(exportation$`Código`))
## Tidying data
export1 <- exportation %>%
pivot_longer(cols = `2007`:`2018`,
names_to = "year",
values_to = "USD",
values_drop_na = T) %>%
mutate(year = parse_double(year))
export <- export1 %>%
pivot_wider(names_from = "Descripción",
values_from = USD,
id_cols = c("state", "year"))
## Save our new data frames in a useful way
write_csv(export1, "exportations_activity_rows.csv")
write_csv(export, "exportations_activity_cols.csv")Now we have 2 csv tables, exportations_activity_rows.csv (created from
export1) which keeps each activity as a row, and
exportations_activity_cols (from export) which transform each
activity into a column using pivot_wider. Each of the 2 can help us to
analyse the data in different ways. We decided to keep them as csv files
to make its access easier for GitHub users.
Now if we start a new session, all we have to do is to call our libraries and load the data from CSV:
library(tidyverse)
library(cowplot)
export.rows <- read_csv("exportations_activity_rows.csv")
export.cols <- read_csv("exportations_activity_cols.csv")As the analysis developed, we found the importance of translating the variable names to English. Not only for the reader to understand, but also due to the long string and complexity of the categories in Spanish. I could first translate and then save the csv files, however I encountered this problem and fixed it once the csv files were already created, and in honour to that, I decided to keep it in the same way for this publication. Therefore, before diving into the EDA, here is the code I used to translate the variables from Spanish (ES) to English (EN).
## Spanish names in a vector
colnames(export.cols)## [1] "state"
## [2] "year"
## [3] "Exportaciones totales"
## [4] "Industria alimentaria"
## [5] "Industria de las bebidas y el tabaco"
## [6] "Fabricación de insumos textiles y acabado de textiles"
## [7] "Fabricación de productos textiles, excepto prendas de vestir"
## [8] "Fabricación de prendas de vestir"
## [9] "Industria del papel"
## [10] "Industria química"
## [11] "Industria del plástico y del hule"
## [12] "Fabricación de productos a base de minerales no metálicos"
## [13] "Industrias metálicas básicas"
## [14] "Fabricación de productos metálicos"
## [15] "Fabricación de maquinaria y equipo"
## [16] "Fabricación de equipo de computación, comunicación, medición y de otros equipos, componentes y accesorios electrónicos"
## [17] "Fabricación de equipo de transporte"
## [18] "Fabricación de muebles, colchones y persianas"
## [19] "Otras industrias manufactureras"
## [20] "Subsectores no especificados"
## [21] "Minería de minerales metálicos y no metálicos, excepto petróleo y gas"
## [22] "Curtido y acabado de cuero y piel, y fabricación de productos de cuero, piel y materiales sucedáneos"
## [23] "Industria de la madera"
## [24] "Impresión e industrias conexas"
## [25] "Fabricación de accesorios, aparatos eléctricos y equipo de generación de energía eléctrica"
## [26] "Extracción de petróleo y gas"
## [27] "Fabricación de productos derivados del petróleo y del carbón"
categorias <- colnames(export.cols)[3:27]
## Shorter equivalents in English
activities.en <- c("Total", "Food", "Drinks and tobacco",
"Textiles", "Textile products", "Tailoring",
"Paper", "Chemistry", "Plastic",
"Minerals based", "Metal industry", "Metal products",
"Machinery", "Electronics", "Transport equipment",
"Furniture", "Other manufactures", "Not specified",
"Mining", "Leather", "Wood",
"Printing", "Electricity", "Petroleum",
"Petroleum products")
## Change column names
colnames(export.cols)[3:27] <- activities.enTo change the values in export.rows we will need to convert the
Spanish expressions into English ones. Here functional programming comes
very handy: first we create our function translate, which will do the
final job, and then we fill the gaps, in this case, we create
equivalent, which will find the equivalent expression in each
language. However, equivalent must be run before translate to work.
translate <- function(vector.es){
vector.en <- c()
for (i in 1:length(vector.es)){
expression.es <- vector.es[i]
expression.en <- equivalent(expression.es)
## Here "equivalent" should take expression.es and return the
## equivalent in English
vector.en <- append(vector.en, expression.en)
}
vector.en
}
equivalent <- function(expression.es){
position <- match(expression.es, categorias)
expression.en <- activities.en[position]
expression.en
}
## Testing our new functions
equivalent("Impresión e industrias conexas")## [1] "Printing"
translate(categorias)## [1] "Total" "Food" "Drinks and tobacco"
## [4] "Textiles" "Textile products" "Tailoring"
## [7] "Paper" "Chemistry" "Plastic"
## [10] "Minerals based" "Metal industry" "Metal products"
## [13] "Machinery" "Electronics" "Transport equipment"
## [16] "Furniture" "Other manufactures" "Not specified"
## [19] "Mining" "Leather" "Wood"
## [22] "Printing" "Electricity" "Petroleum"
## [25] "Petroleum products"
## Creating a new column "Activity" with the English expressions
export.rows <- mutate(export.rows,
Activity = translate(`Descripción`))Now to start with the EDA, let’s look at the totals by state, using
export.cols
export.cols %>%
group_by(state) %>%
summarise(`total export` = sum(Total)) %>%
arrange(desc(`total export`)) %>%
print(n = Inf)## # A tibble: 32 x 2
## state `total export`
## <chr> <dbl>
## 1 Chihuahua 466861927
## 2 Baja California 398935507
## 3 Coahuila de Zaragoza 355638907
## 4 Nuevo León 330267052
## 5 Tamaulipas 284435973
## 6 Campeche 264100465
## 7 Jalisco 213931233
## 8 México 188357470
## 9 Sonora 179661021
## 10 Guanajuato 167191962
## 11 Puebla 127934390
## 12 Tabasco 115797563
## 13 San Luis Potosí 94812554
## 14 Querétaro 88633615
## 15 Aguascalientes 79688240
## 16 Veracruz de Ignacio de la Llave 68556313
## 17 Morelos 37397175
## 18 Zacatecas 34010223
## 19 Ciudad de México 32037661
## 20 Hidalgo 19504479
## 21 Durango 17431796
## 22 Yucatán 14496875
## 23 Michoacán de Ocampo 13411397
## 24 Chiapas 13291536
## 25 Tlaxcala 12987607
## 26 Oaxaca 11023551
## 27 Sinaloa 7825439
## 28 Guerrero 5918438
## 29 Colima 2518028
## 30 Baja California Sur 2303491
## 31 Nayarit 1146388
## 32 Quinta Roo 517674
## Graphical mode
export.cols %>%
group_by(state) %>%
summarise(`total export` = sum(Total)) %>%
ggplot() +
geom_bar(aes(y = `total export`,
x = reorder(state, `total export`, FUN = abs),
fill = `total export`),
stat = 'identity') +
coord_flip()
Now, to do the same but by category, we could simply use export.rows
export.rows %>%
filter(Activity != "Total") %>%
group_by(Activity) %>%
summarise(Total = sum(USD)) %>%
arrange(desc(Total)) %>%
print(n = Inf)## # A tibble: 24 x 2
## Activity Total
## <chr> <dbl>
## 1 Transport equipment 1226859499
## 2 Electronics 747959073
## 3 Petroleum 397933968
## 4 Electricity 208582754
## 5 Other manufactures 147915402
## 6 Machinery 136957553
## 7 Chemistry 133570853
## 8 Metal industry 117915995
## 9 Metal products 82889135
## 10 Food 81653585
## 11 Plastic 80126816
## 12 Mining 52953993
## 13 Not specified 51470567
## 14 Tailoring 43913959
## 15 Drinks and tobacco 31059501
## 16 Minerals based 30584505
## 17 Furniture 19883596
## 18 Petroleum products 14565067
## 19 Paper 13876523
## 20 Leather 9863853
## 21 Printing 6915538
## 22 Textiles 6260722
## 23 Textile products 4954252
## 24 Wood 1959275
export.rows %>%
filter(Activity != "Total") %>%
group_by(Activity) %>%
summarise(Total = sum(USD)) %>%
ggplot() +
geom_bar(aes(y = Total,
x = reorder(Activity, Total, FUN = abs),
fill = Total),
stat = 'identity') +
coord_flip()
Finally, total exportations per year:
export.cols %>%
group_by(year) %>%
summarise(`total export` = sum(Total)) %>%
print(n = Inf)## # A tibble: 12 x 2
## year `total export`
## <dbl> <dbl>
## 1 2007 237809741
## 2 2008 257967777
## 3 2009 198234125
## 4 2010 258504747
## 5 2011 299732519
## 6 2012 320014188
## 7 2013 329562705
## 8 2014 347559680
## 9 2015 337170197
## 10 2016 324901419
## 11 2017 351726063
## 12 2018 387442789
## Visualisation
export.rows %>%
filter(Activity == "Total") %>%
group_by(year) %>%
summarise(Total = sum(USD)) %>%
ggplot(aes(x = year, y = Total)) +
geom_line() +
geom_point() 
## Per state
export.rows %>%
filter(Activity == "Total") %>%
group_by(year, state) %>%
summarise(Total = sum(USD)) %>%
ggplot(aes(x = year, y = Total)) +
geom_line(aes(colour = abbreviate(state, 6)))+
geom_point(aes(colour = abbreviate(state, 6)))
We also would like to see if every year was the same state and/or the same activity making the biggest money by exportation, or if this changed with the time. Unfortunately, our figure with the total exportation per year grouped by state is quite messy and difficult to appreciate due to the big number of states. Thus we need a different approach to that
## Main state per year
export.cols %>%
group_by(year) %>%
filter(Total == max(Total)) %>%
select(year, state, Total) %>%
arrange(year)## # A tibble: 12 x 3
## # Groups: year [12]
## year state Total
## <dbl> <chr> <dbl>
## 1 2007 Baja California 31858677
## 2 2008 Baja California 32988913
## 3 2009 Baja California 26741828
## 4 2010 Chihuahua 34633881
## 5 2011 Chihuahua 38446014
## 6 2012 Chihuahua 41764861
## 7 2013 Chihuahua 43770979
## 8 2014 Chihuahua 45594451
## 9 2015 Chihuahua 40302945
## 10 2016 Chihuahua 43342067
## 11 2017 Chihuahua 46491551
## 12 2018 Chihuahua 51944047
## Activity
export.rows %>%
filter(Activity != "Total") %>%
group_by(year) %>%
filter(USD == max(USD)) %>%
arrange(year) %>%
select(Activity, state, year)## # A tibble: 12 x 3
## # Groups: year [12]
## Activity state year
## <chr> <chr> <dbl>
## 1 Petroleum Campeche 2007
## 2 Petroleum Campeche 2008
## 3 Petroleum Campeche 2009
## 4 Petroleum Campeche 2010
## 5 Petroleum Campeche 2011
## 6 Petroleum Campeche 2012
## 7 Petroleum Campeche 2013
## 8 Transport equipment Coahuila de Zaragoza 2014
## 9 Transport equipment Coahuila de Zaragoza 2015
## 10 Transport equipment Coahuila de Zaragoza 2016
## 11 Transport equipment Coahuila de Zaragoza 2017
## 12 Transport equipment Coahuila de Zaragoza 2018
Our results are quite interesting: the main state until 2009 is Baja California, and then it changes to Chihuahua, but when we look at the main activity exporting USD, the main states are Campeche until 2013 and then it changes to Coahuila. It seems that there is not a big difference between activities in the total USD produced by exportation, and that different combinations could lead to a bigger production in a single state. A good example of this is Campeche, who is the leader in the main exportation activity from 2007 to 2013, however it goes to the 6Th position when it comes to the total exportations overall.
It could be interesting to look at the main states exporting, as well as the leading activities.
We can create specific functions using our data frames to look directly at a given state or activity in order to make our analysis easier and find specific patterns.
## FUNCTION TO CHOOSE MAIN ACTIVITY PER STATE
plot_state <- function(estado, USD_min = 5000000){
export.rows %>%
filter(Activity != "Total") %>%
group_by(state, Activity) %>%
summarise(Total = sum(USD)) %>%
filter(state == estado &
Total >= USD_min) %>%
ggplot() +
geom_bar(aes(y = Total,
x = reorder(Activity, Total, FUN = abs),
fill = Total),
stat = 'identity') +
coord_flip() +
labs(title = estado,
y = "Total USD", x = NULL) +
theme(legend.position="none")
}
## FUNCTION TO CHOOSE MAIN STATES IN A GIVEN ACTIVITY
plot_activity <- function(activity, USD_min = 5000000){
export.cols %>%
select(state, year, activity) %>%
group_by(state) %>%
summarise(Total = sum(!!sym(activity))) %>%
filter(Total >= USD_min) %>%
ggplot() +
geom_bar(aes(y = Total,
x = reorder(state,
Total, FUN = abs),
fill = Total),
stat = 'identity') +
coord_flip() +
labs(title = activity,
y = "Total USD", x = NULL) +
theme(legend.position="none")
}Now we can see particular states and activities. We can start with the main 6 states to see why are there differences in the main activities and main states exporting.
plot_grid(
plot_state("Chihuahua") ,
plot_state("Baja California"),
plot_state("Coahuila de Zaragoza"),
plot_state("Nuevo León"),
plot_state("Tamaulipas"),
plot_state("Campeche", USD_min = 10000),
ncol = 2)
Despite of the fact that different states get their main exportation from different products, Electronics and Transport equipment are generally in the first 2 places. Other activities such as Machinery and Metal products also remain in the top exportations among the first 5 states. In other words, there is a trend among the 5 states with the biggest exportation income with the kind of activities that are responsible for it. However, this trend is totally broken in the case of Campeche, which is on the 6Th place and main exportation activity is from Petroleum, which no other state in the top 5 shares. Nevertheless, petroleum or, Extraction of petroleum and oil as it translates, is in the third position of main activities responsible for exportations.
export.rows %>%
filter(Activity == "Electronics" |
Activity == "Transport equipment" |
Activity == "Petroleum") %>%
group_by(year, Activity) %>%
summarise(`Total per activity` = sum(USD)) %>%
ggplot(aes(x = year, y = `Total per activity`)) +
geom_line(aes(colour = Activity)) +
geom_point(aes(colour = Activity))
It looks
like the extraction of petroleum and oil has, at least since 2007, been
exporting less than the production of electronics and transport
equipment. And another interesting trend is observed in the last plot:
since 2009, the production of transport equipment had a considerable
increase in its exportations.
Coming back to petroleum extraction, we can have a look at the top
petroleum exporters using our plot_activity function
plot_activity("Petroleum")
Only 4 states are exporting more than 5,000 million USD from petroleum.
plot_grid(
plot_state("Campeche", USD_min = 1000000),
plot_state("Tabasco", USD_min = 1000000),
plot_state("Veracruz de Ignacio de la Llave"),
plot_state("Chiapas", USD_min = 1000000))
It seems that the economy of Tabasco, Campeche and Chiapas depends in a big extent on the extraction of petroleum, unlike Veracruz which has other strong activities such as Chemistry, Metalurgical and Food industries.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent