Balassa index for more years about economiccomplexity HOT 1 CLOSED

# Packages ----

# this is for the example itself
library(dplyr)
library(purrr)
library(tidyr)
library(economiccomplexity)

# this is for the data for the example
library(tradestatistics)

# Data ----

# exports from Chile to the World, by product [Y]ear, [R]eporter, product [C]ode
d <- ots_create_tidy_data(years = 2016:2018, reporters = "all", table = "yrc")

# unique years
unique(d$year)

# we can't obtain Balassa index when d has more than 1 year, 
# so we use purrr + dplyr to filter the data and create a 
# new dataset with the correct index

# Filter by year + compute Balassa Index by year ----

# create a new function balassa_index_by_year()
balassa_index_by_year <- function(d) {
  # unique years 
  years <- unique(d$year)
  
  # filter one year at a time, compute the index, repeat and paste
  # all the pieces
  map_df(
    years,
    function(t) {
      d2 <- d %>% filter(year == t)
      
      d2 <- balassa_index(
        data = d2,
        country = "reporter_iso",
        product = "product_code",
        value = "export_value_usd",
        discrete = FALSE
      )
      
      # to use map df, we need d2 to be a data.frame/tibble, not a matrix
      # otherwise, tweak the example and use map(), which returns a list
      # of matrices in this case
      d2 <- as.data.frame(as.matrix(d2))
      d2 <- tibble::rownames_to_column(d2, "country_iso")
      d2 <- d2 %>% 
        gather(product_code, balassa_index, -country_iso) %>% 
        mutate(year = t) %>% 
        select(year, country_iso, everything()) %>% 
        as_tibble()
      
      return(d2)
    }
  )
}

d2 <- balassa_index_by_year(d)

# Final comments ----

# table d already includes a column export_which is different from the obtained now
# why? because tradestatatiscs database used the weighted average of 3 years of
# exports to obtain a more stable Balassa Index