In paul4forest/tradeflows: Process forest products trade flow data from COMTRADE and COMEXT

library(tradeflows)
library(knitr)
opts_chunk$set(message=FALSE, warning = FALSE)
opts_knit$set(root.dir="../..") # file paths are relative to the root of the project directory
library(dplyr)
library(ggplot2)
library(tidyr)

2005 China Malaysia 440349 trade discrepancy issue

chmy <- readdbtbl("validated_flow_yearly") %>%
    filter(reporter == "China" & partner == "Myanmar") %>%
    collect()
mych <-  readdbtbl("validated_flow_yearly") %>%
    filter(reporter == "Myanmar" & partner == "China") %>%
    collect()

chmyr <- readdbtbl("raw_flow_yearly") %>%
    filter(reporter == "China" & partner == "Myanmar") %>%
    collect()
mychr <-  readdbtbl("raw_flow_yearly") %>%
    filter(reporter == "Myanmar" & partner == "China") %>%
    collect()

440724

# Find which product has a high volume issue in the validated data
cam <- readdbtbl("validated_flow_yearly") %>%
     filter((reporter == "Cameroon" | partner == "Cameroon") & 
                unit == "Volume in cubic meters") %>%
     arrange(desc(quantity)) %>%
     head(200)

#' Display a sorted table for the purpose of this analysis
#' @param dtf a data frame of trade flows
littletable <- function(dtf,n=2){
    dtf %>% 
        select(reporter, partner, year, flow, tradevalue, quantity, weight, price) %>%
        arrange(desc(quantity)) %>%
        head(n) %>%
        kable()
}

Raw data shows Kuwait, Botswana and South Africa import extreme quantities

sawn24r <- readdbproduct(440724, "raw_flow_yearly") %>% 
    filterworldeu28() %>%
    addconversionfactorandprice() %>%
    addregion()
sawn24r %>% littletable()
sawn24r %>% filter(regionreporter == "Africa") %>% littletable()

Validated data shows Cameroon Netherlands with extreme volume and price issue

Kuwait, Botswana and South Africa trade flows were corrected in the validated data, but not the export flows between Cameroon and Netherlands remained erroneous

sawn24v <- readdbproduct(440724, "validated_flow_yearly") %>%
    addconversionfactorandprice() %>%
    addregion()
sawn24v %>% littletable()

Archive table to compare once the algorithm has been modified

|reporter |partner | year|flow | tradevalue| quantity| weight| price| |:--------|:-----------|----:|:------|----------:|--------:|--------:|-------:| |Cameroon |Netherlands | 2007|Export | 30350149| 9547163| 52119188| 3.17897| |Cameroon |Netherlands | 2008|Export | 37635852| 5136247| 57231874| 7.32750|

Price extraction issues

Analyse 2007 exports from Africa. Only 18 from 120 entries have a price. All other values are not available (NA). NA values should not be taken into account in the calculation

sawn24r2007afr <- sawn24r %>% filter(regionreporter == "Africa" & year == 2007)
sawn24r2007afr %>% 
    mutate(value = ifelse(is.na(price),"NA","value"),
           value = ifelse(is.infinite(price),"Inf",value)) %>%
    group_by(value) %>% summarise(nrow = n()) %>% kable()

# Change in median price when removing NA values, same as option na.rm
median(sawn24r2007afr$price)
median(sawn24r2007afr$price, na.rm = TRUE)
median(sawn24r2007afr$price[!is.na(sawn24r2007afr$price)])

# Change in median price when removing NA and Inf values
median(sawn24r2007afr$price[!is.na(sawn24r2007afr$price) & !is.infinite(sawn24r2007afr$price)])

price <-  sawn24r %>%
    extractprices()

ggplot(price, aes(x = year, y = medianprice, ymin = lowerprice, ymax = upperprice)) +
    geom_point() + geom_errorbar() +
    facet_grid(unit + flow ~ regionreporter) +
    ggtitle("Regional median prices and price bounds")

ggplot(price, aes(x = year, y = medianprice, color = regionreporter)) +
    geom_line() + facet_grid(flow ~ .) +
    ggtitle("Global median prices")

priceglobal <- sawn24r %>%
    extractprices(grouping = c("flow", "year", "unit"))

ggplot(priceglobal, aes(x = year, y = medianprice, ymin = lowerprice, ymax = upperprice)) +
    geom_point() + geom_errorbar() +
    facet_grid(unit + flow ~ . ) + 
    ggtitle("Global median prices and price bounds")

ggplot(priceglobal, aes(x = year, y = medianprice, color = flow)) +
    geom_line() + 
    ggtitle("Global median prices")

# Try smooth
plot(priceglobal$medianprice[priceglobal$flow == "Import"])
plot(smooth(priceglobal$medianprice[priceglobal$flow == "Import"]))
plot(priceglobal$medianprice[priceglobal$flow == "Export"])
plot(smooth(priceglobal$medianprice[priceglobal$flow == "Export"]))
priceglobal <- priceglobal %>% 
    group_by(flow) %>%
    mutate(mediansmooth = as.numeric(smooth(medianprice)))
ggplot(priceglobal, aes(x = year, y = mediansmooth, color = flow)) +
    geom_line() + 
    ggtitle("Global median prices")

Correct the "shave price" algorithm

Place a breakpoint in the price calculation to see the price data

sawn24c <- sawn24r %>% clean()
# Number of rows and price distribution
# This can be used to export prices and conversion factors to an excel file
# cleandb2excel(440724)

Conversion factor issues

Infinite conversion factor should not be taken into account in the calculation of the mean conversion factor.

conv <- sawn24r %>% 
    extractconversionfactors()
sum(conv$medianconversion)

ggplot(conv, aes(x = year, y = medianconversion)) +
    geom_point() + 
    facet_grid(unit + flow ~ regionreporter )

Comtrade flags in the raw data

nrowflag <- readdbtbl("raw_flow_yearly") %>%
    select(flag, unit) %>%
    group_by(flag, unit) %>%
    summarise(n = n()) %>% collect()

Include comtrade estimates in the price calculation

Functions that extract prices and conversion factors have flag that can disable comtrade estimates if required. This is not activated by default.

paul4forest/tradeflows documentation built on Oct. 8, 2019, 10:35 a.m.