Introduction to exvatools"

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>"
)

Half of today's international trade is composed of intermediate goods, so statistics of gross exports do not reflect the complexity of globalization. Trade flows should be expressed in terms of value added, taking advantage of modern international input-output tables, which show how different sectors in different countries interact with each other. The problem is that handling big input-output tables is complicated and many indicators are not available. This is where exvatools proves useful. Its purpose is double:

International input-output analysis tools

Installing exvatools

Installing exvatools from the CRAN repository follows the usual procedure:

install.packages("exvatools")

To make exvatools available:

library(exvatools)

Using data

exvatools produces basic input-output tables from three types of data:

  1. Raw data of standard international input-output databases

Raw data can be directly downloaded from the web pages of their respective institutions. Four sources are currently supported:

The advantage of these databases is threefold: they are widely used in the economic literature, they are quite rich in terms of countries and sectors, and they are directly downloadable from the web page of their supporting institutions, normally as zipped files containing comma-delimited files (.csv), Excel files (.xlsx) or R data files (.rData).

For instance, if we want to use exvatools with the latest edition of the OECD ICIO tables (2023, with data up to 2020), we must first download to our computer the source file "2016-2020.zip" (95 MB) from the ICIO web page. Then we will use the command make_wio(), specifying the edition, the year and the folder where the source zip file is saved (just the directory). For instance, if we had downloaded it in C:\Users\Username\Documents\R and we wanted the year 2020, we would just need to type:

wio <- make_wio("icio2023", year = 2020, 
                src_dir = "C:/Users/Username/Documents/R")

and exvatools will take care of the rest: it will extract the .csv files from the zip file and produce the basic input-output matrices.

  1. Test datasets included in the package

If we just want to check the features of exvatools, there is no need to download any data. The package includes two sets of fictitious data, one replicating an ICIO-type database (wiotype = "iciotest", with Mexico and China disaggregated) and another one replicating a WIOD-type database (wiotype = "wiodtest").

wio <- make_wio("iciotest")
  1. Custom data

Alternatively, exvatools can use custom data to create basic input-output matrices. In this case, we just need as input a numeric matrix (or data frame) with the intermediate inputs Z and the final demand Yfd, and a vector with the names of the countries (names of sectors and final demand components are optional). The command make_custom_wio() will be used in this case.

wio <- make_custom_wio(df, g_names = c("C01", "C02", "C03"))

For didactic purposes, we will use here ICIO-type fictitious data made with make_wio("iciotest").

The contents of the created wio object can be checked with summary():

summary(wio)

Operating with input-output matrices

exvatools provides multiple commands that make operating with international input-output tables extremely easy: thus, we can multiply a diagonal matrix by an ordinary one with dmult(), an ordinary by a diagonal with multd(), or make a block-by-block Hadamard product of matrices with hmult().

We can also easily obtain a block diagonal matrix with bkd(), a block off-diagonal matrix with bkoffd(), or a diagonal matrix with the sums of all columns with diagcs(),

Additionally, as we are always operating with named rows and columns with names of countries and sectors, we have improved commands to consolidate matrices and provide names, like rsums() to sum rows, csums() to sum columns, sumnrow() to sum every nth row of a matrix, sumncol() to sum every nth column, sumgrows() to sum groups of rows of a particular size, sumgcols() to do the same with columns, etc.

For instance, let us check that the production X is equivalent to the product of the global Leontief inverse matrix B and the final demand Y:

BY <- wio$B %*% wio$Y

We can sum the rows and check that it coincides with the production vector:

BY <- rsums(BY, "BY")
print(cbind(head(BY, 10), head(wio$X, 10)))

In OECD ICIO tables two big industrial countries, China and Mexico, are disaggregated into two. Calculations must be done with disaggregated data, but countries must be later consolidated (e.g., CN1 and CN2 must be converted into CHN). This can be done with the command meld().

For instance, to calculate the value added absorbed abroad (VAX) we need to multiply the value added coefficients matrix V-hat (represented here with W) by the global inverse matrix B by the final demand matrix Y, and then exclude the value added absorbed domestically. This can be easily done with a few commands.

# To calculate all value added induced by demand:
VBY <- dmult(wio$W, wio$B) %*% wio$Y
VBY

We can see that rows for Mexico and China are disaggregated. We can meld them with meld():

VBY <- meld(VBY)
VBY

We just want the value added absorbed abroad (sometimes referred to as value added exported). For that we need the block off-diagonal matrix of VBY, that we can produce with bkoffd():

bkoffd(VBY)

Value added in exports

The model we have used allows us to express value added induced by final demand, but we can also study how gross exports induce value added, not only from final exports but also from exports of intermediates. In this case, we need to consider the effect of the multiple times that intermediate products cross international borders, to avoid double counting.

This is why several methods have appeared to calculate full decomposition of value added in exports, distinguishing what part of value added is pure value added and which one is double counting, and also, within the pure double counting, what part is really exported and what part eventually returns back to the exporting country ('false' exports).

exvatools can then produce key value added indicators (like bilateral value added exports or VAX) that are not currently available in public databases.

Alternative decompositions of value added in exports

There are several methodologies in the economic literature, and exvatools includes the most complete ones: Koopman et al. (2014), Wang et al. (2013), Borin and Mancini (2023) and Miroudot and Ye (2021).

For instance, to create create a full decomposition of value added in exports of Spain using the method of Borin and Mancini (2023), using a source-based approach, we would type::

exvabm <- make_exvadec(wio, exporter = "ESP", method = "bm_src")

The advantage is that, once we have obtained a decomposition, we can play with the results in terms of sectors and countries of destination just using the command get_exvadec_bkdown(). For instance, to select the value added in Spanish exports of services (including construction) to the United States, we just have to type:

get_exvadec_bkdown(exvabm, exporter = "ESP", 
                   sector = "SRVWC", importer = "USA")

An alternative (although with some methodologically limitations) decomposition would be that of Wang et al. (2013). In this case, instead of the normal decomposition, we will use the "terms" output. (that shows the 16 terms that compose the value added in exports):

exvawwz <- make_exvadec(wio, exporter = "all", method = "wwz",
                        output = "terms", quiet = TRUE)

Note that here we have selected export = all, that produces decompositions for all countries (not only a specific one), and we have also used the option quiet = TRUE, that produces a silent output.

We can check any exporting country, any sector, and any destination country. For instance, we can produce the decomposition of the value added in US exports to China for the manufacturing sector:

get_exvadec_bkdown(exvawwz, exporter = "USA", 
                sector = "MANUF", importer = "CHN")

Direction of value added in exports: origin and destination

We have seen that the foreign content in Spanish exports amounts to USD 2501.78 million. Where does it come from? If we do not need a detailed breakdown of the value added, but we are interested in knowing the specific geographical and sector origin of the value added content in exports, we can use the command make_exvadir():

exvadir <- make_exvadir(wio, exporter = "ESP", va_type = "FC", 
                        flow_type = "EXGR")

Please note that the exvadir object that we have obtained is different from the exvadec object, in the sense that 'exporters' in an exvadir object are the different countries and sectors of origin of the value added included in the exports of the country specified with make_exvadir()(in this case, Spain). We can better understand this by typing summary(exvadir):

summary(exvadir)

Get specific data with custom groups

We can use get_data() to summarize the foreign content of Spanish exports, with a breakdown between EU and Non-EU origin (specifying a few countries) and also distinguishing between goods (with utilities) and services. We can also break down the destination of those exports between EU and non-EU:

get_data(exvadir, exporter = c("WLD", "EU27", "FRA",
                               "NONEU27", "USA"),
         sector = c("TOTAL", "GOODSWU", "SRVWC"),
         importer = c("WLD", "EU27", "NONEU27"))

On the other hand, the flexibility of get_data() allows for the creation of custom-made groups of countries and/or sectors.

For instance, let's create a group of countries called LATAM, with Spain and Mexico. We would just have to define the variable in the current environment.

LATAM <- c("ESP", "MEX")

And now we can use it as a normal variable, just introducing it as "LATAM" (with double quotes). We will use the wwz decomposition and extract the domestic value added in intermediates (DVA_INT) from LATAM to USA. Note that, if we use custom groups, we need to select the option custom = TRUE in get_data().

get_data(exvawwz, "DVA_INT", exporter = "LATAM", 
         sector = c("TOTAL", "MANUF", "SRVWC"), 
         importer = "USA", custom = TRUE)

Let us now see an example of the exception marker "x", that allows to define exceptions for countries and for sectors. We can, for instance, calculate the NAFTA exports, both intra-regional and extra-regional, employing just two sectors: non-services and services.

get_data(exvawwz, "EXGR", exporter = "NAFTA", 
         sector = c("TOTAL", "TOTALxSRVWC", "SRVWC"), 
         importer = c("WLD", "NAFTA", "WLDxNAFTA"), custom = TRUE)

Other useful commands

The flexibility of the commands make_exvadir() and get_data() allows for the creation of several ready-made commands in exvatools. One is get_va_exgr(), to get a detailed sector and geographical origin and destination of value added, i.e., how the inputs of specific sectors in specific countries contribute to the value added in exports of a particular sector of a particular country. For instance, if we want to check the value added in US services incorporated in Spanish exports of manufactures, we just have to type:

get_va_exgr(wio,geo_orig = "USA", sec_orig = "SRVWC",
            geo_export = "ESP", sec_export = "MANUF")

Sometimes we are not only interested in the origin, but also in the country of final absorption. For that we have get_va_exgry(). For instance, if we want to know what part of the US value added incorporated in China's exports of manufactures ends up absorbed back in the US, we can type:

get_va_exgry(wio, geo_orig = "USA", geo_export = "CHN",
             sec_export = "MANUF", geo_fd = "USA")

At the beginning we manually calculated the value added induced by final demand. There is also a specific command for that in exvatools called get_va_fd(). This allows, for instance, the calculation of the Chinese total value added (or GDP) induced by US final demand for manufactures:

get_va_fd(wio, geo_orig = "CHN", sec_orig = "TOTAL",
          geo_fd = "USA", sec_fd = "MANUF")

TiVA useful indicators

Finally, if we want to get a list of common trade indicators (exports, imports, value added, production) similar to those of the TiVA database, we could just use make_exvadec() with the method "oecd" and output = "tiva".

exvativa <- make_exvadec(wio, exporter = "all", method = "oecd",
                         output = "tiva", quiet = TRUE)

And then get the decomposition for Spain:

get_exvadec_bkdown(exvativa, exporter = "ESP")

Available sectors and countries in each input-output edition

To check the information about sectors, it suffices to print info_sec():

info_sec("iciotest")

To check the information about available countries, the command is info_geo():

info_geo("iciotest")

These commands do not require to have a wio in the environment, so we can just check what sectors are available in the OECD's ICIO tables, 2023 edition.

info_sec("icio2023")

Additionally, the commands get_geo_codes() and get_sec_codes provide details about the components of the different groups. These commands are also directly applicable for any available input-output table. For instance, for "wiod2016" we would have the following components of EU27:

get_geo_codes("EU27", wiotype = "wiod2016")

And for "icio2023" we have the following components of the business services sector:

get_sec_codes("BIZSV", wiotype = "icio2023")


Try the exvatools package in your browser

Any scripts or data that you put into this service are public.

exvatools documentation built on May 29, 2024, 6:46 a.m.