Fragmentation in

Global Trade
Accounting for Commodities

Marijn A. Bolhuis, Jiaqian Chen, and Benjamin Kett

WP/23/73

IMF Working Papers describe research in

progress by the author(s) and are published to
elicit comments and to encourage debate.
The views expressed in IMF Working Papers are
those of the author(s) and do not necessarily
represent the views of the IMF, its Executive Board,
or IMF management.

2023
MAR
 © 2023 International Monetary Fund WP/23/73

IMF Working Paper

Strategy, Policy, & Review Department

Fragmentation in Global Trade: Accounting for Commodities*

Prepared by Marijn A. Bolhuis, Jiaqian Chen, and Benjamin Kett

Authorized for distribution by Anna Ilyina

March 2023

IMF Working Papers describe research in progress by the author(s) and are published to elicit
comments and to encourage debate. The views expressed in IMF Working Papers are those of the
author(s) and do not necessarily represent the views of the IMF, its Executive Board, or IMF management.

ABSTRACT:
We construct a new database which covers production and trade in 136 primary commodities and 24
manufacturing and service sectors for 145 countries. Using this new more granular data, we estimate spillover
effects from plausible trade fragmentation scenarios in a new multi-country, multi-sector, general-equilibrium
model that accounts for unique demand and supply characteristics of commodities. The results show
fragmentation-induced output losses can be sizable, especially for Low-Income-Countries, although the
magnitudes vary according to the particular scenarios and modelling assumptions. Our work demonstrates that
not accounting for granular commodity production and trade linkages leads to underestimation of the output
losses associated with trade fragmentation.

RECOMMENDED CITATION: Bolhuis A. Marijn, Jiaqian Chen and Benjamin Kett (2023): Fragmentation in
Global Trade: Accounting for Commodities. IMF Working Paper, No. WP 23/73

JEL Classification Numbers: F11, F12, F14, F15, F17, F41, F42, F43, Q17, Q27, Q37, Q43

Commodities; international trade; sanctions; spillovers;

Keywords:
fragmentation

Author’s E-Mail Address: mbolhuis@[Link], jchen@[Link], bkett@[Link]

* The author(s) would like to thank Shekhar Aiyar, Jorge Alvarez, Chikako Baba, Mehdi Benatiya Andaloussi, Wenjie Chen, Stephan
Danninger, Romain Duval, Christian Ebeke, Luc Eyraud, Pierre-Olivier Gourinchas, Nikolay Gueorguiev, Anna Ilyina, Petya Koeva
Brooks, Ting Lan, Chiara Maggi, Papa N’Diaye, Martin Stuermer, Petia Topalova, Daria Zakharova, and Robert Zymek.
IMF WORKING PAPERS Fragmentation in Global Trade

WORKING PAPERS

Fragmentation in Global Trade

Accounting for Commodities

Prepared by Marijn A. Bolhuis, Jiaqian Chen, Benjamin Kett

INTERNATIONAL MONETARY FUND 3

IMF WORKING PAPERS Fragmentation in Global Trade

Contents
1. Introduction ............................................................................................................................................ 5

2. Data ......................................................................................................................................................... 8
Production and trade data .............................................................................................................................. 8
Trade elasticities (short vs. long run) ............................................................................................................. 9
Demand elasticities ...................................................................................................................................... 10

3. Modelling Framework .......................................................................................................................... 11

Model ............................................................................................................................................................ 11
Assessing the costs of fragmentation........................................................................................................... 12

4. Geo-Economic Fragmentation Scenarios ......................................................................................... 14

5. Results .................................................................................................................................................. 16
Main Results................................................................................................................................................. 16
Robustness checks ...................................................................................................................................... 20
Trade Elasticities ................................................................................................................................... 20
Inter-bloc aggregate imbalances ........................................................................................................... 21

6. Conclusion ........................................................................................................................................... 24

References ......................................................................................................................................................... 25

Annex I. Data ..................................................................................................................................................... 28

Commodity list .............................................................................................................................................. 28
Trade Elasticities .......................................................................................................................................... 29

Annex II. Modelling framework details ............................................................................................................ 30

Full Model ..................................................................................................................................................... 30
Equilibrium using exact hat algebra. ............................................................................................................ 34

INTERNATIONAL MONETARY FUND 4

IMF WORKING PAPERS Fragmentation in Global Trade

1. Introduction
Following several decades of a steady increase in global economic integration, globalization has stalled and
may be on the brink of a reversal. The shallow and uneven economic recovery from the Global Financial Crisis
(GFC) coincided with a growing number of military conflicts around the world, a deepening skepticism about the
benefits of globalization, and a growing populism and protectionism (e.g., Brexit, trade war between the United
States and China). The COVID-19 pandemic has further tested international relations. The war in Ukraine
served to split countries along geopolitical lines, further increasing uncertainty over the direction of
globalization. Aiyar and others (2023) documents these developments and coins the term “geoeconomic
fragmentation” (GEF) to describe a policy-driven reversal of global economic integration often guided by
strategic considerations.

Motivated by the rising specter of GEF, a growing number of studies have attempted to gauge the potential
economic effects of possible fragmentation scenarios (Aiyar and others, 2023). This paper aims to quantify the
economic costs of fragmentation from an international trade perspective, with a particular focus on production
and trade of commodities. More specifically, we examine how various fragmentation scenarios affect output in
different country groups by applying a novel multi-country multi-sector model with input-output linkages to a
newly developed dataset that accounts for granular production and trade in commodities.

To effectively account for spillovers from trade fragmentation, we construct a new dataset that covers a
granular level of trade and production in commodities following Fally and Sayre (2019, FS hereafter). FS use
detailed data on production and trade from various sources to construct a rich dataset covering trade,
production, and prices across a range of commodities. Our dataset extends their work along several
dimensions: (i) we update the data to 2019, the most recent pre-pandemic year, from 2016 in FS; (ii) we modify
the list of commodities in order to reflect the most “upstream” products, which is important to ensure the
characteristics of commodities (such as geographical concentration) are appropriately captured; and (iii) we
reconcile our dataset with an otherwise standard input-output (IO) matrix such that the individual commodities
sum up to an aggregate sector in the Eora26 database (Lenzen et al., 2012, 2013). Overall, our database
contains 136 primary commodities along with 24 manufacturing and service sectors for 145 countries.

Exploiting the detail of this dataset requires an adapted model which incorporates commodities. While our
model has similar building blocks, we depart from FS in several ways. Most importantly, we consider sectoral
input-output linkages that allow for feedback loops in production. Intermediate inputs are shown to be important
conduits for transmitting shocks across countries (e.g., Auer et al., 2019; Boehm et al., 2019) and imply
substantially larger losses from increasing trade barriers (e.g., Caliendo and Parro, 2015). Moreover, to reduce
the data burden we follow Cunat and Zymek (2022), making use of proportionality assumptions in the use of
intermediate inputs.

We start from an otherwise standard quantitative multi-country multi-sector trade model and distinguish
between two types of goods, commodities and non-commodities. The production of non-commodities uses
labor, commodities, and non-commodity intermediate inputs, while the production of commodities relies on
labor and non-commodity intermediate inputs. Non-commodities are also consumed as final goods. We
generate a low price elasticity of demand for commodities by introducing a low elasticity of substitution between
commodities and other inputs in the production of non-commodities.

INTERNATIONAL MONETARY FUND 5

IMF WORKING PAPERS Fragmentation in Global Trade

We then use the model to approximate the impact of trade fragmentation on domestic prices and output. We
first show that, up to a first-order approximation, fragmentation has a larger impact on prices in countries and
sectors that lose access to a larger share of their pre-fragmentation supply. This impact is particularly large in
sectors where the elasticity of substitution between foreign and domestic products is lower. This ‘trade
elasticity’ is a crucial parameter measuring the response of trade to changes in trade costs which we discuss in
more detail in the rest of the paper. We also show that the overall effect of fragmentation on a country’s real
GDP can be decomposed into the contributions of (i) the direct effect of import prices on final goods, (ii)
amplification through input-output linkages, and (iii) the effect on prices of commodities.

We calibrate the key model parameters based on the latest existing literature. Our demand elasticities for
commodity sectors are sourced from FS, which conducts a meta-analysis of the literature. The trade elasticities
are calibrated based on recent work by Fontagné et al. (2022) and provide conservative results while being well
in line with other estimates in the literature (see Data section). Notably, the two principal commodity subsectors
have the lowest trade elasticities across all sectors, at 3.4 for Mining and Quarrying, and 2.9 for Agriculture
(compared to an average across sectors of 6). Building on the most recent work on estimating trade
elasticities, we also consider short-run trade elasticities which, due to adjustments costs, are shown by Boehm
et al. (forthcoming) to be significantly lower than long-run elasticities.

Similar to FS, we find that properly accounting for trade and production in commodities (i.e., using sectoral data
disaggregated into product level commodities) substantially increases the adverse economic impact of trade
fragmentation compared with models which implicitly assume perfect substitutes among commodities.
Comparing the baseline equilibrium (2019 global trade barriers) to global autarky, we find that the output losses
more than double for Low-Income Countries (LICs), who are heavily dependent on trade in commodities, while
for Advanced Economies (AEs) and Emerging Market economies (EMs) the welfare losses increase by 4 and
25 percent respectively. Intuitively, trade barriers are much more costly for products which can only be sourced
from a relatively small number of countries and for which demand is relatively inelastic.

The model is used to explore several hypothetical scenarios that illustrate the cost of a more fragmentated
global trade network. We make a distinction between two scenarios. In the ‘mild’ fragmentation scenario
(‘Strategic Decoupling’) there is no trade between the US-EU and Russia, and no trade in high-tech sectors
between the US-EU and China, but the rest of the world (RoW) are free to trade with both groups. In the
‘severe’ fragmentation scenario (‘Geo-economic Fragmentation’), there is no trade between EU-US and
Russia- China and the RoW joins one of the two groups depending on the strength of the trade link with either
the US or China, resulting in zero trade with the other group.

Global output losses are estimated to fall by between 0.3 percent and 2.3 percent in the long run depending on
the fragmentation scenario. Moreover, the impacts are heterogenous across income groups. Specifically, in the
mild fragmentation scenario, LICs benefit from trade diversion as the trade barriers they face remain
unchanged relative to the baseline; however, in the more severe fragmentation scenario, their output drops by
4.3 percent over the long run. These results underscore the vulnerability of LICs to trade barriers and the risks
of forcing them to choose groups. It is, nonetheless, important to note that there is a wide range of potential
estimates of the size of fragmentation-induced losses, depending on the modeling assumptions. For example,
calibrating the model with different estimated trade elasticities in the literature, we find the global GDP loss
ranges between 1.9 and 7.0 percent in the more severe fragmentation scenario. Moreover, these simulations
do not reflect the full effects of global economic fragmentation, as some of the important channels which would
imply larger economic losses are not captured. Another set of scenarios reveals the sensitivity of countries to

INTERNATIONAL MONETARY FUND 6

IMF WORKING PAPERS Fragmentation in Global Trade

trade restrictions in different product groups. In particular, we find that AEs and EMEs are most vulnerable to
disruption in trade in energy and high-tech manufacturing sectors, whereas LICs see the largest output loss
following barriers to trade in agricultural goods.

The fast-growing literature on the costs of sanctions and global fragmentation has generated a wide array of
quantitative estimates, reflecting the consideration of different channels as well as different assumed
fragmentation scenarios. IMF (2022) investigates the effect of eliminating trade in the aggregated high-tech and
energy sectors across rival blocs which are determined based on the vote to condemn Russia’s invasion of
Ukraine at the United Nations General Assembly (UNGA) in March 2022. The results suggest a loss of about
1.2 percent of world GDP, which increases to 1.5 percent when barriers to trade are extended to other sectors
as well. Cerdeiro et al. (2021) employ a set of structural models to examine the costs of three different layers of
fragmentation (trade, sectoral misallocation, and foreign knowledge diffusion), across a range of fragmentation
scenarios. Their estimated welfare costs range from zero (as some countries gain from trade diversion) to 8.5
percent when accounting for all three layers of fragmentation. Bekkers and Goes (2022) focus on knowledge
diffusion across countries, with the global economy divided into an Eastern bloc and a Western bloc based on
UNGA voting records. The results show a range of losses from 0.4 percent of GDP for some countries in a mild
fragmentation scenario to 12 percent for the most affected countries under full technological decoupling.
Javorcik et al. (2022) examine fragmentation through the lens of ‘friend-shoring’, finding output losses of
between 0.1 percent and 4.6 percent of GDP depending on the country and scenario. Our paper differs from
existing papers in that we focus on the international trade channel specifically, and we account for granular
trade and production of commodities, whereas previous studies assume commodities such as copper and
diamond are perfect substitutes. This underestimates the cost of fragmentation for the global economy, with a
particular impact on countries that are more exposed to commodity trade. Moreover, we calibrate our model
with the estimated trade elasticities from the latest available literature and capture both short and long run costs
of trade fragmentation.

Our paper also contributes to the literature on the effects of sanctions related to energy commodities for
European countries. Bachmann et al. (2022) find a Russian gas shut-off (i.e., a 30 percent gas supply shock)
would affect German Gross National Expenditure by -0.7 to -2.3 percent, depending on the elasticity of
substitution used. Applying a similar framework but incorporating uncertainty and second-round effects, Lan et
al. (2022) estimate that Germany’s real GDP will fall by 1.4 percent in 2022 and 2.7 percent in 2023 under the
assumption that households adjust consumption very little. Using a multi-sector, partial equilibrium model,
DiBella et al. (2022) find an average effect of -1.8 percent on EU countries’ GDP. Albrizio el al. (2022) estimate
that by allowing the substitution of Russian gas with LNG from the global market, the adverse economic
impacts are reduced by a factor or five, but with significant global spillovers. While the European impacts are a
special case of the analysis in our paper, we go further by providing global estimates of welfare impacts and
considering broader global fragmentation scenarios. We also go beyond the energy sector to consider a wide
range of commodities which are key inputs to worldwide production processes.

The remainder of the paper proceeds as follows: Section 2 describes the dataset creation and calibration of
elasticities; Section 3 presents the model; Section 4 discusses the fragmentation scenarios; Section 5 presents
the results and the robustness checks; and Section 6 concludes.

INTERNATIONAL MONETARY FUND 7

IMF WORKING PAPERS Fragmentation in Global Trade

2. Data
Production and trade data

The modelling approach used in this paper requires bilateral trade data (including self-trade) for all sectors,
along with detailed disaggregation of commodity sectors, for a wide range of countries. 1 As this data is not
readily available, in particular for self-trade, we combine several data sources (inspired by FS) to produce a
new dataset covering 2019, the latest year prior to the start of the COVID pandemic.

The starting point for the data construction is the Eora26 multi-region input-output table (MRIO), which is a
global input-output table covering 26 sectors and 190 countries. A key advantage of Eora26 is the very broad
country coverage, necessary for the study of global fragmentation, whilst maintaining a uniform and relatively
broad sectoral coverage. However, the sectoral breakdown is too aggregated to effectively take into account
the particular properties of commodities: the two main sectors covering commodities are ‘Mining and Quarrying’
and ‘Agriculture’. The objective of our dataset construction is therefore to break apart these two aggregate
sectors to provide detailed information on individual commodities.

In order to do this, we use both international trade data and production data; self-trade is then calculated as the
difference between production and total exports. The list of commodities is based on that used by FS, with a
small number of modifications to ensure the list uses the most upstream commodities possible and allows the
best possible mapping to both production and trade data. 2 We end up with 54 mining commodities and 82
agricultural commodities with the full list provided in Appendix I.

For international trade data, we use the BACI database produced by CEP II. This database provides bilateral
trade flows for 200 countries at the harmonized system (HS) 6-digit level. While based on Comtrade, BACI
incorporates various operations to improve the consistency of the data across reporters (described in the
associated working paper, Gaulier & Zignago (2010)). Data is provided in both values (USD) and quantities
(metric tons).

We integrate two datasets for production. For agricultural commodities, we use statistics from the Food and
Agricultural Organization (FAO) which include production value data for 173 (raw and processed) agricultural
products across 245 countries. 3

For mining commodities, we use the World Mineral Production dataset produced by the British Geological
Survey (BGS). 4 This dataset includes production data in quantities for more than 70 mineral commodities by
country worldwide. The data is compiled from a range of sources including government departments, national
statistical offices, company reports etc. While FAO data is provided in value units (USD), and hence is directly
compatible with both trade and Eora data, the BGS data is only provided in production quantities (with varying

1
Self-trade is defined as the value of production consumed in the producing country (i.e., not exported).
2
We develop crosswalks by hand which map products in the production data to the commodity list, as well from HS codes in the
trade data to the same commodity list. Production and trade data are therefore linked through the chosen list of commodities (see
Appendix). The crosswalks aim to ensure consistency between the HS code and production data aggregations whilst also ensuring
the most upstream definitions of commodities possible.
3
[Link]
4
[Link]

INTERNATIONAL MONETARY FUND 8

IMF WORKING PAPERS Fragmentation in Global Trade

units). There are two necessary steps to derive value data from the BGS data. First, we convert all quantity
units into a uniform unit, metric tons (as used in BACI), using standard conversion ratios. Second, we derive
exporter-commodity specific prices using unit values from the BACI database. 5 Values are then the product of
quantities and unit values. For natural gas we take trade data from IEA, given the importance of this particular
sector for the European economy and the challenges in measurement of gas exports given the usage of
pipelines and complicated financial transactions. 6

In order the ensure consistency between the production data, trade data, and Eora we employ three main
strategies. First, when the sum of trade values is smaller than the aggregate sector value in Eora, we allow for
a residual equal to the difference. This residual accounts for the fact that our commodity list is not exhaustive,
and hence acts as an ‘other’ category. Second, when the sum of trade data is larger than the figure in Eora, we
scale down exports or self-trade proportionately, leaving the residual at zero. Third, when export value is larger
than production value, we set self-trade equal to zero. This accounts for the possibility that export product
classifications may incorporate some additional value added beyond the raw commodities.

The resulting dataset contains observations in values (USD) at the exporter sector/commodity-importer level
(including self-trade when the importer is also the exporter), with 24 aggregated (manufacturing and services)
sectors, 82 agricultural commodities, and 54 mining commodities, for 145 countries in 2019.

Trade elasticities (short vs. long run)

One key parameter in the model is the trade elasticity, which measures how trade flows respond to trade
barriers (Simonovska and Waugh, 2014). Following Boehm et al. (forthcoming), we make a distinction between
long- and short-run trade elasticities in our results (discussed further below). 7

Estimates of long-run trade elasticities in the literature vary depending on the methods and data used. Our
baseline results rely on state-of-the-art estimates produced by Fontagné et al. (2022). Their paper uses
bilateral tariffs for the universe of country pairs from 2001-2016 to estimate elasticities. It employs a gravity
approach using a fixed effects strategy, which the authors argue benefits from generality, tractability, and
transparency, while being theory-consistent. We take their estimates of elasticities at the sector-level (see
Table 8 in their paper, or Annex I of this paper), for example, we use a value of -3.41 for mining commodities,
and -2.91 for agricultural commodities. These two sectors represent the lowest two values across all sectors,
and compare to a maximum of -10.56, further highlighting the particular nature of commodity-based sectors.
Since Fontagné et al. (2022) only estimate elasticities for one services sector, for the counterfactual analysis
we collapse all services trade into one sector. We also present a range of alternative trade elasticities in the
Robustness section based on alternative sources from the literature.

5
Where multiple HS codes map to a single commodity, we derive prices by dividing the total value of all 6-digit HS codes by the
total quantity. To ensure reasonable bounded prices, we replace observations greater than two standard deviations above the mean
with the median value. While unit values derived from trade data have known limitations such as wide cross-country variance, we
argue that the result remains superior to other sources such as the US Geological Survey which only contains US-specific price
data.
6
For a fuller discussion see [Link]
7
The model employed in this paper is static. However, the difference between results that use short- and long-run trade elasticities
can be interpreted as the difference between the short- and long-run impact of changes in trade costs in a dynamic quantitative
trade model with endogenous firm entry where firm face quadratic adjustment costs. We refer the reader to Boehm et al. (2022) for
details.

INTERNATIONAL MONETARY FUND 9

IMF WORKING PAPERS Fragmentation in Global Trade

Changes in trade costs result in reallocations of inputs across sectors/countries, hence modifying demand for
certain inputs, or reallocation of demand to alternative suppliers. Firms may face adjustments costs, meaning
that these reallocations do not happen immediately. Trade elasticities are typically estimated as long-run values
which are relatively high and hence underestimate the short-run welfare losses of trade costs. Boehm et al.
(forthcoming) address this concern by estimating trade elasticities at short and long horizons. They find that it
can take 7 to 10 years for the elasticity point estimates to stabilize, with elasticities after one year equaling 36
percent of the long-run value. We calibrate the ratio between our short-run and long-run trade elasticities to be
in line with this finding. 8

It is also worth noting that the methodology employed by Boehm et al. (forthcoming) leads to significantly
smaller long-run trade elasticity estimates than are typical in the literature (1.75 to 2.25). They argue this is a
result of omitted variables in typical estimates, where multilateral resistance terms “do not absorb aggregate of
product-specific bilateral taste shifters or other unobserved bilateral gravity variables” (p. 2). The authors
propose time-differencing the traditional gravity specification to resolve this concern which dramatically reduces
their estimates. While we acknowledge the importance of the paper’s methodological innovations, we choose to
use long-run estimates from Fontagné et al. (2022) to ensure conservative results while applying short-run to
long-run ratio from Boehm et al. to generate short-run results.

Demand elasticities

The demand elasticities used in our model are a direct function of the elasticities of substitution between inputs
in the associated production function (with the exact relationship discussed in the model section). Our model
employs a Constant Elasticity of Substitution (CES) aggregator for commodities and other inputs with an
elasticity of substitution equal to 0.2, taking the conservative side of the mode of the demand elasticity
estimates surveyed by FS. 9 Other sectors have a Cobb-Douglas structure which implicitly restricts the elasticity
of substitution to one.

8
In a recent paper, Andersen and Yotov (2023) propose a new reduced-form econometric approach to short- and long-run trade
elasticities that are consistent with existing theories of dynamic adjustment in trade costs. Their estimate of the short-run elasticity is
about 10 percent of the long-run value. To be conservative, we apply the estimates from Boehm et al. (forthcoming) which imply a
higher value of the short-run trade elasticity relative to its long-run value.
9
CES aggregators are standard in the trade literature, with the tractable property that elasticities of substitution between
products/inputs embedded in the aggregator do not change with quantities used.

INTERNATIONAL MONETARY FUND 10

IMF WORKING PAPERS Fragmentation in Global Trade

3. Modelling Framework
We develop a new multi-country, multi-sector, general equilibrium model that accounts for the unique demand
and supply characteristics of commodities, as well as cross-border trade of intermediate inputs. Our starting
point is the canonical input-output model of Caliendo and Parro (2015, hereafter CP). While Fally and Sayre
(2019) also develop a model which accounts for commodities, we argue that their model is limited by the one-
directional production ‘stream’ moving from upstream commodities to downstream final goods. As highlighted
by CP, accounting more fully for the input-output structure of global trade better captures the feedback loops
where the output of one sector is an input to another. These feedback properties therefore capture additional
gains from trade.

Model
We now highlight the main building blocks of the model (see Appendix II for full details).

Production. Consider N countries, indexed by n and m. There are two types of sectors: J commodity sectors
and K non-commodity sectors. Commodities are used as intermediate inputs to produce non-commodities. The
latter are consumed as final goods and used as intermediate inputs (Cunat and Zymek,, 2022). 10 In each
country and sector, there is a representative local producer. Local commodity producers use labor and
intermediate inputs for production while local non-commodity producers use labor, intermediate inputs, and
commodities for production.

International trade. In each country and sector, a representative trading firm acts as a wholesaler and
combines local products from different countries into a composite sectoral good using a CES aggregator. The
𝑘𝑘
composite sectoral good is not tradable. Trade is subject to iceberg trade costs 𝜏𝜏𝑚𝑚𝑚𝑚 .

The price index of a typical sectoral good k in country n can be written as

1
−
𝜃𝜃𝑘𝑘
𝑃𝑃𝑛𝑛𝑘𝑘 = ��(𝜏𝜏𝑚𝑚𝑚𝑚
𝑘𝑘 𝑘𝑘 )−𝜃𝜃𝑘𝑘
𝑝𝑝𝑚𝑚 � (1),
𝑚𝑚

where 𝜃𝜃𝑘𝑘 is the elasticity of substitution between local products from different countries. 𝜃𝜃𝑘𝑘 is a crucial model
parameter that does not only govern the ease with which trading firms can substitute between foreign and
domestic products, but also the sensitivity of trade flows to changes in international trade costs. To see the
latter, note that this model obeys the standard gravity equation. The natural logarithm of 𝑋𝑋𝑚𝑚𝑚𝑚 𝑘𝑘
, the value of
imports by country n on goods from country m in sector k, is given by:

𝑘𝑘 𝑘𝑘 𝑘𝑘
ln 𝑋𝑋𝑚𝑚𝑚𝑚 = −𝜃𝜃𝑘𝑘 ln 𝜏𝜏𝑚𝑚𝑚𝑚 − 𝜃𝜃𝑘𝑘 ln 𝑝𝑝𝑚𝑚 + ln 𝐷𝐷𝑛𝑛𝑘𝑘 + ln Φ𝑛𝑛𝑘𝑘 (2),

10
Assuming that the non-commodity composite good is both consumed and used as the composite intermediate input amounts to
assuming that all sectors source intermediates from other sectors with the same intensity. The overall share of expenditures on
intermediate inputs varies across sectors. This assumption, used in Cunat and Zymek (2022), relaxes the data requirement
substantially such that the empirical analysis does not require data on intermediate input use of detailed commodity sectors.

INTERNATIONAL MONETARY FUND 11

IMF WORKING PAPERS Fragmentation in Global Trade

where 𝑝𝑝𝑚𝑚
𝑘𝑘
is the price of the local product in sector k of country m, excluding trade costs. 𝐷𝐷𝑛𝑛𝑘𝑘 is total demand for
sector k in country n, and Φ𝑛𝑛𝑘𝑘 captures country n’s access to other supplies. 𝜃𝜃𝑘𝑘 is commonly referred to as the
trade elasticity, and we discuss it in detail below and in the Data section.

Inelastic demand for commodities. For each non-commodity sector, we assume a constant elasticity
production function with constant returns to scale. 𝜂𝜂 denotes the elasticity of substitution between commodity
inputs and other inputs, including labor and non-commodities intermediates. The total demand 𝐷𝐷𝑛𝑛𝑘𝑘 for
commodities by sector k in country n is given by:

𝑘𝑘 𝑃𝑃𝑛𝑛𝑜𝑜 1−𝜂𝜂 𝑘𝑘
𝐷𝐷𝑛𝑛𝑘𝑘 = β𝑛𝑛 � � 𝑌𝑌𝑛𝑛 (3),
𝑐𝑐𝑛𝑛𝑘𝑘

𝑘𝑘
where β𝑛𝑛 governs the importance of commodities in production of sector k, 𝑃𝑃𝑛𝑛𝑜𝑜 is the price of the commodity
bundle in country n, and 𝑐𝑐𝑛𝑛𝑘𝑘 and 𝑌𝑌𝑛𝑛𝑘𝑘 are the cost function and total production value of the representative firm. If
a given individual commodity accounts for a small share of total costs, and holding other input costs and
demand constant, the price elasticity of demand of a given commodity j is approximately equal to the elasticity
of substitution 𝜂𝜂:

𝑘𝑘𝑘𝑘 𝑗𝑗
𝜕𝜕 ln�𝑑𝑑𝑛𝑛 /𝑝𝑝𝑛𝑛 �
𝑗𝑗
≈ 𝜂𝜂 (4),
𝜕𝜕 ln 𝑝𝑝𝑛𝑛

𝑘𝑘𝑘𝑘 𝑗𝑗
where 𝑑𝑑𝑛𝑛 is the total demand for commodity j by sector k, and 𝑝𝑝𝑛𝑛 is the price of commodity j.

Households and income. In each country, there is a representative household that supplies labor and
consumes the final good. To account for trade imbalances, households receive and send exogenous
international transfers, such that household income equals labor income plus the trade balance.

Assessing the costs of fragmentation

We interpret fragmentation as policy-driven increases in trade barriers. As is standard in the quantitative trade
literature, we simulate the impact of changes in trade barriers using exact hat algebra (Dekle et al., 2007),
which allows us to study the impact of fragmentation on output without the need to estimate model
fundamentals such as trade costs and productivity levels. We refer the reader to the Appendix for details, but
first derive some results to highlight the role of the trade elasticity in driving the costs of fragmentation.

Impact on domestic prices. Consider two blocs of countries that erect trade barriers such that any trade
between the blocs in sector k is eliminated. The (log) change in the local price index of country n in the sector k
(commodity or non-commodity) can be approximated as:

1 1
𝑑𝑑 ln 𝑃𝑃𝑛𝑛𝑘𝑘 ≈ ln 𝑘𝑘
(5),
𝜃𝜃𝑘𝑘 1 − 𝜋𝜋𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒

where 𝜋𝜋𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒
𝑘𝑘
is the share of initial expenditures of country n on goods from countries that are in the other bloc.
In equation (5), the first-order impact of fragmentation on domestic prices can be summarized using the

INTERNATIONAL MONETARY FUND 12

IMF WORKING PAPERS Fragmentation in Global Trade

exposure to the other bloc (𝜋𝜋𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒

𝑘𝑘
) and the sector-specific trade elasticity 𝜃𝜃𝑘𝑘 . Prices increase more in countries
that lose access to a larger share of supply, which are then forced to source additional products from more
expensive suppliers. This effect is scaled by the inverse of the trade elasticity, as a lower elasticity makes it
more costly to switch suppliers.

Impact on domestic output. The first-order change in output in country n due to fragmentation can be
approximated as:

1 �������������������� 𝛼𝛼𝑛𝑛 (1 − 𝜇𝜇𝑛𝑛 ) 1 �������������������� 𝛼𝛼𝑛𝑛 𝜇𝜇𝑛𝑛 1 ���������������������

𝑗𝑗
ln�1 − 𝜋𝜋𝑘𝑘𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 � + ln�1 − 𝜋𝜋𝑘𝑘𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 � + ln �1 − 𝜋𝜋𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 � (6),
���
𝜃𝜃𝑘𝑘 1 − 𝛼𝛼𝑛𝑛 ���𝜃𝜃𝑘𝑘 1 − 𝛼𝛼𝑛𝑛 𝜃𝜃�𝚥𝚥

where 𝛼𝛼𝑛𝑛 (1 − 𝜇𝜇𝑛𝑛 ) is the country’s average expenditure share on intermediate inputs from non-commodity
sectors and 𝛼𝛼𝑛𝑛 𝜇𝜇𝑛𝑛 is the country’s average expenditure share on intermediate inputs from commodity sectors.
�������������������� ���������������������
𝑗𝑗
ln�1 − 𝜋𝜋𝑘𝑘𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 � and ln �1 − 𝜋𝜋𝑛𝑛,𝑜𝑜𝑜𝑜ℎ𝑒𝑒𝑒𝑒 � are the average (log) shares of initial expenditure of country n on non-
commodity and commodity goods from countries that are in the same bloc. 11

Equation (6) shows that, up to a first order approximation (for exposition only), all the general equilibrium
effects of fragmentation on welfare can be summarized by three terms. The first term captures the direct effect
on prices of final goods, whereas the second term captures the indirect effect through input-output linkages.
The third term then captures the effect of access to commodities. For all terms the effect is increasing in the
average share of initial expenditures on countries in the same trading bloc and decreasing in the average trade
elasticity.

Finally, it is useful to highlight the boundaries to the type of quantitative trade model that we use. First, the
model does not account for any changes in labor productivity due to capital accumulation or technological
change. Short-run disruptions to output due to fragmentation therefore do not lead to long-run losses in labor
productivity. Second, although the model allows for initial aggregate trade imbalances, they are assumed to
remain fixed and do not respond endogenously to changes in trade costs. 12

11
In all cases, these are weighted averages with expenditure shares as weights.
12
However, we do exogenously vary the initial aggregate imbalances in the Robustness section to evaluate the sensitivity of the
results.

INTERNATIONAL MONETARY FUND 13

IMF WORKING PAPERS Fragmentation in Global Trade

4. Geo-Economic Fragmentation Scenarios

Rising geopolitical tensions have increased the specter of protectionism and fragmentation. Russia’s invasion
of Ukraine was followed by far-reaching sanctions by Western countries. 13 These sanctions include measures
which target both individuals (for example freezing of assets held outside Russia, or restrictions on visas for
travel) and the Russian economy. Among those targeting the Russian economy, restrictions on imports and
exports have been central to the policy response. For example, restricted Russia exports include (among
others): crude oil (from December 2022) and refined petroleum (from February 2023), coal, steel, and gold.
Both Belarus and Iran have also faced sanctions for their perceived support of the invasion. At the time of
writing, it is unclear when the sanctions might be lifted and the extent to which the world will return to previous
trading patterns or alternatively stabilize in a new normal with higher global trade barriers.

The intensification of US-China trade tensions in 2018 led to a surge in global trade policy uncertainty and
contributed to a paralysis of the multilateral trade dispute system. The US has recently announced new
measures restricting sales to China of certain high-tech goods, software, and other technology related to
advanced computing and semiconductor manufacturing, as well as restricting activities of “US persons” that
support the development or production of certain technologies in China. 14 These recent measures—motivated
by national security considerations—increase the risk of the US-China high-tech decoupling with potentially
adverse implications for the global economy.

More broadly, data from the Global Trade Alert database shows a rising number of trade restrictions imposed
by countries, notably in high-tech sectors, likely reflecting the importance of these sectors in strategic
competition and national security (IMF, 2022). Furthermore, during the pandemic several countries imposed
export restrictions on medical goods and foodstuffs—with exports bans accounting for about 90 percent of
trade restrictions.

Against this background, this paper considers five illustrative scenarios. The first three scenarios focus on trade
restrictions in key product groups with the broad objective of illustrating which country groups are most
vulnerable to disruption in trade in certain types of products. Specifically, scenarios A, B and C assume the
introduction of barriers that prohibit trade in energy, agricultural goods and ‘high-tech’ sectors respectively. 15
Following Cerdeiro et. al. (2021), high-tech sectors are defined using the classification in OECD (2011), which
is based on sectoral R&D intensities. This methodology highlights two high-tech sectors: electronics and
machinery, and transport equipment. The last two scenarios assume the US and EU impose barriers to prohibit
trade of all goods and services from Russia and vice versa. The strategic decoupling scenario additionally
assumes barriers on trade in high-tech sectors between the US-EU and China but no changes in the trade
relations between US, EU, China, Russia, and the rest of the world (RoW). The Geo-economic fragmentation
scenario goes further by assuming barriers on all trade between the US-EU and China; at the same time,

13
Details of the EU response, for example, can be found here: [Link]
measures-against-russia-over-ukraine/sanctions-against-russia-explained/
14
Details of the US response, for example, can be found here: [Link]
bis/newsroom/press-releases/3158-2022-10-07-bis-press-release-advanced-computing-and-semiconductor-manufacturing-
controls-final/file
15
For simplicity, all counterfactual scenarios assume the introduction of infinite trade costs for the specific sectors mentioned. This
ensures zero trade occurs between the relevant countries/groups, avoids arbitrary assumptions about the finite size of trade
barriers, and keeps the model from becoming too computationally heavy.

INTERNATIONAL MONETARY FUND 14

IMF WORKING PAPERS Fragmentation in Global Trade

countries from the RoW are forced to choose between trading exclusively with either the US and EU or with
Russia and China based on historical trade intensities. A summary of the assumptions is presented in the text
table below. We also analyze the results using an alternative country grouping based on geopolitical
‘closeness’ in the Robustness section.

Table 1: Trade assumptions

p in scenarios
Scenario Country Group I Country Group II ROW Commodities/sectors
A - - - Trade barriers on energy sector

B - - - Trade barriers on high-tech sectors

C - - - Trade barriers on agriculture goods

Strategic Russia Trade barriers on all sectors

USA and EU Free trade
decoupling China Trade barriers on high-tech sectors
Join USA (China) group if country Trade barriers on all sectors
Geo-economic trades more (less) with USA than
USA and EU Russia and China
fragmentation China, resulting in tariffs on all sectors Trade barriers on all sectors
by the other group.

It should be stressed that these simulations do not aim to reflect the full effects of global economic
fragmentation, as some of the important channels, including trade-induced technology spillovers and
uncertainty, are not captured. In addition, the range of possible effects provided by the five scenarios captures
some, but not all, of the possible outcomes for the magnitude and coverage of the trade barriers as well as
possible policy responses. Moreover, how countries decide between joining trading blocs (if necessary), would
depend on many factors that go beyond their historical trade relations. Thus, the scenarios should be viewed
as illustrative.

INTERNATIONAL MONETARY FUND 15

IMF WORKING PAPERS Fragmentation in Global Trade

5. Results
Main Results

Before diving into the results for each scenario, we illustrate the importance of using a more granular dataset to
account for production and trade in commodities. Specifically, we compare the welfare change calculated under
two experiments with and without disaggregated commodity sectors (i.e., the only difference is the input
dataset, with both using the same trade model and elasticities). For simplicity, both experiments consider the
output loss of moving to autarky. 16 The aggregated experiment uses the standard Eora-26 IO matrix and the
disaggregated experiment uses the granular commodity dataset. The estimates use the exact hat algebra
summarized by the algorithm in the Annex. Figure 1 presents the ratio of the output loss incorporating granular
commodity trade to the welfare loss using the standard IO matrix. 17 We find that the loss more than doubles for
LICs, who are heavily dependent on trade in commodities, while for AEs and EMEs the welfare losses increase
by 4 and 25 percent respectively. There are two key insights here. First, aggregated commodity sectors
effectively treat different commodities as being infinitely substitutable, despite the fact that products as diverse
as gold and natural gas are included. In reality, commodities tend to be very specific to downstream production
chains with limited substitutability (hence the low elasticities of demand). Second, aggregate sectors make it
seem like commodities are produced widely across many countries when in fact the production of individual
commodities tends to be geographically concentrated (hence trade barriers may greatly restrict access to
particular commodity inputs).

Figure 1. Relative Output Loss in Model with and without Disaggregated Commodities Sectors
2.5 2.5

2 2

1.5 1.5
Global average

1 1

0.5 0.5

0 0
AE EM LIC
Sources: Fund staff calculations.

16
This counterfactual involves two changes in exogenous variables. We increase all international trade barriers such that there is no
international trade in the new equilibrium. We also set trade imbalances to zero.
17
Throughout the paper, we present impacts in terms of changes in output, or real GDP. In the model these are equivalent to
changes in real factor income. We aggregate to group-level and global impacts using real GDP (PPP) weights from 2019.

INTERNATIONAL MONETARY FUND 16

IMF WORKING PAPERS Fragmentation in Global Trade

Turning to the scenarios, Figure 2 shows the results of the simulations. The charts show the weighted average
impact on real GDP for AEs, EMs and LICs in each scenario. The effects are differentiated by the ‘short-run’,
corresponding to one year after the increase in trade barriers with a low associated trade elasticity, and ‘long-
run’, corresponding to 10 years (shown by Boehm et al. (forthcoming) to be the time required for the elasticity
to stabilize) after the barrier increase and a higher trade elasticity. To be conservative, we only lower the short-
run trade elasticity for strategic sectors (commodities and high-tech) to 36 percent of the long-run value. 18

Under Scenario A, the elimination of energy trade has a significant negative short-run impact on all countries,
ranging from a 10.1 percent decrease in output for EMs to a 7.8 percent decrease for AEs and a 6.3 percent
decrease for LICs. This reflects the different degrees of reliance on energy for production as well as fewer
energy exporters among LICs. Over the long run, the impacts are significantly lower for all country groups,
ranging from 1.1 to 2.1 percent, as countries reallocate inputs to other sectors and find other substitutes for
internationally traded energy.

Scenario B shows the outcome of zero high-tech trade, with AEs experiencing the largest output loss. The
result reflects the broader representation of high-tech trade for AEs. Scenario C underscores the importance of
international trade in agriculture for the global living standards and especially for LICs that are highly dependent
on agricultural goods imports.

In the Strategic Decoupling scenario (SD), the increase in trade barriers between the US/EU and Russia/China
groups reduces output for AEs and EMs by 0.7 and 0.8 percent in the short run respectively, but only by 0.3
percent in the long run. LICs, however, benefit slightly due to trade diversion over both time periods, as
demand and supply from trading partners in the two groups move to countries that still trade freely. Overall, we
estimate a fall in global output of 0.3 percent in the long-run, and 0.8 percent in the short run with our baseline
elasticities.

In the Geo-economic Fragmentation scenario (GEF), when countries in the ROW are forced into exclusive
trade relations with one of the groups and trade barriers are applied across all sectors, the negative impact
across all country groups increases dramatically. In the aggregate, there is a long run reduction in real GDP of
2.3 percent, and a short-run reduction of 4.8 percent. Losses for AEs and EMs are 4.2 and 5.2 percent in the
short run and 2.1 and 2.5 percent in the long run, respectively. LICs would come under significant pressure,
experiencing output losses of 10.8 percent in the short run and 4.3 percent in the long-run. LICs lose most
because they are most exposed to commodity trade, i.e., they import and export commodities at a higher rate
than AEs and EMs. This is because (i) LICs are more commodity intensive (especially in agriculture), (ii) LICs
tend to be small and therefore rely more on commodity imports, and (iii) LICs tend to specialize in the exports
of key commodities, especially metals and energy.

The last panel of Figure 3 provides a breakdown of results by key regions under the SD and the GEF scenarios
(using the baseline, long-run elasticities). Under the SD scenario, we see positive impacts across several
regions due to the diversion of trade that previously occurred between blocs. The largest positive impact occurs
for Middle East and Central Asia, while the largest negative impact occurs for Emerging Market Europe (losses
reduce from 3.8 percent of GDP to 3.0 percent if Russia and Turkey are excluded). As before, the GEF
produces negative impacts across all regions with Latin American and the Caribbean showing the smallest

36 percent is the ratio of the 1-year elasticity to the 10-year elasticity in Boehm et al. (forthcoming). See the Data section for
18

more details.

INTERNATIONAL MONETARY FUND 17

IMF WORKING PAPERS Fragmentation in Global Trade

negative impact and Emerging Market Europe showing the biggest losses. The losses depend largely on the
regions’ pre-fragmentation trade exposure to the other bloc, their overall trade openness, and the concentration
of trade exposure in sectors with low elasticities of substitution.

Figure 2. Estimated Output Losses under Different Scenarios

Scenario A: Tariffs on Energy Sector
(percent deviation from baseline)
0 0
AE LIC
EM AE LIC
EM
-5 -5

-10 -10
Global average

-15 -15

-20 -20

-25 -25

-30 -30

-35 -35
SR LR

Sources: Fund staff calculations.

Scenario B: Tariffs on High-tech Sectors

(percent deviation from baseline)
0 0
AE EM LIC LIC
-5 AE EM -5

-10 Global average -10

-15 -15

-20 -20

-25 -25

-30 -30

-35 -35
SR LR

Sources: Fund staff calculations.

INTERNATIONAL MONETARY FUND 18

IMF WORKING PAPERS Fragmentation in Global Trade

Scenario C: Tariffs on Agriculture Goods

(percent deviation from baseline)
0 0
AE LIC AE EM
EM
-5 -5
Global average

-10 -10
LIC
-15 -15

-20 -20

-25 -25

-30 -30

-35 -35
SR LR

Sources: Fund staff calculations.

Figure 3. Estimated Output Losses under Different Scenarios (continued)

Strategic Decoupling Scenario
(percent deviation from baseline)
0.5 0.5

LIC LIC
0 0
EM AE EM
AE

-0.5 -0.5

-1 -1

-1.5 -1.5
SR LR

Sources: Fund staff calculations.

Geo-economic Fragmentation Scenario

(percent deviation from baseline)
0 0
AE EM LIC AE EM LIC

-5 -5

-10 -10

-15 -15
SR

Sources: Fund staff calculations.

INTERNATIONAL MONETARY FUND 19

IMF WORKING PAPERS Fragmentation in Global Trade

Robustness checks
Trade Elasticities

While our welfare loss estimates are based on trade elasticities sourced from the most recent literature, it is
also important to recognize that the literature has produced a range of different estimates which vary based on
differences in data and methodology used. Table 2 below provides an overview of recent papers providing such
estimates.

Table 2: Selected Trade Elasticity Estimates in the Literature 1/

INTERNATIONAL MONETARY FUND 20

IMF WORKING PAPERS Fragmentation in Global Trade

Figure 4: Range of estimated output losses

(percent deviation from baseline)

0 0
-0.2
-1
-0.3
-1.9
-2
-2.3
-3

-1 -0.9 -4

-5
Baseline
-6
Upper bound
-7
Lower bound -7.0
-2 -8
Strategic Decoupling Geo-economic Fragmentation
(RHS)
Sources: Fund staff calculations.

In this section we present additional results to test the robustness of our results to variations in the choice of
trade elasticities. To do this, we take the variance in long-run trade elasticities by sector from Fontagné (2022)
and scale them up and down based on the ratio of their mean value to the mean value (or point estimate) from
other papers in the literature. We therefore take a maximum value of 6.6 from Broda and Weinstein (2006), and
a minimum value of 2.12 from Boehm et at. (forthcoming). This places our mean value of 6.00 towards the
conservative end of the spectrum of possibilities as higher elasticities, all else equal, lead to smaller welfare
costs.

Figure 4 presents the results for both the strategic decoupling and geo-economic fragmentation scenarios. In
both cases, our baseline results sit on the lower end of the range of results and hence represent conservative
estimates relative to other possibilities in the literature (see the Data section for further explanation of this
choice for baseline elasticities). For example, in the geo-fragmentation scenario welfare costs range from 1.9
percent of GDP for the global economy in the high elasticity case to 7.0 percent of GDP in the low elasticity
case (with a baseline estimate of 2.3 percent of GDP).

Inter-bloc aggregate imbalances

As discussed in the modelling section, our model allows for endogenous bilateral trade imbalances, but
aggregate trade imbalances are assumed to be exogenous and remain unchanged in the various
counterfactuals (as commonly done in the literature, see Costinot and Rodriguez-Clare, 2014). As a robustness
check, we rerun the baseline results with scaled imbalances such that all surpluses and deficits cancelled out
within each bloc by construction (i.e., inter-bloc imbalances are zero).

The results are presented in Table 3, and remain very close to those from the baseline analysis. There are
several explanations for this. First, inter-bloc imbalances are relatively low when countries are assigned to

INTERNATIONAL MONETARY FUND 21

IMF WORKING PAPERS Fragmentation in Global Trade

blocs largely based on economic interests (i.e., trading relationships), the resulting scaling factors are also
therefore relatively small (on the order of 15 percent). Second, nominal wages and prices tend to move
together in the model as country deficits/surpluses are scaled meaning that even though changes in nominal
GDP can be sizable, changes to real GDP are more muted. 19 Third, the model assumes full employment which
also minimizes output changes. Fourth, the model does not contain capital flows or technology impacts and
hence differences in imbalances do not lead to corresponding differences in equilibrium capital stocks or
productive capacity through differences in technology.

Alternative country blocs

The specific grouping of countries in different fragmentation scenarios is fundamentally assumptions driven. In
our baseline, we choose a politically neutral approach by focusing largely on pre-pandemic trading
relationships. Aside from the EU and Russia, who are grouped with the US and China, respectively, countries
are assigned to the US group if they trade more with the US, and to the China group if they trade more with
China.

However, it is reasonable to consider a different approach to assigning countries to different trading blocs
based on political ties (or foreign policy similarities) rather than economic interests. In order to do this, we
employ the ‘Ideal Point Distance’ (IPD) from Bailey et al. (2017). 20 The IPD is based on historical voting
patterns at the UN Generaly Assembly and allows calculation of bilateral political ‘distances’ between two
countries based on how similar their voting patterns have been.

The bilateral IPD scores (averaged over 2017-2021) are calculated for all countries with respect to both the US
and China. Then every country is ranked based on their closeness to each of the two pole countries. If a
country is more highly ranked for the US than China, then it is “assigned” to the US bloc (and vice versa). We
make two manual reassignments, moving Russia and Mali from the US group into the China group. For Russia,
the justification is self-evident. For Mali, we note that this is the only country that would have been placed in the
US group based on the IPD measure that voted against the United Nations General Assembly Resolution on
Ukraine (A/ES-11/L. 7). 21

The last column in Table 3 presents the output loss estimates for the IPD-based country groupings
(Geopolitical blocs). Aggregate losses amount to 3.2 percent of global GDP, significantly more than under the
baseline scenario in which blocs are based on trade intensities. It is intuitive that the economic costs would be
larger under the geopolitical bloc formation, as inherently this implies choosing exclusive trading relations not
solely based on economic interests (i.e., trade intensities). LICs fair somewhat similarly under the robustness
exercise, as their economic and geopolitical preferences are appear to be more aligned. AEs see the largest

19
The small impact of changes in trade imbalances on output is in line with Dekle et al. (2007, 2008) who find that in a
counterfactual world with all current accounts balancing, large changes in nominal variables (wages, nominal GDP), large changes
translate into muted changes in real GDP. The reason for this is that the more a country’s relative wage (and hence nominal GDP)
needs to decline to make exports competitive abroad, the lower its domestic price index.

20
A similar approach is also used in Chapter 3 of the IMF April 2023 WEO (IMF, 2023) except that (i) we assign all countries to one
or other of the blocs based on relative rankings as opposed to allowing for a non-assigned group, and (ii) we assign all countries
individually rather than grouping them together in regions.
21
Our model also has a small number of economies for which no IPD is available, in these cases we manually assigned: New
Caledonia to the US group; Palestine to the US group; Hong Kong SAR to the China group. We also dropped Serbia from the
sample given an absence of a corresponding IPD.

INTERNATIONAL MONETARY FUND 22

IMF WORKING PAPERS Fragmentation in Global Trade

difference, moving from 2.1 percent in losses to 3.4 percent under geopolitical bloc formation. EMs also see
increase in losses from 2.5 percent to 3 percent.

Table 3: GDP Losses for Robustness Checks

Country Grouping Baseline - GEF No inter-bloc imbalances Geopolitical blocs
AE -2.14 -2.09 -3.40
EM -2.48 -2.36 -3.03
LIC -4.25 -4.12 -4.17
Global -2.34 -2.25 -3.20

INTERNATIONAL MONETARY FUND 23

IMF WORKING PAPERS Fragmentation in Global Trade

6. Conclusion
This paper estimates output losses for a range of potential global trade fragmentation scenarios using a newly
constructed dataset that provides detailed trade and production data for commodities, and a novel model which
incorporates the key features of production and trade in commodities. We use trade and demand elasticities
which are both conservative and based on the most up to date literature.

First, we show that not accounting for granular commodity production and trade linkages leads to
underestimation of the output losses associated with trade fragmentation. LICs face losses that are more than
twice compared to a model using aggregate commodity sectors, while AEs and EMs face costs that are 4
percent and 25 percent larger respectively.

Second, we show that output losses tend to be larger the deeper the fragmentation scenario and that LICs
experience larger losses than AEs or EMs. Trade barriers that are limited to specific countries or specific
sectors, whilst allowing the RoW to trade freely, lead to relatively contained GDP losses in the long run as
production processes and source countries adjust and trade diversion provides a boost to countries outside of
the main trade blocs. In contrast, a severe fragmentation scenario leads to larger losses, particularly for low-
income countries that are forced to choose between one bloc or the other. Moreover, we illustrate the
estimated output loss varies widely depending on the assumption on trade elasticities. Formation of blocs by
geopolitical allegiances as opposed to economic interests also increases the losses, while closing inter-bloc
trade imbalances does not materially impact the results.

Third, we exploit recent work on the variation in trade elasticities over time to estimate the potential short run
costs of fragmentation. With trade elasticities as a fraction of their long run values we see costs of 10.8 percent
for LICs in the short run as countries faced adjustment costs which reduce their ability to adapt quickly.
Furthermore, we run robustness checks that use a range of reasonable trade elasticities from the literature to
demonstrate that our loss estimates sit on the conservative side of the spectrum.

Future research will concentrate on several elements. The difference between short and long-run elasticities
can be micro founded by introducing quadratic adjustment costs in a dynamic system. The model could be
additionally extended to introduce endogenous capital formation and technological progress which would
induce long-run welfare effects from short-run adjustment costs. Furthermore, the model currently assumes
exogenous overall trade imbalances, an assumption that could be relaxed in future research. There is also
scope for algorithm of endogenous bloc formation where countries choose their bloc in an iterative process on
the basis of which other countries choose to join and their potential welfare benefits depending on the evolving
bloc composition.

INTERNATIONAL MONETARY FUND 24

IMF WORKING PAPERS Fragmentation in Global Trade

References
Aiyar, S., Chen, J., Ebeke, C., Garcia-Saltos, R., Gudmundsson, T., Ilyina, A., Kangur, A., Rodriguez, S., Ruta,
M., Schulze, T., Trevino, J., Kunaratskul, T. and Soderberg, G. “Geoeconomic Fragmentation and the Future of
Multilateralism”. IMF Staff Discussion Note, SDN/2023/01, International Monetary Fund, Washington, DC.

Albrizio, S., Bluedorn, J. C., Koch, C., Pescatori, A., & Stuermer, M. (2022). Market Size and Supply
Disruptions: Sharing the Pain from a Potential Russian Gas Shut-off to the EU. IMF Working
Papers, 2022(143).

Amiti, M., Redding, S. J., & Weinstein, D. E. (2019). The impact of the 2018 tariffs on prices and
welfare. Journal of Economic Perspectives, 33(4), 187-210.

Anderson, J. E., & Yotov, Y. V. (2023). Estimating gravity from the short to the long run: A simple solution to
the'International Elasticity Puzzle' (No. w30809). National Bureau of Economic Research.

Arkolakis, C., Costinot, A., & Rodríguez-Clare, A. (2012). New trade models, same old gains?. American
Economic Review, 102(1), 94-130.

Auer, R. A., Levchenko, A. A., & Sauré, P. (2019). International inflation spillovers through input
linkages. Review of Economics and Statistics, 101(3), 507-521.

Bachmann, R., Baqaee, D., Bayer, C., Kuhn, M., Löschel, A., Moll, B., ... & Schularick, M. (2022). What if? The
economic effects for Germany of a stop of energy imports from Russia. ECONtribute Policy Brief, 28, 2022.

Bailey, M., Strezhnev, A., and Voeten, E.. 2017. “Estimating dynamic state preferences from United Nations
voting data.” Journal of Conflict Resolution 61(2): 430-56.

Bekkers, E., & Góes, C. (2022). The impact of geopolitical conflicts on trade, growth, and innovation: An
illustrative simulation study. Global Economic Consequences of the War in Ukraine Sanctions, Supply Chains
and Sustainability.

Boehm, C. E., Flaaen, A., & Pandalai-Nayar, N. (2019). Input linkages and the transmission of shocks: Firm-
level evidence from the 2011 Tōhoku earthquake. Review of Economics and Statistics, 101(1), 60-75.

Boehm, C. E., Levchenko, A. A., & Pandalai-Nayar, N. (forthcoming). The long and short (run) of trade
elasticities. American Economic Review.

Broda, C., & Weinstein, D. E. (2006). Globalization and the Gains from Variety. The Quarterly journal of
economics, 121(2), 541-585.

Caceres, C., Cerdeiro, D. A., & Mano, R. (2019). Trade wars and trade deals: Estimated effects using a multi-
sector model. International Monetary Fund.

INTERNATIONAL MONETARY FUND 25

IMF WORKING PAPERS Fragmentation in Global Trade

Caliendo, L., & Parro, F. (2015). Estimates of the Trade and Welfare Effects of NAFTA. The Review of
Economic Studies, 82(1), 1-44.

Cerdeiro, D. A., Mano, R., Eugster, J., Muir, M. D. V., & Peiris, M. S. J. (2021). Sizing up the effects of
technological decoupling. International Monetary Fund.
Costinot, A., & Rodríguez-Clare, A. (2014). Trade theory with numbers: Quantifying the consequences of
globalization. In Handbook of international economics (Vol. 4, pp. 197-261). Elsevier.

Cuñat, A., & Zymek, R. (2022). Bilateral trade imbalances. IMF Working Paper No. 2022/090.

De Bromhead, A., Fernihough, A., Lampe, M., & O'Rourke, K. H. (2019). When Britain turned inward: the
impact of interwar British protection. American Economic Review, 109(2), 325-52.

Dekle, R., Eaton, J., & Kortum, S. (2007). Unbalanced trade. American Economic Review, 97(2), 351-355.

Dekle, R., Eaton, J., & Kortum, S. (2008). Global rebalancing with gravity: Measuring the burden of
adjustment. IMF Staff Papers, 55(3), 511-540.

Di Bella, G., Flanagan, M. J., Foda, K., Maslova, S., Pienkowski, A., Stuermer, M., & Toscani, F. G. (2022).
Natural gas in Europe: the potential impact of disruptions to supply. IMF Working Papers, 2022(145).

Fajgelbaum, P. D., Goldberg, P. K., Kennedy, P. J., & Khandelwal, A. K. (2020). The return to
protectionism. The Quarterly Journal of Economics, 135(1), 1-55.

Fally, T., & Sayre, J. (2018). Commodity trade matters (No. w24965). National Bureau of Economic Research.

Fontagné, L., Guimbard, H., & Orefice, G. (2022). Tariff-based product-level trade elasticities. Journal of
International Economics, 137, 103593.

Gaulier, G. and Zignago, S. (2010) BACI: International Trade Database at the Product-Level. The 1994-2007
Version. CEPII Working Paper, N°2010-23.

Giri, R., Yi, K. M., & Yilmazkuday, H. (2021). Gains from trade: Does sectoral heterogeneity matter?. Journal of
International Economics, 129, 103429.

Goes, C., Bekkers, E., (2022), The Impact of Geopolitical Conflicts on Trade, Growth, and Innovation. Work in
progress.

International Monetary Fund. (2022). Regional Economic Outlook: Asia and Pacific, October 2022. Washington,
DC.

International Monetary Fund. (2023). World Economic Outlook, April 2023 (forthcoming). Washington, DC.

Javorcik, B., Kitzmueller, L., Schweiger, H., Yıldırım M. (2022), Economics Costs of Friend-Shoring, European
Bank for Reconstruction and Development

INTERNATIONAL MONETARY FUND 26

IMF WORKING PAPERS Fragmentation in Global Trade

Kee, H. L., Nicita, A., & Olarreaga, M. (2008). Import demand elasticities and trade distortions. The Review of
Economics and Statistics, 90(4), 666-682.

Lan, T., Sher, G., Zhou, J. (2022), The Economic Impacts on Germany of a Potential Russian Gas Shutoff, IMF
Working Paper WP/22/144

Lenzen, M., Kanemoto, K., Moran, D., & Geschke, A. (2012). Mapping the structure of the world
economy. Environmental science & technology, 46(15), 8374-8381.

Lenzen, M., Moran, D., Kanemoto, K., & Geschke, A. (2013). Building Eora: a global multi-region input–output
database at high country and sector resolution. Economic Systems Research, 25(1), 20-49.

Organization for Economic Cooperation and Development (OECD), 2011, “ISIC Rev. 3 Technology Intensity
Definition,” OECD Directorate for Science, Technology and Industry, Economic Analysis and Statistics Division,
available at [Link]

Ossa, R. (2014). Trade Wars and Trade Talks with Data. American Economic Review, 104(12), 4104-4146.

Ossa, R. (2015). Why trade matters after all. Journal of International Economics, 97(2), 266-277.

Romalis, J. (2007). NAFTA's and CUSFTA's Impact on International Trade. The review of Economics and
Statistics, 89(3), 416-435.

Simonovska, I., & Waugh, M. E. (2014). The elasticity of trade: Estimates and evidence. Journal of international
Economics, 92(1), 34-50.

Soderbery, A. (2018). Trade elasticities, heterogeneity, and optimal tariffs. Journal of International
Economics, 114, 44-62.

INTERNATIONAL MONETARY FUND 27

IMF WORKING PAPERS Fragmentation in Global Trade

Annex I. Data
Commodity list

Agriculture Commodity Mineral Product

Almonds Lettuce Antimony Rare Earths
Apples Linseed Arsenic Salt
Apricots Maize Asbestos Selenium
Asparagus Mandarins Barytes & Strontium Sillimanite
Avocados Mangoes Bauxite Silver
Bananas Mate Beryl Talc
Barley Melons Bismuth Tellurium
Beans Millet Borates Tin
Berries Mushrooms Bromine Titanium
Brazil nuts Mustard seeds Cadmium Tungsten
Buckwheat Natural rubber Chromium Uranium
Canary seed Nutmeg Coal Vermiculite
Carrots Oats Cobalt Wollastonite
Cashews Onions Copper Zinc
Cassava Oranges Crude Oil Zirconium
Cauliflowers Papayas Diamond Other minerals
Cherries Peaches Diatomite
Chestnuts Pears Feldspar
Chickpeas Peas Fluorspar
Cinnamon Pineapples Gallium,indium,rhenium, thallium
Cloves Pistachios Germanium
Cocoa Plums Gold
Coconuts Poppy seeds Graphite
Coffee Potatoes Gypsum
Cotton seeds Rapeseed Iodine
Cucumbers Rice Iron & Steel
Dates Rye Lead
Eggplants Sorghum Lithium
Figs Soy beans Magnesium
Flax Spinach Manganese
Garlic Sugar Mercury
Ginger Sunflower seeds Mica
Grapefruit Sweet potatoes Molybdenum
Grapes Tea Natural Gas
Hazelnuts Tobacco Nickel
Hops Tomatoes Niobium, tantalum, vanadium
Jute Vanilla Palladium
Leeks Walnuts Phosphate
Lemons Wheat Platinum
Lentils Other agriculture goods Potash

INTERNATIONAL MONETARY FUND 28

IMF WORKING PAPERS Fragmentation in Global Trade

Trade Elasticities

Sector – TiVA 2016 Sector(s) - Eora Trade elasticity

Agriculture, hunting, forestry and Agriculture, Fishing -2.91
fishing
Mining and quarrying Mining and Quarrying -3.41
Food products, beverages and tobacco Food & Beverages -4.17
Textiles, textile products, leather and Textiles and Wearing Apparel -4.71
footwear
Wood and products of wood and cork Wood and Paper -8.80
Pulp, paper, paper products, printing -8.21
and publishing
Coke, refined petroleum products and Petroleum, Chemical and Non- -3.67
nuclear fuel Metallic Mineral Products
Chemicals and chemical products -10.56
Rubber and plastics products -6.75
Other non-metallic mineral products -4.79
Basic metals Metal Products -7.39
Fabricated metal products -4.22
Machinery and equipment, nec Electrical and Machinery -5.01
Computer, electronic and optical -5.14
equipment
Electrical machinery and apparatus, -4.11
nec
Motor vehicles, trailers and semi- Transport Equipment -8.92
trailers
Other transport equipment -8.99
Manufacturing nec; recycling Other Manufacturing, Recyling -4.06
Other community, social and personal Construction-Education -8.35
services

Notes: Adapted from Fontagné et al. (2022), Table 8. For Eora sectors that comprise multiple TiVA sectors, we
assign the unweighted average estimate. Wood and Paper, for example, is assigned a trade elasticity of -8.505
(the average of -8.8 and -8.21). We assign an elasticity of -8.35 for all services sectors. Trade elasticities are
defined as the change in trade in response to a change in trade costs.

INTERNATIONAL MONETARY FUND 29

IMF WORKING PAPERS Fragmentation in Global Trade

Annex II. Modelling framework details

Full Model

Environment. There are N countries, denoted by m (exporter) and n (importer). There are K non-commodity
sectors, denoted by k, and J commodity sectors, denoted by j.

Production. In each country, a representative retailer produces composite final good that is aggregated from K
non-tradable non-commodity goods, solving:

max 𝑃𝑃𝑛𝑛 𝑄𝑄𝑛𝑛 − � 𝑃𝑃𝑛𝑛𝑘𝑘 𝑄𝑄�𝑛𝑛𝑘𝑘 (𝐴𝐴. 1),

𝑘𝑘

subject to
𝜎𝜎 𝑘𝑘
𝑄𝑄�𝑛𝑛𝑘𝑘 𝑛𝑛
𝑄𝑄𝑛𝑛 = � � 𝑘𝑘 � (𝐴𝐴. 2),
𝜎𝜎𝑛𝑛
𝑘𝑘

where 𝑄𝑄𝑛𝑛 is total output of the composite final good, 𝑄𝑄�𝑛𝑛𝑘𝑘 are non-commodity inputs, and ∑𝑘𝑘 𝜎𝜎𝑛𝑛𝑘𝑘 = 1.

In each non-commodity sector,

• a representative wholesaler combines varieties from different countries into composite, solving:

max 𝑃𝑃𝑛𝑛𝑘𝑘 𝑄𝑄𝑛𝑛𝑘𝑘 − � 𝑝𝑝𝑚𝑚

𝑘𝑘 𝑘𝑘 𝑘𝑘
𝜏𝜏𝑚𝑚𝑚𝑚 𝑥𝑥𝑚𝑚𝑚𝑚 (𝐴𝐴. 3),
𝑚𝑚

subject to

1+𝜃𝜃𝑘𝑘
1 𝜃𝜃𝑘𝑘 𝜃𝜃𝑘𝑘
𝑄𝑄𝑛𝑛𝑘𝑘 = 𝑘𝑘 )1+𝜃𝜃𝑘𝑘
��(𝜔𝜔𝑚𝑚𝑚𝑚 𝑘𝑘 )1+𝜃𝜃𝑘𝑘
(𝑥𝑥𝑚𝑚𝑚𝑚 � (𝐴𝐴. 4),
𝑚𝑚

where 𝑥𝑥𝑚𝑚𝑚𝑚
𝑘𝑘
are inputs from countries m, 𝜃𝜃𝑘𝑘 is the sector’s trade elasticity, and ∑𝑚𝑚 𝜔𝜔𝑚𝑚𝑚𝑚
𝑘𝑘 𝑘𝑘
= 1. 𝜏𝜏𝑚𝑚𝑚𝑚 are trade costs,
which we describe in more detail below.

• a representative variety producer combines labor with intermediate inputs from the composite final
good and a composite commodity good, solving

max 𝑝𝑝𝑛𝑛𝑘𝑘 𝑞𝑞𝑛𝑛𝑘𝑘 − 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 − 𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑘𝑘 − 𝑤𝑤𝑛𝑛 𝐿𝐿𝑘𝑘𝑛𝑛 (𝐴𝐴. 5),

subject to

INTERNATIONAL MONETARY FUND 30

IMF WORKING PAPERS Fragmentation in Global Trade

𝜂𝜂
𝜂𝜂−1 𝜂𝜂−1
𝑘𝑘
⎡ 𝜂𝜂−1
𝑘𝑘 (1−𝜇𝜇𝑘𝑘 )
𝛼𝛼𝑛𝑛 𝑛𝑛
1−𝛼𝛼𝑛𝑛 𝜂𝜂 ⎤

𝑞𝑞𝑛𝑛𝑘𝑘 = 𝑧𝑧𝑛𝑛𝑘𝑘 ⎢⎢(𝛼𝛼𝑛𝑛𝑘𝑘 𝜇𝜇𝑛𝑛𝑘𝑘 ) 𝜂𝜂 (𝑂𝑂𝑛𝑛𝑘𝑘 )1−𝜂𝜂 + (1 − 𝛼𝛼𝑛𝑛𝑘𝑘 𝜇𝜇𝑛𝑛𝑘𝑘 )1/𝜂𝜂 �(𝑀𝑀𝑛𝑛 ) � ⎥⎥
𝑘𝑘 (1−𝜇𝜇𝑘𝑘 )+1−𝛼𝛼 𝑘𝑘
𝑘𝑘 𝛼𝛼𝑛𝑛 𝛼𝛼 𝑘𝑘 �1−𝜇𝜇𝑛𝑛
𝑘𝑘 �+1−𝛼𝛼 𝑘𝑘
𝑛𝑛 𝑛𝑛 ⋅ 𝐿𝐿𝑘𝑘 𝑛𝑛𝑛𝑛 𝑛𝑛
(𝐴𝐴. 6),
⎢ ⎥
⎣ ⎦

where 𝐿𝐿𝑘𝑘𝑛𝑛 , 𝑂𝑂𝑛𝑛𝑘𝑘 , and 𝑀𝑀𝑛𝑛𝑘𝑘 are labor, composite commodity goods, and intermediate inputs. 𝑧𝑧𝑛𝑛𝑘𝑘 is a productivity
term. 𝜂𝜂 is the elasticity of substitution between composite commodities and other inputs.

In each country, the commodity retailer aggregates a composite from J commodities, solving

𝑗𝑗 𝑗𝑗
max 𝑃𝑃𝑛𝑛𝑜𝑜 𝑄𝑄𝑛𝑛𝑜𝑜 − � 𝑃𝑃𝑛𝑛 𝑄𝑄�𝑛𝑛 (𝐴𝐴. 7),
𝑗𝑗
subject to
𝑗𝑗
𝑗𝑗 𝜎𝜎𝑛𝑛
𝑄𝑄�𝑛𝑛
𝑄𝑄𝑛𝑛𝑜𝑜 = � � 𝑗𝑗 � (𝐴𝐴. 8),
𝑗𝑗
𝜎𝜎𝑛𝑛

𝑗𝑗 𝑗𝑗
where 𝑄𝑄𝑛𝑛𝑜𝑜 is total output of the composite commodity good, 𝑄𝑄�𝑛𝑛 are commodity inputs, and ∑𝑗𝑗 𝜎𝜎𝑛𝑛 = 1.

In each commodity sector,

• a representative commodity wholesaler combines varieties from different countries into a composite,
solving

𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗
max 𝑃𝑃𝑛𝑛 𝑄𝑄𝑛𝑛 − � 𝑝𝑝𝑚𝑚 𝜏𝜏𝑚𝑚𝑚𝑚 𝑥𝑥𝑚𝑚𝑚𝑚 (𝐴𝐴. 9),
𝑚𝑚

1+𝜃𝜃𝑗𝑗
1 𝜃𝜃𝑗𝑗 𝜃𝜃𝑗𝑗
𝑗𝑗 𝑗𝑗 𝑗𝑗
𝑄𝑄𝑛𝑛 = ���𝜔𝜔𝑚𝑚𝑚𝑚 �1+𝜃𝜃𝑗𝑗 �𝑥𝑥𝑚𝑚𝑚𝑚 �1+𝜃𝜃𝑗𝑗 � (𝐴𝐴. 10),
𝑚𝑚

𝑗𝑗 𝑗𝑗
where 𝑥𝑥𝑚𝑚𝑚𝑚 are inputs from countries m, 𝜃𝜃𝑗𝑗 is the sector’s trade elasticity, and ∑𝑚𝑚 𝜔𝜔𝑚𝑚𝑚𝑚 = 1.

• a representative commodity variety producer combines labor with intermediate inputs from the
composite final good, solving

𝑗𝑗 𝑗𝑗 𝑗𝑗
max 𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛 − 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 − 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛 (𝐴𝐴. 11),

subject to
𝑗𝑗 𝑗𝑗
𝑗𝑗 1−𝛼𝛼𝑛𝑛 𝑗𝑗 𝛼𝛼𝑛𝑛
𝑗𝑗 𝑗𝑗 𝐿𝐿𝑛𝑛 𝑀𝑀𝑛𝑛
𝑞𝑞𝑛𝑛 = 𝑧𝑧𝑛𝑛 � 𝑗𝑗
� � 𝑗𝑗 � (𝐴𝐴. 12),
1 − 𝛼𝛼𝑛𝑛 𝛼𝛼𝑛𝑛

𝑗𝑗 𝑗𝑗 𝑗𝑗
where 𝐿𝐿𝑛𝑛 and 𝑀𝑀𝑛𝑛 are labor and intermediate inputs. 𝑧𝑧𝑛𝑛 is a productivity term.

INTERNATIONAL MONETARY FUND 31

IMF WORKING PAPERS Fragmentation in Global Trade

Household. In each country there is a measure of 𝐿𝐿𝑛𝑛 representative households that supply labor, consume the
final good, and receive an exogenous deficit. The budget constraint reads

𝑃𝑃𝑛𝑛 𝐶𝐶𝑛𝑛 = 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛 + 𝐷𝐷𝑛𝑛 (𝐴𝐴. 13),

where ∑𝒏𝒏 𝑫𝑫𝒏𝒏 = 𝟎𝟎.

Trade. Trade between countries is subject to iceberg trade costs where one unit of a tradable good in sector k
shipped from country m to country n requires producing 𝝉𝝉𝒌𝒌𝒎𝒎𝒎𝒎 ≥ 𝟏𝟏 units in m, with 𝝉𝝉𝒌𝒌𝒏𝒏𝒏𝒏 = 𝟏𝟏. The triangular
inequality holds such that 𝝉𝝉𝒌𝒌𝒎𝒎𝒎𝒎 𝝉𝝉𝒌𝒌𝒏𝒏𝒏𝒏 ≥ 𝝉𝝉𝒌𝒌𝒉𝒉𝒉𝒉 for all n, m, h.

𝒋𝒋 𝒋𝒋 𝒋𝒋 𝒋𝒋
Equilibrium. Given 𝑳𝑳𝒏𝒏 , 𝑫𝑫𝒏𝒏 , 𝒛𝒛𝒌𝒌𝒏𝒏 , 𝒛𝒛𝒏𝒏 , and 𝝉𝝉𝒌𝒌𝒎𝒎𝒎𝒎 , 𝝉𝝉𝒎𝒎𝒎𝒎 , an equilibrium is a set of wages 𝒘𝒘𝒏𝒏 , prices 𝑷𝑷𝒏𝒏 , 𝑷𝑷𝒐𝒐𝒏𝒏 , 𝑷𝑷𝒌𝒌𝒏𝒏 , 𝑷𝑷𝒏𝒏 , 𝒑𝒑𝒌𝒌𝒏𝒏 , 𝒑𝒑𝒏𝒏 ,
𝒋𝒋
� 𝒌𝒌𝒏𝒏 , 𝑸𝑸𝒋𝒋𝒏𝒏 , 𝑸𝑸
and allocations 𝑸𝑸𝒏𝒏 , 𝑪𝑪𝒏𝒏 , 𝑴𝑴𝒌𝒌𝒏𝒏 , 𝑴𝑴𝒏𝒏 , 𝑸𝑸𝒌𝒌𝒏𝒏 , 𝑸𝑸 � 𝒋𝒋𝒏𝒏 , 𝑳𝑳𝒌𝒌𝒏𝒏 , 𝑳𝑳𝒋𝒋𝒏𝒏 , 𝒒𝒒𝒌𝒌𝒏𝒏 , 𝒒𝒒𝒋𝒋𝒏𝒏 , 𝑶𝑶𝒌𝒌𝒏𝒏 such that

𝑗𝑗
𝑄𝑄𝑛𝑛 = 𝐶𝐶𝑛𝑛 + ∑𝑀𝑀𝑛𝑛𝑘𝑘 + ∑𝑀𝑀𝑛𝑛 (𝐴𝐴. 22),
𝑄𝑄𝑛𝑛𝑘𝑘 = 𝑄𝑄�𝑛𝑛𝑘𝑘 (𝐴𝐴. 23),
𝑗𝑗 𝑗𝑗
𝑄𝑄𝑛𝑛 = 𝑄𝑄�𝑛𝑛 (𝐴𝐴. 24),
𝑗𝑗
𝐿𝐿𝑛𝑛 = ∑𝐿𝐿𝑘𝑘𝑛𝑛 + ∑𝐿𝐿𝑛𝑛 (𝐴𝐴. 25),
𝑗𝑗 𝑗𝑗 𝑗𝑗
∑𝑝𝑝𝑛𝑛𝑘𝑘 𝑞𝑞𝑛𝑛𝑘𝑘 + ∑𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛 + 𝐷𝐷𝑛𝑛 = 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛 + ∑(𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑘𝑘 + 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 ) + ∑(𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛 )

• in each country, the representative retailer solves the problem summarized by (A1)-(A2) such that

′ ′
∑𝑃𝑃𝑛𝑛𝑘𝑘 𝑄𝑄𝑛𝑛𝑘𝑘
𝑄𝑄𝑛𝑛𝑘𝑘 = 𝜎𝜎𝑛𝑛𝑘𝑘 (𝐴𝐴. 14),
𝑃𝑃𝑛𝑛𝑘𝑘
and

𝑘𝑘
𝑃𝑃𝑛𝑛 = �(𝑃𝑃𝑛𝑛𝑘𝑘 )𝜎𝜎𝑛𝑛 (𝐴𝐴. 15),
𝑘𝑘

• in each country and each non-commodity sector, the representative wholesaler solves the problem
summarized by (A.3)-(A.4) such that

𝑘𝑘 𝑘𝑘 𝑘𝑘 𝑘𝑘
(𝜏𝜏𝑚𝑚𝑚𝑚 𝑘𝑘 )−𝜃𝜃𝑘𝑘
𝑘𝑘
𝑝𝑝𝑚𝑚 𝜏𝜏𝑚𝑚𝑚𝑚 𝑥𝑥𝑚𝑚𝑚𝑚 𝑝𝑝𝑚𝑚𝑚𝑚
𝜋𝜋𝑚𝑚𝑚𝑚 ≡ 𝑘𝑘 𝑘𝑘 𝑘𝑘 = −𝜃𝜃𝑘𝑘
(𝐴𝐴. 16),
∑𝑚𝑚′ 𝑝𝑝𝑚𝑚 ′ 𝜏𝜏𝑚𝑚′ 𝑛𝑛 𝑥𝑥𝑚𝑚′ 𝑛𝑛 𝑘𝑘
∑𝑚𝑚′ �𝜏𝜏𝑚𝑚 𝑘𝑘
′ 𝑛𝑛 𝑝𝑝𝑚𝑚′ 𝑛𝑛 �

and

1
−
𝜃𝜃𝑘𝑘
𝑃𝑃𝑛𝑛𝑘𝑘 = ��(𝜏𝜏𝑚𝑚𝑚𝑚
𝑘𝑘 𝑘𝑘 )−𝜃𝜃𝑘𝑘
𝑝𝑝𝑚𝑚 � (A. 17),
𝑚𝑚

• in each country and each non-commodity sector, the representative producer solves the problem
summarized by (A.5)-(A.6) such that

INTERNATIONAL MONETARY FUND 32

IMF WORKING PAPERS Fragmentation in Global Trade

(1 − 𝛼𝛼𝑛𝑛𝑘𝑘 )𝑝𝑝𝑛𝑛𝑘𝑘 𝑞𝑞𝑛𝑛𝑘𝑘

𝐿𝐿𝑘𝑘𝑛𝑛 = (𝐴𝐴. 18),
𝑤𝑤𝑛𝑛
𝑃𝑃𝑛𝑛𝑜𝑜 1−𝜂𝜂 𝑘𝑘 𝑘𝑘
𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 = 𝛼𝛼𝑛𝑛𝑘𝑘 𝜇𝜇𝑛𝑛𝑘𝑘 ⋅ �� 𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛 (𝐴𝐴. 19),
𝑝𝑝𝑛𝑛𝑘𝑘
𝛼𝛼𝑛𝑛𝑘𝑘 (1 − 𝜇𝜇𝑛𝑛𝑘𝑘 )𝑝𝑝𝑛𝑛𝑘𝑘 𝑞𝑞𝑛𝑛𝑘𝑘
𝑀𝑀𝑛𝑛𝑘𝑘 = (𝐴𝐴. 20),
𝑃𝑃𝑛𝑛
and

1
𝑘𝑘 1−𝜂𝜂 1−𝜂𝜂
𝑘𝑘 (1−𝜇𝜇𝑘𝑘 )
𝛼𝛼𝑛𝑛 1−𝛼𝛼𝑛𝑛
𝑛𝑛 𝑘𝑘 �1−𝜇𝜇𝑘𝑘 �+1−𝛼𝛼 𝑘𝑘
𝑘𝑘 (1−𝜇𝜇𝑘𝑘 )+1−𝛼𝛼 𝑘𝑘 𝛼𝛼𝑛𝑛
𝑝𝑝𝑛𝑛𝑘𝑘 = (𝑧𝑧𝑛𝑛𝑘𝑘 )−1 �𝛼𝛼𝑛𝑛𝑘𝑘 𝜇𝜇𝑛𝑛𝑘𝑘 (𝑃𝑃𝑛𝑛𝑜𝑜 )1−𝜂𝜂 + (1 − 𝛼𝛼𝑛𝑛𝑘𝑘 𝜇𝜇𝑛𝑛𝑘𝑘 ) �(𝑃𝑃𝑛𝑛 ) 𝑛𝑛
𝛼𝛼 𝑛𝑛 𝑛𝑛 ⋅ 𝑤𝑤𝑛𝑛 𝑛𝑛 𝑛𝑛
� � (𝐴𝐴. 21),

• in each country, the representative commodity retailer solves the problem summarized by (A7)-(A8)
such that

𝑗𝑗 ′ 𝑗𝑗 ′
𝑗𝑗 𝑗𝑗 ∑𝑃𝑃𝑛𝑛 𝑄𝑄𝑛𝑛
𝑄𝑄𝑛𝑛 = 𝜎𝜎𝑛𝑛 𝑗𝑗
(𝐴𝐴. 22),
𝑃𝑃𝑛𝑛
and

𝑗𝑗
𝑗𝑗 𝜎𝜎𝑛𝑛
𝑃𝑃𝑛𝑛𝑜𝑜 = ��𝑃𝑃𝑛𝑛 � (𝐴𝐴. 23),
𝑗𝑗

• in each country and each commodity sector, the representative wholesaler solves the problem
summarized by (A.9)-(A.10) such that

𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗 −𝜃𝜃𝑗𝑗
𝑗𝑗 𝑝𝑝𝑚𝑚 𝜏𝜏𝑚𝑚𝑚𝑚 𝑥𝑥𝑚𝑚𝑚𝑚 �𝜏𝜏𝑚𝑚𝑚𝑚 𝑝𝑝𝑚𝑚𝑚𝑚 �
𝜋𝜋𝑚𝑚𝑚𝑚 ≡ 𝑗𝑗 𝑗𝑗 𝑗𝑗 = −𝜃𝜃𝑘𝑘
(𝐴𝐴. 16),
∑𝑚𝑚′ 𝑝𝑝𝑚𝑚 ′ 𝜏𝜏𝑚𝑚′ 𝑛𝑛 𝑥𝑥𝑚𝑚′ 𝑛𝑛
𝑗𝑗
∑𝑚𝑚′ �𝜏𝜏𝑚𝑚 𝑗𝑗
′ 𝑛𝑛 𝑝𝑝𝑚𝑚′ 𝑛𝑛 �

and

1
−
𝜃𝜃𝑗𝑗
𝑗𝑗 𝑗𝑗 𝑗𝑗 −𝜃𝜃𝑗𝑗
𝑃𝑃𝑛𝑛 = ���𝜏𝜏𝑚𝑚𝑚𝑚 𝑝𝑝𝑚𝑚 � � (A. 17),
𝑚𝑚

• in each country and each commodity sector, the representative producer solves the problem
summarized by (A.11)-(A.12) such that

𝑗𝑗 𝑗𝑗 𝑗𝑗
𝑗𝑗 �1 − 𝛼𝛼𝑛𝑛 �𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛
𝐿𝐿𝑛𝑛 = (𝐴𝐴. 18),
𝑤𝑤𝑛𝑛
𝑗𝑗 𝑗𝑗 𝑗𝑗
𝑗𝑗 𝛼𝛼𝑛𝑛 𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛
𝑀𝑀𝑛𝑛 = (𝐴𝐴. 20),
𝑃𝑃𝑛𝑛
and

𝑗𝑗 𝑗𝑗
𝑗𝑗 𝑗𝑗 −1 𝛼𝛼 1−𝛼𝛼𝑛𝑛
𝑝𝑝𝑛𝑛 = �𝑧𝑧𝑛𝑛 � 𝑃𝑃𝑛𝑛 𝑛𝑛 𝑤𝑤𝑛𝑛 (𝐴𝐴. 21),

INTERNATIONAL MONETARY FUND 33

IMF WORKING PAPERS Fragmentation in Global Trade

• markets clear in all countries and sectors:

𝑗𝑗
𝑄𝑄𝑛𝑛 = 𝐶𝐶𝑛𝑛 + ∑𝑀𝑀𝑛𝑛𝑘𝑘 + ∑𝑀𝑀𝑛𝑛 (𝐴𝐴. 22),
𝑄𝑄𝑛𝑛𝑘𝑘 = 𝑄𝑄�𝑛𝑛𝑘𝑘 (𝐴𝐴. 23),
𝑗𝑗 𝑗𝑗
𝑄𝑄𝑛𝑛 = 𝑄𝑄�𝑛𝑛 (𝐴𝐴. 24),
𝑗𝑗
𝐿𝐿𝑛𝑛 = ∑𝐿𝐿𝑘𝑘𝑛𝑛 + ∑𝐿𝐿𝑛𝑛 (𝐴𝐴. 25),
𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗
∑𝑝𝑝𝑛𝑛𝑘𝑘 𝑞𝑞𝑛𝑛𝑘𝑘 + ∑𝑝𝑝𝑛𝑛 𝑞𝑞𝑛𝑛 + 𝐷𝐷𝑛𝑛 = 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛 + ∑(𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑘𝑘 + 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 ) + ∑(𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛 + 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛 ) (𝐴𝐴. 26).

Equilibrium using exact hat algebra.

A variable with a hat represents the relative change of the variable between the counterfactual and observed
𝑗𝑗 𝑗𝑗 𝑁𝑁
equilibrium. Given changes in trade costs 𝜏𝜏̂𝑚𝑚𝑚𝑚
𝑘𝑘
�𝑛𝑛 and prices �{𝑃𝑃�𝑛𝑛 , 𝑃𝑃�𝑛𝑛𝑘𝑘 �
, 𝜏𝜏̂𝑚𝑚𝑚𝑚 , changes in prices 𝑤𝑤 solve: 𝑛𝑛=1
1
1−𝜂𝜂 1−𝜂𝜂
1−𝛼𝛼𝑘𝑘𝑛𝑛
𝛼𝛼𝑘𝑘 𝑘𝑘
𝑛𝑛 (1−𝜇𝜇𝑛𝑛 )
1−𝜂𝜂 𝑘𝑘 �1−𝜇𝜇𝑘𝑘 �+1−𝛼𝛼𝑘𝑘
𝛼𝛼𝑛𝑛
�𝜁𝜁𝑛𝑛𝑘𝑘 �𝑃𝑃�𝑛𝑛𝑜𝑜 � 𝜁𝜁𝑛𝑛𝑘𝑘 ) ��𝑃𝑃�𝑛𝑛 �𝛼𝛼𝑛𝑛(1−𝜇𝜇𝑛𝑛 )+1−𝛼𝛼𝑛𝑛
𝑘𝑘 𝑘𝑘 𝑘𝑘 𝑛𝑛 𝑛𝑛
• 𝑝𝑝̂𝑛𝑛𝑘𝑘 = + (1 − ⋅ 𝑤𝑤
�𝑛𝑛 � � for all non-commodity sectors,

o where 𝜁𝜁𝑛𝑛𝑘𝑘 is a non-commodity sector’s initial expenditure share on commodities.

𝑗𝑗 𝑗𝑗
𝑗𝑗 1−𝛼𝛼𝑛𝑛 𝛼𝛼
• 𝑝𝑝̂𝑛𝑛 = �𝑛𝑛 �𝑃𝑃�𝑛𝑛 � 𝑛𝑛
𝑤𝑤 for all commodity sectors,
1
𝑘𝑘 −𝜃𝜃𝑘𝑘 ]−𝜃𝜃𝑘𝑘
• 𝑃𝑃�𝑛𝑛𝑘𝑘 𝑘𝑘 (𝜏𝜏̂ 𝑘𝑘
= [∑𝑚𝑚 𝜋𝜋𝑚𝑚𝑚𝑚 𝑚𝑚𝑚𝑚 𝑝𝑝̂ 𝑚𝑚 ) for all non-commodity sectors,
1
−𝜃𝜃 −
𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗 𝑗𝑗 𝜃𝜃
• 𝑃𝑃�𝑛𝑛 = �∑𝑚𝑚 𝜋𝜋𝑚𝑚𝑚𝑚 �𝜏𝜏̂𝑚𝑚𝑚𝑚 𝑝𝑝̂𝑚𝑚 � � 𝑗𝑗 for all commodity sectors,
𝜎𝜎 𝑘𝑘
• 𝑃𝑃�𝑛𝑛 = ∏𝑘𝑘(𝑃𝑃�𝑛𝑛𝑘𝑘 ) 𝑛𝑛
𝑗𝑗 𝜎𝜎𝑗𝑗𝑘𝑘
• 𝑃𝑃�𝑛𝑛𝑜𝑜 = ∏𝑗𝑗(𝑃𝑃�𝑛𝑛 )
𝑘𝑘 ⋅𝑝𝑝�𝑘𝑘 −𝜃𝜃𝑘𝑘
𝜏𝜏�𝑚𝑚𝑚𝑚
• 𝑘𝑘
𝜋𝜋�𝑚𝑚𝑚𝑚 =� 𝑚𝑚
� for all non-commodity sectors,
𝑃𝑃�𝑛𝑛𝑘𝑘
𝑗𝑗 𝑗𝑗 −𝜃𝜃𝑗𝑗
𝑗𝑗 𝜏𝜏�𝑚𝑚𝑚𝑚 ⋅𝑝𝑝�𝑚𝑚
• 𝜋𝜋�𝑚𝑚𝑚𝑚 = � 𝑗𝑗 � for all commodity sectors,
𝑃𝑃�𝑛𝑛
𝐷𝐷𝑛𝑛 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛
• 𝐸𝐸�𝑛𝑛 = + 𝑤𝑤
�𝑛𝑛
𝐸𝐸𝑛𝑛 𝐸𝐸𝑛𝑛
o where 𝐸𝐸𝑛𝑛 ≡ 𝑤𝑤𝑛𝑛 𝐿𝐿𝑛𝑛 + 𝐷𝐷𝑛𝑛 denotes total income

𝑗𝑗 𝑗𝑗
• 𝐸𝐸�𝑛𝑛𝑀𝑀 = ∑𝑘𝑘 𝑚𝑚𝑛𝑛𝑘𝑘 𝑌𝑌�𝑛𝑛𝑘𝑘 + ∑𝑗𝑗 𝑚𝑚𝑛𝑛 𝑌𝑌�𝑛𝑛
o where:
𝑗𝑗
 𝐸𝐸𝑛𝑛𝑀𝑀 ≡ +∑𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑘𝑘 + ∑𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛
𝑘𝑘 𝑗𝑗
𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛 𝑗𝑗 𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛
 𝑚𝑚𝑛𝑛𝑘𝑘 ≡ 𝑘𝑘′ , 𝑚𝑚𝑛𝑛 ≡ 𝑗𝑗′
∑𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛 ∑𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛

𝑋𝑋 𝑘𝑘 𝐸𝐸𝑛𝑛 𝐸𝐸𝑛𝑛
• 𝑌𝑌�𝑚𝑚𝑘𝑘 = ∑𝑛𝑛 𝑚𝑚𝑚𝑚
𝑘𝑘 ⋅ 𝜋𝜋
𝑘𝑘
� 𝑚𝑚𝑚𝑚 ⋅� 𝑀𝑀 ⋅ 𝐸𝐸�𝑛𝑛 + �1 − 𝑀𝑀 � ⋅ 𝐸𝐸�𝑛𝑛𝑀𝑀 �
𝑌𝑌𝑚𝑚 𝐸𝐸𝑛𝑛 +𝐸𝐸𝑛𝑛 𝐸𝐸𝑛𝑛 +𝐸𝐸𝑛𝑛
o Where 𝑌𝑌𝑚𝑚𝐾𝐾 denotes production
𝑃𝑃 � 𝑜𝑜 1−𝜂𝜂
• 𝐸𝐸�𝑛𝑛𝑜𝑜 = ∑𝑘𝑘 𝑜𝑜𝑛𝑛𝑘𝑘 ⋅ � �𝑛𝑛𝑘𝑘 � 𝑌𝑌�𝑛𝑛𝑘𝑘
𝑝𝑝𝑛𝑛
𝑗𝑗
𝑗𝑗 𝑋𝑋𝑚𝑚𝑚𝑚 𝑗𝑗
• 𝑌𝑌�𝑚𝑚 = ∑ 𝑛𝑛 𝑗𝑗 ⋅ 𝜋𝜋�𝑚𝑚𝑚𝑚 ⋅ 𝐸𝐸�𝑛𝑛𝑜𝑜
𝑌𝑌𝑚𝑚
• ∑𝑘𝑘 𝑌𝑌𝑛𝑛𝑘𝑘 𝑌𝑌�𝑛𝑛𝑘𝑘 + ∑𝑗𝑗 𝑌𝑌𝑛𝑛𝑗𝑗 𝑌𝑌�𝑛𝑛𝑗𝑗 + 𝐷𝐷𝑛𝑛 = 𝐸𝐸𝑛𝑛 ⋅ 𝐸𝐸�𝑛𝑛 + ∑(𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑘𝑘 + 𝑃𝑃𝑛𝑛𝑜𝑜 𝑂𝑂𝑛𝑛𝑘𝑘 )𝑌𝑌�𝑛𝑛𝑘𝑘 + ∑𝑃𝑃𝑛𝑛 𝑀𝑀𝑛𝑛𝑗𝑗 𝑌𝑌�𝑛𝑛𝑗𝑗

INTERNATIONAL MONETARY FUND 34

Fragmentation in Global Trade: Accounting for Commodities
Working Paper No. WP/2023/075