You are on page 1of 16

How do countries specialize in agricultural production?

A complex
network analysis of the global agricultural product space

Supplementary Information
Mercedes Campi1,* , Marco Dueñas2 , and Giorgio Fagiolo3
1
CONICET & Universidad de Buenos Aires, Instituto Interdisciplinario de Economı́a Polı́tica de Buenos Aires
(IIEP)
2
Department of Economics, International Trade and Social Policy – Universidad de Bogotá Jorge Tadeo Lozano
3
Istituto di Economia, Scuola Superiore Sant’Anna
*
Corresponding author: mercedes.campi@fce.uba.ar

Country and product data

Table SI.1. List of countries and ISO codes.

Country ISO Country ISO Country ISO


Afghanistan AFG Gabon GAB Norway NOR
Albania ALB Gambia GMB North Macedonia MKD
Algeria DZA Georgia GEO Oman OMN
Angola AGO Germany DEU Pakistan PAK
Antigua and Barbuda ATG Ghana GHA Panama PAN
Argentina ARG Greece GRC Paraguay PRY
Armenia ARM Grenada GRD Peru PER
Australia AUS Guatemala GTM Philippines PHL
Austria AUT Guinea GIN Poland POL
Azerbaijan AZE Guinea-Bissau GNB Portugal PRT
Bahamas BHS Guyana GUY Rep. of Korea KOR
Bangladesh BGD Haiti HTI Rep. of Moldova MDA
Barbados BRB Honduras HND Romania ROU
Belarus BLR Hungary HUN Russian Federation RUS
Belize BLZ Iceland ISL Rwanda RWA
Benin BEN India IND Saint Kitts and Nevis KNA
Bermuda BMU Indonesia IDN Saint Lucia LCA
Bolivia (Plurinational State of) BOL Iran (Islamic Rep. of) IRN Saint Vincent and the Grenadines VCT
Bosnia and Herzegovina BIH Iraq IRQ Samoa WSM
Botswana BWA Ireland IRL Sao Tome and Principe STP
Brazil BRA Israel ISR Saudi Arabia SAU
Brunei Darussalam BRN Italy ITA Senegal SEN
Bulgaria BGR Jamaica JAM Sierra Leone SLE
Burkina Faso BFA Japan JPN Slovakia SVK
Cabo Verde CPV Jordan JOR Slovenia SVN
Cambodia KHM Kazakhstan KAZ Solomon Islands SLB
Cameroon CMR Kenya KEN South Africa ZAF
Canada CAN Kiribati KIR Spain ESP
Central African Rep. CAF Kuwait KWT Sri Lanka LKA
Chad TCD Kyrgyzstan KGZ Suriname SUR
Chile CHL Lao People’s Dem. Rep. LAO Swaziland SWZ
China, Hong Kong SAR HKG Latvia LVA Sweden SWE
China, Macao SAR MAC Lebanon LBN Switzerland CHE

1
China, mainland CHN Lesotho LSO Tajikistan TJK
China, Taiwan Province of TWN Liberia LBR Thailand THA
Colombia COL Lithuania LTU Timor-Leste TLS
Congo COG Madagascar MDG Togo TGO
Costa Rica CRI Malawi MWI Trinidad and Tobago TTO
Côte d’Ivoire CIV Malaysia MYS Tunisia TUN
Croatia HRV Maldives MDV Turkey TUR
Cuba CUB Mali MLI Turkmenistan TKM
Cyprus CYP Malta MLT Uganda UGA
Czechia CZE Mauritania MRT Ukraine UKR
Dem. People’s Rep. of Korea PRK Mauritius MUS United Arab Emirates ARE
Denmark DNK Mexico MEX United Kingdom GBR
Djibouti DJI Mongolia MNG United Rep. of Tanzania TZA
Dominica DMA Morocco MAR United States of America USA
Dominican Rep. DOM Mozambique MOZ Uruguay URY
Ecuador ECU Myanmar MMR Uzbekistan UZB
Egypt EGY Namibia NAM Vanuatu VUT
El Salvador SLV Nepal NPL Venezuela (Bolivarian Rep. of) VEN
Estonia EST Netherlands NLD Viet Nam VNM
Ethiopia ETH New Caledonia NCL Yemen YEM
Fiji FJI New Zealand NZL Zambia ZMB
Finland FIN Nicaragua NIC Zimbabwe ZWE
France FRA Niger NER
French Polynesia PYF Nigeria NGA

2
Table SI.2. List of agricultural products.

Crops
Almonds, with shell; Anise, badian, fennel, coriander; Apples; Apricots; Artichokes; Asparagus; Avocados; Bambara beans;
Bananas; Barley; Broad beans, horse beans, dry; Beans, dry; Beans, green; Berries nes; Blueberries; Brazil nuts, with
shell; Buckwheat; Cabbages and other brassicas; Canary seed; Carobs; Carrots and turnips; Cashewapple; Cashew nuts,
with shell; Cassava; Cassava leaves; Cauliflowers and broccoli; Cereals, nes; Cherries; Cherries, sour; Chestnut; Chick
peas; Chicory roots; Chillies and peppers, green; Chillies and peppers, dry; Cinnamon (canella); Fruit, citrus nes; Cloves;
Cocoa, beans; Coconuts; Coffee, green; Cottonseed; Cow peas, dry; Cranberries; Cucumbers and gherkins; Currants Dates;
Eggplants (aubergines); Figs; Fonio; Fruit, fresh nes; Fruit, pome nes; Fruit, stone nes; Garlic; Ginger; Gooseberries;
Grain, mixed; Grapefruit (inc. pomelos); Grapes; Groundnuts, with shell; Hazelnuts, with shell; Hempseed; Hops; Karite
nuts (sheanuts); Kiwi fruit; Leeks, other alliaceous vegetables; Lemons and limes; Lentils; Lettuce and chicory; Linseed;
Lupins; Maize; Maize, green; Mangoes, mangosteens, guavas; Mate; Melons, other (inc.cantaloupes); Melonseed; Millet;
Mushrooms and truffles; Mustard seed; Nutmeg, mace and cardamoms; Areca nuts; Kola nuts; Nuts, nes; Oats; Oilseeds
nes; Okra; Olives; Onions, dry; Onions, shallots, green; Oranges; Oil palm fruit; Palm kernels; Oil, palm; Papayas; Peaches
and nectarines; Pears; Peas, dry; Peas, green; Pepper (piper spp.); Peppermint; Persimmons; Pigeon peas; Pineapples;
Pistachios; Plantains and others; Plums and sloes; Poppy seed; Potatoes; Sweet potatoes; Pulses, nes; Pumpkins, squash
and gourds; Quinces; Quinoa; Rapeseed; Raspberries; Rice, paddy; Roots and tubers, nes; Rye; Safflower seed; Sesame seed;
Sorghum; Soybeans; Spices, nes; Spinach; Strawberries; String beans; Sugar beet; Sugar cane; Sugar crops, nes; Sunflower
seed; Tangerines, mandarins, clementines, satsumas; Taro (cocoyam); Tea; Tomatoes; Triticale; Fruit, tropical fresh nes;
Tung nuts; Vanilla; Vegetables, fresh nes; Vegetables, leguminous nes; Vetches; Walnuts, with shell; Watermelons; Wheat;
Yams; Yautia (cocoyam)
Crops processed
Beer of barley; Oil, coconut (copra); Cottonseed; Oil, cottonseed; Oil, groundnut; Oil, linseed; Oil, maize; Margarine, short;
Molasses; Oil, olive, virgin; Palm kernels; Oil, palm kernel; Oil, palm; Oil, rapeseed; Oil, safflower; Oil, sesame; Oil, soybean;
Sugar Raw Centrifugal; Oil, sunflower; Wine
Livestock Primary
Meat, ass; Beeswax; Meat, bird nes; Meat, buffalo; Milk, whole fresh buffalo; Meat, other camelids; Milk, whole fresh camel;
Meat, camel; Meat, cattle; Meat, chicken; Meat, duck; Eggs, hen, in shell; Eggs, other bird, in shell; Meat, game; Meat,
goose and guinea fowl; Milk, whole fresh goat; Meat, goat; Honey, natural; Meat, horse; Meat, nes; Milk, whole fresh cow;
Meat, mule; Offals, nes; Meat, pig; Meat, rabbit; Meat, other rodents; Meat, sheep; Milk, whole fresh sheep; Snails, not
sea; Meat, turkey
Livestock Processed
Cheese, buffalo milk; Ghee, of buffalo milk; Butter, cow milk; Butter and Ghee; Cheese (All Kinds); Cheese, skimmed cow
milk; Cheese, whole cow milk; Cream fresh; Ghee, butteroil of cow milk; Cheese of goat milk; Lard; Milk, skimmed cow;
Evaporat & Condensed Milk; Milk, skimmed condensed; Milk, skimmed dried; Milk, skimmed evaporated; Milk, whole
condensed; Milk, whole dried; Milk, whole evaporated; Cheese, sheep milk; Butter and ghee, sheep milk; Skim Milk &
Buttermilk, dry; Whey, condensed; Whey, dry; Yoghurt

3
Projected networks, 1993 vs. 2013
Network projections on product-product and country-country layers show similar architectures for the
period 1993–2013. Tables SI.3 and SI.4 show some network statistics for the Agricultural Product
Space Network (APSN) and the Agricultural Country Space Network (ACSN). These networks are
very dense, but, despite that, they have a very robust modular structure, in particular, after applying
the hypergeometric filter.

Table SI.3. Network statistics of the agricultural product space network (APSN) for selected years.

Year 1993 1998 2003 2008 2013


Full Network
Nodes 218 218 219 219 219
Density 0.76 0.76 0.77 0.77 0.77
Av. link weight 0.25 0.25 0.25 0.25 0.24
Link weight skewness 1.18 1.12 1.17 1.18 1.12
Av. node degree 163.95 165.43 167.38 167.80 168.69
Av. node strength 19.48 19.95 20.05 19.90 19.73
Communities 4 3 3 4 4
Modularity 0.22 0.22 0.22 0.21 0.21
Filtered Network*
Density 0.13 0.13 0.13 0.13 0.12
Communities 4 4 4 4 4
Modularity 0.54 0.55 0.56 0.56 0.55
P
Notes: Link weight: (Pkk
P0 > 0). Node degree: N Dk = k0 Akk0 , with Akk0 = 1 if Pkk0 > 0, and zero otherwise.
Node strength: N Sk = k0 Pkk0 . *Links are validated by the hypergeometric filter at the 1% level of significance.
The APSN features 218 products between 1993 and 1999 (where “Fruit, pome nes” is not reported) and 219 since
2000.

Table SI.4. Network statistics of the agricultural country space network (ACSN) for selected years.

Years 1993 1998 2003 2008 2013


Full Network
Nodes 169 169 169 169 169
Density 0.96 0.97 0.98 0.97 0.98
Average link weight 0.30 0.31 0.31 0.31 0.30
Link weight skewness 0.76 0.72 0.84 0.78 0.70
Average node degree 161.91 163.01 163.91 163.63 164.78
Average node strength 21.46 22.09 22.09 22.53 22.68
Communities 2 2 2 2 2
Modularity 0.21 0.21 0.21 0.20 0.20
Filtered Network*
Density 0.15 0.15 0.15 0.14 0.14
Communities 4 4 5 4 4
Modularity 0.56 0.56 0.56 0.55 0.57
P
Notes: Link weight:
P (Cii0 > 0). Node degree: N Di = i0 Aii0 , with Aii0 = 1 if Cii0 > 0, and zero otherwise. Node
strength: N Si = i0 Cii0 . *Links are validated by the hypergeometric filter at the 1% level of significance.

Figures SI.1 and SI.2 show the network representations in 1993 and 2013, for the APSN and ACSN.
The hypergeometric filter validates links at the 1% level of significance and colors represent the detected
communities with the Louvain algorithm in the corresponding year.

4
Apples Apples
Potatoes Plums Potatoes Plums
Carrots HoneyN Carrots HoneyN
HorsesM GrapesGoatMC HorsesM GrapesGoatMC
Leeks Leeks
Apricots Apricots
CheesA Cucumbers ButterCowM CheesA BerriesN Cherries Cucumbers
ButterCowM BerriesN Cherries Quinces
PigsM EggshenS Quinces OnionsD PigsM EggshenS OnionsD
Wheat Wheat
MilkCowWF Peaches MilkCowWF Peaches
MilkCowS Strawberries Figs SheepMWF MilkCowS Strawberries Figs SheepMWF
OliveO ButterGhee Pears OliveO
ButterGhee Pears
CowCheeseWM Barley WalnutsS CowCheeseWM Barley WalnutsS Cabbages
CowCheeseSM Cabbages CowCheeseSM AlmondsS
AlmondsS
SugarB Vetches SheepMC Watermelons SugarB Wine Vetches SheepMC Watermelons
Triticale Wine LettuceChicory Triticale LettuceChicory
CreamF CauliflowersBroccoli CreamF CauliflowersBroccoli SheepM
Oats TurkeysM SheepM Oats TurkeysM
CherriesS Carobs Olives Melons CherriesS Carobs Olives Melons
RapeseedO GrainM Raspberries Mushrooms Kiwi Tomatoes RapeseedO GrainM Raspberries Mushrooms Kiwi Tomatoes
HazelnutsS FruitSN VegetablesL HazelnutsS FruitSN VegetablesL
Rye ChilliesPeppers Rye ChilliesPeppers
LinseedO LinseedO
SkimMButtermilkD FruitPN Artichokes ShhepMBG Tangerines SkimMButtermilkD FruitPN Artichokes ShhepMBG Tangerines
Rapeseed Anise OnionsG Rapeseed PeasD Anise OnionsG GoatMWF
WheyD PeasD SunflowerO GoatMWF WheyD SunflowerO
BuffaloCheeseM CattleM Dates BuffaloCheeseM CattleM Dates
Lard Lard
PoppyS Cranberries Eggplants CamelsM CerealNes PoppyS Cranberries Persimmons Eggplants CamelsM CerealNes
Currants Persimmons Currants
MilkSD WheyC SoybeanO MilkSD WheyC SoybeanO
Chestnut Pistachios BeansG Chestnut Pistachios BeansG CottonseedO
RabbitM CottonseedO MilkWD RabbitM
MilkWD
Gooseberries SunflowerS Spinach CamelMWF Gooseberries SunflowerS Spinach Snails CamelMWF
Snails MargarineS
MargarineS BeansB Cottonseed BeansB Cottonseed
Buckwheat Yoghurt Pumpkins Buckwheat Yoghurt Pumpkins
MilkEC BeerB SafflowerS Garlic MilkEC BeerB SafflowerS Garlic
MustardS MustardS
Hops Hempseed Peppermint MelonseedOkra Hops Hempseed Peppermint MelonseedOkra
Chickpeas AssesM Bambara MilkWE Chickpeas AssesM Bambara
MilkWE ChickensM Linseed ChickensM
Linseed SesameSO Beeswax SesameSO Beeswax
MilkSE BirdMN NutsN MilkSE BirdMN NutsN
MulesM GoatsM MulesM GoatsM
Blueberries MilkSC Lentils SesameS Blueberries MilkSC Lentils SesameS
PeasG CassavaL Karitenuts PeasG CassavaL Karitenuts
MilkWC GeeseGuineaFM Cashewapple MilkWC GeeseGuineaFM Cashewapple
CanaryS Millet CanaryS Millet
Lupins Tungnuts Maize Sorghum Lupins Tungnuts Maize Sorghum
ChicoryR Asparagus Lemons CowpeawsD GameM ChicoryR Asparagus Lemons CowpeawsD GameM
GheeButteroilCM ChilliesPeppersD GheeButteroilCM ChilliesPeppersD
BrasilnutsS Fonio BrasilnutsS Fonio
SafflowerO Tea BeansD GroundnutsS SafflowerO Tea BeansD GroundnutsS
MaizeO Soybeans OffalsN MaizeO Soybeans OffalsN
BuffaloWFM Oranges CashewnutsS BuffaloWFM Oranges CashewnutsS
EggsOBS PotatoesS EggsOBS PotatoesS
BuffaloesBM BuffaloesM GroundnutO NutsK BuffaloesBM BuffaloesM GroundnutO NutsK
Mate StringB Grapefruit Cassava Mate StringB Grapefruit Cassava
Vanilla Ginger PalmK Vanilla Ginger PalmK
RodentsM MaizeGr Cinnamon PigeonP FruitFN PalmfruitOPulsesN RodentsM MaizeGr Cinnamon PigeonP FruitFN PalmfruitOPulsesN
OilseedsN OilseedsN
DucksM RiceP Taro DucksM RiceP Taro
CamelidsM NutsA CoffeeG Yams CamelidsM NutsA CoffeeG Yams
SugarRC Cloves Mangoes SugarRC Cloves Mangoes
Quinoa SugarC1 PalmO Quinoa SugarC1 PalmO
SugarC2 CocoaB1 PalmKO SugarC2 CocoaB1 PalmKO
Yautia Plantains VegetablesFN Yautia Plantains VegetablesFN
Nutmeg Nutmeg
BuffaloM Pepper Bananas BuffaloM Pepper Bananas
MeatN TropicalFruitFN MeatN TropicalFruitFN
Molasses Pineapples Molasses Pineapples
SpicesN Coconuts SpicesN Coconuts
RootTuberN RootTuberN
Papayas CitrusFN Papayas CitrusFN
Avocados CoconutO Avocados CoconutO

Figure SI.1. The Agricultural Product Space Network (APSN). Left: 1993 and right: 2013. Communities:
in blue “Crops and livestock”, in green “Vegetables and fruits”, in purple “Tropical fruits and crops”, and in
orange “Special livestock, oils and crops”. Nodes’ positions are fixed by the 2013 representation in order to
facilitate the comparison.

HND HND
MEX MEX
AUS PER GTM AUS PER GTM
VEN VEN
DNK PAN PAN
NOR NZL DNK NZL
BRB NOR
LVA FIN BOL DOM SLV BRB
DOM SLV
CRI LVA FIN BOL CRI
LTU GBR ISL ZAF JAM GBR ZAF JAM
CUB ECU LTU ISL CUB ECU
CZEEST DEU TTO COL BRA EST DEU TTO COL
SWE USA CZE BRA
CAN URY FJI MYS CAN SWE USA
SVK URY FJI MYS
GUY SVK
BLZ GUY BLZ
AUT FRA POL CHE BLR ARG NCL LCA SUR NIC AUT FRA POL CHE BLR ARG NCL SUR NIC
VCT LCA VCT
IRL PRY PRY
HUN PYF WSM IRL
RUS DMA HUN RUS PYF DMA WSM
SVN NLD MAC GRD THA MAC GRD
UKR MUS SVN NLD MUS
THA
SWZ UKR
HRV KIR HRV SWZ
PHL IDN CIV COG KIR PHL CIV COG
PRT CHL CPV BHS STP CPV BHS IDN
MDA JPN KNA PRT CHL STP
BGR ROU VNM MDG MDA JPN KNA MDG
HTI BGR ROU VNM HTI
BRN VUT GAB ZWE VUT ZWE
KAZ MDV BRN GAB
SLB KAZ MDV
BIH SLB
BMU HKG LBR GHA UGA BIH LBR GHA UGA
ITA ZMB ITA BMU HKG ZMB
LKA KHM LKA KHM
MKD GEO CAF TLS GEO
ESP AZE ISR TWN MKD ISR CAF TLS
ATG LAO GIN SLE ESP AZE TWN
LAO GIN SLE
GRC GRC ATG
BGD MOZ BGD MOZ
KOR KOR
ARM LBN MWI
KGZ TUR BWA GNB TGO NGA KGZ TUR ARM LBN BWA GNB MWI
TGO NGA
MLT DJI LSO GMB CMR MLT DJI LSO CMR
ALB GMB
ALB KWT
UZB MNG KWT AGO UZB MNG AGO
MAR RWA BEN MAR RWA BEN
TKM NAM TCD TKM TCD
CYP CYP NAM
DZA JOR PRK JOR PRK
YEM MMR TZA DZA YEM MMR
IRN BFA TZA
IRN BFA
TJK AFG ARE MRT IND MLI TJK MRT
TUN TUN AFG ARE IND MLI
SEN SEN
IRQ OMN OMN
CHN NPL IRQ CHN NPL
NER KEN KEN
EGY SAU NER
EGY SAU
ETH ETH
PAK PAK

Figure SI.2. The Agricultural Country Space Network (ACSN) and choroplet maps. Left: 1993 and right:
2013. Communities: in red “Tropical I”, in green “Tropical II”, in yellow “Subtropical”, and in blue “Temperate”.
Nodes’ positions are fixed by the 2013 representation in order to facilitate the comparison.

Both network projections link weight distributions are strongly right-skewed: very few products
have a high relatedness, and most of them are weakly related. Fig. SI.3 shows that these link-weight
distributions scale exponentially, quicker than a log-normal, and are best proxied by either a Gamma
or a Weibull density.

5
Figure SI.3. Link weight distributions in the network projections: APSN and ACSN (excluding zeros). Upper
panel: Empirical distributions of product relatedness links (left) and country relatedness links (right), in 1993
and 2013. Lower panel: Fitted distributions of product relatedness links (left) and country relatedness links
(right) in 2013.

Table SI.5 shows statistics of the ACSN that evidence that the interaction within countries of the
communities is much higher than interactions with countries outside the communities. This evidences
the modular structure of the ACSN.
Table SI.5. Within and between communities node degree and node strength in the ACSN. Totals and shares
in 2013.

Node degree
Total Share
Community Subtropical Temperate Tropical I Tropical II Subtropical Temperate Tropical I Tropical II
Subtropical 608 120 32 32 0.77 0.15 0.04 0.04
Temperate 120 968 4 3 0.11 0.88 0.00 0.00
Tropical I 32 4 1026 216 0.03 0.00 0.80 0.17
Tropical II 32 3 216 612 0.04 0.00 0.25 0.71
Node strength
Total Share
Community Subtropical Temperate Tropical I Tropical II Subtropical Temperate Tropical I Tropical II
Subtropical 206.70 37.49 7.45 7.84 0.80 0.14 0.03 0.03
Temperate 37.49 360.06 0.95 0.62 0.09 0.90 0.00 0.00
Tropical I 7.45 0.95 268.64 50.95 0.02 0.00 0.82 0.16
Tropical II 7.84 0.62 50.95 181.49 0.03 0.00 0.21 0.75

The robustness of the network architecture is also reflected in the composition of the communities in
different years. The detected communities in the APSN/ACSN maintain most of the products/countries

6
in different years, as can be observed in Fig. SI.4.
Potatoes NAM
MilkCowWF MRT
CattleM
BeerB LSO
WheyD BWA
WheyC NER
TurkeysM ZMB
Triticale
SugarB UGA
Strawberries TZA
SkimMButtermilkD TGO
Rye
Raspberries TCD
RapeseedO SLE
Rapeseed SEN
PoppyS
PigsM RWA
PeasD NGA
Oats MOZ
MustardS MLI
Mushrooms
MilkWE LBR
MilkWD KEN
MilkWC HTI
MilkSE
MilkSD GNB
MilkEC GMB
MilkCowS GIN
MargarineS
Lard GHA
HorsesM ETH
Hops CMR
GrainM CAF
Gooseberries
Currants BFA
CreamF BEN
CowCheeseWM AGO
CowCheeseSM
ChicoryR ZWE
CheesA MWI
ButterGhee MMR
ButterCowM GAB
Blueberries
Barley COG
Carrots CIV
Wheat NPL
Leeks
CherriesS MDG
Buckwheat LAO
HoneyN IND
EggshenS
BerriesN DJI
Cranberries TLS
OffalsN VNM
Lupins MNG
LinseedO
Linseed PAK
MilkSC STP
Wine SLB
Cabbages
SunflowerS BGD
SunflowerO IDN
RabbitM THA
WalnutsS LKA
Plums
Cherries KHM
Apples SWZ
Watermelons MAC
Vetches
VegetablesL HKG
Tomatoes CPV
Tangerines BRN
Snails
ShhepMBG BHS
SheepMWF ATG
SheepMC PHL
SheepM NCL
Quinces
Pistachios VUT
Pears PRY
Peaches MDV
OnionsG
OnionsD KIR
Olives WSM
OliveO VEN
Melons
LettuceChicory VCT
Kiwi TTO
HazelnutsS SUR
Grapes SLV
GoatMC
FruitSN PYF
Figs PER
Eggplants PAN
Dates
Cucumbers NIC
ChilliesPeppers MYS
Chestnut MUS
CauliflowersBroccoli MEX
Carobs
CamelsM LCA
CamelMWF KNA
Artichokes JAM
Apricots
AlmondsS HND
BeansB
Countries

GUY
Products

Spinach GTM
Pumpkins
Persimmons GRD
GoatMWF FJI
BeansG ECU
Garlic DOM
SesameSO
Anise DMA
BuffaloCheeseM CUB
BirdMN CRI
CanaryS
PeasG COL
Hempseed BRB
MaizeO BRA
FruitPN BOL
SoybeanO
Chickpeas BMU
MulesM BLZ
Lentils TWN
BuffaloesBM
Asparagus ZAF
StringB OMN
Peppermint KWT
OilseedsN
GeeseGuineaFM ARE
Cereales PRK
SafflowerS SAU
SafflowerO KOR
MaizeGr
BuffaloWFM YEM
BuffaloM UZB
Tungnuts TUR
Soybeans
RodentsM TUN
Quinoa TKM
Mate TJK
CamelidsM
BrasilnutsS MLT
Yoghurt MKD
GheeButteroilCM MAR
DucksM LBN
EggsOBS
Maize JOR
SugarC2 ITA
NutsA ISR
ChickensM
Vanilla IRQ
BuffaloesM IRN
Tea GRC
MeatN GEO
Cashewapple
Okra ESP
RiceP EGY
CottonseedO DZA
Cottonseed
Cloves CYP
Cinnamon CHN
Yautia AZE
Yams
VegetablesFN ARM
TropicalFruitFN ALB
Taro AFG
SugarRC KGZ
SugarC1
SpicesN ROU
Sorghum BIH
SesameS KAZ
RootTuberN
PulsesN BGR
PotatoesS MDA
Plantains ARG
Pineapples CHL
PigeonP
Pepper PRT
Papayas HRV
PalmO USA
PalmKO
PalmK URY
PalmfruitO UKR
Oranges SWE
NutsN
NutsK SVN
Nutmeg SVK
Molasses RUS
Millet POL
Melonseed
Mangoes NZL
Lemons NOR
Karitenuts NLD
GroundnutsS
GroundnutO LVA
Grapefruit LTU
GoatsM JPN
Ginger
GameM ISL
FruitFN IRL
Fonio HUN
CowpeawsD GBR
CoffeeG
Coconuts FRA
CoconutO FIN
CocoaB1 EST
CitrusFN
ChilliesPeppersD DNK
CassavaL DEU
Cassava CZE
CashewnutsS CHE
Beeswax
BeansD CAN
Bananas BLR
Bambara AUT
Avocados
AssesM AUS
1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013 1993 1995 1997 1999 2001 2003 2005 2007 2009 2011 2013

Figure SI.4. Composition of the communities in the APSN (left) and in the ACSN (right). 1993 to 2013.
Colors represent communities as in the networks of Figures 1 and 2. White points appear when there is an
additional community detected.

7
Community attaching

Table SI.6. Variables used in the logit estimations and sources.

Name Description Source


Geographical distance Geographical distance in km. between two countries: BACI-CEPII*
ln Distanceij
Diff. in latitudes Distance differences in country latitudes. It proxies BACI-CEPII*
differences in climate and agro-ecological zones:
ln ||lati | − |latj ||
Same region Dummy that indicates if countries belong to the same WDI**
geographical region: East Asia & Pacific, Europe &
Central Asia, Latin America & Caribbean, Middle
East & North Africa, North America, South Asia,
and Sub-Saharan Africa
Diff. in GDP per capita Difference between the GDP per capita of a pair of Penn World Tables
countries: | ln GDP pci − ln GDP pcj | 9.0: [1]
Diff. in human capital Difference in the index of human capital (average Penn World Tables
years of schooling and the returns to education) of a 9.0: [1]
pair of countries: |hci − hcj |
Diff. in political system Difference in the level of democracy of a pair of Systemic Peace:
countries: |polityi − polityj | Polity IV***
Diff. in agricultural labor Difference between the number of economically active ERS-USDA****
adults in agriculture over agricultural land of a pair
of countries: | ln Labori − ln Laborj |
Diff. in agricultural machinery Difference between the total stock of farm machinery ERS-USDA****
over agricultural land of a pair of countries:
| ln M achinei − ln M achinej |
Diff. in fertilizers consumption Difference in fertilizers consumption (in metric ERS-USDA****
tonnes of N, P2O5, and K2O) over agricultural land
of a pair of countries: | ln F ertii − ln F ertij |
Diff. in irrigated land Difference in the total area equipped for irrigation ERS-USDA****
over agricultural land of a pair of countries:
| ln Irrigi − ln Irrigj |
Notes: *www.cepii.fr/CEPII/en/bdd_modele/presentation.asp?id=1, **databank.worldbank.org/source/
world-development-indicators, ***www.systemicpeace.org/inscrdata.html, ****www.ers.usda.gov/
data-products/international-agricultural-productivity (see, [2], for details on the methodology.)

8
Table SI.7. Summary statistics of the variables included in the logit estimations (2013).

Variable Observations Mean Std. Dev. Min Max


Distance (ln) 14,196 8.77 0.78 4.09 9.90
Diff. latitudes (ln) 14,196 2.54 1.12 -5.70 4.16
Same region 14,196 0.19 0.39 0.00 1.00
Diff. in GDP pc 12,403 1.35 0.99 0.00 5.39
Diff. in human capital 8,911 0.79 0.56 0.00 2.54
Diff. in political systems 10,153 0.32 0.27 0.00 1.00
Diff. in agricultural labor 11,935 1.54 1.15 0.00 7.14
Diff. in agricultural machinery 11,935 2.39 1.86 0.00 10.09
Diff. in fertilizers consumption 11,935 2.01 1.59 0.00 7.26
Diff. in irrigated land 11,325 1.79 1.41 0.00 7.58

Table SI.8. Determinants of community attaching. Logit estimations for 1993, 2003 and 2013.

Logit Estimations

Variables 1993 2003 2013


Distance (ln) -0.293*** -0.526*** -0.479***
(0.079) (0.081) (0.075)
Diff. in latitudes (ln) -0.632*** -0.542*** -0.577***
(0.039) (0.039) (0.037)
Same region 1.498*** 1.120*** 1.022***
(0.132) (0.138) (0.126)
Diff. in GDP pc (ln) -0.603*** -1.148*** -0.543***
(0.066) (0.069) (0.065)
Diff. in human capital -1.105*** -0.799*** -0.876***
(0.115) (0.106) (0.106)
Diff. in political systems -0.399** -0.179 -0.989***
(0.157) (0.190) (0.200)
Diff. in agricultural labor (ln) -0.627*** -0.536*** -0.605***
(0.059) (0.058) (0.049)
Diff. in agricultural machinery (ln) -0.288*** -0.234*** -0.011
(0.044) (0.043) (0.037)
Diff. in fertilizers consumption (ln) -0.081** 0.043 -0.163***
(0.041) (0.035) (0.034)
Diff. in irrigated land (ln) -0.720*** -0.332*** -0.500***
(0.042) (0.042) (0.039)
Constant 5.710*** 7.738*** 7.115***
(0.834) (0.867) (0.817)
Observations 7,352 7,405 7,599
Pseudo R2 0.442 0.449 0.400
Notes: The dependent variable is a dummy indicating whether two countries
belong to the same detected community. Estimations include country fixed
effects. Robust standard errors are in parentheses. Significance level: ***
p<0.01, ** p<0.05, * p<0.10.

9
Countries production shares in the communities of the APSN
Our analysis shows that countries concentrate their production on those products that require
capabilities and environmental conditions that they possess. Although countries have capabilities to
produce many products with revealed comparative advantages, the production baskets measured in
Kcal, proteins, and fats are uneven and often concentrated in some products.
The Supplementary file “SF.Production measures” provides yearly information on country total
production and their production shares in each detected product community (measured in Kcal,
proteins, and fats).

Figure SI.5. Countries production shares in Kcal in each community. 1993 and 2013. Colors represent
communities as in the networks of Figure 1. Color intensity represents the share of a country’s total production
in the production of the community.

10
Figure SI.6. Production shares of countries in proteins in each community. 1993 and 2013. Colors represent
communities as in the networks of Figure 1. Color intensity represents the share of a country’s total production
in the production of the community.

11
Figure SI.7. Production shares of countries in fats in each community. 1993 and 2013. Colors represent
communities as in the networks of Figure 1. Color intensity represents the share of a country’s total production
in the production of the community.

12
Diversification of production baskets
The diversification of a production basket can be evaluated by its variety in terms of products, and in
terms of products that reveal a comparative advantage. It is also interesting to consider that even if
production baskets are diversified, they can be concentrated in a relatively low number of products.
The left panel of Figure SI.8 shows the correlation between the number of products that reveal a
comparative advantage and the Gini index for 2013. This measures how much inequality there is in
the distribution of total production among different products of the production basket, revealing that
most countries have high inequality in the distribution of agricultural production.
The right panel of Figure SI.8 shows the diversification/specialization of production baskets
–considering the number of products that reveal a comparative advantage– and the concentration of
production, using the normalized Herfindahl-Hirschman Index for 2013.
An interesting point to notice is that mainly countries in the Temperate and Subtropical communities
have a higher number of products in their production baskets. Conversely, countries in the Tropical
communities are usually less diversified in the variety of products, but their production can have high
and low relative levels of concentration.
We show that some countries that specialize in producing a low number of products can have
high levels of concentration, although, in general, these are small countries in terms of agricultural
production, such as Vanuatu, Macao, or Suriname. Fewer countries are producing a high number of
products with high concentration. Instead, we observe that there are many countries with both a
low and a high number of products and relatively low concentration levels. Some countries that are
very diversified in their production baskets and have a relatively low production concentration are, for
example, Turkey, Portugal, and Spain. Countries with a lower number of products in their production
baskets and relatively low concentration are, in general, smaller countries in terms of agricultural
production.

1.0 BGD Community


Subtropical
VUT
0.6 Temperate
Normalized Herfindahl−Hirschman index

MDG BGR
MYS MUS BGD
PRY GUY
ARG THA
USA SUR Tropical I
PHL IDN CAN HUN
KHM SUR IRQ MAC
PYF
ARE BRA LKA ZAF KAZ SVK UKR WSM Tropical II
SWZ AFG
FRA
HKG LAO VNM SLVUZB KGZ ROU HRV MEX ARE
0.9 MAR
NZL PAK AUS
CZE
VUT WSM URY LVA BOL KHM
ESP CHN AFG
FJI BIH LTU RUS
AGO HND MDA NER PYF
ECU DNK MUS
AZE JPN EGY
Gini index

BLZ CRI NPL GTM KEN 0.4 LKA


TLS GHA ETH TUR HKG VNM
SLE ESTGEOMWIZWE MLI DEU IND GRC KIR
ZMB TWN GBR AGO COG KAZ
BRN RWA BFAMMR FIN SVN KOR ITA TUN LAO OMN IRQ
TKM IRL NIC DZA SWZ TLS MDG
TJK CUBCOL SEN POL PER IRN GUY MMR NPL BGR USA
ALB
MNG BRB PRK CHE TKM BRN MNG
COG VEN SLV UZB ZAF KGZ
BEN TGO NOR CYP TZA AUT PRT MWI
MAC CIV MYS PRY ZMB AZE
KNA MKD LCA LVA BIH ROU IDN
0.8 TCD GIN ARM NGA SWE CMR LSO SLE HUN MAR
NCL BWA KNA GEO HRV LTU
DJI KWT LCA CHL MDA KOR
OMN PAN MLT JAM MDV PRK FJI GBR
SLB SEN TWN FRA
PHL ARG NER
MDV
NAM LBR TTO UGA MOZ BLR 0.2 NCL GHA
TJK SVK
JPN MEX CHN
HTI DOM NLD
Community STP
BLZ
BRA URY THA AUSCAN DZA
UKR IRN
GMB SLB VCT FIN RUS EGY
KIR STP BWA YEM ISR NZL EST BOL ALB
GNB MRT CUB CZE IND
LBN Subtropical ATG GNB BFATGO DNK ZWE DEU GRC
GAB BRB GTM KEN TUR
CAF DJI RWA ARM MKD ITA TUN
LSO
MRT DMA
GRD Temperate NAM GMB LBR BEN
UGA MOZ SVN ETH ESP
BMU TTO GIN AUT
JOR IRL BHS TCD HND NGA VEN TZA PRT
SAU PAK SWE
Tropical I GAB
KWT NIC ISL CAF
PAN CIV MLT CHE YEM POL
PER
0.7 CRI SAU DOM NOR BLR JAM CYP CHL
VCT Tropical II DMA ISR
0.0 CPV HTI GRD ECU COL JOR NLD CMR MLI LBN

0 20 40 60 80 0 20 40 60 80
Number of products (RCA ≥ 1) Number of products (RCA ≥ 1)

Figure SI.8. Correlations between the number of products that reveal a comparative advantage in countries
production baskets and the Gini index (left), and the normalized Herfindahl-Hirschman index (right) in 2013.

We must note that Kcal is not the only measure of interest in food production and, of course,
products are not perfect substitutes. However, this evidence and that presented in the maps allow us
to see that countries tend to exploit intensively only a limited set of products.

13
Robustness of agricultural specialization patterns
Our analysis suggests that specialization patterns of countries in agricultural production are quite
stable. In general, mainly due to changes in the relative prices of products, climatic shocks, or demand
changes, the specialization process is due to the relocation of production factors. To complement our
analysis, we analyze to what extent countries have changed their production baskets between 1993 and
2013. First, we focus on the set of produced goods, which reveal a comparative advantage. Then, we
focus on changes in the vector of the total production of Kcal.
To detect changes in the production basket, we use the Jaccard index, similarly to how we measure
the similarity between countries. Thus, let Ωti be the production basket set of country i in year t, and
let us restrict this set to all products with strictly positive production or only to those products that
reveal a comparative advantage. Thus, the former is defined as Ω̄ti and contains all those k-products
such that RCAtik > 0, i.e., all produced products, and the latter is defined as Ω̃ti , and contains all those
k-products such that RCAtik ≥ 1. Thus, the similarity index between the production basket sets of
two different periods, t and t0 , is defined as the size of the intersection divided by the size of these sets’
union. Using set notation, it is
0
0 |Ωt ∩ Ωti |
∆tt
i = it 0 . (SI.1)
|Ωi ∪ Ωti |

We can use the above-defined production basket sets (Ω̄ and Ω̃) to derive different similarity measures.
The idea of using these two different baskets is to detect to what extent countries have changed their
specialization patterns of agricultural production. Clearly, the set in which we consider only those
products with comparative advantages greater than one is much more susceptible to changes, since it
depends on both internal and external production changes.
Assessing changes in the production patterns might be less straightforward when considering
production volumes. To address this, we measure the angular similarity of the production vector
between 1993 and 2013, which can be done by using the property of the Euclidean inner product of two
vectors, which equals the cosine of the angle between the vectors times the product of the Euclidean
norms of the vectors. More precisely, in our case, let Qti be the vector of production of country i in
year t, production values are transformed from tonnes into Kcal. Thus the cosine of the angle between
the production vectors of country i in t and t0 is:
0
0 Qti · Qti
cos(θitt ) = 0 . (SI.2)
kQti kkQti k

Therefore, the angular distance (AD) and the angular similarity (AS) are defined as:
0
0 θitt
ADtt
i = ; (SI.3)
π
0 0
AStt
i = 1 − ADtt
i . (SI.4)

These measures range between zero and one. However, since Kcal’s production is strictly positive,
the angle θ is constrained to the interval (0, π/2). In other words, the most extreme change occurs
between two orthogonal production baskets. Therefore, in the angular similarity, values close to 0.5
imply dramatic changes, while values close to 1 imply high similarity. For the angular distance case,
the extreme values will be 0 and 0.5, with 0 indicating a null change in the production vector, and 0.5
indicating a maximum change in the production vector.

14
The left panel of Figure SI.9 shows the histogram of country production basket similarity (see: Eq.
(SI.1)) for 1993 versus 2013. When we consider the non-restricted basket set Ω̄, most of the probability
mass is located towards 1 (see blue bars), which implies that typically all those products that were
produced in 1993 are also produced in 2013. Similarly, in the restricted basket set Ω̃, the probability
mass is concentrated for high values of similarity (around 0.7), which means that the sets of products
with comparative advantages are very similar in 1993 and 2013.
The right panel of Figure SI.9 shows the histogram of country production baskets angular similarity
(see: Eq. (SI.4)) for 1993 versus 2013. We confirm that the angular displacements of country production
vectors are typically very small. The histogram shows that the probability mass of the angular similarity
is concentrated close to 1.

Basket restricted to RCA > 0


Basket restricted to RCA ≥ 1 6
7.5

4
5.0
Density

Density

2.5 2

0.0 0
0.00 0.25 0.50 0.75 1.00 0.5 0.6 0.7 0.8 0.9 1.0
Country basket similarity index 1993 vs. 2013 Angular similarity of production baskets, 1993 vs 2013

Figure SI.9. Histograms of country basket similarities between 1993 and 2013. Left: Similarity of the production
baskets. Right: Angular similarity of the vector of production.

In a nutshell, our results indicate that while there were variations between 1993 and 2013, in most
cases, country production patterns are very similar in both years. At the country level, changes in
production baskets can be because countries introduced or stopped producing some products, or shifted
production factors from one sector to others to concentrate production. To show graphically this
evidence, Figure SI.10 shows the map of the angular distances (see: Eq. (SI.3)). Besides, Table SI.9
reports the 20 countries that changed the most between 1993 and 2013. According to the angular
similarity and the angular distance, most countries with relatively greater changes are small countries,
usually, islands like Saint Kitts and Nevis, Saint Lucia, and Saint Vincent and the Grenadines. It is
also interesting to see that a group of countries that were part of the former Soviet Union, such as
Kazakhstan, Uzbekistan, Turkmenistan, Kyrgyzstan, Tajikistan, Lithuania, Latvia, Ukraine, Belarus,
and Bulgaria, are between those that show the relatively greater changes. Finally, some developing
countries, such as Paraguay, Cambodia, Thailand, and Cuba, also display some greater variations. One
notable case is Uruguay, which already had a relatively developed agricultural system in 1993 and
comparative advantages for agricultural production, but appears among the countries with relatively
large changes in their production baskets between 1993 and 2013.

15
Angular distance
0.5
0.4
0.3
0.2
0.1
0.0

Figure SI.10. Angular distance of country production vectors between 1993 and 2013.

Table SI.9. Countries with highest changes in their production baskets between 1993 and 2013 (top 20)
according to the similarity production basket set and the angular distance production vectors.

Similarity production basket set (RCA ≥ 1) Angular distance production vectors


Country ISO Value Country ISO Value
Paraguay PRY 0.38 Saint Kitts and Nevis KNA 0.49
Kazakhstan KAZ 0.39 United Arab Emirates ARE 0.45
Saint Kitts and Nevis KNA 0.42 Trinidad and Tobago TTO 0.44
Guyana GUY 0.44 Maldives MDV 0.39
Cambodia KHM 0.48 Saint Lucia LCA 0.34
United Arab Emirates ARE 0.49 Kuwait KWT 0.32
Uruguay URY 0.49 Uruguay URY 0.31
Croatia HRV 0.49 Latvia LVA 0.29
Bosnia and Herzegovina BIH 0.50 Uzbekistan UZB 0.29
Bahamas BHS 0.50 Belarus BLR 0.28
Thailand THA 0.50 Tajikistan TJK 0.28
Lithuania LTU 0.52 Saudi Arabia SAU 0.27
Cuba CUB 0.53 Ukraine UKR 0.27
Saint Lucia LCA 0.53 Lithuania LTU 0.25
Latvia LVA 0.53 Turkmenistan TKM 0.25
Kuwait KWT 0.53 Cyprus CYP 0.25
Bulgaria BGR 0.53 Saint Vincent and the Grenadines VCT 0.24
Suriname SUR 0.54 Uganda UGA 0.24
Uzbekistan UZB 0.56 Indonesia IDN 0.24
Kyrgyzstan KGZ 0.56 Sierra Leone SLE 0.23

References
[1] Feenstra RC, Inklaar R, Timmer MP. The Next Generation of the Penn World Table. American Economic Review.
2015;105(10):3150–3182.
[2] Fuglie KO. Productivity Growth and Technology Capital in the Global Agricultural Economy. In: Fuglie K, Wang SL,
Ball VE, editors. Productivity Growth in Agriculture: An International Perspective. Wallingford: CAB International;
2012. p. 335–368.

16

You might also like