Package data

I’ll use this demo dataset to illustrate the essential usage of the functions within this package.

library(economiccomplexity)

ec_trade_1962

## # A tibble: 44,435 x 3
## # Groups:   reporter_iso [158]
##    country product   value
##    <chr>   <chr>     <dbl>
##  1 afg     0106        621
##  2 afg     0405       3380
##  3 afg     0407      60807
##  4 afg     0501    1466840
##  5 afg     0504    1466840
##  6 afg     0703      85258
##  7 afg     0712       9057
##  8 afg     0714     164694
##  9 afg     0801    8482381
## 10 afg     0802    8437060
## # … with 44,425 more rows

Revealed comparative Advantage (RCA)

You can obtain RCA with ec_rca().

rca <- ec_rca(
  data = ec_trade_1962,
  c = "country",
  p = "product",
  v = "value"
)

# 5x5 preview
rca[1:5,1:3]

## 5 x 3 sparse Matrix of class "dgCMatrix"
##     0101 0102 0103
## afg    .    .    .
## ago    .    .    .
## alb    .    .    .
## and    .    .    .
## ant    .    .    .

And also you can obtain it in tabular version.

rca_tbl <- ec_rca(
  data = ec_trade_1962,
  c = "country",
  p = "product",
  v = "value",
  tbl = T
)

rca_tbl

## # A tibble: 44,435 x 3
##    country product value
##    <chr>   <chr>   <dbl>
##  1 afg     0106        0
##  2 afg     0405        0
##  3 afg     0407        0
##  4 afg     0501        1
##  5 afg     0504        1
##  6 afg     0703        0
##  7 afg     0712        0
##  8 afg     0714        1
##  9 afg     0801        1
## 10 afg     0802        1
## # … with 44,425 more rows

Another possibility, not used to build networks from bipartite relations, is to obtain RCA as a matrix or tibble without discretization.

rca_decimal <- ec_rca(
  data = ec_trade_1962,
  c = "country",
  p = "product",
  v = "value",
  discrete = F
)

# 5x3 preview
rca_decimal[1:5,1:3]

## 5 x 3 sparse Matrix of class "dgCMatrix"
##     0101       0102 0103
## afg    . .             .
## ago    . 0.01826164    .
## alb    . .             .
## and    . .             .
## ant    . .             .

rca_decimal_tbl <- ec_rca(
  data = ec_trade_1962,
  c = "country",
  p = "product",
  v = "value",
  tbl = T,
  discrete = F
)

rca_decimal_tbl

## # A tibble: 44,435 x 3
##    country product    value
##    <chr>   <chr>      <dbl>
##  1 afg     0106     0.0412 
##  2 afg     0405     0.00944
##  3 afg     0407     0.275  
##  4 afg     0501    14.4    
##  5 afg     0504    20.7    
##  6 afg     0703     0.271  
##  7 afg     0712     0.220  
##  8 afg     0714     3.18   
##  9 afg     0801    34.4    
## 10 afg     0802    48.0    
## # … with 44,425 more rows

Complexity Measures

You can compute both Economic Complexity Index (ECI) and Product Complexity Index (PCI) by using ec_complexity_measures(). The calculations methods are reflections, eigenvalues and fitness (default). See (Hausmann et al. 2014) and (Mariani et al. 2015) for the methodological details.

cm_reflections <- ec_complexity_measures(
  rca = rca,
  method = "reflections",
  tbl = T
)

cm_reflections$complexity_index_c

## # A tibble: 158 x 2
##    country value
##    <chr>   <dbl>
##  1 deu      2.23
##  2 spm      2.20
##  3 fro      2.04
##  4 che      2.03
##  5 swe      1.99
##  6 jpn      1.91
##  7 aut      1.86
##  8 fra      1.86
##  9 ddr      1.84
## 10 gbr      1.83
## # … with 148 more rows

cm_reflections$complexity_index_p

## # A tibble: 991 x 2
##    product value
##    <chr>   <dbl>
##  1 7111     1.73
##  2 9112     1.57
##  3 8804     1.43
##  4 8208     1.40
##  5 9108     1.37
##  6 7107     1.34
##  7 4906     1.33
##  8 8468     1.32
##  9 3813     1.32
## 10 7221     1.31
## # … with 981 more rows

cm_eigenvalues <- ec_complexity_measures(
  rca = rca,
  method = "eigenvalues",
  tbl = T
)

cm_eigenvalues$complexity_index_c

## # A tibble: 158 x 2
##    country value
##    <chr>   <dbl>
##  1 deu      2.23
##  2 spm      2.20
##  3 fro      2.04
##  4 che      2.03
##  5 swe      1.99
##  6 jpn      1.91
##  7 aut      1.86
##  8 fra      1.86
##  9 ddr      1.84
## 10 gbr      1.83
## # … with 148 more rows

cm_eigenvalues$complexity_index_p

## # A tibble: 991 x 2
##    product value
##    <chr>   <dbl>
##  1 7111     1.73
##  2 9112     1.57
##  3 8804     1.43
##  4 8208     1.40
##  5 9108     1.37
##  6 7107     1.34
##  7 4906     1.33
##  8 8468     1.32
##  9 3813     1.32
## 10 7221     1.31
## # … with 981 more rows

cm_fitness <- ec_complexity_measures(
  rca = rca,
  method = "fitness",
  tbl = T
)

cm_fitness$complexity_index_c

## # A tibble: 158 x 2
##    country value
##    <chr>   <dbl>
##  1 deu     28.8 
##  2 fra     21.4 
##  3 usa     14.1 
##  4 nld     11.5 
##  5 che     10.5 
##  6 gbr      9.81
##  7 ita      9.61
##  8 swe      7.35
##  9 aut      7.34
## 10 jpn      7.27
## # … with 148 more rows

cm_fitness$complexity_index_p

## # A tibble: 991 x 2
##    product value
##    <chr>   <dbl>
##  1 7111     75.1
##  2 2850     55.9
##  3 8804     32.1
##  4 9307     24.7
##  5 9112     20.2
##  6 3505     19.6
##  7 3813     14.3
##  8 7507     12.6
##  9 2927     12.1
## 10 2810     11.8
## # … with 981 more rows

Proximity

Proximity matrices are used to create both country-country and product-product networks. Using proximity_matrices() is straightforward.

pro <- ec_proximity(
  rca = rca,
  d = cm_fitness$diversity,
  u = cm_fitness$ubiquity,
  tbl = T
)

pro$proximity_c

## # A tibble: 10,432 x 3
##    from  to      value
##    <chr> <chr>   <dbl>
##  1 ago   afg   0.0962 
##  2 alb   afg   0.219  
##  3 arg   afg   0.120  
##  4 aus   afg   0.111  
##  5 aut   afg   0.00690
##  6 bdi   afg   0.0714 
##  7 bel   afg   0.00938
##  8 ben   afg   0.107  
##  9 bfa   afg   0.0816 
## 10 bgr   afg   0.0945 
## # … with 10,422 more rows

pro$proximity_p

## # A tibble: 417,822 x 3
##    from  to    value
##    <chr> <chr> <dbl>
##  1 0102  0101  0.346
##  2 0103  0101  0.263
##  3 0104  0101  0.421
##  4 0106  0101  0.2  
##  5 0201  0101  0.381
##  6 0203  0101  0.474
##  7 0204  0101  0.158
##  8 0205  0101  0.368
##  9 0206  0101  0.316
## 10 0207  0101  0.368
## # … with 417,812 more rows

Networks

The proximity_networks() function is designed to use igraph for the internal computations and also to pass proximity-based networks to igraph, ggraph or export to Cytoscape by saving the output as csv/tsv.

To create some reduced networks I’ll use a high proximity cutoff, so that the networks are limited to the spanning trees for clarity and fast rendering.

net <- ec_networks(
  pc = pro$proximity_c,
  pp = pro$proximity_p,
  cutoff_c = 1,
  cutoff_p = 1,
  tbl = T
)

net$network_c

## # A tibble: 157 x 3
##    from  to    value
##    <chr> <chr> <dbl>
##  1 irn   alb   0.375
##  2 irn   afg   0.5  
##  3 irn   tur   0.348
##  4 npl   afg   0.214
##  5 syr   afg   0.357
##  6 ago   ben   0.339
##  7 ago   per   0.358
##  8 ago   sen   0.308
##  9 sur   ant   0.304
## 10 sur   and   0.130
## # … with 147 more rows

net$network_p

## # A tibble: 990 x 3
##    from  to    value
##    <chr> <chr> <dbl>
##  1 9614  4202  0.529
##  2 9614  0101  0.579
##  3 0102  0201  0.5  
##  4 0102  0104  0.423
##  5 0103  0505  0.647
##  6 1102  1005  0.5  
##  7 1102  0105  0.417
##  8 4803  4707  0.615
##  9 4803  0105  0.417
## 10 0106  1302  0.514
## # … with 980 more rows

Just two basic examples with ggraph.

library(igraph)
library(ggraph)
library(magrittr)

set.seed(200100)

net$network_c %>%
  graph_from_data_frame(directed = F) %>%
  ggraph(layout = "kk") +
  geom_edge_link(aes(edge_alpha = value, edge_width = value),
                 edge_colour = "#a8a8a8") +
  geom_node_point(color = "darkslategray4", size = 8) +
  geom_node_text(aes(label = name), vjust = 2.2) +
  ggtitle("The Country Space") +
  theme_void()

net$network_p %>%
  graph_from_data_frame(directed = F) %>%
  ggraph(layout = "kk") +
  geom_edge_link(aes(edge_alpha = value, edge_width = value),
                 edge_colour = "#a8a8a8") +
  geom_node_point(color = "darkslategray4", size = 4) +
  geom_node_text(aes(label = name), vjust = 2.2) +
  ggtitle("The Product Space") +
  theme_void()