Use case: migration flows

In this use case we show how to use tidycenso to visualize the most important internal migration flows over the last 10 years.

First download the number of people by province of residence in 2011 and 2020.

library(tidycenso)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

df <- get_censo("per.ppal", c("ID_RESI_2011_N2", "ID_RESI_2020_N2"), "SPERSONAS")

head(df)
#>   ID_RESI_2011_N2     ID_RESI_2020_N2 SPERSONAS
#> 1     02 Albacete         02 Albacete    324174
#> 2     02 Albacete 03 Alicante/Alacant      3849
#> 3     02 Albacete          04 Almería       366
#> 4     02 Albacete      01 Araba/Álava        75
#> 5     02 Albacete         33 Asturias       165
#> 6     02 Albacete            05 Ávila        39

We then do some data wrangling estimate the net flows, and select the largest ones.


df <- 
  df %>% 
  filter(
    ID_RESI_2011_N2 != "No consta",
    ID_RESI_2020_N2 != "No consta",
  ) %>% 
  mutate(across(c(ID_RESI_2020_N2, ID_RESI_2011_N2), ~substr(., 1, 2))) 

entries <- 
  df %>% 
  rename(to = ID_RESI_2020_N2, from = ID_RESI_2011_N2)

exits <- 
  df %>% 
  rename(from = ID_RESI_2020_N2, to = ID_RESI_2011_N2) %>% 
  mutate(SPERSONAS = -SPERSONAS)


net_flows <- 
  rbind(entries, exits) %>% 
  group_by(from, to) %>% 
  summarise(SPERSONAS = sum(SPERSONAS)) %>% 
  filter(
    SPERSONAS > 0,
    !from %in% c("Ex", "No"),
    !to %in% c("Ex", "No")
  ) %>% 
  ungroup() %>% 
  slice_max(SPERSONAS, prop = .10)
#> `summarise()` has grouped output by 'from'. You can override using the
#> `.groups` argument.

With the estimated net migration flows, we can now visualize them We’ll use tidygraph and ggraph to create the graph, ggplot2 to visualize it and mapSpain to download the geometries of the provinces.


library(tidygraph)
#> 
#> Attaching package: 'tidygraph'
#> The following object is masked from 'package:stats':
#> 
#>     filter
library(ggraph)
#> Loading required package: ggplot2
library(mapSpain)
library(ggplot2)
library(sf)
#> Linking to GEOS 3.12.1, GDAL 3.8.4, PROJ 9.4.0; sf_use_s2() is TRUE

sf_provs <- esp_get_prov() %>%
  select(cpro, ine.prov.name)

nodes <- 
  sf_provs %>% 
  st_centroid() %>% 
  st_coordinates() %>% 
  as.data.frame() %>% 
  mutate(id = sf_provs$cpro)
#> Warning: st_centroid assumes attributes are constant over geometries

graph <- tbl_graph(nodes, edges = net_flows, node_key = "id", directed = T)

coord_layout <- 
  create_layout(
    graph = graph,
    layout = "manual", 
    x = nodes$X,
    y = nodes$Y
  )

Once we have created the graph and the layout for it, we can use ggplot2 to visualise it.


p <- 
  ggraph(coord_layout) +
  geom_sf(data = sf_provs, color = "black",  fill = "grey90") +
  geom_edge_arc(
    aes(
      width=SPERSONAS,
      alpha = after_stat(index),
      color = after_stat(index)
    ),
    lineend = "round", n = 200,
    strength = .1) +
  scale_edge_width_continuous(range=c(0.1, 1.5)) +
  scale_edge_color_gradientn(colors = scales::brewer_pal("seq", direction = 1)(5)[3:5]) + 
  scale_edge_alpha(range = c(.05, 1)) +
  theme_void() +
  theme(legend.position = "none")

p