Tutorial

Introduction

The insee package gathers tools to easily download data and metadata from insee BDM database.

It uses SDMX queries under the hood. Have a look at the detailed SDMX webservice page.

The first version of the package was published on CRAN 2020-07-29.

Proxy issues

Requirement for INSEE employees

In order for someone working behind a proxy server to be able to use insee, it is necessary to modify system variables as follow.

Sys.setenv(http_proxy = "my_proxy_server")
Sys.setenv(https_proxy = "my_proxy_server")

Installation & Loading

You can easily install insee with the following code :

# Get the development version from GitHub 
# install.packages("devtools")
devtools::install_github("InseeFr/R-Insee-Data")

# CRAN version
# install.packages("insee")

# library Loading
library(insee)
library(tidyverse)

Functionalities

This section will give you an overview of what you can do with insee.

Series have two identifiers the SDMX identifier and the so called idbank. Both can be used to download data.

Datasets List

INSEE BDM database offers more than 200 Datasets. The get_dataset_list() function returns the datasets catalogue :

insee_dataset = get_dataset_list() 

Series Keys List

INSEE BDM database currently offers more than 150 000 series. The get_idbank_list function returns the series catalogue from a dataset name.

idbank_list = get_idbank_list('BALANCE-PAIEMENTS')

Find a series key

The best way to download data is to find the right series key (idbank), but how ? Indeed, in some cases it is not easy to understand what are the differences among series, especially for non-French speakers. To make the search easier, the best way is to use the get_idbank_list function with a dataset name, then it can be helpful to filter with the columns FREQ, NATURE, UNIT etc. Moreover, the insee package provides the function add_insee_title to get titles from idbanks, either in English or in French. It is not advised to use the function on the whole idbank dataset, as each SDMX query has 400-idbank limit. Then, add_insee_title function splits the list into several lists of 400 idbanks each. Thus, the user should filter the idbank dataset before using the function to avoid as much as possible this bottleneck as the following example shows. After the data retrieval, it is really nice to use the split_title function on the dataframe to get more readable titles easy to use in plots and add_insee_metadata to get the metadata with the data.

idbank_list_selected =
  get_idbank_list("IPI-2015") %>% #industrial production index dataset
  filter(FREQ == "M") %>% #monthly
  filter(NATURE == "INDICE") %>% #index
  filter(CORRECTION == "CVS-CJO") %>% #Working day and seasonally adjusted SA-WDA
  #automotive industry and overall industrial production
  filter(str_detect(NAF2,"^29$|A10-BE")) %>% 
  add_insee_title()

Another way to find a series key is to perform a keyword-based search with the function search_insee. Beware that this function uses package internal data which might not be the most up-to-date. See the following examples :

# search multiple patterns
dataset_survey_gdp = search_insee("Survey|gdp")

# data about paris
data_paris = search_insee('paris')

# all data
data_all = search_insee()

Download data

Download using a list of idbanks

The get_insee_idbank function should handle up to 1200 idbanks. It is then advised to narrow down the idbanks list used as argument of the function. Otherwise, put the limit argument to FALSE to ignore the function's idbank limit.

library(tidyverse)
library(insee)

# the user can make a manual list of idbanks to get the data 
# example 1

data = 
  get_insee_idbank("001558315", "010540726") %>% 
  add_insee_metadata()

# using a list of idbanks extracted from the insee idbank dataset
# example 2 : household's confidence survey

df_idbank = 
  get_idbank_list("ENQ-CONJ-MENAGES") %>%  #monthly households' confidence survey
  add_insee_title() %>% 
  filter(CORRECTION == "CVS") #seasonally adjusted

list_idbank = df_idbank %>% pull(idbank)

data = 
  get_insee_idbank(list_idbank) %>%
  split_title() %>% 
  add_insee_metadata()

# example 3 : get more than 1200 idbanks

idbank_dataset = get_idbank_list()

df_idbank = 
  idbank_dataset %>%
  slice(1:1201)

list_idbank = df_idbank %>% pull(idbank)

data = get_insee_idbank(list_idbank, firstNObservations = 1, limit = FALSE)

Download using a dataset name

For some datasets as IPC-2015 (inflation), the filter is necessary.

insee_dataset = get_dataset_list() 

# example 1 : full dataset
data = get_insee_dataset("CLIMAT-AFFAIRES")

# example 2 : filtered dataset 
# the user can filter the data
data = get_insee_dataset("IPC-2015", filter = "M+A.........CVS.", startPeriod = "2015-03")

# in the filter, the + is used to select several values in one dimension, like an "and" statement
# the void means "all" values available

# example 3 : only one series
# by filtering with the full SDMX series key, the user will get only one series
data = 
  get_insee_dataset("CNA-2014-CPEB",
                    filter = "A.CNA_CPEB.A38-CB.VAL.D39.VALEUR_ABSOLUE.FE.EUROS_COURANTS.BRUT",
                    lastNObservations = 10)

Support

Feel free to open an issue with any question about this package using https://github.com/pyr-opendatafr/R-Insee-Data Github repository