The goal of this project is to provide a simple and uniform interface to data from Statistics Sweden (SCB, Statistiska Centralbyrån) directly from the IPython Notebook. Specifically, it is intended to be used together with pandas for iterative data analysis and manipulation.

import pandas as pd
from SCB import SCB
        
scb = SCB('http://api.scb.se/OV0104/v1/doris/sv/ssd/BE/BE0101/BE0101H/FoddaK')
df = scb.filter(code='Tid', kind='item', values=['2014']).get()

This section describes the general outline of the API provided by SCB. The original docs can be found here (Swedish).

For any given table, the URL for the table has the following format:

API-NAME/API-VERSION/LANGUAGE/DATABASE-ID/<LEVELS>/TABLE-ID

where API-NAME currently translates to http://api.scb.se/OV0104/. The current version is 1, and possible languages are sv or en, and the DATABASE-ID is ssd. The depth of the number of levels depend on the requested table. So, effectively, any given table URL will have the following structure:

http://api.scb.se/OV0104/v1/sv/ssd/<LEVELS>/TABLE-ID

It's rarely obvious where the data you're looking for is actually located. For that reason, this library provides a cursor-like object which allows you to quickly jump around between the levels of the database. Once you find the table you're looking for, you're provided with a number of filters. SCB implements a limit of 100,000 rows per request so filtering is often needed.

Here's a workflow example. Note that these operations are executed within the notebook environment.

We start by listing all the top level categories in the database. Calling the describe method on the scb instance outputs a nicely formatted HTML table with all the categories.

from SCB import SCB
scb = SCB('http://api.scb.se/OV0104/v1/doris/sv/ssd')
scb.describe() # outputs an HTML table

We select the top level category "Arbetsmarknad" by using the go method on the instance and passing in the proper level id. We continue to do so until we reach our destination table.

scb.go('AM')
scb.go('AM0502')
scb.go('ArbOrsakSyssSNI2007')

At this point, scb points to an actual table and we can call describe to get a list of possible filters.

scb.describe() # list possible filters
scb.filter(code='Tid', kind='item', values=['2014']) # filter the results
df = scb.get() # save the data in a pandas dataframe

The actual HTTP request is not made until we call get on the instance. Boom! From this point on df is just your ordinary pandas dataframe. Have fun!

Just clone this repository and import the SCB module.