Data

You learn on a twin of a real research database, and you can use the real one for your thesis, free.

What you work with in this course

The labs and the report use a firm-year panel of about 1,550 large Swedish firms (at least 250 employees in some year), 2018–2024. It combines three kinds of data that researchers use to study AI and work:

• firm accounts from Serrano, the Swedish House of Finance’s database of company financials (employees, salaries, value added, sales, capital, industry, county);
• job advertisements from Platsbanken, the Swedish Public Employment Service, counted per firm and year, with the ads that ask for AI skills identified by the AI-Econ Lab’s AI Monitor;
• occupational AI exposure (DAIOE, a percentile rank from 0 to 100) and O*NET ability scores for the occupations the firm advertises for (SSYK 2012).

The file you get is a synthetic twin of this research panel: the same variables, names, units and structure, and statistical properties close to the real data, but every firm in it is simulated, so it can be shared openly. The quirks of real accounts data are kept on purpose (negative values, firms with zero employees, outliers, firms that change industry), because handling them is part of the job: garbage in, garbage out.

Download

• The course dataset for the labs: na310g-course-data.zip (Stata and CSV, with the codebook and the occupation file). Put na310g_data_course.dta in data/raw/ in your project folder.
• Your own dataset and research question for the report: page Your data, with your personal code from Blackboard.

Serrano for your bachelor thesis: free for you

Serrano is a database with financial history at the company level since 1997, built on annual reports from the Swedish Companies Registration Office (Bolagsverket), company data from Statistics Sweden and bankruptcy information, and covering most legal forms of Swedish firms. Örebro University holds a licence, so as a student here you can download it yourself, at no cost, by logging in with your university account.

• Log in and download: data.houseoffinance.se/serrano (choose Örebro University when asked for your institution).
• User manual: how to download and unpack the files. The yearly Stata release comes in seven parts that you unzip together (on a Mac, use for example The Unarchiver or Keka; on Windows, 7-Zip). The files are large (the main Stata file was about 19 GB in the 2023 release), so check your disk space first.
• Help: support.data.houseoffinance.se.
• Cite it as: Weidenman, P. The Serrano Database for Analysis and Register-Based Statistics. Swedish House of Finance Research Data Center.

After this course you already know the variables, the panel structure and the checks they need, so starting a thesis on Serrano is a small step.

Public data you can combine with it

• Statistics Sweden’s database: aggregates by industry, region and year, for example to add industry or county controls.
• Platsbanken historical job ads (JobTech, open data): more than 7 million ads published on Platsbanken since 2006.
• O*NET: what each occupation involves, the basis of most AI-exposure measures.