The data was originally assembled for the project "The Politics of Administrative Design" (financed by the Swedish Research Council through grant 2014-947), which focused on how partisan shifts in government can affect the staff, structure, and process of public bureaucracies. The purpose of the dataset is to provide a quantitative catalogue of Swedish agencies for public use.
For the instruction data, one observation corresponds to one agency instruction, while the variables cover factors such as the agency's management structure and formal functions. For the agency head data, one observation corresponds to one agency head, while the variables cover factors such as the head's education and background experience. For the agency budget data, one observation corresponds to one fiscal year, while the variables cover factors such as the amount of funds that were allocated and withdrawn from the agency during that year. Finally, all three datasets contain relevant time-period indicators to enable users to determine the temporal coverage of an observation (e.g., enactment and revocation dates for the instructions). In total, the database covers 1925 agency instructions, 2315 agency heads, and 7102 fiscal years.
Importantly, while each dataset has its own unit of analysis, we have also included a unifying agency identification number that can be used to link variables across all three datasets. For example, suppose that we are interested in the School Inspectorate ("Skolinspektionen"). We can then start our investigation in either of the three datasets, and use the agency's ID number to match the agency's various instructions, heads, and budgets over time. Note, however, that because the units do not perfectly overlap (e.g., one fiscal year may cover more than one agency head), merging the datasets will require substantive decisions about how to structure the information in each dataset. For this reason, we have left it to the user's discretion to decide whether and how to merge the datasets, rather than impose any particular structure on all three datasets. In total, the database contains information on 664 unique agencies.