What Fafo Actually Is
It is a research institute based in Oslo that focuses on labor market issues, welfare policy, and social inequality. Full name is Fafo Institute for Applied International Science, though nobody outside academic circles calls it that. They produce reports on unemployment trends, immigration integration, minimum wage effects, and public sector efficiency. Their work shows up in Norwegian parliamentary debates and media commentary fairly often. Their research methodology is fairly standard social science. Large-scale survey data, register statistics from Statistics Norway, and econometric modeling. What makes them worth paying attention to is their access to Norwegian administrative registers. Those registers contain nearly complete population-level data on employment, education, health outcomes, and income going back decades. Not every institute gets that kind of access. Fafo has maintained it for years, which means their publications tend to have a level of empirical grounding that other think tanks lack. I used to work with a lot of their datasets when I was doing my own analysis on labor market participation rates. The tricky part is that their raw microdata files are massive and the documentation, while thorough, is written in Norwegian. You will save yourself hours if you learn to navigate their register-based variables rather than relying on the clean summary tables they publish. The published reports are fine for a general picture. The actual data tells you something different sometimes.
How Their Research Gets Used
Policy makers in Norway cite Fafo heavily because their work tends to be methodologically conservative. They do not chase headlines. A typical Fafo report will present a regression result, show confidence intervals, and acknowledge the limitations of their identification strategy. That is actually useful. A lot of think tanks out there present correlational findings as causal. Fafo usually does not cross that line without a very clear caveat attached. One thing most people miss is that their work on immigration and labor market outcomes has shifted significantly over the last decade. Early 2000s research on refugee settlement patterns showed different findings than their post-2015 work. The methodology improved, the sample sizes grew, and the policy context changed. If you are citing Fafo, you need to check the date. A 2008 Fafo report on integration labor market outcomes is not going to apply cleanly to the current situation. I learned that the hard way when I tried to use older data in a model and got results that looked obviously wrong for the present period. The workaround was to build a weighted composite using only their post-2014 publications and verify against Statistics Norway contemporaneous figures.
Getting Their Publications
All Fafo reports are available free on their website. There is no paywall. The PDFs are well formatted and most have English summaries even when the full text is in Norwegian. Their working paper series is where the more technical material lives. Those are the ones researchers actually read and cite in academic work. The policy briefs are written for a broader audience and sometimes oversimplify findings that are more nuanced in the full papers. If you are mining their data for your own analysis, go straight to the register-based datasets. They require an application through the Norwegian Central Bureau of Statistics, but the approval process is straightforward for academic or institutional researchers. Individual researchers can get access. It takes about six to eight weeks from application to data transfer. Plan ahead.
Get the Full Details
Where Their Work Falls Short
Fafo is strong on descriptive and quasi-experimental analysis. They are less impressive on predictive modeling or forecast work. If you need someone to predict what unemployment will look like in two years, their track record is average at best. They are better at explaining why something happened after the fact. That is a real limitation if you are using their research for forward-looking policy planning. Their models tend to lag behind structural breaks in the economy, which Norway has experienced several times recently. Another issue is that their scope is almost entirely Norway-focused. They have some international comparative work, but it is thin. If you are looking for cross-country labor market analysis, you are better off going to the OECD or the Nordic Council of Ministers publications. Fafo will not be your primary source for anything outside Norway. Their funding comes partly from the Norwegian government, which some people view as a conflict of interest. In practice it rarely shows up in the research itself. Their funding agreements are transparent and they maintain editorial independence. But you should be aware that their commissioning bodies can influence which topics get funded. Research on politically sensitive subjects sometimes takes longer to materialize than work on less contentious areas. It is not censorship. It is just resource allocation.
Practical Tips for Using Fafo Data
The register variables use Norwegian coding standards. Age is stored differently than in international datasets. Income variables come in multiple forms: gross, net, taxable, and equivalized. Pick the one that matches your analysis before you start merging, because converting between them later is tedious. The documentation explains the differences, but you will waste time if you assume one format is interchangeable with another. Their geographic variables are based on municipality codes that have changed over time due to municipal mergers. If your analysis spans multiple years, you need to account for those boundary changes. Fafo provides reconciliation tables, but they are not always up to date. I cross-reference their tables with the Statistics Norway geographic history database and manually adjust for any discrepancies. It adds maybe thirty minutes to the setup phase but prevents systematic errors in your results. If you are not familiar with Norwegian administrative data structures, the learning curve is steeper than it needs to be. Most of their codebooks assume you already know how Norwegian registers work. A basic understanding of how person IDs, municipality codes, and yearly observation windows are structured will save you significant debugging time. The alternative is spending days figure out why your panel data has gaps that look random but are actually just recording artifacts from administrative system changes.
