Stefan B. Andrade, Jens-Peter Thomsen
Sociological Science, February 14, 2018
DOI 10.15195/v5.a5
Abstract
An overall finding in comparative mobility studies is that intergenerational mobility is greater in Scandinavia than in liberal welfare-state countries like the United States and United Kingdom. However, in a recent study, Landersø and Heckman (L & H) (2017) argue that intergenerational educational mobility in Denmark and the United States is remarkably similar. L & H’s findings run contrary to widespread beliefs and have been echoed in academia and mass media on both sides of the Atlantic Ocean. In this article, we reanalyze educational mobility in Denmark and the United States using the same data sources as L & H. We apply several different methodological approaches from economics and sociology, and we consistently find that educational mobility is higher in Denmark than in the United States.
An overall finding in comparative mobility studies is that intergenerational mobility is greater in Scandinavia than in liberal welfare-state countries like the United States and United Kingdom. However, in a recent study, Landersø and Heckman (L & H) (2017) argue that intergenerational educational mobility in Denmark and the United States is remarkably similar. L & H’s findings run contrary to widespread beliefs and have been echoed in academia and mass media on both sides of the Atlantic Ocean. In this article, we reanalyze educational mobility in Denmark and the United States using the same data sources as L & H. We apply several different methodological approaches from economics and sociology, and we consistently find that educational mobility is higher in Denmark than in the United States.
 | This work is licensed under a Creative Commons Attribution 4.0 International License. |
- Citation: Andrade, Stefan B., and Jens-Peter Thomsen. 2018. “Intergenerational Educational Mobility in Denmark and the United States.” Sociological Science 5: 93-113.
- Received: December 7, 2017
- Accepted: January 9, 2018
- Editors: Jesper Sørensen, Stephen Morgan
- DOI: 10.15195/v5.a5
In a response to this article, Kristian Karlson (2021) demonstrates that the authors’ findings are an artifact of the use of a crude education measure which does not differentiate at the lower end of the educational attainment distribution. Given Sociological Science’s mission to create a platform for scientific exchange, I think it is important to point this out here. It brings up at least two important points, IMO.
1) If sociology were to take itself more seriously as a scientific endeavor, we might expect that overturning the Andrade and Thomsen (A&T) findings should lead to a retraction. This is not the case here for this study. This motivates me to write this comment. Studies that will not be retracted but may contain flawed or mis-modeled analyses should be flagged, or at least quick and casual readers should have easy to find information that suggests they may not want to quickly cite this study without a careful consideration of the evidence. This is my opinion, but I see good reason to do this at it would help reduce promotion of citations of flawed studies. And the comment area seems to be the ideal place to do this, hence my comment.
2) Karlson shows that differentiating at the bottom of the distribution of education, in particular among parents, explains ‘away’ A&T’s findings. This brings up a theoretical problem. Just because we can alter a measurement and get different results does not mean we should. If sociology’s goal is to explain phenomena, and in the best cases identify causality, then a data-driven measurement discussion should also be a theoretical discussion. Namely, are there meaningful theoretical differences between those in a 3- or 5-category ranking of education and can these be meaningfully compared between the U.S. and Denmark. Karlson points out these problems. But what if switching to a 7-category or 11-category ranking would reinstate A&T’s findings, just like switching to a 5-category measure removed them? The literature Karson cites (from Pfeffer and Breen for example, and especially Schneider) suggest finer grained measures are better but again these face limitations across educational systems. And we should all know the cost of dichotomization or trichotomization of variables to which there is more information, as it throws away potentially meaningful variance (e.g., MacCallum et al 2002). Thus, there should be a strong theoretical basis behind our modeling decisions regardless of what the data and different robustness checks say. Otherwise we can tinker with models and produce the results we ‘need’ to defend a position. At least this is becoming clear in the findings from recent many analyst and multiverse approaches (e.g., Schweinsberg et al or Breznau et al).
Regarding point 2 – it is now my hope that an expert in education measurement and theory will join the discussion, and we can make our sociology more interactive, taking advantage of this platform!
Cites:
Karlson, Kristian Bernt. 2021. “Is Denmark a Much More Educationally Mobile Society than the United States? Comment on Andrade and Thomsen, ‘Intergenerational Educational Mobility in Denmark and the United States’ (2018).” Sociological Science 8:346–58. doi: 10.15195/v8.a17. https://sociologicalscience.com/articles-v8-17-346/
MacCallum, Robert C., Shaobo Zhang, Kristopher J. Preacher, and Derek D. Rucker. 2002. “On the Practice of Dichotomization of Quantitative Variables.” Psychological Methods 7(1):19–40.
Schweinsberg, Martin, et al. 2021. “Same Data, Different Conclusions: Radical Dispersion in Empirical Results When Independent Analysts Operationalize and Test the Same Hypothesis.” Organizational Behavior and Human Decision Processes.
Breznau, Nate, Eike Mark Rinke, Alexander Wuttke, et al. 2021. “Observing Many Researchers Using the Same Data and Hypothesis Reveals a Hidden Universe of Data Analysis.” https://osf.io/preprints/metaarxiv/cd5j9/
Hi Nate,
Thank you for your interest in our work.
Have you read our rejoinder?
Here, we show that not only does Karlson completely refrain from dealing with our critique of Landersø & Heckman’s analysis, he also does not engage with our data at all, but remain in the speculative realm. Please see our table 4 in the rejoinder where we top up NLSY with GSS-data and show that our results are robust, and by no means an artefact of a crude education measure. In fact, when we include additional parental information from the National Longitudinal Survey of Youth 1997 (NLSY97) for the United States, as suggested by Karlson, the gap between Denmark and the United States increases.
We await your comment.
Here is a link to our rejoinder published together with Karlson’s comment:
https://sociologicalscience.com/articles-v8-18-359/
As science operates at the speed of science, it will take me some time to read your rejoinder in detail and especially to link the three papers (your 1st, Karlson’s and your 2nd) to the original Landersø and Heckman paper. As this could take years, it seemed prudent to offer a quick reply.
1) Based on some conversations in social media, my original comment may have given the impression that I was calling for “retraction” of your paper. This was certainly not my intention, and I apologize if this in some way has caused you professional or personal harm. I only wanted to call attention to the fact that a reader of your article would not automatically be aware that there were some problems with your paper. Perhaps this is an improvement that Sociological Science could make to their platform to ‘flag’ papers which contain some (potentially serious) mistakes.
2) I have casually read your new paper and, as you know, started doing some initial work with the US data which I thank you for sharing along with your well documented replication code. Your rejoinder seems like an entirely new paper with many new analyses. That is great. Pushing forward our knowledge base! It doesn’t change the fact that the first paper has problems, but I am happy that there is an exchange going on with you and Karlson that should ideally move science forward.
3) However, I must say that I think your rejoinder title (Yes, Demark is a More Educationally Mobile Society…) is more overconfident than any scholar should be about quantitative social science. We know that this work is messy, where data-generating models are not always known, and data never drawn from a fully random sample. Also, your making a conclusion that appears to transcend space and time based on results from a small segment of time in history (a single ‘cohort’ roughly speaking). We are seeing from an outpouring of recent social science that involves many analysts and multiverse approaches, that we should have more humility given the great deal of uncertainty in our point estimates and their confidence intervals.
4) In fairness, your title and work is far more humble and scientific in its conclusions, than Landersø and Heckman, who, after believing they showed that there is no difference in educational mobility between Denmark and the U.S. came to the extreme conclusion that this was evidence that the Danish welfare state is a waste of government investment. I will refrain from saying much, but certainly radical spatial and social group inequality, mass incarceration, extreme amounts of citizen and police gun violence, lower life expectancy lower levels of subjective well-being in the U.S. all speak against the Landersø and Heckman conclusion (certainly as much or more than a few cohort estimates of educational mobility that they use to support their conclusion).
I appreciate your academic work, and thank you for taking the time to have an exchange with me here on this platform.