Research Overview

Is there a general support for the welfare state attitude? One we can observe across countries and time? In a study of public opinion during the 1990s in five countries (Breznau 2010), I find some confirmatory evidence for a latent attitude that encompasses support for government provision of health, old-age, family and educational support including both price regulation and subsidies. 

Using different data from many more countries spanning the over three decades, I again find confirmatory evidence for a similar underlying attitude toward government provision of social welfare, or what scholars often label 'the welfare state' (Breznau 2019)

I also look at how existing levels of inequality, presumably as a result of varying social policies, shape public preferences for redistribution. The basic theory is that more inequality should push the public to shift toward preferences for more government redistribution. However, in looking at within-country associations both raw and adjusting for other variables, I see no generalizable effect in my join work with Carola Hommerich (Breznau and Hommerich 2019).

Recently, I led a team of researchers in an attempt to better understand what sometimes appears as inconsistent public support for economic redistribution. Many individuals believe that income inequality in their country is too high, but do not support the government engaging in redistributive policy to reduce it. We hypothesized that this has to do with government affect. That people who think that their government is corrupt, untrustworthy or ineffective are less likely to support government policies in general. This government affect should offer them a heuristic cue for answering the survey question about their support for redistribution. As we were uncertain about the data generating model we ran a multiverse analysis to test for a moderating effect of government heuristics on the link between perceptions of inequality and support for redistribution. These results show that regardless of how we model it, that there is a negative moderating effect - those country-time points where perceptions of corruption and lack of trust is higher, have a reduced partial correlation between perceptions of inequality and support for redistribution (below in Figure 4 from Breznau et al. 2024)

Throughout my matriculation and career as a researcher, I experienced many unreliable and antithetical-to-science features of sociology and the social and behavioral sciences in general. Researchers refusing to answer emails, refusing to share their data and code and engaging in behaviors intended to maximize their chances of getting publications. I discuss these experiences in many posts in my blog Crowdid, in my article 'Does Sociology Need Open Science?' and in a recent talk I gave at Cornell. Motivated by an experience in my graduate studies where I could not replicate nor gain access to original materials from a well-known study (Breznau 2015), and the experience of the norms and rent-seeking motives that create perverse incentives, I became an advocate for open science. 

The published paper is the 'currency of our trade' (to quote a former mentor), and leads academics to sometimes do anything to try and get something published. By doing this we place status-seeking above science. This is what Aage B. Sørensen (1996) defines as a practice that causes "ego-mania" and "much destruction" and a practice labeled by Bruno Frey (2003) as "Publishing as Prostitution". Profit-seeking publishing firms have capitalized on this status-seeking phenomenon in academia and have themselves contributed much destruction to the scientific enterprise. Elsevier has been one of the most draconian. We are bound in a system of paying double-rent for our own science. The universities and funding agencies pay for our research and simultaneously pay for access to this research in published format. 

Motivated by this I began sharing all code and materials and began using strategies to minimize p-hacking and HARKing. Multiverse analysis is something I advocate as a robustness strategy. Motivated by the Open Science Conference 2018 that I organized with colleagues Alexander Wuttke and Eike Mark Rinke, we three developed a study that we hoped would contribute to doing and thinking differently about science. This was our now well-known "Hidden Universe of Uncertainty" study where we had 73 teams independently test the same hypothesis with the same data and come to results all over the place (below is our Figure 1 from the paper). 

Broad variation in the findings from 73 teams testing the same hypothesis with the same data. The distribution of estimated AMEs across all converged models (n = 1,253) includes results that are negative (yellow; in the direction predicted by the given hypothesis the teams were testing), not different from zero (gray), or positive (blue) using a 95% CI. AME are xy standardized. The y axis contains two scaling breaks at ±0.05. Numbers inside circles represent the percentages of the distribution of each outcome inversely weighted by the number of models per team. 

I have presented this study in a number of forums. The Spark Talk was recorded and available online

I am leading a project looking at the 'Role of Theory in The Reproducibility Crisis'. In this project we are developing a method that uses theories-as-data and attempts to identify where a subfield is most in need of theoretical development. 

This is still in its early stages but has a working paper and an R app. Below is a figure using simulated data, about how this would work.

As machine learning and the use of AI is exploding into scientific research, I also caution about the biases inherent in trained algorithms and try to remind social scientists that their are in the business of explanation not prediction, except to use prediction in the service of explanation (Breznau 2022). 

I maintain an open science bibliography on Zotero that is public.

I teach courses and workshops in open science (materials available on OSF) and recently published an article on how to teach replication (Bauer et al. 2024). I also started a blog called 'Crowdid' where I could quickly output positions or findings relating to open science, such as outing a fake journal, using AI for open science, the role of software in irreproducibility, a close look at Sci-Hub, and much more (full list on my Public Science subpage). 

Presentation of our 'Hidden Universe' paper.

Discussion of Multiverse Models at Metascience 2021

With George J. Borjas we identified a strong statistical association of researchers' political ideology and their findings in a policy-relevant area. Paper presented recently at the Harvard Seminar in the Economics of Science & Engineering, Feb. 7th, 2025. 

Our paper "Ideological Bias in the Estimates of the Impact of Immigration" finds that in the many-analysts study I co-led, that teams with more anti-immigration policy preferences, were more likely to find that immigration had a negative impact on support for social policy, and vice-versa for pro-immigration and positive impacts.