In this article, I wish to answer a question that has been posed by a colleague almost immediately after I submitted my PhD thesis: which parts of the various papers did I enjoy writing the most, and which one did I enjoy the least? It is a natural question and one that I instinctively believed to be fairly easy.
I thought for a few seconds, and then responded that I couldn’t really point my finger to it, and that I probably enjoyed everything somewhat equally. However, I noticed an oddity. I did enjoy writing the chronologically first and third papers more than the second one.
I struggled for about a week to find a proper hook for this article, until — in the dark of night, in a heavily delayed German train — I had the necessary insight. Yes, I did enjoy writing all my papers, but the fact that I did enjoy the second paper a little less was because that fell into a time when something in my writing changed: I went from writing pure theory pieces to writing methods-papers.
At first, my writing was heavily influenced by the more classical, German sociological tradition. It shows in my chronologically first paper, as it is somewhat sparse in methods, but rich in theory. As I grew accustomed to the toolbox of analytical sociology and computational social science, however, this began to change. The chronologically second paper falls into this twilight zone, where my writing adapted to accommodate more complex methodological setups, but wasn’t quite there yet. And the chronologically last paper in my dissertation then manages the switch.1
This is what I want to focus on today: My academic provenance in theory, the acquisition of methodological skills during my PhD, and how this has completely changed how I do research and what I enjoy about it.
Theory and the Historical Sciences
During my B.A. and M.A., I heavily focused on historical and social theory, and until I joined IAS, this didn’t change much. This was not the least due to the fact that I simply lacked methodological experience, and felt uncomfortable with having to perform analyses that others could trust. I knew the basics of survey methodology, and what an OLS regression looks like. I learned about PCA, correspondence and factor analysis2 but neither did I ever work with actual data, nor did I have a feeling for what an OLS was telling me.3
During my B.A., theoretical work was essentially built-in, because that’s what history is about: collecting historical evidence and then building a theory that can explain how historical events likely went about based on this evidence.4 In history, theorizing about past events is really the only way to tell a cohesive story, because historical evidence does not lend itself to straight-forward causal claims. But very quickly I realized that purely historical methodology is inadequate for explaining how history really happened.
The inflection point for this insight came in a seminar on medieval history. In my term paper, I was tasked with exploring the founding history of the diocese of Verden, a small German town close to the city of Bremen. Like with any other source of data, the farther you go into the past, the less reliable the data becomes. And we’re talking early European medieval ages (around 800 AD), so you can probably imagine how bad it was. By 2013, when I took the course, diplomatics researchers had already done a good job at sifting through the archival evidence, and we had a good understanding about the available facts. But those facts were so sparse that any statement about what might have happened back then amounted to little more than tea-leaf reading.
So I decided to throw in a bit of sociological theory to see if it stuck. I chose neo-institutionalism,5 because that was what I just learned at that time. And indeed, by applying this theory to the data, a somewhat better picture of the founding history of that diocese emerged. What I didn’t think about, however, was that I was dealing with a medievalist, and not a sociologist. The term paper ended up being the worst-graded thing I wrote during my entire undergrad. However, when I applied the same theoretical framework to a similar historical episode in my bachelor thesis, I received a very decent A-. That confirmed my suspicion that my bad grade in the seminar was mostly due to my undergrad hubris and failing to read the room, and not a fundamental flaw in my thinking.
At this point I realized that history was not my vocation. I was too interested in understanding social change and how society works, and had too little interest to continue on my path of becoming a historian. So, for my graduate studies, I switched to sociology.
From Historical to Sociological Theory
In the fall of 2014, I started my graduate studies in sociology. Or rather, as my university called it, “Societies, Globalization, and Development” (Gesellschaften, Globalisierung und Entwicklung, often abbreviated GGE). I realized only after I started the program that the reason we didn’t have a “Sociology” program was that the institute was simply lacking personnel for it.6
However, I am very happy that I received the GGE-package, and not a straight-forward sociology package. The downside of the program was certainly its lack of methodological education, but this lack was compensated by a huge amount of sociological theory and sheer breadth of knowledge. I took courses in international relations, domestic policy, area studies, systems theory, history of sociological science, and more. I was taught the primary schools in international relations, economic sociology, how finance works in Ethiopia and Bangladesh, and the history of Western sociology. In the evenings, I was a frequent guest at lectures from the philosophy department to bring myself up to speed on the philosophy of science.
All of this meant that I was well-equipped with broad world knowledge, but little methodological knowledge. There were a few methods courses that mostly revolved around surveys and simple OLS regression, but those made up only part of the study program. I received the biggest push in methods education only in the final year of my graduate studies, when I took a course on advanced methods. The closest I got to proper applied quantitative methods during that time was when I was employed in a project on food security in Ethiopia, which required me to wade through a few Gigabytes of quantitative survey data using Stata.
What I am eternally grateful for is that the program helped me develop a good theoretical sense and critical thinking skills. This helped me publish my very first peer-reviewed paper in 2019 on crowd science and the sociology of violence.7 I ended the program with a 100-page theoretical thesis on the cycle of violence between British and U.S. imperialism in the Middle East and radical Islamism that culminated in the founding of ISIS.
Entering the Lab
With the start of my employment at the Institute for Analytical Sociology, the days of me being a theory-nerd were over. Now it was all about quantitative methods: data, precision, methodological rigor. And, more than some of my colleagues, I really had to sit down and learn. I lacked what colleagues from more traditional quantitative programs already brought to their PhD. Given that I made it this far, I appear to have been successful, although this is for my committee to decide.
The papers that comprise my dissertation are approximately chronologically ordered, and across all three, one can clearly see a trajectory of me acquiring methodological skills. The first paper, Policymaking in Times of Crisis, is very theory-heavy. I spent a lot of time scouring through the qualitative literature to understand the case I was dealing with. The methodological part is comparatively short. In the end, it turns out to be the most straight-forward methods framework of all my papers. I spend pages introducing and discussing the underlying theory, and then a few paragraphs on collecting data, and testing my hypotheses using very simple fixed-effect OLS regressions.
My second paper, Measuring Issue-Level Polarization, is somewhat of a middle child. I got a bit more adventurous with my methods, and my theoretical part shrank substantially. However, the paper does not feel holistic. Methodologically-wise, I was still learning, and the theory did not lend itself nicely to the approach. That paper was all about a methods-heavy approach where theory came second. It almost appears as if the paper’s methods and theory parts are in a form of “competition” for attention. I will come back to this below.
The third paper then, Brittle Parties?, appears to be better both in terms of methods and in terms of theory. Re-reading it now, I get the feeling that it’s one example for what I learned during my PhD. The theory part is more solid than in paper number two, and it leads naturally to a set of measurements I will need in order to answer my research question. The theory properly binds my research question to my methods. And the methods, in turn, are suitable to answer the questions. They neither attempt to look as fancy as possible, nor are they standard run-of-the-mill.
While none of them are perfect, I think the papers show an evolution from an abstract theoretical writing and research style to a methodologically heavy, quantitative one. And the more I think about it, the more I believe that it would be weird if all my papers were equally good. It appears almost “honest” to have a paper somewhere in the middle, which is neither my best nor my worst work, as it serves as evidence for the development that has happened in my thought processes over the past five years.
Looking back, it is interesting to see how my interests have completely changed. I still enjoy theory and thinking about concepts, but I am equally fascinated by applying state-of-the-art methods. At first, my theory was an end in itself. Now, it serves the purpose of answering a research question, and rests on equal footing with the methodological approach.
But it was a long journey. In the first two years of my PhD, I was happy if I got a simple, linear regression going. At the same time, my colleagues were pushing for complex causal models; simulations; and agent-based models. Gradually, this changed. As I participated in more and more courses (one of the requirements of earning a Swedish PhD involves accumulating 60 ECTS in course work8), my confidence in applying methods grew. By now, I am comfortable applying run-of-the-mill statistics as well as top-notch text analysis methods from LDA through word2vec to Encoder-models and generative AI.
There is one more thing, though. It is not just that I learned to weave methods into my theoretical thinking. In order to properly include quantitative methods in my research process, I had to change the way I approach theory.
Theory ≠ Theory
This brings me to a final point for today’s reflection. The way one writes theory, and how one approaches theoretical problems is partly a function of the discipline, and partly a function of the methods. I mentioned that my second paper is a middle child. The reason for this is that the theoretical part does not appear to fully fit the methodological approach. In that paper, I used the same, theory-heavy writing style like I did in my first paper, but then added a heavy methodological part on to it. Now I realize that the theory one writes for theory-papers is of a different kind than the theory required to write methodological papers.
Indeed, no theoretical approach is the same. Historical theory is fundamentally different from political theory, which in turn differs from sociological theory, which differs from the theory you need for a quantitative paper. I never excelled at historical or political theory; and I believe this is partly due to the fact that historical theory never fully “clicked” for me and that political theory has its very own language which I do not speak.
But even within sociology, theory can differ tremendously; both in form and content. Gabriel Abend has written an excellent piece on “The Meaning of Theory” (2008)9 that tries to disentangle the various notions of theory, and how to justify one’s own choice in a paper. Abend argues that, which definition, or concept, of “theory” one uses is partly a political question because there is no “real or objective referent for ‘theory’” (p. 176). I argue that the way you write theory is partly defined by the type of paper you write, hence a function of the methodological approach.
In my paper on the sociology of violence, I rely on research literature to argue that the conceptual perspective of crowd science has affected the sociology of violence, and is visible in many concepts the latter uses. I spend many paragraphs arguing that the choice of words from the 19th century crowd science was derogatory and subjective, and that sociology can – and should – do better.
In my dissertation papers, I do not produce an argumentative chain to discuss concepts and their drawbacks. Instead, I focus on particular phenomena, and produce theoretical insight only insofar as it is relevant to the case at hand. Instead of weighing theoretical notions and their pros and cons, I draw on theory for a set of expectations and relevant concepts, and then let the data speak. However, and this is important, my methodological choices and my measurement instruments are all derived from theoretical insight. I use measurements not because they are convenient, but because theoretical accounts argue that they are the correct ones.
Doing so requires a learning process The reason that the theory-part of my chronologically first paper, Policymaking in Times of Crisis, looks substantial is because I had been working theoretically on the concept for a few years, so I already had a good grasp on it. The methods are fully in the service of the theoretical part, and are used merely as a test for some of the theoretical assumptions I make. The core of that paper is theory, and since I had ample experience with theory writing, this is a potential explanation for why it looks good (at least for me – again, the dissertation committee may come to a different conclusion).
The reason for why my chronologically third paper, Brittle Parties?, appears (to me) to have an equally good match between methods and theory is that the type of theory used here is entirely different from the one I used in my first paper. While in the first paper, my methods were in the service of my theory, in the third paper theory is in the service of my methods. Theory serves as the connecting thread that ties my research questions to my measurements and methodological challenges.
Both my first and third papers, therefore, feel well-rounded to me because the fit between theory and methods is better than in my second paper. The style of theory differs completely, but it matches the corresponding methods. Thus, the reason for why my chronologically second paper, Think Alike, Talk Alike? looks much weaker in comparison, is because I wrote a theory part similar to my first paper, while at the same time applying methods like in my third paper. Theory and methods thus do not match. This leads to the impression that theory and methods appear to be in some kind of competition for attention. What should the reader focus on? The theory? Or the methods? Or both? It’s hard to determine this just from looking at the outline.
Final Thoughts
So, what do I enjoy most — theory or methods? This might be the wrong question. No theory can survive without a methodological approach, regardless of whether it is qualitative or quantitative. Likewise, methods cannot live without theory, as has become clear during the debate on “post-theory” from the late 2000s. But depending on how one weighs each part, the style of theory must adapt. Writing theory-papers where the “data” is other research literature requires a different process than writing methods-papers.
This change that I lived through was subtle. Through my rich PhD education, I learned to value the ability of being able to test theories using quantitative data. Instead of restricting myself to theoretical thoughts in the proverbial ivory tower, I gained the ability to test theories with data. And while this means that my papers now have a large methodological part which necessarily makes the theoretical parts smaller, this approach comes with its own challenges. And this trajectory shows in my dissertation.
To answer the question: I enjoyed all parts of the writing process. What I enjoyed less was the at times painful learning process that enabled me to write proper research papers in the first place.
1 At least that’s what I hope. I haven’t defended it yet.
2 This is primarily thanks to Prof. Andreas Schmitz, who joined my university during my graduate studies, and who did an amazing job in stuffing our brains with fundamental methodological knowledge in a short amount of time. More so, he will even be able to verify first-hand whether I learned properly as part of my dissertation committee. The world is small.
3 Several years ago, I was in Stockholm together with my supervisor, and on the train ride back I uttered concerns about my regression models, and that I felt uncomfortable with reporting its results. When he asked why, he was surprised to learn I did not run a proper regression before.
4 As our teachers never went tired of telling us: Following Leopold von Ranke, we were supposed to tell history “wie es eigentlich gewesen” (“as it really happened”).
5 Hasse, R., Krücken, G., & Meyer, J. (2005). Neo-Institutionalismus (2., vollst. überarb. Aufl). Transcript-Verl.
6 Fortunately, things have changed. By now there is not only sufficient sociological personnel, but word has it that there are plans for splitting up the combined political science and sociology department into one for political science, and another one for sociology.
7 Erz, H. (2019). Der lange Schatten von Gustave Le Bon. Zum sprachlichen Einfluss der Crowd Science auf die Soziologie der Gewalt. Soziologiemagazin, 2019(2), 71–88. https://doi.org/10.3224/soz.v12i2.06.
8 I accidentally accumulated over 70, because I forgot to count.
9 Abend, G. (2008). The Meaning of ‘Theory.’ Sociological Theory, 26(2), 173–199. https://doi.org/10.1111/j.1467-9558.2008.00324.x.