empirical political research |

Communication works for those who work at it.

Fri, 20 Nov 2020 00:00:00 +0000

Communicating information in science is a challenging task, so after giving feedback to undergraduate research designs, I realized that some simple bits of advice might help young researchers convey their points more straightforwardly. Are you curious about them? Then you are on the right track here.

Remind the reader of your arguments before jumping into operationalization and measurement. If we can forget our' parents' birthdays, do you think the memory is suddenly going to improve for some fancy hypotheses? No, it won’t, and it is almost a deterministic claim. Summarize your main points in one or two sentences and refer to particular hypotheses when naming your variables. Remember (from my perspective): your reader is lazy and does not want to overthink - instead, he/she lets you convince them. So, make your writing smooth and cogent.
Phrase your hypotheses as concisely as possible, so operationalize your statements properly. For instance, if you intend to examine educational performance, give it a number - in grades, hours of studying, or even the annual amount of money spent on tutoring. We do not need a perfect measure, but having a plausible one is a must.
Avoid using killer words, such as “different,” “broad,” “many,” “multiple,” and their synonyms, as they add unnecessary vagueness to your story. If you are trying to stratify the sample, use more precise descriptors to justify your choice. “Different questions” or “different levels of measurement” would NOT work out. Good practice: fill your research design writing with adjectives and quantifiers (numbers) to make it more succinct and more comfortable to read for a knowledgeable audience.
Write out numbers and entities, restrict abbreviations to the necessary minimum or avoid them at all: we expect “three hours of individual participation” and not “3 h.” Of course, your reader will understand your intention, but well-trained academic community uses reader-oriented writing style guidelines that you should stick to as well. Numbers always catch the reader’s attention by distorting the reading flow, so make sure that you use numbers only if they must be emphasized and not because of carelessness. Remember: the reader (or reviewer) is the king.
Justify, justify, and, once again, justify. Obviously, even sound research design strategies have shortcomings, so try to always attach a “because” to your claims even if they seem self-evident. By doing so, you will be less harshly criticized for flawed arguments. For instance, writing “I select 1000 respondents from each federal state” will not help. Writing “Sample sizes will be determined, proportional to the population of every federal state to increase sample representativeness” does.
Finish by discussing the limitations of your research design and speculating about alternative ways of studying your research problem. Social science is about cumulative efforts, so there is no need to expect that you invent something new. Instead, aim to enhance and deepen existing knowledge about how the world works than re-tell the entire mankind’s history in one scientific paper.

Want to know more research design insides? Take a look at my presentation from Fall 2020.

All life is an experiment. The more experiments you make the better.

Tue, 20 Oct 2020 00:00:00 +0000

During the session on experimental research design, students are expected to learn about the advantages and limitations of experiments in social sciences and experimental ethics. How do I design an experiment? What questions could I answer by setting up an experiment? How do I maximize external and internal validity? Empirical political research addresses these questions (and many more), giving excellent insides into survey and experimental methods.

First things first: experiments, if simplified, can be divided into three categories: laboratory, field, and natural. This typology builds on the extent to which a researcher is involved in the experimental process: the lab offers most for controlling and manipulating, while in a natural setting, the researcher takes the observer’s role. The second important distinction that should be made is between control and treatment groups. The former embodies participants who do not receive a stimulus, which should induce changes in the respondents' behavior. The latter is treated with a stimulus and is compared to the placebo group. Traditionally, groups are assigned randomly to restrict self-selection and other biases that can distort inference.

Experimental research is not capable of answering every single question in PolSci. However, some hypotheses can be reasonably tested with experimental or quasi-experimental designs. So, it is worth asking prospective graduates, who do not have much research experience yet, about their vision and ideas for experimental designs. Thereby we can collect inspiring ideas that might be developed into papers later.

Ann-Katrin and Anastasia suggested that the more a person listens to country music, the more likely he or she is to fall into stereotypical views of men and women. They proposed a laboratory experiment, which uses visual aids, e.g., videos and pictures that frame interactions between genders, as a treatment for pre-and post-testing of experiment participants. Building inference based on both static and dynamic portrayals of inter-gender interactions would increase the external validity of the laboratory experiment conducted both in static and dynamic settings.
Dima, Natalja, Rebecca, and Martha came up with an idea of testing their hypothesis that men are more likely to be hired for professional positions than women with a natural experiment. They suggested that assessing the HR’s official behavior would not entirely explain the male dominance in top positions because researchers would then not control for the structural social bias of women being primary caregivers of their families. This bias reduces the amount of available free time for women and, therefore, their motivation and capabilities acquire additional qualifications. However, as the social value system is currently in transition, we will soon become participants of a natural experiment that would flip the hiring chances in favor of women. Would you agree?
Tilman investigated whether online social networks make political views more extreme with a natural experiment. He suggested that the critical difference between virtual and genuine verbal communication is the premise of anonymity that makes legal prosecution difficult, when not impossible in a partially law-free zone. He proposed that a natural experiment would be appropriate for collecting and evaluating data obtained from chats, forums, Twitter, Instagram, and TikTok streams. The quasi-experimental approach could then be reasonably combined with non-random sampling in the form of a snowball strategy, allowing to reach hidden groups.

Find out more by taking a look at my presentation from Fall 2020.

If you are the effect of it, you caused it.

Tue, 20 Oct 2020 00:00:00 +0000

Even when we were kids and did not (yet) know the word causality, probabilistic causal relationships have been sneaking into our lives with remarkable frequency. Think about it for a minute: a seven-year-old you weighing the relative attractiveness of Milky-way against TWIX. What would you pick and why?

Political science is not very different from the chocolate bars' tradeoff - the only distinction is, probably, that the scientists are a tiny bit older and more concerned with the big picture rather than individual eating habits (which does not imply that scientists dislike chocolate - instead, we love it). So, probabilistic causality occurs if a single event or a sequence of repeating episodes like the child’s addition to TWIX bars is somewhat likely to happen, but the probability will lie below 100 percent. Thus, going to the gym or eating a second hamburger for lunch, preferring Joe Biden or Donald Trump (debatable!) - tradeoffs as such are examples of probabilistic but not deterministic outcomes that might generate a significant change in our perceived reality, which we, academics, aim to comprehend with the scientific method.

Identifying and explaining causal mechanisms underlying socio-economic and cultural turbulences are pivotal for social sciences. A researcher should always bear in mind four conditions of causality - association, causal mechanism, time ordering, and no alternative explanations - when approaching complex issues.

The road of causality is long and winding, as Olaf Dekkers once said, which is why one day scholars came up with the so-called “wheel of science” - a plausible narrative for conveying and validating arguments. The wheel is set in motion with a relevant question that needs a defensible answer. To pinpoint the answer, one needs to develop a testable theory that accounts for the association between X and Y - our factors of interest, which we operationalize in hypotheses. After setting up a model that can capture anticipated relationships between X-s and Y-s (remember, the confectionery industry has far more to offer than just Milky-way and TWIX) and designing an appropriate empirical test (will kids always choose TWIX, or can we make Milky-way more lucrative?), one should be able to draw some conclusions accurately. The latter is what we conventionally call inference - a conclusion justified by evidence and logical reasoning.

The next step would be discussing the offered arguments' strengths and weaknesses, showing both sides of the coin. Lastly, we should step out of the comfort zone and conduct robustness checks to show how convincing our results are by confronting them with alternative why-s.

Find out more about finding causality in social sciences by taking a look at my presentation from Fall 2020.