Fellow Econometricians, it is about time I make my debut in our wonderful journal, the SECTOR. Just after I finished my last course of the Econometrics and Data Science Master here at the VU, I managed to be relevant enough to write for the respected readers of the SECTOR. Until now, the most honorable publication was a hardly funny article in the ‘Ecotribune’, Kraket’s informal journal. All this is changing rapidly. This publication in the SECTOR is not the only publication I am striving for. The world where publications rule, academia, is calling me.
Some, maybe most, of you already know me. I entered the Econometrics and Operations Research bachelor’s program here at the Vrije Universiteit in September 2018. To most of us Econometrics students at the VU, the first day of our studies is a vivid memory. I remember entering the Wiskunde&Natuurkunde building for the first time. Everyone was completely lost. Eventually, we managed to reach the lecture hall. Seated in the lecture room, I realized what ‘being lost’ actually meant. Some sadistic soul in the scheduling department, probably embodied by a computer, found it funny to hook us up with Analysis I as our first course. I did not get any of what was told during this first lecture. What was this weird ‘U’ symbol doing on the blackboard? What was this woman doing on a blackboard in the first place? I was six years old when they replaced the blackboards with digital versions at my primary school.
This was only the first impression of what was going to be a difficult year. Just behind getting my driver’s license, which took me five years, obtaining my ‘propedeuse’ was the hardest thing I have ever accomplished. Even the course Linear Algebra, I found difficult at that time. Many of my fellow students experienced the same. We were told over and over that it all would get better after the first half of the year. From time to time the struggling even made me feel insecure. It made me question whether I chose the right studies. Luckily, I found plenty of friends around me in the lecture rooms. We started to find out that studying is not something to be done alone, it is rather a group effort. Mathematics cannot be learned from a book alone. It needs to be discussed, questioned even, to truly grasp it. With this new approach to studying, I managed to pass all my first-year courses within the first year.
During my second year, studying was more natural. I got higher grades with less effort. However, I was still in doubt about my choice of studies. During this time, I got very invested in playing bridge. The card game you might have heard about during your probability course. I became very passionate about bridge, and I wanted to feel this same passion for my studies. It must be something I want to be doing for the rest of my life. The courses so far did not make me feel passionate, but I also did not know what could bring me this feeling of passion. I doubted whether I should apply for a different program or I should focus moreon my side activities. I chose the latter.
This made me decide to put my studies in second place during my third year. Out of convenience, I chose to do the Econometrics minor at the VU, since I knew I would be focusing on other things, like bridge, anyway. Due to the pandemic, I could put more time into my studies than I anticipated. The course Computational Methods taught by Marc Nientker seemed all right, so I decided to take it on. Apparently, I was in for a treat. The course was amazing. The professor knew exactly how to challenge his students. For example, the assignments were very well designed. A good assignment covers much of the content taught in the course, it challenges your programming skills and it helps to come to a deeper understanding of the material. These assignments checked all the boxes.
The course was also the first time I understood why we had to do so much calculus, linear algebra and statistics training in the previous years. Combining all the concepts, made me understand the wonderful concept of bootstrap.
I will take you back to the exact moment, I had this realization and my interest in Econometrics was born. Nientker said: “If you are following all that I am saying, you should be amazed by the results I have just told you about”. The result was about the concept that you can perform an exact test using bootstrap.
Exactly the moment Nientker spoke out these words, I was feeling amazement. That feeling has never left.
After my bachelor's, I continued with the master's in Econometrics and Data Science. Since I enjoyed programming a lot, I chose to do the Data Science track. I was looking forward to my master, but the realization that this was going to be my last year of studying left a bittersweet taste. Because of that, I decided to take on a student job for two days a week and to follow only one of the two courses per period. It was nice to have this variety, but I knew that the student job was not for long. After a couple of months I quit, such that I had time to write my master thesis.
Choosing the right topic is of key importance. Some of you wrote a thesis already and others will have to write one soon. For both your bachelor and master thesis you have very limited time. It might feel like a long period, but I assure you: it is not. Therefore, it is very important to balance challenging yourself and finding something that can be finished in this limited time. The topic of your thesis will have a great impact on the result.
For my bachelor thesis, I chose a topic I was not comfortable with. It was quite a stressful time. For my master thesis, the topic was much more suitable.
For choosing the topic, I had three criteria in the back of my head. Firstly, the topic must be something I am somewhat familiar with. When you choose a topic you know nothing about, it is very hard to oversee the possible challenges. Especially the programming could take much longer than anticipated. That is generally not good for the overall quality of your report. Secondly, the data must be easily available. Finding and processing data yourself is potentially the most cumbersome task of the whole project. You could certainly opt for a topic that does not need empirical data, but if you do want to use empirical data, I would advise using an already existing dataset. Suppose you want to use the data of a certain paper, you could google the title of the table followed by “GitHub”. If the data is openly available, 90% of the time you can find it on the paper’s GitHub. Thirdly, preferably the supervisor has some affection for the topic of interest. Ideally, your supervisor is doing research related to your thesis topic. Although you are working on your thesis individually, you can get more in-depth feedback, if your supervisor is working in the same field.
Keeping this in mind, I eventually settled down on the topic: Growth-at-Risk forecasting with neural networks with Julia Schaumburg as my supervisor.
A quick minilecture on Growth-at-Risk (GaR): Growth-at-Risk is the Value-at-Risk equivalent to GDP growth rates. A 5%-GaR of -2% means that the GDP growth rates are going to be lower than -2% only 5% of the time. The current standard for GaR prediction is the Linear Quantile Regression. My goal was to investigate whether neural networks could be a viable alternative to Linear Quantile Regression.
Since I skipped a couple of courses in the first half of the year, I was not thinking too much about what to do after my master. Usually, students enter the job market after finishing their master thesis, but for me, that was not the case. I was confident something would come on my path naturally. That happened indeed. After my thesis defense, my supervisor offered me a position as a research assistant. We would work with Lukas Hoesch to adapt my thesis into something publishable in an econometrics journal. It was not something I even considered while writing my thesis, but since I enjoyed working on my thesis so much, I happily accepted the position.
We certainly talked things through, before I accepted, but still, it was all very new for me. Quickly, I realized that writing a thesis is something totally different than writing a master thesis. The level of perfection is brought to a whole other level. It is a year ago and we are still working on making it publication ready. Every single choice you make, every tiny thing you did not think of including, might be a reason for a journal to reject your paper. It needs to be perfect. This strive for perfection was quite new for me. It was not something that had been needed before. During my studies, there were usually a couple of mistakes in the reports I handed in and still, you could get a very high grade for it. It is something I have been working on improving. With this, you can really tell the experience of more senior researchers. They have a great eye for detail.
As mentioned, I worked on this project for almost over a year now and since September we have been working on this with the three of us. Slowly but steadily, we are converging into something publishable. Now we are this far, it is common practice to start presenting your work at conferences. The reason behind this is twofold. First, presenting our work at conferences gives fellow researchers the chance to provide you with feedback. It is a great opportunity to show your work to many people and help each other. Second, you want people to know you are working on this project and you are almost finished such that other people who might have the same idea will not bother starting on it as well. It is every researcher’s worst nightmare that someone else comes up with the same idea and comes to a publication faster than you do. Then all the work has been done for nothing.
Not too long ago, I got the opportunity to attend the 2023 edition of the National Econometrics Study Group. This is a yearly conference, attended by researchers from all the Dutch universities with an Econometrics program and the KU Leuven (Belgium). It was the first scientific conference I have ever attended. I got the opportunity to present our work at the poster session. The day consisted of five one-hour blocks of presentations. In between the blocks, there were 20–30-minute breaks where people got the opportunity to walk by the posters. I put a lot of effort into preparing the poster and I practiced an elevator-pitch-like presentation. Every time someone walked past our poster, I had to give a quick introduction to it. There was only limited time, so I tried to explain everything as quickly and efficiently as possible. I felt a bit tense in advance, but it felt very natural when I started giving the mini-presentations. The NESG was a great opportunity to present our work for the first time. Besides, it was wonderful to meet so many econometrics researchers. The conference had a very pleasant atmosphere due to the approachability of everyone involved.
My academic journey does not end here! I will start with a Ph.D. in September. Traditionally, pursuing a Ph.D. in Econometrics involves completing a research master at the Tinbergen Institute. However, alternative paths can also lead to a Ph.D., and I have chosen one such route. Although I have not taken the research master, I will still have the opportunity to take relevant courses at the Tinbergen Institute.
While the exact focus of my research is still somewhat open, it will center around the intersection of time series econometrics and machine learning. The field of machine learning is rapidly expanding, and it is crucial for time series econometrics to keep pace with the latest advancements. My objective during the Ph.D. program is to contribute to this task.
The potential of machine learning is highly promising, and I firmly believe that it can make significant contributions to solving econometrics problems. However, it is crucial to acknowledge the serious concerns associated with its use. One of the main challenges of machine learning lies in its lack of interpretability and explainability. In domains such as healthcare, where machine learning techniques consistently save lives, the benefits often outweigh the drawbacks. However, in the field of econometrics, quantifying the risks of applying machine learning algorithms can be considerably more challenging. In hindsight, there have been many instances where machine learning algorithms have caused more harm than progress. As practitioners, it is our ethical duty to use machine learning responsibly and cautiously.