AØKK08411U Seminar: Applications of Machine Learning in Economics
In recent years, the use of machine learning has become an increasingly popular tool in the toolbox of the modern-day economist, and it has found a significant role to play in many fields, including theoretical econometrics, monetary policy, structural modelling and in particular in applied work.
While a narrow definition may emphasize that the term machine learning covers a set of algorithms that allows a system to improve prediction from data (or experience more broadly), the practical applications and developments associated with machine learning often go hand in hand with new methods for collecting such data. These methods may include webscraping and use of non-tabular datasources such as images and text.
Through participation in this seminar, students will both be given an opportunity to explore advanced applications of machine learning in economics as well as exploring how new sources of data can provide new insights to problems that have already been studied with more traditional empirical approaches and data.
With this seminar, the goal is essentially twofold. First and foremost, we want to give participants an opportunity to apply methods from machine learning and social data science to carry out an independent research project. We believe that having having prior experience with independently applying theoretical knowledge learned from associated courses (see “recommended academic qualifications”) in an empirical project will put participants in a stronger position when it comes to appropriately working with machine learning techniques in a master’s thesis, as part of a PhD or in another type of job.
Secondly, we hope that participants – through working on their research project – will obtain a deeper understanding of the extent at which methods from machine learning and social data science complements empirical approaches from traditional statistics and econometrics. A potential pitfall of applying new approaches that receive lots of praise from academia and the industry is to apply these techniques to poblems where more traditional approaches are already perfectly suitable. Learning to be able to spot a problem where machine learning techniques can generate new insights has a high value and is an end in itself in this seminar.
Participants of the seminar will get a chance to work on a variety of projects of applied nature. Examples of seminar projects include the replication of existing studies using machine learnings methods for causal inference, extending existing research by collecting and analyzing new non-tabular data, and implementing novel machine learning driven methods and showing the relevance of their application in economics.
After completing the seminar the student is expected to be able to fulfill the learning outcome specified in the Master curriculum and to be able to:
• Have reviewed and understood the relevant literature related to the topic of their seminar paper to an extent where the student understands relevant state-of-the-art machine learning approaches while also being able to discuss possibilities and limitations of these.
• Know of state-of-the-art software packages for designing machine-learning systems.
• Can account for the defining features of prediction- and inference problems, and can identify economic problems that can be fruitfully approached via machine learning methods.
• Combine machine-learning methods with established economic methods in e.g. econometrics or structural modelling to enhance methods or generate new insights.
• Choose suitable software to code up specialized machine learning algorithms. This can include using deep learning libraries such as tensorflow, pytorch etc.• Write high-quality code in Python, R or other relevant languages including documentation and testing. (Students should develop the coding skills required to participate in open-source software development)
• Carry through independent research in the intersection of machine learning, data science and economics.
The following list of literature is includes a series of
celebrated examples from the economic literature where machine
learning techniques has yielded new insights. Furthermore, the
papers can hopefully serve as a source of inspiration for new
• Athey, S., Chetty , R., Imbens, G. W., & Kang, H. (2019) The Surrogate Index: Combining Short-Term Proxies to Estimate Long-Term Treatment Effects More Rapidly and Precisely. NBER Working Paper No. w26463.
• Athey, S., & Imbens, G. W. (2019). Machine Learning Methods That Economists Should Know about. Annual Review of Economics, 11, 685–725.
• Athey, S., Tibshirani, J., & Wager, S. (2019). Generalized random forests. Annals of Statistics, 47(2), 1179–1203.
• Belloni, A., Chernozhukov, V., & Hansen, C. (2014). High-dimensional methods and inference on structural and treatment effects. Journal of Economic Perspectives, 28(2), 29–50.
• Gebru, T., Krause, J., Wang, Y., Chen, D., Deng, J., Aiden, E. L., & Fei-Fei, L. (2017). Using deep learning and google street view to estimate the demographic makeup of neighborhoods across the United States. Proceedings of the National Academy of Sciences of the United States of America, 114(50), 13108–13113.
• Hansen, S., McMahon, M., & Prat, A. (2018). Transparency and deliberation within the FOMC: A computational linguistics approach. Quarterly Journal of Economics, 133(2), 801–870.
• Hartford, J., Lewis, G., Leyton-Brown, K., & Taddy, M. (2017). Deep IV: A flexible approach for counterfactual prediction. In 34th International Conference on Machine Learning, ICML 2017 (Vol. 3, pp. 2257–2268). International Machine Learning Society (IMLS).
• Hastings, J. S., Howison, M., & Inman, S. E. (2020). Predicting high-risk opioid prescriptions before they are given. Proceedings of the National Academy of Sciences of the United States of America, 117(4), 1917–1923.
• Jean, N., Burke, M., Xie, M., Davis, W. M., Lobell, D. B., & Ermon, S. (2016). Combining satellite imagery and machine learning to predict poverty. Science, 353(6301), 790–794.
• Kouw, W. M., & Loog, M. (2018). An introduction to domain adaptation and transfer learning. https://arxiv.org/abs/1812.11806
• Nie, X., & Wager, S. (2017). Quasi-Oracle Estimation of Heterogeneous Treatment Effects. Retrieved from http://arxiv.org/abs/1712.04912
- Advanced Microeconometrics
- Seminar in Advanced Microeconometrics
- Dynamic Programming - Theory, Computation, and Empirical Applications
- Topics in Social Data Science
- Social Data Science: Econometrics and Machine Learning
- Summer School in Social Data Science
Note: Several of the Social Data Science courses have changes names over the years. If you have followed an equivalent course under a different name that is equally valid.
Finally, we expect that all seminar projects will have a strong programming components. Students are expected to be able to
work in Python, R and other relevant languages depending on their projects requirements.
- identify and clarify a problem,
- seek and select relevant literatur,
- write a academic paper,
- present and discuss own paper with the other students at the seminar.
Mandatory activities in the seminar:
- Kick-off meeting
- Finding literatur and defining the project
- Writing process of the seminar paper
- Presentation of own project and paper
- Giving constructive feedback to another student´s paper
- Actively participating in discussions at the presentations and other meetings.
The aim of the presentations is, that you use the presentation as an opportunity to practice oral skills and to receive feedback at the paper. The presentations are not a part of the exam and will not be assessed.
The seminar project paper must be uploaded in Absalon before the presentations, as the opponents and the other seminar participants have to read and comment on the paper. It is important that you upload a paper that is so finalized as possible due to the fact that the value of feedback and comments at the presentation is strongly associated with the skill level of the seminar paper.
The teacher defines what materials may be used for the presentations.
After the presentations, you can with a few corrections improve the seminar paper by including the feedback and comments emerged during the presentations. It is NOT intended that you rewrite or begin the writing of the seminar paper after the presentation has taken place.
In case of a pandemic like Corona the teaching in this seminar may be changed to be taught either fully or partly online. For further information, see the course room on Absalon.
• Kick-off meeting: February 9, at 10.15-12.00
• Extra days of introducing teaching: February 25, at 10.15-12.00
• Workshops/ Presentations meetings: May 11-12
It is strongly recommended that you think about and search for a topic before the semester begins, as there is only a few weeks from the kick-off meeting to the submission of the project description/agreement paper.
There is no weekly teaching/lecturing and the student cannot expect guidance from the teacher. If the teacher gives a few introduction lectures or gives the opportunity for guidance, this as well as other expectations are clarified at the kickoff meeting.
All information regarding the seminar is communicated through Absalon including venue. So it is very important that you by yourself logon to Absalon and read the information already when you are registered at the seminar.
- Project work
• Each student receives individually oral feedback on the paper
and at the presentation from peers and teacher.
• The teacher gives the students collective oral feedback and individual guidance.
- 7,5 ECTS
- Type of assessment
- Written examinationA seminar paper in English that meets the formal requirements for written papers stated in the curriculum of the Master programme and at KUNet for seminars.
- Exam registration requirements
- All aids allowed
for the seminar paper.
The teacher defines the aids that must be used for the presentations.
- Marking scale
- 7-point grading scale
- Censorship form
- External censorship
- Exam period
Deadline for submitting the final seminar paper: June 1, 2021 before 10 AM
The seminar paper must be uploaded to the Digital Exam portal. More information will be available from the middle of the semester.
The reexam is a written seminar paper as stated in the Master curriculum.
Criteria for exam assesment
Students are assessed on the extent to which they master the learning outcome for the seminar and can make use of the knowledge, skills and competencies listed in the learning outcomes in the Curriculum of the Master programme.
To receive the top grade, the student must with no or only a few minor weaknesses be able to demonstrate an excellent performance displaying a high level of command of all aspects of the relevant material.