Seminar "Predictability"


The seminar is open to Master students.

To successfully pass the seminar you need to write a paper and give a presentation. Papers can be written in either German or English and should have a length of 15-20 (team of two) or 20-25 pages (team of three). For hints on how to write a paper see our guidelines. You need to hand in a printed version and also a digital one (PDF). The seminar talks should be given in English.

The two main parts of your paper and presentation will be (i) explanation of the analytical methodology and (ii) replication of analyses from the key references. You should also provide an introduction, a short summary of the literature (which can be part of the introduction), and some concluding remarks. For most topics, you will need to use a software such as R or Matlab.

Please contact your supervisor to discuss the outline of your paper, your empirical part, and any questions that you may have. For organizational questions, please ask Nenad Ćurčić.

FAQ & Organisational matters

  • Do we get a grade? Yes. Your paper and your presentation will be graded and lead to one grade (equally weighted). Both the paper and presentation have to be passed.
  • What do we have to hand in? An outline of your paper to discuss the content of your paper and your final paper one week before the presentation.
  • Who is responsible? For content-related questions, please contact your supervisor. For organizational questions, please ask Nenad Ćurčić.

Time Table

  1. 29.01.2018 - 03.02.2018 Submission of your seminar preferences via online platform:
  2. 04.02.2018 First round of seminar matching
  3. 11.02.2018 Second round of seminar matching
  4. 15.02.2018 Topic allocation (sort the topics on Taddle until 22.02. 11:55 pm)
  5. 27.02.2018 General information about Seminar, introductory meeting, 4:30 pm, room: He18, 1.20
  6. 03.04.2018.-17.04.2018. Registration at the Higher Services Portal
  7. until 20.04.2018. Meet your supervisor to discuss the outline of the paper
  8. 14.06.2018. Submission of the paper until 11:59 am, HeHo 18, room 1.00 (to Nenad)
  9. 22.06.2018.-23.06.2018 Presentations, HeHo 18 (exact schedule and room tba)


1.       OOS Tests

Explain the differences between in-sample and out-of-sample (OOS) tests and introduce different statistics used to evaluate OOS model performance. Then show how to assess the economic significance out-of-sample. Illustrate your entire explanations by updating the OOS analyses from Table 1 and 3 in Goyal and Welch (2008) for the variable dividend price ratio. Usage of Excel is fine.

Key references:

  • Welch, I., & Goyal, A. (2008). A Comprehensive Look at the Empirical Performance of Equity Premium Prediction. Review of Financial Studies, 21, 1455-1508.
  • Campbell, J. Y., & Thompson, S. B. (2008). Predicting Excess Stock Returns Out of Sample: Can Anything Beat the Historical Average? Review of Financial Studies, 21, 1509–1531.

supervisor: Nenad Ćurčić

students: Javier Alberto Almeida García and Kevin Pubantz


2.     Predictability in Efficient Markets

Goyal and Welch (2008) showed that many predictors of the stock market fail after some time. Your first task is to update Table 1 of Goyal and Welch (2008) for all years with data available. Up to which date does each variable show a positive OOS R-squared? Discuss this question by updating Figure 1. Your second task is to debate possible causes why predictors fail after some time. In your discussion, you can focus on predictability patterns that can arise in a  market that is fully or largely efficient.

Key references:

  • Mclean, R.D., & Pontiff, J. (2016). Does Academic Research Destroy Stock Return Predictability?, The Journal of Finance, 71(1), pp. 5-32.
  • Welch, I., & Goyal, A. (2008). A Comprehensive Look at the Empirical Performance of Equity Premium Prediction. Review of Financial Studies, 21, 1455-1508.
  • Timmermann, A., & Granger, C. (2004). Efficient market hypothesis and forecasting. International Journal of Forecasting, 20, 15-27.

supervisor: Nenad Ćurčić

students: Jhonatan Lopez and Shaurya Shaurya


3.       Imposing restrictions

In order to increase prediction performance, some authors impose economically motivated restrictions. Your first task is to explain and compare restriction approaches with the focus on restrictions for stock return regression forecasts proposed by Campbell and Thompson (2008) as well as forecasts involving valuation ratios by Ferreira and Santa-Clara (2011). As a second task you should update the analysis done for the dividend-price ratio and reported in Table 1 in Campbell and Thompson (2008). Usage of Excel is recommended.

Key references:

  • Campbell, J. Y., & Thompson, S. B. (2008). Predicting Excess Stock Returns Out of Sample: Can Anything Beat the Historical Average? Review of Financial Studies, 21, 1509–1531.
  • Ferreira, M.I., & Santa-Clara, P. (2011). Forecasting stock market returns: the sum of the parts is more than the whole. Journal of Financial Economics, 100, 514–537.

supervisor: Nenad Ćurčić

students: Hanliang Guo, Wenqi Wu and Xiaoyu Di


4.       Forecast combination

Many economic variables with in-sample predictive ability for the equity premium fail to deliver consistent out-of-sample forecasting gains relative to the historical average. However, forecasts combination is a relatively simple forecasting technique that seems to work well for equity premium prediction. You shall review the evidence and then update panel A from Table 1 in the study of Rapach et al. (2010) using data until this year focusing on the OOS R-squared for individual predictive regression models and the mean combination method. Moreover, you should also update the first graph from Figure 2. Data sources and an Excel template (recommended software is Excel) will be provided to you. Time permitting, you can update more.

Key reference:

  • Rapach, D., et al. (2010). Out-of-Sample Equity Premium Prediction: Combination Forecasts and Links to the Real Economy. Review of Financial Studies, 3, 821-62.

supervisor: Nenad Ćurčić

students: Aleksei Minkov and Artem Chekin


5.       Predicting stock market returns with the least-angle regression (LARS) approach

When using an OLS approach for prediction, it might be tempting to use all sort of available data as explanatory variables. This approach is often referred to as the “kitchen sink” approach. A major drawback of this approach is that it is likely to suffer from overfitting and thus leads to poor predictive performances. Shrinkage approaches like the ones used by Li & Tsiakas can tackle this issue and improve the predictive performance.

A different but surprisingly very similar way of addressing the issue is a least-angle regression (LARS).

Your task is to introduce the concepts of LARS on the one hand and shrinkage approaches on the other. The focus should be on a detailed explanation of LARS. Then use LARS to predict returns for the same data and time period as in section 3 of the 2017 paper from Li and Tsiakas. Compare your results to those from Li and Tsiakas.

Usage of R is recommended. The following R package might be useful: lars

Key references:

  • Li, J., & Tsiakas, I. (2017): Equity premium prediction: The role of economic and statistical constraints. Journal of Financial Markets, 36, pp. 56-75.
  • Efron, B. et al. (2004): Least angle regression. The Annals of Statistics 32(2), pp. 407-499

supervisor: Carsten Schäfer-Siebert

students: Ivan Subota and Oleg Ryzhkov


6.       Tower building and stock market returns

Summarize the empirical findings on the relationship between skyscraper building on the one hand, and economic activity and stock market returns on the other hand. Then update Table 1, Figure II, Figure III, Table 2 (without MAE) and variation 6 from Table 5 of Löffler (2013). Time permitting you can update more. If you have trouble implementing the Hodrick estimator you can use Newey/West. For stock market data, use the data sources used in the paper and possibly others according to the directions we give in the template. Tower building data will be provided to you. Both R and Excel are fine.

Key reference:

  • Löffler, G. (2013). Tower building and stock market returns. Journal of Financial Research, 36(3), 413-434.

supervisor: Prof. Dr. Gunter Löffler

students: Christoph Vagts and Yannick Schwarz


7.       Predicting stock returns with idiosyncratic volatility

Idiosyncratic volatility belongs to the class of implicit data. This means that idiosyncratic volatility is not directly observable but must be estimated based on a model, which will be a key part of your analysis.
Since one can theoretically get rid of idiosyncratic volatility using diversification, it is not obvious that it should be priced. Yet, there are quite a few papers which link idiosyncratic volatility to the cross‑section of returns.

Your task is to replicate parts of the key reference paper: Investigate the predictive power of idiosyncratic volatility for the stocks listed in the STOXX Europe 600 index in January 2015. Use the STOXX Europe 600 as your regional reference and to calculate the Fama-French factors, as indicated in the template that will be handed to you. Use daily data from 2015 up to now.
You shall also give a brief overview about the meaning of idiosyncratic volatility for asset pricing and discuss why idiosyncratic volatility might have predictive power for stock returns.

Usage of R is recommended.

Key reference: 

  • Ang, A. et al. (2009): High idiosyncratic volatility and low returns: International and further U.S. evidence. Journal of Financial Economics 91(1), pp. 1-23

available at

supervisor: Carsten Schäfer-Siebert

students: Di Chen and Linjie Li


8.       Predicting index volatility with GARCH models

When prediction is performed through parametric models, one has to choose how exactly the model should look like. For example, there are a lot of different GARCH models available, which all could be used to predict the volatility of a stock index.

Have a look at the study of Awartani and Corradi. Use daily DAX total return index data to replicate parts of their study: Compare the predictive performances of the introduced GARCH specifications over different periods, forecast horizons and for different test statistics. You shall introduce RiskMetrics and at least 5 different further GARCH specifications, may restrict yourself to 3 different choices to model the conditional mean and may assume p = q = 1 for all considered GARCH models. You don’t need to implement White’s reality check.

Usage of R is recommended. The following R packages might be useful: rugarch, forecast

Key reference:

supervisor: Carsten Schäfer-Siebert

students: Dorin Calugaru and Emanuele Luzzi


All relevant information to be found on this page. Students will be contacted via e-mail occasionally.

Dates and Room

Please note the detailed timetable.

Module description

This seminar is open for Master students.

Module description