Workshop‎ > ‎

BZ 2016

Pension Finance, Asset-liability Management, Asset Allocation under Parameter Uncertainty
organized by
Alex Weissensteiner, Free University of Bozen-Bolzano, Italy

venue: Free University of Bolzano - Room E-423
date: Sep 2, 2016
map: see below

Invited speakers
Thomas Dangl - TU Wien
Andreas Hamel - Free University of Bolzano
Michael Hanke - University of Liechtenstein
Holger Kraft - Goethe University Frankfurt
Teemu Pennanen - King's College London
Rolf Poulsen - University of Copenhagen
Mogens Steffensen - University of Copenhagen
Sebastian Stoeckl - University of Liechtenstein

Thursday, Sep 1 come together

19:30 meeting in the lobby of the Hotel Stiegl  
20:00 dinner at Torgglhaus 

Friday, Sep 2 scientific program

09:00 - 09:15 Welcome
09:15 - 10:00 Financial Turbulence and Aggregate Stock Returns (S. Stoeckl)
10:00 - 10:45 Predictors and Portfolios over the Life Cycle (H. Kraft)
10:45 - 11:15 coffee break
11:15 - 12:00 ROM Simulation with Exact Means, Covariances, and Multivariate Skewness (M. Hanke)
12:00 - 12:45 Value-At-Risk for Multivariate Positions - A Set-Valued Approach (A. Hamel)

12:45 - 14:15 lunch break

14:15 - 15:00 Hedge Funds Don’t Hedge – And 50-Odd Other Odd Things in Finance (R. Poulsen)
15:00 - 15:45 Asset Valuation via Optimal Investment (T. Pennanen)
15:45 - 16:15 coffee break
16:15 - 17:00 Long-term Asset Allocation under Parameter Uncertainty and Ambiguity Aversion (T. Dangl)
17:00 - 17:45 On the Separation of Preferences for Risk and Substitution (M. Steffensen)
17:45 - 18:00 discussion

19:30 dinner at Voegele 


Abstract of invited talks:

Long-term Asset Allocation under Parameter Uncertainty and Ambiguity Aversion
Thomas Dangl (Technical University of Vienna) and Alex Weissensteiner
This paper focuses on the implications of predictability and parameter uncertainty on the optimal asset allocation of ambiguity-averse long-term investors. We study the term-structure of the multivariate risk-return trade-off in a VAR model under full consideration of parameter uncertainty and decompose the predictive covariance along different sources of risk/uncertainty. We calibrate the model to real returns of US stocks, US long-term government bonds and real-estate using term spread and dividend yield as additional predictive variables. While over short periods the model-implied conditional covariance structure of asset-class returns determines the optimal allocation, we find that over longer horizons, optimal asset allocation is primarily determined by the covariance structure induced by estimation errors. As a consequence, the ambiguity-avers long-term investor tilts her portfolio not simply towards the global minimum-variance portfolio but shrinks portfolio weights towards a seemingly inefficient portfolio which shows maximum robustness against estimation errors.

Value-At-Risk for Multivariate Positions - A Set-Valued Approach
Andreas Hamel (Free University of Bozen-Bolzano) and Daniel Kostner
What is the median (a.k.a. the 50% quantile) of an at least bivariate random variable? Surprisingly, there does not seem to be a "standard" answer to this simple question. At the heart of the problem lies the non-totalness of order relations in higher dimensional spaces. In this talk, a new cdf-like function for multivariate random variables is proposed which is inspired by Tukey depth functions. It involves a general ordering cone and leads to set-valued quantiles which share basically all properties with scalar quantiles. A corresponding set-valued value-at-risk is then defined and compared to existing concepts from the literature. It has a convincing financial interpretation and also shares all the (dis)advantages with its scalar little brother. Finally, first order stochastic dominance for multivariate random variables is introduced. The basic methodology consists in applying recently developed concepts from set optimisation to statistics. This shows that "going set-valued" admits to obtain concepts and results strikingly parallel to the scalar case which cannot not be obtained using "vector orders/constructions."

ROM Simulation with Exact Means, Covariances, and Multivariate Skewness
Michael Hanke (University of Liechtenstein), Spiridon Penev, Wolfgang Schief and Alex Weissensteiner
We develop a simulation algorithm that generates multivariate samples with exact means, covariances, and multivariate skewness. If required for financial applications, absence of arbitrage can be ensured. We use the Kollo measure of multivariate skewness, which is more informative than the Mardia skewness previously used in this context. Potential applications include the simulation of risk factors for the risk management of financial institutions.

Predictors and Portfolios over the Life Cycle
Holger Kraft (Goethe University Frankfurt), Claus Munk and Farina Weiss
We show that the net corporate payout yield predicts both the stock market index and house prices and that the log home rent-price ratio predicts both house prices and labor income growth. We incorporate the predictability in a rich life-cycle model of household decisions involving consumption of both perishable goods and housing services, stochastic and unspanned labor income, stochastic house prices, home renting and owning, stock investments, and portfolio constraints. We find that households can significantly improve their welfare by optimally conditioning decisions on the predictors. For a modestly risk-averse agent with a 35-year working period and a 15-year retirement period, the present value of the higher average life-time consumption amounts to roughly $179,000 (assuming both an initial wealth and an initial annual income of $20,000), and the certainty equivalent gain is around 5.5% of total wealth (financial wealth plus human capital). Furthermore, every cohort of agents in our model would have benefited from applying predictor-conditional strategies along the realized time series over our 1960-2010 data period.

Asset Valuation via Optimal Investment
Teemu Pennanen (King's College London)
We study optimal investment and the valuation of assets whose payouts cannot be replicated by trading other assets. Our market model allows for constraints and illiquidity effects that are encountered in practice. We review two hedging-based notions of asset value and relate them to the classical notions of risk neutral and net present values. Many classical results e.g. on attainability and duality are extended to the illiquid market setting. The techniques are illustrated with applications to pensions, options and money markets.

Hedge Funds Don’t Hedge – And 50-Odd Other Odd Things in Finance
Rolf Poulsen (University of Copenhagen)
The information in this paper is basically a wealth of quantitative finance trivia. It may seem frivolous or pedantic, i.e. very much inside baseball stuff. However, quantitative finance is an area with a large inflow of people with diverse backgrounds, and while this absence of old school ties is generally refreshing, it does mean that there is little common ground, and that may sometimes cause misunderstandings that are not simply amusing, but a genuine obstacle.
In rough terms, the information can be grouped into three categories of secret handshakes: (1) Conceptual misnomers or infrequently asked questions. These are things or terms that we have grown so used, that we forget why do or say so, and that the latter might make little sense. (2) False friends and (my) pet peeves. (3) Plain coincidences, especially regarding people and names.

On the Separation of Preferences for Risk and Substitution
Mogens Steffensen (University of Copenhagen)
We formalize a global objective under separation of preferences for risk and intertemporal substitution. We discuss its connection with stochastic differential utility (time-continuous recursive utility) which is based on local separation. For a Merton market the optimal decisions with respect to consumption and investment coincide. We consider two more general markets and characterize the solutions for these markets. In one case we study an incomplete market by adding an extra state process. In another case, we study the effects from an uncertain lifetime and access to life insurance. The latter gives new insight in how, possibly, an endogenous demand for hump-shaped consumption can arise even with 'fair' pricing of insurance. This has important implications for product design and product advice in the pension industry. Finally, we discuss briefly how frictions in the insurance and pension market may, or may not, alter the conclusions and how to elicit preferences from policyholders.

Financial Turbulence and Aggregate Stock Returns
Sebastian Stoeckl (University of Liechtenstein)
In this paper I show that financial turbulence (FT) is as strong a predictor for the equity premium as short interest (SI), which is the best performing predictor up to date (Rapach et al., 2016). However, FT has the advantage of not needing external (fundamental) information for its calculation, as it depends solely on the distribution of the cross-section of stock returns. Additionally, FT substantially outperforms a multitude of popular variables (including SI) in predicting future volatility both in-sample as well as out-of-sample. Combining these two findings, I show that a mean variance investor relying on predictions from FT achieves significantly higher certainty equivalent returns and Sharpe ratios than by using any other of these variables.

Participation is free of charge - but please register by sending an email to:

Arrival via train: both the university and the hotel are within walking distance
Alex Weissensteiner,
Jun 29, 2016, 9:39 AM