Contributions published at Financial Economics (Felix Kübler)

Deficit finance, aka pay-go policy, is free when growth rates routinely exceed safe government borrowing rates. Or so many say. This note presents four counterexamples based on four versions of a simple OLG economy. In each version the growth rate exceeds the safe rate for one of four reasons – uninsured idiosyncratic risk, uninsured aggregate risk, policy uncertainty, and imperfect financial intermediation. Deficit finance does not directly address any of these problems. What works, respectively speaking, is progressive taxation, bilateral intergenerational risk-sharing, early policy resolution, and improved intermediation. The four examples thus show that seemingly free deficits may be more costly than they appear. Indeed, inefficient pay-go policy can even lower the government’s borrowing rate, encouraging yet more deficit finance.

We introduce deep equilibrium nets---a deep learning-based method to compute approximate functional rational expectations equilibria of economic models featuring a substantial amount of heterogeneity, significant uncertainty, and occasionally binding constraints. Deep equilibrium nets are neural networks that directly approximate all equilibrium functions and that are trained in an unsupervised fashion to satisfy all equilibrium conditions along simulated paths of the economy. Since the neural network approximates the equilibrium functions directly, simulating the economy is computationally cheap, and training data can be generated at virtually zero cost. We demonstrate that deep equilibrium nets can solve rich and economically relevant models accurately by applying them to solve three different models, all featuring a very high-dimensional state space. Specifically, we solve two overlapping generations models with aggregate and idiosyncratic uncertainty, illiquid capital, a one-period bond, and occasionally binding constraints. Additionally, we solve a Bewley-style model with a continuum of agents, aggregate and idiosyncratic risk, borrowing constraints, and recursive preferences.

One of the approximately 700 things that make climate change a knotty problem is that fighting it requires people living today to do things for the benefit of future generations. And, you know, what have future generations ever done for us?

null

Das Umschichten von Portfolios kann unter gewissen Umständen einen Mehrwert gegenüber dem einfachen Kaufen und Halten, auch Buy and Hold Ansatz genannt, von Anlageprodukten generieren. In der daliegenden Bachelorarbeit werden elf Portfolios, bestehend aus den Anlagekategorien Liquidität, Obligationen, Aktien sowie Alternative Anlagen zusammengestellt. Für jede Portfoliozusammensetzung wird der Buy and Hold Ansatz, das monatliche, quartalsweise sowie jährliche Rebalancing anhand von drei Kennzahlen beurteilt. So werden optimale Rebalancing-Strategien anhand der kumulierten Rendite, der Volatilität sowie der Sharpe Ratio definiert. Die Resultate zeigen, dass ein zu häufiges Rebalancing, wie die monatliche Rebalancing-Frequenz, bei allen Portfoliozusammensetzungen angesichts der Rendite, des Risikos sowie der Sharpe Ratio zu schlechteren Ergebnissen führt. Ab einem gewissen Aktienanteil im Portfolio weist die jährliche Rebalancing-Strategie gegenüber dem Buy and Hold Ansatz eine Outperformance auf. Bei den Untersuchungen der Volatilität kann nur bei Portfolios mit einem hohen Anteil an Obligationen eine Risikominderung durch das Rebalancing gezeigt werden. Weist das Portfolio einen hohen Anteil an Aktien auf, so steigt die Volatilität und die Buy and Hold Strategie weist die geringste Standardabweichung auf. Bei der risikoadjustierten Performancekennzahl der Sharpe Ratio zeigte die jährliche Rebalancing- Strategie die besten Resultate.

This work examines deep reinforcement learning for optimal asset allocation in the US stock market using a custom implementation of the proximal policy optimization algorithm (PPO). Two different policy distributions, namely the Gaussian and Dirichlet distribution are tested as well as two different network architectures, namely a feed-forward neural network and the latter combined with an additional LSTM module, with the aim to provide the agent with more information about the past. The portfolio rebalancing is done under consideration of constraints such as transaction costs, budget constraints, no short-selling constraints and no fractional shares trading. The state space is constructed including technical momentum indicators, past returns, volatility and the volatility index (VIX). The best-performing strategy is PPO using a Gaussian policy with a feed-forward neural network, while the Dirichlet policy is found to enforce equal weights under all circumstances. I

Erdöl und Gold, zwei Rohstoffe, welche ihren Marktsektor seit Jahrzehnten dominieren, gehören zu den meistgehandelten Gütern der Welt. Beide verfügen jedoch über eine Eigenschaft, die ihre zukünftige Existenz gefährdet. Bei den Rohstoffen handelt es sich um nicht erneuerbare Ressourcen, welche während unserer Lebzeit nicht natürlich wiedererscheinen werden. Der Rückgang in Erdöl- und Goldlagerstätten ist in den letzten Jahren vermehrt ins Rampenlicht der Wissenschaft geraten. In der Forschung von Speirs et al. (2015) wird auf die Unsicherheit bei der Verfügbarkeit von natürlichen Ressourcen wie Erdöl und Metalle eingegangen. Weitere Studien versuchen die Trends von Lebenszykluskurven bei der Produktion von nicht erneuerbaren Ressourcen zu modellieren und vorherzusagen (Semenychev et al., 2014). Diese Arbeit versucht den Fokus auf die direkt vom Rückgang betroffenen Produzenten des jeweiligen Rohstoffes zu legen. Die unmittelbare Abhängigkeit von aussterbenden Gütern machen die Produktionsunternehmen zu interessanten Forschungsobjekten. Die komplexe Beziehung zwischen einem Basisobjekt und seinem Produzenten wurde bereits exzessiv untersucht. Im Rahmen dieser Arbeit wird auf bereits bestehenden Studien aufgebaut, beispielsweise auf die Studien von Baur (2014) und Sadorsky (2008), welche die erwähnte Beziehung bei Gold respektive Erdöl beobachteten. Durchgeführte Forschungen bestätigen die Existenz einer Abhängigkeit zwischen den Gütern und ihren Produzenten. Ein erhofftes Nebenergebnis dieser Arbeit wird die direkte Vergleichbarkeit von zwei wichtigen Gütern sein. Das Hauptergebnis befasst sich mit der Beantwortung der Nullhypothese «Handelt es sich bei Aktienkursen von Rohstoffproduzenten um vorauseilende Indikatoren für den Preis ihres jeweiligen Basispreises?». Ein vorauseilender Indikator ist eine Variante der volkswirtschaftlichen Indikatoren. Per Definition handelt es sich bei ihnen um Messgrössen, welche die Analyse und Vorhersage von wirtschaftlicher Leistung ermöglicht (Frumkin, 2005).

Viele Anleger halten Gold aus Diversifikationsgründen in ihrem Portfolio. In diesem Zusam-menhang soll Gold nicht nur unter normalen Marktbedingungen als Hedge gelten, sondern ge-rade in Krisenzeiten auch Safe-Haven-Eigenschaften haben, um Preiseinbrüche im Portfolio ausgleichen zu können. Verschiedene Studien (Areal, Oliveira, & Sampaio, 2013; Baur und McDermott,2010) zeigen, dass Gold sowohl ein Hedge als auch ein Safe-Haven für US-ameri-kanische und europäische Aktienmärkte ist. Anderseits liefern andere Untersuchungen Belege dafür, dass die Korrelationen zwischen finanziellen Vermögenswerten, in Zeiten von Finanz-krisen tendenziell zunehmen (Campbell, Koedijkund, & Kofman, 2002; Erb, Harvey, & Vis-kanta, 1994; Chesnay & Jondeau, 2001). Aufgrund dieser unstimmigen Ergebnisse wird in die-ser Bachelorarbeit die Hedge-Eigenschaft und insbesondere die Safe-Haven-Eigenschaft von Gold und Goldminenaktien rund um den Zeitraum der Covid-19 Krise untersucht. Dabei wur-den im Multifaktorenmodell die Goldrisikoprämie auf die Faktoren von Fama und French (1993) regressiert. Die Ergebnisse konnten zeigen, dass Gold ein Hedge gegenüber Aktien-märkten war, allerdings kein Safe-Haven während der Covid-19 Periode. Nichtsdestotrotz konnten Gold und Goldminenaktien, Safe-Haven-Eigenschaften gegenüber Large-Cap Aktien und Value Aktien aufweisen.

null

Anthropogenic climate change produces two conceptually distinct negative economic externalities. The first is an expected path of climate damage. The second, the focus of this paper, is an expected path of economic risk. To isolate the climate-risk problem, we consider three mean-zero, symmetric shocks in our 12-period, overlapping generations model. These shocks impact dirty energy usage (carbon emissions), the relationship between carbon concentration and temperature and the connection between temperature and damages. By construction, our model exhibits a de minimis climate problem absent its shocks. However, due to non-linearities, symmetric shocks deliver negatively skewed impacts, including the potential for climate disasters. As we show, Pareto-improving carbon taxation can dramatically lower climate risk, in general, and disaster risk, in particular. The associated climate-risk tax, which is focused exclusively on limiting climate risk, can be as large as, or larger than, the carbon average-damage tax, which is focused exclusively on limiting average damage.

This paper examines the presence of calendar anomalies in the Chinese and Hong Kong stock markets during the period of January 2000 and December 2019. The logarithmic daily returns are tested for evidence of the day-of-the-week effect and the month-of-the-year effect by the methods of linear regression model and EGARCH model. The findings show that the calendar effects are present in both the Chinese and Hong Kong stock markets. While the Tuesday, Thursday and February effects are observed in the Chinese stock markets, the Monday effect and positive returns in the first month of each quarter are observed in the Hong Kong stock market. Additionally, there is strong evidence of leverage effect in both the Chinese and Hong Kong stock markets. Further investigation reveals that investor sentiment can at least partly explain the calendar effects in the Chinese stock markets.

Carbon taxation is mostly studied in social planner or infinitely lived‐agent models, which obscure carbon taxation's potential to produce a generational win win. This article's large‐scale, dynamic 55‐period, overlapping generations model calculates the carbon tax policy delivering the highest uniform welfare gain to all current and future generations. Our model features coal, oil, and gas, increasing extraction costs, clean energy, technical and demographic change, and Nordhaus' carbon/temperature/damage functions. Assuming high‐end carbon damages, the optimal carbon tax is $70, rising annually at 1.5%. This policy raises all generations' welfare by almost 5%. However, doing so requires major intergenerational redistribution.

The replacement of positive with normative economics has left climate policy in its sorry state – as a fight between generations, across regions, and even among economists over climate justice. This column uses a multi-region, overlapping generations model of climate change to study climate policy as an externality whose resolution can uniformly and equally benefit all humankind, regardless of year or place of birth. The optimal uniform welfare-improving policy, implemented via a time-varying global carbon tax plus region- and generation-specific net transfers, can materially limit global emissions, dramatically shorten the use of fossil fuels, and raise the welfare of all current and future agents by over 4%.

We replicate the approach of Chen, Pelger, and Zhu (2019) and apply a generalized non-linear asset pricing model to a large set of U.S. equity and macroeconomic data. We conrm a major part of their results for an extended time period and reduce the portfolio turnover of the model. Furthermore, we leverage a subset of their neural network architecture and propose an idea for an alternative approach to traditional portfolio optimization, which does not rely on estimated mean returns and covariance matrices as input parameters. We compare both models in detail. Based on the obtained results, our model cannot compete with the original approach in terms of realized Sharpe ratios, however it does exhibit a signicantly lower turnover ratio and higher explained variation. I

Diese Arbeit beinhaltet die Anwendung zweier aktiven Anlagestrategien. Anhand der Titel des UBS 100 Index wird versucht, eine Value und eine Simple Moving Average (SMA) Strategie umzusetzen, um zu sehen, ob durch aktives Anlegen der Index outperformt werden kann. Für die Berechnungen wird ein Zeitraum von 18 Jahren (2002 - 2020) verwendet. Die Ergebnisse der Anlagestrategien werden mit der Performance des UBS 100 Index verglichen. Ausserdem werden insbesondere die erzielten Renditen in der Finanz- und der Coronakrise betrachtet. Ist eine der beiden Strategien speziell geeignet für die Anwendung in einer schwierigen Wirtschaftslage? Für die Anwendung der Value Strategie wurde durch das Price-to-Book Ratio entschieden, welche Aktien gekauft werden. Bei der Anwendung der SMA Strategie wurde der 50- und 200-Tage Durchschnitt verwendet. Während die Resultate der Value Strategie sehr positiv sind, ist die SMA Strategie stark von den Transaktionskosten abhängig. In den beiden untersuchten Krisen schneiden allerdings sowohl die Value wie auch die SMA Strategie nicht sonderlich gut ab.

This thesis investigates the market timing ability of the VSMI with two di↵erent strategies for the Swiss market using data from July 2011 until June 2019. The first strategy is based on size and di↵ering between small and big stock portfolios, whereas the second strategy consists of alternating between value and growth stock portfolios and therefore is constructed on the style of the stocks. For both strategies, the portfolios are for robustness formed in two ways according to Fama and French (1993), with the first setting being a 2-by-3 approach and the second a 5-by-5 approach. The empirical results of this thesis then suggest a profitable strategy based on size, where the big stocks outperform the small stocks on days following an increase in the VSMI. This superiority is especially profitable on days following a particularly high increase in the VSMI of over 30% and results in a significant payo↵ when implementing a zerocost portfolio. Overall the strategy based on size clearly outperforms the market portfolio. The reasoning for this is that the VSMI is a valid timing instrument for size investing and indicates a ”flight to quality” and ”flight to liquidity” movement. On the other hand, the empirical results show that the style investing strategy reacting to changes in the VSMI is not equally profitable and delivers mixed results that are not significant. According to the results, the value stocks outperform the growth stocks on days following minor positive increases in the VSMI, but on days following huge increases, the opposite is the case. This observation might be due to the popularity of style investing and the fact that these portfolios react faster to changes in the VSMI, and therefore, the VSMI loses its timing ability.

The findings of this master’s thesis provide evidence on the profitabil-ity of momentum strategies when focusing on liquid stocks in aggre-gated market data. However, when potentially illiquid stocks are also considered, the profitability of most strategies vanishes favoring the use of contrarian strategies. The thesis could not corroborate momen-tum in every country and shows large discrepancies across markets. While Indian stocks delivered the best returns, momentum in U.S. stock returns could not be found. The main driver of negative perfor-mances has been identified to be past losers which exhibit a trend re-versal and deliver strong positive returns. In general, the strategies’ performances in BRIC states are superior to those in developed na-tions. Further, delayed portfolio formation did not alter the results and shows no signs of return reversal in the long run. The results have important implications for the efficiency of markets and the collective behavior of market participants in times of quantitative easing and high frequency trading.

This paper explores the impact of information traders and noise traders on China stock market by applying the the information adjusted noise model (IAMN).The result of empirical test shows China stock market is completely inefficient and both the two types of the investors are contributing to the noise trader risk, which increase the volatility of the market. The Noise trader risk is higher in Bullish period than in Bearish market and the risk is increasing over time. The study of the interaction between information traders and noise traders shows that a part of the noise trader risk is attributed to information traders. They can be overreacting to the mispricing that caused by noise traders and in many cases even join them, which expand the noise trader risk in the market.

I perform a comparative analysis of machine learning methods for measuring equity risk premiums in the Swiss stock market. The implemented methods are partial least squares (PLS), gradient boosted regression trees (GBRT), random forest (RF), deep neural networks (NN), and an equal-weighted ensemble of these models (ENS). Empirical ndings are promising. Machine learning methods surpass the linear benchmark model in terms of out-of-sample predictive R2. In addition, there are large economic gains|in the form of increased portfolio Sharpe ratios|to investors using machine learning forecasts. Furthermore, I identify the best-performing methods (ENS, NN, and PLS) and acknowledge the potential of machine learning for improving risk premium measurement.