Essays in Financial Economics
ISBN 978-91-7731-078-5 Doctoral Dissertation in Finance Stockholm School of Economics Sweden, 2018
Essays in Financial EconomicsErik Sverdrup • 2018
Erik Sverdrup
Essays in Financial Economics
ISBN 978-91-7731-078-5 Doctoral Dissertation in Finance Stockholm School of Economics Sweden, 2018
Essays in Financial EconomicsErik Sverdrup • 2018
Erik Sverdrup
Essays in Financial Economics
Erik Sverdrup
Akademisk avhandling
som för avläggande av ekonomie doktorsexamen vid Handelshögskolan i Stockholm
framläggs för offentlig granskning måndagen den 4 juni 2018, kl 15.15,
Swedish House of Finance,
Drottninggatan 98, Stockholm
Essays in Financial Economics
Essays in Financial Economics
Erik Sverdrup
Dissertation for the Degree of Doctor of Philosophy, Ph.D., in Finance
Stockholm School of Economics, 2018
Essays in Financial Economics SSE and Erik Sverdrup, 2018c
ISBN 978-91-7731-078-5 (printed) ISBN 978-91-7731-079-2 (pdf)
This book was typeset by the author using LATEX.
Printed by:
BrandFactory, Gothenburg, 2018 Keywords:
Financial intermediaries, networks, hedge funds, computational linguistics, empirical asset pricing.
Foreword
This volume is the result of a research project carried out at the Department of Finance at the Stockholm School of Economics(SSE).
This volume is submitted as a doctoral thesis at SSE. In keeping with the policies of SSE, the author has been entirely free to conduct and present his research in the manner of his choosing as an expression of his own ideas.
SSE is grateful for the financial support provided by the Jan Wallander and Tom Hedelius Foundation, which has made it possible to carry out the project.
G¨oran Lindqvist Magnus Dahlquist
Director of Research Professor and Head of the Stockholm School of Economics Department of Finance
Stockholm School of Economics
Acknowledgements
I am grateful to the entire faculty and staff at the Stockholm School of Economics. I am particularly grateful to my supervisor Michael Halling for his support during these years.
I am also grateful to Valeri Sokolovski and Magnus Dahlquist for being my coauthors.
I owe special thanks to Daniel Metzger for allowing me to use his conference call tran- script data for my paper. I am thankful to my fellow PhD students, particularly Markus and Yingjie who were in the same cohort, and Katarına and Alberto, who shared ther same office. Finally, I gratefully acknowledge the financial support of The Swedish Bank Research Foundation.
Stockholm, April 20, 2018 Erik Sverdrup
Contents
Introduction 1
1 Hedge Funds and Financial Intermediaries 5
1 Introduction . . . 6
2 Data . . . 10
2.1 Hedge fund data . . . 10
2.2 Other data . . . 11
3 Descriptive statistics . . . 12
4 Methodology and results . . . 13
4.1 Financial intermediary risk . . . 13
4.2 Individual prime broker channel . . . 17
4.3 Prime broker and hedge-fund network . . . 19
4.4 Prime broker fixed effect . . . 23
5 Robustness . . . 25
5.1 Hedge fund portfolio returns during extreme factor realizations . 25 5.2 Double sorts . . . 25
6 Conclusion . . . 26
Bibliography . . . 27
Tables and Figures . . . 38
A-I Data . . . 46
A-II Robustness . . . 47
A-II.A Intermediary-beta-sorted portfolios . . . 47
A-II.B Cross-sectional regression . . . 50
A-III Prime broker and hedge fund network . . . 51
A-IV Prime broker fixed effect . . . 54 xi
2 Information in Corporate Conference Calls 57
1 Introduction . . . 57
2 Data . . . 60
3 Topic decomposition . . . 63
4 Firm returns and conference call topics . . . 67
4.1 Section differences . . . 69
4.2 Return predictability . . . 70
5 Aggregate measures and uncertainty indexes . . . 72
6 Conclusion . . . 75
Bibliography . . . 76
Tables and Figures . . . 83
A-I Definitions . . . 107
3 The Benchmark Currency Stochastic Discount Factor 109 1 Introduction . . . 109
2 Definitions, theoretical background and literature review . . . 112
2.1 Background theory . . . 112
2.2 Factors proposed in the literature . . . 115
3 Data . . . 118
4 Descriptive Statistics . . . 118
5 Empirical Methodology . . . 119
6 Estimation results . . . 121
7 Cross-sectional asset pricing . . . 124
8 Conclusion . . . 127
Bibliography . . . 127
Tables and Figures . . . 132
Introduction
This thesis consists of three independent papers in financial economics. The unifying theme is the goal to address a number of active areas in financial research with new and interdisciplinary methodology.
The first paper, Hedge Funds and Financial Intermediaries(ongoing joint work with Magnus Dahlquist and Valeri Sokolovski), contributes to the growing field of inter- mediary asset pricing by evaluating intermediary risk in an important and often poorly understood asset class − hedge funds. The field of intermediary asset pricing seeks to shed light on the role played by financial intermediaries in the determination of asset prices and risk premiums. In a recent and important study, He, Kelly and Manela(2017) find that a factor based on the shocks to the health of a small group of key financial interme- diaries(a group comprised mostly of large global investment banks that are predominant traders in many markets) has significant explanatory power for the cross-sectional varia- tion in expected returns across multiple asset classes. Hedge funds are essentially portfo- lios of, often exotic, securities of multiple asset classes, hence they offer an ideal set of test assets for the evaluation of the intermediary factor’s pricing power. Additionally, hedge funds are also interesting as test assets because of their direct link to the financial inter- mediaries through their prime brokerage relationships(prime brokers are hedge funds’
primary gateway to the financial market and the source of financing).
We find that the He et al. (2017) factor is a strong determinant of the cross-section of hedge fund returns. Hedge funds with high intermediary risk exposure significantly out- performs, on average, the low-exposure funds by around 7% per year on a risk-adjusted basis. Moreover, we take advantage of a unique and rich hedge fund dataset and extend the He et al. (2017) measure by linking each hedge fund to their prime broker. Method- ologically, we adopt tools from network theory to gauge the relative importance of each
1
2 ESSAYS IN FINANCIAL ECONOMICS
prime broker through their centrality rank in a client-broker network. This approach, not only allows us to easily visualize the core-periphery market structure prevalent in the hedge fund sector, but also to identify in a precise and economically intuitive way the most important intermediaries for that sector.
The advent of better and cheaper computer power and the ballooning set of un- conventional data sources and algorithms for finding structure, has sparked interest in applying tools from machine learning(ML) to economic research questions. In a survey article, Athey(2017) argues that the adoption of ML related tools in empirical work is an important intermediate step for creating variables of economic interest. I pay heed to her insight, and in my second paper, The Information Content in Corporate Con- ference Calls: A Computational Linguistics Approach, I adopt a clustering algorithm to create new units of observation from a collection of text transcripts of US corporate conference calls.
Clustering is a way to systematically partition a data set into groups or clusters such that the elements of each group are similar according to some predefined metric. Then one can assign semantic labels to each cluster based on the nature of the representative elements. In the case of the document-clustering technique(Latent Dirichlet Allocation) that I employ, the similarity metric is a vector of weights that describes a mixture of topics. Hence, the procedure reduces the large set of conference calls transcripts to a manageable set of themes or topics.
I find that the measures constructed from this composition are related to economic tangibles. There is a significant relation between the fraction of conversation devoted to certain topics during a conference call and the firm’s stock return on conference call days.
Another early adopter, Hansen, McMahon and Prat(2017), uses the same methodology on transcripts from the Federal Open Market Committee meetings. One of their findings is that when they consider topic clusters that plausibly relate to economic uncertainty and aggregate these topics to form an index, this index strongly co-moves with the manu- ally constructed Economic Policy Uncertainty index by Baker, Bloom and Davis(2016).
I obtain similar results with conference call transcripts. In particular, some aggregated topic proportions(across firms every quarter), appear related to indices that measures economic sentiment. This suggests that clustering techniques can be useful for creating new variables of interests, and that corporate conference calls are informative and when
aggregated are able to capture some of the prevalent economic sentiment among agents in the economy.
The third and final paper, The Benchmark Currency Stochastic Discount Fac- tor(joint work with Valeri Sokolovski), investigates the currency carry trade (the puz- zling historical fact that an investor consistently obtains a positive risk adjusted return when investing in high interest rate currencies and borrowing in low interest currencies) through the lens of the information theoretic framework of Ghosh, Julliard and Taylor (2016). The framework proposes an alternative way to recover an important object of interest in financial economics − the stochastic discount factor(SDF). This is the equilib- rium discount rate that equalizes the prices of all assets in the economy(absent arbitrage opportunities) with the sum of their discounted payoffs. A non-parametric SDF recov- ered using this new framework can better price the cross section of currency portfolios than the existing models.
Moreover, in our cross-sectional asset pricing tests we adopt new methodology from Bryzgalova(2015) designed to overcome the rank deficiency problem in the original two- pass Fama-MacBeth procedure. This Penalized Fama-MacBeth estimator utilizes modern results from the high dimensional statistical learning literature where a suitably weighted l1penalty on the asset pricing coefficients of interests yield an estimator that both simul- taneously selects and eliminates weak(or irrelevant factors). Our non-parametrically ex- tracted SDF overcomes this additional hurdle. Hence, we adopt a new methodology to recover a non-parametrically extracted SDF, in the spirit of the recent work contributed to by, among others, Kozak, Nagel and Santosh (2017), and we employ modern tests that consider the statistical difficulties arising from testing many competing factor mod- els similar to the growing literature contributed by Feng, Giglio and Xiu(2017) among others.
3
4 ESSAYS IN FINANCIAL ECONOMICS
References
Athey, S.(2017). The impact of machine learning on economics, Economics of Artificial Intelligence, University of Chicago Press.
Baker, S. R., Bloom, N. and Davis, S. J.(2016). Measuring economic policy uncertainty, Quarterly Journal of Economics 131(4): 1593–1636.
Bryzgalova, S. (2015). Spurious factors in linear asset pricing models, Working paper, Stanford University.
Feng, G., Giglio, S. and Xiu, D.(2017). Taming the factor zoo, Working paper, University of Chicago.
Ghosh, A., Julliard, C. and Taylor, A. P. (2016a). What is the consumption-CAPM missing? An information-theoretic framework for the analysis of asset pricing models, Review of Financial Studies 30(2): 442–504.
Ghosh, A., Julliard, C. and Taylor, A. P.(2016b). An information-theoretic asset pricing model,Working paper, London School of Economics.
Hansen, S., McMahon, M. and Prat, A.(2017). Transparency and deliberation within the FOMC: A computational linguistics approach, Quarterly Journal of Economics 133(2): 801–870.
He, Z., Kelly, B. and Manela, A.(2017). Intermediary asset pricing: New evidence from many asset classes,Journal of Financial Economics 126(1): 1–35.
Kozak, S., Nagel, S. and Santosh, S.(2017). Shrinking the cross section, Working paper, National Bureau of Economic Research.
