The Book of Alternative Data: A Guide for Investors, Traders and Risk Managers

Preface xv Acknowledgments xvii Part 1 Introduction and Theory 1 1 Alternative Data: The Lay of the Land 3 1.1 Introduction 3 1.2 What is "Alternative Data"? 5 1.3 Segmentation of Alternative Data 7 1.4 The Many Vs of Big Data 9 1.5 Why Alternative Data? 11 1.6 Who is Using Alternative Data? 15 1.7 Capacity of a Strategy and Alternative Data 16 1.8 Alternative Data Dimensions 19 1.9 Who Are the Alternative Data Vendors? 23 1.10 Usage of Alternative Datasets on the Buy Side 24 1.11 Conclusion 26 2 The Value of Alternative Data 27 2.1 Introduction 27 2.2 The Decay of Investment Value 27 2.3 Data Markets 29 2.4 The Monetary Value of Data (Part I) 31 2.4.1 Cost Value 34 2.4.2 Market Value 34 2.4.3 Economic Value 35 2.5 Evaluating (Alternative) Data Strategies with and without Backtesting 35 2.5.1 Systematic Investors 36 2.5.2 Discretionary Investors 38 2.5.3 Risk Managers 39 2.6 The Monetary Value of Data (Part II) 39 2.6.1 The Buyer's Perspective 40 2.6.2 The Seller's Perspective 41 2.7 The Advantages of Maturing Alternative Datasets 45 2.8 Summary 46 3 Alternative Data Risks and Challenges 47 3.1 Legal Aspects of Data 47 3.2 Risks of Using Alternative Data 50 3.3 Challenges of Using Alternative Data 51 3.3.1 Entity Matching 52 3.3.2 Missing Data 54 3.3.3 Structuring the Data 55 3.3.4 Treatment of Outliers 56 3.4 Aggregating the Data 57 3.5 Summary 58 4 Machine Learning Techniques 59 4.1 Introduction 59 4.2 Machine Learning: Definitions and Techniques 60 4.2.1 Bias, Variance, and Noise 60 4.2.2 Cross-Validation 61 4.2.3 Introducing Machine Learning 62 4.2.4 Popular Supervised Machine Learning Techniques 64 4.2.5 Clustering-Based Unsupervised Machine Learning Techniques 70 4.2.6 Other Unsupervised Machine Learning Techniques 71 4.2.7 Machine Learning Libraries 71 4.2.8 Neutral Networks and Deep Learning 72 4.2.9 Gaussian Processes 80 4.3 Which Technique to Choose? 82 4.4 Assumptions and Limitations of the Machine Learning Techniques 84 4.4.1 Causality 84 4.4.2 Non-stationarity 85 4.4.3 Restricted Information Set 86 4.4.4 The Algorithm Choice 86 4.5 Structuring Images 87 4.5.1 Features and Feature Detection Algorithms 87 4.5.2 Deep Learning and CNNs for Image Classification 89 4.5.3 Augmenting Satellite Image Data with Other Datasets 90 4.5.4 Imaging Tools 91 4.6 Natural Language Processing (NLP) 91 4.6.1 What is Natural Language Processing (NLP)? 91 4.6.2 Normalization 93 4.6.3 Creating Word Embeddings: Bag-of-Words 94 4.6.4 Creating Word Embeddings: Word2vec and Beyond 94 4.6.5 Sentiment Analysis and NLP Tasks as Classification Problems 96 4.6.6 Topic Modeling 96 4.6.7 Various Challenges in NLP 97 4.6.8 Different Languages and Different Texts 98 4.6.9 Speech in NLP 99 4.6.10 NLP Tools 100 4.7 Summary 102 5 The Processes behind the Use of Alternative Data 105 5.1 Introduction 105 5.2 Steps in the Alternative Data Journey 106 5.2.1 Step 1. Set up a Vision and Strategy 106 5.2.2 Step 2. Identify the Appropriate Datasets 107 5.2.3 Step 3. Perform Due Diligence on Vendors 108 5.2.4 Step 4. Pre-assess Risks 109 5.2.5 Step 5. Pre-assess the Existence of Signals 109 5.2.6 Step 6. Data Onboarding 110 5.2.7 Step 7. Data Preprocessing 110 5.2.8 Step 8. Signal Extraction 111 5.2.9 Step 9. Implementation (or Deployment in Production) 112 5.2.10 Maintenance Process 113 5.3 Structuring Teams to Use Alternative Data 114 5.4 Data Vendors 116 5.5 Summary 118 6 Factor Investing 119 6.1 Introduction 119 6.1.1 The CAPM 119 6.2 Factor Models 120 6.2.1 The Arbitrage Pricing Theory 122 6.2.2 The Fama-French 3-Factor Model 123 6.2.3 The Carhart Model 124 6.2.4 Other Approaches (Data Mining) 125 6.3 The Difference between Cross-Sectional and Time Series Trading Approaches 126 6.4 Why Factor Investing? 126 6.5 Smart Beta Indices Using Alternative Data Inputs 127 6.6 ESG Factors 128 6.7 Direct and Indirect Prediction 129 6.8 Summary 132 Part 2 Practical Applications 133 7 Missing Data: Background 135 7.1 Introduction 135 7.2 Missing Data Classification 136 7.2.1 Missing Data Treatments 137 7.3 Literature Overview of Missing Data Treatments 139 7.3.1 Luengo et al. (2012) 139 7.3.2 Garcia-Laencina et al. (2010) 143 7.3.3 Grzymala-Busse et al. (2000) 146 7.3.4 Zou et al. (2005) 147 7.3.5 Jerez et al. (2010) 147 7.3.6 Farhangfar et al. (2008) 148 7.3.7 Kang et al. (2013) 149 7.4 Summary 149 8 Missing Data: Case Studies 151 8.1 Introduction 151 8.2 Case Study: Imputing Missing Values in Multivariate Credit Default Swap Time Series 152 8.2.1 Missing Data Classification 153 8.2.2 Imputation Metrics 154 8.2.3 CDS Data and Test Data Generation 154 8.2.4 Multiple Imputation Methods 157 8.2.5 Deterministic and EOF-Based Techniques 160 8.2.6 Results 164 8.3 Case Study: Satellite Images 173 8.4 Summary 176 8.5 Appendix: General Description of the MICE Procedure 178 8.6 Appendix: Software Libraries Used in This Chapter 179 9 Outliers (Anomalies) 181 9.1 Introduction 181 9.2 Outliers Definition, Classification, and Approaches to Detection 182 9.3 Temporal Structure 183 9.4 Global Versus Local Outliers, Point Anomalies, and Micro-Clusters 184 9.5 Outlier Detection Problem Setup 184 9.6 Comparative Evaluation of Outlier Detection Algorithms 185 9.7 Approaches to Outlier Explanation 189 9.7.1 Micenkova et al. 189 9.7.2 Duan et al. 191 9.7.3 Angiulli et al. 192 9.8 Case Study: Outlier Detection on Fed Communications Index 194 9.9 Summary 201 9.10 Appendix 202 9.10.1 Model-Based Techniques 202 9.10.2 Distance-Based Techniques 202 9.10.3 Density-Based Techniques 203 9.10.4 Heuristics-Based Approaches 203 10 Automotive Fundamental Data 205 10.1 Introduction 205 10.2 Data 206 10.3 Approach 1: Indirect Approach 211 10.3.1 The Steps Followed 212 10.3.2 Stage 1 213 10.4 Approach 2: Direct Approach 223 10.4.1 The Data 223 10.4.2 Factor Generation 224 10.4.3 Factor Performance 225 10.4.4 Detailed Factor Results 229 10.5 Gaussian Processes Example 238 10.6 Summary 239 10.7 Appendix 240 10.7.1 List of Companies 240 10.7.2 Description of Financial Statement Items 241 10.7.3 Ratios Used 242 10.7.4 IHS Markit Data Features 243 10.7.5 Reporting Delays by Country 244 11 Surveys and Crowdsourced Data 245 11.1 Introduction 245 11.2 Survey Data as Alternative Data 245 11.3 The Data 247 11.4 The Product 247 11.5 Case Studies 249 11.5.1 Case Study: Company Event Study (Pooled Survey) 249 11.5.2 Case Study: Oil and Gas Production (Q&A Survey) 252 11.6 Some Technical Considerations on Surveys 254 11.7 Crowdsourcing Analyst Estimates Survey 255 11.8 Alpha Capture Data 256 11.9 Summary 256 11.10 Appendix 256 12 Purchasing Managers' Index 259 12.1 Introduction 259 12.2 PMI Performance 261 12.3 Nowcasting GDP Growth 262 12.4 Impacts on Financial Markets 263 12.5 Summary 266 13 Satellite Imagery and Aerial Photography 267 13.1 Introduction 267 13.2 Forecasting US Export Growth 269 13.3 Car Counts and Earnings Per Share for Retailers 271 13.4 Measuring Chinese PMI Manufacturing with Satellite Data 277 13.5 Summary 280 14 Location Data 283 14.1 Introduction 283 14.2 Shipping Data to Track Crude Oil Supplies 283 14.3 Mobile Phone Location Data to Understand Retail Activity 287 14.3.1 Trading REIT ETF Using Mobile Phone Location Data 288 14.3.2 Estimating Earnings per Share with Mobile Phone Location Data 291 14.4 Taxi Ride Data and New York Fed Meetings 295 14.5 Corporate Jet Location Data and M&A 296 14.6 Summary 298 15 Text Web Social Media and News 299 15.1 Introduction 299 15.2 Collecting Web Data 299 15.3 Social Media 300 15.3.1 Hedonometer Index 302 15.3.2 Using Twitter Data to Help Forecast US Change in Nonfarm Payrolls 305 15.3.3 Twitter Data to Forecast Stock Market Reaction to FOMC 308 15.3.4 Liquidity and Sentiment from Social Media 309 15.4 News 309 15.4.1 Machine-Readable News to Trade FX and Understand FX Volatility 310 15.4.2 Federal Reserve Communications and US Treasury Yields 316 15.5 Other Web Sources 320 15.5.1 Measuring Consumer Price Inflation 321 15.6 Summary 322 16 Investor Attention 323 16.1 Introduction 323 16.2 Readership of Payrolls to Measure Investor Attention 323 16.3 Google Trends Data to Measure Market Themes 325 16.4 Investopedia Search Data to Measure Investor Anxiety 328 16.5 Using Wikipedia to Understand Price Action in Cryptocurrencies 330 16.6 Online Attention for Countries to Inform EMFX Trading 330 16.7 Summary 333 17 Consumer Transactions 335 17.1 Introduction 335 17.2 Credit and Debit Card Transaction Data 336 17.3 Consumer Receipts 337 17.4 Summary 340 18 Government, Industrial, and Corporate Data 341 18.1 Introduction 341 18.2 Using Innovation Measures to Trade Equities 342 18.3 Quantifying Currency Crisis Risk 344 18.4 Modeling Central Bank Intervention in Currency Markets 346 18.5 Summary 348 19 Market Data 351 19.1 Introduction 351 19.2 Relationship between Institutional FX Flow Data and FX Spot 351 19.3 Understanding Liquidity Using High-Frequency FX Data 355 19.4 Summary 357 20 Alternative Data in Private Markets 359 20.1 Introduction 359 20.2 Defining Private Equity and Venture Capital Firms 360 20.3 Private Equity Datasets 362 20.4 Understanding the Performance of Private Firms 363 20.5 Summary 364 Conclusions 365 Some Last Words 365 References 367 About the Authors 373 Index 375