Deep learning networks are applied to stock market analysis and prediction.A comprehensive analysis with different data representation methods is offered.Five-minute intraday data from the Korean KOSPI stock market is used.The network applied to residuals of autoregressive model improves prediction.Covariance estimation for market structure analysis is improved with the network. We offer a systematic analysis of the use of deep learning networks for stock market analysis and prediction. Its ability to extract features from a large set of raw data without relying on prior knowledge of predictors makes deep learning potentially attractive for stock market prediction at high frequencies. Deep learning algorithms vary considerably in the choice of network structure, activation function, and other model parameters, and their performance is known to depend heavily on the method of data representation. Our study attempts to provides a comprehensive and objective assessment of both the advantages and drawbacks of deep learning algorithms for stock market analysis and prediction. Using high-frequency intraday stock returns as input data, we examine the effects of three unsupervised feature extraction methodsprincipal component analysis, autoencoder, and the restricted Boltzmann machineon the networks overall ability to predict future market behavior. Empirical results suggest that deep neural networks can extract additional information from the residuals of the autoregressive model and improve prediction performance; the same cannot be said when the autoregressive model is applied to the residuals of the network. Covariance estimation is also noticeably improved when the predictive network is applied to covariance-based market structure analysis. Our study offers practical insights and potentially useful directions for further investigation into how deep learning networks can be effectively used for stock market analysis and prediction.

Deep learning networks for stock market analysis and prediction

https://econ.duke.edu/~boller/Published_Papers/jfe_15.pdf

Tail Risk Premia and Return Predictability

We investigate lead-lag relationships among country stock returns and identify a leading role for the United States: lagged U.S. returns significantly predict returns in numerous non-U.S. industrialized countries (after controlling for national economic variables and countries' own lagged returns), while lagged non-U.S. returns display little predictive ability with respect to U.S. returns. The predictive power of lagged U.S. returns is robust across a number of dimensions, including out-of-sample tests. Information frictions seem a ready explanation of the predictive power of lagged U.S. returns; indeed, structural estimation of a news-diffusion model indicates that return shocks emanating in the United States are only fully reflected in equity prices outside of the United States with a lag. Overall, our results indicate that predictive regressions for non-U.S. countries should be augmented with lagged U.S. returns to capture an important source of international return predictability.

https://papers.ssrn.com/sol3/Delivery.cfm/SSRN_ID2065056_code16976.pdf?abstractid=1508484

International Stock Return Predictability: What is the Role of the United States?

How Rating Agencies Achieve Rating Stability

The Limits of International Law

Recent empirical evidence suggests that the variance risk premium predicts aggregate stock market returns. We demonstrate that statistical finite sample biases cannot “explain” this apparent predictability. Further corroborating the existing evidence of the U.S., we show that country specific regressions for France, Germany, Japan, Switzerland and the U.K. result in quite similar patterns. Defining a “global” variance risk premium, we uncover even stronger predictability and almost identical cross-country patterns through the use of panel regressions. These new findings are broadly consistent with the implications from a two-country general equilibrium model explicitly incorporating the effects of time-varying economic uncertainty.

Stock Return Predictability and Variance Risk Premia : Statistical Inference and International Evidence

Recent empirical evidence suggests that the variance risk premium, or the difference between risk-neutral and statistical expectations of the future return variation, predicts aggregate stock market returns, with the predictability especially strong at the 2-4 month horizons. We provide extensive Monte Carlo simulation evidence that statistical finite sample biases in the overlapping return regressions underlying these findings can not “explain” this apparent predictability. Further corroborating the existing empirical evidence, we show that the patterns in the predictability across different return horizons estimated from country specific regressions for France, Germany, Japan, Switzerland and the U.K. are remarkably similar to the pattern previously documented for the U.S. Defining a “global” variance risk premium, we uncover even stronger predictability and almost identical cross-country patterns through the use of panel regressions that effectively restrict the compensation for world-wide variance risk to be the same across countries. Our findings are broadly consistent with the implications from a stylized two-country general equilibrium model explicitly incorporating the effects of world-wide time-varying economic uncertainty.

/pdf/stock-return-predictability-and-variance-risk-premia-29k5ko7ace.pdf

Stock Return Predictability and Variance Risk Premia: Statistical Inference and International Evidence

The impact of undiversified idiosyncratic risk on value-at-risk and expected shortfall can be approximated analytically via a methodology known as granularity adjustment (GA). In principle, the GA methodology can be applied to any risk-factor model of portfolio risk. Thus far, however, analytical results have been derived only for simple models of actuarial loss, i.e., credit loss due to default. We demonstrate that the GA is entirely tractable for single-factor versions of a large class of models that includes all the commonly used mark-to-market approaches. Our approach covers both finite ratings-based models and models with a continuum of obligor states. We apply our methodology to CreditMetrics and KMV Portfolio Manager, as these are benchmark models for the finite and continuous classes, respectively. Comparative statics of the GA reveal striking and counterintuitive patterns. We explain these relationships with a stylized model of portfolio risk.

/pdf/granularity-adjustment-for-mark-to-market-credit-risk-models-1yfe4zinn0.pdf

Granularity adjustment for mark-to-market credit risk models

Archaeological looting correlates with a number of problems, including the destruction of stratigraphic data and the damage and loss of artifacts. Looting is also understood to generate revenue, but systematic analysis of this issue is challenged by its opacity: how can we study the economic effects of archaeological looting when the practice is rarely directly observable? To address this problem, we estimate the market value of archaeological sites where artifacts have been previously excavated and documented, using a machine-learning approach. The first step uses 41,587 sales of objects from 33 firms to train an algorithm to predict the distribution channel, lot packaging, and estimated sale price of objects based on their observable characteristics. The second step uses the trained algorithm to estimate the value of sites in which a large number of artifacts have been legally excavated and documented. We make an out-of-sample prediction on two Syrian sites, Tell Bi’a and Dura Europos.

/pdf/a-site-level-market-model-of-the-antiquities-trade-4i4bqgu3s9.pdf

A Site-Level Market Model of the Antiquities Trade

The illicit antiquities market is fueled by a well-documented rise in looting at archaeological sites and a fear that the proceeds of such looting may be financing terrorism or rogue states. In this report, the authors compile evidence from numerous open sources to outline the major policy-relevant characteristics of that market and to propose the way forward for developing policies intended to disrupt illicit networks.

/pdf/tracking-and-disrupting-the-illicit-antiquities-trade-with-3vw6g0ql1k.pdf

James V. Marrone

Papers

Stock Return Predictability and Variance Risk Premia : Statistical Inference and International Evidence

Stock Return Predictability and Variance Risk Premia: Statistical Inference and International Evidence

Granularity adjustment for mark-to-market credit risk models

A Site-Level Market Model of the Antiquities Trade

Tracking and Disrupting the Illicit Antiquities Trade with Open Source Data