Abstracts: May 3-5, 2018 Conference on Market Microstructure and High Frequency Data

Knut Are Aastveit, Norges Bank, Norway

Multivariate Bayesian Predictive Synthesis in Macroeconomic Forecasting

We develop the methodology and a detailed case study in use of a class of Bayesian predictive synthesis (BPS) models for multivariate time series forecasting. This extends the recently introduced foundational framework of BPS to the multivariate setting, with detailed application in the topical and challenging context of multi-step macroeconomic forecasting in a monetary policy setting. BPS evaluates– sequentially and adaptively over time– varying forecast biases and facets of miscalibration of individual forecast densities, and– critically– of time-varying inter-dependencies among them over multiple series. We develop new BPS methodology for a specific subclass of the dynamic multivariate latent factor models implied by BPS theory. Structured dynamic latent factor BPS is here motivated by the application context– sequential forecasting of multiple US macroeconomic time series with forecasts generated from several traditional econometric time series models. The case study highlights the potential of BPS to improve of forecasts of multiple series at multiple forecast horizons, and its use in learning dynamic relationships among forecasting models or agents.
Torben Andersen, Northwestern University

Volatility, Information Feedback and Market Microstructure Noise: A Tale of Two Regimes

We extend the classical ”martingale-plus-noise” model for high-frequency prices by an error correction mechanism originating from prevailing mispricing. The speed of price reversal is a natural measure for informational efficiency. The strength of the price reversal relative to the signal-to-noise ratio determines the signs of the return serial correlation and the bias in standard realized variance estimates. We derive the model’s properties and locally estimate it based on mid-quote returns of the NASDAQ 100 constituents. There is evidence of persistent local regimes of positive and negative serial correlation, arising from lagged feedback effects and sluggish price adjustment. The model performance is decidedly superior to existing stylized microstructure models. Finally, we document intraday periodicities in the speed of price reversion and noise-to-signal ratios (Torben G. Andersen, Ilya Archakov, Gokhan Cebiroglu, and Nikolaus Hautsch)
Andrew Caminschi, University of Western Australia

Benchmark auctions and continuous trading: The case of the London gold and silver fixings.

This empirical market microstructure study examines the interactions of a financial benchmark auction, the London fixings, with continuously trading spot and futures markets. It finds strong empirical evidence that the fixing produces a biased benchmark price, impacts the continuous trading markets, and increases an exploitable information asymmetry. The trade advantage, found to be significant and economic, increased with the transition of electronic futures trading which occurs midway through the study period, 2000 to 2013.
Rong Chen, Rutgers University

Modeling High-Dimensional Dynamic Traffic Networks with Matrix Factor Models, with an Application to International Trade Flow Time Series

Dynamic network analysis has found an increasing interest in the literature because of the importance of different kinds of dynamic social networks, biological networks and economic networks. Most available probability and statistical models for dynamic network data are deduced from random graph theory where the networks are characterized on the node and edge level. In this paper, we investigate the analysis of dynamic traffic network using matrix valued factor models. The traffic network can be represented by a time series of matrices. The corresponding matrix factor model then has a network community interpretation. We applied the proposed method to the monthly international trade flow data from 1981 to 2015. The results unveil an interesting evolution of the latent trading network and the relations between the latent entities and the countries.
Jianqing Fan, Princeton University

High-Frequency Large Volatility Estimation via Matrix Completion

When estimating co-volatilities based on high-frequency data, one of the crucial challenges is non-synchronization for illiquid assets, which makes their co-volatility estimators inaccurate. In this talk, we study how to estimate the large integrated volatility matrix without using co-volatilities of illiquid assets. Specifically, we pretend that the co-volatilities for illiquid assets are missing, and estimate the low-rank matrix using a matrix completion scheme with a structured missing pattern. To further regularize the sparse volatility matrix, we employ the principal orthogonal complement thresholding method (POET). We also investigate the asymptotic properties of the proposed estimation procedure and demonstrate its advantages over using co-volatilities of illiquid assets. The advantages of our methods are also verified by an extensive simulation study and illustrated by high-frequency financial data for constituents of the S&P 500 index. (Joint with Donggy Kim)
David Finer, Chicago Booth

What Insights Do Taxi Rides Offer into Federal Reserve Leakage?

How do markets learn about central banks? Employing narrative evidence and futures data, Cieślak, Morse and Vissing-Jørgensen (2018) argue that private, informal channels systematically carry Federal Reserve information to markets around monetary-policy meetings. I complement their work with an analysis of potential channels: New York Fed insiders’ interactions with commercial bankers. Using taxi data, I find highly statistically significant evidence that late-night meetings at the New York Fed and lunchtime offsite interactions increase around FOMC meetings. These results suggest increased opportunities for Federal Reserve information to flow to markets along informal or discreet channels.
Eric Ghysels, University of North Carolina at Chapel Hill

Portfolio Choice and Heterogeneous Agents – An Empirical Study

We use a unique data set of retail investor brokerage accounts over a 10 year span of over 10000 individuals to study the empirical determinants of portfolio choice. While the original data set contains all the trading transactions, we focus on monthly snapshots of portfolio allocation across the large cross-section of individual and large cross-section of stocks. Our data also contain individual socio-economic characteristics such as education, gender, age, income, as well as a measure of risk aversion. In the first part of the paper we apply simple machine learning techniques to uncover patterns of portfolio choice. In the second part of the paper we propose a new interaction model for individual and stock characteristics and estimate key parameters of interest which measure the sensitivity of portfolio weighting schemes to individual investor characteristics.
Marlene Haas, Cornerstone Research

(No) Limits to Short-Selling Arbitrage and Stock Market Quality

This paper studies whether equity short sales and options are complements or substitutes and finds that they are substitutes when the underlying market faces short-selling constraints. This substitutability is associated with the following implications for stock price volatility and stock market liquidity: Stocks that are subject to short-sale restrictions and have exchange-traded options are more volatile and less liquid than both constrained stocks without options and unconstrained stocks. Traders predominantly use at-the-money options to substitute equity short sales. (Joint with Angel Tengulov)
Yingying Li, Hong Kong University of Science & Technology, Hong Kong

Volatility of volatility: estimation and tests based on noisy high frequency data

This paper proposes a volatility of volatility estimator in the high frequency setting with noise and jumps. A feasible central limit theorem with a rate of convergence $n^{1/8}$ is established. To our knowledge, this is the first inference result for volatility of volatility under this general setup. We further find that the rate of convergence can be improved to $n^{1/5}$ under the null that volatility process has a bounded variation. This yields a more powerful test for the presence of diffusion component in the volatility process. Finite sample performance of the estimator and test statistic are examined by simulation studies. Empirical analysis shows that volatility processes in general appear to have diffusion components. (Joint with Guangying Liu and Zhiyuan Zhang)
Yuan Liao, Rutgers University

Uniform inference for conditional factor models with instrumental and idiosyncratic betas

We study conditional factor models where betas depend on observed instruments such as firm specific characteristics. We specify the factor betas as functions of observed instruments that pick up long-run beta fluctuations, plus an orthogonal idiosyncratic component that captures high-frequency movements. It is often the case that researchers do not know whether or not the idiosyncratic beta exists, or its strengths, and thus uniformity is essential for inferences. It is found that the limiting distribution of the estimated instrument effect on the betas has a discontinuity when the strength of the idiosyncratic beta is near zero, which makes usual inferences either very conservative or invalid. We propose an inference that is valid uniformly over a broad class of data generating processes for idiosyncratic betas with various signal strengths and degrees of time-variant, and show that a cross-sectional bootstrap procedure is essential.
Andrew Patton, Duke University

Realized SemiCovariances: Looking for Signs of Direction Inside the Covariance Matrix

We propose a new decomposition of the realized covariance matrix into four “realized semicovariance” components based on the signs of the underlying high-frequency returns. We derive the asymptotic distribution for the different components under the assumption of a continuous multivariate semimartingale and standard infill asymptotic arguments. Based on high-frequency returns for a large cross-section of individual stocks, we document distinctly different features and dynamic dependencies in the different semicovariance components. We demonstrate that the accuracy of portfolio return variance forecasts may be significantly improved by exploiting these differences and “looking inside” the realized covariance matrices for signs of direction (by Tim Bollerslev, Andrew Patton and Rogier Quaedvlieg)
Mathieu Rosenbaum, Ecole Polytechnique, France

Optimal make-take fees for market making regulation

We consider an exchange who wishes to set suitable make-take fees to attract liquidity on its platform. Using a principal-agent approach, we are able to describe in quasi-explicit form the optimal contract to propose to a market maker. This contract depends essentially on the market maker inventory trajectory and on the volatility of the asset. We also provide the optimal quotes that should be displayed by the market maker. The simplicity of our formulas allows us to analyze in details the effects of optimal contracting with an exchange, compared to a situation without contract. We show in particular that it leads to higher quality liquidity and lower trading costs for investors. (Joint with Omar El Euch, Thibaut Mastrolia and Nizar Touzi)
Gustavo Schwenkler, Boston University

Efficient Inference and Filtering for Multivariate Jump-Diffusions

This paper develops estimators of the transition density, filters, and parameters of multivariate jump-diffusions in the presence of latent factors and measurement errors. The drift, volatility, jump intensity, and jump magnitude are allowed to be general functions of the state. Our density and filter estimators converge at the canonical rate typically associated with exact Monte Carlo estimation. Our parameter estimators have the same asymptotic distribution as maximum likelihood estimators. The results of this paper enable the empirical analysis of previously intractable models of asset prices and economic time series.
Michael Sørensen, University of Copenhagen, Denmark

Estimating functions for stochastic differential equations with jumps

Asymptotic theory for approximate martingale estimating functions for stochastic differential equations with finite-activity jumps is presented. Low frequency sampling as well as high frequency sampling with increasing time horizon is considered. We present a class of jump-diffusions, generalising the Pearson diffusions, for which explicit martingale estimating functions exist that work in both asymptotic scenarios. A primary aim of the lecture is to shed light on the question of rate optimal and efficient estimation for jump diffusions. It is demonstrated how Godambe-Heyde optimal estimating functions for continuous diffusion processes can be used to construct rate optimal and efficient estimating functions for diffusions with jumps. (Joint with Nina Munkholt Jakobsen and Mathias Schmidt)
Viktor Todorov, Northwesten University

Time-Varying Periodicity in Intraday Volatility

We develop a nonparametric test for whether return volatility exhibits time-varying intraday periodicity using a long time series of high-frequency data. Our null hypothesis, commonly adopted in work on volatility modeling, is that volatility follows a stationary process combined with a constant time-of-day periodic component. The limit distribution of the developed test depends on the error in recovering volatility from discrete return data and the empirical process error associated with estimating volatility moments through their sample counterparts. In an empirical application to S&P 500 index returns, we find strong evidence for variation in the intraday volatility pattern driven in part by the current level of volatility. When volatility is elevated, the period preceding the market close constitutes a significantly higher fraction of the total daily integrated volatility than during low volatility regimes. (Joint with Torben G. Andersen and Martin Thyrsgaard)
Rene Wells, University of Calgary

Choice of Order Size and Price Discovery: The Last Digit Puzzle

I claim that uninformed traders prefer ending the size of their orders with a zero (e.g. 110 shares) but it is not the case for informed traders, creating an information channel and providing a signal. I propose the Last Digit Hypothesis (LDH): i) some traders exhibit a last digit preference for the digit 0 and other traders do not while ii) the latter are better able to trade on information than the former. The LDH predicts that a trade arising from a marketable order with a size ending with a 0 on average contributes less to price discovery than other trades. My empirical findings support the LDH. However, the LDH is not an equilibrium since informed traders have an incentive to mimic the preferences of uninformed traders to avoid detection and face little constraints or costs to do so. It is puzzling that I find no evidence of such mimicking. I offer plausible explanations for this finding.
B.J.M. Werker, Tilburg University, The Netherlands

Liquidity premiums in various asset classes

The cost of illiquidity may be twofold: suboptimal asset allocation and suboptimal consumption. We show that only the latter generates significant liquidity premiums. This insight explains the gap between the empirical and theoretical literature on liquidity premiums. Assets which generate a relative stable income stream are most likely to carry a liquidity premium. We link our findings to the empirical evidence for liquidity premiums in different asset classes. Consistent with the empirical literature we find liquidity premiums to be present in certain asset classes, whereas absent in others. (Kristy Jansen and Bas J.M. Werker)
Lan Zhang, University of Illinois at Chicago

The Five Trolls Under the Bridge: Principal Component Analysis with Asynchronous and Noisy High Frequency Data

We develop a principal component analysis (PCA) for high frequency data. As in Northern fairly tales, there are trolls waiting for the explorer. The first three trolls are market microstructure noise, asynchronous sampling times, and edge effects in estimators. To get around these, a robust estimator of the spot covariance matrix is developed based on the Smoothed TSRV (Mykland et al (2017)). The fourth troll is how to pass from estimated time-varying covariance matrix to PCA. Under finite dimensionality, we develop this methodology through the estimation of realized spectral functions. Rates of convergence and central limit theory, as well as an estimator of standard error, are established. The fifth troll is high dimension on top of high frequency, where we also develop PCA. With the help of a new identity concerning the spot principal orthogonal complement, the high-dimensional rates of convergence have been studied after eliminating several strong assumptions in classical PCA. As an application, we show that our first principal component (PC) closely matches but potentially outperforms the S&P 100 market index, while three of the next four PCs are cointegrated with two of the Fama-French non-market factors. From a statistical standpoint, the close match between the first PC and the market index also corroborates this PCA procedure and the underlying S-TSRV matrix. (Joint with Dachuan Chen and Per Mykland)
Zhengjun Zhang, University of Wisconsin at Madison

Informed trading in the Bitcoin market 

Bitcoin’s price sensitivity to the material events makes informed trading very profitable in this new market. We propose a novel indicator to assess informed trades ahead of cryptocurrency-related events. Using trade-level data of USD/BTC exchange rates, we find evidence of informed trading in the Bitcoin market prior to large events: Quantiles of the order sizes of buyer-initiated (seller-initiated) orders are abnormally high before large positive (negative) events, compared to the quantiles of seller-initiated (buyer-initiated) orders. When examining the timing of informed trades, we further notice that informed traders prefer to build their positions two days before large positive events and one day before large negative events. The profits of informed trading in the Bitcoin market are estimated to be considerably large. (Joint with Wenjun Feng and Yiming Wang)


Hao Zhou, Thsingua University, China

Leverage Network and Market Contagion

Using daily account-level data that track hundreds of thousands of margin investors’ leverage ratios, trading activities, and portfolio holdings, we examine the effect of margin-induced trading on stock prices during the recent market turmoil in China. We start by showing that individual margin investors have a strong tendency to scale down their holdings after experiencing negative portfolio shocks. Aggregating this behavior across all margin accounts, we find that returns of stocks that share common margin-investor ownership with the stock in question help forecast the latter’s future returns. This transmission mechanism is present only in market downturns, suggesting that idiosyncratic, adverse shocks to individual stocks can be amplified and transmitted to other securities through a de-leveraging channel. As a natural extension, we also show that the previously-documented asymmetry in return comovement between market booms and busts can be largely attributed to deleveraging-induced selling in the bust period. Finally, we show that stocks that are more central in the margin-holding network have significantly larger downside betas than peripheral stocks.