Working Papers

HighFrequencyCovariance: A Julia Package for Estimating Covariance Matrices Using High Frequency Financial Data (with Margaryta Klymak)

High frequency data typically exhibit asynchronous trading and microstructure noise, which can bias the covariances estimated by standard estimators. While a number of specialised estimators have been developed, they have had limited availability in open source software. HighFrequencyCovariance is the first Julia package which implements specialised estimators for volatility, correlation and covariance using high frequency financial data. It also implements complementary algorithms for matrix regularisation as well as functions to estimate a covariance matrix blockwise and combine the results. This paper first presents the issues associated with exploiting high frequency financial data. We then describe the volatility, covariance and regularisation algorithms and demonstrate their implementation in the HighFrequencyCovariance package. We perform a Monte Carlo experiment, which shows the accuracy gains that are possible in different settings. Finally, we show that different estimators can be combined to form an ensemble estimator which can produce more accurate estimates with lower variance than any of the individual estimators.

JEL Codes: C22, C58, G11, G12

Keywords: covariance estimation, correlation, volatility, high-frequency financial data, Julia

This was presented at JuliaCon 2021. The video is here.

Comparative Advertising: The role of prices

In markets where firms sell similar goods to their competitors, firms may be able to free-ride off the costly price signalling of competitor firms by engaging in price comparative advertising. As the goods are similar consumers can reason that if one good is high quality (revealed for instance through price signalling) then so is the other. This paper models this phenomenon and finds that in equilibrium there will be firms price signalling as well as freeriding firms that signal through advertising. Surplus is strictly higher in markets where advertising firms are active relative to pure price signalling markets. In some cases advertising markets can be even more efficient than full information markets as advertisers surrender market power to avoid costly price signalling.

JEL Codes: D82, D83, M37

Keywords: Comparative advertising, Price Signalling

Version as MPRA Paper No. 79872

For a nontechnical explanation of this paper please press here

It's good to be bad: A model of low quality dominance in a full information consumer search market (with Margaryta Klymak)

This paper examines a consumer search market exhibiting vertically differentiated firms, heterogeneous consumers and endogenous consumer market entry. In an asymmetric information setting high and low quality firms make equal sales and profit in this market. Conversely when there is full information, search frictions induce an unravelling mechanism that leads to a unique refined equilibrium where all consumers approach low quality firms and high quality firms make no sales or profit. This presents a rationale for why low quality firms may disclose their quality and high quality firms may not even when disclosure is costless.

JEL Codes: D82, D83, L15

Keywords: Quality Disclosure, Consumer Search

Version as Edinburgh School of Economics discussion paper 280

For a nontechnical explanation of this paper please press here

Work in Progress

Paying over the odds at the end of the fiscal year: Evidence from Ukraine (with Margaryta Klymak)