High-frequency financial econometrics and low frequency investment management

当前的低收益环境对许多机构投资者构成了相当大的挑战。因此，一些投资者放宽了投资限制，允许进行风险更高的投资。然而，高风险投机也需要更严格的风险控制，以符合预先设定的投资目标。在解决这些需求时，定量投资管理（QIM）中遵循的系统方法是非常合适的。qm起源于资产定价理论，关注的是利用计量经济学技术对资产风险和回报的相关驱动因素进行建模和预测。因此，QIM是一个高度技术性、经验性和数据驱动的学科。认识到资本市场的适应性，景顺定量策略（IQS）目前正在寻找高频新闻分析的新数据，这些数据将为候选人提供更精确的资本市场模型。虽然量化投资经理已经在分析大量数据，但对大量数据集的分析带来了一些挑战，使高频计量经济学成为金融领域最活跃的研究领域之一：数据不仅规模大，而且通常相当复杂，达到高分辨率的高速，并且容易受到噪声的影响。特别地，我们提出的研究旨在探讨新闻流数据和高频数据在低频QIM中的使用。对于新闻流数据，我们会考虑提供实时新闻分析服务。这些服务实时收集新闻，并根据关键词和短语到预定义情感值的映射对新闻进行分类。因此，我们提出的研究将通过建立计量经济学、计算机科学和数据科学等技术来调查这些新数据集与QIM的相关性。

The current low yield environment poses considerable challenges to many institutional investors.Hence some investors loosen investment constraints to allow for more risky investments.Nevertheless, speculating on higher risk levels also demands for stricter risk control to comply withpredefined investment targets. In addressing these needs, a systematic approach as followed withinquantitative investment management (QIM) is highly suitable. QIM originates in asset pricing theoryand is concerned about modeling and forecasting the relevant drivers of assets' risk and return usingeconometric techniques. Consequently, QIM is a highly technical, empirical and data-drivendiscipline. Acknowledging the adaptive nature of capital markets Invesco Quantitative Strategies(IQS) is currently sourcing new data on high-frequency (HF) news analytics that will be available forthe candidate for a more precise modelling of capital markets. While quantitative investmentmanagers are already analyzing large amounts of data, the analysis of huge data sets poses severalchallenges rendering HF econometrics one of the most active research areas in finance: Not only isthe data large in size, it is usually fairly complex, arrives at high speed at high resolution, and issubject to noise. In particular, our proposed study aims to investigate the use of news flow data aswell as HF data for low-frequency QIM. As for news flow data, we will consider the provision of realtimenews analysis services. These services collect news in real time and classify them based on amapping of key words, and phrases to pre-defined sentiment values. Therefore, our proposed studywill investigate the relevance of these new data sets for QIM by building on techniques fromeconometrics, computer science, and data science alike.