Missing Data in Financial Modeling

The global economy consists of markets with varying levels of interdependency. Taking these levels into account will lead to improvement of price predictions both globally and in local markets. However markets are open at different times so there are temporal alignment issues that must be addressed before estimating these correlation structures and then using them to improve predictions. To address these issues, we develop the Global Price Syncer: a variation of the Expectation Maximization Algorithm, which can impute large amounts of missing data. The efficiency of the model was tested on a dataset, consisting of three months worth of equity index futures from four of the biggest markets globally.