Journal of Risk & Control

Multiple Imputation for Missing Values with an Empirical Application

  • Pdf Icon [ Download ]
  • Times downloaded: 8566
  • Abstract 

    Missing data are the most common problem in many research areas. For cross-section and time-series data, imputation can be a challenging problem. The most widely used method for filling missing observations is the multiple imputation which increase the number of the available data and thereby reducing biases that may occur when observations with missing values are simply deleted. The main purpose of this paper is to employ a bootstrapping expectation–maximization (EM) algorithm in order to impute missing values mainly to economic data. In the application we use a dataset that is consisted by annual panel data for the 27 countries of the European Union covering the period 2000-2017. The data were obtained from the databases of World Bank and Eurostat namely the Global Financial Development Database, The Standardized World Income Inequality Database by Solt (2019) and the World Development Indicators. Different indicators were chosen representing the development of banking system and stock markets, economic growth, economic inequality, innovation, fiscal policy, physical and human capital, and trade openness. Finally, diagnostic tools are used inspecting the imputations that are created.

     

    Keywords: Multiple imputation, Amelia II, Economic data, Financial development, Inequality.