“Eureka, there’s a 98 percent correlation!” Many users of time series data do not realize that trends and common movements of observations of variables over time, such as employment and prices over seasons and business cycles, present special problems. Improper tests can find statistical correlations that do not indicate causality. Also, mutilation problems arise from a commonly used smoothing procedure: moving averages. This is a user-friendly course in applied statistics in which estimates of economic activity including national income estimates, employment survey data, and financial data will be described. Attempted solutions will be presented for transforming time series for testing purposes. The methods presented include Granger causality tests, autoregressive filter procedures, and cointegration procedures. Control problems that occur when a variable is under the control of the government will be addressed. There will be a final statistical project.
Spring 2013 - 63070 - PA397C - Advanced Empirical Methods for Policy Analysis
Analyzing Government Economic Statistics