Three Essays on Error in Economic Data

Kevin Brancato

Advisor: Richard E Wagner, PhD, Department of Economics

Enterprise Hall, 318
April 28, 2005, 08:00 PM to 07:00 PM


"These three essays focus on error in macroeconomic data as currently produced in the United States. The first examines why error is unavoidable, and why analysts using macroeconomic data must incorporate their own expectations of error in the data into their evaluations. The second dissects the structure and assembly of Gross Domestic Product, to better understand the causes of error. The third demonstrates that it is possible to develop rule-of-thumb estimates of error for GDP to assist in economic judgments. Using publicly available revisions, the minimum error in GDP is calculated to be |1|%, or about 35% of the average quarterly growth rate. Instead of converging to the ItrueO growth rate, the sequence of revisions appear to increasingly diverge from the initial IAdvanceO estimate. The methods used are unavoidably heterodox: a combination of analogy, straight-forward calculation, regression, and visualization. Stories about the demand, use, and quality of data in different disciplines frame and supplement explicit analysis of the source data, construction methods, and revision processes of GDP. "