International data revision: Theory and analysis
This paper examines the effects data revision has on the ability to make policy decisions for a developing economy. I start by examining the literature on data revision through the lens of econometrics, international data comparisons, and accounting. Each of these literatures points to certain pitfalls of inaccurate data, and methods to detect the difference between random errors and systematic revisions. Systematic revisions can be due to changes in variable definitions, overall changes in data collection, or deliberate misrepresentation.
David Runkle (1998), uses US estimates of growth in output and inflation to discuss how changes in methods of data collection as well as time can affect how researchers look at the way policies are made.
Oskar Morgenstern (1963) writes, “there are many reasons why one should be deeply concerned with the ‘accuracy’ of quantitative economic data and observations…[I] t is a difficult thing to have to make decisions on the basis of information of greatly mixed quality.” Morgenstern finds three types of errors: (1) “errors introduced in the basic data,” (2) “error produced independently of enumeration or sampling difficulties,” and (3) error “introduced in trying to fill in the gaps for those industries and years where estimates are not known.” Morgenstern (1963) concludes by questioning the feasibility of any international comparisons due to errors in data.
Patterson and Heravi (1991) find that revised data are similar if they are co-integrated and dissimilar if not. The hypothesis that initial data is an unbiased forecast of final values takes for granted that the error term is stationary.
The accounting literature is also concerned with data revisions, particularly data manipulation. In accounting, accounts are kept on the accrual basis, which allows some discretion in when accruals are reported. Guay, Kothari, and Watts (1996), discuss the effectiveness of the various models used to detect the use and motivation behind managerial discretion in financial reporting.
For my study, I examine the recent revision of the data from the World Bank and International Monetary Fund utilizing tests of both the economics and accounting literature. We examine the extent of data revision, and any patterns that exist. These patterns give insight into the econometric problems that could be added due to errors in data. We look for regional attributes that cause higher data revision, as well as data series that are more prone to errors. Political, legal, and economic factors are also included as contributing factors to data revision.