John Rush’s Final Honours Year thesis Data Revisions and Some Econometric Consequences was submitted to the then Department of Economic Statistics at the University of Sydney and was instrumental in his being awarded a First Class Honours degree in Economics in 1973.
In his thesis, Rush examined an area of research that had received virtually no public attention in Australia at that time: the magnitude and effects of revisions to data published in the Australian quarterly national accounts. Rush noted that some major economic data series published in a current quarterly publication would often be revised in subsequent publications. As economic policy advisors often used these series to “measure” and “fine tune” the economy, it was clear that the first published estimate of an important time series may well give a misleading impression, thereby leaving the advisors in a position similar to a sailor attempting to navigate with faulty charts.
At the time Rush undertook his data revisions study, econometricians devoted much time to the “best” method of statistically estimating the parameters of their models. Different estimation methods might produce differing model parameters, and the differences could be significant. Interestingly, these same theoretical econometricians tended to give far less attention to the “errors in variables” problem resulting from data revisions to the time series they used in their models.
Rush therefore decided to investigate whether, for a small Australian econometric model, model parameter estimates derived from two distinct data sets (first published data versus “final” data) produced greater parameter variation than did the parameter differences produced by estimation using two distinct methods of estimation (ordinary least squares versus two stage least squares).
- data revisions made to important “indicator” time series focused upon by economic policy advisors could be large and were not always systematic
- the implied effects of data revisions on estimated econometric model parameters could outweigh the changes in parameter estimates caused by varying the method of model estimation
- economic policy advisors should give serious attention to the data revision problem when formulating policy measures based upon “first published” data.
Rush later presented his university thesis as an abridged paper, co-authored with his thesis supervisor, Stephen Harrison (then a lecturer in Economic Statistics at the University of Sydney), to Section 24 - Economics of the 45th Congress of the Australian and New Zealand Association for the Advancement of Science (ANZAAS) in Perth, Western Australia in August 1973. That paper, Data Revisions and Some Econometric and Policy Consequences, is available for download here (5.2MB).
Rush’s work on data revisions (as noted in the Rush/Harrison ANZAAS paper) was later drawn upon and explicitly cited at paragraph 3.1 on page 5 by Prof Warren P Hogan of the University of Sydney in writing his Presidential Address to the New South Wales Branch of the then Economic Society of Australia and New Zealand. Hogan’s address, titled How Do We Know Where We Are Going, When We Don’t Know Where We Have Been, Let Alone Where We Are? (2.7MB), was delivered on 18 October 1978. (Rush was at that time Secretary of the New South Wales Branch of the then Economic Society of Australia and New Zealand.)
Warren Hogan was a distinguished Professor of Economics at the University of Sydney from 1968 to 1998. Following his death in December 2009, the University instituted the Warren Hogan Memorial Lecture and the Inaugural Lecture was delivered at the University by Glenn Stevens, Governor of the Reserve Bank of Australia, on 8 December 2011.
In his Hogan lecture, Stevens said "One Hogan paper of that era that really struck a chord with me then I have always recalled its title was about the issue of revisions to economic statistics. It was entitled How Do We Know Where We Are Going, When We Don’t Know Where We Have Been, Let Alone Where We Are?"
"These days the study of revisions to data is an art in itself (though the quality of national accounts data is better than it was in the 1970s). Central banks put huge effort into what has come to be called ‘nowcasting’ that is, working out what the national accounts probably will, or should, say about the current quarter in three or four months’ time when they are released. Some central banks actually ‘backcast’ the recent past, deciding their own version of GDP, not necessarily the same as the one published by the statisticians, in effect foreshadowing revisions to come." (Reserve Bank of Australia Bulletin December Quarter 2011 p84; the full Lecture text is published here)
Stevens was therefore indirectly drawing upon Rush’s 1972 data revisions work in his Hogan Lecture. Almost 40 years after first being recognised, the area upon which Rush focused is now central to data quality problems needing to be addressed at the highest level internationally by policy advisors.
Data revisions have become a well-defined area of economic study, known as the real-time data problem (being the area to which Stevens alludes in the above quotation). The Reserve Bank of Australia first gave public recognition to this problem in 2002 with the publication of Research Discussion Paper No 2002-05 titled Real-time National Accounts Data.
To gain an appreciation of the importance of the real-time date problem in contemporary international economic policy analysis, go (for example) to the Organisation for Economic Co-operation and Development (OECD) Web site and insert the term “data revisions” in the OECD home page search box. Numerous studies relevant to that search will be returned.
John Rush’s view in 1972 that data revision was not only a fertile ground for economic research but also an area of significant importance to economic policy makers and analysts has indeed proved prescient!