Professor Stephen Turnovsky

Professor Stephen Turnovsky

MA (Victoria University of Wellington), PhD (Harvard University), Doctorat Honoris Causa (Aix-Marseille University), Hon. DLitt (Victoria University of Wellington), Fellow Econometric Society, FASSA

Discipline: Economics

Year Elected: 1976

It was a pleasant surprise to learn recently that I was being honoured as a Jubilee Fellow of the Academy of Social Sciences in Australia.  But I must also confess some embarrassment, since my connection with the Academy has been very limited, following my departure for the United States just a few years after my election to ASSA.

While my contact with the Academy was restricted to attending a few Annual General Meetings in the years immediately after my election, the 10 years I spent at the ANU formed a crucial part of my career, and were a personally productive period on which I look back with great satisfaction and enjoyment.  Indeed, the ANU at that time provided an extraordinarily vibrant environment, with some of the most distinguished and talented individuals that one could hope to meet anywhere.  I remember that leaving the ANU and Canberra was difficult, and I suspect that with today’s virtually instantaneous communication I likely would have remained in Australia.

I characterize my research interests in economics as focusing on the dynamic adjustments of the macro economy, or “macrodynamics”.  I was first introduced to that discipline back in 1963 when, as a Junior Lecturer in the Department of Mathematics at Victoria University of Wellington in New Zealand, I presented a departmental seminar based on material in RGD Allen’s (1956) seminal book, Mathematical Economics.  The topics discussed included various business cycle models, developed by Paul Samuelson and Sir John Hicks, and the application of the stabilization rules pioneered by Bill Phillips.  This was all very appealing to me, since it seemed like an intriguing use of the methods and techniques that I had learned as an undergraduate majoring in mathematics.

The training I received taught me early on how to construct small dynamic models and to examine their dynamic characteristics.  This background was consolidated in the graduate training I received at Harvard University, as well as in my first academic appointments at the University of Pennsylvania and the University of Toronto.  This interest was pursued further when I joined the ANU and began working with John Pitchford and various students and coauthors, culminating in our co-edited 1977 North-Holland book, Applications of Control theory to Economic Analysis.  These were early days in the application of control methods to economics, and it has always been a great source of pride to me that the ANU was right at the forefront of these important developments.

I was fortunate to enter the economics profession during a period of profound transition.  The extensive application of formal modeling was just in its infancy when I began graduate school in 1964.  My background in mathematics placed me in a good position to ride that wave.  Over the last half century these methods have become the norm, although they have evolved in many dimensions.

In developing formal models of the economy, an important issue concerns the scale of the model.  This issue first arose in the 1960s with the development of the large scale Brookings Model and has surfaced periodically since.  I have always worked with “small analytical models”, being primarily interested in the insights that they have to offer in helping us understand economic principles and processes, with a particular interest in dynamics.  In my view, the main advantage of small models is that they discipline our thinking, enabling us to identify regularities that would surely be obscured in large complex models.

When I began studying technical economics in the mid 1960s, the focus was on developing models with the objective of using them to help us understand the economic responses to specific structural or policy changes.  The typical approach was to begin with a small model and then to extend it in various dimensions, trying to make it more realistic and more all-encompassing.  Today models are much more “data driven”, and motivated in terms of explaining observed empirical episodes and, in some cases, resolving “puzzles”.  Nowadays, the model is primarily a vehicle employed to study a specific economic problem.  The importance of the analysis rests entirely on the significance of the economic problem and the appropriateness of the model to address that issue, rather than in any intrinsic interest in the model, itself, or the formal technical skills that it may embody.  This reflects a natural maturing of the discipline, and is a welcome consequence of the substantial advances made to macrodynamic theory since the pioneering days.

As macro models have become enriched and their dynamic dimension increased, even small models have become less tractable analytically.  This has inevitably led to the increased application of numerical simulation methods, which in turn has led to the role of calibration, whereby specific functional forms and numerical values are assigned to key relationships and parameters.  This raises a couple of issues that are not necessarily dealt with satisfactorily in the literature.  First there is no well articulated methodology for choosing the required parameter values.  If the numerical simulation requires a certain number of  parameters to be set, this is typically done by reconciling the model with an equal number of (arbitrarily chosen) conditions, taken from various stylized facts, empirical estimates, equilibrium constraints, and optimality conditions.  There are usually many other constraints that could be considered, but are not.  We need to develop some rigorous systematic procedure that uses all available and relevant information in choosing the parameter values employed for our simulations.

Second, much of the current literature employing calibration aims at replicating a specific economy over some episode or time period.  My view of numerical simulations is a little different.  I regard them as a way of conducting comparative statics or comparative dynamics, in situations where the model is too complex to accomplish this analytically with any insight.  But this needs to be done over a range of plausible and realistic parameter values.  The parameters need not replicate any specific economy or period, but rather must be within the bounds of reasonableness, recognizing that parameter values do vary over time and between economies.

The study of economic dynamics has been (and still is) an exciting activity, and it has been a privilege and a pleasure to have had the opportunity to participate in that process.