Peter Charles Bonest Phillips is a Sterling Professor of Economics and Professor of Statistics at Yale University. His research interests include econometric theory, financial econometrics, and applied macroeconomics. He is the founder and editor of “Econometric Theory”, and has advised over 60 Ph.D students in econometrics. In 2013, he was named a Thomson Reuters Citation Laureate for his contributions to the field of economics.
NB: Interview answers have been truncated.
Yale Economic Review (YER): What drew you to the field of econometrics?
Peter C.B. Phillips (PP): The University of Auckland in New Zealand offered an advanced graduate course in econometrics taught by Rex Bergstrom, who was New Zealand’s leading econometrician at the time.. Bergstrom did his Ph.D at Cambridge under Richard Stone, a student of John Maynard Keynes and a later Nobel Laureate in Economics. … it was extremely valuable to have a teacher like Bergstrom who had extensive overseas experience and an international reputation in the field.
As it happened I was the only graduate student in economics that year. So our classes were intensive sessions of discussion and analysis based on readings each week. Econometrics had a youthful invigorating energy that invited engagement and beckoned discovery. In the 1960s the frontier of the discipline seemed within easy reach for anyone with strong mathematics training and I felt myself being drawn in by the allure of its possibilities. Before long, I was in the grip of an academic passion and the research adventure had begun.
Two years later in 1971 I moved to the LSE and completed my doctoral work under Denis Sargan. He was First Wrangler (top of the class) in the Mathematics Tripos at Cambridge and the only economist ever to attain that distinction. The most important lesson I learned from Sargan was that no problem is too difficult to tackle. While much of the ongoing advanced work in mathematical statistics and econometrics at that time was concerned with asymptotic theory, Sargan was making massive headway in the more complex and challenging field of finite sample theory and exact distributions, which soon captivate[d]my own interest.
My dissertation work was on continuous time econometrics, a new field in econometrics at that time…which has since realized some of its enormous potential in the rapidly developing disciplines of mathematical finance and financial econometrics. Financial markets now generate vast datasets of high frequency observations, leading to market microstructure analysis, high frequency econometric methods, and algorithmic trading strategies that have transformed the industry.
YER: What prompted you to found the journal Econometric Theory, and what does this journal seek to achieve?
PP: The 1980s were a transformational period for econometrics. A revolution occurred in our approach to nonstationary time series. For the first time, we had a theory to explain regressions with stochastically nonstationary data, we developed tests to assess whether a regression was spurious, and we had a new formal framework called cointegration for modeling co-movement in economic time series. These new methods exploded into the profession and completely transformed empirical practice. On another edge of the discipline, the subject of microeconometrics had begun to emerge, using individual-level data across section and over time to learn about economic behavior and the policy impact of treatment[s].
The time was ripe for a journal outlet devoted to these developments. I proposed the idea to several publishers, received enthusiastic interest, and selected Cambridge University Press. I called the journal Econometric Theory to emphasize its goal of promoting innovative theoretical developments in econometrics.
YER: Could you tell us a little about your current research? Are there any problems or topics you would like to tackle in the future?
PP: My research is largely in the toolroom of econometrics where we forge the methods of empirical research. The tools are vehicles through which we force economic ideas to face the reality of observations, discarding what doesn’t work and identifying what does. Much of my own focus has been on trends. Trends are ubiquitous in economic discourse and they figure prominently in media commentary and congressional testimony on economic affairs. Trends also play a major role in economic theory, where we seek to understand such things as growth mechanisms in national economies, and in empirical work, where we seek to discover the main engines of change in economic activity.
One of the most significant developments in econometrics over the last 30 years has been the recognition that trends are stochastic and produce a random non- stationarity where data meander rather like the course of a river or a coastline. These meanderings substantially complicate the process of modeling and forecasting. The main body of my work has concentrated on developing this line of research and embodying the methods into the practical toolkit of applied research.
I have been interested in economic trends and non-stationary systems since the 1970s when I learnt how badly standard methods of inference work in a non-stationary environment. Since that time I have been involved in developing econometric methods for dealing with trends and studying their various forms. Trends are the hamlet of econometrics – inscrutable, unfathomable, and often unpredictable. You never know what they might do next. No one understands them, but everyone sees them in the data. Financial bubbles are another manifestation of non-stationarity in economics. Again, no one really understands them, but they keep appearing in the data. A fascinating field of research. I’m proud to be involved in it.
YER: Yale has one of the best econometrics programs in the world. Could you tell us what is special about econometrics at Yale?
PP: Yale econometrics IS special. Its lineage is unique, interwoven with the famous roots of the Cowles Foundation, the Econometric Society, and Yale’s own great tradition of applied economics and economic policy, going back through James Tobin to Irving Fisher.
The present program is distinctive in its technical strength, its coverage of the many different edges of econometrics, its quant support for economics and finance, and the quality of our students. Successive waves of brilliant students have passed through our program, adding to its versatility and widening the horizons of the subject. The productivity of both students and faculty has been enhanced by the support of the Cowles Foundation, which continues to play a major role in promoting econometrics as an empirical anchor to the discipline of economics.
YER: Econometrics is a technical field of economics that is unfamiliar to many students. How can a Yale student learn more about the field of econometrics? Also, how can a Yale student prepare for a career as an econometrician?
PP: Econometrics is one of the central pillars of the economics discipline, providing the avenue through which we confront economic ideas with observation. Evidenced-based analysis of this type is fundamental to all the sciences, including social science and business. Like much of the subject of economics, econometrics often seems technical to beginners and demands some training in mathematics, probability and statistics.
Within this setting, it is natural for students to find the subject technically demanding. But as in most disciplines, the ideas themselves are usually easy to grasp and intuition goes a long way to enrich and sharpen understanding. Additionally, econometric teaching programs now benefit from the availability of packaged software with a vast menu of options that enable easy implementation of techniques discussed in the classroom. These aids have transformed the teaching of econometrics, so that students engage with data analysis at an early stage in their studies. This engagement typically motivates us to learn more and look deeper at the methods. As our interest grows, the capacity to absorb more abstract technical work strengthens.
In the past, advanced undergraduate students in economics have often been motivated to take some of our graduate offerings in econometrics. They are then in a strong position to move directly into second year graduate courses and start research early. Students wanting to pursue a career in econometrics often take supplemental courses in statistics and mathematics.
But the frontiers of the subject are not all technically demanding. Some are more applied, many involve empirical model building and forecasting, others are algorithmic and computational, still others are conceptual. The edges of the subject multiply and, as they do, so does the demand for econometricians. Most departments of economics now have several econometricians on their faculty, each covering different subfields of the subject. Outside the education sector, the scope is large with a host of career opportunities in industry, banking, finance, government agencies, and international economic organizations.
YER: Despite the development of new analytical tools, we were unable to foresee the dot-com bubble, the housing bubble and the recent financial crisis. How have such incidents influenced the development of econometrics? Did this prompt the introduction of new tools and variables, or did these incidents simply reveal the intrinsic limitations of analytical tools?
PP: Financial bubbles are a form of nonstationarity, one where there is explosive or mildly explosive instability in the system followed by collapse. In his book Lords of Finance; The Bankers who Broke the World, Liaquat Ahamed tells us there have been 60 financial crises in the last 400 years. That’s 16 crises a century or 4 a generation. In spite of this history, financial theory has little to say about bubbles and the phenomenon is poorly understood by economists. Every crisis has its own special character, which often leads to the delusion during a run-up that ‘this time is different’ – there won’t be a collapse.
Global financial stability is a public good, like clean air and a healthy environment. After the recent crisis, it is now widely appreciated that we need an early warning device to alert central banks, financial institutions, and fiscal authorities of asset price exuberance, excessive credit creation, and potential threats to financial stability.
My own research has focused on building real time econometric detectors that can be used as a warning alert system. These detectors enable us to date-stamp crisis periods ex post, but their principal advantage is as a real time diagnostic in market surveillance.