Why I Became a Forecaster
In the last semester of my Bachelor of Commerce (Honours) at Monash University, I received an offer of employment with one of the major banks in Australia (the ANZ bank). At the time, I was getting ready to work in a large corporation. However, at the completion of the semester, the then head of department Professor Maxwell L. King approached me and my good friend and classmate Ashton de Silva (also an academic, now at RMIT University, Melbourne, Australia) and asked us to continue our studies by enrolling in the Master of Philosophy degree in Econometrics, as well as offering us a fractional teaching contract. This was a very sudden and unforeseen turn of events. It was completely “unforecastable”, but also very exciting. I had always wanted to become an educator, and this was the aspect of the proposition that really attracted me at first. I decided that I was going to continue, and called the bank to turn down their offer.
After completing the compulsory coursework in the Master’s degree, my next task was a research project. Inspired and guided by Professor Farshid Vahid (who subsequently became my PhD supervisor), I wrote a paper on Australian income inequality, which was published in the Economic Record. Seeing my work published in an academic journal was a thrill. I was offered a PhD scholarship, and expected to write my PhD dissertation on something related to income inequality, welfare, poverty.
Then, in 2003, the late Professor Sir Clive W. J. Granger visited Monash. It was his suggestion that prompted a change of direction to pursue the path of forecasting. My new topic of interest became multivariate ARIMA modeling, with a particular focus on forecasting. In my thesis, entitled “Essays on alternative methods of identification and estimation of Vector Autoregressive Moving Average Models”, I developed a methodology for identifying and estimating VARMA models. My 2008 publication in the Journal of Business and Economics Statistics, entitled “VARMA versus VAR for macroeconomic forecasting”, was the first in the literature to perform such a multivariate forecasting competition and to provide evidence that the models identified by the VARMA methodology proposed in my thesis forecast more accurately than the ubiquitous but theoretically unattractive VARs.
In 2006, I started a post-doctoral position with Professor Rob J. Hyndman. The topic of research was “tourism forecasting”. The aim was to build forecasting models with a focus on forecasting various aspects of Australian tourism. My research started to have an impact on industry, which made it very exciting. I started building forecasting models that for the first time were used by industry and had a direct impact on policy making at Tourism Australia. The first paper written based on this research (with Rob) reviewed the forecasts generated for Australian domestic tourism by the Tourism Forecasting Committee (an independent body responsible for generating consensus forecasts for Australian tourism) and published by Tourism Research Australia (the body responsible for providing research support to Tourism Australia and publishing Australian tourism forecasts). Our paper forecasted a decline in domestic tourism, in contrast to TRA’s published forecasts. We suggested that the published forecasts, especially the long-term forecasts, might be over-optimistic. These results triggered TRA to reconsider its forecasting practices and to revise their future forecasts downward.
In our second publication (with Rob and Roman Ahmed, our PhD student), we looked at forecasting methods for forecasting hierarchical times series. This paper proposed two new methods for forecasting time series data that belong to a hierarchical structure. The paper applied these methods to the forecasting of Australian domestic tourism demand disaggregated by geographical regions and by purpose of travel. It showed that the proposed methods produce more accurate forecasts than the typical bottom-up and top-down approaches. These methods were adopted fully by TRA for producing detailed regional forecasts for Australian domestic tourism flows.
I think that the feeling of excitement and satisfaction that you get from overcoming demanding research challenges, which I first experienced during my post-graduate studies, and then discovering the impact of my post-doctoral research on industry, is what attracted me to become a forecaster and has driven me to continue working on various aspects in the stimulating and ever changing field of forecasting. Besides my ongoing research project, I am currently writing (with Rob) a forecasting textbook entitled “Forecasting: principles and practice”. The entire book is online and free-of-charge at http://otexts.com/fpp. A print version and a downloadable e-version will be available soon to purchase on Amazon. We are writing the book for three audiences: (1) people who find themselves doing forecasting in business when they may not have had any formal training in the area; (2) undergraduate students studying business; (3) MBA students doing a forecasting elective. We use it ourselves for a second-year subject for students who are undertaking a Bachelor of Commerce degree at Monash University, Australia. Improving industry practices and seeing students flourish and pick up on cutting edge forecasting technology is very fulfilling.
Finally, being part of a large international forecasting community which thrives on innovation is also very stimulating. I look forward to the International Symposium on Forecasting every year. I have attended almost every meeting since my first one in 2004, which was in Sydney, Australia. In February, I organized (with Farshid) a two-day workshop on “Forecasting Multivariate Time Series” which was funded by research funds from Monash University and also supported by the IIF. The response was overwhelming. We had a two-day program filled with presentations from researchers from across the world, featuring their current research on issues and challenges related to multivariate modeling and forecasting. We are looking forward to the special issue of the International Journal of Forecasting with the same title, which will include some papers from this workshop (the end date for submissions to the special issue is June 30, 2013).
I hope to keep working in this field for a long time to come.