Research Unit for Statistical and Empirical Analysis in Social Scienses

Hi-Stat Vox No. 21 (November 24, 2011)

The 2011 Nobel Prize for economics has been jointly awarded to Professor Christopher Sims and Professor Thomas Sargent for their research on macroeconomics and econometrics. Although I immediately sent Professor Sims an email to congratulate him, as he used to be my supervisor at Yale University, I would like to offer my sincerest congratulations once again. It may seem somewhat disrespectful to address a Nobel laureate by his first name, but it feels odd to address Professor Sims as such because I have always known him as Chris, so that is what I will call him for the rest of this article.

After gaining a PhD in Economics at Harvard, Chris went on to teach at the University of Minnesota and Yale University, before arriving at his current position as professor at Princeton University. He apparently graduated second in his class from the Harvard Mathematics Department. I was fortunate enough to be taught by him as a graduate student at Yale, just after he had moved there from Minnesota.

Chris is well known for his work as an econometrician, which is primarily Bayesian in nature. In Bayesian estimation, model parameters are regarded as random variables, and the first step is to set the prior distribution, that is, the distribution of parameters before data are observed. This is then updated to the posterior distribution, which is calculated based on Bayes' theorem after data have been observed, in order to infer parameters. I once asked Chris why he favored the Bayesian approach. He replied by pointing to the Lucas critique, which argues that when government and central bank policies change, so do the model parameters, so that they should be regarded not as constants but as stochastic variables. Whereas Bayesian estimation was not commonly used in the past because it was difficult to derive the posterior distribution, it has started to become more widespread since the development of a simulation technique called the Markov chain Monte Carlo (MCMC) method, as explained below.

When Chris became my supervisor, I was conducting research on financial stochastic volatility (SV) models rather than macroeconomic models. As it is difficult to analytically calculate likelihoods using such models however, I spoke to Chris and he advised me to use the MCMC approach instead. That was what started me using MCMC. Used as a general term for methods whereby values are sampled and then used to sample the next set of values, MCMC methods make it possible to sample parameters from the posterior distribution even when using more complex models, thereby making it possible to use Bayesian estimation. MCMC methods had previously been used in physics, but they had only just started to be applied in econometrics at that time, so there was very little literature available. I learnt almost everything directly from Chris.

To estimate parameters using MCMC methods, we have to generate a lot of random numbers, so it takes a considerable amount of time. It might seem alien to young people nowadays, but back then computer CPUs were still at the Intel i386 stage, meaning that they struggled to perform calculations with any speed. Fortunately, Chris allowed me to use the lab next to his own, which was equipped with an i486 33MHz computer, the fastest available back then. Even so, estimating SV models still took the best part of a week. With Chris' help however, I managed to finish my dissertation and successfully completed my PhD. I felt like giving up on MCMC after that, because it just took too long. As computers have become significantly faster however (the time taken to estimate SV models has gone from nearly a week to around 15 minutes), I still use MCMC methods to this day. I don't know what sort of research I would be working on if I hadn't met Chris. He helped to make me the researcher I am today.

Chris taught a class in macroeconomics that was compulsory for first year graduate students at Yale and allowed me to work as his teaching assistant (TA). His lectures were hard to follow, to be honest, to the extent that he focused on transversality conditions for an entire term at one point. However, the TA sessions were fairly straightforward, as all I had to do was explain the contents of Chris' lectures in simpler terms.

Although Chris' work has touched on a wide range of areas, his most important piece of research to date has probably been on vector autoregressive (VAR) models. Using such models, it is possible to run simulations to illustrate policy effects. For instance, when the central bank lowers interest rates, you can simulate the effects that will have on macroeconomic variables over time. VAR models are therefore important to policymakers as well as researchers. In order to estimate VAR models however, identifying restrictions are needed. In fact, Chris has written numerous papers on what sort of identifying restrictions to impose. He is also using VAR models in order to carry out important research on causality between variables.

One of Chris' most important papers on VAR models, titled “Macroeconomics and Reality,” was published in the journal Econometrica in 1980. To commemorate the 30^th anniversary of the publication of that paper, we organized the “Journal of Economic Dynamics and Control Conference on Frontiers in Structural Macroeconomic Modeling: Thirty Years after ‘Macroeconomics and Reality’ and Five Years after ‘Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy’,” an international conference co-hosted by the Hitotsubashi University Global COE Program “Research Unit for Statistical and Empirical Analysis and Social Sciences” and the Bank of Japan's Institute for Monetary and Economic Studies. Chris delivered a keynote speech at the conference, which took place on January 23 and 24, 2010, at Hitotsubashi University's Mercury Tower, and I had the honor of chairing the event. I compiled a summary of the conference for the Global COE Hi-Stat Newsletter, so please feel free to take a look. It would be difficult to get a Nobel laureate to come and give such a speech, so I'm glad we invited him before he was awarded the prize. While he was here, we also had a great time going to see sumo wrestling, which Chris really seemed to enjoy. Many of the participants of the conference now regret that they did not take the opportunity to have their photo taken with Chris.

Following on from our conference, Seoul National University and the Korean Development Institute hosted a conference titled “Recent Developments in Dynamic Analysis in Economics - 30 Years after Macroeconomics and Reality” from May 25 to 27, 2011. The event was attended by students who had previously been taught by Chris at Minnesota, Yale and Princeton. Even Professor Thomas Sargent, the other winner of the Nobel Prize in economics, was able to attend. I was also invited to give a paper on VAR models, Chris' specialist subject. He asked me some penetrating questions that made me go pale and took me right back to my days as a student. I am currently revising my paper based on Chris' comments.

My research to date has revolved primarily around financial econometrics, focusing particularly on asset price volatility. In recent years however, I have finally started to engage in macroeconomic analysis, thanks to the increasing use of MCMC methods for the estimation of macroeconometric models, based on Chris' research amongst others. I may have a long way to go to live up to Chris himself, but I am determined to contribute in some way to the research he has carried out to date.

Original text in Japanese