Introductions (Chinese) and instrumental music ♪ ♪ ♪ Xiaohong Chen, I’ll be discussing some of her research connected to what I think of a sieve bath based methods of conditional moment estimation; copula based methods of her time series econometrics; and temporal dependence and nonlinear models. Now you know there’s a bunch of jargon attached into these different titles and I like and I’m would like to kind of try to indicate why I think I think these methods and these ideas are really truly important and I in economic practice. Xiaohong Chen; she developed methods for so-called that were so called it’s semi-parametric her along with others a prominent person in the audience here Peter Robinson has been a leader in this field. And this is very interesting very important piece of this and which you allow for some explicit structure you allow for imposing economic structure along some dimensions key parameters that have direct interpretations which is often important But flexibility along others is combination of putting together flexibility along some dimensions and and in specific structure along others How do you take a complex environment allow for flexibility some places but still try to tease out and still try to extract out the really key insights? So these contributions I view as being very complementary. The enhance our understanding of statistical complexity in different and very complimentary ways I learned about Xiaohong Chen from someone who would be really proud of you, Hal White I asked him about a problem on stochastic proximation and he said, “Oh, you should talk to Xiaohong Chen, this kid. She’s a brilliant and you should see what she’s doing.” And then that’s how I met her; I met her before I met her, just like Gregory. And so what she was doing was studying learning. She was applying stochastic approximation to models with agents inside who were trying to learn and they were learning by doing what Milton Friedman said. In a context much more ambitious than Friedman had in mind

So that’s just the beginning of Xiaohong’s work, but it had a theme that’s carried through, and Lars captured this really well. We’re learning about stuff, but Xiaohong’s works… A lot of the stuff that Gregory taught us how to do was learning about, and and for a good reasons, learning about models that were really tightly parameterized And there were good reasons for that which I’ll come back to. And a theme of Xiaohong Chen’s work was, “Are we paying too big a price for parameterizing as tightly as we had been doing?” And, “How would you learn about how big that price is?” Because there’s something that lurks behind this is, you cannot be totally general. Go back to Fineman You can’t solve Fineman’s problem unless you bring some prior information into it. A good example about this is around around 1900 or 1910, there’s some guys in physics trying to build models of the electron They’re really the smartest guys around They get nowhere because there’s not enough theory about the electron to tell them what to look for And that continues on to this day. So one of the themes is, the only thing that we can learn about, that statisticians or econometricians can learn about, is a model. It’s a model that restricts the data. And fundamentally it’s going to be finite dimensional. Or something like finite dimensional Xiaohong has taught me this; tried to teach it to me If you say, “I’m gonna let the data speak by themselves and not get in their way,” they’re gonna tell you nothing So the only thing you can learn about is a model. The message of Xiaohong’s work is a model both enables learning and it sets us up for mistakes. The kind of mistakes Lars was talking about. Sets us up for approximation errors So, what do you do in the face of that? So Xiaohong sent me off learning about information geometry, you know, divergence measures and things like that. So in terms of characterizing the kinds of errors you can make and how you can reduce them in a clever way as more data come in. So I think Lars mentioned this; Xiahong’s work tells us a lot about what we can get from big data and what we can’t get. And it teaches us things about how to design strategies for acquisition and more data and where that’ll help us and where it’s not going to be. And then Lars mentioned this there’s ambitious applications to nonlinear data reduction techniques. Some work with Lars. And she’s done work on asset pricing estimation of preferences and technology parameters in environments where she’s taught us what relaxing previous restrictions and parameters can bring some big insights There’s a lot that unites the work of Gregory and Xiaohong And I think it’s the best that economics has to offer It’s a tight language in terms of stochastic processes explicit models There’s no loose talk. Occasionally there’s a conjecture. It’s real science about matching models to data. It’s teaching the objects and we can learn about It shows the benefits of free trade from my country Because think how much I’ve learned from two

people that weren’t originally born where I was born And somehow I got to read their work and then I was lucky enough to meet them Thanks a lot unfortunately I’ve only met professor Chao a few times over the years, but by contrast, I’ve met professor Xiaohong Chen countless times In fact, for a couple of glorious years she and I were colleagues when I managed to persuade her to come to the LSE. This was too good to last and she returned to the US where her career has certainly flourished. Now the award of the China economics prize to Xiaohong has nothing to do with gender balance Certainly she’s had to work all the harder in order to succeed in a male-dominated profession. I can think of nobody male or female who’s more high-powered technically in econometrics than Xiaohong. And this is evidenced by her work on sieve estimation, copula, and other areas of nonparametric and semi-parametric econometrics. She’s already a very prominent figure in econometrics not only in terms of research, but in terms of her service to the profession and I think my prediction that she’ll become much more prominent has zero mean squared error. So finally let me offer my congratulations to both these distinguished scholars for this prestigious award. This takes me now; it’s a natural segue from Gregory’s work in recruiting brilliant Chinese economists, and brilliant Chinese students to study economics in the West, to one such example of a student that he recruited namely Xiaohong Chen who is a brilliant and highly productive scholar in mid-career In the case of Gregory there is a long life with many many achievements. In the case of Xiaohong there’s a relatively short life. she’s been out for maybe 20-25 years She’s mid career but she also has a very impressive body of work. And in some ways that body at work overlaps with Gregory’s interest in economic dynamics Like Gregory she’s worked on macro time series, she’s worked on the econometrics of learning the econometrics of financial models, just dynamic economic models, but also on a broad range of econometric tools. And in some sense Xiaohong represents the next generation of economists. She is the very best example of body of work her work is some of the very best example of what we now think of as called semi-parametric and non-parametric econometrics. When econometrics got going back in the 1930s and 40s – when Gregory was still a graduate student undergraduate – understanding linear equation systems was really a big deal and it was important. The linear models provided the economists with some initial insights as to how the economy would proceed And so linear simultaneous equations were the tools of the trade But as economists get better data and as they probe more deeply into the structure of the economy, they became aware of fundamental nonlinearities in the economy. And even some nonlinear phenomena where dynamics can be quite dramatic where there could be phased periods, transition periods, or the whole structure in the economy shifts. And Gregory, of course, contributed to that but a growing problem in the field of econometrics that came out I think of the application of these models in the 1980s was a concern that many of the empirical results coming out of these models were artifacts of assuming particular functional forms making particular assumptions about distributions and were essentially maybe artifacts and maybe not to be trusted There’s still an ongoing discussion in econometrics; some people call it the credibility revolution some people doubt that it’s the credibility

revolution is all that credible but that’s a separate issue. I think what is the case is that economists became much more aware about the importance of being careful about distributional and functional foreign assumptions. And so when Xiaohong was a graduate student and then in her early years she was deeply trained in the mathematical statistics that took economists to the next frontier. Tools that allowed for understanding the structure of the economy, but doing it without imposing a lot of parametric or distributional structure And therefore allowing much more flexible explorations of the economy. And so Xiaohong has been a pioneer in applying and developing and extending what is called non-parametric or semi-parametric econometrics Understanding how to build non-parametric and semi-parametric models that are more robust. Understanding how to test them Understanding how to apply them and extend them to time series settings with high degrees of non non-linearity and the processes generating the stochastic processes of the economy. And so Xiaohong’s work which continues unabated to this moment has led to a series of very influential papers that influenced the discussion of the in economics econometrics today. So for example, if you go through her work on nonparametric instrumental variables and moment conditions she has done some fundamental pioneering work. Measurement error which used to be the province only in linear models has now, in the hands of Xiaohong and some of her co-authors, become a major subject for investigation in nonlinear settings. Empirically nonlinear measurement error is found to be important validation the reliability studies are conducted. But until the Xiao Hong came along some of her co-workers and peers it was not possible to really consider it nonlinear measurement error but chohan’s work has made that possible and so in a series of studies not only measurement error but understanding the sensitivity of estimators understanding how to pool data from different sources in models that are highly nonlinear and semi-parametric and non-parametric Studying models of finance; volatility understanding how financial markets operate. Looking at models of learning which are essential to the macro economy When Gregory was writing models of adaptive, expectations were still quite current. Gregory also worked on rational expectation models. But Xiaohong has gone a little further, taking those rational expectations models, challenging them at times, considering them in nonlinear settings, and extending the work in a variety of fields. Not just in finance and macroeconomics, but also with application and microeconomics to the theory of the firm, and to understanding consumer behavior. So some of her work, for example, in finance is equally irrelevant to understand the habit formation in macro finance models as it is in habit formation in micro models of addiction, for example, or micro models of the formation of character and of skills and of beliefs. So Xiaohong’s work is kind of probe very deeply, but she’s probed deeply, widely and created a toolbox a whole series of tools that economists, including this economists, use. I mean both of your recipients have created tools that I’ve used and many other economists have used. But in particular, Xiaohong, now is mid-career and developing a very exciting body of work in a number of fields. And in that sense I think that maybe the prize that you’re giving today might be given again in 20-30 years when Xiaohong is a more senior scholar than she is now; she’s still quite young, still quite productive, and still quite active. I hope that you will consider maybe giving her a second prize because I’m confident in the next 20 or 30 years we’ll see a period of productivity that will equal or maybe even excel what Xiaohong has done