Financial market interdependence

Francis Diebold, Kamil Yilmaz interviewed by Romesh Vaitilingam, 20 March 2009

Unfortunately the file could not be found.

Open in a pop-up window Open in a pop-up window


Download MP3 File (7.3 MB)




View Transcript

<p><em>Romesh Vaitilingam interviews Francis Diebold and Kamil Yilmaz for Vox</em></p>
<p><em>January 2009</em></p>
<p><em>Transcription of an VoxEU audio interview []</em></p>
<p><strong>Romesh Vaitilingam</strong>: Welcome to Vox Talks, a series of audio interviews with leading economists from around the world. My name is Romesh Vaitilingam, and today's interview is with Professor Kamil Yilmaz from Koc University in Istanbul and Professor Francis Diebold from the Wharton School at the University of Pennsylvania.<br />
We met at the American Economic Association's annual meetings in San Francisco in January 2009, where we spoke about their research on financial market interdependence, how markets move together in terms of their returns or their volatility.</p>
<p><strong>Francis Diebold</strong>: That's exactly right. It could be different markets, or it could be different assets in the same market, like different stocks in a stock market, or it could be cross country stock markets, or it could be a stock market and bond market and foreign exchange market, and so on, so any sort of combination. That's right, we&rsquo;re interested in how things move or don't move together, both returns on the one hand and volatilities on the other, and those things are actually quite different and the patterns are quite different.</p>
<p><strong>Romesh</strong>: So tell me how you go about doing the research. What are the data you look at? What are the techniques you use for trying to assess what is happening?</p>
<p><strong>Francis</strong>: Well, we use a technique that goes way back in empirical modeling called variance decompositions, and the variance decomposition is a standard tool in what is called vector autoregressive analysis, which most people would view as the great work horse, really, of modern empirical macroeconomics. A variance decomposition just answers the question how much of the variance of a forecast error of something is due to shocks in various other things, and that is what we use to assess these spillovers across financial market returns and return volatilities.<br />
We ask how much of the forecast error variance in forecasting a return is due to shocks in other returns, or how much of the variance in forecasting the volatility of returns is due to shocks in the volatilities of other returns.<br />
And we build up a natural spillover measure, based on that, by effectively adding that up across all the different assets that we are looking at.</p>
<p><strong>Romesh</strong>: And the data you are looking at?</p>
<p><strong>Francis</strong>: In the EJ paper, it was something like 17 or so.</p>
<p><strong>Kamil Yilmaz</strong>: 19.</p>
<p><strong>Francis</strong>: 19.</p>
<p><strong>Romesh</strong>: Kamil, talk us through it, then.</p>
<p><strong>Kamil</strong>: Six of them were developed markets, we can call, or hub markets; and the remaining 13 basically were a random choosing of stock markets in Latin America and East Asia. And in that regard, we covered perhaps 90% of the whole world's stock markets' capitalization. But, the data is basically the base indices in these countries. The base in the U.K. is FTSE 100, in the U.S, Dow Jones or S&amp;P 500 it doesn't matter what would be used. So, in this paper, basically we used stock market indices and the volatility of those indices.<br />
For the volatility, we have basically the range estimates where you have an opening price, the maximum price during the day or the lowest price, as well as the closing price of the day. So, basically, from these four data points, we calculate the range estimate for the weeks. So, our data is weekly, in that regard.<br />
So, we do a separate VAR for returns of those 19 countries, and a separate VAR for volatility, and we plot them together, the indices together.<br />
And basically, when we undertake the VARs and then look at the variance decomposition, this is like we work with 200 week windows, which are moving over time for returns as well as for volatility, and follow the value of the indexes, one single index for every VAR that we obtain, and then, we plot that over time.</p>
<p><strong>Francis</strong>: The results are quite interesting, actually, in that we find not much variation in spillovers over time for returns. There is some, but it tends to be smooth and trending gradually upward.</p>
<p><strong>Romesh</strong>: Over what kind of period?</p>
<p><strong>Francis</strong>: The saimple range here is what?</p>
<p><strong>Kamil</strong>: We go all the way to 1992.</p>
<p><strong>Francis</strong>: '92 to the present, more or less the present.</p>
<p><strong>Kamil</strong>: Weekly data.</p>
<p><strong>Francis</strong>: And due, perhaps, to increased financial markets integration, gradually as you move through the '90s and the '00s...</p>
<p><strong>Romesh</strong>: I was going to say, I assume the process of financial globalization would lead to closer&hellip;.</p>
<p><strong>Francis</strong>: Yeah, you see that, and if financial integration was smoothly and slowly improving in the '90s and '00s, which arguably it was, you might expect to see that in terms of gradually increased spillovers in returns. But, the thing that is really interesting is that the behavior of volatility spillovers is completely different. Volatility spillovers shoot up in various crises, whether that might be the Asia crisis of '97 or the Russian default of '98 or 9/11, and very much September and October of 2008.<br />
Although probably, I might be blending different things in my head because the EJ paper, for example, does not go all the way up. Or does it?</p>
<p><strong>Kamil</strong>: It goes until November 2007.</p>
<p><strong>Francis</strong>: 2007. But, in other work, we have done 2008.</p>
<p><strong>Kamil</strong>: But, we are following this. In the EJ paper, we also give a link. So every week, we recalculate this index. You can find the EJ paper link, which is at my university. So we update. Perhaps this weekend it is not being calculated because I am away. So basically, we are calculating this every week. So this index captures all these waves of this financial crisis, starting from February 2007, the first initial wave of the collapse of sub prime mortgage companies. So then, August 2007, and you can see in each case, especially in the volatility spillover case, the jumps in the index.</p>
<p><strong>Francis</strong>: So that is how we envision it. We envision this being used as a tool for people to monitor markets in real time, real time here being weekly, but almost real time in terms of tracking spillovers in returns and return volatilities and using that to inform judgments having to do with asset allocation decisions or macroeconomic policies or whatever. And it really is quite interesting to see the volatility spillovers jumping. And also coming back down, it is very interesting to see how these things evolve and how they evolve compared to what the popular perceptions are, based on arguably less sophisticated or less comprehensive sorts of measures.</p>
<p><strong>Romesh</strong>: How might you use this kind of information in macroeconomic policy?</p>
<p><strong>Francis</strong>: A big issue in macroeconomics is contagion or herd behavior, there are lots of different words that more or less mean the same thing, or spillovers, for that matter. There are a lot of debates ranging from do they exist, does contagion exist &ndash; reasonable people actually fight bitterly about that, as you probably know &ndash; to well, it exists, but how strong is it and what exactly are the patterns over time?<br />
What we're trying to do is just inform people in a rigorous replicable sort of way that really is nothing more than a transformation of a vector auto regression, which people fit routinely.<br />
So, it's just taking a standard dynamic model that people use all the time and transforming it in hopefully a clever way to get at this issue of contagion or spillovers, which people really didn't, I don&rsquo;t think, realize could come out of a vector auto regression so easily.<br />
There's a variety of ways in which policy makers might respond to such an index. Policy makers in particular are very, very concerned with systemic risks and systemic events, meaning basically situations where there is a spillover across a wide variety of assets within a class or asset classes within a country or countries within the world. Situations like that, for example, across countries would call presumably for more coordinated action across central banks.<br />
So, central banks for example, might need to know whether and when and where to be paying extra attention to coordinating versus situations where there is less need, things like that.</p>
<p><strong>Kamil</strong>: I agree with, Frank, especially, like the index tells you how fast the volatility spreads around a region. In East Asia there is actually some studies that are using the same, Hong Kong Monetary Authority using our framework to do this in the East Asia case. We did a similar study for Latin America. And so you can just take pockets of different regions to analyze, but how fast volatility spreads around this region or around the world and how much coordination is needed in that regard.<br />
So, is it just an idiosyncratic shock that stays in one country or is it going to other countries? So is it... So at the moment, for example, is it a systemic shock we are living through? And what Frank said about the returns spillovers, for example, which I think maybe he did not see the latest, but since September, even in the returns spillovers we observed a significant jump which we did not observe before.</p>
<p><strong>Romesh</strong>: That's basically historically unprecedented. It's very interesting. For the first time ever&hellip;</p>
<p><strong>Kamil</strong>: Yeah. We did not have such a jump before. And so, in that case, you cannot use this index for forecasting, I guess. But, it tells you what happens as of now and using that... So like we have weeks for the US, and it's the barometer of how risky the investment is, but this basically tells you how that riskiness spreads around the world.</p>
<p><strong>Francis</strong>: It's interesting. I was at a meeting the other day at the Central Bank of Chile; and this guy Claudio Borio from Bank of International Settlements in Basel had an interesting distinction. He likes to talk about barometers versus thermometers. Barometers have a forecasting aspect to them. They tell you if it's likely to rain or something tomorrow. Whereas thermometers just tell you what is going on now. What we're really doing and this is exactly what Kamil meant is a thermometer, as opposed to a barometer.<br />
Another buzz word that I like that people use is kind of doing nowcasting as opposed to forecasting. And nowcasting is very important because so many traditional economic methods or frameworks don't give you a picture of where you are now until way after the fact.<br />
The classic example is the NBER's designation of recession dates. They often wait a year or two years; and by the time that they finally announce it, it's of no value at all to anybody really trying to make plans and do things.<br />
So we're not forecasting, but we think that just by having a very accurate assessment of where we are at this instant, that actually is quite useful.</p>
<p><strong>Romesh</strong>: And very unusual to have an Economic Journal article that is right up to date. Nice to be able to do that if you could link to the latest numbers.</p>
<p><strong>Francis</strong>: Well, that is really our hope and intention and it&rsquo;s worked out well. As Kamil mentioned, the Hong Kong Monitoring Authority, other people at the IMF, are beginning to use this framework and put it to work in different contexts. What we've done so far, as we were saying is, in the EJ paper, a broad set of 19 equity industries around the world. In a different paper, we looked at a set of equity indices focusing in on Latin America, as well as the US, so North America basically. In work that we are going to present in about two weeks that is finished but not quite circulating yet, we look across asset classes. So, in everything that we've discussed so far, it's all equities, but in the new work, we're looking at equities, bond markets and FX.<br />
For example, you see very clearly a radically enhanced... let me just back up. We do something else in the new work, which is add to the framework in such a way that we can look at the direction of spillovers, not just how much spillovers are happening, but where are they coming from and where are they going to?<br />
And you can see very clearly, for example, in the recent 2008 events, a heightening of volatility spillovers from the credit markets to others. You can see exactly when that happens and you get a quantitative feel for how much is this spillover going up.<br />
So, that's where we're going with it. Also just in terms of the research program, one can go even further and we're working on this too, which is looking at spillovers not only in asset markets but in business cycles, in real economic activity across countries. People for many years have talked about the global business cycle and argued, &quot;Is there a global business cycle and, if so, how strong is the global component?&quot; and so on.<br />
Thinking about this spillover technology, but applied to things like GDPs or industrial productions of countries as opposed to asset returns, let's address that too.</p>
<p><strong>Kamil</strong>: Basically we are continuing with this&hellip; and that's over the last four months, from July onwards, the index is picking up substantially in terms of spillovers. So, that's basically showing that G6 countries, excluding Canada, they are really going into recession all together. So, I think there is a need perhaps even like if we can have weekly data for all these countries like... or there's a delay in the monthly industrial production data, so we can only use November data just in a week or so.<br />
But, it may be become possible at some date to use more timely data to check out whether really the recession, the shocks into the business cycles is really spilling over to other major economies.</p>
<p><strong>Romesh</strong>: Of course, there is a big discussion now about whether the emerging economies can hold up the declining economies of Europe and the US. You can inform that up to date.</p>
<p><strong>Francis</strong>: Hopefully, yeah.</p>
<p><strong>Kamil</strong>: We don't have China and India because of their times, but I think we can add them in a separate work, we can add them as well. At the moment, we are working on the G6 countries, but it's possible to add China; the only problem is data problems with China and India, it&rsquo;s not going as far back.</p>
<p><strong>Romesh</strong>: That's great. Kamil Yilmaz, Frank Diebold, thank you very much.</p>
<p><strong>Kamil</strong>: Thank you.</p>
<p><strong>Francis</strong>: Thank you.</p>

Topics:  Financial markets Macroeconomic policy

Related Articles

Paul F. and Warren S. Miller Professor of Economics, and Professor of Finance and Statistics, University of Pennsylvania

Professor of Economics, Koç University


CEPR Policy Research