Tuesday, September 30, 2014

Return Connectedness in Credit Default Swap Markets

Credit default swaps (CDS) on private and sovereign debt have become one of the major investment vehicles for investors since early 2000s.  That is why we applied our connectedness analysis framework to CDS markets as well. More specifically, we study the return connectedness among credit default swaps for sovereign debt of a group of developed and emerging market economies for the period from January 2010 to July 2014.  We focus on return connectedness rather than volatility connectedness in CDS markets, because the daily returns in CDS markets measure the fear in sovereign bond markets.

The full-sample static connectedness graph depicts a very interesting picture.  Returns for CDS on developed and emerging market government bonds form two sub-networks.  There are 15 countries in the developed country network sub-graph.  As denoted by node colors most of the developed country sovereign debt CDS generate low return connectedness to others. Among the developed countries, Belgium and Italy had the highest connectedness to others, followed by Spain, France, Netherlands and Austria.  U.S. has very low return connectedness to others, which means that it is likely to be a net recipient of return shocks from others.  The Southern, Spain, Italy, and Portugal, and the Northern European countries, Sweden, Denmark, Finland and Norway, form two subgroups that have high pairwise connectedness among each other.

The second sub-network graph includes Central and Eastern European economies as well as the major emerging market economies.  Among the 24 countries in this group there are only two developed economies: Japan and Israel.  While having very weak ties with this group of countries, Japan has no pairwise connectedness to other developed economies. Similarly, being located in a politically volatile region Israel also has stronger connectedness with emerging market economies than developed economies of Europe and the North America.

South Africa, Russia and Turkey are located at the center of the emerging market group and they have the highest return connectedness  to others over the 2010-2014 period, as shown by scarlet red color of their nodes.  In the emerging market group, Slovakia, Croatia, Bulgaria, Kazakhstan and Poland are located closer to the developed country group, which indicates that they are generating connectedness to developed country groups more than other countries in the emerging market group.  From their location, we can contemplate that Turkey, South Africa and Russia have more significant return connectedness to other emerging market economies in the CDS markets.

Central and Eastern European countries are mostly (with the exception of Kazakhstan and Poland) located in the northern half of the emerging market sub-network graph.  Among these countries Croatia, Romania, Hungary, Poland, Kazakhstan and Bulgaria generate more return connectedness to others. Latin American countries are located in the southern of the emerging market sub-graph.   Among the Latin American countries Brazil, Mexico, Panama and Colombia have higher connectedness to others compared to Chile, Venezuela and Peru. 

Wednesday, September 17, 2014

Episodes of high volatility connectedness among the global stock markets

Over the last two weeks we’ve being working on network graphs for 8 episodes in our dynamic global stock market connectedness analysis.  All these important dates are associated with significant increases in the total volatility connectedness index. In this rather long blog post, I analyze the pre- and post-event network graphs for each episode briefly.

     A.      Fed’s unexpected rate hike and the unwinding of carry trades – May 2006
     B.      First signs of the subprime mortgage crisis, late February-March 2007
     C.      Liquidity crisis, July-August 2007
     D.      As bank losses mount worldwide Fed was forced to cut rates drastically, Jan 22, and Jan 30, 2008
     E.       The U.S. financial crisis went global after the Lehman Bankruptcy on Sep. 15, 2008
     F.       EU's failure to act to contain the Greek crisis led to frenzy in the European markets, May 2010
     G.      Following the troubles of Italy and Spain, S&P’s decision on , August 5, 2011 to downgrade U.S. government’s credit rating from AAA to AA+ had worldwide impact
     H.      Bernanke’s speech on May 21, 2013 about the possibility of the start of QE tapering in late summer caught the markets off guard.

A. Let me start with the first episode, namely with the Fed’s unexpected decision to raise the interest rates an additional 25 basis points on May 10, 2006 and its indication that it would raise the policy interest rates one notch in its subsequent meeting on June 2006. This decision caught the markets off-guard, as the majority of market participants were expecting an end to the rate hikes in its May 2006 meeting.  While the bulk of the impact of this decision was felt in FX and bond markets, stock markets were also affected from the decision.  As a result, the volatility connectedness among the European stock markets increased significantly, as can be seen from the increase in the number of European stock markets with red and scarlet nodes as well as the increased thickness of the edges among the European stock markets.  European stock markets moved towards the center of the network with stronger volatility connectedness with each other as well as to the EU periphery and the major emerging market economies.  The “to” connectedness of the major emerging market economies also increased. U.S. stock market’s volatility connectedness to others, however, did not increase as a result of the decision.  This is expected because most of the carry trade originated from the U.S. and led to investments in the European and emerging market economies.  When the unexpected rate hike decision was taken it implied capital outflows from target countries toward the home country, the U.S.

B. Next in our list is the emergence of first signs of the subprime crisis in February 2007.  In the last week of February news about the troubles of major mortgage lenders, including the New Century Financial Corp, one of the biggest, became public.  While the total connectedness index jumped from 50% on Feb. 25, to 62% on February 28, 2007, while the net volatility connectedness of the U.S. stock market jumped from -3% to 20%.   In the network graphs, the U.S. stock market’s node turned from brown on Feb. 22 to red on Feb. 28, with significant increase in its volatility connectedness to Brazil, Canada, Mexico, and Argentina.  Major European stock markets’ nodes continued to stay red, but as can be seen from thick edges among them, high to-connectedness of the European stock markets was due to high pairwise volatility connectedness within Europe.

C. Liquidity crisis of July-August 2007 was a serious game changer.  For one thing, the connectedness index continued to climb from 60% in March to 70% level by late July 2007. This is the period when banks started to feel the pressure from the rapidly declining subprime mortgage market. By mid-June 2007, two subprime hedge funds of Bear Stearns had lost nearly all of their value. Then, in late July BNP Paribas decided to close three hedge funds of its own. This was the trigger that initiated the sell-off bank assets in the U.S. and Europe. The connectedness index jumped from 70% in late July to 78% by early September 2007. The “to” connectedness of the U.S., European, and Latin American stock markets were already high by the end of June 2007, as can be understood from red and dark red colors of their respective nodes.  By September 4, 2007, they all were turned in darker red colors and the edges among them became thicker.

D. In the last quarter of 2007, major U.S. banks worked hard to raise equity from investors all around the world. At the same time, they replaced their CEOs who were seen responsible for the large losses they suffered. But these moves were not sufficient to solve their problems. In the first couple of weeks of the new year, major banks started to announce further losses they suffered in the last quarter of 2007.  Finally grasping how desperate the situation of the U.S. financial system, Federal Reserve held an unscheduled meeting on an official holiday (January 22) and decided to cut the federal funds target rate by 75 basis points. Then, one week later, in their scheduled January 30 meeting, they decided to cut the fed funds target rate by another 50 basis points to 4.0%.  The comparison of the network graphs for January 7 and 28 reveal a situation we observe during the liquidity crisis episode. While the U.S., European and the Latin American stock markets (all on the “western hemisphere” of the network graph) had very high volatility connectedness to others on both dates, there is a definite increase in the “to” connectedness of these markets between the two dates.  Furthermore, the pairwise connectedness of these markets with each other also increased from January 7 to January 28, as shown by the thickness of the edges. Furthermore, the group of stock markets that generated high “to” connectedness moved slightly to the center of the network graph.

E. The total connectedness index declined steadily for the next 7 months; even during the takeover operation of Bear Stearns. By the end of August 2008, the index declined to 62%. With the U.S. Treasury’s decision to take Government Sponsored Enterprises Fannie Mae and Freddie Mac under full government custody in the first week of September led to an increase in the index. Then came the Lehman bankruptcy on September 15, 2008. The index jumped to 70% on September 15, and all the way up to 78% by November 27, 2008. These were the major steps toward the globalization of the U.S. financial crisis.  As of October-November 2008, all major stock markets and economies of the world were affected from the whirlwind of the financial crisis. Three of the four network graphs we display in this episode of the crisis definitely show the worsening of the situation in the rapidly increasing total and directional volatility connectedness measures. By November 27, the edges at the center of the network graph were so thick that the white background can be seen only as white dots.

F.   After hitting a low of 60% at the end of June 2009, the total connectedness index increased in the rest of the summer to reach 66% by mid-November. While earlier on troubles in European banks led to the increase in connectedness, starting from November, it was an outcome of the revelations about the Greek fiscal deficit and debt stock. As the newly elected Greek government revealed that the government budget deficit and the debt stock was actually much higher than announced by the incumbent government, it became clear that the losses of the European banks holding the Greek debt would mount to billions of euros. Markets reacted and the volatility connectedness index moved up by 5 percentage points from December to late April 2010.  However, after a brief lull in April 2010, the index jumped by 6 percentage points in May 2010, as it became evident that the EU leaders were not really serious about a resolution to protect the European banks. 

G. The index did not stay low for too long, thanks to the increased worries about the sovereign debt and banking problems in Italy and Spain, two EU members with sizable economies compared to the members that had problems before. Then following very intense political negotiations U.S. Republicans and Democrats finally agreed on August 2 to raise the debt ceiling temporarily. A couple of days later S&P decided to lower the credit rating of the U.S. government from AAA to AA+. In this episode, the connectedness index went up from around 60% to 76% in just a couple of days. The pre- and post-downgrading network graphs clearly show the increased number of stock markets that had high volatility connectedness to other countries.

H. The last episode we focus on is the one that occurred in May-June 2013.  In his speech to the U.S. Congress, on May 21, 2013 Federal Reserve Chairman Ben Bernanke caught market participants by surprise by spelling out that the tapering of the quantitative easing program can be sooner than markets anticipated.   While he wanted to prepare the market participants for the eventual initiation of tapering, the markets showed significant reaction to his speech.  For a couple of weeks financial market gyrations increased. During this period, the index increased from 52% to 60%. Aside from the usual European suspects, countries such as Austria, Czech Republic, South Africa, Russia, Chile and Brazil started generating substantial volatility connectedness to other countries during this episode. 

Tuesday, September 2, 2014

Financial and Macroeconomic Connectedness website is up today !!!

I started this blog more than two years ago. At the time my aim was to build the Financial and Macroeconomic Connectedness (FandMC, for short) website and use this blog to analyze important developments in financial markets around the world as revealed by applications of the Diebold-Yilmaz Connectedness Index (DYCI) methodology.  The website is finally up thanks to the hardwork of my student/assistant/coauthor Mert Demirer.

At the moment, the FandMC website presents the static (full sample) and dynamic connectedness analyses for global stock, bond, foreign exchange and CDS markets.  The website aims to provide academics and policy makers with a powerful tool to analyze how return or volatility shocks to individual assets/markets other assets/markets in a country or around the world.  If you are not already familiar with the DYCI framework I would recommend that you have a look at our Journal of Econometrics (DY 2014) paper listed in the Research section of the website.

In 2014, with the help of shrinkage techniques, such as lasso and elastic net, we have expanded the DYCI framework from small to large scale network analysis.  Three of the four (excluding bonds) financial connectedness analyses presented in the Indices section are estimated using the elastic net procedure.  As the dimension of the network increases, the connectedness analysis through tables becomes rather difficult.  The volatility connectedness analysis of Bond market Thanks to Gephi, a powerful network graphical display program, we can now generate network graphs for the full sample (“static network) and for each instance in our dynamic analysis. 

Obviously it is impossible to create network graphs for every day, but from now on we will try to create network graphics that display the connectedness of assets/markets before, on or after important dates, such as September 15, 2008, that witnessed the bankruptcy announcement of Lehman Brothers. To give an example, in our dynamic volatility connectedness index graph for global stock markets, we highlighted four days. When you click on letter C you will see the network graphics that show how volatility shocks spread among the major stock markets of the world on September 25, 2008, ten days after Lehman’s collapse.  
In addition to the static network graphs,  FandMC website presents graphs of directional connectedness indices (to, from and net) for each asset/market that allows for the dynamic analysis of connectedness in financial markets. The displayed graphs are small, but by clicking the enlarge button it can be viewed quite easily.  Furthermore, all graphs in our website can be printed or downloaded.  The data  underlying the static and dynamic connectedness graphs can be download from the Data section.

All in all, we have tried to include as much as possible in this first run of the FandMC  website.  We will continue building and developing the website further.  In this endeavor we welcome your comments, and suggestions.