Through the Conover-Iman test, this thesis has ranked risk magnitude among banks, insurers and NFIs. After this, this section would like to investigate risk change timing among the three sectors by comparing the dates of their risk turning points using structural break tests. Meanwhile, the structural break test aims to identify which data type provides earlier risk signal.
Following Arsov et al. (2013), autoregressive regressions and Quandt-Andrews break point (QABP) tests are performed to examine an unknown structural break point of each firm- level risk metric by sector. A break date obtained from one test is the time point when the risk of a firm, indicated by one risk measure, transformed from being in a peaceful mode to a volatile status. For each risk method, frequencies of all significant break dates obtained from
the 96 tests are grouped by sector, as demonstrated in Error! Reference source not found..
According to Error! Reference source not found., companies in each firm-level risk methodology have shown various break points, which implies that a shock may affect market participants at different levels. If the break date with the highest frequency (the highest bar in
the plot) represents the turning point of one metric, as can be seen from Error! Reference
source not found., the turning points are 2008 Q3 for volatility, 2009 Q1 for VaR, 2008 Q3 for
expected shortfall, 2009 Q1 for beta, 2011 Q4 for CDS spread, and 2008 Q3 and 2008 Q4 for Z-Score. According to the sector composition of the highest bar in each of the six sub-plots,
more banks experienced a structural change than insurers and NFIs. Comparing the dates of the highest bars in the six sub-plots, volatility and expected shortfall provide roughly earlier risk signals than other risk metrics. However, if focus on the earliest significant break quarter (the first bar) for each of the six risk evaluations, the turning points are 2007 Q4 for volatility, 2008 Q3 for VaR, 2008 Q3 for expected shortfall, 2008 Q3 for beta, 2007 Q4 for CDS spread and 2008 Q3 for Z-Score. In terms of the sector composition of the first bar in each of the six sub-plots, roughly more insurers22 transformed their risk modes relative to the other two sectors. Comparing the earliest significant turning dates of the six risk measures, volatility and CDS spread altered their risk status in the same earlier quarter, while all others moved to volatile modes later when Lehman Brother filed bankruptcy. Although volatility and CDS spread have shown the same earliest break point in 2007 Q4, the number of companies that changed structures in this quarter is higher indicated by CDS than by volatility. In other words, CDS delivers relatively earlier signals than equity- and accounting-based risk indicators.
Figure 7 Frequencies of All Significant Break Points for Each of the Firm-Level Risk Measures
The figure presents the frequencies of all significant structural break dates, grouped by industry, for every firm- level risk method–volatility, VaR, expected shortfall, beta, CDS spread and Z-Score. Break points for each risk method are obtained by the autoregressive regressions of the 96 firms in the sample and the Quandt-Andrews break point (QABP) tests on the regressions.
22 Exceptions: same number of insurers changed risk mode to that of banks (VaR), to that of NFIs (Beta), and more banks
than insurers experienced the break (expected shortfall)
Volatility
2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0
10 20
Banks Insurers Non-Financial Institutions
VaR 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0 5 10 Expected Shortfall 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0 5 10 15 Beta 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0 2 4 6 CDS Spread 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0 10 20 Z-Score 2005Q1 2005Q2 2005Q3 2005Q4 2006Q1 2006Q2 2006Q3 2006Q4 2007Q1 2007Q2 2007Q3 2007Q4 2008Q1 2008Q2 2008Q3 2008Q4 2009Q1 2009Q2 2009Q3 2009Q4 2010Q1 2010Q2 2010Q3 2010Q4 2011Q1 2011Q2 2011Q3 2011Q4 2012Q1 2012Q2 2012Q3 2012Q4 2013Q1 2013Q2 2013Q3 2013Q4 2014Q1 2014Q2 2014Q3 2014Q4 0 2 4 6