ISSN ECONOMICS DISCUSSION PAPER SERIES. International yield curve comovements: impact of the recent financial crisis

Size: px

Start display at page:

Download "ISSN ECONOMICS DISCUSSION PAPER SERIES. International yield curve comovements: impact of the recent financial crisis"

Deborah Griffin
5 years ago
Views:

1 ISSN 7-7 ECONOMICS DISCUSSION PAPER SERIES International yield curve comovements: impact of the recent financial crisis Simeon Coleman, Kavita Sirichand. WP 7 School of Business and Economics Loughborough University Loughborough LE TU United Kingdom Tel: + () 9 7 Fax: + () 9 9

2 International yield curve comovements: impact of the recent financial crisis Simeon Coleman a, and Kavita Sirichand b b School of Business and Economics, Loughborough University, Loughborough, LE TU, UK. s.coleman@lboro.ac.uk b School of Business and Economics, Loughborough University, Loughborough, LE TU, UK. k.sirichand@lboro.ac.uk Abstract Empirical evidence on international yield comovement is sparse and lacks consensus. Employing a dynamic correlation approach, we show that during the recent global financial crisis euro area yields have ceased to comove with the yields of the other international markets - Canada, UK and US. Some implications of our results are discussed. Keywords: Interest rates, comovement. JEL Codes: E, F. Introduction During the recent global financial crisis (GFC), there appears to be significant similarity in the monetary policy being adopted across several key economic areas (see Figure ). Some existing studies suggest that if we observe similarities in the monetary policy responses of Central Banks, we should observe a high degree of comovement across international yields (see Jothikasthira et al., ; Sutton, ). However, there is also evidence to suggest that the dynamics of both short and long rates may be both maturity- and country-specific and any a priori generalizing assumptions may be misleading (see Coleman and Sirichand, ). So, where does this leave the stakeholder in international markets? In this paper, we investigate the linkages in yields across four international markets. In particular, we seek to ascertain the similarity in behavior and the degree of comovement between the UK, US, Canadian and euro area yields and examine the dynamics of their comovements over a period which includes the recent GFC. Corresponding author K. Sirichand, address: k.sirichand@lboro.ac.uk Phone: +()97.

3 Figure : Central Bank Base Rates () March 997 to February 7 US UK Can euro Source: Data compiled from Bloomberg () The macroeconomics literature points to the interest rate channel as being key to linking the stock markets of different countries. For example, Kim (), Canova (), Wongswan (), and Ehrmann and Fratzscher (9) show the transmission of US monetary policy to equity markets of other countries through the reaction of US short rates, foreign short rates and exchange rates. Moreover, studying the similarities in comovement between yields may help us better understand the channels through which crises and contagion spread i.e., comovement and contagion rely on similar linkages between economies (see Dungey et al., ; Zaharia, ; Bai and Zhang, ; Baur, ). Moreover, comovement has implications for portfolio allocation, in that fundamental theory suggests that diversification can successfully reduce unsystematic risk when assets are imperfectly correlated. Many studies have investigated stock market and business cycle comovement (including Campbell (99), Karolyi and Stulz (99), Ammer and Mei (99), Yetman (a)). A widely held view is that there is common movement of returns on individual financial assets within national and amongst international markets (see Barberis et al. () for stocks; and Sutton () and JLL for bonds). However, there also appears to be a lack of consensus regarding the extent of comovement in international equity markets (see Laopodis, ). Further, JLL highlight several periods (UK-US in 997; and UK and US versus Germany in 99/) when there was comovement at the long end, even when there was divergence at the short end. Against this background, how should an international portfolio be managed to maximize the gains of diversification? Conspicuously, limited attention has been paid to international yield curve comovement.

4 Recently, the relationship between interest rates across maturities has come under more scrutiny, with some focus on its impact on monetary policy formulation (see Rudebusch et al., ; Atkenson and Kehoe, 9). Furthermore, an international perspective has been suggested, with empirical evidence, pointing to the importance of global factors across countries (see Bernanke, 7; JLL; and Kose et al., ; ) and Sutton, ). They argue that macroeconomic shocks impact the term structure of interest rates through two main channels: the policy response channel and the risk perception channel. The reasoning, under the policy response channel, based on the rational expectations hypothesis, is that, if national policymakers formulate their policies in reaction to global factors, then a global business cycle is likely. Under such circumstances, central banks mandate to ensure macroeconomic stability would suggest similar policy reaction functions and imply comovement amongst the short-term rates of these regions. Further to this policy response channel, JLL argue that macroeconomic shocks also impact the term structure of interest rates through a risk perception channel, where the risk compensation demanded changes with investors perception of risk as there are changes in the economic and financial environment. We observed a global shock in the recent GFC, to which we would expect similar monetary policy responses, which is, in fact, what we do observe, see Figure. Given the similar monetary policy responses, empirical findings from the existing literature suggest that we should observe high degrees of comovement across international yields. In this paper, we investigate if this is the case. We consider whether there is comovement between the term structures of the UK, US, Canada and the euro area. If there is, what is the extent of this comovement? and finally, we investigate if, and how, any such comovement has changed over time and analyze separate pre- and post-global financial crises samples. The analyses also attempts to determine whether the euro members considered to be less risky (based on rating agency rankings) differ in their comovement with the other international markets considered. Broadly speaking, studying yield comovement could help us better understand the channels through which crises and contagion spread. More specifically, our analyses should inform the portfolio composition decisions of an investor seeking to limit their exposure to unsystematic risk by diversifying across these international markets. We find robust evidence of increasing comovement only amongst three of the four international markets considered. Intuitively, at least from the policymakers viewpoint, the expectation is that the reaction to and impacts of these policies, vis-à-vis the global shocks, would also be similar. One of the aims of this study is to ascertain if international yields comove and posit potential reasons for any observed patterns of comovement, as such, we do not explicitly model the yields.

5 The remainder of the paper is as follows: Section summarizes the empirical methods and the main results, and Section concludes.. Methods and results. Data We employ weekly yield data for the UK, US, Canada, and the euro area, and consider four points on their government liability curves. Specifically, maturities of,, and months (hereafter,, and ), over //997 - // for Canada, the UK and US; and /9/ - // for the euro area (due to data availability). We consider two sets of yields for the euro area: (i) euro area which considers yields for central government bonds for all member states and (ii) euro AAA which considers the yields for the AAA-rated euro area central government bonds only, i.e., debt securities with the most favorable credit risk assessment. We make this distinction aiming to identify any differences between core and peripheral countries in the euro area. Differences between the comovements with the euro area compared to those with the euro AAA group will underscore the importance of the risk perception channel in this type of analyses. It is worth noting, however, that the composition of countries included in the euro AAA group is not static, and is subject to change. As can been seen from Appendix A, in 997 the AAA-rated group would have contained many of the euro area member states, but, interestingly, only five states debt had the top level AAA rating in January. Table and Figure provide summary statistics and a graphical representation respectively. Visually, we observe a general comovement between the regions for each of the maturities, and an overall decline across all yields over the sample. Moreover, the yields appear to exhibit closer comovement at the long end (y and y) than at the short end (y and y). Furthermore, a clear divergence between the euro and euro AAA yields during the crisis period, across all maturities, is apparent. This divergence in government yields is a reflection of the debt crisis affecting many euro area member states. We employ Wednesday observations of continuously compounded nominal government spot rates i.e. official central bank yield curve data from: Bank of Canada: Bank of England: the US Treasury: and the European Central Bank:

6 Table : Summary Statistics of Yields Series Mean Std. dev Maximum Minimum Obs. UK month Can US euro area euro AAA month -month -month UK Can US euro area euro AAA UK Can US euro area euro AAA UK Can US euro area euro AAA Notes: Descriptive statistics are presented for weekly observations of the -, -, - and -month yields over //997 to // for the UK, Canada and the US (7 observations), and over /9/ to // for the euro area (7 observations). UK -month has a slightly shorter sample, ending // (7 observations).

7 Figure : Plots of Yields 997 to 7 month 7 month 7 month 7 month UK Can US Euro Euro_AAA month month

8 month month Euro Euro_AAA. Empirical methods Our aim is to investigate the potential relationships between the yields of the different regions by considering their comovement. Typically, the Pearson Correlation Coefficient (hereafter PCC) has been a popular measure to analyze comovement, partly due to its simplicity. However, a major limitation, particularly with high frequency data, is the narrowness of the time-specific information it provides. In this study, seeking to analyze dynamic comovement between the markets, we follow Yetman (b), who highlights the PCC s inefficiencies at adequately describing business cycle comovement at higher frequencies and proposes an improvement based on a z-score. In brief, the comovement, at time t, between series y of regions i and j can be estimated by the product of their respective z-scores:. () In this study, aiming to establish the extent of any comovement, we construct for each area-pairing (i.e., C pairs) and maturity using equation (). Positive values of imply comovement, i.e., and are either both positive or both negative; whereas, a negative See Ammer and Mei (99), Karolyi and Stulz (99) and Bin et al. () for some applications; and Yetman (a) for a theoretical background. As Yetman (b) points out, up to one-degree of freedom correction, the average of the comovements in each time period as given by Equation () will equal the PCC, i.e., where the PCC follows a Student s t-distribution. 7

9 value of implies movement in opposite directions. For any pairing, the relative size of is indicative of the extent of the observed comovement. Further, we examine how these comovements have evolved over time, both graphically and by regressing on a time trend: () Equation () is run over two sample periods: (i) pre-crisis (Pre), //997 - //7 ( obs.) for the UK, US and Canada; and /9/ - //7 ( obs.) for the euro area. and (ii) post-crises (Post), includes the crisis period and spans //7 - // ( obs.) for all the four areas. We note that, for the UK, there is a slightly shorter post-crisis sample, until //, for y. Splitting the sample up in this way allows us to comment on the GFC s impact on the observed comovement. A positive slope coefficient suggests the comovement is increasing with time.. Results and discussion We summarize our analyses on comovement as follows: Table presents descriptive statistics of ; Table summarizes the proportion of positive comovements; Figure and Table then illustrate how these comovements have evolved over time. The estimated comovement statistics, Table suggest some interesting patterns, we highlight the salient ones. First, over the entire sample (Table, Column ), the UK yield curve exhibits strong comovement with both the Canadian and US curves, with average exceeding. across all maturities. This is particularly noteworthy, since by contrast, and interestingly so, the UK yield curve displays significantly less comovement with that of its geographically closer neighbour, the euro area, with coefficients of.9 and. for y and y respectively, and negative correlations of -. and -. for the longer end, y and y respectively. This finding suggests that geographical proximity, economic ties and even trade links do not necessarily infer similar yield curve behavior. Further, we find strong yield comovement between the US and Canadian with average values greater than.9 across all maturities. It is important to note here that we do not suggest that significant economic ties should imply similar market behavior, however we are positing that such close ties would present a higher potential for financial market integration.

10 Table : Descriptive Statistics of Yield Comovements Series Mean Std. dev Maximum Minimum Mean Full Sample Pre Post -month -month -month -month UK-Can UK-US UK-euro UK-euro AAA Can-US Can-euro Can-euro AAA US-euro US-euro AAA UK-Can UK-US UK-euro UK-euro AAA Can-US Can-euro Can-euro AAA US-euro US-euro AAA UK-Can UK-US UK-euro UK-euro AAA Can-US Can-euro Can-euro AAA US-euro US-euro AAA UK-Can UK-US UK-euro UK-euro AAA Can-US Can-euro Can-euro AAA US-euro US-euro AAA Notes: Summary statistics are presented for the pairwise comovements,, of each n-month yield,,, and, over the full sample. Average comovements for Pre (//997 to //7 ( obs.) for the UK, US and Canada, and /9/ to //7 ( obs.) for the euro area) and Post (//7 to // ( obs.) for all) sub-samples are reported in Columns and. Entries in bold indicate the lower comovement over the pre- and post-crisis periods. The UK -month has a slightly shorter post-crisis sample, ending // ( observations). 9

11 Their respective comovement with the euro area yields are low and of similar magnitude to those between the UK and the euro area. Interestingly, the comovement between the euro area and the other areas varies across maturities, with positive comovement at the short-end and negative comovement at the longend of the yield curve. Despite the similarity in the central bank policy response of these areas to the recent crisis (as in Figure ), yield comovement between the UK, US and Canadian pairings with the euro area is found to have become more negative following the GFC. We posit that this is due to the risk perception channel becoming more dominant. That is, the observed decrease in comovement relative to the euro area may be explained by the change in investors perception of risk in relation to their view of the long term prospects of the euro area e.g. owing to the troubling sovereign debt of some member states and its implications for the future of the union. The flight to safety as a result of such changing perceptions may explain the observed negative yield comovements between the euro area and the other regions. Table : Percentage () of positive yield comovements Series -month -month -month -month UK-Can 7 (7) 7 () (7) (7) UK-US 7 () (7) (7) (7) UK-euro () 77 () () () UK-euro AAA 9 () 9 () (9) (9) Can-US 9 (77) 9 (77) (7) (7) Can-euro 7() (9) () () Can-euro AAA 7 () 7 () () (77) US-euro 7 () () () () US-euro AAA 79 () 7 (7) () () Notes: The values above express the number of positive comovements observed as a percentage of the total. A positive value suggests that the yields for both countries moved in the same direction. The values in the parentheses (.) give the actual number of positive comovements. The total number of observations is 7 for the UK (UK -month is 7), Canada and the US, and 7 for the euro area. From Table we note that the UK, US and Canadian yields comoved 7-9 of the time. 7 For the UK-euro area pairing y and y move together 79 and of the time, but only and of the time for y and y. Similarly, the Canada-euro area and the US-euro area pairings exhibit positive comovement between and 7 of the time at the short end, and between and at the long end. When considering how closely the UK, US and Canadian curves move with that of the euro area, there is a clear disparity between 7 Total number of observations is 7 for the UK, Canada and the US, and for the euro area.

12 the short end and the long end of the yield curve. In each of these pairings, there is higher degree of comovement between the euro-aaa area and the UK, US and Canada, than there is between the pairings of entire euro area and each of the other markets. Arguably, this observation underscores the importance of the risk perception channel. In fact, our finding suggests that even within the euro area alone there is noticeable difference between members based on how risky they are perceived to be. Martin and Zhang () provide some evidence to support the existence of a flight to quality from the periphery to the core member states of the euro area. Against this background, our finding of different dynamic comovement estimates for the euro area and the euro AAA group with the other markets is credible. Figure : Comovement,, over th March 997 to th October month month month month UKCan UKUS UKEuro UKEuroAAA CanUS CanEuro CanEuroAAA USEuro USEuroAAA

13 To see how changes over time, both in terms of direction and magnitude, Figure plots for each country pair. Comovement between each UK, US and Canada pairing is positive throughout, indicating yield movement in the same direction. The magnitude of the comovement is high pre-march and post-march 9. Moreover, the comovement between UK-euro, Can-euro and US-euro yields is positive through and (and in for short rates only). This is followed by negative comovement over -, which highlights that over the main part of the crisis period the yields, in Canada, the UK and US all moved in the opposite direction to those in the euro area. Further, the magnitude of these negative estimates appear to increase with time, suggesting an increasing divergence. We find that 9 confidence interval bands for each of the above comovement coefficients (see Appendix B), indicate that the lengths of the intervals are of the same magnitude. Hence, the inferences we made above, based on the averages, remain valid. Table : Regression of on time -month -month -month -month Pre- Post- Pre- Post- Pre- Post- Pre- Post- UK-Can -. ***. *** -. ***. *** -. ***. *** -. ***. *** UK-US -. ***. *** -. ***. *** -. ***. *** -. ***. *** UK-Euro. ***. ***.9 *** -. ***. *** -. ***. *** -. *** UK-EuroAAA.7 *** -. ***.9 *** -. ***. ***. ***. *** -. *** Can-US -. ***. *** -. ***. *** -. ***.9 *** -. ***. *** Can-Euro. *** -. ***. *** -. ***. *** -. *** -. *** -. *** Can-EuroAAA. *** -. **. ***. **. ***. *** -. ***. *** US-Euro. ***.. *** -. ***. *** -. ***. *** -. *** US-EuroAAA. ***. *. ***. **. ***. ***. ***. Notes: The regression of the comovement coefficient on a time trend,, is run separately over the Pre and Post sample periods. The table above presents the estimated coefficient with its corresponding p-value in parenthesis (.). The Pre period runs from //997 to //7 ( obs) for the UK, US and Canada, and from /9/ to //7 ( obs) for the euro area. The Post period runs from //7 to // ( obs) for all countries. Table reports the estimates of the slope coefficient in Equation () in each case. We note that, though small in size, almost all are significant, which is unsurprising since there are prolonged periods where is either increasing or decreasing with time. For UK-Can, UK-

14 US and Can-US, the coefficients are negative in the Pre period and positive in the Post period, across all maturities and, generally, the coefficients are of a smaller magnitude prior to GFC. For UK-euro, US-euro and Can-euro, the opposite is true the slope coefficients are negative, suggesting that during Post period, comovement has decreased with time.. Conclusions In this paper, we analyze dynamic yield comovement between the UK, US, Canada and the euro area, over a period which includes the recent financial crisis. We find that the Canadian, UK and US yields comove, but they do not comove with the euro area to the same extent; and, strikingly, an increasing divergence unfolds over the recent crisis. Our results also point to a higher degree of comovement between the UK, UK and Canada with the euro AAA group, and much less so with the entire euro area. Although similar monetary policy responses of these regions to the crisis suggest positive yield comovement, our finding of decreasing, and even negative, comovements lead us to posit that changes in risk perception have been the dominant channel through which the crisis has impacted comovement, and our results also clearly show that the crisis has changed the interaction of yields in these international markets. The evidence from this analysis underscores the need for targeted investigation of the dynamics of yields prior to undertaking diversification and portfolio rebalancing for effective risk management, with adequate emphasis being given to the potential for global shocks to have asymmetric impacts across international markets, evidenced by the change in comovement we observe pre- and post-crisis, particularly with the euro area.

15 References Ammer, J. and Mei, J. (99). Measuring International Economic Linkages with Stock Market Data, Journal of Finance, (), 7-7. Atkenson, A. and Kehoe, P.J. (9). On the Need for a New Approach to Analyzing Monetary Policy, NBER Macroeconomics Annual,, 9-. Bai, Y. and Zhang, J. (). Financial integration and international risk sharing, Journal of International Economics,, 7-. Barberis, N., Shleifer, A. and Wurgler, J. (). Comovement, Journal of Financial Economics, 7, -7. Baur, D. (). Financial contagion and the real economy, Journal of Banking & Finance,, -9. Bernanke, B.S. (7). Globalization and Monetary Policy, speech at the Fourth Economic Summit, Stanford Institute for Economic Policy Research, Stanford, California. Bin, F-S., Chen, D-H., and Chiu, P-C. (). The Nonlinear Dynamic Spot-Futures Relationship surrounding the Tick Size Reduction on the Taiwan Stock Exchange: A Threshold VECM Approach, Journal of Money, Investment and Banking,, 7-. Campbell, J., Shiller, R., (99). Yield spreads and interest rate movements: A bird s eye view. Review of Economic Studies, 9. Canova, F., (). The transmission of us shocks to Latin America. Journal of Applied Econometrics (), 9. Coleman, S. and Sirichand, K. (). Fractional integration and the volatility of UK interest rates, Economics Letters,, -. Dungey, M., Fry, R., Gonz alez-hermosillo, B and Martin, V. (). Contagion in international bond markets during the Russian and the LTCM crises, Journal of Financial Stability,, 7. Ehrmann, M., Fratzscher, M., (). Global financial transmission of monetary policy shocks. ECB Working Paper. Jotikasthira, P, Le, A. and Lundblad, C. T. (). Why Do Term Structures in Different Currencies Comove? UNC Kenan-Flagler Research Paper No. -. Available at Karolyi, G. A. and Stulz, R. M. (99). Why Do Markets Move Together? An Investigation of U.S.-Japan Stock Return Co-movements, Journal of Finance, (), 9-9. Kim, S., (). International transmission of U.S. monetary policy shocks: Evidence from VAR s. Journal of Monetary Economics (), 9 7.

16 Kose, M. A., Otrok, C., and Whiteman, C. H. (). International business cycles: World, region, and country-specific factors. American Economic Review 9(), 9. Laopodis, N. T. (). Portfolio diversification benefits within Europe: Implications for a US investor, International Review of Financial Analysis,, -7. Martin, F. and Zhang, J. (). Correlation and volatility on bond markets during the EMU crisis: does the OMT change the process? Economics Bulletin, (), 7-9. Rudebusch, G. D., Swanson, E.T. and Wu, T. (). The Bond Yield Conundrum from a macro finance perspective, Monetary and Economic Studies,, -9. Sutton, G. D. (). Is there excess comovement of bond yields between countries? Journal of International Money and Finance, 9, 7. Wongswan, J., (). The response of global equity indexes to U.S. monetary policy announcements. FRB International Finance Discussion Paper No.. Yetman, J. (a). The Decoupling of Asia-Pacific?, in Yin-Wong Cheung, Vikas Kakkar, Guonan Ma (ed.) The Evolving Role of Asia in Global Finance (Frontiers of Economics and Globalization, Volume 9), Emerald Group Publishing Limited, -. Yetman, J. (b). Exporting recessions: International links and the business cycle. Economics Letters,,. Zaharia, M. (). Euro govt-greek contagion fear puts German yields at new low, Rueters article, May. (Online, Available at Last Accessed June )

17 Appendix A Sovereign credit rating Member state Lowest (Highest) rating since March 997 Rating (Jan ) Austria AAA (AAA) AAA Belgium AA- (AA+) AA Cyprus RD (AA) CCC Estonia A- (A+) A+ Finland AA+ (AAA) AAA France AA+ (AAA) AA+ Germany AAA (AAA) AAA Greece RD (A+) B- Ireland BBB+ (AAA) BBB+ Italy BBB+ (AA) BBB+ Latvia BBB- (A) BBB+ Luxembourg AAA (AAA) AAA Malta A (AA-) A The Netherlands AAA (AAA) AAA Portugal BBB- (AAA) BB+ Slovenia BBB+ (AA) BBB+ Slovakia BBB+ (A+) A+ Spain BBB (AAA) BBB Note: Entries based on Rating Agency Fitch s criteria. Source:

18 Appendix B 9 Confidence Intervals for Series -month -month -month -month LB UB LB UB LB UB LB UB UK-Can UK-US UK-euro UK-euro AAA Can-US Can-euro Can-euro AAA US-euro US-euro AAA Notes: 9 confidence intervals are presented for the pairwise comovements,, of each n-month yield,,, and, over the full sample, with LB and UB being the lower and upper bounds of the interval respectively. 7

Some Historical Examples of Yield Curves

3 months 6 months 1 year 2 years 5 years 10 years 30 years Some Historical Examples of Yield Curves Nominal interest rate, % 16 14 12 10 8 6 4 2 January 1981 June1999 December2009 0 Time to maturity This