The correlation between bond yields and stock prices in
crisis periods: Safe and risky countries
Mirthe Kuenen 10201858
Economie en Bedrijfskunde, Finance and Organization Thesis coordinator: R.J. Doettling MSc
2 February 2015
Abstract
This thesis aims to show the difference between ‘safe’ investment countries and riskier countries during crisis periods. In this thesis it becomes clear whether the correlation between stock prices and bond yields is different in crisis periods. Furthermore, the difference of this effect of crisis periods will be investigated for both high rated countries and low rated countries. The interaction coefficient between recession times and the riskiness of a country turns out to have a negative effect on the correlation between stock prices and bond yields, indicating that risky countries in recession experience a smaller effect of a flight-to-quality than safe countries in recession.
Table of contents
Page
1. Introduction 1
2. Literature 3
2.1 Results of previous research 3
2.2 Stock-bond relationship during crisis periods 4 2.3 Stock-bond relationship for high and low rated countries 6
3. Data 7 4. Method 9 5. Results 10 6. Conclusion 15 7. Bibliography 16 8. Appendix 17
1 Introduction
There has been research on the relationship between stocks and bonds, but not that much is applied to crisis periods. In this thesis the correlations between stock prices and bond yields of either high rated countries and low rated countries will be observed over the period from 2004-2014 in North America and the European Union. This period includes the global financial crisis and the Euro crisis. The aim of this paper is to show the difference between ‘safe’ investment countries and riskier countries during crisis periods. In the recent financial crisis there have been government bonds that before the crisis were seen as an investment grade, but were during the crisis downgraded to junk bonds. In this unstable environment investors take different decisions. For example they could switch to government bonds of safer countries. With these points kept in mind the following research question has been formulated: Does the correlation between stock prices and bond yields become larger or smaller during recession times and is this different for safe countries compared to high risk countries?
The stock markets also suffered from the economic downturn. To hedge for the risk of investing in stocks, investors could choose to invest more in government bonds instead of stocks. In some cases where governments are in a very bad financial position it might even be more attractive to invest in stocks though.
It is interesting to see whether more stable economies have a different bond yield – stock price interaction than less stable economies in crisis periods. Investigating the correlation between stock prices and bond yields is relevant for different reasons. First, correlation between different kinds of securities plays an important role in portfolio optimization because of reducing risk and/or increasing returns through allocating assets. Second, watching the correlation between stock prices and bond yields move over time gives interesting insights in the factors that are of importance for stock and bond valuation. Especially during crisis periods when markets are disturbed a better understanding of the stock price and bond yield movement is of great importance. During the recent crisis some countries were hit harder than other countries, driving bond yields generally more up for the riskier countries than for safe countries.
The different assets addressed are stocks and government bonds for both safe and risky countries. Stocks are generally riskier than government bonds. The bond yield reflects the
risk that a country might be unable to repay its debts. If this probability is large, the yield the government must offer on its bonds must be high to continue to attract investors. The
probability of a high rated country to repay its debts is generally substantial, so bond yields are low. In crisis periods all assets tend to become riskier, also government bonds. In the recent crisis some countries had to be saved by bailout packages of the International Monetary Fund and other European countries. This influences the bond yields of those
countries. Bonds of low rated countries during crisis periods are no safe investments resulting in high bond yields, for example Greece offered a bond yield of 48.6% in March 2012. The period concerned in this thesis starts in 2004 when almost all the countries observed experienced economic growth. In 2007 the United States entered into a severe crisis, in September 2008 Lehman Brothers fell and the country was in a recession for a year. This crisis was also noticeable in Europe, additionally the European debt crisis hit the continent causing many countries to come into recession (again).
This thesis studies stock index prices and 10 year bond yields and calculates the correlation between those two of all the countries concerned. Additionally, macroeconomic data on the countries is included in the dataset. Panel regressions will be run on this dataset with the dependent variable being the correlation between stock prices and bond yields. The aim is to confirm or reject the hypothesis, which is: The interaction term between recession times and the riskiness of a country is zero. From the empirical research follows that recession has a positive effect on the correlation. This means that when stock prices fall, bond prices rise, so bond yields also fall. This can be explained by a flight-to-quality where investors choose to invest in bonds, which causes bond yields to fall and thus stock prices need to fall too to attract investors. On the other hand, the interaction coefficient between recession times and the riskiness of a country has a negative effect on the correlation. This can be explained by the poor credit rating of the countries, which requires a higher bond yield to compensate for the additional risk of investing in that country, even when a flight-to-quality occurs. The H0 hypothesis can thus be rejected, since the interaction term between recession and the riskiness of a country is not zero.
This paper will start with elaborating on the existing literature on this topic. Second, the data will be described. Third, the method for the investigation is explained. After that, the results will be discussed and ultimately, the thesis will be concluded.
2 Literature
In this section the existing literature on this topic will be addressed. First the results of previous research will be described. Second, the reasons for assuming a different stock-bond relationship during crisis periods will be discussed and finally, the reasons for assuming a different stock-bond relationship for either high-risk countries or low-risk countries will be discussed. Based on these findings a hypothesis for this study was formulated.
2.1 Results of previous research
Intuitively a positive relationship between stock prices and bond yields can be expected. As stock prices rise this becomes a good investment, people shift from bonds to stocks, bond prices fall, so bond yields rise to attract new investors. Beltratti and Shiller (1992) say that it can also easily be explained that the stock price-bond yield relationship is negative, because an increase in expected long-term bond yields would seem to make long term bonds a more attractive investment, and so stock prices would have to fall to induce people to hold stocks (p. 1). So there are reasons to expect a positive correlation or a negative correlation. Both ways of reasoning discussed above include the already existing expectations. For example, a bond yield becomes low if there are already a lot of people investing in bonds. Under normal circumstances the correlation hence can either be positive or negative and fluctuations between positive and negative values are common.
In developing and lower rated countries as defined by credit rating agencies such as S&P, Moody’s and Fitch, bond yields are generally higher than in safer, developed countries, since lending money to these countries is riskier. This however does not need to imply that the correlation between stock prices and bond yields is smaller for less safe countries. Since the correlation depends on the movement of both stock index prices and bond yields over time, it can also remain constant. A country with high bond yields might constantly remain at that yield, with stock prices also staying equal, the correlation does not change. In general stocks tend to outperform bonds in times of economic expansion when stocks rise and bonds tend to outperform stocks in times of recession, because bonds are a lot less risky.
Ilmanen (2003) conducted a research regarding historical data on stock and bond returns, he concluded that this correlation has not been stable (p. 56). He also found that government bond yields reflect expectations of future short-term rates and the required bond risk
premium as compensation for bearing additional risk. This makes stocks riskier than bonds, so in less stable environments this is a reason to prefer bonds over stocks. This confirms the expectation that in crisis periods a flight-to-quality occurs where investors prefer bonds over stocks.
Bansal et al. (2014) concentrated on the correlation between bond returns and stock returns and concluded that the same economic forces that drive equity-risk changes, also in large part are the forces that drive the negative stock-bond return correlation they observed in their sample period. They found that equity risk either explains in part excessive bond returns and forward-looking term risk premia (p. 723). It is important to note that the stock-bond
correlation Bansal et al. address in their study is a different correlation than the stock price-bond yield correlation addressed in this study. The stock return is defined as: (!!!!!!!)!!"#
!!!!
and the bond return is defined as: !"!!!∗ (!!!!!!!"!!!!) ∗!!!!
(!!!!!!!!!!!"!!!) ∗!!!! where RI is the bond return,
A is the accrued interest to “normal” settlement date, CP compensates for the large drop in interest when bonds go ex-dividend and G is the coupon payment received from t-1. An increasing bond return includes an increasing bond price, which results in a decreasing bond yield. So a rising bond return actually means a falling bond yield. Still it is of interest for this thesis that equity risk influences bond yields and thus explains a connection between bond yields and the stock market. According to Ammer and Campbell (1993) excess bond returns can be forecast using the same variables that help to predict excess stock returns. A forecast of future excess returns has less effect on bond returns than on stock returns, because the returns in the bond market generally move slower than returns in the stock market (p. 32). This is another factor that helps to explain the relation between stock prices and bond yields.
2.2 Stock-bond relationship during crisis periods
Investors could shift their attention from the private sector to the public sector during crisis periods in search for safer investments. The bond-stock return correlation as Bansal et al. also use in their study might be negative in periods of this “flight to quality” say Christiansen and Ranaldo (2007, p. 445). A high stock return induces a high stock price and a high bond return induces a low bond yield. So a negative bond-stock return correlation means a falling stock price and a falling bond yield, which is a positive stock price-bond yield relationship or vice versa. This positive stock price-bond yield correlation during crisis periods is in line with the outcomes of this thesis. Ilmanen (2003) found that stock and bond discount rates move in
opposite directions due to differences in equity and bond risk premiums. Flight to quality episodes according to Ilmanen often raise required equity risk premiums (reduce stock prices) and reduce bond risk premiums (bond yields) (p.58). This causes a positive correlation
between stock prices and bond yields during flight to quality periods.
Compared to other asset classes like equities and corporate bonds, Barrios et al. (2009) state that the adjustment in government bond yield differentials between European countries during the period of intense financial market turbulences appears to have been proportionally stronger. They think this can be explained by a risk transfer effect from the private to the public sector due to the announcement of national financial rescue packages (p. 24). So even though in a flight-to-quality, investors prefer government bonds over stocks as a less risky investment, in times of financial rescue packages the bond yields are subject to stronger differentials between the European countries. In this case the bond yields of risky countries will go up to compensate for the risk of not being able to repay its debts, whereas for the safe countries there is no need to offer high bond yields to attract investors.
Cheng et al. (2011) researched the S&P 500 and different U.S. Treasury bond yields and found that the stock market leads bond yields. Cheng et al. assign this to the central bank policy that is reacting to the stock market and is following it. Additionally, they found that before the crisis in 2007 the short-term yields lead the long-term yields and that this
relationship inversed in the period after the crisis. In another study Connolly et al. found that in times of uncertainty on the stock markets, investors’ assessments of both stock risk and the relative attractiveness of stocks versus bonds are more frequently revised (2005, p. 189). This could cause the correlation between stock prices and bond yields to fluctuate more. When the stock market is in an uncertain period, the bond yield-stock price correlation is more likely to be positive in the future.
Bansal et al. (2014) suggest in their study that equity-risk movements are associated with a flight-to-quality to Treasury bonds. Also they state that bond-risk changes and 10-year
Treasury bond excess returns are negatively related (p. 701). Hartmann et al. (2004) looked at the phenomenon of contagion between stock and bond markets, where a crash in one market spills over to another market. They found out that this phenomenon cannot be commonly occurring in the G-5 countries (p. 323). So when the stock market crashes and a flight-to-quality occurs, it is reasonable to assume that the bond market will not crash due to the stock market crash, causing bond yields to stay low.
Christiansen and Ranaldo (2007) performed a research on the effect of macroeconomic announcements on bond-stock return correlation and they found out that in times of
expansion an announcement typically strengthens the stock-bond return comovements (so a negative correlation in this thesis). In recession times the markets react stronger to
announcements, but the influence on the correlation between bond and stock returns depends on the type of the announcement (p. 468).
Overall, in times of crisis a flight-to-quality effect causes investors to prefer bonds over stocks which results in the correlation between stock prices and bond yields to be positive. In case of government rescue packages investors might however prefer stocks over bonds, because the risk transfers from the private sector to the public sector. This is especially the case for riskier countries and it explains why there is a difference between risky and safe countries. Also will the relative attractiveness between either stocks or bonds more frequently be revised during crisis periods.
2.3 Stock-bond relationship for high and low rated countries
Allen et al. (2012) discuss that developing countries behave differently in times of crises than developed countries. “We show that in most emerging markets, which generally do not have well-developed stock markets, changes in the financial structures are of a smaller magnitude compared to countries with balanced structure of the financial system after a crisis. In contrast, we find that in developed countries, the changes are more significant and of longer duration.” (p. 2961). This assumption is interesting for this thesis, since the stock and bond prices will be studied before, during and after crisis periods. If there are changes in the financial structures, this may be reflected in the relationships between stocks and bonds. Furthermore, Barrios et al. (2009) found that high debt countries and, foremost, countries with large current account deficits are found to experience the highest bond yield increases as consequences of deteriorating public finances and increase in general risk aversion (p. 3). If this does not coincide with a just as high drop in stock prices, this might imply that the correlation between stock prices and bond yields becomes smaller or (more) negative during crisis periods for lower rated countries than for higher rated countries.
Hamori and Tamakoshi focused on the recent financial crisis in Europe and discussed that the significant unidirectional causality-in-mean from bank stock returns in Greece to Greek long-term bond yields arises only during the sovereign debt crisis period. Also they found that
Portugal, Italy and Spain bank stock returns were influenced by Greek sovereign bond yields (p. 641). Furthermore, they suggest that future studies should consider analyzing the
causalities among sovereign bond yields in different markets.
3 Data
This section describes the countries included in the dataset and the variables included in the regression.
The countries used for this research are European Union and North American countries, since these regions include many developed countries where data on stocks and bonds is available for the time span considered and several countries in these regions suffered heavily from the crisis periods. For some countries no data could be found and unfortunately they then had to be removed from the dataset. An overview of the used and the removed countries can be found in appendix 1. Especially bond yield data was sometimes difficult to find. For some countries only monthly or semi-annual data could be found which makes calculating the semi-annual correlations less precise or even impossible. For other countries the required data only started after 2008. To be able to see any effect of the crisis period the data has to be available from at least the beginning of 2007. After this selection 24 countries remained, 16 of which are high rated countries and 8 are low rated countries as of December 2014, this can be found in appendix 2.
For these countries daily market stock index prices and 10-year government bond yields are downloaded from Thomas Reuters DataStream. Half yearly correlations between stock prices and bond yields are calculated and included for each country.
Apart from that, the foreign currency rating which is the risk of sovereign default on foreign currency obligations for each country is included. These credit ratings are obtained from S&P, Moody’s and Fitch. All countries are observed over time and the country credit ratings are updated every half year. To be able to work with these data a numeric weight is given to the country credit ratings. These weights can be found in appendix 3, the highest possible weight is 21 and the lowest possible weight is 1. The distance between each of the possible ratings is uniform and I gave this distance a value of 1. There was no reason to assume the distances not to be uniform, so that is why this scale is chosen. The mean of the three credit
ratings for each country in each period of time is calculated and will be used as an explanatory variable in the regressions.
Apart from this rating variable in the regression an interaction coefficient between recession times and the riskiness of a country is included. Just multiplying the rating mean as described above with the recession times dummy gives suspicious results. Since the rating mean is a high value for high rated countries such as the U.S. and a low value for low rated countries such as Greece. This means that for the U.S. in recession, the value of the interaction coefficient is 21, whereas for Greece in recession the value is around 5. On the other hand when the U.S. or Greece are not in recession both values are 0. To get an interaction coefficient that is easier to interpret, a dummy will be constructed for risky countries. This value is 1 for risky countries and 0 for safe countries. The difference between high rated countries and low rated countries in this thesis will be drawn between the A-(A3) rating and the BBB+(Baa1) rating, whose values are 15 and 14 in this thesis as can be seen in appendix 2. Every average value lower than 15 is graded as a ‘Risky’ country. This division is chosen because all values from BBB+ and lower are less likely to meet their obligations, although BBB is still a higher rating than a junk bond. S&P define the BBB rating: “Adequate capacity to meet financial commitments, but more subject to adverse economic conditions”. To still remain two substantial groups in this thesis, the division is defined as between BBB+ and A-. This division between the two groups of countries is used for creating the riskiness dummy that is included in the regression. This riskiness dummy is used to generate an interaction coefficient between the countries that are rated as risky and that were in recession. Additionally, the real GDP in millions of euros and debt to GDP percentage for every country, every half year is retrieved from Thomas Reuters Datastream.
Finally, from the GDP data, the recession time dummy is constructed. A recession in this study is defined as two consecutive quarters of GDP decline and the end of the recession will be defined as the first quarter of GDP growth. According to Gaski (2012) this is safer than the NBER definition and insulated from suspicion because of its objectivity (p. 120). If a country was in a recession at the moment of observation it gets a 1, if it was not in recession it gets a 0.
4 Method
The existing literature elaborates on the effect of crisis on the stock-bond relationship or the stock-bond relationship in low and high rated countries. These effects have however not yet been combined. In this research it will be tested if the interaction term between a recession time period and the riskiness of a country is a significant explanatory variable of the correlation between stock prices and bond yields. The hypothesis that will be tested is: H0: The interaction term between recession times and the riskiness of a country is zero (Recession*Risky = 0).
H1: The interaction term between recession times and the riskiness of a country is not zero (Recession*Risky ≠ 0).
To start I will plot a graph of the average half year correlations between stock index prices and 10-year government bond yields of both high rated countries and low rated countries. This shows whether there are trends that can immediately be seen from the graph. The
countries are defined either high rated or low rated based on the average credit rating of S&P, Moody’s and Fitch in December 2014 (appendix 2). The graph is included in the results section. After this graphical illustration, a panel regression will be executed on the data on bond yields and stock prices of the selected countries. The following panel regression is used:
!!"#$%&',!"#$ = !!+ !!∗ !!"#"$$%&'+!!∗ !!"#$%!,!+ !!∗ !"#$%&!,!+ !!
∗ !"#!!"#!,!+!!∗ !"#!,!+ !!∗ !!"#"$$%&'∗ !!"#$%!,! + !!"#!"#$
The dependent variable is the correlation between stock prices and bond yields and the explanatory variables are explained below.
D_Recession: dummy variable which is 1 for recession time periods and 0 if the country is not in recession.
D_Risky: dummy variable which is 1 if the country was defined as a risky country at that moment in time and 0 if not (Every value lower than 15 in appendix 3 is defined as risky). Spread: the bond yield spread with German government bonds.
Debt_GDP: government debt as a percentage of real GDP. GDP: real GDP (yearly observation) in millions of euros.
Ratingmean: Average of the Fitch, Moody’s and S&P foreign currency credit ratings at the given period in time. Appendix 3 shows the weights given to the different ratings.
The variable Ratingmean is not included in the final panel regression, but was used before to see how including that variable works out. As discussed in the data section, an interaction coefficient between Ratingmean and the recession times dummy gives results that are very difficult to interpret and that are insignificant. For that reason in this thesis I choose not to include the Ratingmean variable, but instead use the riskiness dummy.
The final objective of the panel regression is to find the effect of the interaction coefficient between the dummies for recession times and risky on the correlation between stock prices and bond yields.
5 Results
First, I will discuss the graph that shows the correlations of high rated countries and low rated countries over time. It immediately draws the attention that the correlations are not
consistent, it fluctuates a lot between positive and negative values. The lower rated countries seem to lead the higher rated countries in the beginning of the sample period. Around 2008 this trend fades away and the high and low rated countries seem to move apart. This could confirm the hypothesis that the correlations of high rated countries are not the same as the correlations of low rated countries during crisis periods. Also the difference between the correlations of high rated countries and low rated countries is included. This shows that until halfway 2014 the correlations of the high rated countries are higher than those of the low rated countries. It is important to note that the recession periods haven’t been defined yet for all the separate countries, so at this moment it is not possible to tell anything about the effect of recession.
The regression that is used in this data research aims on investigating which variables influence the correlation between stock prices and bond yields. When the correlation is positive, this means stock prices and bond yields rise or fall together. In other words, stock prices increase, for example because of a higher demand for stocks and bond yields also increase which means that the government needs to offer a higher yield to attract investors. This could be the case if the country is in a risky position or when the demand for bonds is low. A positive explanatory variable thus stimulates the positive correlation between stock prices and bond yields and a negative explanatory variable stimulates a negative correlation. To see whether adding more variables increase the explanatory power of the regression a few different regressions have been performed. The variables used are explained in the method section.
Table 1: Regression outputs
Regression 1 2 3 4 5 6 7 Recession .1376⁺ (.0695) 0.2574** (.0570) 0.1561* (0.0689) 0.2646** (0.0594) 0.2614** (0.0578) 0.2709** (0.0578) 0.2630** (0.0598) Rating mean 0.0077 -‐1 -‐0,8 -‐0,6 -‐0,4 -‐0,2 0 0,2 0,4 0,6 0,8 1 30/06/2004 31/12/2004 30/06/2005 30/12/2005 30/06/2006 29/12/2006 29/06/2007 31/12/2007 30/06/2008 31/12/2008 30/06/2009 31/12/2009 30/06/2010 31/12/2010 30/06/2011 30/12/2011 29/06/2012 31/12/2012 28/06/2013 31/12/2013 30/06/2014 31/12/2014
High rated countries Low rated countries High-‐low rated
(0.0266) Risky -0.2705** (.0628) -0.0671 (0.1562) -0.2094* (0.0852) -0.0586 (0.1470) SpreadDE -0.0179 (0.0118) -0.0033 (0.0122) -0.0048 (0.0113) Debt_GDP -0.0047** (0.0016) -0.0030 (0.0031) -0.0032 (0.0023) -0.0031 (0.0024) GDP 2.45e-6* (1.16e-6) 2.14e-6⁺ (1.09e-6) 2.15e-6⁺ (1.06e-6) 1.74e-6⁺ (9.10e-7) 2.15e-6⁺ (1.06e-6) Recession*Risky -0.4146** (0.1302) -0.4139** (0.1308) -0.4122** (0.1412) -0.3922* (0.1283) -0.4015* (0.1540) R² (within) 0.0105 0.0597 0.0507 0.0727 0.0726 0.0683 0.0728 Constant included, robust standard errors in parenthesis.
⁺10% significance, *5% significance and **1% significance
The influence of a recession, the riskiness of a country and the interaction coefficient between these two coefficients on the correlation between stock prices and bond yields are the key factors in this study. The dummy variable Recession remains positive and significant throughout all the regressions. This is in line with the existing literature where they talk about a flight to quality during crisis periods. In a flight-to-quality the required equity risk
premiums rise, which makes stock prices fall. Since investors prefer to invest in bonds during crisis periods the bond yields also fall, because governments don’t need to offer a high bond yield to attract investors. So bond yields fall and stock prices also fall, resulting in a positive correlation. In regression 1 Recession is the only explanatory variable and the R-squared is very low. Adding more explanatory variables makes the recession dummy more significant and the R-squared higher. This makes that regression 1 is not the best predictor of the model, because the correlation between stock prices and bond yields depends on more factors than just the occurrence of a recession.
In regression 2 the main variables of interest are included: Recession, Risky and
Recession*Risky. All three are significant at a 1% level. Recession has a positive influence on the correlation between stock prices and bond yields like I already discussed. This first outcome would imply that a country in recession generally has a higher correlation than a
country that is not in recession, but a risky country that is in recession has a negative effect on the correlation. This can be interpreted in the following way. For example a risky country that is in recession experiences a positive effect on correlation for being in recession, which can be explained by a flight-to-quality. On the other hand, since it is a risky country it also experiences the negative effect from the interaction coefficient for risky countries in
recession and the variable for being a risky country, because a risky country is less likely to be able to repay its debt. The government of a risky country thus needs to offer higher yields to attract investors. This makes that the stock prices fall, but the bond yields rise, causing a negative correlation.
In the following regressions more variables are included to see whether the correlation between stock prices and bond yields also depends on other factors and if the three variables discussed before remain significant.
In regression 4 the Ratingmean is used instead of the riskiness dummy, to see its effect on the correlation. The results of regression 4 are only slightly different from the results in the final regression 7. Since I chose to use the riskiness dummy in the interaction coefficient as explained in the data and the method sections, it is more in line with the other coefficients to use the riskiness dummy instead of Ratingmean. The mean of all the credit ratings is thus only used to distinguish high rated countries from low rated countries.
The riskiness dummy does not remain significant when other variables are added. When other macroeconomic factors are included the riskiness dummy is not significant anymore. A cause for this could be that the GDP and the debt to GDP percentage capture part of the country’s health which is also included in the riskiness dummy.
The interaction term between the dummy variables recession and risky remains negative and significant through all the regressions. This is an important factor in this thesis, since this explains the difference in difference of a safe country in recession and a risky country in recession. Like explained before, risky countries need to offer a higher bond yield to attract investors, especially during recession periods when the probability of repaying its debts becomes smaller. This makes the correlation between stock prices and bond yields negative or less positive, if the stock prices also decrease.
To control for other explanatory variables macroeconomic factors such as real GDP and the debt to GDP percentage are included. GDP is an indicator of the performance of a country.
Since GDP is a large number in millions of euros, the coefficient is very small, but significant in all the regressions. The coefficient on GDP is positive, which implies that a high real GDP causes a higher correlation between stock prices and bond yields. This implies that in larger countries, as those are the ones with high GDP’s, the correlation between stock prices and bond yields is more likely to be positive. It could be possible to ascribe this to the high probability that a big country repays its debt, which is generally larger than for a smaller country. The possibility that a very large country in terms of real GDP goes bankrupt for example is much smaller than the possibility that a small country goes bankrupt.
The debt to GDP percentage coefficient is generally not significant. Only in the regression where the riskiness and the interaction coefficient are excluded it is significant. So when there is already accounted for risk, the debt to GDP ratio is not significant anymore. The bond yield spread with Germany has been taken up in the regression, but turns out not to be
significant.
These results are straightforward related to graph 1. In the recession times period, which roughly started in 2007 until around 2012 the correlations of the high rated countries rose, whereas the correlations of the low rated countries fell. This is in line with the results from the table. Recession generally causes the correlation to increase, but for risky countries the correlation decreases during recession.
Overall, the main conclusions that can be drawn from the results are first, that recession has a significant positive effect on the correlation between stock prices and bond yields. This effect has to do with the flight-to-quality in times of recession. Second, GDP has a positive effect on the correlation, because the bond yields of a country with a high GDP are not expected to rise dramatically, as the possibility of repaying its debt is considerably big for larger
countries. And finally, the interaction coefficient between riskiness and recession has a significant negative effect on the correlation. This is because risky countries, especially when they are in recession, are less likely to repay their debts, thus bond yield need to be high to attract investors who are willing to take the risk of investing in the risky country.
6 Conclusion
This study analyzed if the correlation between stock prices and bond yields in recession time periods becomes higher or lower and to see if this correlation was different for high rated countries or low rated countries. Taken all results together and relating this to the existing literature gives some interesting outcomes.
From the literature it was expected that a recession period would positively influence the correlation between stock prices and bond yields because of a flight-to-quality. That is, the required equity risk premiums would rise, which makes stock prices fall and bond yields would also fall, because investors prefer to invest in bonds. This is in line with the results found in this thesis.
The interaction coefficient between the recession times dummy and the riskiness dummy is negative. This makes that the H0 hypothesis, which stated that the interaction coefficient would be zero, can be rejected. So although a recession causes a positive bond yield – stock price relationship, possibly because of the flight-to-quality, this effect is less positive or even negative for risky countries that are in recession. A possible explanation is that the bond yields do not decrease as much as they do for high rated countries. If a country has a poor credit rating, they still have to offer high bond yields to attract investors. This makes the flight to quality effect weaker, because in this case the stock prices still fall, but the bond yields do not decrease that much.
To answer the research question it can be concluded that the correlation between stock prices and bond yields becomes larger during recession times periods because of a flight-to-quality, but this positive effect is smaller or even becomes negative for high risk countries in
recession, because they still need to offer high bond yields.
To complete this study, government rescue packages can be included in the regressions, to see if this has a significant effect on the correlation apart from just the recession dummy and the interaction coefficient.
A possible additional study could be to look at other emerging countries outside of Europe and North-America. Or the study could be limited to one country for which all the
developments over the period investigated in this thesis needs to be watched closely to get a more specific understanding of what makes the correlation change.
7 Bibliography
Allen, F., Gu, X., & Kowalewski. (2012). Financial crisis, structure and reform. Journal of
Banking and Finance, 2960-2973.
Baele, L., Bekaert, G., & Inghelbrecht, K. (2009). The determinants of stock and bond return
comovements. Cambridge, MA: National Bureau of Economic Research.
Bansal, N., Connolly, R., & Stivers, C. (2014, June). The Stock-Bond Return relation, the term structure's slope and asset-class risk dynamics. Journal of financial and
quantitative analysis, 43(3), 699-724.
Barrios, S., Iversen, P., Lewandowska, M., & Setzer, R. (2009). Determinants of intra-euro
area government bond sreads during the financial crisis. European Commission,
Economic and Financial affairs, Brussels.
Beber, A., Brandt, M., & Kavajecz, K. (2006). Flight-to-quality or flight-to-liquidity?
Evidence from the Euro-area bond market. Cambridge, MA: National Bureau of
Economic Research.
Beltratti, A., & Shiller, R. (1992, October). Stock prices and bond yields: Can their
comovements be explained in terms of present value models? Journal of Monetary
Economics, 30(1), 25-46.
Cheng, S., Guo, K., Sornette, D., & Zhou, W. (2011, August 10). The US Stock Market Leads the Federal Funds Rate and Treasury Bond Yields. Plos One, 6(8). Christiansen, C., & Ranaldo, A. (2007, March 15). Realized bond-stock correlation:
macroeconomic announcement effects. Journal of futures markets, 27(5), 439-469. Connolly, R., Stivers, C., & Sun, L. (2005, March). Stock market uncertainty and the
stock-bond return relation. Journal of financial and quantitative analysis, 40(1), 161-194. ESMA. (2014). Credit Rating Agencies’ 2014 market share calculations for the purposes of
Article 8d of the CRA Regulation. ESMA.
Fama, E., & French, K. (1993). Common risk factors in the returns on stocks and bonds.
Journal of Financial Economics, 33(1), 3-56.
Gaski, J.F. (2012). On the competing definitions of recession. Society, 49(2), 118-121. Hamori, S. & Tamakoshi, G. (2012, September 9). Causality-in-variance and
causality-in-mean between the Greek sovereign bond yields and Southern European banking sector equity returns. Journal of Economics and Finance, 38, 627-642.
Hartmann, P., Straetmans, S., & Vries de, C. (2004, February). Asset market linkages in crisis periods. Review of Economics and Statistics, 86(1), 313-326.
Hotchkiss, E., & Ronen, T. (2002). The Informational Efficiency of the Corporate Bond Market: An intraday analysis. Review of Financial Studies, 15(5), 1325-1354. Ilmanen, A. (2003). Stock-bond correlations. Journal of fixed income, 55-66.
Weaver, B., & Wuensch, K. (2013, Vol.45(3)). SPSS and SAS programs for comparing Pearson correlations. Behavior Research Methods, 880-895.
8 Appendix
Appendix 1: Countries included in the thesis
Included Excluded Reason for exclusion
Austria Estonia No data
Belgium Latvia No daily data
Czech Republic Lithuania No daily data
Denmark Slovakia No daily data
Finland Cyprus Data starts 2008
France Romania Data starts 2008
Germany Mexico Data starts 2010
Ireland Luxembourg Malta Netherlands Poland Sweden United Kingdom United States Canada Bulgaria Croatia Greece Hungary Italy Portugal
Slovenia Spain
Appendix 2: Country credit ratings December 2014
S&P Moody’s Fitch
EU
Austria AA+ Aaa AAA
Belgium AA Aa3 AA
Bulgaria BB+ Baa2 BBB-
Croatia BB Baa3 BB
Cyprus B+ B3 B-
Czech Republic AA A1 A+
Denmark AAA Aaa AAA
Estonia AA- A1 A+
Finland AA+ Aaa AAA
France AA Aa1 AA
Germany AAA Aaa AAA
Greece B Caa1 B Hungary BB Ba1 BB+ Ireland A Baa1 A- Italy BBB- Baa2 BBB+ Latvia A- Baa1 A- Lithuania A- Baa1 A-
Luxembourg AAA Aaa AAA
Malta BBB+ A3 A+
Netherlands AA+ Aaa AAA
Poland A- A2 A-
Portugal BB Ba2 BB+
Romania BBB- Baa3 BBB-
Slovakia A A1 A+
Spain BBB Baa2 BBB+
Sweden AAA Aaa AAA
United Kingdom AAA Aa1 AA+
North-America(Nafta)
United States AA+ Aaa AAA
Canada AAA Aaa AAA
Mexico BBB+ A3 BBB+
Appendix 3: Conversion table
S&P and Fitch Moody’s Numeric value
AAA Aaa 21 AA+ Aa1 20 AA Aa2 19 AA- Aa3 18 A+ A1 17 A A2 16 A- A3 15 BBB+ Baa1 14 BBB Baa2 13 BBB- Baa3 12 BB+ Ba1 11 BB Ba2 10 BB- Ba3 9 B+ B1 8 B B2 7 B- B3 6 CCC+ Caa1 5 CCC Caa2 4 CCC- Caa3 3 CC Ca 2
C C 1
