Investigation into the commonality in liquidity and
sovereign debt ratings relationship
August 15, 2014
MSc Thesis in Business Economics
Specialization: Finance
Author: Kanykei Dordoeva Supervisor: Patrick Tuijp
Abstract
Commonality in liquidity is a well-documented phenomenon when individual stock liquidity is partially driven by market liquidity. The driving forces of such occurrence are yet to be understood. This study examines the relationship between sovereign debt ratings and commonality in liquidity. By using the daily data of 611 unique stocks from Greece, Italy, Portugal and Spain, I estimate panel VAR model. To investigate direction of such relationship, Granger causality test is employed. I find that there is a significant one-way relationship between these two variables, specifically, sovereign debt rating Granger-causes commonality in liquidity. This research contributes to the sovereign debt rating literature by finding a new evidence of a fortifying element of the rating on the market, and to the commonality in liquidity literature by finding a new factor that affects commonality in liquidity.
Table of Contents
1. INTRODUCTION ... 2 2. LITERATURE REVIEW ... 4 4. DATA ... 13 5. METHODOLOGY ... 15 6. DESCRIPTIVE STATISTICS ... 19 7. EMPIRICAL RESULTS ... 21 8. ROBUSTNESS CHECKS ... 23 9. CONCLUSION ... 24 10. REFERENCES ... 26 11. APPENDICES ... 281. Introduction
Traditional asset pricing models are based on the idea that the only systematic fluctuation that stocks face is market risk. Thus, the key to success is diversification of the portfolio based on market returns. However, recent studies have shown that liquidity is important for asset pricing, and it can create a source of systematic risk (Lee, 2011; Acharya and Pedersen, 2005). In times of substantial market downturns, or high volatility, financial intermediaries decrease the supply of liquidity, which consequently leads to dry-up of market liquidity and increase in commonality in liquidity (Brunnermeier and Pedersen, 2009). Therefore, understanding how liquidity risk is priced is an important concern for investors to make better investment decisions in terms of timing, as well as for regulators and policy makers to prevent liquidity shocks.
Commonality in liquidity, the phenomenon where individual stock liquidity is partially explained by market liquidity, is relatively new, unexplored area of research. First manifested by Chordia et al. (2000) it is defined as co-movement of changes in individual and changes in market liquidity. The concept of commonality in liquidity is based on the notion that individual liquidity has common underlying determinants across securities. Subsequently, liquidity can create a source of systematic risk.
There is evidence that commonality in liquidity varies over time. As it is evidenced by Rösch and Kaserer (2013), it increases during times of market uncertainty and peaks during the financial crisis. However, the drivers of commonality in liquidity have not yet been well identified. Karolyi et al. (2012) attempt to investigate what drives commonality in liquidity. They find that demand determinants, such as investor sentiment, correlated trading activity of international and institutional trading, affect commonality stronger than supply determinants, which is funding liquidity.
My research aims to shed light on what drives the variation of commonality in liquidity over time. Specifically, I want to investigate whether the sovereign debt rating, which is known as an approximation for country risk, affects commonality in liquidity. The sovereign debt rating has been known not only as an indicator of credit risk, but also as a fortifier of economic boom-bust cycle (Reinhart, 2002). Previous research shows that in times of market uncertainty the relation between credit risk and uncertainty risk intensifies, namely, credit risk affects liquidity risk more strongly. (Rösch and Kaserer, 2013). Intuitively, in times of higher market volatility investors are more risk averse, which leads to the flight-to-quality effect in the sense of higher required risk premium. Therefore, I want to analyze whether the increased market uncertainty due to downgrade of the sovereign debt rating affects commonality in liquidity.
Research question: Do changes in sovereign ratings affect commonality in liquidity?
Analyzing commonality in liquidity is important for several reasons. First of all, understanding how liquidity risk is priced is an important concern for investors to make better investment decisions in terms of timing of transactions. Secondly, commonality in liquidity poses a question to diversification strategies that are based on stock returns. If stock returns and market liquidity are correlated, the common liquidity effects source could represent non-diversifiable risk factor. Thirdly, commonality is important for policy makers and regulators. Commonality in liquidity shocks may affect financial market as a whole, as shocks to liquidity can have a market-wide effect. The first hypothesis is thus following:
Hypothesis 1. Commonality in liquidity exists and varies over time.
Concluding from the prior research mentioned in the literature review, sovereign debt ratings do not only provide a benchmark of a country risk, but also affect the economic environment in number of ways. Firstly, changes in sovereign debt ratings may act as a self-fulfilling prophecy and exacerbate country’s economic cycle (Schumacher, 2014). Secondly, re-grading of countries may influence investors’ sentiment, as it is shown in the dynamics of the level of private investment in re-rated country (Chen et al., 2013). And lastly, sovereign debt rating news spill over on countries abroad (Fereira and Gama, 2007). What can be argued is that a downgrade in country’s sovereign debt rating will give a negative signal to the market, and as a result of uncertainty and asymmetric information, it may impose flight-to-quality effect and thus will give a rise in increased commonality in liquidity. This reasoning leads to the next hypothesis:
Hypothesis 2. The change of sovereign debt rating affects the liquidity commonality within a country.
The aspect in which my study contributes to the existing literature is that I will measure the effect of sovereign ratings on commonality in liquidity To the best of my knowledge, there is no research conducted that examines this relationship.
I analyze stocks listed on the major stocks exchanges of Greece, Italy, Portugal and Spain over a period of January 1995 – December 2013. The sample consists of daily stock-related time series data, which comprises a panel dataset with 4 countries in it. The sovereign debt long-term ratings for these countries are taken from Moody’s agency. Since these four countries have been affected by the recent financial crisis, Moody’s re-graded the sovereign debt several times. I analyze Panel Vector Autoregressive model with fixed effects using monthly commonality in liquidity measures for a
given country, computed following the methodology of Karolyi et al. (2012), and linearly-transformed sovereign debt ratings. Subsequently, I test the coefficients of the model for the Granger causality. I find that sovereign debt rating Granger-causes commonality in liquidity, while commonality in liquidity does not Granger-causes the changes in sovereign ratings.
Section 2 provides some background on commonality in liquidity and sovereign debt ratings. Section 3 describes the objectives of the research and main hypotheses. Section 4 specifies data. Section 5 outlines proposed methodology and section 7 discusses the results. Section 9 concludes.
2. Literature Review
This literature review provides some background on the concepts of commonality in liquidity and sovereign debt ratings. The first part of the literature review provides the introduction to the concept of commonality in liquidity and discusses the main theories and empirical findings related to it. The second part briefly discusses the role of the credit rating agencies, the notion of sovereign debt ratings and the recent research on the impact of sovereign debt ratings.
2.1. What is liquidity?
According to frictionless economic models, all assets can be traded freely and for free. In reality, securities trading always incur costs. Costs can be explicit, i.e. directly related to the transaction, such as brokerage fees or taxes, and implicit. Implicit costs include bid-ask spread1, adverse selection costs, and opportunity costs that arise from technology or management inefficiencies, or government regulations. Thus, the general definition of a liquid asset is the asset that can be traded easily, i.e. without significant barriers on the market and with minimal costs.
Market liquidity it is the aggregate liquidity of all the securities. In real world, financial markets are perceived as liquid if trading costs are minimized and trading volumes are high. Kyle (1985) develops the definition of liquid market, decomposing it into three market properties: (i) “tightness” – the turnover of a position over a short period of time costs little; (ii) “depth” – the large-sized orders have small impact on trading costs; (iii) “resiliency” - prices are corrected quickly after a random shock.
2.2. Determinants of liquidity
Individual stock liquidity is generally attributed to explicit trading costs, such as brokerage fee or government taxes. Implicit liquidity costs are less straightforward, and thus have been heavily
studied in the financial market microstructure framework. Among others, these costs include opportunity costs, imperfect technology, management and adverse selection problem. Adverse selection problem occurs when buyers and sellers have asymmetric information. In such circumstances, “losing” stocks are likely to be selected and unexpected losses arise. Consequently, the information that traders own or do not own may impact the outcome of the trade (Kyle, 1985; Glosten and Milgrom; 1985).
Glosten and Milgrom (1985) develop the market-making model of sequential singe asset trade and two types of investors: the rational, informed insider and uninformed noise trader. The expected payoff from the single asset is either 1 or 0; and only the insider knows the outcome. If expected payoff is 0, he will sell the asset, if it is 1 then he will buy it. Thus the direction in which the trade will go depends on informed trader. After each period of trade market maker revises bid-ask quotes. Market is then semi-strong efficient as quotes reflect past and current information. It can be explained by the fact that ask quote, the price at which a counterparty can buy the asset, can incorporate information about the direction of the trade. The main result of the model is that bid-ask spread equals to the fraction of informed traders; it is merely the information effect. Subsequently, as the proportion of informed traders rises, spread widens.
Another model that aims to explain the impact of information on asset pricing is the model of Kyle (1985). This is a model of sequential trading with three participants: a risk neutral insider and, a uniformed noise trader, and a market maker who is not in competition with other market makers. Only the insider possesses the information on the value of the asset, and he takes the advantage of it. Market maker cannot distinguish the insider and noise trader but can guess if the overall direction of the trade is on the buy side (undervalued asset) or the sell side (overvalued asset). This model demonstrates that asset prices are positively dependent on order flow, i.e. the greater is the excess of buy orders over sell orders, the greater is the price. The degree of that effect increases with the informational advantage of insiders.
There are a number of studies that demonstrate the link between the stock returns and liquidity. Amihud and Mendelson (1986) study the effect of bid-ask spread on respective stock returns. Bid-ask spread is the cost of immediate execution and is perceived as a proxy for illiquidity. The study shows that the function of stock returns - both market returns and net of spread returns to shareholders - upturns with bid-ask spread. In addition to that, a “clientele effect” is observed, where stocks with higher spreads have longer holding periods and thus are their returns are less sensitive to spread. This results in the concave slope of spread-return function. As the study suggests, firms with
higher-spread securities will have higher required rate of return, and thus are incentivized to increase liquidity of these securities. As a result, liquidity-increasing actions will lead to higher value of the firm.
Another factor that can impact liquidity is so called inventory risk, or demand pressure. Inventory risk arises when there is a risk of not selling shares when the price goes down. In other words, when there are not enough volunteers to buy shares and the dealer is forced to sell the shares for a price less than its fundamental value in order not to lose more money. Stoll (1978a) develops a continuous market-making model, where a single risk-averse dealer is exposed to risk related with holding securities. This inventory risk fluctuates with the market and affects bid and ask quotes negatively, along with inventory level and risk aversion of the dealer.
2.3. Measures of liquidity
Liquidity is ambiguous concept, and it cannot be defined in single dimension. Aitken and Winn (1997) summarize 68 existent liquidity measures. This suggests that there is no consensus on single definition among researchers. They divide liquidity measures into two categories: trade-based measures and order-based measures. Measures based on trade include turnover value, volume and ratio. The advantage of these measures is that they are readily available for investors; however since they are ex-post, they reflect what had been in the past and this does not necessarily reflect the future. Measures based on order-book analysis involve bid and ask quotes. These measures provide more insight on the cost of immediate trading of an asset2. In case of large trades, order-based measures do not express the implicit costs of liquidity, such as opportunity costs of other market trades and the market impact. For instance, if market-maker wants to purchase a large amount of stocks but faces the deficit of stocks on the trading floor, he needs to raise his bid offer in order to attract more sell offers. Therefore, ex-ante order-based measures tend to deliver underestimated values of liquidity costs.
2.4. Commonality in liquidity
The literature that analyzes the individual stock liquidity and its importance for asset pricing is vast. But what if individual features of stocks such as bid-ask spread or trading volume have common underlying drivers across the market? Recent literature addresses this problem to identify the co-movements of individual assets’ liquidity.
2 For example, calculating relative spread – bid-ask spread as a fraction of security price, can compare securities.
Commonality in liquidity, the occurrence when individual stock liquidity is partially explained by market liquidity, is relatively new, yet well-documented phenomenon in the scientific literature (among others, Chordia et al. (2000), Huberman and Halka (2001), Hasbrouck and Seppi (2002)). This phenomenon was first recognized by Chordia, Roll, and Subrahanyam (2000). Chordia et al. (2000) defines it as “co-movement between changes in individual stock liquidity and changes in market liquidity”. By analyzing bid-ask spreads of roughly one thousand U.S. stocks, and controlling for prevalent liquidity elements such as volume, volatility and price, Chordia, Roll, and Subrahanyam (2000) find a common factor of changes in individual liquidity and changes in market liquidity.
Hasbrouck and Seppi (2002) take a different approach by examining common factors of order flows in returns and liquidity. They aggregate 15-minutes interval trades of 30 stocks listed on Dow index, and find significant common factors in order flows, and relatively small commonality in liquidity proxies – market depth and bid-ask spreads. In contrast to CSR approach, they analyze the effect of levels of market liquidity on individual rather then innovations of liquidity.
2.5. Sources of commonality on liquidity
Considering the facts stated above, one might ask - what drives commonality in liquidity? Until now, there is no straightforward answer to that question. Liquidity is a complex concept with many determinants, and which factors may affect the common factor of market and individual liquidity has still been discovered. This problem was posed in the recent study of Karolyi et al. (2012). They attempt to determine what forces drive commonality in liquidity, and why it fluctuates over time. By analyzing 40 developed and emerging countries, they showed that liquidity commonality varies across countries and varies over time, and that it is driven strongly by demand determinants than supply determinants. They find that “commonality in liquidity is greater in countries with more market volatility, large presence of international investors and more correlated trading activity”. As an additional test, they examine how U.S. default spread affects liquidity commonality, which results in negative correlation.
2.5.1. Supply determinants
Trades cannot be executed without available capital. That is, traders are dependent on the capital accessibility, and the ease with which they can get financing is called funding liquidity. Brunnermeier and Pedersen (2009) propose a model that links market liquidity with funding liquidity. According to the model, tight funding liquidity and declining market liquidity fortify one another, and
that occurrence may lead to phenomena called liquidity dry-up and flight-to-quality. As a result of decreased market liquidity, commonality in liquidity increases. Rösch and Kaserer (2013) explore the dynamics and the drivers of liquidity commonality during the financial crisis, using a data from Deutsch Boers. They show that “commonality in liquidity increases during economic downturns, peaks at major crises and becomes weaker if analyzed from the limit order book perspective”. They demonstrate the consistence between theoretical concept of spiral between funding and market liquidity (Brunnermeier and Pedersen, 2009) and find that “in times of tight funding liquidity commonality in liquidity increases which leads to market liquidity dry-ups”, and that there is a positive correlation between liquidity and credit risk.
2.5.2. Demand determinants
Prior research suggests commonality in liquidity is driven by demand-side factors, namely potential determinants of institutional and individual investors’ behavior. From the one hand, it is argued that commonality in trading behavior of institutional and individual investors is a potential force of growth in liquidity commonality (Kamara, Lou, and Sadka, 2008; Koch, Ruenzi, and Starcks, 2009). From the other hand investors’ sentiment and feeble inducements to trade individual securities in another explanation for analyzed phenomenon (Morck, Yeung, and You, 2000). Karolyi et al. (2012) provide evidence of strong explanatory power of demand-side factors, compared with supply-side factors.
2.6. Implications of commonality in liquidity
2.6.1. Asset pricing
According to the traditional asset pricing theory, expected stock returns are linked to how individual stock is sensitive to the undiversifiable systematic risk, or market risk.3 Nevertheless, it is seems plausible that liquidity plays an important role in asset pricing. Investors require a higher rate of return from stocks that have lower aggregate liquidity, since liquidation of such asset is then more costly. Therefore, if commonality in liquidity exists, then it may create the source of systematic, undiversifiable risk.
It was evidenced that market liquidity impacts the pricing of securities. Among studies that tested whether liquidity does have an impact on asset pricing was the study of Pastor and Sadka (2003). This study analyzes whether the market liquidity have impact on asset pricing, by analyzing
cross-sectional differences of expected return to the variations of market liquidity4. The study concludes that after controlling for market returns, and for company-specific size, value and momentum characteristics, stocks that are exposed to more sensitivity to market liquidity have significantly higher expected returns.
If market liquidity plays a role in asset pricing, can commonality in liquidity be an important part of it? Acharya and Pedersen (2005) develop an asset pricing model in which expected stock returns are not only dependent on market systematic risk, but also on market and individual liquidity. The standard CAPM beta, correlation between individual and market risk is transformed to liquidity adjusted beta, which can be decomposed into three relationships: co- movement of single stock and market liquidity (commonality in liquidity), co-movement of stock returns and market liquidity, and co-movement of market returns and stock illiquidity (Acharya and Pedersen, 2005). They find that the liquidity-adjusted CAPM better explains the expected rate of return than the standard CAPM. However the impact of commonality in liquidity amounted for only 0.8% of premium. This was also tested and evidenced by Lee (2010). Controlling for risk in financial markets of 50 countries, he finds that systematic liquidity risk is priced impartially of market risk, and furthermore, the U.S. financial market is a driver of global liquidity risk. The study concludes that how liquidity risk is priced is dependent on geographical, economic and political characteristics of a country. He finds that in emerging markets, the local commonality beta is priced at the 4.8% of premium at the 1% significance level, however it is not significantly priced in developed markets and global panel.
In overall, this liquidity-adjusted CAPM suggests that systematic liquidity risk is priced, because investors during market declines would pay more for the liquid stock that would allow that to cash-out more easily. Furthermore, if securities are illiquid in absolute terms have high market liquidity risk, then “flight to liquidity” effect may arise. Namely, such stocks tend to have high values of all three liquidity betas: commonality in liquidity, co-movement of stock returns with market liquidity, and co-movement of stock liquidity with market returns. This effect is also investigated by Brunnermeier and Pedersen (2009) and will be discussed in the section below.
2.7. Flight to quality, flight to liquidity
Flight to quality and flight to liquidity effects are related to the phenomenon when investors move their capital for less uncertain assets. Specifically, flight to quality effect arise when, during market uncertainty or market downturns, investors move away their capital by selling of what they perceive riskier assets and buying safer assets, such as government bonds or gold. Flight to quality is
closely related to the occurrence when investors relocate their capital from what they recognize as less liquid assets towards more liquid assets.
Brunnermeier and Pedersen (2009) develop a model which demonstrates that how funding liquidity, i.e. the ease with which financing can be obtained from trading, relates to the market liquidity. The ability to trade depends on the capital availability, and vice versa, the aggregate market liquidity impact funding liquidity. This mechanism was explained on the basis of margin borrowing; if margin requirements increase during market downturns, then they induce the process called “liquidity spiral” which can lead to liquidity dry-up and flight-to-liquidity. The important implication of this model is that liquidity has common underlying drivers that lead to market-wide consequences, which can be empirically tested.
Another comprehensive study of Vayanos (2004) provides a theoretical model that investigates flight to quality and flight to liquidity effects on asset pricing. The framework of the model is based on the idea that investors are fund managers and they keep track of their fund performance. During the times of higher market volatility, the probability that the fund performance will deteriorate increases. As in the times of market uncertainty liquidity risk rises, investors are less willing to hold illiquid assets and thus require higher rate of return. Consequently, uncertainty in the market is allied with flight to liquidity.
An interesting result emerges from this model. During times of high volatility, investors are more risk averse. Subsequently, with higher risk aversion there is a flight to quality effect, which is the time when investors move towards safer assets.
Rösch and Kaserer (2013) provide an empirical evidence of commonality in liquidity and flight-to-quality effect during the financial crisis. Their findings corroborate with proposition of Brunnermeier and Pedersen (2009) model. There is a relation between liquidity risk and credit risk, and scarce funding liquidity causes liquidity dry-up. In times of market downturns investors shift towards less risky assets, which is the evidence of flight to quality effect. What is important, they find that commonality liquidity exists and varies over time. The effect increases during market declines and peaks during the financial crisis, and during tight funding liquidity.
The existent research on market-wide liquidity shows that liquidity phenomena such as commonality in liquidity, flight to quality, flight to liquidity and liquidity dry-up are closely related and, intuitively, have common determinants. When markets decline, market liquidity decreases.
Market declines in return and liquidity induces flight to liquidity, i.e. more investors transfer their capital into safer assets.
3. Sovereign credit ratings
3.1. Credit Rating Agencies
The credit rating agencies (CRA) are private companies that measure the creditworthiness, i.e. the debtor’s ability to repay its debt, of sovereign nations, private companies, non-profit organizations and state or local governments. The CRAs rate debt instruments such as government bonds, corporate bonds, preferred stocks, mortgage-backed securities and collateralized debt obligations5. The biggest CRAs in the world are Standard & Poor’s, Fitch Ratings and Moody’s with the with a market share of 95% as for 2013, which makes them the “ Big Three” of the industry6. The
business model of these companies is based on the scheme where issuer of the debt pays for the rating to be assigned.
The mission of credit rating agencies is to facilitate the trading of securities on the secondary market by evaluating the credit risk of borrowers. Their judgment is based on quantifiable factors, such as economic stability, as well as qualitative, as political situation. Thus, their ratings are rather subjective; and their measure does not indicate absolute probability. The agencies use both public and private information to assess the credit risk, and therefore provide valuable proxy for the credit risk. However, the trustworthiness and value of the ratings have been widely interrogated, as well as consequences of the newly assigned ratings7.
The ratings evaluate solely the credit risk, and do not include market or liquidity risk (De Haan and Ambenbrink, 2011). Standard and Poor’s describe their measure of the credit risk as “Standard & Poor’s credit ratings are designed primarily to provide relative rankings among issuers and obligations of overall creditworthiness; the ratings are not measures of absolute default probability. Creditworthiness encompasses likelihood of default and also includes payment priority, recovery, and credit stability” (Standard and Poor, 2012).
3.2. The concept of sovereign credit ratings
Sovereign debt is the government bond that is issued in a foreign currency to foster a country’s economy growth. The ability to repay this debt instrument provides a benchmark for a
5 Source: “The Credit Rating Controversy”, Council on Foreign Relation. URL:
http://www.cfr.org/financial-crises/credit-rating-controversy/p22328
6 Source: same as above.
country’s economy climate. It is the function of credit rating agencies is to assess a country’s ability to repay its sovereign debt. Thus, a proxy for country credit risk, sovereign risk, indicates the capability and willingness of the country to repay its sovereign debt.
3.3. Assessment methodology
The general methodology for measuring sovereign debt risk is to weigh factors that have impact on the creditworthiness of the country to repay the debt. S&P’s use five-factor model to quantify sovereign risk. This includes economic structure and growth prospects, institutional and governance effectiveness and security risks, international investment position, fiscal performance, external liquidity, and monetary flexibility (Standard and Poor, 2013). Moody’s use four-factor model and provides weighs for each of these factors. It comprises economic strength, institutional strength, fiscal strength and susceptibility to event risk. Credit rating agencies publish rationale of their methodology, however the exact weighs on each factor are not given.
Figure 1 Moody’s sovereign debt ratings measurement methodology.
Information that is used for assessing the credit risk is mainly publicly available, historic data from such sources as World Bank. In some cases, however, credit agencies rely on additional information that cannot be disclosed8. Because it is nearly impossible to accurately predict future events, such
ratings are reviewed on regular basis.
3.4. Implications of the sovereign debt ratings
Although sovereign debt ratings provide a benchmark for country risk, they have been widely criticized during the recent financial crisis. As an example, in 2011 Standard and Poor’s unprecedentedly downgraded United States from AAA to AA+, while before the crisis they
8 Source: SEC, 2002, Statement of Standard and Poor’s on credit rating agencies. URL:
http://www.sec.gov/news/extra/credrate/standardpoors.htm ; Moody’s, 2013, Sovereign Bond ratings.
favorably graded bankrupting institution Lehman Brothers9. This stimulated the overall panic and contributed to the breakdown of the U.S. housing market. The controversy of the credit rating agencies lies not only in the occasional inability to accurately assess the credit risk, but also in effect that it has on the financial markets. Therefore, it is important to understand the structure of sovereign rating changes.
Prior research analyzed the role of sovereign rating and its effect on markets. In general, sovereign ratings play an essential role in detecting countries’ ease to get a capital in international capital markets, and that rated ability is useful in forecasting sovereign defaults (Reinhart, 2002). However, the existent research provides evidence that changes in ratings may have an impact on country’s economic cycles. Chen et al. (2013) identified the effect of sovereign debt rating changes on private investment. According to the article, upgrade in a rating results in rise of private investment, whereas downgrade consequences in temporary decline. The authors suggest that sovereign rating may have an influence on cost of capital. That is, when the country’s sovereign debt is downgraded, it makes negative signals to the market and brings uncertainty. As a result, a flight-to-quality effect might occur – investors will transfer their capital to less risky countries.
There is an evidence of self-fulfilling prophecy of sovereign ratings, which implies that changes in ratings have impact on macroeconomic environment (Reinhart, 2002; Ferreira and Gama, 2007). Ferreira and Gama (2007) provide an empirical evidence of significant, asymmetric effect of a country’s credit ratings downgrade on another countries’ return spread, and the magnitude of the effect depends on geographic proximity and whether this country is emerging or developed. This evidence suggests that news related to the country’s sovereign risk spill over to countries abroad. The recent study of Schumacher (2014) provides evidence that there is a significant two-way interaction between macroeconomic conditions and changes in sovereign ratings. That is, changes to ratings, which reflect country’s macroeconomic and political situation, can themselves fortify a country’s economic cycle.
4. Data
1.1. Data sources and screens
The initial data on trading activity is gathered from Datastream. The main variables are: daily adjusted official closing price, returns, turnover in volume and number of common shares outstanding. The sample includes stocks listed on stock exchanges of Greece, Italy, Portugal and
Spain, respectively: Athens Stock Exchange, Milan Stock Exchange, Lisbon Stock Exchange, and Madrid Stock Exchange. The period covered is January 1995 - December 2013. Daily returns are calculated as relative change in Total Return Index (RI), which is defined as theoretical growth in value of a share over a period of one day, assuming dividends are re-invested. Turnover by volume is the number of shares that is traded on a particular day, expressed in thousands.
5.2. Liquidity measure
To compute individual stock liquidity, Amihud (2002) ILLIQ measure is used. It is defined as a daily proportion of absolute stock returns to the trading dollar volume and can be understood as “daily price response associated with one dollar of trading volume” and can be interpreted as a price impact. This measure was chosen because of relative accessible data in comparison with order book based measures, such as quoted or effective bid-ask spread.
Following Karolyi et al. (2012), initial ILLIQ is log-transformed to decrease the effect of outliers, and multiplied -1 to reach values that are increasing in individual stocks liquidity: 𝐿𝑖𝑞𝑢𝑖𝑑𝑖𝑡𝑦!,!,! = −𝑙𝑜𝑔 !"#$%&'#:!"""|!"#$%&| , where Turnover is price multiplied by volume and Return is a daily percentage change of total return index. To reduce the effect of outliers, I dropped the observations with daily liquidity in the top or bottom 1% of cross-sectional distribution of all stocks. In addition to individual stock liquidity, market liquidity computed. Market liquidity is daily cross-sectional value-weighted average, calculated using market capitalization weighs. To prevent biased estimators of commonality in liquidity, in the computations of aggregate market liquidity stock i was excluded. Thus, daily market liquidity measure is slightly different for each observation.
In line with Karolyi et al. (2012), I use following data screening procedures. Trading days with more than 90% of individual stock returns equal to zero are defined as non-trading days. I discard stock-month observations with a stock price, turnover and monthly return in the 1st or 99th percentile of the cross-sectional distribution of all stocks. Following the approach used in Ince and Porter (2006) and Karolyi et al. (2012), daily returns are set to missing if previous or current value of Total Return Index (RI, Datastream) is less than 0.01, and stock-month observations are dropped if (1+Ri,t) x (1+Ri,t-1)-1<0.5, where Ri are monthly t returns of a stock I, Ri,t orRi,t-1 is more than or equal to 300%.
1.2. Sovereign debt ratings
Sovereign debt long-term ratings for Greece, Portugal, Italy ans Spain are taken from Moody’s official website. In the sample, only assigned rating were considered, i.e. without watchlists and outlooks. The rating announcement of a particular day is attributed to the month in which it happened. The ratings are linearly transformed from 0 to 21 for Aaa10. The linear transformation of sovereign debt ratings for causal analysis and significance check was previously used by, among others, Ferri et al. (2009), Reinhart (2002), Afonso et al. (2007), Schumacher (2014). Table below summarizes the assigned values for Moody’s credit ratings.
5. Methodology
The first part of the methodology describes the model to compute the commonality in liquidity. The second part concerns the effect of sovereign debt ratings on commonality in liquidity.
5.1. Estimating commonality in liquidity
5.1.1. Liquidity measure
I use ILLIQ measure proposed by Amihud (2002). Following Karolyi et al. (2012) the daily stock liquidity innovation is the logistic transformation of the Amihud’s liquidity multiplied by -1, to reduce outliers:
𝐿𝑖𝑞!,! = −log (1 + !!,!
!!,!!"#!,!) (1)
Where 𝑅!,! is the daily return for each stock i and 𝑉𝑜𝑙!,! is the volume. Amihud (2002) interprets the measure as “a daily price response associated with $1 of trading volume”.
The Amihud’s measure was chosen for various reasons. First of all, this measure does not require intraday data, which management is difficult to execute within the given time frame. In addition to that, Amihud’s measure is empirically proved to be a plausible proxy for liquidity11.
5.1.2. Commonality measure
10 As D-grade does not exist in Moody’s grading, but does exist in S&P’s and Fitch, Moody’s C is set it to 1 to make it
more compatible with S&P’s and Fitch.
11 The empirical study of Lesmond (2005) discovers that the Amihud’s illiquidity measure is highly correlated with
bid-ask spreads in 23 emerging markets. Moreover, the significant amount of literature that analyzes commonality in liquidity and systematic liquidity risk. It is used, among others, by Karolyi et al. (2012), Acharya and Pedersen (2005), Kamara, Lou, Sadka (2008).
Chordia et al. (2000) first popularized commonality in liquidity, defining it as “co-movement of changes in individual stock and changes in market liquidity”. They capture this effect by calculating time series regressions of individual stock liquidity on the aggregate market liquidity. This approach was further developed by several studies. Chordia et al. (2005) report significant cyclical effects of days of the week and days around holidays. Therefore following Hameed, Kang, and Viswanathan (2010) I first calculate residuals of the time series filtering OLS time series regressions for each stock i built upon d observations in a period of each month t:
𝐿𝐼𝑄!,!,! = 𝛼!,!𝐿𝐼𝑄!,!,!!!+ ! 𝛽!,!,!𝐷! + 𝛿!,!𝐻𝑂𝐿𝐼𝐷!,!+ 𝜀!,!,!
!!! , (2)
where 𝐿𝐼𝑄!,!,! denotes the daily d individual liquidity of each stock i within a month t, 𝐷! is a set of dummy variables for each day of the trading week from Monday through Thursday, and 𝐻𝑂𝐿𝐼𝐷!,! is a dummy for days around holidays12. The lagged value 𝐿𝐼𝑄!,!,!!! was added to measure the innovations of daily individual liquidity; it is needed to further estimate the commonality measure.
To measure the commonality, the residuals 𝜀!,!,! from regression of equation (2) are used. They capture the seasonality effect and take the daily innovations in liquidity for each stock within a month. Subsequently, I follow the approach proposed by Karolyi et al. (2012), using single-factor market model of OLS regression:
𝜀!,!,! = a!,!+ !!!!!b!,!,!𝜀!,!,!!!+ ω!,!,! , (3)
where 𝜀!,!,!!! denotes the innovations in market liquidity, calculated as a cross-sectional market – value weighted average of residuals from (2) of all stocks i within a month t, excluding the stock i to control for potentially misleading constraint. Thus, the monthly measure of market liquidity is slightly different for each observation. Market value weighs are taken from the market capitalization of each company in the end of each month. Leading and lagging value of market liquidity is included as in Chordia et al. (2000), to detect any lagged modification in commonality in liquidity.
Inspired by Roll (1988), the R2 of regressions of Eq.3 are used to compute the monthly measure of commonality in liquidity. It is computed as equally weighted average of the 𝑅! across individual stocks within a given month, with restriction of minimum 10 daily observations within each month.
12 The dummy was calculated based on the following logic. If a holiday is on Friday, then for subsequent Thursday
HOLID is set to 1, and if a holiday falls on Monday, then Tuesday is set to 1; on any other weekdays HOLID is set to 1 for the day prior and subsequent to the holiday.
In line with Morck, Yeung and Yu (2000) I will use a logistic transformation of the commonality measure, and thus commonality in liquidity is defined as:
𝑐𝑜𝑚𝑙𝑖𝑞!= ln (!!!!!!) (4)
5.2. Testing for the relationship between commonality in liquidity and sovereign debt ratings
To estimate the causal relationship between sovereign debt ratings and liquidity commonality, time series approach will be used. The econometric methodology that will be employed is a dynamic VAR framework and it is applied as follows; first the commonality and rating time series are checked for a unit root, then the OLS panel VAR is estimated with fixed effects. After that, Wald F-test will be performed to test the Granger Causality of the both commonality in liquidity and rating variables.
5.2.1. Panel Unit Root test
There is empirical evidence that panel unit root tests, in comparison with individual series unit root tests provide higher statistical power. Recently developed tests comprise Im, Pesaran and Shin (IPS) (2003), Breitung (2000) (Breitung), Hardi (2000) (Hardi), Levin et al. (LLC) (2002). The basic model for these tests is the standard AR(1) process
𝑦!" = 𝜌!𝑦!"!!+ 𝛿!𝑋!"+ 𝜀!" (5)
where i=1, 2,….,N is the number of cross-sections over a period t=1,2,…,T, 𝜌! is the autoregressive coefficient, and 𝑋!" is the exogenous variable that reflects any trend or fixed effects. The null and alternative hypotheses are defined as follows:
𝐻!: 𝜌! = 1, 𝑖 = 1,2, … , 𝑁 (6) 𝐻!!: 𝜌! < 1, 𝑖 = 1,2, … , 𝑁; 𝑖 = 1,2, … , 𝑁 (7) 𝐻!!: 𝜌! < 0, 𝑖 = 1,2, … , 𝑁!; 𝜌! = 1, 𝑖 = 𝑁!+ 1, 𝑁!+ 2, … , 𝑁 (8)
For all the tests mentioned above, under the null hypothesis, each time series is non-stationary, i.e. contains a unit root. However, there are two types of alternative hypotheses for these tests. The first is so called homogenous alternative 𝐻!! that assumes that autoregressive coefficient 𝜌! is identical for all the cross-sections. LLC, Breitung and Hardi tests are based on this hypothesis. The second type is heterogeneous alternative 𝐻!!, under the significant fraction of autoregressive coefficients does not have a unit root. IPS test employs this alternative.
The sample I analyze consists of stock exchanges from four different countries, with different level of economic development, and financial markets development. Consequently, heterogeneity may arise. Therefore I adopt IPS panel unit root test that allows for heterogeneity among autoregressive coefficients.
5.2.2. Model estimation
The VAR(p) system has the following form:
𝑟𝑎𝑡𝑖𝑛𝑔! = 𝑎! + 𝑎! !𝑟𝑎𝑡𝑖𝑛𝑔!!!+ 𝑏! !𝑐𝑜𝑚𝑙𝑖𝑞!!!+ 𝑢! (9) 𝑐𝑜𝑚𝑙𝑖𝑞!= 𝑎!+ 𝑎! !𝑐𝑜𝑚𝑙𝑖𝑞!!! + 𝑏! !𝑟𝑎𝑡𝑖𝑛𝑔!!!+ 𝑣! (10)
where 𝑟𝑎𝑡𝑖𝑛𝑔! is a numerical value of Moody’s sovereign long-term debt ratings, recoded as 1 for C to 21 to Aaa. 𝑐𝑜𝑚𝑙𝑖𝑞! is the commonality in liquidity measure computed in the previous step. I identify he number of lagged variables p by minimizing Schwarz criterion (SC). The variables are taken in levels or first differences, depending on the output of the panel unit root test.
Vector Autoregression has proven to be a powerful tool for forecasting. However, structural inference based on VAR models is tricky; it requires distinguishing correlation from causation. Therefore, VAR method for structural inference is usually used as a tool for computing Granger causality, impulse response functions and variance decomposition (Stock and Watson, 2001).
5.2.3. Testing for causality
Having established the stationarity of variables and estimated the panel VAR model of Eq. 9 and 10, the final step is to check whether there is a significant relationship between commonality in liquidity and sovereign debt ratings. In order to identify the direction of this relationship, the Granger causality test is employed. Specifically, the Granger causality is estimated using Wald F-test that assess the joint significance of coefficients of lagged independent variables. Under the null hypotheses, lagged comliq (rating) coefficients are jointly zero and thus comliq (rating) does not Granger-cause rating (comliq). The rejection of the null hypothesis gives us the result when one variable Granger-causes another. Since the term causality has many definitions and is highly philosophical, Granger-causality test does not provide clear-cut results. Instead, it helps to understand whether the independent variable x helps to predict dependent variable y based on its past values.
6. Descriptive statistics
This section presents the descriptive statistics of stock-related data, liquidity measures and commonality in liquidity estimates. Stock-related data and liquidity are presented in descriptive statistics, and commonality in liquidity graph depicts its variation over time.
Table 1 shows the statistics for trading-related data for Greece, Italy, Portugal and Spain. The presented statistics is calculated using the initial data for the period of January 1995 – December 203 gathered from Datastream. The number of unique stocks does not necessarily reflect the size of the market; it is the number of the analyzed firms for which the data were collected to the Datastream database. Return, turnover and size ratios are calculated for each individual stock.
Table 1. Descriptive statistics of stock data
This table reports descriptive statistics of the data related to trading activity over the period of 1:1995-12:2013. The sample consists of daily data of stocks listed on the major stock exchanges of four countries. The first two columns present number of unique stocks and daily observations in the sample. Return is daily percentage change in total return index (RI) for a given stock, Turnover of individual stocks is calculated as percentage ratio of shares traded and number of common shares outstanding for the beginning of a given year. Size is daily market capitalization of a given stock, expressed in millions.
#Unique stocks
#Daily
obs. Return, % Turnover, % Size, mln.
mean median st.dev. mean median st.dev. mean median st.dev. Greece 168 464997 0.31 -0.12 8.34 1.80 0.21 18.62 502.45 68.14 1427.28
Italy 272 731836 0.15 -0.05 5.64 2.60 0.39 24.99 2270.80 286.96 7548.66 Portugal 59 128942 0.15 0.04 5.35 0.66 0.16 2.32 1946.06 434.39 3101.61 Spain 112 307565 0.16 0.05 5.29 3.57 1.07 8.41 4323.56 830.62 10992.49
As we can see from the table above, the biggest size, i.e. daily capitalization of a given has Spain with average of 4323.56 mln.€, while Greece has drastically lower average of 502.45 mln.€. Spain has also the highest returns, 0.16% in average over the period 14 years and the highest turnover – 1.07% shares traded over shares outstanding. Over the analyzed period the individual stock return on average were decreasing in Greece and Italy, with median -0.12% and -0.05%, while in Portugal and Spain returns were increasing, 0.15% and 0.16% of an increase subsequently.
Table 2 presents the descriptive statistics of market liquidity measures and commonality in liquidity estimates. Market liquidity is calculated as a aggregate monthly average of individual stocks liquidity. Individual stock liquidity is Amihud ILLIQ measure, which is log-transformed to reduce the effect of outliers, and multiplied by -1. Hence, the measure is increasing with increasing liquidity. Table 2 shows the evidence that the most illiquid and volatile in liquidity market over the analyzed period is Greece. With the average market liquidity is -3.56, this drastically differs from the
situation in the rest of the countries. This can be explained by the fact that the financial crisis hit Greece and deeply altered the market liquidity.
The average of individual month commonality measures is taken to compute commonality in liquidity, with a restriction of minimum of 10 stocks within a given month. The average commonality in liquidity in all four countries falls in range of 20-30%, as it is presented in Table 2. With the common variation in market and individual liquidity to be in that range, it is also very volatile – 4-8%.
Table 2. Market liquidity and commonality in liquidity measures
Market liquidity is aggregate monthly commonality in liquidity for the country, calculated as mean of monthly
liquidity commonality measures of all stocks. Commonality in liquidity is defined as R2 of OLS regression
𝜺𝒊,𝒕,𝒅= 𝐚𝐢,𝐭+ 𝒋!!𝟏𝟏 𝐛𝐢,𝐭,𝐣𝜺𝒎,𝒕,𝒅!𝒋+ 𝛚𝐢,𝐭,𝐝, where 𝜺𝒊,𝒕,𝒅 is a residual from filtering regression from Eq.2, defined as
daily innovation of individual stock liquidity, and 𝟏𝒋!!𝟏𝜺𝒎,𝒕,𝒅!𝒋 are added contemporaneous, leading and lagging
innovation in market liquidity.
Market liquidity Commonality in liquidity mean median st.dev. mean median st.dev. Greece -3.4262 -3.5647 0.8962 0.2239 0.2103 0.0580 Italy -0.3658 -0.3605 0.1584 0.2685 0.2542 0.0742 Portugal -0.4943 -0.0538 0.9421 0.2184 0.2122 0.0455 Spain -0.1844 -0.1795 0.0767 0.2120 0.2024 0.0588
Monthly time series that are presented on Figure 2 allow us to investigate the time-variation of liquidity commonality. As can be seen on the graph, the liquidity commonality in all four countries is rather very volatile. However, an interesting pattern can be noticed. The commonality increases markedly during the recent financial crisis. In Greece, for instance, the two peaks of commonality over the period of 2007-2012 happened exactly on, or 1 month before the Moody’s sovereign debt downgrade announcement. For example, the liquidity commonality raised from 23% in November 2009 to 38% in December 2009, when Moody’s announced the downgrade of the Greece rating from A1 to A2.
Nonetheless, these patterns could also represent other underlying determinants of liquidity commonality, or just a statistical noise. Therefore, the final step of my empirical research is to test whether the sovereign debt rating affects commonality in liquidity.
Figure 2. This figure depicts the aggregate commonality in liquidity monthly time series, calculated as an average of
monthly individual stock liquidity commonalities. The aggregate commonality liquidity is calculated with minimum 10 stocks within a month. The blue reference line indicates the upgrade of Moody’s sovereign debt rating for a given country, while the light-brown line represents the downgrade of that rating.
7. Empirical results
This section contains the empirical results of the main hypothesis, which is what is the effect of sovereign debt ratings on the commonality in liquidity. The first step is to test the commonality in liquidity and sovereign debt ratings for the non-stationarity.
Table 3 shows that the IPS panel unit root test is conclusive for rating, with 1% of the level of significance; comliq is stationary. After taking the first differences from both variables,the rating and comliq coefficients do not contain unit roots, hence are stationary.
Table 3. IPS Panel unit root tests
This table presents the results of the IPS panel unit root test for the full sample. The lag length is selected using Schwarz criterion. *** represents the significance at 1% level.
Variable Level First difference
Intercept Intercept and trend Intercept Intercept and trend
comliq -23.64 -30.13 -27.22 -28.42
rating 4.88*** 4.64*** -15.422 -22.31
The output of the Panel VAR model with fixed effects is presented in Table 4. The Panel VAR with 4 lags was estimated using OLS method and variables are first-differenced before the estimation was done. 2 equations are formulated in order to check for the direction of causality, further testing for Granger causality. As we can see from Table 4, the coefficients of the independent variables are not significant in both of the equations. The reason for that may be the small size of the sample, i.e. only four countries that gives a low variation in sovereign debt rating changes. However, the conclusions cannot be drawn from the PVAR coefficients; they are computed for further Granger causality test.
Table 4. Panel VAR(4) with fixed effects model of the commonality in liquidity and sovereign debt ratings model estimation output
This table presents the output of the panel VAR model with fixed effects of eq. 9 and 10. The data gathered is from the estimated commonality in liquidity for four countries, and Moody’s sovereign debt ratings over a period of 1:1995-12:2013. The panel consists of 4 cross-sections and with 223 observations in each and 892 in total. The lag length is selected using Schwarz criterion. The *, **, *** represents significance at 10%, 5% and 1% level respectively.
Lags
Variables 1 2 3 4
Dependent variable - rating
constant 0.0061 (0.102) d.rating 0.0197 (0.597) 0.0727** (2.165) 0.0149 (0.443) 0.2143*** (6.379) d.comliq 0.0359 (0.917) -0.0438 (-1.116) 0.0129 (0.329) 0.0231 (0.596)
Dependent variable - comliq
constant -0.5329*** (-10.56) d.comliq 0.1163*** (3.495) 0.1277*** (3.835) 0.1030*** (3.115) 0.0909*** (2.759) d.rating -0.0282 (-1.005) 0.0107 (0.375) 0.0499 (1.750) -0.0003 (-0.009)
As a final step, the Granger causality is tested using Wald test on VAR coefficients estimated in the previous step. Under the null hypothesis, all of the 4 coefficients of the explanatory variables are jointly zero. This means that explanatory variable does not Granger-cause dependent variable. Under the alternative hypothesis, all of the coefficients of explanatory variable are different from zero, hence, this variable Granger-causes dependent variable.
Table 5. Wald-F test is performed using the coefficients estimated in PVAR(4) model. The lag length is selected using
Schwarz criterion. *** represents the significance at 1% level.
Null Hypothesis F-statistic p-value
Rating does not Granger-cause commonality in liquidity
4.08*** 0.0068
Commonality in liquidity does not Granger-cause rating
0.58 0.6751
Table 5 presents the results of the Wald test for coefficients of the PVAR(4) model, presented in Table 4. The test is conclusive for sovereign ratings: the sovereign debt rating Granger-cause commonality in liquidity at 1% of significance level. This means that past values of sovereign debt rating have strong predictive power for commonality in liquidity. From the other hand, commonality in liquidity does not Granger-cause changes in sovereign debt ratings.
8. Robustness checks
This section presents the robustness tests that are performed to check for the valid interpretation of results. The first robustness check is to estimated Panel VAR(p) model from Eq. 9 and 10 with different lag length. Selecting the lag length is an essential step because may be sensitive to the lag structure. In general, choosing too many or too few lags is both problematic for the estimated coefficients. From the one side, selecting too few lags leads to omitting potentially important observations and thus to biased results. From the other side, too many lags lead to using observations that are not necessary, hence to the high standard error and imprecise outcomes.
Table II of the Appendix presents the results of the Wald tests that are tested on coefficients for Panel VAR using 2-8 lags. As it is shown in the table, results are consistent with the main result for 2-5 lags, while for 6 and 8 lags results were insignificant. In other words, PVAR model with up to 5 lags leads to the result when rating Granger-causes commonality in liquidity and not the other way around, while from lag 6 to 8 the results show that there are now Granger-causality from both directions.
9. Conclusion
This research studied 611 unique stocks listed on the major stock exchanges of four countries – Athens Stock Exchange of Greece, Milan Stock Exchange of Italy, Lisbon Stock Exchange of Portugal and Madrid Stock Exchange of Spain. Out of the trading-related activity data for the period of January 1995 – December 2013 collected from Datastream, the monthly commonality in liquidity estimates were computed. Prior to the estimation of the commonality in liquidity, daily individual stock liquidity was regressed on its lag, to take the daily innovations, and controlled for possible seasonality and cyclicality effects – using day of the week and holidays binary variables. I have found that commonality in liquidity is existent and it varies over time.
Following that, the assigned sovereign debt ratings were linearly transformed. Based on the panel VAR model with fixed effects, Wald F-test was performed to examine the Granger causality of commonality in liquidity and ratings. It was found that sovereign debt ratings Granger-cause commonality in liquidity, but commonality in liquidity does not Granger-cause sovereign debt ratings. Basically, these results imply that past values of sovereign debt ratings help to predict future values of commonality in liquidity, but not the other way around. Performing the robustness test using different lag lengths, I found that there is a significant evidence of Granger causality of ratings only for 2-5 lags, while the test loses its power starting from the lag 6. This may imply that sovereign debt ratings do not have a long-term impact on commonality in liquidity.
The main aim of this research was to shed light on factors that affect commonality in liquidity. The material commonality in liquidity may create a source of non-diversifiable, systematic risk and thus is important for the asset pricing field of research. The determinants of commonality in liquidity have not yet been fully identified; however, researchers divide the determinants into supply and demand determinants (Karolyi et al., 2012; Rosch and Kaserer, 2013). Demand determinants are related to the investors correlated trading activity, investor sentiment; supply determinants are mainly associated to the funding liquidity of the market.
As the main hypothesis of this thesis it whether sovereign debt ratings affect commonality in liquidity, this research analyzes the drivers of commonality in liquidity from the new light. From the one hand, sovereign debt ratings act as trigger to the change in investor sentiment, and from the other hand, as the reflection of a country’s economic and political situation (which may impact funding liquidity). Indeed, credit rating agencies assign a sovereign debt rating to a country and that rating should reflect the true economic and political situation in the country. However, the problem with such rating is that it is forward-looking assessment of a country’s capability to repay the sovereign
debt. And because future is unpredictable, there is a certain subjectivism in every CRA’s assessment. As it was previously evidenced, sovereign debt rating can fortify a country’ s economic cycle and shift private investments (among others, Reinhart, 2002; Chen et al., 2013). But if the evidence of my research is correct then sovereign debt ratings have impact on not only economy, but on the deeper level of financial markets.
My research has its limitations and more extensive and thorough research can be carried out. Firstly, the small size of my sample can be extended. Small sample size may constitute a threat to internal validity of the critical values of the hypothesis test. This is because the conclusion based on normality of critical values are validated with large-sample approximation (Stock and Watson, 2012). Thus, more countries can be added. Another limitation is the method of calculating liquidity. Prior research suggests, that order-book-based measures of liquidity provide more precise approximation (Lesmond, 2005). Moreover, it was shown that commonality in liquidity gets weaker the deeper data is analyzed in limit order books (Korajczyk and Sadka, 2008). Therefore, intraday order-book-based data can be analyzed to get a different view on commonality in liquidity.
Finally, because commonality in liquidity is still a developing area of the financial research, no clear implications for the business and policy makers can be drawn. However, despite the limitations of the research, this study contributes to the existent literature by providing an evidence of a new factor that affects commonality in liquidity.
10. References
Acharya, V. & Pedersen, L., 2005. Asset pricing with liquidity risk. Journal of Financial Economics. 77(2), p. 275-410.
Afonso, A., Gomes, P., Rother, P., 2011. Short-and long-run determinants of sovereign debt credit ratings. International Journal of Financial Economics. 16 (1), 1–15.
Amihud, Y., 2002. Illiquidity and Stock Returns: Cross-section and Time-series Effects. Journal of Financial Markets. 5(1), p. 31-56.
Breitung, J., 2000. The local power of some unit root tests for panel data. Advances in Econometrics 15, 161–177.
Brockman, P ; Chung, DY ; Perignon, C., 2002 Commonality in Liquidity: A Global Perspective. Journal Of Financial And Quantitative Analysis, Vol.44(4), pp.851-882
Brunnermeier, M.K., Pedersen L.H., 2008. Market Liquidity and Funding Liquidity. Review of Financial Studies. 22(6), p. 2201-2238.
Chen, S., Chen, H., Chang C., Yang, S., 2013. How do sovereign credit rating changes affect private investment? Journal of Banking and Finance 37, 4820-4833.
Chordia, T., Roll, R., Subramanyam, A., 2000. Commonality in liquidity. Journal of Financial Economics. 56(1), p. 1-28.
Dittmar, R.F., Yuan, K., 2008. Do sovereign bonds benefit corporate bonds in emerging markets? Review of Financial Studies 21, 1983–2014.
Ferreira, M.A., Gama, P.M., 2007. Does sovereign debt ratings news spill over to international stock markets? Journal of Banking and Finance. 31 (10), 3162–3182.
Gande, A., Parsley, D.C., 2005. News spillovers in the sovereign debt market. Journal of Financial Economics 75, 691–734.
Hameed, A.; Kang, W. ; Viswanathan, S., 2010. Stock Market Declines and Liquidity. The Journal of Finance, Vol.65(1), pp.257-293
Kamara, A. ; Lou, X. ; Sadka, R., 2008. The divergence of liquidity commonality in the cross-section of stocks. Journal of Financial Economics, Vol.89(3), pp.444-466
Korajczyk, R.A., Sadka, R., 2008. Pricing the commonality across alternative measures of liquidity. Journal of Financial Economics. 87, p. 45-72.
Im, KS ; Pesaran, MH ; Shin, Ym 2003. Testing for unit roots in heterogeneous panels. Journal Of Econometrics, 2003, Vol.115(1), pp.53-74
Lee, K., 2011. The world price of liquidity risk. Journal of Financial Economics, Vol.99(1), pp.136-161
Lesmond, D. A., 2005. Liquidity of emerging markets. Journal of Financial Economics, Vol.77(2), pp.411-452
Levin, A., Lin, C.F., Chu, C., 2002. Unit root tests in panel data: asymptotic and finite-sample properties. Journal of Econometrics 108, 1–24.
Markowitz, H., 1952. Portfolio selection. The Journal of Finance, Vol.7(1), pp.77-91
Morck, R.,Yeung,B.,Yu,W.,2000.The information content of stock markets: why do emerging markets have synchronous stock price movements? Journal of Financial Economics 58, 215–260. Reinhart, C.M., 2002. Default, currency crises, and sovereign credit ratings. World Bank Economic Review. 16 (2), 151–169.
Rösch C.R, Kaserer K., 2012. Market Liquidity in the Financial Crisis: The Role of Liquidity Commonality and Flight-to-Quality, Journal Of Banking & Finance. Vol.37(7), pp.2284-2302
Standard and Poor, 2013. Sovereign Government Rating Methodology and Assumptions.
Stock, J. H., & Watson, M. W. (2012). Introduction to Econometrics: Global Edition. Pearson Education
11. Appendices
Appendix I. Sovereign debt ratings
Table A 1. Linear transformation of Moody’s rating scale
Interpretation Rating Numeric value
Investment grade
Minimal credit risk Aaa 21
Very low credit risk
Aa1 20
Aa2 19
Aa3 18
Low credit risk
A1 17
A2 16
A3 15
Moderate credit risk
Baa1 14
Baa2 13
Baa3 12
Substantial credit risk
Ba1 11
Ba2 10
Ba3 9
Speculative grade
High credit risk
B1 8
B2 7
B3 6
Very high credit risk
Caa1 5 Caa2 4 Caa3 3 Near default Ca 2 Default C 1 - 0
Source: Own compilation based on Moody’s Ratings & Defintions, 2009
