Liquidity, liquidity commonality and the business cycle : an analysis on the relationship between financial markets and the real economy in the Eurozone between 1996 and 2014

Liquidity, liquidity commonality and the business cycle

An analysis on the relationship between financial markets and the real economy in the Eurozone

between 1996 and 2014

Author: Thijs Beudeker

10837841

Supervisor: Patrick Tuijp

Date: 22-07-2015

___________________________________________________________________________

Abstract

This paper analyzes the relationship between financial markets and the real economy business cycle in the Eurozone. Using panel data and principal component analysis, we find that stock market liquidity and bond market liquidity are not strong explanatory variables of business cycle indicators such as GDP growth and growth in unemployment. However, we show that liquidity commonality explains growth in GDP well in a number of specifications. The introduction of a common currency and common interest rate significantly changes liquidity commonality between stock markets. In the period after the introduction of the Euro and before the fall of Lehman Brothers, we show that liquidity commonality increases. Additionally, there are signs that the trend in liquidity commonality changes after the outbreak of the financial crisis. A regression model using interaction terms of liquidity commonality and time dummies confirms these results.

JEL-classification: G10; G15

Keywords: liquidity; liquidity commonality; business cycle; Eurozone

2

Statement of Originality

This document is written by Student Thijs Beudeker who declares to take full responsibility for

the contents of this document.

I declare that the text and the work presented in this document is original and that no sources

other than those mentioned in the text and its references have been used in creating it.

The Faculty of Economics and Business is responsible solely for the supervision of completion

of the work, not for the contents.

3 I. Introduction

In finance, one branch of literature is concerned with a phenomenon experienced every day by traders, investors, and governments in financial markets: market liquidity (Gennotte and Leland, 1990; Brunnermeier and Pedersen, 2009; Bhide, 1993; Chordia, Roll and Subrahmanyam, 2001). Liquidity in financial markets is important to governments, central banks and all other market participants alike for various reasons. Traders and investors are better off in a liquid market as it becomes easier to trade small and large amounts of securities. The costs of a roundtrip trade are smaller when spreads are smaller, and they can find counterparty to trade with quickly when a market is very liquid. Furthermore, informed traders can more easily hide between uninformed traders in a liquid market (Kyle, 1985)

Market liquidity impacts the core activities of (a government and) a central bank in three ways. First, market liquidity will have an impact on the formation of monetary policy and the implementation activities of their policies (Adrian and Shin, 2009). However, market liquidity also determines how quickly new information is priced, i.e. how quickly prices move towards the fundamental value. Additionally, low levels of market liquidity may harm the transmission of monetary policies via open market activities (Bernanke and Gertler, 1995).

Second, market liquidity can dry up during systemic crises, while a liquidity freeze can also cause disruptions in the market (Gravelle, 1999). Liquidity and solvency problems for market participants – such as market makers - can arise due to amplification rather than dampening of shocks to financial markets as a result of specific levels of liquidity. Naes, Skjeltorp and Odegaard (2011) show that portfolio holdings shifts coincide with deterioration in liquidity. In turn, these problems may lead to liquidity spirals and a credit crunch (Brunnermeier and Pedersen, 2009) or payment system disruptions (Muranaga and Shimizu, 1997). Thus, variation in market liquidity potentially interferes with the central bank’s activities as a lender of last resort as well as in its supervision of financial stability.

Third, for investors a liquid market means being able to trade for fair value and at favorable conditions. With a small bid-ask spread, both informed and liquidity traders will not buy or sell for prices that deviate much from the fair value of the financial security. In a deep market, they are able to more easily hide their true intentions of trading as their price impact is small. Additionally, a resilient market ensures that when they trade, the observed prices are close to the fundamental value as information is quickly incorporated into the price of an asset (Kyle, 1985).

4 Finally, in a liquid market, governments are able to issue sovereign debt instruments more easily and at better prices without disturbing the market too much. Central banks implement monetary policy via open market operations, for which liquidity a key factor (Carpenter and Demiralp, 2006). Liquidity in the secondary market (for government securities such as notes, bills and bonds) is very important: if liquidity is sufficiently high, governments are able to issue large amounts of debt at low cost in the primary market. As investors are confident in trading the securities at favorable conditions in the secondary market, they are willing to provide liquidity in the primary market.

Academic literature in finance concerning market liquidity has historically focused more on the United States of America (Chalmers and Kadlec, 1998; Chordia, Roll, Subrahmanyam, 1999; Acharya and Pedersen, 2003; Pastor and Stambaugh, 2003; Amihud, 2002), while Europe has been attracting more attention in recent years (Jankowitsch et al., 2002; Beber, Brandt and Kavajecz, 2008; Manganelli and Wolswijk, 2009). This research motivates why researching liquidity in Europe is interesting: with the introduction of the Euro, we have a natural experiment we can exploit to research the consequences for market liquidity when a common currency is introduced. It therefore fits in well in the literature with a focus on Europe. The Euro was introduced in physical form on January 1st, 2002 in 12 countries and has since then been introduced as the common currency and sole legal tender in 7 more countries. One of the main arguments for a common currency in the Eurozone was market integration: the Euro removed exchange rate risks between the involved countries, increasing trade and in turn leading to higher economic growth. One of the elements that should be affected is liquidity via a common interest rate set by the European Central Bank, as interest rate (risk) is an important driver of liquidity (Lucas, 1990; Hellwig, 1994; Brunnermeier, 2008). Additionally, we are also able to test the results of Naes, Skjeltorp and Odegaard (2011) for the Eurozone, who find a strong relationship between liquidity in the stock market and the real economy business cycle in the US.

This thesis aims at answering the following questions: Are liquidity and liquidity commonality related to

the real economy business cycle in the Eurozone? Has the introduction of the Euro changed liquidity commonality in the Eurozone? With the introduction of a common currency we would expect business

cycles to converge as trade between countries increases (no exchange rate risk) and a common monetary policy is introduced. When we look at GDP growth in Figure 1, we see that many countries are growing in a similar trend and apparently converge in the years following the introduction of the Euro. However, with the outbreak of the financial crisis we see a large divergence between several Eurozone countries only to converge in the following years. Several studies show evidence for convergence of business

5 cycles (Artis and Zhang, 1999), while De Haan, Inklaar and Jong-A-Pin (2008) show that business cycles of many Euro countries are still substantially out of sync. Massmann and Mitchell (2004) confirm that business cycles in the Eurozone diverge and converge over time.

Figure 1 GDP growth in the Eurozone between January 1996 and December 2014, displayed in percent per year. GDP growth is

calculated by taking log differences of nominal GDP data provided by the OECD.

Another indicator of the real economy business cycle is unemployment. As we can see in Figure 2, there are large differences in Europe in terms of unemployment. The same, but even clearer pattern emerges as in Figure 1 (GDP growth): convergence after introduction of the Euro and large divergence after the outbreak of the financial crisis. It is clear that business cycles in the Eurozone are not fully synchronized and seem to be sensitive to (negative) shocks. Therefore it is important to investigate what the consequences have been for the real economy and financial markets in the Eurozone with the introduction of a common currency. Had liquidity improved in the Eurozone, and has liquidity commonality increased?

6

Figure 2 Unemployment in the Eurozone 1996-2014, as a percentage of total workforce (seasonally adjusted). Data provided by

the OECD.

The short-term (3 month) and long-term (10 year) interest rates paid by governments are shown in Figure 3 and Figure 4. As expected, the interest rates converge quickly after introduction of the Euro. After the outbreak of the financial crisis, we see a clear divergence between two groups of countries. A pattern of flight-to-quality can be observed: both the short-term and long-term interest rates op Greece and Portugal increase substantially, while those of Germany, Austria and The Netherlands decline. However, from 2011 onwards the rates apparently converge, which coincides with new EU policy (EFSF, financial support programs, EFSM, open market operations by the ECB).

7

Figure 3 Long-term interest rates paid by Eurozone central governments, i.e. the rates paid on 10-year government securities.

Data provided by the ECB.

Figure 4 Short-term interest rates paid by Eurozone central governments, i.e. the rates paid on 3-month government securities.

8 This thesis is relevant for several strands of literature: first, we add to the literature on liquidity commonality by showing how liquidity commonality evolves over time in the Eurozone. The introduction of a common currency lets us research whether this impacts liquidity commonality. Second, our research is relevant for literature on how financial markets and the real economy are linked as we research explanatory variables for the real economy business cycle. Specifically, we offer a new explanatory variable in the form of liquidity commonality between stock markets. This thesis answers several questions: First, does a relationship between liquidity and the business cycle exist in the Eurozone? Second, does the commonality in liquidity between stock markets contain useful information about future economic growth? Finally, has the introduction of a common currency and interest rate policy changed liquidity?

This thesis is structured as follows: in Section II we review the relevant literature in different fields of academic research. In Section III we take a closer look at our data, provide sample statistics and show how the data is manipulated to conduct econometrically robust research in order to answer our research questions. In Section IV, the methodology is further explained and in Section V we show our empirical results. Here, we also employ several robustness checks and show whether the results hold under different specifications of our model. Section VI offers concluding remarks.

II. Literature review a. Liquidity

In the past decades, liquidity in financial markets has been a widely discussed topic. Liquidity in financial markets is defined as the degree to which an order can be executed within a short time frame at a price close to the security’s consensus value (Foucault, Pagano and Roell, 2013). This definition is sometimes extended to include ‘without causing large price movements’. Empirical studies show that liquidity is time-varying and persistent (Amihud, 2002; Chordia et al., 2000; Pastor and Stambaugh, 2003). In most models, liquidity consists of three different elements: tightness, depth and resilience. When a market is tight, the costs of trading small amounts is small (small bid-ask spread). When a market is deep, the costs of trading large amounts is small (large trades do not cause large price movements). Finally, a market is resilient if prices converge quickly to consensus value, i.e. deviations from true value are small and are corrected quickly.

A number of studies have introduced new measures of liquidity, using a range of variables and involving different frequencies of data observation. Tick-by-tick data most accurately models what participants in

9 the financial markets experience (for all three elements of liquidity described above). However, as this data is scarcely available to researchers there have been widely adopted proxies for liquidity used throughout the literature. Goyenko and Uhkov (2009) and Goyenko, Holden and Trczinka (2009) show that these measures of liquidity using low-frequency data hold up well empirically in explaining several elements of liquidity such as price impact and spread.

One of the most used measured is the illiquidity measure of Amihud (2002). In his influential paper he proposed a proxy for illiquidity based on stock returns, volume traded and number of observations in a particular time period. This measure is based on the framework of Kyle (1985), who proposed that the fact market makers are not able to distinguish whether order flow originated from an informed or a liquidity trader1 results in them setting prices as an increasing function of order imbalance. In turn, this creates a positive relationship between order flow and price change, more commonly known as price impact. This is precisely what this measure proxies: the absolute (percentage) price change per dollar of daily trading volume, or the daily impact of the order flow. Amihud shows that this illiquidity proxy is positively and strongly related to high-frequency measures of illiquidity.

The relationship between illiquidity and returns is positive, as proposed by Amihud and Mendelson (1986). In their paper the relationship between the bid-ask spread and returns on NYSE/AMEX stocks between 1960 and 1981 is researched, in which they suggest that the market-observed expected return is an increasing and concave function of the relative bid-ask spread. Their results are consistent with the model, indicating that investors require higher returns on high bid-ask spread securities. Brennan and Subrahmanyam (1996) measure illiquidity not by the bid-ask spread, but use the price response to order flow (size) and the fixed costs of trading as a measure of price impact. They find that these measures of illiquidity are positively correlated with stock returns.

A more in-depth research on several components of liquidity (risk) is done by Acharya and Pedersen (2005). In a liquidity-adjusted capital asset pricing model, they show that an asset’s required return is dependent on its own liquidity level as well as the covariances of its own return with the market return and liquidity. They model four different elements of liquidity risk: 1) return commonality, 2) illiquidity commonality, 3) return sensitivity to market illiquidity and 4) illiquidity sensitivity to market returns. They show that the required return of a security i is increasing in the covariance between its illiquidity and the market illiquidity. Their model provides a unified framework for understanding the various channels through which liquidity risk may affect asset prices. Even though their findings are insightful,

1

In this setting, a liquidity trader is a market participant that trades for any other reason than trading on having better/more information than other market participants.

10 the empirics are not fully conclusive. As they state themselves: “we must also entertain the possibility that not all of these risk factors are empirically relevant”.

Not only are there different elements to liquidity in financial markets, Brunnermeier and Pedersen (2008) postulate that there is a distinct difference between market liquidity and funding liquidity. The former is defined as the ease with which assets are traded, the latter as the ease at which a financial market participants can obtain funding for their activities. They show that under specific circumstances margins are destabilizing and market liquidity and funding liquidity are reinforcing, leading to liquidity spirals. Their model shows an explanation to several phenomena concerning liquidity: a) liquidity can suddenly dry up, b) there is commonality of liquidity across securities, c) liquidity is related to volatility, d) liquidity is subject to ‘flight to quality’ and e) an asset’s liquidity co-moves with the market.

Compared to many of the aforementioned liquidity measures/proxies, the illiquidity proxy of Amihud (2002) is very appealing as it does not rely on intraday observations of microstructure data. Its simplicity of interpretation, the availability of the required data and the fact it holds up very well empirically show why this measure is used frequently throughout the literature.

b. Liquidity commonality

Commonality in liquidity is defined as liquidity co-movements across assets or markets. Most empirical studies regarding liquidity commonality can be classified as either research on commonality between individual asset’s liquidity and market liquidity and liquidity commonality between markets.

Using several measures of liquidity (quoted spreads, quoted depth and effective spreads), Chordia, Roll and Subrahmanyam (1998) find that these measures significantly co-move with both market- and industry-wide liquidity. Controlling for well-known determinants of liquidity such as volatility, volume and price the results remain significant. Marshall, Nguyen and Visaltanachoti (2013) show that these general results also hold for the commodity futures markets. Between 1997 and 2003, when commodities prices were relatively stable, they show that there is a strong systematic liquidity factor in commodities. However, they find no evidence of commonality in liquidity between stocks and commodities.

Several studies have attempted to explain commonality in liquidity between different markets. Goyenko and Ukhov (2009) examine long-run commonality in liquidity between stock and bond markets. Their findings suggest the effect of stock illiquidity on bond illiquidity is consistent with flight-to- quality or flight-to-liquidity episodes. The evidence indicates that bond illiquidity acts as a channel through which monetary policy shocks are transferred into the stock market. These effects are observed across illiquidity of bonds of different maturities and are especially pronounced for illiquidity of short-term maturities. The

11 paper provides evidence of illiquidity integration between stock and bond markets. Mancini, Ranaldo and Wrampelmeyer (2013) study commonality in liquidity between several FX markets and find that commonality in liquidity is strong in FX markets, liquidity is priced and that they present evidence that their findings may be triggered by liquidity spirals (Brunnermeier and Pedersen, 2009).

c. Liquidity and the real economy business cycle

A number of researchers have attempted to explain the relationship between financial markets and the real economy. For this thesis in particular, we are interested in the relationship between financial market liquidity and the real economy business cycle. From both a theoretical and an empirical point of view the idea that market declines cause asset illiquidity has received attention (Christiano and Eichenbaum, 1995; Naes et al., 2011). They argue that spells of illiquidity occur because market participants engage in panic selling (demand effect), market makers withdraw supply in liquidity (supply effect) or a combination of both. Hameed, Kang and Vishwanathan (2010) show in their empirical research (using data on NYSE stocks between January 1988 and December 2003) that negative market returns decrease stock liquidity, especially in times of tightness in the funding market. Using the proportional bid-ask spread (bid-ask spread as a proportion of the price of an asset) as one of their key measures of liquidity, they find that changes in spreads are negatively related to market returns. Negative market returns reduce liquidity more when there are also large declines in the aggregate balance sheets of financial intermediaries or in the market value of the investment banking sector.

Not only is the business cycle in non-crisis times shown to be related to liquidity in financial markets. Rosch and Kaserer (2013) examine the dynamics and drivers of market liquidity during the most recent crisis (subprime mortgage and sovereign debt crisis). This paper sheds light on a different feature of market liquidity: liquidity commonality and flight-to-quality. Their research shows that liquidity commonality varies over time, peaks at major crisis events and becomes weaker the deeper we look into the order book. Their findings are also in line with recent theoretical models that argue for a spiral effect between funding liquidity and market liquidity (Brunnermeier and Pedersen, 2009): they find that funding liquidity tightness induces an increase in liquidity commonality which in turn leads to market-wide liquidity dry-ups.

Related to this thesis is Naes, Skjeltorp and Odegaard (2011). Using data for the United States over the period 1947 and 2008, they examine the relationship between stock market liquidity and the real economy business cycle. This provides us with insights into the dynamics of liquidity and business cycles. First, an in-sample test of predictive ability of market illiquidity is performed and they show that – using increasing numbers of controls in different specifications – market illiquidity is a consistent and

12 significant explanatory variable for several macro variables related to economic growth (GDP, unemployment, real investment and real consumption). Second, Granger causality tests show that causality runs from illiquidity to the business cycle indicating that liquidity can be a leading indicator for recessions). Third, an event study is performed around recessions to see how liquidity in the US stock market moves in relationship to recessions and to several key variables such as term spread and credit spread. Finally, they show that – as expected – the effect of illiquidity and the business cycle is larger for small firms than for large firms, highlighting a flight to quality effect.

III. Data a. Liquidity variables

In the relevant academic literature, several measures/proxies (discussed below) of liquidity have been proposed in order to avoid having to rely on detailed tick-by-tick data. Tick-by-tick data can most accurately model liquidity in terms of the three before discussed elements (tightness, depth, resilience), as it models precisely what financial market participants see and experience. However, as this data is not readily available researchers often use proxies to model liquidity. Widely used proxies are based on seminal work done by Amihud (2002), Roll (1984), Lesmond, Ogden and Trczinka (1999) and many others. In most cases, widely available data such as price, return and volume traded are used to create liquidity proxies. As Goyenko et al. (2009) show, the Amihud (2002) measure of illiquidity holds up well empirically while using relatively basic and readily available data. Therefore, we will focus on this proxy for modeling liquidity in the largest stock markets in Europe. As stated in the literature section, the Amihud (2002) measure reads as follows:

𝐼𝐿𝐿𝐼𝑄𝑖𝑦 =_𝐷1 𝑖𝑦∑ |𝑅𝑖𝑦𝑑| 𝑉𝑂𝐿𝐷𝑖𝑣𝑦𝑑 𝐷𝑖𝑦 𝑡=1 . (1)

For the sovereign bond market, we follow Bernoth et al. (2004) and Gravelle (1999) in their approach by using the supply of debt as a proxy for liquidity in the bond market for government securities. The correlation between bid-ask spreads and the supply of debt is significantly negative, suggesting that the volume of supply has a positive effect on liquidity (Gravelle, 1999). We will use ECB data on debt issued by central governments to model liquidity in the sovereign bond market, for which higher amounts of debt issued imply more liquidity in the market.

13 b. Liquidity data

To model liquidity for the European stock markets in our sample we use daily data on stock prices, returns and volume traded from COMPUSTAT from 1990 to 2015. We have used the largest index per country as a representative market for that country and used data from COMPUSTAT to include all stocks in each index at each point in time in our dataset. A list of the markets in our sample is given in Table 7. This allows us to calculate the Amihud (2002) measure per stock and then take the equally weighted average of all stocks in each market. High values on this measure indicate high market illiquidity, as the price impact of trades is high. Figure 5 shows the average value of the Amihud (2002) measure for each market over time between 1990 and 2015. It can be noted that not for every market data is available over the entire time period, restricting the number of observations and resulting in an unbalanced panel. For our tests we will exclude countries for which we have missing observations from 1996 to 2014. We will then check for robustness of our results by adding these countries to the sample later on.

Figure 5 Average illiquidity in stock markets 1996-2014. Illiquidity is measured by the Amihud (2002) measure as given in (1),

which measures the absolute (percentage) price change per dollar of daily trading volume, or the daily impact of the order flow. Higher values imply more illiquid markets while lower values depict times of liquid markets. Data provided by COMPUSTAT.

14 We use debt issued by central governments to proxy liquidity in the market for government securities (bonds, excluding stocks and derivatives). The government bond market is of particular interest, as within a monetary union governments are able to issue debt but do not have the possibility to monetize and inflate away excessive debts. It is also the market in which central banks (mainly on behalf of governments) perform their monetary policy operations, where they extract information on market movements and expectations, and where governments raise funds. Additionally, Gravelle (1999) argues the (almost) risk-free nature of government securities results in this market being a pricing benchmark for several other fixed-income securities and serve as collateral for various financial intermediaries, enabling them to finance their operations. In essence, a greater understanding of the market for government securities provides us with a better understanding of other fixed-income markets.

Figure 6 Debt issued (via notes, bills and bonds) by European central governments, 1996-2014. Debt issued measured in millions

15 Using data provided by the European Central Bank, we use monthly observations of debt issued by central governments to model liquidity in the bond market. Except for Greece, we have complete time series from 1990 to 2015 for all countries in our sample. Figure 6 depicts our liquidity measures for all countries in our sample. As we can see, governments issue debt on a regular interval making these time series rather cyclical and tests show evidence of a unit root. Therefore, we will take log differences of this variable in our regression specifications.

c. Macro data

Following Naes et al. (2011), we model the real economy business cycle using several macro variables. For each country in our sample, GDP and seasonally adjusted unemployment data are used to proxy the business cycle. Additionally, we use a number of financial variables that contain valuable information on the business cycle. As in Naes et al. (2011), these are excess market return, volatility, the credit spread and the term spread. The excess market return is calculated as the difference between return on the market index used and the 3 month interest rate on government bonds for each particular country. Volatility is the periodic average per country of the standard deviation of daily returns on each stock. Our two spread variables are calculated as the difference between the long term (10 year) and short term (3 month) interest rates (term spread) and the difference between the long term interest rate and investment grade corporate bonds (credit spread). Whereas Naes et al. (2011) use Moody’s Baa benchmark for determining the credit spread, we must rely on a different measure for the European corporate bond sector. Data availability taken into account, the most appropriate benchmark was the Bank of America Merill Lynch Euro Large Capital Investment Grade (in euro), hence the yield on that security will be used.

d. Time-series adjustments of series

Our sample covers a period from the start of 1996 up to and including 2014. In order to avoid running spurious regressions we check for stationarity by running unit root tests (panel unit root tests, Dickey-Fuller and Levin-Lin-Chu). We reject the null hypothesis of the existence of a unit root for the term spread, volatility and excess return. For both liquidity variables (Amihud (2002) and debt issued) we are not able to reject the null, which is also the care for the credit spread, GDP and unemployment. Following Naes et al. (2011) we take log differences of non-stationary time series. Additionally, we have excluded penny stocks (defined as a price below 5 euro), observations with daily volume traded below 1000 euros and have excluded stocks when there were missing observations in a particular month. All stock data was obtained via COMPUSTAT, bond and macro data was obtained via DataStream and the OECD. For an

16 overview of variables, definitions and data sources we refer to Table 8 in the Appendix. Table 1 shows contemporaneous correlations for the liquidity, market and macro variables.

Table 1 Correlation table of the macro, market and liquidity variables in our sample. Each variable used is the time-series

adjusted variable (i.e. Amihud is the log-differences version of the Amihud (2002) measure of illiquidity, for instance). T-statistics are displayed below each correlation coefficient between parentheses.

Liquidity

variables Market variables Macro variables

Amihud Debt issued Term spread Credit spread Volatility Excess return GDP growth UE growth Term spread 0.0475 0.0002 1 (0.00) (0.84) Credit spread 0.0004 0.0048 0.0055 1 (0.69) (0.00) (0.00) Volatility 0.0512 0.0031 0.0067 0.0000 1 (0.00) (0.00) (0.00) (0.98) Excess return -0.1629 -0.0038 -0.0306 -0.0003 -0.1622 1 (0.00) (0.00) (0.00) (0.76) (0.00) GDP growth -0.0477 -0.0052 -0.0389 -0.0026 -0.0571 0.3070 1 (0.00) (0.00) (0.00) (0.02) (0.00) (0.00) UE growth 0.0209 0.0015 0.0057 -0.0006 0.0030 -0.1103 -0.4083 1 (0.00) (0.17) (0.00) (0.61) (0.01) (0.00) (0.00)

Consistent with both literature and our expectations, the Amihud (2002) measure for illiquidity in the stock market is negatively correlated with dGDP and positively with dUE. This implies that high illiquidity correlates with a decrease in GDP and higher unemployment, consistent with the findings of Naes et al. (2011). For the stock market all correlations except Term are according to our expectations: illiquidity negative with excess return, positive with volatility and positive with dCredit. Thus, when market liquidity is low, volatility is high and returns are low: in line with both Hameed, Kang and Vishwanathan (2010) and Naes et al. (2011). However, while Amihud (2002) measures illiquidity, dDEBT should measure liquidity as governments issue more in liquid markets. If this is the case, we expect signs opposite to what we find for Amihud (2002), which is not the case. All signs are similar and - except for Term and dUE - significant. Most other signs in the correlation matrix are as expected: positive and significant for excess return and dGDP, and negative and significant for excess return and dUE. Finally, as expected the macro variables are significantly related to each other with the expected sign (negative).

17 IV. Methodology

a. Functional form

For our in-sample tests for liquidity and the business cycle we follow Naes et al. (2011) and add a dummy for the introduction of the Euro as well as an interaction term between our variable of interest (liquidity) and the Euro dummy. The regression models take on the following functional form:

𝐵𝐶_𝑖,𝑡 = 𝛼_𝑖+ 𝛽𝐵𝐶_{𝑖,𝑡−1}+ 𝛾𝐿𝑖𝑞_{𝑖,𝑡−1}+ 𝛿𝐹𝑖𝑛_{𝑖,𝑡−1}+ 𝜃𝐸𝑈𝑅𝑂_𝑡+ 𝜗𝐿𝑖𝑞_{𝑖,𝑡−1}× 𝐸𝑈𝑅𝑂_𝑡+ 𝜀_𝑖,𝑡, (2) where BC is a the business cycle variable of interest (either GDP or unemployment), Liq is market liquidity indicator for quarter t-1, Fin is a vector of control variables including excess return, volatility, the term spread and the credit spread, EURO is a dummy indicating the time period after the introduction of the euro and 𝜀 is an error term assumed to be i.i.d and mean zero. Since we are researching the explanatory power of liquidity in financial markets for the real economy business cycle, we use lagged values for our liquidity and control variables. To see whether the effect of liquidity on the business cycle has changed after the introduction of the Euro we introduce an interaction term. As in Naes et al. (2011), we also include the lag of the dependent variable.

The main focus will be a panel regression with country fixed effects. The coefficients are estimated by OLS. Additionally, robust standard errors are used to control for heteroscedasticity and autocorrelation. Robustness checks will be performed to check how sensitive our results are to using different specifications in Section 5d.

b. Liquidity commonality measurement

To measure commonality in liquidity between the stock markets of the different countries in our sample, we use principal component analysis (PCA). A natural method of extracting common information in different variables is using PCA, like Mancini, Ranaldo and Wrampelmeyer (2013) do for various FX markets. First, we extract seven principal components shown in Figure 8 in the Appendix. As we can see the first two are above 1, with the first being at approximately 3.5 explaining roughly half of the variation in the data. Following Mancini et al. (2013), the first principal component is used as a measure for liquidity commonality and will be added to our specifications. Like before, we will show different specifications of our base model.

To estimate liquidity commonality over time, we use our daily stock market data (before, we were limited to quarterly observations due to the observation frequency of macro data). We calculate average Amihud (2002) values for each index per day, which is subsequently used as input for PCA. Like before, we

18 assume the first principal component to measure commonality in liquidity between the markets in our sample. We take a 100 day window and roll this forward by 1 day for the complete sample giving us 6750 observations from April 1st 1996 to December 31st 2014. This results in a new time series that shows how much of the variation in market liquidity is explained by a common component assumed to measure liquidity commonality. The screeplot of the principal components and eigenvalues is shown in Figure 9 in the Appendix. With daily data, the first principal component explains approximately half of the variation in market liquidity levels for our sample.

c. Liquidity commonality over time

In order to find whether liquidity commonality has changed over time, we perform several tests. First, we report statistics on the explained variation of stock market liquidity by liquidity commonality for three different periods: 1996-2001, 2002-2007 and 2008-2014, using the quarterly data we constructed for the panel regressions of section 4a and b. Then, we follow up by using daily data and composing a new time series by rolling forward a 100 day window of principal component analysis. This new time series shows how much variation in stock market liquidity is explained by liquidity commonality, just like for the quarterly data. Finally, we check for what the introduction of the euro has attributed to that commonality, using the following functional form:

𝐿𝑖𝑞_𝑖 = 𝑎_𝑖+ 𝛽_𝑖𝐿𝑖𝑞𝐶 + 𝛾_𝑖𝐿𝑖𝑞𝐶 × 𝐷_{𝐸𝑢𝑟𝑜}+ 𝛿_𝑖𝐿𝑖𝑞𝐶 × 𝐷_{𝐿𝑒ℎ𝑚𝑎𝑛}+ 𝐷_{𝐸𝑢𝑟𝑜}+ 𝐷_{𝐿𝑒ℎ𝑚𝑎𝑛}+ 𝜀_𝑖 (3) where 𝐿𝑖𝑞_𝑖 is our stock market liquidity measure based on Amihud (2002), 𝐿𝑖𝑞𝐶 is the first principal component representing liquidity commonality and we add two interaction terms using two dummies: one to indicate the period after the introduction of the Euro (January 2002 onwards, 𝐷𝐸𝑢𝑟𝑜) and one to

indicate the financial crisis after the fall of Lehman (September 2008 onwards, 𝐷𝐿𝑒ℎ𝑚𝑎𝑛). The error term

is assumed to be i.i.d and mean zero. This model allows us to see whether liquidity commonality has changed significantly in terms of explaining stock market liquidity in two distinct periods.

V. Results

a. Stock and bond market liquidity and the business cycle

Table 2 summarizes the results from various regressions specifications for the stock market, each time including more controls. As we can see no specification shows evidence of a strong predictive ability of liquidity for GDP growth or growth in unemployment. Table 3 summarizes results from several specifications for liquidity in the markets for government securities as an explanatory variable for the real

19 economy business cycle. Like liquidity in the stock markets, we are unable to find conclusive evidence of strong predictive ability of liquidity with regards to the real economy business cycle.

Table 2 Results regressions of stock market liquidity and business cycle indicators. For both dependent variables –

unemployment growth and GDP growth – we have four specifications. Each specification adds more variables, starting with the term and credit spread, following up with volatility and excess return, and the final specification adds a dummy variable for the introduction of the Euro. T-statistics are reported below each coefficient between parentheses.

In both cases – liquidity in the stock market and liquidity in the sovereign bond market – we see no evidence such as in Naes et al. (2011), implying that there is no economically significant information about the real economy business cycle. However, in many specifications we find several other variables that do contain valuable information about future macroeconomic conditions. As Naes et al. (2011) state, the main purpose of adding additional variables is to check whether liquidity provides us with additional insight in future economic growth. First we add two non-equity controls: the term spread and the credit spread. The term spread is positively related to future economic growth and negatively related to future

GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth(t-1) 0.465*** 0.463*** 0.389*** 0.296** (5.86) (5.80) (4.49) (3.27) UE growth(t-1) 0.558** 0.557** 0.538** 0.483** (3.08) (3.07) (3.00) (2.62) dLiq(t-1) -0.045** 0.057 -0.045** 0.056 -0.018 -0.040 -0.028 -0.015 (-2.64) (0.61) (-2.64) (0.60) (-1.34) (-0.42) (-1.83) (-0.13) Term(t-1) 0.210*** -0.855*** 0.206*** -0.777*** 0.273*** -0.954*** (4.06) (-3.95) (4.86) (-4.02) (5.52) (-4.18) dCred(t-1) -0.036 -0.151 -0.034 -0.157 -0.065** -0.066 (-1.43) (-0.72) (-1.65) (-0.78) (-2.60) (-0.29) Vol(t-1) -0.030 0.042 -0.096* 0.236* (-1.01) (0.38) (-2.05) (2.18) Er(t-1) 0.016*** -0.054*** 0.015*** -0.049*** (5.41) (-7.74) (6.36) (-8.12) Euro 0.005** -0.015** (3.47) (-3.06) dLiq×Euro(t-1) 0.039 -0.102 (1.77) (-0.99) Constant 0.002*** -0.000 0.002*** -0.000 0.004** -0.003 0.004** -0.003 (6.67) (-0.58) (6.63) (-0.67) (3.53) (-1.24) (3.52) (-1.32) N 514 514 514 514 514 514 514 514 adj. R-sq 0.226 0.31 0.225 0.309 0.278 0.336 0.33 0.355

20 growth in unemployment. This relationship is not as expected, but this is not surprising as we already found earlier that the term spread correlates in opposite fashion to expectations. Interestingly, the credit spread is not found to be significant in any specification. Second, we add market financial market variables such as excess return and volatility, and a lag of the dependent variable of interest. Both excess return and the lag of the dependent variable are significantly related to the real economy business cycle as expected: higher excess returns on the market predict higher economic growth and lower unemployment, and there is a positive relationship between the dependent variable of interest and its one period lag. Finally, only when we include a dummy for the introduction of the euro is volatility a strong explanatory variable for both growth in GDP and unemployment.

b. Liquidity commonality between European stock markets and the business cycle

Table 4 summarizes results for liquidity commonality between stock markets as an explanatory variable for the real economy business cycle. Similar to our findings for the relationship between stock market liquidity and the business cycle, liquidity commonality contains valuable information in several specifications. However, in the specifications controlling for volatility, excess return and the introduction of the euro the coefficients of interest are not significantly different from zero. Similar to liquidity in the stock and bond market, we find no evidence of liquidity commonality to contain any valuable information on future unemployment growth. Summarizing, we do not find conclusive evidence in favor of liquidity commonality as an explanatory variable for the real economy business cycle.

c. Liquidity commonality over time

Part of this thesis is aimed at answering the question whether after the introduction of the Euro as common currency, and a common interest rate set by the ECB, liquidity commonality changed. Liquidity is in part driven by interest rates, because these directly influence the margins of market makers in financial markets. With a common monetary policy set by the central bank, we would therefore expect liquidity to move together in a stronger fashion than with every central bank pursuing its own monetary policy. To get a first impression of how strong liquidity commonality explains variation in market liquidity in the Eurozone, we identify three periods: before the introduction of the euro (April 1996 to December 2001), and two periods after the introduction of the Euro (convergence period from January 2002 to August 2008 and a divergence period during the financial crisis from September 2008 to December 2014). Table 5 summarizes how much variation is explained by liquidity commonality in each period, using our quarterly measures of market liquidity and running PCA on three subsamples. Factor loadings for the first principal component between stock market liquidity levels in Europe are given in Table 9 in the Appendix.

21

Table 3 Results regression bond market liquidity and business cycle. For both dependent variables – unemployment growth and

GDP growth – we have four specifications. Each specification adds more variables, starting with the term and credit spread, following up with volatility and excess return, and the final specification adds a dummy variable for the introduction of the Euro. T-statistics are reported below each coefficient between parentheses.

GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth(t-1) 0.471*** 0.469*** 0.391*** 0.298** (6.20) (6.14) (4.64) (3.35) UE growth(t-1) 0.558** 0.557** 0.539** 0.484** (3.08) (3.06) (3.00) (2.61) dLiq(t-1) 0.015 0.002 0.015 0.002 0.024 -0.037 0.026 -0.053 (0.33) (0.03) (0.33) (0.33) (0.46) (-0.44) (0.53) (-0.66) Term(t-1) 0.210*** -0.857*** 0.206*** -0.780*** 0.277*** -0.955*** (3.93) (-4.05) (4.77) (-4.09) (5.35) (-4.56) dCred(t-1) -0.034 -0.154 -0.035 -0.153 -0.065** -0.062 (-1.40) (-0.73) (-1.75) (-0.75) (-2.55) (-0.27) Vol(t-1) -0.031 0.041 -0.098* 0.237* (-0.99) (0.36) (-2.03) (2.12) Er(t-1) 0.017*** -0.053*** 0.016*** -0.049*** (5.11) (-11.54) (6.00) (-12.82) Euro 0.005** -0.015** (3.52) (-3.03) dLiq×Euro(t-1) 0.014 -0.121* (0.98) (-1.96) Constant 0.002*** -0.000 0.002*** -0.000 0.004** -0.003 0.004** -0.003 (6.88) (-1.72) (6.86) (-1.83) (3.43) (-1.08) (3.49) (-1.18) N 514 514 514 514 514 514 514 514 adj. R-sq 0.219 0.31 0.218 0.309 0.277 0.336 0.329 0.355

22

Table 4 Results regression stock market liquidity commonality and the real economy business cycle. For both dependent

variables – unemployment growth and GDP growth – we have four specifications. Each specification adds more variables, starting with the term and credit spread, following up with volatility and excess return, and the final specification adds a dummy variable for the introduction of the Euro. T-statistics are reported below each coefficient between parentheses.

Table 5 Explained variation in market liquidity by liquidity commonality. The quarterly data from the regression model in

Section 5a and 5b is used. The three periods indicate the base period before the introduction of the Euro, the initial period after introduction, and the financial crisis.

GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth(t-1) 0.459*** 0.457*** 0.389*** 0.298** (5.75) (5.69) (5.43) (3.30) UE growth(t-1) 0.557** 0.556** 0.538** 0.482** (3.07) (3.05) (2.99) (2.60) dLiq(t-1) 0.008 0.009 0.008 0.006 0.015 -0.021 0.006 0.022 (0.40) (0.07) (0.38) (0.04) (0.75) (-0.16) (0.26) (0.12) dLiqC(t-1) -0.070** 0.074 -0.070** 0.077 -0.045* -0.014 -0.038 -0.040 (-2.74) (1.02) (-2.69) (1.07) (-1.97) (-0.19) (-1.60) (-0.40) Term(t-1) 0.215*** -0.861*** 0.210*** -0.775*** 0.279*** -0.959*** (4.15) (-3.93) (4.75) (-3.94) (5.24) (-4.05) dCred(t-1) -0.030 -0.158 -0.031 -0.157 -0.063* -0.053 (-1.17) (-0.72) (-1.44) (-0.77) (-2.32) (-0.24) Vol(t-1) -0.029 0.031 -0.096* 0.235* (-0.92) (0.28) (-2.00) (2.11) Er(t-1) 0.015*** -0.054*** 0.015*** -0.049*** (5.73) (-8.19) (7.03) (-8.71) Euro 0.005** -0.015** (3.48) (-3.21) dLiq×Euro(t-1) 0.018 -0.099 (0.83) (-0.65) Constant 0.002*** -0.000 0.002*** -0.000 0.003** -0.003 0.004** -0.003 (6.55) (-0.78) (6.40) (-0.86) (3.33) (-1.13) (3.45) (-1.32) N 509 509 509 509 509 509 509 509 adj. R-sq 0.237 0.31 0.236 0.309 0.281 0.334 0.331 0.355

Fixed effects Yes Yes Yes Yes Yes Yes Yes Yes

Start End n Explained variation April 1996 December 2001 22 0.2823 January 2002 August 2008 26 0.5610 September 2008 December 2014 26 0.6332

23 We see that since the introduction of the Euro the explained variation has increased substantially from 28% to over 55% in the following period, in line with expectations. It is however surprising to see the explained variation increase further during the financial crisis. To give us a better insight into the dynamics of liquidity commonality, we use daily measures of liquidity. At a higher frequency we are better able to show how liquidity evolves over time by showing more detailed levels of liquidity and commonality in liquidity. Rolling forward our PCA test in a 100 day window, we create a new time series of the explained variation in market liquidity in our sample, which consists of 6750 observations.

Figure 7 Explained variation in market liquidity by liquidity commonality using daily observations for liquidity and liquidity

commonality. Liquidity is measured by the Amihud (2002) measure of illiquidity and liquidity commonality is the first principal component between all the stock market liquidity variables, assumed to represent liquidity commonality. The red line shows the trend of the new time series which depicts how much variance in liquidity is explained by liquidity commonality, created by applying a Hodrick-Prescott filter.

Our measure ranges from 0.1992 to 0.8632 with a standard deviation of 0.1388. Experiencing a low around 2000 after swinging up and down around the Russian default, we see the explained variation of market liquidity by liquidity commonality rise gradually over the following period. After the bankruptcy of Lehman Brother and the outbreak of the financial crisis, liquidity commonality explaining market liquidity levels experiences a downward trend. This graph implies that indeed liquidity commonality has increased during the period where GDP, unemployment and interest rates (on government securities) converged, while apparently dropping slightly during the financial crisis.

24 In Table 6 we show the results for the regression model in (3), which is run for each country separately. For each specification we see the first interaction term is significant, indicating that after the introduction of the Euro liquidity commonality has changed significantly in explaining market liquidity. The same holds for the second interaction term, indicating that this has changed again during the financial crisis. Note however, that the signs of the coefficients change. This indicates that the effect is different per country. These results confirm what we already show in Figure 7, where we see different trends around the fall of Lehman Brothers in September 2008.

Table 6 Regression results liquidity on liquidity commonality, using daily observations between April 1st 1996 and December

31st 2014. The first seven specifications show regressions per country, the last specification is a panel regression for all countries in our sample using fixed effects. Liquidity commonality is measured as the first principal component between stock market liquidity variables in our sample. All liquidity measures are transformed by taking log-differences. T-statistics are reported below each coefficient between parentheses.

d. Robustness checks

In order to see whether our results depend on various details in our specification, we run a number of robustness checks. First, we include the three excluded countries (Austria, Greece and Italy). Second, we will add a financial crisis dummy in order to control for the two major financial crashes in our sample: the Russian default of 1998 and the subprime mortgage/sovereign debt crisis of 2008. The results for the regression specifications are shown in Tables 10 to 12.

LiqNL LiqBE LiqFR LiqGE LiqES LiqFI LiqPO Liq

LiqC -0.174* -1.487*** 1.313*** -4.791*** 1.236*** -0.378*** 10.38*** 0.872 (-1.69) (-13.50) (14.04) (-57.21) (13.55) (-4.70) (101.29) (0.49) LiqC×D2002 1.050*** -1.086*** -0.232*** 3.915*** 1.836*** -3.389*** -4.790*** -0.385 (14.07) (-13.61) (-3.43) (64.50) (27.76) (-58.21) (-64.51) (-0.34) LiqC×D2008 -0.812*** 3.179*** -0.901*** 0.861*** -3.328*** 4.723*** -5.546*** -0.261 (-8.90) (32.60) (-10.88) (11.60) (-41.18) (66.40) (-61.12) (-0.19) D2002 0.005 0.005 0.001 -0.013 0.007 -0.001 -0.012 -0.001 (0.35) (0.36) (0.10) (-1.16) (0.62) (-0.09) (-0.94) (-0.33) D2008 0.008 0.001 0.000 -0.008 0.002 0.000 -0.007 -0.001 (0.59) (0.09) (-0.04) (-0.68) (0.13) (-0.01) (-0.50) (-0.30) Constant -0.0258*** 0.0235*** -0.00278 -0.0528*** -0.0427*** 0.0679*** 0.0509*** 0.003 (-4.83) (4.11) (-0.57) (-12.15) (-9.01) (16.27) (9.56) (0.17) N 4548 4548 4548 4548 4548 4548 4548 31836 adj. R-sq 0.619 0.558 0.681 0.782 0.699 0.769 0.747 0.472

25 Our robustness checks show that in many specifications our results hold. For both stocks and bonds, market liquidity does not contain valuable information about the real economy business cycle. Including a dummy to control for financial crises does not change out results: the coefficients are still not significant. Excluding countries one by one from our sample does not change this, as we do not find a significant coefficient for liquidity in any subsample. We are therefore confident in rejecting the hypothesis that market liquidity in stock and bonds markets contains economically significant information regarding future macroeconomic conditions in the Eurozone.

In regards to liquidity commonality, these results are in line with most specifications of our base model: in many specifications liquidity commonality between stock markets in the Eurozone is a good explanatory variable for growth in GDP. Including a dummy to control for financial crises however has an influence: liquidity commonality does not contain valuable information when we control for financial crises. These results confirm our earlier statements about liquidity commonality and its relationship to the real economy business cycle, however we do find some conflicting evidence.

VI. Conclusion

This thesis aims at answering two questions. First, does liquidity (commonality) in the Eurozone stock and bond market contain valuable information for future macroeconomic conditions? Second, has liquidity commonality since the introduction of the Euro increased among Eurozone stock markets? In answering these questions, this thesis adds to several strands of the finance literature, specifically for literature with a focus on liquidity (commonality), and literature with a focus on the relationship between financial markets and the real economy.

In contrast to earlier work by Naes et al. (2011), we find no evidence of an economically meaningful relationship between liquidity and the real economy business cycle. For both the stock market and market for government securities, we are not able to establish that such a relationship exists for the Eurozone between 1996 and 2014. Controlling for non-equity and financial market variables – term spread, credit spread, excess return and volatility, as well as including a time dummy for the introduction of the euro – no liquidity variable contains economically significant information that can be used to explain business cycle movements (measured in terms of GDP and unemployment). Similar to our findings for both the stock and bond markets, we do not find conclusive evidence that liquidity commonality between stock markets contains valuable information for the business cycle. In a model equal to that of Naes et al.

26 (2011) we find a significant relationship, however when controlling for the time after introduction of the Euro this is nog longer significant at 5%.

We have also looked at liquidity commonality in stock market liquidity at a higher frequency. A rolling principal component analysis shows that after the introduction of the Euro two distinct periods had different trends in liquidity commonality. First, a period of convergence in GDP growth, unemployment and interest rates in government securities in which commonality in liquidity experienced a strong upward trend. Second, the financial crisis after the fall of Lehman Brothers has shown liquidity commonality to experience a slight downward trend. However, this is not conclusive evidence that the Euro has led to a significant change in liquidity commonality. Therefore, we run regressions of stock market liquidity on liquidity commonality and interaction terms of liquidity commonality and time dummies. Our results show that the estimated coefficients are significant, thus liquidity commonality has changed significantly in the two identified time periods.

There are several limitations to this thesis that need to be addressed. First, like Naes et al. (2011) we are limited largely by the fact that macroeconomic data is published on a quarterly basis. This means that we are forced to average our daily liquidity measures and potentially exclude interesting up and down spikes in market liquidity. Second, our measure for liquidity in the market for government securities (debt issued by a central government) is not ideal as it does not use data directly from financial markets such as prices, volume and spreads. Additionally the correlations with other variables are not as expected for this measure of liquidity. Third, by not having access to tick-by-tick data we are not able to research specific components of liquidity (tightness, depth and resilience) and show how these develop over time and how they are related to the real economy business cycle. Last but not least, although using principal component analysis to extract a common component that represents liquidity commonality is used more often in the literature (Mancini, Ranaldo and Wrampelmeyer, 2013), the coefficients derived from PCA are less straightforward to interpret. Hence, it could be interesting to consider the Karolyi, Lee and Van Dijk (2012) R²-measure of liquidity commonality, which has a more clear-cut interpretation. Future research might be able to address these issues and expand on our research.

27 References

Acharya, V. V., & Pedersen, L. H. (2005). Asset pricing with liquidity risk. Journal of Financial

Economics, 77(2), 375-410.

Adrian, T., & Shin, H. S. (2009). Money, liquidity, and monetary policy. FRB of New York Staff Report, (360).

Amihud, Y. (2002). Illiquidity and stock returns: cross-section and time-series effects. Journal of

financial markets, 5(1), 31-56.

Amihud, Y., & Mendelson, H. (1986). Asset pricing and the bid-ask spread. Journal of financial

Economics, 17(2), 223-249.

Artis, M. J., & Zhang, W. (1999). Further evidence on the international business cycle and the

ERM: is there a European business cycle?. Oxford Economic Papers, 51(1), 120-132.

from the euro-area bond market. Review of Financial Studies, 22(3), 925-957.

Bernanke, B. S., & Gertler, M. (1995). Inside the black box: the credit channel of monetary policy

transmission (No. w5146). National bureau of economic research.

Bernoth, K., Von Hagen, J., & Schuknecht, L. (2004). Sovereign risk premia in the European government bond market.

Bhide, A. (1993). The hidden costs of stock market liquidity. Journal of financial economics, 34(1), 31-51.

Brennan, M. J., & Subrahmanyam, A. (1996). Market microstructure and asset pricing: On the compensation for illiquidity in stock returns. Journal of financial economics, 41(3), 441-464. Brunnermeier, M. K. (2008). Deciphering the liquidity and credit crunch 2007-08 (No. w14612) National

Bureau of Economic Research.

Brunnermeier, M. K., & Pedersen, L. H. (2009). Market liquidity and funding liquidity. Review of

Financial studies, 22(6), 2201-2238.

Carpenter, S., & Demiralp, S. (2006). The liquidity effect in the federal funds market: evidence from daily open market operations. Journal of Money, Credit and Banking, 901-920.

Chalmers, J. M., & Kadlec, G. B. (1998). An empirical examination of the amortized spread. Journal of

Financial Economics, 48(2), 159-188.

Chordia, T., Roll, R., & Subrahmanyam, A. (2000). Commonality in liquidity. Journal of Financial

Economics, 56(1), 3-28.

Chordia, T., Roll, R., & Subrahmanyam, A. (2001). Market liquidity and trading activity. Journal of

28 Christiano, L. J., & Eichenbaum, M. (1995). Liquidity effects, monetary policy, and the business cycle.

Journal of Money, Credit and Banking, 1113-1136.

Foucault, T., Pagano, M., & Röell, A. (2013). Market Liquidity: Theory, Evidence, and Policy. Oxford University Press.

Gennotte, G., & Leland, H. (1990). Market liquidity, hedging, and crashes. The American Economic

Review, 999-1021.

Goyenko, R. Y., Holden, C. W., & Trzcinka, C. A. (2009). Do liquidity measures measure liquidity?.

Journal of financial Economics, 92(2), 153-181.

Goyenko, R. Y., & Ukhov, A. D. (2009). Stock and bond market liquidity: A long-run empirical analysis.

Journal of Financial and Quantitative Analysis, 44(01), 189-212.

Gravelle, T. (1999). Liquidity of the government of Canada securities market: stylized facts and some

market microstructure comparisons to the United States treasury market. Bank of Canada.

De Haan, J., Inklaar, R., & Jong‐A‐Pin, R. (2008). Will business cycles in the euro area

converge? A critical survey of empirical research. Journal of economic surveys, 22(2),

234-273.

Hameed, A., Kang, W., & Viswanathan, S. (2010). Stock market declines and liquidity. The Journal of

Finance, 65(1), 257-293.

Hellwig, M. (1994). Liquidity provision, banking, and the allocation of interest rate risk. European

Economic Review, 38(7), 1363-1389.

Jankowitsch, R., Mösenbacher, H., & Pichler, S. (2006). Measuring the liquidity impact on EMU government bond prices. The European Journal of Finance, 12(2), 153-169.

Kyle, A. S. (1985). Continuous auctions and insider trading. Econometrica: Journal of the Econometric

Society, 1315-1335.

Lesmond, D. A., Ogden, J. P., & Trzcinka, C. A. (1999). A new estimate of transaction costs. Review of

Financial Studies, 12(5), 1113-1141.

Lucas, R. E. (1990). Why doesn't capital flow from rich to poor countries?. The American Economic

Review, 92-96.

Mancini, L., Ranaldo, A., & Wrampelmeyer, J. (2013). Liquidity in the foreign exchange market: Measurement, commonality, and risk premiums. The Journal of Finance, 68(5), 1805-1841. Manganelli, S., & Wolswijk, G. (2009). What drives spreads in the euro area government bond market?.

Economic Policy, 24(58), 191-240.

Marshall, B. R., Nguyen, N. H., & Visaltanachoti, N. (2013). Liquidity commonality in commodities.

29

Massmann, M., & Mitchell, J. (2003). Reconsidering the evidence: are Eurozone business cycles

converging (No. B 05-2003). ZEI working paper.

Muranaga, J., & Shimizu, T. (1999). Market microstructure and market liquidity (Vol. 99). Institute for Monetary and Economic Studies, Bank of Japan.

Næs, R., Skjeltorp, J. A., & Ødegaard, B. A. (2011). Stock market liquidity and the business cycle. The

Journal of Finance, 66(1), 139-176.

Pastor, L., & Stambaugh, R. F. (2001). Liquidity risk and expected stock returns (No. w8462). National Bureau of Economic Research.

Rösch, C. G., & Kaserer, C. (2014). Reprint of: Market liquidity in the financial crisis: The role of liquidity commonality and flight-to-quality. Journal of Banking & Finance, 45, 152-170.

30 Appendix

Table 7 Countries and markets used in sample. For each country the main/largest stock index was used. Data provided by

COMPUSTAT, index names as given by COMPUSTAT.

Country Index

Austria ATX Index

The Netherlands Amsterdam AEX - Index

Greece Athens Stock Exchange General Index

Belgium Belgium 20 Index

France CAC 40 Index

Germany Deutscher Aktienindex (DAX) Index

Spain IBEX 35 Index

Italy Milan MIB 30 Index

Finland OMX Helsinki 25 Index

Portugal PSI 20 Index

31

Table 8 Data source and definitions

Variable Source Description

dAmihud Compustat As in Amihud (2002), period average (daily obs) of absolute value returns over volume traded. Log differences.

dDebt ECB Log difference of debt issued by central government

Term DataStream Difference between 10y yield and 3m yield on government securities

Credit spread

DataStream Difference between 10y yield and investment grade yield. Log differences.

Volatility Compustat As in Naes et al. (2011), volatility of daily prices of stocks in sample

Excess return

Compustat Difference between stock return and market return per period

dGDP OECD Log difference of nominal GDP

dUE OECD Log difference of unemployment, seasonally adjusted

Euro - Dummy indicating period after introduction euro (January 2002 onwards)

Figure 8 Screeplot after principal component analysis using quarterly observations, eigenvalue per principal component.

32

Figure 9 Screeplot after principal component analysis using daily observations, eigenvalue per principal component. Horizontal

line at eigenvalue of 1, below this line components do not significantly explain variation in the data.

Table 9 Factor loadings for the first principal component between stock market liquidity levels in Europe, where liquidity is

measured using the Amihud (2002) measure of illiquidity.

Variable Component 1 LiqNL 0.4036 LiqBE 0.4567 LiqFR 0.4830 LiqGE 0.4092 LiqES 0.1987 LiqFI 0.3454 LiqPO 0.2626

33

Table 10 Results regressions of stock market liquidity and business cycle indicators using 10 countries in sample, now including

Austria, Greece and Italy. For both dependent variables – unemployment growth and GDP growth – we have four specifications. Each specification adds more variables, starting with the term and credit spread, following up with volatility and excess return, and the final specification adds a dummy variable for the introduction of the Euro. T-statistics are reported below each coefficient between parentheses. GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth(t-1) 0.351*** 0.350*** 0.192* 0.101 (3.60) (3.78) (2.21) (1.25) UE growth(t-1) 0.422** 0.419** 0.392** 0.344* (2.53) (2.49) (2.44) (2.11) dLiq(t-1) -0.046 0.047 -0.046 0.047 -0.017 -0.044 -0.020 -0.038 (-1.39) (0.54) (-1.40) (0.53) (-0.59) (-0.58) (-0.86) (-0.49) Term(t-1) 0.095 -1.382 -0.497 0.420 -0.409 0.120 (0.14) (-0.65) (-0.62) (0.19) (-0.55) (0.05) dCredit(t-1) -0.001 -0.016 -0.003 -0.011 -0.008 0.004 (-0.12) (-0.46) (-0.45) (-0.32) (-1.34) (0.13) Vola(t-1) -0.161*** 0.196 -0.285*** 0.516** (-4.48) (1.61) (-4.78) (2.62) er(t-1) 0.022*** -0.062*** 0.021*** -0.054*** (4.80) (-3.95) (4.87) (-4.11) Euro 0.007*** -0.018*** (4.84) (-3.84) Constant 0.002*** 0.000 0.002*** 0.000 0.008*** -0.007** 0.010*** -0.011** (5.55) (1.14) (6.11) (0.90) (0.89) (-2.60) (8.30) (-2.53) N 399 399 399 399 399 399 399 399 adj. R-sq 0.131 0.169 0.131 0.17 0.264 0.212 0.326 0.236

34

Table 11 Results regression bond market liquidity and business cycle using 10 countries in sample, now including Austria,

Greece and Italy. For both dependent variables – unemployment growth and GDP growth – we have four specifications. Each specification adds more variables, starting with the term and credit spread, following up with volatility and excess return, and the final specification adds a dummy variable for the introduction of the Euro. T-statistics are reported below each coefficient between parentheses. GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth UE growth GDP growth(t-1) 0.422*** 0.400*** 0.395*** -0.28 (8.46) (7.70) (8.17) (7.64) UE growth(t-1) 0.565*** 0.562*** 0.557*** 0.531*** (4.61) (4.37) (4.51) (4.21) dLiq(t-1) -0.035 0.067 -0.031 0.057 -0.025* 0.015 -0.021 -0.029 (-1.86) (0.28) (-1.77) (0.24) (-1.94) (0.06) (-1.55) (-0.12) Term(t-1) -0.139*** 0.337 -0.087*** 0.053 -0.082*** -0.012 (-6.87) (0.94) (-9.33) (0.15) (-6.94) (-0.03) dCredit(t-1) 0.194*** -0.547 0.153*** -0.325 0.152*** -0.277 (5.32) (-0.66) (4.70) (-0.38) (4.97) (-0.34) Vola(t-1) 0.015* 0.103 -0.035** 0.516* (1.98) (0.42) (-3.24) (1.86) er(t-1) 0.022*** -0.118** 0.020*** -0.107* (9.40) (-2.34) (8.63) (-2.15) Euro 0.004*** -0.033*** (6.40) (-4.47) Constant 0.002*** 0.003*** 0.002*** 0.002*** 0.003*** -0.003 0.003*** -0.003 (11.31) (5.22) (10.53) (3.53) (7.14) (-0.62) (8.34) (-0.60) N 702 699 702 699 702 699 702 699 adj. R-sq 0.180 0.328 0.2 0.327 0.294 0.34 0.318 0.349