What Drives the Stock Market Reaction to ECB Monetary Policy?
Reinder K. Haitsma
a,†a University of Groningen A R T I C L E I N F O A B S T R A C T
Article History: In this paper the transmission channels underlying the stock market reaction to both the conventional and unconventional monetary policy of the European Central Bank (ECB) are studied. First, we observe that stocks react to monetary policy surprises in a heterogeneous manner. Second, we find that there is no clear evidence for the presence of an interest-rate channel. Third, there is mild evidence for a credit channel before the recent crisis. During the crisis, we find convincing evidence for the presence of a credit channel for unconventional monetary policy surprises. In particular, stocks of firms that are either highly leveraged or have fewer means to service their debt respond stronger to policy surprises. Finally, we present some evidence that value and past loser stocks show a larger reaction to monetary policy surprises. First version: 7 April 2015
Second version: 22 May 2015 Final version: 16 June 2015 JEL classification: E44 E52 G14 Keywords: Monetary Policy Stock Market ECB Transmission Channels 1 Introduction
Do European stocks react to monetary policy of the ECB? If so, what drives the stock market reaction? These are important questions for both policy makers as well as investors. Based on the transmission mechanism of monetary policy, policy makers might want to revise their plans and investors might want to adjust their portfolios in anticipation of changes in monetary policy. Over the past two decades the subject of stock market reactions to monetary policy has gained more and more interest in
This paper is submitted as a master thesis for the programmes Economics (EBM877A20) and Finance (EBM866B20). Supervisor: prof. dr. De Haan.
†
2
research. Nevertheless, most of this research (e.g. Bernanke and Kuttner, 2005; Ehrmann and Fratzscher, 2004) has focused on the stock market reactions to monetary policy in the United States. Especially the evidence as to what underlies the relationship between policy rates and stock returns is centred on the United States and the federal funds rate. It therefore appears that an investigation of this subject for the euro area and the ECB may be a fruitful venture. This paper aims to identify whether a transmission channel is present between ECB monetary policy and European stocks, and moreover, which stock characteristics influence such a transmission channel. We study these relationships both for the conventional monetary policy measures as well as the unconventional monetary policy instruments used in recent years.
1.1 What Might Influence Stock Reactions?
First, we outline why stock returns might respond to monetary policy and identify which characteristics may influence this relation. Bernanke and Kuttner (2005) identify three ways in which monetary policy might influence stock returns. First, it might influence the discount factor that is used to calculate the present value of the cash flows of a firm, thereby influencing its valuation. Second, increased interest rates might signal larger payments to debt holders thereby decreasing the flows to equity holders, for instance in the form of dividends. Finally, the preferences of investors might alter due to monetary policy changes. If interest rates increase, investors might find it more profitable to switch to fixed income instruments in lieu of stocks, thereby reducing the demand for stocks and their respective prices. Although there are several ways in which monetary policy might influence stock returns, the overall pattern that emerges is clear. An increase in interest rates set by the central bank is expected to decrease stocks prices. However, especially in times of crisis, the response might be different. Hosono and Isobe (2014) note that a decrease in the policy rate might actually signal to investors that economic conditions are worsening. If this is the case, the mechanism might work the other way around, i.e., a loosening of monetary policy might decrease stock returns.
3
negatively affected. The credit channel is often studied by investigating how underlying debt-characteristics of certain stocks or stock portfolios influence the relation with monetary policy.
Having presented the two channels at work, we are able to identify the factors that might influence them. First turning to the interest-rate channel, Peersman and Smets (2005) argue that the effect differs per industry. The demand of some industries is highly dependent on interest rates, whereas the demand of others is not. They therefore try to gauge the heterogeneity in responses to monetary policy between industries. A factor that might also be relevant in this respect is the durability of the products. It is often hypothesised that the consumption of durable goods is more dependent on interest rates, since the purchase of such a good is to some extent similar to an investment (see e.g. Hamburger, 1967). For the credit channel several papers turn to measures of financial structure, since such measures may identify the dependence on bank funds. Peersman and Smets (2005) research several variables for European industries that may influence financial structure. First, the size of a firm may matter, as Peersman and Smets argue that larger firms often have better access to financial markets. Second, financial leverage is included, to proxy the reliance on debt. Third, the maturity structure of debt is considered, since firms with short-term debt might also be more dependent on monetary policy. Fourth, the financing need for working capital can also be regarded to measure reliance on short-term debt. Finally, they include a coverage ratio of cash flows over interest payments. This ratio measures the extent to which firms are able to meet short-term interest payments. Thus all these measures try to gauge to what extent an industry is dependent on bank funds.
In addition to the factors outlined above, there are some other factors that may influence the relation between stock returns and monetary policy. For instance, Kontonikas and Kostakis (2013) investigate factors from the Carhart four-factor model (1997), next to measures of size. That is, they form portfolios on the basis of market-to-book ratios, earnings-to-price ratios, and momentum. It is argued that value stocks are more sensitive to rising interest rates than growth stocks, since value stocks rely on high cash flows relative to their stock price. Momentum is included as past performance might reflect the stock’s sensitivity to overreactions. In other words, the worst performing stocks may be more prone to the anxiety of investors. In addition, stocks that perform worse might find it more difficult to obtain financing. Hence past losers are expected to react stronger to monetary policy surprises than past winners.
1.2 Contribution
4
The contribution of this paper to the existing stock of research is twofold. First, to our knowledge, since the implementation of the ECB, there have not been any papers truly investigating the credit channel of monetary policy transmission with respect to stocks in the euro area. Even though research on the United States is ample, we believe you cannot extrapolate these results to Europe. Abundant research has shown that the relationship between monetary policy surprises and stocks may differ across countries and central banks (see e.g. Sondermann et al., 2009). We therefore aim to fill this gap and provide evidence on European monetary policy and European stocks. Second, we will investigate the influence of the crisis and the unconventional monetary policies. Even though there has been quite some research on unconventional policies conducted by the ECB (and other central banks), we do not know of research investigating to what extent sectors or underlying stock characteristics affect the relationship. We therefore aim to add a comprehensive review of the influence of ECB monetary policy on stock markets to the literature. We find that it is difficult to discern a pattern in the stock market reaction to pre-crisis ECB monetary policy. Although we find significant responses for some portfolios, most of these parameters do not fit in the frameworks outlined above. We find that there is no clear evidence for the presence of an interest-rate channel. There is only some mild evidence for the presence of a credit channel and a larger response of loser stocks. In contrast, we find that during the crisis the credit channel is clearly at work for unconventional monetary policy announcements. That is, firms that are reliant on debt show a stronger response. A clear interest-rate channel remains absent. In addition, it appears that value and loser stocks respond stronger to unconventional monetary policy surprises than growth and winner stocks respectively. Before we continue, a short outline of the monetary policy conducted by the ECB is provided in the next section.
1.3 ECB Monetary Policy
Since the enactment of the European Monetary Union in 1999, the ECB has had the primary objective of promoting price stability. In order to do so, the ECB has control over the main interest rates in the euro area. These are the interest rates on the main refinancing operations, deposit facilities, and the
marginal lending facilities. Using these instruments the ECB can steer interest rates in the euro area
5
in unconventional monetary policy; both to keep inflation at a reasonable level and to stimulate the economies of its members. The ECB engaged in several actions to achieve these goals. Among those actions were long-term refinancing operations, a securities market program, and outright monetary
transactions. We will not go into the exact details of these programs, since we do not make a
distinction between them in our empirical analysis. However, it should be clear that the conduct of monetary policy changed drastically. It has to be noted that the ECB engaged in both unconventional monetary policy as well as conventional monetary policy during the crisis. As Cour-Thimann and Winkler (2013) point out, the unconventional measures of the ECB are aimed at accommodating the transmission of regular monetary policy in lieu of being a purely monetary stimulus. The two types of policy should therefore be regarded as complements that consequently coexist. This stands in contrast to the unconventional measures conducted by most other central banks.
The remainder of this paper is structured as follows. In section 2 we will review the literature that has been published on the subject thus far. A discussion of our methodology will be provided in section 3. This is followed by an outline of our data in section 4. In section 5, the results of our research are discussed. The paper is concluded in section 6.
2 Literature Review
In this section we will discuss the existing literature on stock market reactions to monetary policy. As discussed in the introduction, most literature on stock market reactions to monetary policy is based on the monetary policy of the Federal Reserve (FED). Hence the literature review is also biased towards the United States. Nevertheless, we aim to incorporate studies analysing the euro area as well. Although there are several other channels through which monetary policy can be transmitted (e.g. bond yields, and output), we mostly focus on research that gauges stock markets reactions, as this is most closely related to our research. We will first discuss some papers that deal with stock market reactions to monetary policy in general. Then we will go into the two transmission channels identified in the introduction. We end the section with a discussion of papers that deal with the crisis.
2.1 General Findings
6
and expected policy changes and find that an unexpected tightening in monetary policy has a significant negative effect on stock prices, whereas expected changes do not yield significant results.
Even though research has been done on the subject before the 1990s, from the 1990s onwards interest in the relationship increased. A substantial fraction of these papers focuses on the impact of monetary policy on broad stock indices. For instance, Patelis (1997) researches the relation between excess stock returns of the NYSE index and the federal funds rate. He finds that tightening monetary policy will lead to lower excess returns in the short-run, however the opposite effect is found when looking at the long-run. Patelis then turns to a variance decomposition to identify the channels through which these effects operate and finds that the influence of unexpected monetary policy on excess returns is mainly driven by shocks in the dividend yield. Galí and Gambetti (2014) confirm the former findings by Patelis (1997) using a vector autoregressive model for the S&P500. Kuttner (2001) takes a novel approach in disaggregating the overall change in the federal funds rate in an expected and unexpected part using futures data. As this approach will also be applied in this paper, a further discussion is provided in section 3.1. In applying this method, Kuttner finds that the interest rates of bills and bonds of different maturities respond statistically and economically significantly to monetary policy surprises, whereas responses to expected changes were only small. Kuttner does not investigate stock market reactions. A paper that applies the same definition of monetary policy surprises, but does examine stock market reactions is by Chuliá et al. (2009). In their research, Chuliá et al. distinguish between positive and negative surprises and find heterogeneous responses. They show that the average response for 94 stocks of the S&P100 index is larger for positive surprises, that is, a surprise tightening of monetary policy. For negative surprises, the occurrence of the news seems to be the most relevant factor, whereas the magnitude of the loosening of the federal funds rate is found to be of little impact on the stock market response. Rigobon and Sack (2004) also investigate stock market reactions and apply yet a different technique to gauge surprises based on the heteroskedasticity of policy shocks. They find that stock prices of the S&P500, Dow Jones Industrial Average, and NASDAQ decrease following a tightening of monetary policy.
7
line with Bohl et al. (2008); positive changes in policy rates lead to significant negative effects on stock prices. In addition, Filbien and Labondance (2013) investigate the learning of stock market participants in the euro area. They find that learning has occurred and ECB monetary policy has therefore become less surprising over time. Another quite similar paper is by Hussain (2011), who examines returns and volatilities of both EU and US market indices using intraday data. He finds that for both areas there is a significant effect of monetary policy surprises on returns and volatility. In addition, Hussain shows that the press conferences held on the days of ECB monetary policy announcements influence volatilities. It can therefore be concluded that the ECB also releases information during such press conferences. Andersson (2010) examines intraday data for both the US and EU markets and finds that volatility goes up for both markets following monetary policy announcements. However, it appears that the response for the US is larger than for the EU. The author provides three possible explanations for this. First, as the FED also releases forward-looking statements in their policy announcements, it might be that these announcements therefore simply contain more information. Second, it can be argued that price stability is a higher priority for the ECB than for the FED. This anchors expectations and decreases surprises. Finally, there is some uncertainty about the precise time of the FED announcements, whereas the timing of the ECB announcements is fixed. Furthermore, Ioannidis and Kontonikas (2008) research thirteen OECD countries, including nine euro members: France, Germany, Italy, Belgium, the Netherlands, Finland, Sweden, Spain and Switzerland. For most of the stock markets in these countries it is also found that tightening monetary policy has a significant negative effect on stock returns. However, the sample is from 1972-2002, hence only a small part of the period that the ECB conducted monetary policy in the euro area is included.
2.2 The Interest-Rate Channel
8
findings of Jensen et al. (1997). A much cited article that also goes into differences between industries is by Bernanke and Kuttner (2005). Using Fama-French industry portfolios, they show that for all industries unanticipated changes in the federal funds rate lead to a decrease in excess industry returns. However, Bernanke and Kuttner also find quite some variation in the magnitude of the responses. In particular, the sectors for high-tech, telecom and durable goods respond quite strongly, whereas energy, utilities and nondurables only show a mild reaction. Chuliá et al. (2010) investigate industry responses and confirm that the utilities sector responds only weakly. In addition, the financial sector is also found to respond quite strongly. Ehrmann and Fratzscher (2004) find a significant intermediate response of the financial sector. They confirm the findings of Bernanke and Kuttner (2005) for the other sectors. Basistha and Kurov (2008) confirm much of the findings discussed above by investigating FED policy, individual stocks, and macroeconomic conditions in the period 1990 to 2004. Again more interest rate-sensitive sectors (e.g. durables) show a larger response to monetary policy than sectors that are known for lower levels of sensitivity (e.g. non-durables and utilities).
9
and Ehrmann (2003) also investigate several European industries by use of the EURO STOXX industry indices. Seven out of ten industry stock returns turn out to have a significant negative relation with ECB monetary policy. Angeloni and Ehrmann can confirm the results for the telecom industry found by Kholodilin et al. (2009), but they also find a significant relation for the technology and utility industries. Thus, even though telecom is often found to be the most responsive sector, which sectors respond to ECB monetary policy changes and which do not is still not completely clear.
2.3 The Credit Channel
10
There is far less evidence on the credit channel for the European Union. One of the papers that examines the European market is by Peersman and Smets (2005). However, they look at output of industries in lieu of stock market returns. They do this for the period of 1980 to 1999 and find that a tightening of monetary policy leads to a contraction in output. As is shown in Perez-Quiros and Timermann (2000), Peersman and Smets (2005) establish that the effects are larger in times of recession as compared to times of growth. After having identified the heterogeneous effects of monetary policy on industries, as outlined above, Peersman and Smets try to explain why industries react differently by an examination of industry characteristics (other than durability). They confirm the existence of the credit channel, since firm size and financial structure are shown to matter for the transmission. To be more precise, smaller firms and those with poorer financial structures are affected most by monetary policy. Dedola and Lippi (2005) confirm the findings on size and financial structure of Peersman and Smets (2005). Sondermann et al. (2009) investigate stock markets in Europe and their response to ECB monetary policy. They find that the stock market indices of the several European countries react differently to the monetary policy changes. Thereafter, they identify that the stock markets with the highest stock market capitalisations are the ones that react significantly to monetary policy surprises. This finding runs counter to the common assumption that small cap stocks react the most to policy surprises.
2.4 The Crisis
In the final section of the literature review, we will go into the crisis. The crisis has spurred new research, since, as discussed in section 1.3, unconventional measures were taken by central banks during the crisis. Since we also include the crisis and the unconventional policy measures in our analysis, some background on the findings with regard to such measures is provided.
As discussed above, from 2007 onwards, the policy rates of the ECB started to decline towards zero. To increase inflation in the euro area the ECB normally decreases the interest rate. However, at the zero lower bound this is no longer possible. Consequently, the ECB had to apply other measures to conduct monetary policy. In order to do so unconventional measures were taken. Examples of these measures are long-term refinancing operations and the securities market program. It is interesting to see how these unconventional policies influenced markets. Not surprisingly, several papers have looked at these measures and estimated their influence on stock markets and other economic indicators.
11
Ciccarelli et al. (2013) do extensive research on the transmission of ECB monetary policy on output. They are one of the few that look at the influence of bank and firm size in the euro area on the transmission during the crisis. Ciccarelli et al. find that the transmission mechanism has changed. For the size effects, it is concluded that unconventional measures taken by the ECB did not help in fostering credit availability for small firms, even though it did for small banks, which are argued to be the credit providers for small firms. It therefore appears that a size effect is at play during the financial crisis. Ricci (2014) also looks at European banks during the crisis. Using an event-study approach, Ricci concludes that the stock prices of banks are more responsive to unconventional measures than to regular policy changes by the ECB. This is in line with the findings on equity markets of Fiordelisi et al. (2014) for the euro area, US, UK, and Switzerland. We would like to note that this could be attributed to the unconventional nature of monetary policy, but also to the bearish features of stock markets during the crisis; e.g. Chen (2007) shows that in the US monetary policy has larger effects on stocks in bear markets. In addition, Ricci (2014) shows that banks with weak balance sheets and/or high-risk operations are more responsive to policy announcements. In contrast, Lambert and Ueda (2014) find that unconventional monetary policy surprises affect the stock valuation of banks in the euro area in the opposite way. An easing of monetary policy negatively affects stock returns. However, additional analysis shows insignificant results. In their extensive research on the effects of ECB unconventional monetary policy, Fratzscher et al. (2014) find that the policies have relieved tensions on European equity markets, since the announcements led to a positive response in stock prices. A quite drastic change in the stock response is found by Wang and Mayes (2012), who report that instead of the negative response to a surprise policy rate increase before the crisis, during the crisis stock markets responded positively to such changes, especially when interest rates were close to the zero lower bound. However, they do not look at unconventional measures. Rogers et al. (2014) find that the announcements of unconventional monetary policy of the ECB led to positive stock reactions during the crisis thereby easing financial conditions. In contrast, Hosono and Isobe (2014) conclude that eurozone stock markets react negatively to a surprise loosening of ECB unconventional monetary policy. The latter provide as possible explanation for this that the policies might be too expansionary, thereby leading markets to believe that the economy is in a worse state than expected before the announcement. This finding and explanation is confirmed by Kontonikas et al. (2013) for funds rate cuts in the US. On the other hand, whereas Hayo and Niehof (2011) find a significant effect of ECB monetary policy on European equity markets, the relationship does not change significantly during the crisis.
2.5 Summary
12
drive such a relation. However, the study of the credit channel for ECB monetary policy and the heterogeneity in stock market reactions during the crisis has not received a lot of attention. We therefore aim to fill this gap. As our literature review is quite extensive, we aim to both summarize the topics covered by the discussed papers on ECB monetary policy and clearly show what we add to the literature in Table 1. An “x” indicates the topic is covered in the respective paper. The stock market index column indicates whether the paper went into the effects of ECB monetary policy on a broad stock market index (or several indices). Whereas the interest-rate channel, credit channel, and momentum columns reflect to what extent characteristics influencing the studied relationship are covered. The final column shows whether or not the recent crisis is included. Table 1 clearly shows that the papers included do not perform such an extensive investigation as we do. Almost all papers cover two out of five topics. In particular, the credit channel has received very limited attention for ECB monetary policy. Moreover, the papers that go into this subject are either on output (Dedola and
Table 1
A comparison of previous papers on the effects of ECB monetary policy to ours.
Market Index Interest-Rate Channel Credit Channel Value vs.
Growth Momentum Crisis
Angeloni and Ehrmann (2003) x x
Bohl et al. (2008) x
Bredin et al. (2007) x x
Dedola and Lippi (2005) x x
Filbien and Labondance (2013) x
Fiordelisi et al. (2014) x x
Fratzscher et al. (2014) x x
Hayo and Niehof (2011) x x
Hosono and Isobe (2014) x x
Hussain (2011) x
Jardet and Monks (2014) x x
Kholodilin et al. (2009) x x
Peersmans and Smets (2005) x x
Ricci (2014) x
Rogers et al. (2014) x x
Sondermann et al. (2009) x x
Wang and Mayes (2012) x x
Haitsma (2015) x x x x x x
13
Lippi, 2005; Peersman and Smets, 2005) or only briefly look at size (Sondermann et al., 2009). We would like to note that Ciccarelli et al. (2013) investigate the credit channel, however, their research does not have a link with stock markets, hence we omit the paper from the table. To our knowledge, there are no papers investigating the influence of value and growth stocks, and momentum in Europe up until now. Finally, in combination with the crisis the research has been quite limited. The eight papers on ECB monetary policy in the crisis either only deal with broad stock market indices or only with the banking sector and its characteristics (Fiordelisi et al., 2014; Ricci, 2014; Lambert and Ueda, 2014). Fratzscher et al. (2014) investigate transmissions channels during the crisis, but these are different than ours.
3 Methodology
Having presented our research aims and an overview of the literature, we are ready to discuss the methodology of our paper. Since we discern between expected and surprise changes in monetary policy we will first discuss the estimation of the two. After this, we will present the regression model. We end the section with a brief overview of the hypotheses.
3.1 Surprises
14
, (1)
where Δrt u
represents the policy surprise at day t, m represents the number of days in the month, and
represents the discrepancy between the futures spot rate at day t and the prevailing rate at the day before the announcement, t-1. We use continuous three-month Euribor futures rates, since it is shown by Bernoth and Von Hagen (2004) to be a reliable predictor for the policy rates of the ECB. However, since we will utilize such futures rates, the first caveat will be less problematic and we will therefore follow Bredin et al. (2007) and simply use the discrepancy between the daily futures rates. Equation (1) then becomes:
. (2)
The futures rates are calculated by subtracting the daily settlement price from 100, which provides us with an implied expectation for the policy rate. The surprise factor therefore reflects how much investors alter their beliefs about the policy rate after the announcement as reflected in the three-month Euribor futures rates. A positive surprise then implies that the announced policy rate was higher than expected, or in other words that monetary policy is tighter than expected, and vice versa. The expected part of the policy change (Δrt
e
) can be represented by the difference between the actual rate change (Δrt) and the unexpected part calculated above:
15
changes in German bund futures prices to approximate surprises. However, as Rogers et al. (2014) note, the monetary policy of the ECB during the crisis often only led to a decrease in interest rates for troubled countries in lieu of lowering interest rates for more stable countries like Germany. That is, the spreads between the troubled and stable countries often decreased, but German rates may have remained the same or even increased. Using only German rates would then bias surprise estimations, hence we opt for the approach of Rogers et al. (2014).1 The surprise factor for the unconventional measures during the crisis, Δrtu,c, can then be represented as follows:
, (4) where and are the Italian and German government bond yields at day t respectively.
3.2 Regression Model
To study the relationship between stock portfolios and monetary policy surprises, we adopt a simple econometric model as in Kuttner (2001), Ehrmann and Fratzscher (2004), and Bernanke and Kuttner (2005). Although there are several methods used in the literature to estimate the effects of monetary policy on stock returns (e.g. a VAR or event study approach), it appears that most methods yield quite similar results. We therefore opt to use the most straightforward (and popular) approach: a regression analysis. As with the surprise factor, the investigation of the crisis requires some adjustments to the model. To be precise, we want to model the effects of the surprises pre-crisis as well as during the crisis. Finally, we also want to be able to perform a Wald test on the found parameters for conventional monetary policy surprises to test whether the relation has changed significantly due to the crisis. The regression model can be represented as follows:
, (5)
where Rt i
represents the returns on day t of a certain stock (portfolio) i, α is a constant, Ct is a dummy
that takes a value of zero pre-crisis and one thereafter, Δrt u
, Δrt e
, and Δrt u,c
are respectively the conventional monetary policy surprise, the expected policy rate change, and the unconventional monetary policy surprise on day t, Xt is a vector of control variables on day t, and εt is the error term on
day t. β1 represents the effects of the monetary policy surprise on stock returns pre-crisis, whereas β2
shows the effects after the start of the crisis. Since the unconventional monetary policy surprise is set to zero pre-crisis it does not warrant multiplication with the crisis dummy. Even though the efficient market hypothesis would suggest that the expected change in the policy rate should not lead to a stock
1
16
market response, we do control for any possible response to expected changes. Since it is (for now) impossible to devise an expected change measure for the unconventional monetary policy announcements, we cannot control for this. The vector of control variables includes two variables. First, it includes the MSCI World Index excluding Europe to control for general economic movements in the rest of the world. Second, we also include the crisis dummy separately for reasons of consistency. Whereas Ehrmann and Fratzcher (2004) then continue to estimate a panel model in which time-varying, firm-specific characteristics can be included as interaction with the monetary policy surprise variable, we will use portfolios based on sectors and firm-specific characteristics to estimate the heterogeneity in responses (see e.g. Guo, 2004; Kontonikas and Kostakis, 2013), thereby altering the dependent variables in lieu of the independent variables. The former approach would require data on the firm-specific characteristics for every point of observation, which is not feasible for the stocks under our consideration. In adopting the latter method, the above model suffices for all regressions.
3.3 Hypotheses
We end the section with a short discussion of our hypotheses. First of all, we believe that pre-crisis the conventional policy surprise will yield less significant results compared to the unconventional surprises during the crisis. The ECB has a very clear two per cent inflation target, it will therefore not surprise the market very often in normal times. However, during the crisis more surprising policy measures were taken. We therefore expect stocks to be more responsive to such measures. It has to be noted that we expect contractionary monetary policy to reduce stock returns in normal times. This means that we expect negative parameter estimates for the monetary policy surprises. As explained in section 1.1, during the crisis the sign might be reversed, since a loosening of monetary policy might signal worsening economic conditions. The signs of both the conventional surprise during the crisis and the unconventional surprise in general are therefore difficult to predict due to these opposing forces. For the interest-rate channel, we expect the different sectors to react in a heterogeneous fashion, with some yielding significant and others insignificant coefficients. Moreover, there may be differences in the magnitude of the significant parameters. The sectors most reliant on interest rates are expected to yield the highest parameter estimates. Third, we hypothesise that a credit channel is present, which implies that the portfolios that are most reliant on debt in lieu of other means of financing (e.g. a portfolio with a high debt-to-equity ratio) will be affected more by monetary policy surprises than those that do not rely on such financing. Finally, based on Kontonikas and Kostakis (2013), we expect both value stocks and past loser stocks to be more responsive to monetary policy.
4 Data
17
announced on the 22nd of August 2007 and the period of crisis the time thereafter. Often the fall of Lehman Brothers on the 15th of September 2008 is considered to be the starting point of the recent financial crisis. However, we believe that the start of the unconventional measures is an appropriate starting date for Europe, as such measures will only be used in times of trouble. In addition, it allows us to investigate the effects of the coexistence of conventional and unconventional policies more accurately.2 We are aware that the financial crisis had already ended before the 27th of February 2015. Nevertheless, the ECB still conducted unconventional monetary policy at the end of the sample, which might indicate that the Euro Crisis had not yet ended, we therefore define the whole second period to be the crisis period. Although the usage of intraday data has gained interest in recent research (e.g. Andersson, 2010; Hussain, 2011), we do not possess such data. However, our time window of one day does not seem to be too large and is therefore not very likely to be influenced by other information released on the days of announcement. Especially for the unconventional measures, the announcements are likely to be the most important news events of the day. In addition, daily time windows will still capture the relevant effects when markets need some time to digest the announcements, which might especially be relevant for the unconventional monetary policy announcements. Since we also control for movements in the world by including a broad market index, the MSCI World Ex Europe index, we believe to get consistent results. The data for this index is obtained from Datastream. The remainder of the data used to perform the research as outlined in section 3 is gathered from several sources. We will outline the data requirements, the sources and, if necessary, the construction per variable below. We end the section with a short discussion of the descriptive statistics.
4.1 Policy Surprises
To estimate policy surprises we first need to identify the timing of monetary policy announcements. The ECB has monthly press conferences in which the monetary policy is outlined. The website of the ECB lists the dates of such announcements.3 We use these dates for the conventional monetary policy decisions. It has to be noted that in the early years of the European Monetary Union there were sometimes several monetary policy press releases in one month. The unconventional measures taken by the ECB in recent years did not always correspond to the regular announcement dates and are therefore partly extracted from other sources. Specifically, we use the dates provided by Rogers et al. (2014) for the period up to April 2014 and the database of press release of the ECB up to and including February 2015.4 We would like to stress that we include all the press releases, even when there was no change in monetary policy. If pre-announcement the market suspects that there might be a change in the policy rates of the ECB, the market will reflect these expectations. When the press
2
A robustness check using the fall of Lehman Brothers as the starting point of the crisis period is presented in section 5.5.5.
3
See: http://www.ecb.europa.eu/press/govcdec/mopo/previous/html/index.en.html.
4
18
release is presented and it appears that there is no change in policy rates, this comes as a surprise to the market, consequently moving the market. It is therefore also important to include such announcements. We end up with a sample that consists of 226 conventional monetary policy announcements and 23 unconventional announcements of which seventeen were made during the regular press releases on monetary policy decisions. Of the unconventional measures, two occurred in the weekend.5 We could use the Monday after to proxy the surprise for these monetary policy announcements, but the measure might then include a substantial amount of noise over the weekend. We therefore decide to exclude these announcements from our sample and end up with a final sample of 21 unconventional policy measures.
Secondly, the data for Euribor futures, and German and Italian bonds are needed to estimate the surprises on the announcement dates. Datastream provides the relevant data for the continuous three-month Euribor futures, which is available from 1998 onwards. Since Datastream does not provide bond yields, the 10-year bond yields for German and Italy are obtained from investing.com. The surprise component is then calculated for every announcement date as described in section 3. The expected changes are calculated using Equation (3). We follow Wang and Mayes (2012) and calculate the actual policy rate change as the published Euribor rate at day t+1 minus the Euribor rate at day t, thereby controlling for the time of publication. Euribor rates are extracted from the EMMI website.
4.2 The Interest-Rate Channel
Having estimated the surprises, it is required to gather data on stock returns. All returns are calculated as:
, (6)
where pti is the closing price of stock or index i on day t. Our first regressions will have a broad stock
market index as the dependent variable to get a sense of the general effects. For this purpose, we obtain the daily closing prices of the EURO STOXX 50 index from Datastream for the sample period. The EURO STOXX indices only include stocks from twelve eurozone members6, which is preferred since results will then not be influenced by the admission of other members to the monetary union during the sample period. After this, we want to gauge the stock market response per sector. Datastream provides the relevant sector indices that cover the stocks of euro area countries only. To cover the most appropriate array of industries for our purposes, we use the nineteen ‘supersector’
5
Namely, the security markets programme announcement on the 9th of May 2010 and the acknowledgement of the security markets programme on the 7th of August 2011.
6
19
indices, as defined by The International Classification Benchmark. Table 2 provides an overview of the included sectors.
Table 2
The sectors that are studied for the interest-rate channel.
Automobiles & Parts Media
Banks Oil & Gas
Basic Resources Personal & Household Goods
Chemicals Real Estate
Construction & Materials Retail
Financial Services Technology
Food & Beverage Telecommunications
Health Care Travel & Leisure
Industrial Goods & Services Utilities
Insurance
4.3 The Credit Channel
The credit channel is studied using certain firm characteristics. First, we want to look at the influence of size on the studied relation. We therefore collect size portfolios for the eurozone from Datastream, namely EURO STOXX Large, Mid, and Small. We then look at other factors that might influence the reliance on debt. However, it is quite hard to find the required characteristics for European stocks, hence we will only focus on the 50 stocks of the EURO STOXX 50. The individual stock prices are retrieved from Datastream. Only 44 stocks cover the whole sample period, hence we drop six stocks.7 Morningstar provides the relevant statistics on the firm characteristics and we retrieve the following ratios for our sample of stocks over the period 2004 to 2013: the interest coverage ratio, the free cash flow to income ratio, the current ratio, the financial leverage ratio, and the debt-to-equity ratio. All of these ratios either give a sense of how leveraged a firm is or they reflect the ability of the firm to pay interest. The financial ratio is defined as assets over equity, hence the higher the ratio, the higher the dependence on leverage. We take the median of the measures over the sample period and sort the stocks on their score, we thereafter divide the stocks into three groups: high, mid, and low. Finally, we create the portfolio returns by taking the average of the daily returns of the sorted groups of stocks. It has to be noted that for both the coverage and current ratio there is no data on the ten financial stocks. For the free cash flow ratio there are four missing observations and for the debt-to-equity ratio one observation lacks any statistics. In addition, we create portfolios for the free cash flow to income ratio, the financial leverage ratio, and the debt-to-equity ratio excluding financials, since the banks included in the EURO STOXX 50 often score highly similar on the chosen metrics thereby often constituting
7
20
the lion’s share of one of the sorted portfolios. It is therefore important to also investigate the influence excluding such stocks, as sectoral heterogeneity might otherwise drive the results.
4.4 Value versus Growth and Momentum
Finally, we explore two other stock characteristics which might influence monetary policy transmission, namely value versus growth stocks and momentum. A value stock can be defined as a stock with a relatively low price when taking its fundamentals into account. It can therefore be characterised by a low market-to-book and price-to-earnings ratio. The opposite holds for a growth stock. Data on the market-to-book and price-to-earnings ratios for the period 2004 to 2013 are retrieved from Morningstar. We again create a high, mid, and low portfolio for each ratio. These portfolios are created in a similar way to the credit channel portfolios. As explained in section 4.3, we estimate the portfolios both with and without financials. For the momentum factor we can simply use the returns on the 44 stocks and sort them based on past performance. We use three different time-spans to determine the performance of the stocks. In particular, we look at the returns in the previous month, three months, and twelve months and then sort the stocks on their relative performance. Finally dividing them up in two equal groups of past winners and losers. The portfolios are updated daily, hence they constantly reflect the best or worst performing stocks of the past period. Two notes are in order. First, we use data from January 1998 onwards for the construction of the momentum portfolios, which allows us to construct a yearly momentum portfolio from January 1999 onwards. The required data are obtained from Datastream. Second, we drop the stock of Daimler from the six- and twelve-month momentum portfolios, since it only has data from November 1998 onwards.
4.5 Descriptive Statistics
Due to the large number of equity returns studied, the descriptive statistics of the data are presented in appendix A. The descriptive statistics of the conventional surprise show that on average the surprise is zero, thus positive and negative surprises average out over the sample. The unconventional surprise has a negative mean, which shows that most of the announcements loosened monetary conditions, which may be represented by a decreasing spread. Three extreme observations are removed from the automobiles and parts sector returns, which appeared to be a problem with the data. The remainder of the stock return data does not show any serious outliers. It be can seen that the stock portfolios with higher leverage and lower free cash flows have higher standard errors. These portfolios therefore appear more risky, as theory would predict.
5 Results
21
standard errors. The Durbin-Watson statistics do not point towards the presence of autocorrelation, we therefore do not need to apply Newey-West standard errors. In addition, Table 8 in appendix B shows no signs of multicollinearity of the independent variables. In what follows, we do not report the estimated parameters for the control variables. In general, the MSCI World Ex Europe index showed highly significant positive parameter values, whereas the crisis dummy was mostly found to be insignificant. In addition to our regression results, we report Wald test t-statistics, which tests the null hypothesis of equal coefficients for the pre-crisis conventional monetary policy surprise and the conventional monetary policy surprise during the crisis.
5.1 General Results
As a start, we will discuss the results for the broad EURO STOXX 50 index and for 44 of the 50 separate stocks included in the index. The results for the broad index are presented in Table 3. It can be observed that pre-crisis there appears to be a weakly significant influence of monetary policy surprises on the broad stock returns of the EURO STOXX 50. The negative sign of the surprise variable is as expected, indicating that a surprise loosening of monetary policy leads to an increase in
Table 3
Regression output for the EURO STOXX 50. The regression model is given by Equation (5). The MSCI World Ex Europe index and the crisis dummy are included as control variables. All regressions are estimated with White's consistent standard errors. Number of observations: 4215. The t-statistics are provided in parentheses. * Denotes significance at the 10%, ** denotes significance at the 5% level, and *** denotes significance at the 1% level.
EURO STOXX 50
Conventional Surprise Pre-Crisis -0.071*
(-1.913)
Expected Change Pre-Crisis -0.146***
(-2.690)
Surprise Conventional Crisis 0.071
(0.880)
Expected Change Crisis 0.042
(0.441) Surprise Unconventional -0.069*** (-2.961) Constant 0.000 (0.195) R2 0.283
Wald Test t-Statistic -1.560
22
stock returns in regular times. To be precise, a 0.25% surprise cut in the policy rate is expected to lead to an increase in the index returns of almost 1.78%, which is therefore also economically significant. The expected change shows to have a highly significant negative influence on the stock index, which contradicts our expectations. If we look at the results during the crisis, we see that the conventional surprise and the expected change do no longer have significant parameters. It is, however, interesting to see that the parameter values turn positive. On the other hand, the null hypothesis that the pre-crisis parameter is equal to the crisis parameter cannot be rejected. We find a highly significant, negative influence of the unconventional monetary policy surprise. This indicates that an announcement that leads to a decrease in the yield spread between German and Italian government bonds leads to an increase in the returns of the index. Since we cannot use the most common 0.25% policy rate change here to give an impression of the economic significance, we use the average spread change as presented in appendix A. A monetary policy announcement that caused a decrease in the German-Italian yield spread of 0.06% on average led to an increase in EURO STOXX 50 returns of almost 0.5%. Announcements that relieve tensions for (certain) governments thus also benefit the stock market.
Before we investigate the heterogeneous impact of monetary policy on a large range of different stock portfolios, we look at the distribution of the estimated surprise parameters for 44 of the 50 EURO STOXX stocks to get a first idea about the heterogeneous response of different stocks. Figure 1 reports the histogram for the conventional policy surprise, pre-crisis. The graph shows that stocks are clearly influenced to a different extent. The parameter estimates range from negative values to positive values. It has to be noted that most of the positive parameter estimates lack significance. In addition, another
Figure 1
Distribution of the parameters for the pre-crisis conventional monetary policy surprises for 44 stocks.
23
confirmation of the heterogeneous response of different stocks is the fact that 25 out of 44 parameter estimates are significant. The stocks with the highest significant estimates are Vivendi and Deutsche Telekom. Both companies can be affiliated with the telecom sector, which was often found to be the most responsive sector to monetary policy.
Figure 2 reports the histogram for the conventional policy surprise during the crisis. It can be seen in Figure 2 that there are almost no negative coefficients in the sample anymore, whereas these constituted the lion’s share of the significant pre-crisis parameter estimates. This finding could be seen as preliminary evidence that the crisis has changed the relationship between stock returns and (conventional) monetary policy. It has to be noted that only two of the estimated coefficients are significant. The two significant results can be found for the stock returns of Danone and Unibail, it appears difficult to find a reasonable explanation for these significant results.
Figure 2
Distribution of the parameters for the conventional monetary policy surprises for 44 stocks during the crisis.
Finally, we look at the distribution of the parameter estimates for the unconventional monetary policy surprises in Figure 3. For the unconventional surprises, the coefficients are almost all negative and 31 out of 44 estimations are significant. It therefore appears that this surprise yields the most significant results across the board. The spread is still quite large, with the monetary policy surprise having a twenty times larger influence on the stocks with the smallest (negative) estimates as compared to those with the largest estimated parameter values. The largest significant coefficients are exclusively found
24
for banks, a sector heavily affected by the crisis.8 It therefore seems that this sector is most responsive to unconventional monetary policy surprises.
Figure 3
Distribution of the parameters for the unconventional monetary policy surprises for 44 stocks.
5.2 The Interest-Rate Channel
Having estimated the general effects and having presented some preliminary evidence for a heterogeneous response of individual stocks, as well as a different reaction before and after the start of the crisis, we are in a position to present the results for the interest-rate channel. That is, the estimates for different sector portfolios. The results are presented in Table 4. The pre-crisis conventional surprise variable shows to be significant in ten out of nineteen cases. All significant parameters have the expected negative sign, indicating that surprise monetary policy tightening leads to lower stock returns for those sectors. For instance, stock returns in the automobiles & parts sector are expected to decrease by 1.7% when the ECB announces a surprise policy rate increase of 0.25%. The largest parameter estimate is found for telecommunications, which corresponds to our findings in section 5.1 and earlier research (see e.g. Kholodilin et al., 2009; Angeloni and Ehrmann, 2003). A surprise policy rate cut of 0.25% by the ECB is expected to lead to an increase in telecom stock returns of 2%. It appears that these effects are therefore also economically significant. When looking at the characteristics of the sectors that are influenced, we would expect sectors that rely most on interest rates (e.g. durables and banks) to show the largest responses. It appears that this expectation cannot be confirmed by the data. Although a durable sector like automobiles & parts shows a significant result, as well as the banking sector, other sectors with significant results cannot be classified as durable or
8
To be precise, the five highest parameter estimates are found for: Banca Intesa, Banco Bilbao Vizcaya Argentaria, BNP Paribas, Société Générale, and Unicredit, which are all banks.
25 Table 4
Regression output for the interest-rate channel. The regression model is given by Equation (5). The MSCI World Ex Europe index and the crisis dummy are included as control variables. All regressions are estimated with White's consistent standard errors. Number of observations per regression: 4215, except automobiles & parts: 4212. The t-statistics are provided in parentheses. * Denotes significance at the 10%, ** denotes significance at the 5% level, and *** denotes significance at the 1% level.
Conventional Surprise Expected Change Conventional Surprise Crisis Expected Change Crisis Unconventional Surprise Constant R 2 Wald Test t-Statistic
Automobiles & Parts -0.068*** -0.078*** 0.099 -0.040 -0.073*** 0.000 0.079 -1.568
(-3.331) (-2.454) (0.946) (-0.354) (-3.081) (0.689) Banks -0.043** -0.062*** 0.170 0.131 -0.146*** 0.000 0.071 -1.737* (-2.056) (-2.764) (1.408) (0.892) (-3.029) (0.904) Basic Resources 0.013 0.040 0.322 0.183 -0.089*** 0.001** 0.058 -1.405 (0.477) (1.144) (1.478) (0.723) (-3.387) (2.437) Chemicals -0.039* -0.027 0.233* 0.131 -0.075*** 0.000 0.063 -2.198** (-1.928) (-0.958) (-1.908) (0.917) (-2.942) (1.385)
Construction & Materials -0.031** 0.004 0.228* 0.135 -0.095*** 0.000** 0.050 -1.913*
(-1.991) (0.266) (1.697) (0.894) (-3.186) (2.110)
Financial Services -0.007 0.004 0.150* 0.092 -0.044** 0.000** 0.020 -1.709*
(-0.541) (0.252) (1.651) (0.934) (-2.200) (2.105)
Food & Beverage -0.036* -0.033* 0.141* 0.039 -0.060** 0.000 0.017 -2.046**
(-1.696) (-1.787) (1.679) (0.400) (-1.992) (1.116)
Health Care -0.044*** -0.039** 0.141** 0.071 -0.043 0.000 0.031 -2.634***
(-2.574) (-1.962) (2.072) (0.977) (-1.394) (0.857)
Industrial Goods & Services -0.028 -0.061** 0.191 0.115 -0.057** 0.000 0.089 -1.823*
(-1.378) (-2.169) (1.613) (0.891) (-2.338) (1.372)
26 Table 4 - Continued Conventional Surprise Expected Change Conventional Surprise Crisis Expected Change Crisis Unconventional Surprise Constant R 2 Wald Test t-Statistic Insurance -0.069*** -0.128*** 0.190 0.114 -0.103*** -0.000 0.084 -2.032** (-2.678) (-3.200) (1.522) (0.731) (-2.692) (-0.026) Media -0.042 -0.074** 0.172* 0.116 -0.047 0.000 0.056 -2.052** (-1.522) (-2.098) (1.712) (1.057) (-1.469) (0.127)
Oil & Gas -0.028 -0.050 0.232* 0.146 -0.102*** 0.000 0.050 -2.060**
(-1.048) (-1.502) (1.880) (1.024) (-3.154) (1.332)
Personal & Household Goods -0.044*** -0.046** 0.194** 0.115 -0.053* 0.000 0.053 -2.588***
(-2.724) (-2.247) (2.141) (1.139) (-1.663) (1.149) Real Estate 0.001 0.025** 0.165** 0.089 -0.079*** 0.000** 0.012 -2.390** (0.048) (1.983) (2.435) (1.165) (-2.972) (2.363) Retail -0.032* -0.035 0.181** 0.129 -0.067** 0.000 0.037 -2.496** (-1.864) (-1.371) (2.167) (1.430) (-2.482) (0.506) Technology -0.026 -0.124* 0.137 0.050 -0.037 0.000 0.127 -1.108 (-0.473) (-1.833) (1.005) (0.340) (-1.305) (0.060) Telecommunications -0.080** -0.136*** 0.147* 0.118 -0.073** -0.000 0.059 -2.491** (-2.570) (-3.313) (1.714) (1.237) (-2.409) (-0.047)
Travel & Leisure -0.023 -0.040* 0.158 0.099 -0.059** 0.000 0.052 -1.758*
(-1.265) (-1.879) (1.561) (0.790) (-2.315) (0.761)
Utilities -0.025* -0.015 0.238** 0.187* -0.081** 0.000 0.030 -2.491**
(-1.770) (-0.964) (2.278) (1.750) (-2.472) (1.430)
27
sensitive to interest rates, like food & beverage. The parameter estimates for expected changes are significant in over half of the cases.
After the results for the pre-crisis monetary policy surprise, we look at the results for the crisis. The estimated parameters for the conventional surprise variable are now positive, as was the case for the EURO STOXX 50 index. However, whereas the latter lacked significance, eleven out of nineteen coefficients are significant now, albeit only at 5% and 10% levels. Although telecommunications still yields a significant parameter estimate, the largest influence can now be observed in the utilities sector. The effects in this sector (and the others) are also highly economically significant, with a 0.25% policy rate cut leading to an expected decrease in chemical sector stock returns of over 4%. A link between significance of the sector and its durability again appears to be absent. Looking at the Wald test, which indicates to what extent the crisis has altered the relation between stock returns and conventional monetary policy announcements, we see that in most cases it has changed. Sixteen out of nineteen portfolios show a t-statistic that can significantly reject the null hypothesis that the parameters are equal. The most interesting effects are found in the results for the unconventional monetary policy surprises. Sixteen sectors are significantly influenced by such announcements and most of these parameters are highly significant. In contrast to the conventional surprise during the crisis, the estimates carry a negative sign. There is quite some heterogeneity in the magnitude of the parameters. The highest parameter estimate is again found for the banking sector. Apparently, this sector benefited the most from announcements that eased financial conditions (i.e. decreased yield spreads) of euro countries. Again using the average change in yield spreads on announcement days, a monetary policy announcement that caused a decrease in the German-Italian yield spread of 0.06% on average led to an increase in banking stock returns of almost 1%. The results hint mildly towards a link with durability, since durable sectors like construction & materials, real estate, and automobile & parts score relatively high. Whereas non-durable sectors like personal & household goods, travel & leisure, and media are at the lower end of the spectrum. However, the sectors most affected after banks are insurance and oil & gas, which in turn do not share a link with durability. The connection to durability therefore seems quite weak. It is therefore again difficult to discern a relation with the exposure to interest rates. In contrast to the results for the pre-crisis period, expected changes are now found to be insignificant for all portfolios, except one. The constant turns out to be significant in four places in Table 4, possibly indicating excess returns of the corresponding portfolio.
5.3 The Credit Channel
28 Table 5
Regression output for the credit channel. The regression model is given by Equation (5). The MSCI World Ex Europe index and the crisis dummy are included as control variables. All regressions are estimated with White's consistent standard errors. Number of observations per regression: 4215. The t-statistics are provided in parentheses. * Denotes significance at the 10%, ** denotes significance at the 5% level, and *** denotes significance at the 1% level. Conventional Surprise Expected Change Conventional Surprise Crisis Expected Change Crisis Unconventional Surprise Constant R 2 Wald Test
EURO STOXX 50 Large -0.061* -0.128*** 0.075 0.046 -0.064*** 0.000 0.292 -1.565
(-1.843) (-2.700) (0.934) (0.480) (-2.991) (0.225)
EURO STOXX 50 Mid -0.020* -0.058*** 0.075 0.060 -0.068*** 0.000 0.268 -1.284
(-1.679) (-3.876) (1.025) (0.701) (-4.013) (1.186)
EURO STOXX 50 Small -0.005 -0.021* 0.090 0.061 -0.059*** 0.000* 0.228 -1.309
(-0.401) (-1.731) (1.262) (0.717) (-2.924) (1.651)
Interest Coverage High -0.083** -0.175*** 0.073 0.040 -0.044* 0.000 0.281 -2.065**
(-2.399) (-3.543) (1.086) (0.501) (-1.763) (0.308)
Interest Coverage Mid -0.044** -0.080*** 0.137 0.102 -0.066*** 0.000 0.243 -1.796*
(-2.004) (-2.686) (1.392) (0.902) (-3.135) (1.346)
Interest Coverage Low -0.073** -0.137** 0.042 0.020 -0.071*** 0.000 0.268 -1.235
(-2.031) (-2.490) (0.491) (0.197) (-3.047) (0.531)
Current Ratio High -0.053 -0.151*** 0.075 0.032 -0.043* 0.000 0.287 -1.387
(-1.417) (-2.970) (0.890) (0.327) (-1.778) (0.738)
Current Ratio Mid -0.070** -0.100** 0.089 0.046 -0.069*** 0.000 0.247 -1.619
(-2.462) (-2.229) (0.948) (0.415) (-3.571) (1.023)
Current Ratio Low -0.075*** -0.133*** 0.076 0.078 -0.067*** 0.000 0.242 -1.966**
(-2.822) (-3.529) (1.056) (0.965) (-2.838) (0.311)
29 Table 5 - Continued Conventional Surprise Expected Change Conventional Surprise Crisis Expected Change Crisis Unconventional Surprise Constant R 2 Wald Test
Free Cash Flow High -0.069** -0.147*** 0.052 0.030 -0.047** 0.000 0.273 -1.391
(-1.967) (-2.898) (0.654) (0.331) (-2.081) (0.444)
Excluding Financials -0.074** -0.158*** 0.047 0.033 -0.045** 0.000 0.264 -1.486
(-2.065) (-3.097) (0.642) (0.397) (-2.406) (0.269)
Free Cash Flow Mid -0.064** -0.106*** 0.110 0.079 -0.078*** 0.000 0.253 -1.892*
(-2.477) (-2.789) (1.248) (0.778) (-2.800) (1.224)
Excluding Financials -0.059** -0.107*** 0.091 0.052 -0.067** 0.000 0.265 -1.542
(-2.296) (-2.817) (0.970) (0.475) (-2.525) (1.205)
Free Cash Flow Low -0.064* -0.127*** 0.079 0.067 -0.107*** 0.000 0.237 -1.342
(-1.783) (-2.716) (0.787) (0.529) (-3.526) (0.634)
Excluding Financials -0.061** -0.113** 0.095 0.063 -0.061*** 0.000 0.259 -1.794*
(-2.001) (-2.497) (1.169) (0.664) (-3.234) (0.911)
Financial Leverage High -0.080** -0.159*** 0.084 0.080 -0.114*** 0.000 0.219 -1.521
(-2.110) (-3.025) (0.834) (0.614) (-3.041) (0.198)
Excluding Financials -0.085*** -0.158*** 0.081 0.048 -0.073*** 0.000 0.252 -1.618
(-2.822) (-3.333) (0.826) (0.407) (-2.977) (0.290)
Financial Leverage Mid -0.067** -0.120** 0.062 0.029 -0.062*** 0.000 0.275 -1.297
(-2.069) (-2.495) (0.660) (0.274) (-2.820) (0.652)
Excluding Financials -0.071** -0.126** 0.070 0.053 -0.066** 0.000 0.258 -1.552
(-1.982) (-2.390) (0.840) (0.562) (-2.478) (0.217)
30 Table 5 - Continued Conventional Surprise Expected Change Conventional Surprise Crisis Expected Change Crisis Unconventional Surprise Constant R 2 Wald Test
Financial Leverage Low -0.057** -0.123*** 0.104 0.073 -0.056** 0.000 0.278 -1.983**
31
the large and mid cap stocks. However, the estimated parameters are only weakly significant. The large cap has a three times higher coefficient than the mid cap. The small cap stocks were in fact the ones expected to be influenced the most, but the estimated coefficient is highly insignificant. The expected pattern therefore appears to be reversed. It is difficult to reconcile such results with the theoretical framework of a credit channel. For the interest coverage variable, we find that all portfolios show a significant negative parameter, but there does not appear to be a clear pattern in the magnitude of these results. The results for the current ratio confirm the hypothesis that those firms with fewer means to meet short-term obligations should be influenced the most by monetary policy. The parameters for the low and mid level current ratios have negative and significant estimates, with the former having the largest and most significant coefficient. The free cash flow portfolios show significant negative parameters pre-crisis, however, a pattern in the magnitude once more appears to be absent. Additional evidence for the presence of a credit channel on European equity markets can be found in the portfolios sorted on the financial leverage ratio. The portfolio with the highest leverage shows large and highly significant parameter estimates, whereas the estimates for the portfolio with low financial leverage are both smaller and less significant. Finally, the debt-to-equity ratio portfolios show some mild evidence for the presence of a credit channel. The two portfolios with high ratios have the largest coefficients as well as significance levels. Nevertheless, the portfolios with low ratios comes close in terms of magnitude. Whereas the mid portfolios have relatively low parameter estimates. In contrast to our expectations, we observe that the estimated parameters for the expected changes are pre-crisis almost exclusively highly significant.
