Has the ECB’s Monetary Policy Prompted Companies to Invest or Pay Dividends?

This paper focuses on how the European Central Bank’s (ECB) monetary policies influenced non-financial firms. The paper’s two main contributions are, first, to shed light on non-financial firms’ decisions on leverage, and how the ECB’s conventional and unconventional policies may have affected them. Second, the paper also examines how these policies influenced non-financial firms decisions on capital allocation – primarily capital spending and shareholder distribution (for example, dividends and shares repurchases). Towards this end, we use an exhaustive and unique dataset comprised of income statements and balance sheets of leading non-financial firms that operate in the European Economic and Monetary Union (EMU). The main results suggest that ECB’s monetary policies have encouraged firms to raise their debt burden especially after the global recession of 2008. Finally, the ECB’s policies, mainly after 2011, seem to have also stimulated non-financial firms to allocate more resources towards not only capital spending but also shareholder distribution. <br><br>


Introduction
In recent years, one of the main problems faced by developed countries has been the combination of slowing economic growth and lack of inflation in an environment of zero lower bound on interest rates. Summers (2013) brought back the term Secular Stagnationfirst coined by Hansen (1939) to describe the economic environment in the United States (US) since the 2008-2009 Global Financial Recession. This term implies that central banks cannot slash interest rates enough to boost investment and consumption. Indeed, the situation where a central bank is hitting the zero-lower bound is known "liquidity trap" and has fostered vast literature where the effectivity of different fiscal and monetary policies (central bank's extraordinary monetary measures, among them) to boost economic activity has been examined. See, for example, Krugman (1998), Krugman and Eggertsson (2012), Orphanides (2004), Bernanke and Reinhart (2004), and Koo (2011Koo ( , 2013, to name a few. On the other side of the Atlantic, the European Economic and Monetary Union (EMU) countries -who, unlike the US, are not part of a fiscal union, but only of a monetary one faced a similar plight. So, the responsibility of the European Central Bank (ECB) to stimulate the euro area economy has been higher than that of the Federal Reserve and has therefore been translated into a full bunch of different conventional and unconventional monetary policies. Summing up, in 2011-2012, after the worst years of the European sovereign debt crisis, the ECB tried to boost liquidity in financial markets by introducing the Securities Markets Program (SMP)first announced in May 2010whose objective was to inject funds into specific market segments that were suffering from insufficient liquidity and depth 1 . The SMP, unlike a quantitative easing program, only injected funds to small and somewhat fewer liquid markets that engulfed with high-risk premium. On July 26, 2012, Mario Draghi (who entered office as President of the ECB on November 2011), promised to do "whatever it takes" to preserve the Euro with the aim to rekindle economic growth in the EMU (Draghi, 2012). Since then, the ECB has introduced several conventional and unconventional stimulating monetary policy measures. Some of these policies include slashing interest rates (including cutting its cash rate to zero and the deposit rate to -0.4% by March 2016), implementing both the longer-term refinancing operations (LTRO) and targeted longer-term refinancing operations (TLTRO), and introducing quantitative easing programs or QE. The main QE programs introduced include the public sector purchase program (PSPP), the asset-backed securities purchase program (ABSPP), a covered bond purchase program (CBPP3), and the corporate sector purchase program (CSPP). As of January 2018, the PSPP was the most massive program among all the assets purchase programs the ECB has implemented with over 1.9 trillion euros in holdings, and it accounts for over 82% of the total asset purchase programs. Table 1 summarizes the most significant announcements regarding the conventional and unconventional monetary policies implemented by the ECB during the recent period.
[Insert Table 1 here] In this context, this paper aims to examine whether ECB's conventional and unconventional monetary policies in times of crisis influenced non-financial firms' decisions. Specifically, the paper focuses on three critical issues: Leverage, investments and shareholders distribution (which comprises primely of dividends and shares buybacks). The contribution of this paper to the existing literature is twofold. First, it examines how ECB monetary policies in times of crisis have affected non-financial firms' decisions on leverage. Second, it analyzes how those policies have influenced non-financial firms' decisions on capital allocationprimarily capital spending and shareholder distribution (for example, dividends and shares repurchase). To the best of our knowledge, this is the first paper to take such a deep dive into the study of the effects of the ECB's policies on non-financial firms. To that end, we use an exhaustive and unique dataset comprised of income statements and balance sheets of leading nonfinancial firms that operate in EMU countries.
The main results suggest that the ECB's conventional and unconventional policies encouraged firms to raise their debt burden especially after the global recession of 2008. Moreover, the ECB's monetary policiesmainly after 2011 in the wake of the European economic crisis and as the ECB shifted its monetary policy as Mario Draghi entered officeseem to have also stimulated non-financial firms to allocate more resources towards not only capital spending but also shareholders distribution.
The rest of the paper is organized as follows. Section 2 includes a literature review of the effects of the ECB's monetary policies on non-financial firms. Section 3 presents the analytical framework. Section 4 describes the data used in the paper. Section 5 explains the econometric methodology while Section 6 reports the empirical results. Finally, Section 7 presents the concluding remarks and suggests some possible policy implications.

The ECB's monetary policies' effects on non-financial firms
An extensive literature has studied the impact of ECB's policies since 2011 from different perspectives and using different methodologies; however, only a few papers have focused on its effects on non-financial corporations despite its crucial role in the economy 2 . Lenza et al. the ECB's balance sheet and macroeconomic conditions. They estimate a panel of eight advanced economies and show that a surprise rise in a central bank's balance sheetmostly via QE programswould raise liquidity (supply side), mainly in countries where central banks are already hitting the zero-lower bound and under the prevailing conditions following the global economic crisis of 2008. Cycon and Koetter's (2015) research suggests that the ECB's unconventional monetary policy reduces refinancing costs. Although it does not lower loan rates; the ECB's policy does mitigate the rise in loan prices because of higher credit demand.
Besides this extensive literature, only a few papers attempted at showing the link between non-financial corporations' investments in the EMU and the ECB's monetary policy. Kanga and Levieuge (2017) assess the effects of the different ECB's unconventional monetary policies on the cost of credit of non-financial firms in each EMU country. Daetz et al. (2016) focus, albeit not exclusively, on the impact of the ECB's LTROs on non-financial firms' cash holdings and concluded that the ECB's measures were most beneficial to corporations from peripheral countries. Darracq-Paries and De Santis (2015), who look at the effects of the 3-year long-term refinancing operations (LTROs) by considering them as a credit supply shock, show that LTROs have helped to elevate the growth rate of real GDP and to raise the prospects of loan provisions for non-financial firms. Arce et al. (2018) show that the ECB's CSPP appear to encourage Spanish firms to issue more bonds and use these funds to increase real investment. Finally, according to Ferrando et al. (2015) small and medium enterprises that are more reliant on local bank creditare more harshly impacted by the euro area's credit crisis than large companies that were able to seek funding aboard. This result is more evident in the stressed countries (Spain, Italy Greece Portugal, and Ireland) than in the rest of the EMU countries.
On the whole, the existing literature that has already focused on the effects of ECB's unconventional monetary policy on non-financial corporations is not only scarce but has not focused on how the different types of policy measures affected companies' decisions on capital structure and capital allocation. This paper will try to fill this gap in the literature.

Analytical framework
In order to better analyze how the ECB's monetary policy affected non-financial firms, in this Section, we first review the literature on the optimal choice of the firm's capital structure in order to examine whether those models might shed some light on the relationship between interest rates and companies' leverage. Then, we examine more deeply how interest rates could influence a firm's decision to allocate its capital between investments and profits distributionvia dividends and buybacks, or a combination of the two.

Capital structure
One of the first studies on the optimal choice of the firm's capital structure is the seminal paper by Modigliani and Miller (1958) who propose the "leverage theorem". The theorem states that, in a context of asymmetric information between companies and investors, a firm determines its leverage ratio based on the capital cost and access to finance. However, since then a couple of other alternative theories was proposed by other authors later [Myers (1984), Kraus and Litzenberger (1973) or Merton (1974), to name a few]. Myers (1984) frame a company's choice under the "pecking order" theory which points out that firms prefer internal funds such as retained earnings to external financing, and debt to equity. Kraus and Litzenberger (1973) offer a competing view (the "trade-off" theory); this view assumes every company achieves an optimal capital structure (a "debt target") at any point in time and trade off tax advantages from debt against refinancing risk. Other authors consider market conditionsincluding interest ratesas a variable that might influence companies' decision on their capital structure. Merton (1974), for example, examine from a theoretical perspective how changes in macroeconomic conditions influence companies on matters such as debt, while Barry et al. (2008) examine this subject albeit empirically. Based on their research, in general, lower interest rates should allow companies to increase their leverage as they reduce their borrowing costs. The theories mentioned above have different implications, not only in the reasons underneath the company's decision to issue more debt but also in the effects that interest rate changes have on that decision. Although there is no consensus on the effect that interest rate changes have on capital structure decisions, our aim in this paper is not to explore the accuracy of those models. However, we aim to use them as a background to build up an econometric framework to examine how those changes may impact firms' leverage decisions.

Capital spending, dividends, and buybacks
One of the ECB's goals through its extraordinary monetary policies was to boost investment. This goal has a simple underlying logic that investments and interest rates are negatively correlated. This logic is prominent in a simple Keynesian IS-LM model where interest rate and its coefficient of interest sensitivity determine investment: = ̅ + In the above equation d>0 stands for the coefficient of interest sensitivity and, under normal economic conditions, falling interest rates should lead to higher investments and lift the aggregate demand to a higher equilibrium. This relationship between interest rates and investment has mainly been examined from an empirical perspective in the literature and its evolution in EMU countries from 1999 until the present is in Figure 1. This Figure shows that it is not clear-cut in the euro area since it only suggests a limited relationship between investments and yields (the correlation over the period is not significant, although the fall in interest rates since 2014 coincided by a steady rise in investment in EMU countries).
[Insert figure 1 here] Nonetheless, the aim of this paper goes beyond that relationship, since the goal is to analyze not only the effect of interest rates on investments but also on dividends and buybacks. To the best of our knowledge, this is the first paper that examines how companies change their capital allocation among investments, buybacks, and dividends due to changes in interest rates. We present below a simple analytical framework to better understand those relationships and the underlying assumptions behind them.
Let us consider that a company, which already took on a debt obligation, needs to decide how to allocate its resources. Specifically, consider a company that needs to evaluate how much to invest in a particular projectnoted as Iversus how much it should allocate towards returning capital to shareholdersin the form of dividends or buyback and noted asover a timeframe of two periods: is the added value to the company's stock price, which the firm aims to maximize. The firm has a budget constraint of: This constraint means that the company has to use all its resources towards an investment I in a particular project or paying its shareholders via dividends or buybacks noted as Vor a combination of both (we are assuming that there are no other alternatives, for example, keeping the capital in cash).
The investment I will yield a return in time one of a profit of ( )a convex, continuous function of I -(let us assume that the company can allocate any portion it desires towards a particular project). This profit will need to be discounted with 1 + . Where r in this equation stands for the company's cost of debt. For simplicity, assume that r stands for the prevailing market interest rates (in other words, the company's risk premium over the market is zero). Conversely, the company can allocate towards shareholders via dividends or buybacks. This shareholder distribution has a positive and constant return set to . This parameter represents the added value associated from a company repurchasing its stocks back or paying dividends to its investors. Put differently, we consider that profit distribution creates value to its shareholders in the form of a signaling mechanism about the positive prospects of a company's future returns -especially if the company's value is undervalued according to the company's management. This positive correlation could be explained by agency costs, information asymmetries, and market irrationality, as Fairchild (2006) points out. In other words, a shares buyback could signal to investors a company is doing well, and its stock is undervalued. This signal will justify a specific return for the company, over time, for these profits' distributions 3 . In this vein, the empirical research that has been done on this matter has also shown the positive relationship between buybacks and stock prices [see Wang et al. (2008), McNally (1999) and Gup and Nam (2001)]. However, with regards to the relation between dividends and firm valuation (as Black and Scholes (1974) examine in detail), the empirical research is not conclusive. In particular, Denis and Osobov (2008) use an international comparison to show minimal empirical evidence for a signaling effect for dividend-paying companies; while Wood and Frankfurter (2002) and Bernhardt et al. (2005) call into question the validity of signaling theories for dividends 4 . In any case, for our model, we consider shares buybacks and their more established positive relation with firm's value to justify a company's decision to allocate capital towards them over investing 5 . In the econometric estimation, however, we use a broader term: "shareholder yield" that includes dividends, buybacks and deleveraging. With these methods, firms can return value to investors as a signaling mechanism.
Given these assumptions, we can solve the firm's maximization problem 6 to know how it distributes its capital in time zero between V and I, based on prevailing market interest rates. The Lagrangian equation is: The First order condition (FOC) for the investment is: While the FOC for the shareholder distribution is: These two FOCs before accounting for the λ budget constraint leads to: 3 Dividends tend to be "stickier"; furthermore, even if market conditions are not good, a company will be more incline to maintain its dividend not to alarm investors from a possible selloff of the stock. Conversely, if a company faces a transitory gain then it will be more incline to distribute its windfall through buybacks rather than raise dividend and thus lift expectations about future d ividend.
That could explain the rise in the prevalence of buybacks as they have become more ubiquitous in recent years mainly, however not solely, in the United States. 4 Conversely, the research done by Hussainey et al (2011) and Garrett, and Priestley (2002) supported the positive relation between dividends and share prices. 5 Even when interest rates fall, both investments and shareholders yield could remain subdued due to low productivity and earnings growthbecause expected lower growth leads to lower growth in investments. If companies face higher capital costs due to heightening risk in the markets (as measured, in one way, via Weighted Average Cost of Capital, WACC), they may opt out of taking risk even as interest rates continue to decline. Conversely, falling equity pricing tend to prop up dividend yields. Lower equity prices could also lead to repurchasing stocks over investing for some companies. Finally, when it comes to banks, they have been more reluctant to provide loans, in part, because of the new capital restrictions (e.g. Basel III) and for them being more prudent after the financial global crisis of 2008 (this could explain the rise in the variance of the interest rates on loans in recent years). 6 This maximization problem does not account for the difference between growth companies and value companies. Where the former tends to allocate more towards investments and the latter tends to prioritize shareholder distribution.
Electronic copy available at: https://ssrn.com/abstract=3317598 The solution shows that a company assesses a project based on two parameters: the company's return to shareholders, and . Therefore, a company divides its resources between investments and shareholders distribution until the discounted marginal return on a given project is equal to the added value a dividend or buyback has on a company's stock price. This is the framework that might help us understand how monetary policy changes could impact non-financial firms' decisions on capital expenditure and shareholder yield 7 .

Data
We gathered the data directly related to the companies' financials from Bloomberg. We focus on non-financial firms listed in the leading stock exchanges from the four largest economies in the EMU: Germany, France, Spain, and Italy (and fair out as a good representation for the entire EMU because their aggregate GDP accounted for roughly 75% of EMU's GDP in 2017) from 2000 to 2017 [Deutsche Börse (DAX), BME Spanish Exchanges (IBEX35), Borsa Italiana (FTSE MIB), Euronext Paris (CAC40)]. Explicitly, we gather quarterly data from a total of 62 non-financial firms (banks and insurance companies are excluded) that register a market capitalization of 2 trillion euros at the beginning of 2017 (which represents nearly a third of the total market capitalization of non-financial firms in the four leading stock exchanges). Therefore, our analysis focuses on large-cap companies since, although their number is not high, they represent a sizable portion of the market value of publicly traded non-financial firms in the EMU.
For our analysis, we use three main dependent variables: "CapEx-to-sales", "Debt-toequity" and "Shareholder yield 8 " that capture capital spending, leverage, and capital distribution to shareholders, respectively. Figures 2 and 3 show the high correlation between the first two variables behavior in the 62 companies included in the sample and in the four largest economies in the EMU (Germany, France, Spain and Italy) while a detailed description of them, together with the rest of the variables used in our analysis, is in Appendix A.
[Insert Figures 2 and 3  As for the monetary policy independent variables, we use changes to the ECB's assets and the 3-month Euribor interest rate. The ECB's assets are used because they show the different policy measures the ECB has employed over the years concerning changes to its balance sheet. This variable does not distinguish the different policy schemes such as LTRO, TLTRO, PSPP, ABSPP, CBPP3, and CSPP 9 . These programs have different targets, starting points, budgets and some have even winded down in recent years. However, all these policies aim to boost liquidity and reduce borrowing costs. Moreover, since late 2014 the majority of the growth in the ECB's assets is attributed to the PSPP. Appendix B includes further details about their progression. As such, we pick the changes to the ECB assets to show how these conventional and unconventional policies, without distinction, affect companies' decisions. We then use the 3-month Euribor as a proxy of the ECB's direct impact on interest rates. We decided to use this variable rather than the ECB's deposit rate because it has a more direct connection to the interest rates faced by companies (nonetheless, Appendix C shows that they are highly correlated).
Finally, we also take into account that two substantial economic events occured during our sample period: (1) the global economic recession of 2008 and (2)

Econometric Estimation
Based on the theoretical framework laid out in Section 4 we estimate the econometric models to examine the role of monetary policy in determining firms' capital spending, leverage and shareholder payouts. Our panel data analysis relies on Blundell and Roulet (2013) who looked at 4,000 global companies and examined the impact of low interest rateswhich directly resulted from the monetary policies of central banks including the Federal Reserve, the ECB and Bank of Japan in recent years-on their investments. They conclude that, since capital spending depends on the cost of equity and uncertainty, low interest rates and tax benefits incentivizes long-term investment (because debt finance is cheap, companies have an incentive to borrow and carry out buybacks -also known as de-equitation-) 10

Leverage
Two of the most widely models used in the literature to analyze the way a company decides on its capital structure are the tradeoff model of Kraus and Litzenberger (1973) and the pecking order model of Myers (1984). The former model looks at a company's aim to raise its debt load until it reaches a specific debt ratio target, whilst according to the latter model, a company will first exhaust its internal funds (available cash) before raising funds from debt and equity. However, these two models neither analyze the relationship between interest rates and the company's decisions on debt as described in Section 3.1, nor examine the role of macroeconomic or monetary policy factors (such as QE programs) on the capital structure of firms. Therefore, following Kühnhausen and Stiber (2014) 11 , in our model, we incorporate external variables that could influence a company's decision on its debt-to-equity ratio ( , is the dependent variable in the model which measures the company's debt burden or leverage): , = , + 1 * , −1 + 2 * −1 + 3 * −1 + , (7) As equation 7 shows, our model includes three prime independent variables. The first (X vector) corresponds to microeconomic variables that are attributed to each companythey are also realted to tradeoff and pecking order models-. The second (Y vector) comprises macroeconomic variables that may proxy the changes in the economy. Finally, the third (Z vector) includes variables which are directly or indirectly related to ECB's monetary policy and proxy supply-side developments 12 .
For our purposes, the monetary policy variables (Z vector) are the most important ones. They include the ECB's asset levels -a proxy to the ECB's asset purchase programs 10 The issue of leveraged buybacks in sovereign debt has been examined by Baglioni (2015), however the process of leveraged buybacks of non-financial firms in times of QE programs is less researched. 11 Their model is based on Rajan and Zingales (1995) and include five macroeconomic factors: GDP per capita, the growth rate of GDP (in constant local currency), inflation rate, interest rate, and tax rate. 12 All independent variables, except WACC, lag the dependent variable by one period. and loans-and changes in 3-month Euribor interest rate. Since the ECB added more funds to the economy and brought down interest rates to encourage companies to take on more loans, we should expect a negative correlation between companies' leverage and interest rates and a positive correlation with the changes in the ECB' assets.
Regarding the microeconomic variables (X vector), three variables are included in the model: profitability (EBITDA-to-sales), growth in profits (growth in earnings per share or EPS) and WACC. We include the variables profitability and growth in profits as they play an important role in determining the leverage of a company as described by both Myers (1984) and Kraus and Litzenberger (1973) 13 while the cost of capital (estimated by the Weighted Average Cost of Capital (WACC)) is a critical variable in this kind of models and a negative relationship between it and the leverage ratio should be expected. Finally, as regards the macroeconomic variables (Y vector), we have included the inflation rate in the EMU because, since inflation depreciates the debt value in real terms, we should expect a positive relationship between inflation and leverage.

Capital spending and shareholder's yield
To analyze the relationship between ECB's monetary policy and the developments of capital spending and shareholder yields we have adjusted the Blundell and Roulet's (2013) model who conducted a panel data analysis and estimated two regressions (one for capital spending per sales and another for dividends and buybacks per sales). Therefore, we have also estimated two equations (an investment equation (8) and a shareholder yield equation (9)), but have adjusted their model by including variables that show how monetary policy affects capital expenditure and dividends/buybacks: , = , + 1 * −1 + 2 * −1 + 3 * −1 + 4 * −1 + 5 * , −1 + 6 * , −1 + , (8) , = , + 7 * −1 + 8 * −1 + 9 * , −1 + 10 * , −1 + , (9) In equation (8), the dependent variable is the company's capital spending divided by sales ( , ). The regression also includes the two main ECB policy variables -the cost of debt (it-1 which is proxied by 3-months Euribor rate) and the changes in the ECB's assets (ECBt-1)plus another four independent variables: the cost of capital (ki,t-1, measured by the WACC), changes in profits (Ei,t-1 proxied by EBITDA-to-Sales), the inflation rate in the EMU (Pt-1), and the spread between long-term and short-term yields (St-1) 14 .
By including the last two variables, we aim to test changes to the economy and market expectations that are directly linked to the ECB's policies all awhile still including variables related to the ones Blundell and Roulet use in their analysis. In particular, inflation serves as a proxy for changes in demand and monetary policy. Nonetheless, the 13 The empirical evidence is also divided: Fama and French (2002) show that companies with higher profits tend to be less leveragedputting the pecking order model right on this issue; Wald (1999) reaches a similar conclusion. On the other hand, Frank and Goyal (2008) show the opposite. 14 The spread between 10-years weighted average of sovereign bond yields of all EMU countries and 3-month Euribor rate.
relationship between inflation and capital spending is not clear. On the one hand, higher inflationary pressures may lead the real returns (see Fama and Gibbons, 1982) on projects to be less profitable 15 , but on the other, a rise in the rate of inflation might also indicate higher economic activity. With regards the spread between long-and short-term rates, it is used as a proxy of economic conditions. According to Baumeister and Benati (2010), the compression of long-term bond spread may even impact GDP and inflation. Furthermore, this compression tends to indicate a fall in the term premium. The decline in the term premium could be due to lower expectations of either sudden inflation eruptions or future lower interest rates because of slower economic activity in the future. In other words, a contracting spread, or the flatting of the yield curve, may correspond with companies reducing capital spending as economic activity deteriorates. Therefore, we would expect a positive relationship between capital expenditure and bond yield spread.
As stated before, our model includes an investment equation (8) and a shareholder yield equation (9) where the variables that may affect the shareholder yield (yi,t) are explored. Likewise equation (8), equation (9) also includes the two main ECB policy variablesthe cost of debt (it-1) and the changes in the ECB's assets (ECBt-1)plus another two independent variables: the cost of capital (ki,t-1 measuread by WACC) and changes in profits (Ei,t-1 proxied now by earnings per sale or EPS of each company). A positive relationship is expected for the former variable (if the cost of retaining a euro to invest relative to the cost of bonds rises, a company is better off repurchasing its sharesand reducing its relative rising cost of capital). Finally, regarding the latter variable, although Blundell and Roulet (2013) use earnings yield in their model, we decided to use changes in EPS because it isolates the changes in a company's fundamentals by not including the variations in its underlying stock price (which could shift based on changes to liquidity in the markets, supply and demand changes and more). As for the expected relationship, even though there is no consensus in the literature 16 , we still expect rising earnings leading to higher returns to investors.

Empirical results
In this section, we first discuss the results from the panel data analysis applied to the leverage, the investment and the shareholder yield regressions. Concretely, we consider two basic panel regression methods: the fixed-effects (FE) method and the random effects (RE) model 17 . To determine the empirical relevance of each of the potential methods for our panel data, we test FE versus RE. We do so by using the Hausman test statistic to analyze the non-correlation between the unobserved effect and the regressors. 15 Although Rappaport and Taggart (1982) alluded that companies might not include inflation in their decision process when evaluating a project. 16 According to French and Fama (2002), more profitable firms tend to have higher dividend payments. Eije and Megginson (2008) looked into European companies and showed that rising earnings didn't raise the chances of increases in cash payoffs to investors. Even Miller and Modigliani (1961) point out that rising profits do not necessarily lead to a rise in dividend paymentit will depend on other factors such as the payout ratio. 17 Estimations were also performed by the Arellano-Bond GMM approach, rendering similar quantitative results.
This test indicates that the fixed effects estimators are more appropriate for all the timeframes in the leverage and the investment regression. However, in the shareholder yield model, the Hausman test shows that the best method (FE or RE) to be used changes depending on the subsample. Subsequently, we also present the results corresponding to a cross-country and a cross-sector analysis for the whole period (it has also been estimated using panel data techniques and in each case the Hausman test has been used to select the best methodology -FE or RE) in order to examine whether companies from different countries or industrial sectors have different reactions to ECB's policies.

Panel unit root tests
A dependent stationary variable cannot be explained using non-stationary variables since the statistical properties of the former (mean, variance, autocorrelation, et cetera) remain constant over time while the statistical properties of the latter change over time. Therefore, to assess the statistical characteristics of our variables, we perform a variety of unit roots tests in panel datasets. In particular, we use the Levin-Lin-Chu (2002), Harris-Tzavalis (1999), Breitung (2000), Im-Pesaran-Shin (2003), and Fisher-type (Choi 2001) tests. The results from these tests (which are shown in Appendix D) decisively reject the null hypothesis of a unit-root for all the variables except for the ECB assets. Therefore, while the rest are found to be stationary in levels, the latter can be treated as the first-difference stationary. So, in the different empirical estimations it will be transformed into a stationary variable by differencing it.

Leverage: Empirical results
The results regarding the main drivers of the leverage ratio are presented in Table 2.
[Insert Table 2 here] These results indicate that interest rates and changes to the ECB's balance sheet have a positive impact on companies' leverage. For the entire period (column 5), a one percentage point fall in 3-months Euribor tends to lift the debt-to-equity ratio, on average, by 3.46 percentage points. Moreover, for every 1 trillion euros the ECB adds to its balance sheet, via the various LTRO and QE programs, companies are likely to raise their debt ratio, on average, by 0.17 percentage point. A closer examination of the results also reveals that the ECB's policies have a stronger marginal effect on companies' debt-to-equity ratio after 2011Q3 (column 2) and 2008Q1 (column 4). Specifically, the 3-months Euribor coefficients in column 4 (-3.968) and column 2 (-2.495) are much lower than the coefficients in column 3 (-0.655) and column 1 (-1.174).
As for changes in the ECB's assets, the coefficients are much higher in column 4 and 2 compared to column 1 and 3. The inflation rate, which is another variable that is indirectly affected by monetary policy, also presents positive and significant coefficients across the different time frames. Finally, the overall regressions' fit is satisfactory as measured by the R 2 values. They range from 66.7% to 83.3% for the various time frames.

Capital spending: Empirical results
The results corresponding to the investment equation are presented in Table 3. It can be observed that ECB's policies (both changes in interest rates and balance sheet assets) have a significant and stimulating impact on a company's capital spending across the different time frames. In particular, from 2001 to 2017 (column 5) for every 1 trillion euros buildup in the ECB's assets, the capital-spending-to-sales ratio rises, on average, by 2.98 percentage points. As for interest rates, a decline of one percentage point in the 3-months Euribor tends to raise the CapEx-to-sales ratio, on average, by 1.5 percentage points.
[Insert Table 3 Table 4 presents the results of the panel data analysis for the shareholder yields model.

Shareholder yield: Empirical results
[Insert Table 4 here] The results indicate that changes in the ECB's policies have a stimulating and significant impact on companies' shareholder yield across the different time samples. In particular, from 2011 to 2017, for every 1 trillion euros the ECB adds to its balance sheet, shareholder yield rises, on average, by 1.33 percentage point (column 5). Moreover, for every one percentage point decline in the 3-months Euribor, shareholder yield increases, on average, by 0.912 percentage point. We also find that after 2011Q3 (column 2) the ECB's policiesmainly related to changes in interest rates (3-months Euribor)seem to have a stronger marginal impact on shareholder yield compared to before. The results of the regressions are significant according to the F-tests and the R 2 values throughout the different time frames. The R 2 values range from 59.1% to 73.5%. Finally, these results also suggest, as indicated in Section 3.2, that lower interest rates do not crowd out dividends or buybacks in favor of investments. This finding implies that the added value for companies for returning capital to shareholders may have been high enough to encourage them to allocate more funds not only to investments but also to shareholder distribution.

A Cross-Country Analysis
In order to analyze how companies from different countries react to ECB's policies, we have also conducted a cross-country analysis. To this end, we have separated the companies in our sample according to their country of origin (based on where their head offices are located): Germany, France, Italy, and Spain. The results of the panel data analysis for the entire period (2000-2017) are presented in Appendix E. These results show that, for the debt-to-equity ratio (Table E1), the coefficients for the ECB assets are positive and significant across the different countries. However, the ECB's balance sheet variable appears to have the strongest stimulating effect on German companies: For every 1 trillion euros the ECB adds to its balance sheet, a German company's debtto-equity ratio rises, on average, by 4.7 percentage points. Conversely, Italian companies have the lowest coefficient of 1.51. Moreover, the 3-months Euribor coefficients are all negative and significant. However, Spanish and French companies have the lowest coefficients at -9.8 and -8.3, respectively. German companies recorded the highest 3-months Euribor coefficient. This result suggests that Spanish and French companies are more sensitive to changes in interest rates than German companies are.
Regarding the CapEx-to-sales ratio regressions (Table E2), German companies are the least sensitive to changes in the ECB assets or interest rates while Spanish and French companies are the most sensitive to the ECB's policies. Finally, as for shareholder yields (Table E3), Italian companies are the least sensitive to changes in the ECB's assetsthe coefficient is only 0.267; while the coefficient of Spanish companies is the highest in the sample at 3.06. Conversely, Spanish companies are the least sensitive to changes in interest rateswith a coefficient of -0.452; while the coefficient of Italian companies is the lowest at -1.437. These findings indicate that both Italian and Spanish companies are more sensitive to only one (and different) form of the ECB's policies compared to companies from other countries.

A Cross-Industry analysis
Finally, we have also conducted a cross-sector analysis in order to examine whether the ECB's policies have affected differently according to the economic sector. So, we break down the sample into 12 industrial sectors 18

Concluding remarks
In this paper, we have examined whether the ECB's monetary policies have encouraged non-financial firms to raise their debt burden, invest more and boost their shareholder distribution. The main results indicate that the answer to all three is yes. However, the results also show that these policies seem to have a stronger marginal impact on these companies' decisions not only after the global recession of 2008 but also after late 2011 as the Euro debt crisis was unfolding and Mario Draghi entered office, and dramatically changed the ECB's policies. We also find that French and Spanish companies appear more sensitive to changes in the ECB's policies on issues of investments and leverage. This finding might have policy implications: The ECB's main asset purchase program (PSPP) allocates its funds based on a country's size (GDP) rather than its need. The results suggest that the ECB's policies could boost investments of non-financial firms more efficiently if the ECB were to allocate more funds to countries, such as France and Spain, where companies react more strongly to its policies. Finally, one of the main goals the ECB set out to do via its stimulative monetary policies was to encourage companies to invest in the economy, which should lead to higher economic growth. As in every empirical analysis, the results must be regarded with caution, since they are based on a set of countries and companies over a certain period and a given econometric methodology. Nonetheless, we show that while the ECB's policies seem to have done so, the policies may have also encouraged companies to use the low interest rate environment to distribute capital to their shareholders. Even though share buybacks and dividends could play a role in boosting economic activity 19 , their stimulative impact on the economy is indirect and unclear.