Financialisation and Real Investment in the European Union: Beneficial or Prejudicial Effects?

ABSTRACT This article presents an empirical analysis of the relationship between financialisation and real investment for non-financial corporations using panel data composed of 27 European Union countries over 19 years (1995 to 2013). On the one hand, financialisation leads to a rise in financial investments, diverting funds from real investments (‘crowding out’ effect); on the other, pressures from shareholders to intensify financial payments restrict the funds available for new real investments. We estimate an aggregate investment equation with the traditional variables (lagged investment, profitability, debt, cost of capital, corporate savings and output growth) and two further measures of financialisation (financial receipts and financial payments). The findings demonstrate that financialisation has damaged real investment in European Union countries, mainly through the channel of financial payments, either by interest or dividend payments. It is also found that the prejudicial effects of financialisation on investment were more severe in the pre-2007 crisis period. It is concluded that financialisation contributed to a slowdown of real investment by 1 to 8 per cent in the full and pre-crisis period, respectively. During the pre-crisis period, financialisation was the main driver of the slowdown of investment in the European Union.


Introduction
Conventional economic theory maintains that the growth of finance fosters economic growth due to the positive association between savings and investments (e.g. Levine 2005). Nevertheless, scholars of financialisation (for example, Orhangazi 2008aOrhangazi , 2008bHein and van Treeck 2010;Hein 2012;Hein and Dodig 2015) postulate that the increasing growth of finance harms the real investment of non-financial corporations (NFCs) through two channels. The first channel involves the greater engagement of NFCs in financial activities, which tends to divert funds from real investments (the 'crowding out' effect). The second channel is caused by strong pressures on NFCs to increase their financial payments (interest, dividends and/or stock buybacks) to the financial markets and shareholders, respectively, which leads to lower retention ratios and fewer funds for long-term productive projects.
In light of this, some empirical studies have been conducted in recent years to assess the relationship between financialisation and real investment. Most of these derive and estimate investment equations that find statistical evidence of the prejudicial effects of the phenomenon on real investment (for example , Stockhammer 2004;Orhangazi 2008aOrhangazi , 2008bvan Treeck 2008;Onaran, Stockhammer, and Grafl 2011;Tori and Onaran 2015; Barradas and Lagoa 2017).
This article examines the impact of financialisation on the real investment of NFCs in the European Union (EU) countries during the period 1995 to 2013 inclusive, contributing to the literature in two new ways. First, the focus is on EU countries, whereas most studies are oriented to the specificities of large, highly developed and financialised countries like the United States or the United Kingdom. Second, a panel data econometric analysis is used rather than the time series econometric analysis more usual in empirical studies on this matter. This allows us to understand whether the prejudicial effects of financialisation have been generalised and are transversal to a large set of countries or, alternatively, are specific to certain countries. The use of a panel data econometric analysis also permits a higher number of observations, sample variability and less collinearity, and thus improves the accuracy and reliability of estimates.
EU countries represent an interesting case study as they share common economic rules because they belong to the same economic and political region. However, these countries have some diversity in terms of financialisation, probably due to their different types of financial system ('market-based' or 'bank-based', in the typology of Sawyer 2013), distinct growth models in the era of financialisation ('debt-led consumption boom', 'domestic demand-led' and 'export-led mercantilist', in the classification of Hein 2012) and diverse models/varieties of capitalism ('liberal market economies', 'coordinated market economies', 'hybrid/mixed market economies' and 'transition economies', in the spirit of Hall and Soskice 2001, and other related works in the field of comparative political economy). These dissimilarities could explain the differences in the levels of financialisation across countries (see Figures A7 and A8 in the Appendix). Despite this heterogeneity, there has been a downward trend in the investment rate in most of these countries ( Figure A1), concurrent with a rise in financial receipts and financial payments (Figures A7 and A8). It is therefore interesting to determine whether there is a disruptive relationship between financialisation and real investment and to what extent the increase of both financial receipts and financial payments have contributed to the slowdown of real investment.
Accordingly, we estimate an investment equation using standard variables (lagged investment, profitability, debt, cost of capital, corporate savings and output growth) and two additional variables linked to financialisation (financial receipts and financial payments). We estimate an aggregate investment function given our interest in studying a macroeconomic issue.
It is concluded that financialisation exerts a negative influence on the real investment of EU countries, mainly through the second channel and especially in the pre-2007 crisis period. This confirms our presupposition that the disruptive relationship between financialisation and real investment is a generalised phenomenon with a negative effect on EU countries from a macroeconomic point of view. It is also found that these negative effects of financialisation on real investment are more severe in the pre-crisis period. During the pre-crisis period, financialisation was the main driver of the slowdown of investment in the EU.
The remainder of the article is organised as follows. In Section Two, an overview is presented of the theoretical and empirical literature on the relationship between financialisation and the real investment of NFCs. An investment equation is built in Section Three. The data and the econometric methodology are described in Section Four. Section Five discusses the main findings. Section Six concludes.

The relationship between financialisation and real investment
It is generally accepted that higher rates of physical capital accumulation are crucial to sustain more dynamic economic growth and employment creation (Orhangazi 2008a(Orhangazi , 2008b. Conventional economic theory advocates that the growth of finance is generally a positive phenomenon supporting the real investments of NFCs given the linkage between savings and investments. This idea has been reinforced by some empirical studies which found a positive relationship between the growth of finance and economic growth (Levine 2005;Ang 2008;Arestis, Chortareas, and Magkonis 2015).
Nonetheless, the strong growth of finance in the last three decades has cast doubts on the 'finance-growth nexus'. Empirical studies confirm a decrease or even a reversal in the relationship between financial development and economic growth (Prasad, Rajan, and Subramanian 2007;Kose et al. 2009;Rousseau and Wachtel 2011;Cecchetti and Kharroubi 2012;Barajas et al. 2013;Dabla-Norris and Srivisal 2013;Beck, Degryse, and Kneer 2014). These studies recognise that the size of the financial sector has an inverted U-shaped effect on economic growth. This means that, from a certain threshold, a further enlargement of the financial sector can reduce real economic growth.
Against this backdrop and according to the financialisation literature, framed in the post-Keynesian tradition, the growth of finance can be prejudicial to the real investments of NFCs through two distinct channels, as is theoretically discussed by Orhangazi (2008aOrhangazi ( , 2008b, Hein and van Treeck (2010), Hein (2012) and Hein and Dodig (2015), among others. Figure 1 exhibits the channels (and factors that feed each of them) associated with the effects of financialisation on real investment. 1 The first channel involves NFCs' increasing investment in financial activities and financial assets, which takes funds from real and productive activities. This is labelled by Hein (2012) and Hein and Dodig (2015) as the 'management's preference channel'. As both external and internal funds are limited, NFCs can only use these funds to invest in financial or real activities, since financial investments and real investments are considered perfect substitutes (Tobin 1965). Thus, NFCs have fewer funds for real and productive investments when they increase their financial investments, and this is commonly referred to as the 'crowding out' effect on real investment. Krippner (2005) confirms that NFCs in the US have increased their involvement in financial investments, as revealed by the growing importance of financial revenues and profits vis-à-vis the revenues and profits from real investments. Similarly, Cingolani (2013) argues that this behaviour expresses a higher accumulation of financial assets to the detriment of productive accumulation. In that sense, Botta, Caverzasi, and Tori (2015) build a new monetary circuit taking into account the real financial macroeconomic interactions in the financialised era; namely, incorporating the increasing acquisition of financial assets by NFCs. The literature on financialisation offers several explanations to describe this stance by NFCs.
First, Crotty (2005) advocates that the rise in financial investments (normally in the form of buying financial subsidiaries or expanding an existing one) has resulted from the incompatibility of the shorter planning horizons of NFCs with the pursuit of longterm real projects. This short-termism mirrors a tendency among investors to sacrifice long-term investment projects in order to increase short-term profits (Aspara et al. 2014). According to Samuel (2000), this focus on short-term profits instead of longterm expansion reflects a certain 'managerial myopia'.
Second, Crotty (1990) concludes that shareholders are more concerned with current profitability than long-term expansion or, ultimately, a corporation's actual survival. Orhangazi (2008aOrhangazi ( , 2008b stresses the strong pressures (essentially exerted by shareholders) on managers to achieve higher short-term returns. These pressures encourage financial investments, which tend to produce larger and more speculative short-term profits rather than real investments that normally involve more uncertainty and only produce profits in the medium and long term. This is the so-called 'rent-seeking behaviour' of NFCs. In fact, Hein (2012) and Hein and Dodig (2015) stress that NFCs face a 'growthprofit trade-off' because shareholders are orientated mainly toward short-term profitability. Once again, this discourages the implementation of real capital projects. Baud and Durand (2012) also state that NFCs intensify their financial investments during bull markets which produce higher levels of profit and respond to the pressures of shareholders. Levy-Orlik (2012) notes that NFCs sometimes repurchase their own shares in order to prevent hostile takeovers, which also leads to a rise in share prices and increases short-term profits. Tomaskovic-Devey, Lin, and Meyers (2015) even stress that the performance of managers is no longer evaluated in terms of market share but rather according to their ability to generate short-term profits and higher share prices.  Orhangazi (2008aOrhangazi ( , 2008b, Hein and van Treeck (2010), Hein (2012) and Hein and Dodig (2015).
Third, Crotty (2005) and Orhangazi (2008aOrhangazi ( , 2008b argue that NFCs may be engaging more in financial activities in reaction to the downward trend of profits from the real sector and the increase in external funding costs since the 1980s. Baud and Durand (2012) confirm that the involvement of US retailers in financial activities is the result of the decline in profitable opportunities in real investments, motivated by the maturation of markets, low profitability rates, stricter regulations, sluggish consumer demand and increasing competition. Soener (2015) notes that this is the political economy perspective in which NFCs are becoming more financialised so as to remain viable. Crotty (2005) terms this the 'neoliberal paradox'; he claims that shareholders tend to coerce NFCs to remain competitive and profitable even in downturn environments, thus inducing managers to move from productive to financial investments. According to Lin and Tomaskovic-Devey (2013), this behaviour reflects the NFCs' strategy of 'fast growth in a slowgrowth economy'.
Fourth, Akkemik and Özen (2014) advocate that macroeconomic uncertainty and heightened risks, together with institutional changes in corporate governance, are the main reasons behind the rise in financial investments by NFCs. They tested these hypotheses using a panel data econometric analysis for 41 corporations quoted on the Istanbul stock exchange for the period 1990 to 2002. However, they found this channel to be mainly determined by highly uncertain macroeconomic environments and by the characteristics of corporations (such as dimension), whilst institutional features (e.g. close ties with the government, family ownership, discretion of managerial power and unionisation) do not have a statistically significant impact on financialisation. In fact, Baud and Durand (2012) stress that there is a greater preference for liquid assets in business environments characterised by high levels of uncertainty, as the financial investments of NFCs represent a kind of 'wait-and-see' strategy. In turn, NFCs involve themselves in financial activities through financial instruments which hedge several risks against uncertainty (Soener 2015).
Following a neo-institutionalism perspective, Soener (2015) adds two further explanations to describe the growing importance of NFCs' financial investments. First, he stresses that NFCs learn to financialise with other corporations, i.e. so-called 'mimetic behaviour'. Second, he emphasises that some actors (like financial executives or independent consultants) influence investors and managers to make more financial investments. Here, there is an institutional transmission of knowledge and practice from the know-how of these actors in the corporate finance field to the respective investors and managers.
Nonetheless, some authors (e.g. Fazzari, Hubbard, and Petersen 1988;Gertler and Gilchrist 1994;Ndikumana 1999) emphasise that the increase in financial receipts due to investments in financial activities and financial assets could exert a positive influence on productive investments if (and when) NFCs channel these financial incomes to make real investments. Orhangazi (2008aOrhangazi ( , 2008b recognises that this could be a more relevant mechanism in the case of small and medium-sized corporations since they face higher financial constraints and are therefore forced to use all incomes (even financial) to undertake real investments. The same argument is presented by Tori and Onaran (2015), who stress that the effect of financial receipts on investment depends on the individual NFC's dimension and/or sector. Accordingly, it exerts a negative influence in the case of large NFCs because they re-invest these financial incomes in other financial investments, but it has a positive influence in the case of small and medium-sized corporations because they are more likely to be financially constrained.
The second channel is associated with the strong pressures on NFCs to increase their financial payments (interest, dividends and/or stock buybacks) to the financial markets and the respective shareholders. This limits the funds available for real investments, which is commonly referred to as the 'profit without investment' hypothesis (Stockhammer 2006;Cordonnier and Van de Velde 2015). According to Stockhammer (2006), this hypothesis represents a stylised fact in the neoliberal era after the mid-1980s, reflecting a trend of lower investment and output growth along with higher profits due to increasing shareholder power. As noted by Aglietta and Breton (2001) and Duménil and Lévy (2004), the higher levels of dividend payout ratios reduce the funds available for real investments made by NFCs, which has had a negative effect on the execution of long-term investment projects including activities such as innovation, research and development. Hein (2012) and Hein and Dodig (2015) term this the 'internal means of finance channel'. Once again, the literature on financialisation presents several explanations for the low financial retention ratios of NFCs.
First, Orhangazi (2008aOrhangazi ( , 2008b focuses on the high levels of NFCs' indebtedness, which lead to a rise in financial payments in the form of interest. Second, and regarding financial payments through dividends, the author notes that managers are encouraged to raise short-term payout ratios and it is in fact in their interest to do so as their remuneration is based on the short-term evolution of stock prices. Their strategy is therefore to distribute high dividends, because this tends to drive a short-term increase in stock prices. On the other hand, this is simply a response to pressures from shareholders who, in some cases, are institutional investors that seek constant appreciations in stock value and high payout ratios. If NFCs do not make these financial payments in the form of dividends, their stocks could decline sharply, as demand for them will fall and supply increase, which could ultimately lead to a takeover. Third, Lazonick and O'Sullivan (2000) and Stockhammer (2010) argue that this growing trend in financial payments by NFCs over the last three decades is associated with a new design of corporate governance that favours the maximisation of shareholder value: the so-called 'shareholder value orientation'. Aglietta (2000) and Van der Zwan (2014) note that this has become 'the norm of the transformation of capitalism' and is responsible for the dissemination of policies and practices that tend to favour shareholders over the other constituents of corporations. Lazonick and O'Sullivan (2000) suggest there has been a shift from 'retain and reinvest' to a 'downsize and distribute' strategy-that is, a transfer from a strategy oriented towards profit retention and reinvestment in the growth of corporations to one of downsizing of corporate labour forces and the distribution of profits to shareholders. Levy-Orlik (2012) emphasises that investors' decisions based exclusively on the aim of maximising shareholder value target the reduction of production costs and increase in stock prices to the detriment of employment, income equality, innovation and industrialisation.
Conversely, and as referred to by Orhangazi (2008aOrhangazi ( , 2008b, some authors claim that the increase in financial payments could be positive for NFCs' real investment on the grounds that higher levels of financial payment depend on higher profits and solvency. As such, these corporations will probably have access to more funding at lower costs, which could increase the implementation of new productive investments. However, this has not been confirmed empirically (e.g. Orhangazi 2008aOrhangazi , 2008bvan Treeck 2008;Onaran, Stockhammer, and Grafl 2011;Tori and Onaran 2015;Barradas and Lagoa 2017), probably because those pressures to raise short-term financial payments are so strong and constant that NFCs cannot implement new real investments.
Despite the growing body of theoretical work on the effects of financialisation on real investment, there are few empirical studies on the subject, as emphasised by Onaran, Stockhammer, and Grafl (2011). Nevertheless, some empirical studies estimate investment functions for several countries in order to make an econometric analysis of financialisation's impact on real investment; most of these find it to be harmful. 2 Stockhammer (2004) estimates an investment equation for Germany, France, the UK and the US using a time series econometric analysis for each country individually. He uses interest and dividends received (the so-called rentier income) by NFCs to measure financialisation, and concludes that it has led to a deceleration in real investment, particularly in the US and the UK. In the case of the UK, the negative effect is more modest, particularly because there has not been a decline in accumulation due to low accumulation rates in recent decades. The insignificant effect in Germany is explained by the recent nature of the 'shareholder value orientation' phenomenon in this country. Orhangazi (2008aOrhangazi ( , 2008b also identifies the deleterious effect of financialisation in the US. He conducts a time series econometric analysis using aggregate data for NFCs as a whole from 1961 and 2004 and a panel data econometric analysis using micro-data from 1973 and 2003, analysing by sector (manufacturing versus non-manufacturing corporations), industry (durable versus non-durable producers) and dimension (small versus large corporations). In both studies, he applies financial profits (interest and dividends) and financial payments (interest, dividends and stock buybacks) to measure the two channels of financialisation. He finds a negative relationship between these two channels and investment in both studies. van Treeck (2008) also performs a time series econometric analysis for the US, for the period 1965 to 2004. He concludes that interest and dividend payments exert a negative influence on long-term non-financial investment in the US. Onaran, Stockhammer, and Grafl (2011) estimate an investment function using a time series econometric analysis for the US from 1960 to 2007. They find evidence supporting the claims that financialisation (proxied by interest and dividend payments) has suppressed the level of investment. Tori and Onaran (2015) analyse the effect of financialisation on physical investment in the UK, using micro-data at the corporation level from 1985 to 2013. They find evidence supporting the claims that financialisation exerts a negative influence on physical investment through financial receipts and financial payments. Barradas and Lagoa (2017) perform a time series econometric analysis focused on Portugal from 1977 and 2013, using a vector error correction model and financial receipts and financial profits of NFCs as proxies to capture financialisation. They conclude that financialisation has hurt real investment, mainly through financial payments and particularly in the long term.
The literature has focused on large and highly developed countries through time series econometric analyses. Here, we aim to make an empirical assessment of the relationship between financialisation and real investment of NFCs using a large set of countries, that is, EU countries. To the best of our knowledge, this is the first panel data econometric analysis of a group of countries over time regarding financialisation and real investment. This approach will allow us to perceive if the prejudicial effects of financialisation have been generalised and are transversal to this large set of countries or if they only affect specific countries from a macroeconomic point of view. 3

Financialisation and real investment: an economic modelisation
Empirical studies of real investment are particularly difficult when they are carried out through econometric estimations of investment functions. Effectively, 'estimation of investment functions is a tricky and difficult business and the best posture for any of us in that game is one of humility' (Eisner 1974, p. 101). In the same vein, Davidson (2000) emphasises that investment equations do not follow a stable functional expression over time, since investment decisions are constantly affected by the exogenous 'animal spirits' of investors.
Nevertheless, there are several theoretical and empirical studies of real investment in the literature that reveal the main determinants of investment decisions. Chirinko (1993) and Kopcke and Brauman (2001) surveyed the traditional literature on this issue, identifying five determinants of investment: the level of profitability, the degree of indebtedness, the cost of capital (or interest rates), cash-flow (or internal funds) and output (or sales).
In what follows, we estimate an equation where investment is a function of the prevalent variables in the explanation of investment decisions of investors in NFCs: lagged investment, profitability, level of debt, cost of capital, corporate savings and output growth. Additionally, we incorporate two further variables (financial receipts and financial payments) to account for the two channels related to financialisation's prejudicial effects on productive investments, as described previously.
Accordingly, our investment function takes the following form: where i is the country, t is the time period (years), LS is the labour income share of country i at time t, P is the profitability of NFCs of country i at time t, D is the corporate debt of NFCs of country i at time t, CC is the cost of capital of country i at time t, CS is the corporate savings of NFCs of country i at time t, OG is the output growth of country i at time t, FR are financial receipts of NFCs of country i at time t and FP are financial payments of NFCs of country i at time t.
The two-way error term component is given by: where h i accounts for unobservable country-specific effects and l t accounts for timespecific effects. The term 1 i,t is the random disturbance in the regression, varying across countries and years.
We use lagged values for the independent variables because of the time lag between investment decisions and the respective capital expenditures (the decision to invest in t was based on information at t − 1), the role of the formation of investors' expectations (adaptive expectations) and the need to avoid potential problems of simultaneity and reverse causation (i.e. endogeneity problems) (Orhangazi 2008a(Orhangazi , 2008b. 4 On the one hand, profits (where financial receipts and financial payments are included) of a certain year are only available for investments in the following year; on the other hand, investors only know the lagged values of output growth when they make investment decisions.
We also include the lag of the dependent variable, taking into account the degree of persistence that is present in the majority of macroeconomic variables. The inclusion of lagged investment helps to control for dynamic effects; namely, the gestation in time (investment projects usually take over one year to be implemented), irreversibility in investment decisions and inertia (higher/lower investment normally leads to higher/ lower subsequent investment) (Orhangazi 2008a(Orhangazi , 2008b. This is corroborated by Kopcke and Brauman (2001), who confirm that years of high/low investment tend to follow years of high/low investment. In fact, there is a 'path dependency' of real investment over time, as stated by Tori and Onaran (2015). Note also that Ford and Poret (1990) argue that the best determinant of current investment is its own past history.
All NFC variables (investment, profitability, debt, corporate savings, financial receipts and financial payments) are expressed as ratios of the respective gross value added (both the numerator and denominator are in volume). This permits the comparison of variables expressed in different currencies, making exchange rates unnecessary for conversion to the same currency that could skew results due to the respective movements in international financial markets. This also allows the respective coefficients to be interpreted in percentage points.
It should be noted that we propose to estimate an aggregate investment function, similar to the approach taken by Stockhammer (2004), Orhangazi (2008a), van Treeck (2008), Onaran, Stockhammer, and Grafl (2011) and Barradas and Lagoa (2017). Stockhammer (2004) emphasises that the respective results should be analysed with care as we are addressing a macroeconomic issue-i.e. the slowdown of real investmentalthough the theory of NFCs' investment decisions is supported by microeconomic fundaments. This strategy implies the assumption of a representative corporation.
In addition, we recognise some limitations to this approach, since the use of an aggregate investment function does not reflect different financialisation levels among NFCs or potential dissimilarities in the behaviour of NFCs from different countries, sectors, industries, dimensions and/or ownerships. It should also be noted that, as a panel data econometric analysis estimates an average effect of several countries, it does not account for the historical, social and economic circumstances responsible for real investment in each country. However, Orhangazi (2008a) states that this approach has several advantages; namely, by allowing us to look beyond the particularities of different NFCs and to identify the main relationships that dominate real investment. Here, we follow a macroeconomic perspective to assess whether financialisation has been beneficial or prejudicial to real investment in the EU. Thus, if the two channels of financialisation are found to have a macroeconomic effect, we cannot determine whether it is due to the impact of some corporations/countries or is more generalised across all corporations/countries. If we do not find any macroeconomic effect, we cannot exclude that they affect a subset of corporations/countries, which, however, is not enough to create a macroeconomic effect in all countries.
Accordingly, lagged investment, profitability, corporate savings and output growth are expected to exert a positive influence on investment, while cost of capital and financial payments are expected to influence NFCs' investment negatively. The level of debt and financial receipts could have a positive or a negative influence on investment. Thus, coefficients of these variables are expected to have the following signs: Profitability is expected to exert a positive influence on real investment, mirroring the demand conditions that are crucial to determine the viability of investment projects. Following a Keynesian argument, profitability tends to have a positive effect on real investment by functioning as a source of internal funds (Stockhammer 2004). Kopcke and Brauman (2001) note that expectations for future demand conditions and future profitability have the strongest influence on investment. Nonetheless, Kuh and Meyer (1955) and Minsky (1975) state that, given uncertainty about the future, it is past demand conditions and past profitability rather than future expectations that are the major influencers of investment. This seems to prevent the anticipation of future demand conditions and future levels of profitably, in a context where these expectations are normally formed on the basis of the past.
The debt level has an undetermined effect on investment (Orhangazi 2008a(Orhangazi , 2008b. A positive effect is expected when the debt level is perceived to be safe. Here, a rise in debt may have no effect or even a positive effect on investment by increasing the available funds. A negative effect occurs when the debt level is perceived to be unsafe, as it signals greater financial fragility and makes it more difficult to obtain further funding. In that situation, future profits may be insufficient to repay existing debt, increasing the probability of bankruptcy. The cost of capital (normally measured by the level of real long-term interest rates) is expected to exert a negative effect on real investment, reflecting the funding costs or the respective opportunity costs.
In addition, real investment is expected to depend positively on the respective corporate savings, because a higher level of corporate savings tends to be associated with a higher level of retained profits, which is necessary to support real investments. The internal means of finance rests on the 'principle of increasing risk' (Kalecki 1937). This means that corporations, in imperfectly competitive financial markets, need their own funding sources to attract external capital in terms of credit, bonds or equity issues for investment purposes. This also means that corporations constrain the utilisation of external capital for investment purposes to minimise the risk of illiquidity and insolvency. On the other hand, the internal means of finance support real investment due to the 'financial hierarchy', according to which internal funding has a cost advantage over external funding (Fazzari, Hubbard, and Petersen 1988).
On the other hand, real investment also depends positively on output growth. This follows the Keynesian argument of the 'accelerator principle', i.e. that most investors exhibit a higher propensity to invest in periods of economic growth than during downturns, since these periods are commonly associated with a higher demand. The accelerator principle postulates that an acceleration/deceleration of the GDP will accelerate/decelerate real investment, given a multiplier higher than zero. Lopes (2003) confirms that real investment in the EU and the US is strongly pro-cyclical in relation to the respective business cycle. Indeed, Sørensen and Whitta-Jacobsen (2005) highlight the existence of two stylised facts of business cycles in relation to investment: investment is strongly positively correlated with the business cycle and it is the most volatile component of aggregate demand.
Finally, and as discussed in the previous section, financial receipts have an undetermined effect on real investment, given their dual role as a source of funding in the function of the NFCs' dimension. Financial payments tend to lower real investment because they reduce the funds available for these real investments.

Data
In order to analyse the role of financialisation in real investment of EU countries, we collected annual data from 1995 to 2013 for a set of 27 countries (Austria, Belgium, Bulgaria, Cyprus, Czech Republic, Denmark, Estonia, Finland, France, Germany, Greece, Hungary, Ireland, Italy, Latvia, Lithuania, Luxembourg, Netherlands, Norway, Poland, Portugal, Romania, Slovakia, Slovenia, Spain, Sweden and the UK). Malta was the only EU country excluded, due to the lack of data. Table 1 shows the sample period and the number of observations (and those missing) per country. This is the period and the frequency for which all data are available and they are suitable for the study for two reasons. First, financialisation became more prevalent in the 1990s (Van der Zwan 2014); second, investment by corporations is a long-term decision (it usually takes over a year to implement investment projects and recover the invested capital), and therefore annual data are likely to capture the determinants of real investment better than higher frequency data.
We constructed a panel of data (or longitudinal data), since we collected data for a set of 27 cross-sectional units (N = 27) that were observed over time between 1995 and 2013 (T = 19). Nonetheless, we obtained unbalanced panel data because it was impossible to collect data for all years for each country. We have 64 missing values and our sample is therefore composed of a total of 449 observations. Table A1 in the Appendix contains the descriptive statistics of the data and Table 2 presents the corresponding correlation matrix between all variables. The most important finding is that the absolute values of all correlations are lower than 0.8, which is crucial to exclude the existence of severe multicollinearity between the variables of our model (Studenmund 2005). In addition, profitability, corporate savings and output growth are the only variables positively correlated with investment; indeed, they are precisely the three variables expected to have a positive effect on investment. The variables of debt, cost of capital, financial receipts and financial payments are negatively correlated with investment, which could signal a negative effect on investment. This also seems to confirm our presupposition that financialisation has hampered real investment through the two aforementioned channels.
Regarding data definition, we used gross fixed capital formation of NFCs divided by the respective gross value added to delineate the investment of NFCs. The ratio between these two variables is usually known as the investment rate of NFCs.
We used the gross operating surplus of NFCs divided by the respective gross value added as a proxy of profitability; the ratio between these two variables is commonly referred to as the profit share of NFCs. 5 −0.413 * * * 0.020 0.467 * * * 0.017− − 0.597 * * * −0.152 * * * 0.692 * * * 1 * * * Significance at 1% level. 5 According to Eurostat (2015), 'gross operating surplus can be defined in the context of national accounts as a balancing item in the generation of income account representing the excess amount of money generated by incorporated enterprises' operating activities after paying labour input costs. In other words, it is the capital available to financial and non-financial corporations which allows them to repay their creditors, to pay taxes and eventually to finance all or part of their investment' (http://ec.europa.eu/eurostat/statistics-explained/index.php/Glossary:Gross_operating_surplus_(GOS)_-_NA).
In order to measure the corporate savings, we used the gross savings of NFCs divided by the respective gross value added.
Financial receipts correspond to the sum of interest and the distributed income of corporations (where dividends are included) received by NFCs. 6 These were divided by the gross value added by NFCs.
The sum of interest and the distributed income of corporations (where dividends are included) paid by NFCs were used as a proxy for financial payments. These were divided by the gross value added by NFCs.
It should be noted that the variables of gross fixed capital formation, gross value added, gross operating surplus, gross savings, financial receipts and financial payments of NFCs were collected from the European sector accounts (at current prices and in millions of national currency) available at Eurostat. When not available on Eurostat, observations of these variables were obtained with data from the sector accounts of the national statistics office of each country. 7 The proxy of the debt level used here was the total financial liabilities of NFCs divided by the respective gross value added. The total financial liabilities of NFCs were obtained from the financial accounts (at current prices and in millions of national currency) available at Eurostat. If not available at Eurostat, observations of these variables were obtained from data of the financial accounts of each country's central bank. 8 The long-term real interest rates (deflated by the GDP deflator) from the AMECO database to measure the cost of capital of NFCs were used. For some countries, the short-term real interest rates (deflated by the GDP deflator) were also used for several years (Bulgaria for 2000 andCzech Republic from 1995Estonia for 1996Estonia for , 1997Estonia for , 2011Estonia for , 2012Estonia for and 2013Hungary from 1995Hungary from to 1998Lithuania for 1999 andNorway for 2011 andRomania from 1998Romania from to 2005Slovakia from 1995Slovakia from to 1999and Slovenia for 2001) because, in the case of these countries, long-term real interest rates are not available for all years. 9 We chose this strategy instead of using only the short-term real interest rates for all countries because investment is a long-term decision and is therefore more dependent on long-term interest rates. Thus, this variable contains 414 observations with longterm real interest rates, which contrasts with only 35 observations with short-term real interest rates.
Finally, we applied the usual variable of gross domestic product to describe the evolution of output growth. This was collected from European national accounts, available at Eurostat (at current prices and in millions of national currency), and was deflated using the GDP deflator (2005 = 00), available on the AMECO database. Subsequently, the respective annual growth rate was calculated.

Methodology
As we have a dynamic panel data model, due to the presence of a lagged dependent variable among the independent variables, we cannot use the traditional panel data estimators (e.g. pooled ordinary least squares, least squares dummy variables, fixed-effects, randomeffects) since they are biased and/or inconsistent because the lagged dependent variable is correlated with fixed effects in the error term (Nickell 1981;Baltagi 2005;Cameron and Trivedi 2009). Accordingly, there are three different approaches to deal with dynamic panel data. First, we can take the first differences of our model as proposed by Anderson and Hsiao (1982). This instrumental variables (IV) approach removes the constant term and the individual effects and produces consistent estimates. However, Holtz-Eakin, Newey, and Rosen (1988), Ahn and Schmidt (1995) and Baltagi (2005) stress that this methodology does not produce necessarily efficient estimates because it does not use all available moment conditions. Second, we can perform a generalised method of moments (GMM) (Hansen 1982) following the estimator proposed by Arrelano and Bond (1991). This estimator is also based on the first differences of our model ('difference GMM'), albeit using a large set of instruments, and is therefore more efficient than the former. However, Roodman (2009) claims that this estimator has a weakness, namely, magnifying gaps in the case of unbalanced panels, which is our case. Third, we can implement the estimator proposed by Arrelano and Bover (1995) and Blundell and Bond (1998), which allows the introduction of more instruments (not only the lagged levels but also lagged differences), thus improving its efficiency ('system GMM'). This estimator subtracts the average of all future observations of a certain variable ('forward orthogonal deviations' or 'orthogonal deviations') instead of subtracting the previous observation from the contemporaneous one, i.e. the 'first-difference transformation' (Arrelano and Bond 1991). Therefore, this estimator tends to minimise data loss in the case of unbalanced panels by using the aforementioned 'orthogonal' transformation (Roodman 2009).
Nevertheless, these three estimators have proved to be severely biased and imprecise in macro panels where the cross-sectional dimension, N, is relatively small (Bruno 2005a(Bruno , 2005b, which is our case. Therefore, they are not appropriate to carry out our estimates. Against this backdrop, we rely on the least squares dummy variables bias corrected (LSDVBC), introduced by Nickell (1981), Kiviet (1995) and Bun and Kiviet (2003), which was extended by Bruno (2005aBruno ( , 2005b for unbalanced panels. 10 Note that Monte Carlo evidence has shown that this estimator outperforms the IV-GMM estimators in terms of bias and root mean squared error (efficiency) when we are in the presence of macro panels (Kiviet 1995;Judson and Owen 1999;Bruno 2005aBruno , 2005b. Additionally, Monte Carlo evidence has also revealed the good performance of this estimator in terms of bias and efficiency even when we have endogenous variables among the independent variables (Behr 2003), which could be our case with regard to the variables linked to financialisation.
Therefore, we use the LSDVBC estimator for our investment function. 11 This estimator works in two steps (Bruno 2005a(Bruno , 2005b. First, it generates consistent estimates which 10 We follow the 'xtlsdvc' instruction from Stata software. 11 Note that we assume the stationarity of our data for the following three reasons. First, plots of our eight variables ( Figure  A1 to Figure A8 in the Appendix) already seem to indicate that all variables are stationary in levels. Second, our variables require the definition of an initial matrix of starting values that can be performed using three different consistent estimators (Anderson and Hsiao 1982;Arrelano and Bond 1991;Blundell and Bond 1998). However, the choice of one of these three estimators has only a marginal impact on the LSDVBC estimates (Bun and Kiviet 2001;Bruno 2005aBruno , 2005b. Second, it corrects the bias by undertaking a set of multiple replications to bootstrap the standard errors.
In the next section we analyse the estimations of our investment equation for EU countries and undertake some robust analyses to assess whether the results exhibit some sensitivity to other specifications and/or sub-samples. In all of the specifications, we use the Arrelano and Bond estimator to initialise the LSDVBC estimator and a number of replications equal to 250. Time dummies are included in all specifications, as well as a WALD test to assess the joint significance of them all.

Empirical results and discussion
This section presents our estimates. We start by estimating our baseline investment function, as specified in Section Three (Model I). Then, and to obtain a better understanding of the effects of financialisation on real investment in EU countries, we re-estimate the investment function, splitting financial receipts into interest and dividend receipts (IR and DR, respectively) and dividing financial payments into interest and dividend payments (IP and DP, respectively) (Model II). We re-estimate the investment function in order to assess the global net effect of financialisation, through the difference between financial payments and financial receipts-net financial payments (NFP), on investment (Model III). Finally, we re-estimate the investment function taking into account the net effects of both channels, i.e. the difference between interest payments and interest receipts-net interest payments (NIP)-and the difference between dividend payments and dividend receipts-net dividend payments (NDP), on productive investment (Model IV). The respective results are presented in Table 3.
In relation to Model I, all variables are statistically significant at the conventional significance levels, with the exceptions of profitability and debt. On the other hand, all coefficients of the statistically significant variables have the expected signs, with the exception of corporate savings. In fact, NFCs' investment is negatively influenced by corporate savings, which may suggest that internal means of finance are used for other purposes (e.g. to constitute reserves or even to finance financial investments) rather than to implement new real projects. This could be explained by the strong reduction in external funding costs (visible in the fall of real interest rates) in the majority of EU countries since the mid-1990s ( Figure A4 in the Appendix), confirming that the cost advantage of internal funding over external funding was not so significant during that time.
are in fact defined in ratios (in the case of investment, profitability, debt, corporate savings, financial receipts and financial payments) or in growth rates (in the case of output growth); intuitively, it is plausible to assume that these variables do not exhibit a unit root. Third, the traditional panel unit root tests have low power and perform very poorly in the presence of panels where the cross-sectional dimension, N, is higher than the period dimension, T, as recognised by Baltagi (2005), Hlouskova and Wagner (2006) and Cameron and Trivedi (2009), among others. These authors also emphasise that for small T, there is the potential risk of concluding that the whole panel is non-stationary even when the panel has a large proportion of stationary data, since the conventional panel unit root tests tend to assume that T 1. These are the reasons why we do not perform panel unit root tests.
This result does not confirm the 'principle of increasing risk' (Kalecki 1937). The nonsignificance of profitability and the negative impact of corporate savings on investment appear to be consistent with the 'profit without investment' hypothesis (Stockhammer 2006). As expected, lagged investment and output growth exert a positive influence and the cost of capital negatively influences real investment. 12 These results are corroborated by the literature; namely, by confirming that the lagged investment is a strong determinant of current investment (Ford and Poret 1990) and that investment is strongly pro-cyclical in relation to output growth (Lopes 2003;Sørensen and Whitta-Jacobsen 2005). This is also consistent with Keynesian predictions, according to which demand is a key determinant of investment. In the same vein, the 'accelerator principle' is confirmed, since coefficients of output growth are always higher than zero. Financial receipts exert a positive influence on real investment, which seems to exclude the 'crowding out' effect. This positive relationship reveals that financial investments have not been detrimental to real investments, probably because NFCs use financial returns to finance real investments. This is a relatively expected result, taking into account that the structure of the European productive system is characterised essentially by small and medium-sized corporations. 13 These corporations face higher financing constraints in comparison with large corporations, being more dependent on financial receipts to make new real investments. This result has essentially been found in studies that use micro-data at the corporation level, where the effect of financial receipts on investment in the case of small and mediumsized corporations is regularly positive (Orhangazi 2008b;Tori and Onaran 2015).
Finally, real investment is negatively influenced by financial payments in line with the literature on financialisation. This is a quite common result in similar studies of the relationship between financialisation and investment for other economies and/or periods (Stockhammer 2004;Orhangazi 2008aOrhangazi , 2008bvan Treeck 2008;Onaran, Stockhammer, and Grafl 2011;Tori and Onaran 2015;Barradas and Lagoa 2017).
With regard to Model II, results do not change dramatically, with the exception of profitability that becomes statistically significant, exerting a positive effect on real investment. Financial receipts maintain their positive influence on real investment, but only dividend receipts are statistically significant. This suggests that not all financial investments are beneficial to real investments, in a context where returns from shares (dividends) seem to be more decisive in real investment than returns from deposits and/or from debt securities (interest).
Both interest and dividend payments continue to exert a negative influence on real investment, but only the former is statistically significant. This seems to illustrate that the servicing of debt has been more harmful for real investment than the paradigm of 'shareholder value orientation'. In fact, the literature regards the majority of EU countries as latecomers in the development of 'shareholder value orientation', in a context where this paradigm is still a recent phenomenon among these countries (Barker 2010). Despite a certain convergence with Anglo-Saxon countries since the mid-1990s, the author stresses that NFCs of EU countries follow a business strategy that is still broadly focused on the achievement of the interests of stakeholders rather than the pure maximisation of shareholder interests. A similar conclusion was reported by Stockhammer (2004) to justify the insignificant effect of financialisation on real investment in Germany.

12
Note that results do not change dramatically (in terms of statistical significance and signs) if we exclude observations with short-term real interest rates. Results are available from the author upon request. 13 According to the European Commission (2016), small and medium-sized corporations are the backbone of the EU, representing 99 per cent of all business (https://ec.europa.eu/growth/smes_en).
Regarding Model III, results are quite similar in comparison with the other two models. Here, profitability loses its statistical significance. Net financial payments are statistically significant and represent a negative determinant for real investment. This indicates that the beneficial role of financial receipts does not compensate for the prejudicial effect of financial payments, stressing that the global effect of financialisation in EU countries is harmful to real investment. Indeed, all countries exhibit a higher level of financial payments vis-à-vis financial receipts ( Figures A7 and A8 in the Appendix).
From Model IV, we conclude that both net interest payments and net dividend payments are negative, but only the former are statistically significant. This finding illustrates that the global and harmful effect of financialisation on productive investment is caused essentially by interest payments due to debt servicing rather than by dividend payments due to the recent emergence of shareholder-oriented NFCs in EU countries. Despite the non-significance of debt in the explanation of investment, the negative impact of net interest payments represents an indirect effect as the NFCs' indebtedness can be prejudicial to real investments. Indeed, this effect represents an important fragility of NFCs and a strong constraint to boosting real investment in EU countries, in a context where the high level of NFCs' indebtedness and debt service increases the vulnerability of NFCs to interest rate shocks. Real investment is affected simultaneously by the price of debt (cost of capital) and the volume of debt and the corresponding debt service (interest payments).
We evaluate now the robustness of these results. First, we re-estimate the four specifications of the investment function, excluding the recent years of deep economic crisis. Note that output growth, financial receipts and financial payments exhibit a strong decline in 2008 in the majority of EU countries ( Figures A6-A8 in the Appendix). However, as we have a dynamic panel model explaining the current investment through the lagged values of the remaining variables, we only exclude the years after 2009. 14 Results are presented in Table 4.
Overall, the results do not change significantly. The most important changes are profitability and debt, which become statistically significant in all four models, both positively influencing real investment. The cost of capital loses its statistical significance, albeit maintaining a negative coefficient. 15 Financial receipts and financial payments maintain their statistical significance and their positive and negative coefficients, respectively. However, interest receipts and dividend payments are now also statistically significant with a positive and negative coefficient, respectively. This suggests that all financial investments by NFCs were beneficial to real investments in the pre-2007 crisis period and that the debt service and philosophy of 'shareholder value orientation' were both prejudicial. The global net effect of financialisation continues to exert a negative impact on real investment. However, the respective coefficients are more than double those of the same coefficients in the full period, suggesting that the deleterious effects of financialisation were more pronounced in the pre-crisis period. Indeed, an interaction of a dummy variable for the years of crisis (2009)(2010)(2011)(2012)(2013) with the global net effects of financialisation reveals that during these years these negative effects 14 Note that results are quite similar if we exclude the years after 2008 instead of the years after 2009. Results available upon request. 15 Once again, results do not change greatly if we exclude observations with short-term real interest rates. Results available upon request.
were smoother but not sufficient to cancel them in full. 16 This is also reported by Tori and Onaran (2015), who reiterate that the negative effects of financialisation on investment were stronger in the pre-crisis period. Second, we re-estimate the four specifications of the investment function excluding some countries that are outliers in order to avoid strongly biased estimates. We apply the traditional strategy in two steps in order to detect outliers (Chirinko, Fazzari, and Meyer 1999;Orhangazi 2008b). We begin by calculating the average of each country in relation to all independent variables. Then, countries whose averages lie in the 1 per cent or the 99 per cent tail of distribution of these variables were excluded. 17 Following this approach, Belgium, Denmark, France, Greece, Luxembourg and Sweden were not included. Results without these countries are presented in Table 5.
Once again, results do not suffer considerable alteration. Nevertheless, the cost of capital loses its statistical significance in all models albeit maintaining a negative coefficient, as opposed to the level of debt, which gains statistical significance in all models, exerting a marginal positive effect on investment. 18 Financial receipts maintain a positive significant impact on investment, whilst financial payments maintain their negative significant impact. The global net effect of financialisation (through net interest payments and net dividend payments, respectively) continues to influence real investment negatively, albeit with a higher magnitude in comparison to results for all countries. This can reveal that the prejudicial effects of financialisation are worse in the remaining countries than in the outlier countries. This conclusion is supported by a dummy interaction for outlier countries with the global net effects of financialisation. This interaction proved to be positive and statistically significant; reinforcing that in these countries the negative effects are not so strong. 19 These previous analyses seem to confirm that our results are relatively robust in relation to the presence of both years of crisis and outlier countries, mainly with regard to the impact of financialisation on real investment. This conclusion is reinforced if we exclude simultaneously years of crisis and outlier countries, as demonstrated in Table 6.
Here, the most important finding is related to the global net effect of financialisation, which remains as a negative and significant determinant of real investment, albeit with a higher intensity. Effectively, the respective coefficient is around three times that of the same coefficient for all years/countries, which confirms our previous results demonstrating that the effects of financialisation on real investment are worse for the full period and all countries. Our next step is to assess whether financialisation has affected different countries in the same manner and/or to the same degree, taking into account the heterogeneity existing within EU countries in relation to three different characteristics shared between them; namely, the type of financial system ('market-based' and 'bank-based' countries), their presence in the euro area (euro area and non-euro area countries) and their model/ variety of capitalism ('liberal market countries', 'coordinated market countries', 'hybrid/ mixed market countries' and 'transition countries'). 20 In all cases, we create dummy 17 As pointed out by Chirinko, Fazzari, and Meyer (1999), Greene (2000) and Orhangazi (2008b), we do not take into account outliers in relation to the dependent variable in order to avoid a censored regression bias. The average of all countries with regard to the dependent variable falls between 1 per cent and 99 per cent of the distribution, which suggests that there is no outlier in relation to the dependent variable. 18 Note that results would be quite similar excluding observations with short-term real interest rates. Here, the most important change is related to the variable of financial receipts that loses its statistical significance, albeit maintaining a positive coefficient. Results available upon request. 19 Results available upon request. 20 As advocated by Cingolani (2013), Sawyer (2013) and Botta, Caverzasi, and Tori (2015), we recognise that the dichotomy between 'market-based' and 'bank-based' is objectionable by underestimating other important particularities in the functioning of financial system (e.g. types of banks and/or the role of banks in the money-creation process). Here, we use this Table 5. Estimations of the investment function without outlier countries (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013) 0.000 * * * 0.000 * * * 0.000 * * * 0.000 * * * Note: Standard errors in (), z-statistics in []. Coefficients, standard errors and z-statistics for the year dummies are not reported. * * * Statistical significance at 1% level. * * Statistical significance at 5% level. * Statistical significance at 10% level. distinction because this is the traditional way of clustering different countries in terms of their financial systems as presented in the literature.
variables and interact these with the global net effect of financialisation (Model III) in order to determine whether there is any relationship between the effect of financialisation on real investment and these specific characteristics of countries. 21 Table 6. Estimations of the investment function without outlier countries (1995)(1996)(1997)(1998)(1999)(2000)(2001)(2002)(2003)(2004)(2005)(2006)(2007)(2008)(2009) Against this backdrop, six dummy variables were created. First, we created a dummy variable to reflect 'market-based' countries (D Market−based ). 22 Second, we constructed a dummy variable to take into account euro area countries (D Euro Area ). 23 Third, four further dummy variables were constructed translating the model/variety of capitalism of each country (D Liberal , D Coordinated , D Hybrid and D Transition ). 24 Results of these interactions are illustrated in Tables 7-12. D Market−Based takes the value 1 for Belgium, Finland, France, Netherlands, Sweden and the UK; and the value 0 for the remaining countries. The distinction between 'market-based' and 'bank-based' countries was made following Bijlsman and Zwart (2013) and Haan, Oosterloo, and Schoenmaker (2015).  Hall and Soskice (2001) and other related works in the field of comparative political economy.
We do not find any statistically significant difference in relation to the impact of financialisation on real investment according to the type of financial system. This result is consistent with the literature, according to which both financial systems have proved to be supporters of financialisation and give a weak boost to investment (Sawyer 2015). In fact, Levine (2002) concludes that neither 'market-based' nor 'bank-based' financial systems are particularly effective at promoting economic growth. In relation to the interaction for the Eurozone countries, the respective coefficient is only statistically significant in the pre-crisis period and without outlier countries. In the remaining three cases, it is almost significant at the 10 per cent significance level. However, the coefficient is positive in all four cases, which gives a tenuous indication that the negative impact of financialisation in the euro area countries could be less.
Moreover, in all four cases the positive coefficient of the interacted term is not sufficient to compensate the negative coefficient of the non-interacted term, confirming that financialisation also exerts a negative impact (albeit smoother) in the euro area countries.
Finally, and regarding the model/variety of capitalism, we do not find any statistically significant difference on the effects of financialisation. In general terms, this analysis shows us that there are no significant differences of the impact of financialisation on different groups of countries, which shows that EU countries are all negatively affected to the same degree despite their institutional differences.
Finally, we present the economic significance of our statistically significant estimates (McCloskey and Ziliak 1996;Ziliak and McCloskey 2004) for Model I and Model III and for the full period and for the pre-crisis period. This allows us to identify to what extent financialisation could explain the slowdown of real investment in EU countries since 1995. Results are presented in Tables 13 and 14.
In the period from 1995 to 2013, the cost of capital was the main driver of the rate of investment, whilst output growth had the most severe impact. In fact, the fall in real interest rate favoured acceleration in the investment rate by around 16-17 per cent. The slowdown of output contributed to a decline in the investment rate by about 19-20 per cent. The decrease in financial receipts was prejudicial to real investment. The investment rate would have been higher by about 4 per cent if there had not been a fall in financial receipts. This supports our previous conclusion that the NFCs of EU countries use returns from financial investment to fund real investment, excluding the 'crowding out' effect. The decrease in these returns was therefore harmful for real investment. This effect did not compensate for the concomitant decline in financial payments, which only contributed to an increase of investment rate by about 0.4 per cent. In fact, the global net effect of financialisation was marginally negative, being responsible for a decline in investment rate of around 1 per cent. In the pre-crisis period, the global net effects of financialisation on real investment were relatively worse. The investment rate would have been higher by about 8 per cent without an increase in net financial payments. During that time, financialisation was the main driver for the slowdown of investment rate.

Conclusion
This article sought to determine whether financialisation has had beneficial or prejudicial effects on real investment in EU countries by conducting a panel data econometric analysis for 27 EU countries from 1995 and 2013, using macroeconomic annual data.
Contrary to mainstream economics, the literature on financialisation refers to two channels through which the increasing importance of finance could be disruptive to the real investment of NFCs. First, the rise in financial investments made by NFCs diverts funds from productive investments, causing a type of 'crowding out' effect on real investment. Second, the funds available to support real investments have decreased due to the strong pressure from shareholders on NFCs for financial payments in the form of interest, dividends and/or stock buybacks.
We estimate an investment equation to describe real investment using macroeconomic annual data through the standard variables (lagged investment, profitability, debt, cost of capital, corporate savings and output growth) and two other variables to reflect the two channels of financialisation (financial receipts and financial payments).
Using a dynamic macro panel, we estimated the investment equation using the LSDVBC estimator. We identified financial payments as a negative determinant of real investment in EU countries, in accordance with the predictions of the literature on financialisation and with other empirical studies around this issue (Stockhammer 2004;van Treeck 2008;Orhangazi 2008aOrhangazi , 2008bOnaran, Stockhammer, and Grafl 2011;Tori and Onaran 2015;Barradas and Lagoa 2017). Both interest and dividend payments negatively influence real investment, especially in the pre-crisis period. Nevertheless, financial receipts of dividends exert a positive impact on real investment, contrary to the financialisation literature. This seems to illustrate that NFCs in EU countries are using financial incomes to fund real investments, which excludes the 'crowding out' effect that can be explained by the large number of small and medium-sized EU corporations that face high funding constraints and therefore use all incomes (even financial) to fund new investments. This is a traditional result in studies that use micro-data at the corporation level, where the effect of financial receipts on investment in the case of small and medium-sized corporations is regularly positive (Orhangazi 2008b;Tori and Onaran 2015). However, this positive effect linked to financial receipts does not compensate for the negative effect from financial payments. In net terms, the process of financialisation has been prejudicial to real investment of NFCs in EU countries.
We also conclude that the prejudicial effects of financialisation on real investment were far more severe in the pre-crisis period during which financialisation was the main driver for the slowdown of investment in the EU. We do not find any statistically significant difference in relation to the impact of financialisation on real investment for 'marketbased' countries, Eurozone countries, 'liberal market countries', 'coordinated market countries', 'hybrid/mixed market countries' and 'transition countries'. This seems to suggest that the harmful effects of financialisation are similar and transversal to all EU countries despite their institutional differences.
Our findings also suggest that the prejudicial effects of financialisation on real investment are not peculiar to the most developed and financialised economies, such as the US 0.000 * * * 0.000 * * * 0.000 * * * 0.000 * * * Note: Standard errors in (), z-statistics in [] * * * * * . Coefficients, standard errors and z-statistics for the year dummies are not reported. * * * Statistical significance at 1% level. * * Statistical significance at 5% level. * Statistical significance at 10% level. and the UK. Instead, it seems to be a generalised phenomenon that negatively affects most EU countries over time.
A possible extension of this work could be the use of different measures to capture the two channels of financialisation. The proportion of financial assets to total assets of NFCs and the level of payout ratios of NFCs appear to be two interesting alternatives.
This article also does not take into account the effects of globalisation on real investment of NFCs, which could represent a limitation of our study and a good starting point from which to perform further research around this issue. The idea of this channel is that NFCs from developed countries are increasing their external activities in low-wage (developing) countries in order to have both lower production costs and taxation and higher profit margins, which also impair the real investment at a domestic level (Durand and Gueuder 2016;Fiebiger 2016).
Finally, future research on this topic could focus on assessing the causes and determinants of financialisation of the NFCs of EU countries, following the approach developed by Akkemik andÖzen (2014) and/or Soener (2015). In fact, it seems reasonable to understand first why the NFCs of EU countries are financialised in order to derive policy implications that could mitigate these prejudicial effects of financialisation on the slowdown of real investment. Note: The long-term coefficient is obtained by dividing the short-term coefficient (estimated coefficient) and one minus the coefficient of the autoregressive estimation (estimated lagged investment coefficient). The actual cumulative change corresponds to the growth rate of the correspondent variable. The economic effect is the multiplication of the long-term coefficient by the actual cumulative change. Note: The long-term coefficient is obtained by dividing the short-term coefficient (estimated coefficient) and one minus the coefficient of the autoregressive estimation (estimated lagged investment coefficient). The actual cumulative change corresponds to the growth rate of the correspondent variable. The economic effect is the multiplication of the long-term coefficient by the actual cumulative change.