Evolving Corporate Financing Patterns in Europe: Is There Convergence?

Abstract

One of the main objectives behind the formation of the European Union (EU) is attainment of financial integration among member countries. Greater financial integration is expected to facilitate financial sector efficiency, macroeconomic stability and effective implementation of monetary policy in the EU (Trichet 2006).

These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.

5.1 Introduction

One of the main objectives behind the formation of the European Union (EU) is attainment of financial integration among member countries. Greater financial integration is expected to facilitate financial sector efficiency, macroeconomic stability and effective implementation of monetary policy in the EU (Trichet 2006). While there have been several attempts at assessing the progress of financial integration in the EU,¹ one important aspect of the process, namely corporate financing, has been largely under-researched. This paper attempts to investigate convergence in corporate financing patterns in the EU and thereby provide insights into the larger issue of European integration. In this respect, the paper extends the work of Murinde et al. (2004), which tested for convergence in corporate financing patterns in the EU using a 1972–1996 dataset, by using more recent data for 1972–2004 and by studying a slightly larger set of EU countries, namely Finland, France, Germany, Italy, The Netherlands, Spain, Sweden and the UK.² In addition, this study invokes more innovative econometric techniques by using modern panel unit root tests, following Evans and Karras (1996), and further by employing more appropriate (GMM) methodology for testing convergence in panel data, following Islam (1995) and Nerlove (1996).

In a previous study, Murinde et al. found little evidence of convergence in bank and bond finance and some evidence of convergence in equity and internal corporate finance, with strong growth in the latter in line with other previous studies (Bertero 1994; Corbett and Jenkinson 1996). Our extended dataset, as summarized in Fig. 5.1, indicates that the EU has continued to witness convergence in corporate financing with a clear shift from bank financing towards market based financing. The graphs suggest an ongoing switch from bank to equity and bond debt finance and indicate that internal finance is no longer growing strongly. In this paper we undertake formal empirical testing to confirm the above convergence.

Fig. 5.1

Evolving corporate financing patterns in the EU Note: Mean values are plotted for each year. Leverage refers to the ratio of debt (bank + bond) financing to equity financing; bank to market financing ratio refers to the ratio of bank financing to market (equity + bond) financing

Mullineux (2007, 2010) inter alia examines the impact of financial sector convergence, postulating the evolution of a hybrid model in which financial conglomerates have evolved in the US and Japan (where they were prohibited by regulations until the regulatory reform in those countries in the mid to late 1990s) similar to the prevailing European universal banking and bancassurance models. And capital markets have become more important in Europe and Japan, leading to some catch up with the US (and the UK). The euro denominated corporate bond market has grown rapidly, since the introduction of the euro in 1999 and overtook the US dollar (USD) denominated market in the middle of the subsequent decade. By 2007 it was approximately 50% bigger in the USD market.

Our findings confirm the growth in the NFC bond financing and suggest that it is at the expense of bank financing. A similar disintermediation (‘securitisation’) involving a switch by large corporates (but not bank dependent SMEs) from bank loan to bond (and shorter term rate and commerce paper) finance can be seen in Japan and the US. The situation of medium sized companies is under-researched and more complex as they are gaining access to alternatives to bank finance through the issuance of ‘junk’ (below investment grade) bonds and venture capital. However our data set pertains to NFCs of all size classes since the disaggregated data by asset size was not available.

The observed convergence on a hybrid model (‘hybridisation’) consisting of large global financial conglomerate banks (and smaller local banks and specialist institutions), is increasingly market orientated and thus a less bank orientated (insider) and more market orientated (outsider) system of corporate governance is evolving (Mullineux 2007). Precisely, corporate governance is becoming more Anglo-Saxon. However, the term ‘Anglo-Saxon’ itself is a bit of a misnomer if it is taken to include both the US and the UK since their corporate governance models are very different (Mullineux 2010) and indeed the UK system is now closer to that of the rest of Europe, in terms of the role of institutional shareholders in the corporate governance system, than to the US; which is increasingly looking like an outlier. Indeed since the passing of the Sarbanes-Oxley Act in 2002, which was necessary to underpin internal controls after the Enron debacle, New York appears to be losing business to London, which has a lighter corporate regulatory and corporate governance regime. Regulation in the UK is conducted by the Financial Services Authority on a ‘risk-based’ basis. London’s success is not entirely driven by its low regulatory ‘tax’, it is also at the centre of the rapid growth in euro denominated financial markets and is laying claim to the Islamic bond market too. The lack of institutional shareholder proxy voting rights in the US may also be putting New York at a disadvantage and the regulator, the Securities and Exchange Commission was reviewing the situation in 2007.

More generally, financial sector integration is being encouraged in the EU as a way of improving financial service provision, deepening markets and reducing the cost of capital, including the cost of payments and settlement. The introduction of the euro in 1999, aimed to facilitate this process, along with the Financial Services Act Plan (FSAP) adopted by the European Council in March 2000. Thirty eight of its 42 members had been widely adopted by the end of 2004. There is some frustration about the continuing high costs of cross-border payments, but the EC is pressing for a single European Payment Area (SEPA) to be completed before the end of the decade. The European Central Bank (ECB) is seeking permission to launch a Europe-wide system itself, as a means of bypassing the obstacles. It should be noted that the FSAP and SEPA cover retail banking initiatives as well as corporate finance (cost of capital and money and capital market liquidity) issues. European regulatory convergence was achieved to a large extent before the end of the 1990s. Revisions are underway in preparation for the implementation of Basel II and there remains a concern that the country level regulatory and supervisory authorities may not implement the EU level regulation conformably. Significant tax differences also remain.

The development of a single European financial market, to the extent that remains possible under globalization, also has implications for the conduct of monetary policy (Trichet 2006). Changes in interest rates have more similar effects in the various parts of the EU. Harmonisation of home loan markets may reduce financial instability and lower costs of capital may increase investment and growth. Therefore, financial integration in the EU can be expected to boost economic growth and aid financial stability. In this context, the issue of whether financial systems across EU countries are converging becomes important.

The remainder of the paper is organized as follows. Section 5.2 provides an overview of empirical modeling of convergence. Section 5.3 presents the results from panel unit root tests of convergence followed by the results from GMM regressions in Sect. 5.4. Finally, Sect. 5.5 summarizes and concludes the paper.

5.2 Modeling Convergence

Bulk of the literature on convergence can be traced back to Barro and Sala-i-Martin (1992) who developed regression based tests for growth convergence. Subsequent developments in the convergence literature have proceeded in two broad directions. The first is the approach of Evans and Karras (1996) who developed a formal test of convergence that is based on panel unit root tests. The second is the work by Islam (1995) and Nerlove (1996) who extended the Barro and Salai-i-Martin framework to allow for testing of convergence in a panel framework. We conduct both the above types of tests in this paper and hence provide a brief overview of each.

While the classical growth regression approach is quite popular in the literature, it has faced criticism on account of ignoring time-series properties of the data. Evans (1996) recommended exploiting both the time-series and the cross-section information provided by panel data in order to evaluate the convergence hypothesis. Evans and Karras (1996) showed that, economies can be said to converge if and only if there exists a common trend such that E_t (y_n,t+1 − a_t+1) = μ _n, where a _t is the common trend and μ _n is a constant. Evans and Karras (1996) posited this question as a test of stationarity of the mean-differenced series, y _n,t+i  − ${\bar{y}_{t + i}}$. In this paper we employ a variety of modern panel unit root tests to conduct tests of convergence on the mean-differenced data on bank financing, equity financing, bond financing and internal financing respectively for the EU countries in our data-set where we consider the share of each mode of financing in total financing as the relevant endogenous variable.

For conducting panel unit root tests, we first use the test given by Levin, Lin and Chu (LLC 2002). This is essentially a pooled Augmented Dickey-Fuller (ADF) test but is general enough to allow for individual fixed effects as well as time effects. Next we employ Breitung’s (2000) test which is similar to LLC expect for the way in which it uses proxies to estimate the auto-regressive coefficients. However the major limitation of these tests is that each cross-section in the panel is assumed to share the same auto-regressive coefficient. Thus rejection of the null of non-stationarity implies that the rate of convergence is same across all units. This assumed homogeneity of the unit root was overcome by Im et al. (2003, 1997) who estimated individual-specific ADF tests and then computed the mean of the different t-statistics. Thus, the IPS test does not assume that all series are stationary under the alternative, but is consistent under the alternative that only some of the series are stationary. Therefore, we employ the IPS test as a robust means of testing our convergence hypothesis. We also use the Fisher-ADF test and the Fisher-PP test (Madalla and Wu 1999; Choi 2001). The Fisher tests are similar to the IPS test in the sense that they combine independent unit root tests (conducted as chi-square tests in this case) of the individual series.

Finally, we apply a stationarity test, viz. the Hadri test (Hadri 2000) which is a residual based Lagrange multiplier test with the null hypothesis of stationarity of the series. This test has high power and has the advantage of being robust to non-normality. We also provide results from a variant of the Hadri test that is heteroscedasticity consistent. In sum, we employ five panel unit root tests and two panel stationarity tests.

However the above framework provides only an examination of unconditional convergence where different initial conditions among the countries cannot be controlled for. The classical growth regression approach of Barro and Sala-i-Martin allows testing for conditional as well as unconditional convergence and the original framework was extended by Islam (1995) and Nerlove (1996) to fit panel data. Their framework starts by defining a Cobb–Douglas production function, y = A^1−αk^α, where y is the per capita output, k is the capital intensity and A denotes productivity. Following the Solow model’s equality between savings and investment, we can derive the following formulation:

$$ \log {y_t} = \frac{{\alpha (1 - \gamma )}}{{1 - \alpha }}[\log s - \log (n + g + \delta )] + (1 - \gamma )\log {A_t} + \gamma \log {y_{t - 1}} $$

(5.1)

where s is the savings rate, n is the population growth rate, g is the productivity growth rate and δ is the depreciation rate of capital stock. It can be shown that the rate of convergence of per capita output to the equilibrium level is inversely proportional to γ. If γ is smaller than 1 then there is such convergence, and convergence speed increases as γ decreases. The model yields the following estimable formulation:

$$ {y_{it}} = \gamma {y_{it - 1}} + x_{it}^\prime \beta + {\mu_i} + {u_{it}} $$

(5.2)

where t = 1 to T, represents year and i = 1 to N represents country; y _it stands for GDP per capita for country i in year t; x _it stands for all the determinants of growth; μ _it represents the country-specific effects; and u _it is white noise.

In this paper, our objective is to adopt the above framework for testing convergence in growth rates of corporate financing patterns of EU countries. The flow of funds data that we use decomposes the aggregate economic activity of a country to the flow of funds in the government, private, household and financial sectors. Therefore, a convergence in national economic growths does not automatically imply that there will be convergence in each of the disaggregated sectors of the economy. This motivates the modeling of convergence in the disaggregated components of the aggregate economy, i.e. financial sector in our case. Analogous to the neo-classical production function that is typically assumed for the macro-economy, we conceptualise the economic activity in the financial sector in terms of corporate financing being produced by employing different inputs such as those implied by monetary policy and other control variables (Murinde et al. 2004). Therefore, we replace GDP per capita in the traditional growth model by the types of corporate financing. In other words, in (5.2) we replace y_it by the share of corporate financing from a particular source. Consequently, we are able to test for convergence among EU countries in terms of their corporate financing patterns.

However the main problem with the model outlined in (5.2) is that the lagged dependent variable y_it−1 and the country-specific effects μ_it are correlated, which means that the usual panel estimators are biased and inconsistent. The Generalized Method of Moments (GMM) methodology is a convenient means of estimating this model where instrument variables are used for y_it−1 and moment conditions are exploited in the estimation (Hansen 1982; Arellano and Bond 1991). In this paper we follow Arellano and Bover’s (1995) methodology of orthogonal deviation that removes the unobserved country-specific effects. The orthogonal deviation transformation expresses each observation as the deviation from the mean of future observations for the same country and it weights each deviation to standardize the variance. The advantage of using this transformation is that the transformed errors will be serially uncorrelated and homoskedastic.

We apply the above methodology to estimate four different equations. First, we estimate the convergence model for bank financing based on the following equation:

$$ BAN{K_{it}} = \gamma BAN{K_{it - 1}} + {\beta_1}BM{Y_{it}} + {\beta_2}E{R_{it}} + {\beta_3}I{R_{it}} + {\beta_4}OPE{N_{it}} + + {u_{it}} $$

(5.3)

where BANK is bank financing by the NFCs, BMY is a financial deepening variable calculated as the ratio of money supply to GDP, ER is the nominal exchange rate, IR is the nominal interest rate and OPEN is a measures of the degree of openness calculated as the ratio of exports and imports to GDP. While the control variables BMY and IR are proxies for monetary policy and are consistent with the idea of monetary convergence as stipulated by the European Commission (Murinde et al. 2004), ER and OPEN are expected to control for the influence of trade policy and terms of trade on corporate financing. In the above equation, if the estimated γ turns out to be less than one, then we can deduce that there is convergence in bank financing by the NFCs across the countries in our sample and over the time period considered. Moreover, the inverse of γ indicates the speed of convergence.The second equation that we estimate is based on the role of equity markets in providing finance to NFCs:

$$ EQUIT{Y_{it}} = \gamma EQUIT{Y_{it - 1}} + {\beta_1}BM{Y_{it}} + {\beta_2}E{R_{it}} + {\beta_3}I{R_{it}} + {\beta_4}OPE{N_{it}} + + {u_{it}} $$

(5.4)

where EQUITY is equity financing by the NFCs and the control variables are the same as before. We also estimate an equation based on bond financing of NFCs:

$$ BON{D_{it}} = \gamma BON{D_{it - 1}} + {\beta_1}BM{Y_{it}} + {\beta_2}E{R_{it}} + {\beta_3}I{R_{it}} + {\beta_4}OPE{N_{it}} + + {u_{it}} $$

(5.5)

where BOND is bond financing by the NFCs and the control variables are the same as before. Finally, we test convergence in the use of internal finance by NFCs by estimating the following equation:

$$ INTER{N_{it}} = \gamma INTER{N_{it - 1}} + {\beta_1}BM{Y_{it}} + {\beta_2}E{R_{it}} + {\beta_3}I{R_{it}} + {\beta_4}OPE{N_{it}} + + {u_{it}} $$

(5.6)

where INTERN is internal financing by the NFCs and the control variables are the same as before.

The data for this study are taken from the OECD flow-of-funds tables and covers the period 1972–2004 for eight EU member countries, viz. Finland, France, Germany, Italy, the Netherlands, Spain, Sweden and the UK. We define the financing variables as percentages of the total. The data on macroeconomic variables are collected from IMF’s International Financial Statistics database. Mean values of the corporate finance data are plotted in Fig. 5.1. The evolving pattern provides preliminary indication of a shift from bank based financing to market sources. Table 5.1 presents a correlation matrix of all the main variables that we study.

Table 5.1 Correlation matrix of corporate financing and macroeconomic variables for eight EU countries, 1972–2004

5.3 Results from Panel Unit Root Tests

Panel unit root tests have the advantage that they take account of time-series properties of the variables while examining convergence. Thus we employ a variety of panel unit root tests for convergence in corporate financing in the EU. Accordingly we conduct five panel unit root tests and two panel stationarity tests on each corporate financing variable and the results for all the eight countries in our sample are presented in Table 5.2. For bank financing, the results are overwhelmingly in favour of convergence as the null hypothesis of non-stationarity of the data gets rejected by all the five unit root tests (albeit marginally for the LLC test at the level of 10%). Furthermore, the Hadri tests of stationarity do not reject the null hypothesis of stationarity. However for equity financing, the results do not indicate convergence. Although the Fisher-PP test rejects the null hypothesis, however the rest of the tests do not provide evidence to suggest a convergence in equity financing among the sample countries.

Table 5.2 Panel unit root and stationarity tests: results for eight EU countries, 1972–2004

The bond financing results are in favour of convergence as indicated by the unanimous rejection of the null hypothesis in all the unit root tests. It may be noted that although the Hadri test rejects the null hypothesis of stationarity at the 10% level, the heteroscedasticy consistent version of the test provides evidence for convergence. The results for internal finance are not entirely unanimous. While the LLC test fails to reject the null hypothesis of non-stationarity, the Fisher-PP test rejects it only at the 10% level of significance. However all the other tests provide results to indicate convergence in internal financing. In sum, the above tests strongly indicate that there has been convergence in the corporate financing patterns of the EU countries in terms of bank and bond financing. Our results provide weak evidence in favour of convergence in terms of equity and internal financing.

We also conduct the panel unit root tests for only the countries that have adopted euro as their currency. The results are presented in the Appendix Table 5.6. The results are almost similar to those obtained for the full sample, except for mixed results in the case of equity financing. Hence, the convergence hypothesis appears to hold for bank finance and bond finance whereas the results are mixed for equity finance and internal finance (the Hadri tests of stationarity indicate convergence whereas the unit root tests do not). These results reinforce our earlier findings for the EU countries. Hence, our results strongly suggest that NFCs in euro countries have converged in terms of their bank and bond financing patterns, whereas the results for equity and internal finance are mixed. Finally we conduct the panel unit root tests on our sample only for the countries that were EU members from the beginning of our data period, i.e. 1972. Hence, we leave out the new entrants, viz. Finland, Spain and Sweden, from our sample. The results are presented in Table 5.7. These results suggest that these countries exhibited convergence in terms of bank finance and bond finance, thereby re-affirming our previous results.

$$ \Delta CORPFI{N_{it}} = \beta CORPFI{N_{t - 1}} + \sum\limits_{j = 1} {{\lambda_j}\Delta CORPFI{N_{t - j}} + {\alpha_i} + {u_{it}}}, $$

where CORPFIN is the mean differenced endogenous variable for corporate financing measured as bank financing, equity financing, bond financing and internal financing by NFCs as a ratio of total financing. We employ a variety of modern panel unit root tests based on the above formulation. Probabilities for Fisher tests are computed using an asymptotic Chi-square distribution. All other tests assume asymptotic normality.

To summarize, our panel unit root tests indicate that the EU countries have exhibited convergence in their corporate financing patterns in terms of their bank financing and equity financing. This pattern of convergence has been consistent across the countries that have adopted the euro as their currency as well as for the founder EU member countries.

5.4 Results from GMM Regressions

While the above tests examined financial convergence only in an unconditional sense, we now move to the formal testing for convergence based on the modification of the classical regression approach as outlined in (5.3)–(5.6). These set of regressions, based on the dynamic panel GMM methodology, allow us to assess unconditional as well as conditional convergence. The results of the estimation of (5.3)–(5.6) all the eight countries in our sample are presented in Table 5.3. The coefficients of the lagged financing variables are less than one in all cases except for internal financing. This indicates that there has been convergence in corporate financing patterns in terms of bank, equity and bond financing across the eight countries in our sample over the period 1972–2004. However, the speed of convergence varies across the source of finance. Considering un-conditional convergence, bond finance appears to have exhibited the quickest convergence followed by equity and bank finance in that order. This pattern is repeated even when factors affecting financial convergence are controlled for, i.e. in the case of conditional convergence once again it is bond finance that exhibits the fastest convergence followed by equity and bank finance in that order. Our results indicating slow convergence of bank finance are comparable with the results of Murinde et al. (2004) who observed a lack of convergence in financing from this source.

Table 5.3 GMM estimation results for eight EU countries 1972–2004

Therefore, based on a more recent and expanded dataset, we observe that the EU countries have begun to converge in terms of the use of bank financing by NFCs, although the speed of convergence is the slowest for this source of finance. Another interesting finding is that EU countries exhibit convergence in bond financing and in fact this variable shows the fastest conditional as well as unconditional convergence, whereas Murinde et al. (2004) did not observe any convergence in financing from this source. This indicates that in recent years, NFCs in European countries have shown a tendency to source similar proportions of their total financing requirements from the bond markets. Our GMM results indicate convergence in equity financing whereas the panel unit root tests do not suggest convergence in this source of financing. Hence, while EU countries did not exhibit a common trend in terms of equity financing, there was convergence in the sense suggested by the growth regression approach, i.e. countries with lower initial levels of equity financing exhibited higher growth in financing from this source.

The role played by introduction of the euro in 1999 in the patterns of corporate financing is examined by including a dummy variable (EURODUM) for the year 1999 in the above specifications.³ For the bank finance models, the results for the impact of the introduction of the euro suggests that while there has been convergence in bank financing, the introduction of the euro has actually led to an increase in dependence on bank borrowings. Similarly, the coefficient of the dummy variable is positive and significant in all cases of the equity financing and bond financing models suggesting that NFCs in the European countries increased their financing from equity issues and bond markets subsequent to the introduction of the Euro. However the introduction of the euro appears to have reduced the dependence on internal financing. These results indicate a convergence towards a variant of the Anglo-Saxon model of corporate financing characterized by increased importance of market based sources of finance and reduced role of internal finances in providing funds to the NFCs.

Whether entry of a country into the EU mattered in terms of the patterns of corporate financing is examined by including a dummy variable (ENTRYDUM) for the years of EU entry in the above specifications. While the results for the bank finance, bond finance and internal finance models do not show any impact of entry, the coefficient of the dummy variable is negative and significant in almost all specifications of the equity finance models. These results indicate that entry into the EU was characterized by NFCs of member countries reducing their dependence on equity financing. One possible explanation for this is that membership of the EU provided countries with an immediate reduction in risk premium thereby making debt financing less expensive.

We also re-estimate the model specifications only for the countries that have adopted the euro thus leaving out Sweden and UK in these set of estimations. The results from these estimations are reported in Table 5.4. The results are almost the same as obtained for the entire sample earlier. Hence, the euro countries have exhibited both unconditional and conditional convergence in their corporate financing patterns. Bond finance appears to have exhibited the quickest convergence, in this case followed by bank finance and equity finance in that order.

Table 5.4 GMM estimation results excluding non-Euro EU countries, 1972–2004

We then re-estimate the model specifications only for the countries that were EU members from the beginning of our data period, i.e. we leave out Finland, Spain and Sweden, from our sample. See Table 5.5. Once again we observe that there has been convergence in corporate financing patterns across this sample of countries, including internal financing in this case. Considering un-conditional or conditional convergence, bond finance appears to have exhibited the quickest convergence followed by equity finance, bank finance and internal finance in that order.

Table 5.5 GMM estimation results for expanded EU, 1972–2004

5.5 Concluding Remarks

Recent studies, based on micro as well as macro level approaches have shown that the EU is undergoing financial integration (Baele et al. 2004; Gaspar et al. 2003; Kiehlborn and Mietzner 2005). In this context, the present paper examines a particular aspect of the financial integration process, namely corporate financing patterns. We examine convergence in the corporate financing patterns of European countries during the period 1972–2004. Employing a number of modern panel unit root tests, we find evidence for convergence in bank and bond finance, but we do not obtain unanimous results for equity finance and internal finance. We then apply the dynamic panel variant of the traditional growth regression approach.

Our results suggest that NFCs in Europe are converging in terms of the proportion of funds they access from banks, equity issues and bond markets. In sum, it appears that financial integration in EU has been characterized by NFCs increasingly taking recourse to bond and equity markets for their financing needs. Hence to some extent this indicates a move from bank-based financing to the Anglo-Saxon mode of market based financing. Whether this has also been accompanied by a shared reduction in internal financing is however not consistently borne out by our results.

Appendix

Table 5.6 Panel unit root and stationarity tests: results excluding non-Euro EU countries, 1972–2004

Table 5.7 Panel unit root and stationarity tests: results for expanded EU, 1972–2004