Standard-Nutzungsbedingungen: Die Dokumente auf EconStor dürfen zu eigenen wissenschaftlichen Zwecken und zum Privatgebrauch gespeichert und kopiert werden. Sie dürfen die Dokumente nicht für öffentliche oder kommerzielle Zwecke vervielfältigen, öffentlich ausstellen, öffentlich zugänglich machen, vertreiben oder anderweitig nutzen. Sofern die Verfasser die Dokumente unter Open-Content-Lizenzen (insbesondere CC-Lizenzen) zur Verfügung gestellt haben sollten, gelten abweichend von diesen Nutzungsbedingungen die in der dort genannten Lizenz gewährten Nutzungsrechte. Terms of use: Documents in EconStor may be saved and copied for your personal and scholarly purposes. You are not to copy documents for public or commercial purposes, to exhibit the documents publicly, to make them publicly available on the internet, or to distribute or otherwise use the documents in public. licence. www.econstor.eu ECONOMIC AND FINANCIAL REPORTS are preliminary material circulated to stimulate discussion and critical comment. Quotation of material in the Reports should be cleared with the author or authors. If the documents have been made available under an Open Content Licence (especially Creative Commons Licences), you may exercise further usage rights as specified in the indicated The views expressed are those of the individual authors, and do not necessarily reflect the position of the EIB. Individual copies of the Reports may be obtained free of charge by writing to the above address or on-line from www.eib.org/efs/pubs.htm * We acknowledge comments from seminar participants at the EIB. Abstract While it is commonly believed that companies issue non-current debt in order to finance capital expenditures, the relationship among these two variables is in practice much more complicated, and it depends on the overall real and financial flows related to companies' activity. Looking at such flows reveals that internal sources are higher than capital expenditures for listed companies in France, Germany and Italy. UK companies on the contrary run a financial deficit. Yet, French companies issue more debt than UK companies do. The anomaly is explained by our econometric model, revealing that lagged leverage is the main determinant of debt issues. As French companies display the highest leverage, they also issue the most debt. Collateral is found to positively influence debt issues in all countries except France. There is also evidence that debt issuance by UK companies is positively affected by size and liquidity, and negatively affected by profitability. Size, liquidity and profitability are not found to affect issuance for continental companies, perhaps because these companies run a financing surplus. Do capital expenditures determine debt issues? Companies finance their investment each year out of a mix of internal and external sources, the former consisting mainly of retained earnings, the latter of debt. Investment accumulated over the life of the firm constitutes the firm's capital stock. Accumulated earnings and debt, together with shares issued over the life of the firm, determine its capital structure. While often the firm's main assets, the capital stock is by no means the only asset that needs to be financed. Cash, financial investments, other tangible as well as intangible assets are also financed by the capital structure of the firm and thus the issue of how investment is financed cannot easily be addressed. If one looks at companies' balance sheets, it is the total of the firm's assets that is financed by the total of its liabilities and shareholder's equity, that is, by its capital structure. Thus, investment in any given year is just one element (although a very important one) within a more general balance sheet optimisation problem. Taking a flow of funds perspective seems to be more promising. After all, investment is a "capital expenditure" that is ultimately financed by some cash outflow. Why not study the relationship between capital expenditures, internally generated funds, issues of new loans, etc. as reported in the cash flow statements? Unfortunately, these statements do not have unambiguous interpretation either. Capital expenditures may be financed through the sale of an existing asset, for instance. New debt can be issued to refinance expiring loans even in the absence of any investment activity. Thus, even in a cash flow statement, it is only the sum of sources that equates the sum of uses. Attempting to match individual sources to individual uses is unlikely to be appropriate. A healthy scepticism should however not deter us from analysing flow of funds data in the search for broad regularities in investment financing. We do so in this paper by relying on company-level sources, and by focussing our attention on listed companies in the UK, Germany, France, and Italy over the last decade. Indeed we do find broad regularities. Firstly, we observe that internal funds are by themselves sufficient to finance almost all capital expenditures in the UK and even exceed capital expenditures in the other countries. Debt and share finance play only a limited role in financing capital expenditures when compared to internal funds. This evidence is puzzling especially since an analysis of balance sheet data for our sample, revealing that debt in various forms accounts for between 30% and 40%, is consistent with the commonly held view that debt is primarily issued to finance investment. In the second part of the paper, we try to address the puzzle through an econometric analysis of the relationship between borrowing and investment flows. We are able to identify a robust statistical relationship between debt issues and capital expenditures, characterised by a reasonable fit and which is unlikely to be spurious. In particular, debt issues depend on investment in a way that is consistent with the consensus in the capital structure literature. 4 The paper is organised as follows. The first section describes the aggregate accounts of the companies in our sample and presents the most relevant summary statistics. In the second section, after proposing an empirical specification that is broadly consistent with the main theories of capital structure, we discuss the results from estimation of the model. The third and final section concludes. Excellent survey material on capital structure can be found in Harris and Raviv (1991) and the recent "red book" of the Deutsche Bundesbank and the Banque de France edited by Sauvé and Sheuer (1999). Therefore, we do not feel compelled to discuss the existing literature in general terms. Rather, we refer to selected specific papers as the need arises in the discussion. DESCRIPTION OF THE SAMPLE We concentrate on companies that are listed on major European stock markets and that report comprehensive cash flow data. As national characteristics are thought to be an important factor in corporate finance, we find it appropriate to conduct our analysis by country. This leaves us with a dilemma, as the number of listed companies varies greatly across countries and in particular there are insufficient numbers of observations to support an econometric analysis for all European countries. Four countries satisfy the joint requirement of providing a large enough number of companies that at the same time report a long enough history of data. These are the UK, France, Germany and Italy. Sample selection criteria There are three main criteria for a company to be included in our sample. Firstly, we recognize that financing decisions are taken at a group level, rather than at a subsidiary level. Hence, we require accounts to be reported at a consolidated level and exclude consolidated subsidiaries from the sample to avoid double counting. Thus, for instance, Volkswagen AG is included but not Audi AG, as the latter is consolidated into the former. There is one obvious conceptual difficulty with applying this criterion for listed subsidiaries of foreign companies. One would think for instance that a UK subsidiary of a US parent company does not take autonomous financing decisions and should therefore be excluded from the sample. In practice this does not turn out to be a cause for concern as the few cases that fall into this category fail to satisfy other sample selection criteria 1 . Our second selection criterion is based on availability of relevant cash-flow information for a sufficient number of years. All companies included for the UK and France report 1 An open question concerns the potential biases, which may result from omission of joint ventures and Special Purpose Vehicles, domestic or foreign. These entities are included in our sample only if they are consolidated into another listed company or they are listed independently. Otherwise, their omission implies a loss of information on the overall volume of the borrowing and investing activities of the companies in our sample, since these entities are typically set up to raise funds in order to finance specific capital expenditures. Whether a systematic bias in our analysis would arise is a question that deserves further research. 5 non-missing values for issues of debt and capital expenditures 2 over the period 1989-99. In the case of Italy and Germany the sample has been shortened to the period 1994-99 in order to obtain a sufficient number of companies. Thirdly, we trim from the sample a few companies based on the fact that they report negative net worth, and/or large outlier values for profitability and market to book 3 . Previous literature has restricted attention to companies in the mining, construction and manufacturing sectors, usually defined by their primary two-digit SIC sector code between 10 and 39 (a table of two-digit SIC sectors is reported in the Annex). For each country, we combine all companies across such sectors, and refer to this group as the "manufacturing group". Coupled with the above criteria, this yields 192 companies for the UK 4 , 60 for France, 36 for Germany and 43 for Italy. We decided however to look also at companies in the transport, utilities, trade and services sectors (two-digit SIC from 40 to 59 and from 70 to 89). In the UK, there are 120 such companies satisfying our criteria 5 . We refer to this second group as the "services group". An insufficient number of non-manufacturing companies are listed in other countries. We are thus left with five groups, namely UK "manufacturing" (UKM), UK "services" (UKS), French, Italian and German "manufacturing" (FRM, ITM, GEM respectively). All data are extracted from the Worldscope Database compiled by Bureau Van Dijk. Asset composition in the balance sheets The composition of assets for our five groups of companies is plotted in 2 Availability of other data, in particular from balance sheets, that are needed for the estimation turns out not to be restrictive for companies that report complete cash flow data. We will explain below how debt issues include all issues of bonds, loans, capitalized leasing, etc. with maturity at issue above one accounting period. 3 Using a method presented and compared with alternative methods in Kremp (1995), firms with observations for profitability and market to book outside the interval defined by the first and third quartiles minus or plus five times the inter-quartile range were discarded. 4 Of these, 160 are in the manufacturing sector (two-digit SIC 30-39), 30 in construction and only 2 in the mining sector. The prevalence of manufacturing is even more important in the other countries. Debt instruments of various kind and maturities account for 30% to 40% of the balance sheet in The long-term debt data at our disposal lump together all interest bearing financial obligations excluding amounts due within one year. Included are thus mortgages, bonds, long term bank overdrafts, medium term loans, capitalized lease obligations, revolving credit, etc. Excluded are current portion of long term debt, pensions, deferred taxes and minority interest. Thus there is no way to distinguish between market and bank debt. Naturally, beyond the broad similarities across countries, there are also differences that are driven by country specific features. Firstly, the proportion of equity is much higher for the UK, amounting to over 45% of the total balance sheet, compared to approximately 35% in continental Europe, despite the overall de-leveraging process that swiped on the continent in the 1990s (see below). Secondly, continental European companies have a much higher proportion of provisions. In Germany, provisions include a large amount of company-based supplementary old age pension schemes. These have a long tradition, especially encouraged by special tax regulations. In Italy, companies accumulate withdrawals from employees' monthly paychecks in the form of "provisions for termination indemnity", which are subject to a favourable tax treatment and are only paid when an employee leaves the firm. The amount of provisions in France is higher than in the UK, but lower than in Germany and Italy. Cash flow statements: sources of funds Evolutions of balance sheet components are best examined through plots of the cash flow accounts. According to standard practice we report separately sources of funds from uses. As the composition of flows is very volatile across time, we have computed unweighted averages across both companies and time periods. The results for the sources of funds are shown in Notably in Italy, half as much short-term debt has been repaid as other sources have been raised. Secondly, long-term borrowing has been much smaller compared to internal sources in all groups. Long-term borrowing comprises instruments whose maturity exceeds the normal reporting period of one-year and it represents the amount received from issuance of long-term debt (thus including both new loans and issues of bonds, convertible and not convertible), increase capitalised lease obligations, and debt acquired from acquisitions. Unlike short-term borrowings for which only the net change is reported, long-term borrowings are reported gross of repayments, the latter being 8 classified as a use of funds in the cash flow statements. Thirdly, share issues were a significant although small source of funds only in the UK, while elsewhere they were negligible. Finally, other financial sources are important only in Germany where they are the main component of provisions. The prevalence of retained earnings, coupled with the aforementioned reductions in short-term borrowings explains why Common Equity has increased for all groups in 1999 from the beginning of the sample (see Cash flow statements: uses of funds In France, reductions in Long-term Debt represent over 30% of total uses of funds, and this implies lower levels of capital expenditures than in the UK, relative to total uses. The dynamics are well known, as French companies reduced investment and repaid debt in the (successful) effort to improve their equity base in the 1990s. Germany and especially Italy present capital expenditures in percentages of total uses that are even lower than in France. Over 50% of funds in both countries are used to accumulate cash and short-term investments and to repay long-term debt. Financing gap versus financing surplus An examination of the relative amounts of capital expenditures and retained earnings provides a key interpretation of capital structure dynamics over the period. The comparison is drawn directly in DETERMINANTS OF DEBT ISSUES It appears from the descriptive analysis in section 1 that companies issue debt only when internal resources are not sufficient to cover their financing needs. This observation is consistent with two very different theories of capital structure. According to the "tradeoff theory" companies try to attain an optimal level of leverage, which in turn depends on their profitability, stability of cash flows, assets usable as collateral, taxes, etc. Observed financing decisions reflect a tendency towards an unobservable optimal leverage level. If this level is constant actual leverage will be close to its optimal level and companies will only issue small amount of debt to refinance expiring loans or bonds. Otherwise, suppose that such an optimal level is not constant over time. As long as adjustment costs are small 11 enough, the observed actual leverage should be close to the optimal level in any given period and therefore companies are not likely to issue or retire important amounts of debt. Our discussion in Section 1, showing that the amount of long-term debt has not changed too much over the last decade, could then be taken as evidence supporting a "trade-off theory". Quite differently, the "pecking order" theory of capital structure states that firms always prefer to finance their assets out of internal sources as asymmetric information creates a conflict of interest between insiders (managers, holders of existing loans, bonds or shares) and outsiders (investors, banks, etc.) No optimal level of leverage exists according to this second theory. Our discussion in Section 1, showing that companies rely mostly on internal funds, could support the "pecking order" theory too. This ambiguity is now widely recognized in the literature. While a large number of studies has developed and applied empirical methodologies to tests the implications of both theories, the consensus is that tests of either hypothesis in general lack power against the alternative. Shyam-Sunder and Myers (1999) and Chirinko and Singha (1999) provide a state of the art discussion of the issue. Rather than entering the debate, we choose to take an eclectic approach and include elements from both theories in our model of debt issues. Determinants of debt issues Several determinants of leverage have been investigated in the empirical literature. Harris and Raviv (1991) present a comprehensive survey of the results: "… studies generally agree that leverage increases with fixed assets, non-debt tax shields, growth opportunities, and firm size and decreases with volatility, advertising expenditures, research and development expenditures, bankruptcy probability, profitability and uniqueness of the product". The survey in Harris and Raviv (1991) has been the starting point for most recent studies. Rajan and Zingales (1995) and more recently the papers in Sauvé and Scheuer (1999) provide a thorough discussion of recent developments. Among these, a most notable development has been the estimation of dynamic leverage functions, which include lagged leverage as a regressor. Kremp and Stöss (2001) estimate dynamic leverage models for France and Germany while Ozkam (2001) studies UK companies. Drawing on this literature, we select five main determinants of debt issues. Firstly, gross debt issues should depend positively on the lagged level of leverage (Lev). Both the "trade-off" and the "pecking order" theory are consistent with a positive relationship. If the former theory applies, we can assume that companies with observed high levels of leverage also have high unobservable leverage targets. They will issue new debt in order to remain close to their target as existing debt expires. If the latter theory applies, companies may display high levels of debt as a result of consistently weak cash inflows. Such companies will show both high levels of leverage and high issuance activity. We choose to measure Lev as the ratio between long-term debt and long-term debt plus equity, to be consistent with the fact that our debt issue variable includes only non-current debt instruments. Secondly, debt issuance may depend on the size of the company (Size), however the sign of such an effect is uncertain a priori. On the one hand, it is widely believed that smaller 12 companies are faced with tighter borrowing constraints due to asymmetric information. The argument is discussed by a wide literature, and in particular by Fazzari, Hubbard and Petersen (1988) and Gilchrist and Himmelberg (1994). Also, larger companies typically have better access to debt markets. Both arguments would imply a positive sign. On the other hand, at least in principle, if a small and a large company face the same investment opportunity, the latter is more likely to have sufficient internal sources to cover the full expenditure while the smaller company will have to resort to credit (pecking order theory). In this case, one should observe a negative sign. We find it simplest to measure Size as the natural logarithm of the number of employees. Thirdly, the availability of fixed assets to be used as collateral should be positively related to debt issuance, as it reduces the risk for the lender. This is particularly true if the lender is subject to both ordinary business risk and moral hazard/ adverse selection (the latter due to asymmetric information). Following standard practice (see Kremp and Stöss 2001) we include the ratio of fixed assets plus inventories divided by total assets among our explanatory variables (Colta). Fourthly, firm profitability (EBTA) may influence issues of debt in two different directions. High profitability may signal high marginal product of capital, in which case companies are likely to have a high demand for investment. This in turn generates high demand for credit, which is gladly met by lenders and bond investors seeing low risks in lending to a profitable firm. Conversely, highly profitable firms will be able to generate internally most of needed funds and will not need to issue debt. If the pecking order theory applies these companies will not distribute their profits as dividends, but rather use them to finance their capital expenditures 6 . Concerning the measure to use, EBITDA is generally thought to be the best widely available proxy for cash flow as it takes into account all funds that are generated internally by the company. We follow standard practice of normalizing by total assets to obtain a measure of profitability (henceforth EBTA). It would be preferable to split EBITDA into its three components Earnings (before amortization and depreciation), interest and taxes so that one could use the three separately as explanatory variables in the econometric model. This would allow testing the separate effects of such components. However, insufficient availability of data on interest and tax payments prevents us from exploiting this approach. The fifth and final explanatory variable is the "current ratio", that is, the ratio between current assets and current liabilities (Curr). One reason why this variable may be important is that it is a proxy for liquidity. According to Ozkan (2001), firms with higher liquidity can support a relatively higher leverage as lenders have greater reassurance that obligations will be met when they fall due. This will imply a positive relation between issues of debt and the current ratio. Another reason for a positive sign, relevant in our set-6 Note the difference between the argument given above to include lagged leverage under the pecking order theory (i.e. sustained low profitability implies both high leverage and high issue) and the argument given here (i.e. uncharacteristically low profitability in a given period forces the firm to issue an uncharacteristically higher amount of debt). Separate identification of these two effects would require specifying a fully dynamic model of the company balance sheet, which is beyond the scope of this paper. 13 up is that sustained high issues of non-current debt allow for reduction in current liabilities and thus increase the current ratio for a given level of current assets. However, if a company generates cash flows in excess of expenditures, one may observe an increase in the volume of liquid assets in the absence of debt issues, possibly giving rise to a negative statistical relationship 7 . A final word about potentially omitted variables as listed in the above quote from Harris and Raviv (1991). The magnitude of non-debt tax shields, other than depreciation is not available, too few firms report R&D expenditures and in any case these are not reported separately from advertising expenditures. Finally, there are too few observations to compute a meaningful measure of earning volatility 8 . The econometric model The following general model, whose variables have already been introduced above, is estimated for each group of companies separately: Three main observations are in order concerning the dependent variable and the lagged structure of the model. Firstly, as already mentioned, the LTB variable represents the gross amount received by the company from issuance of long-term debt (convertible and not convertible), increase capitalized lease obligations, and debt acquired from acquisitions. The choice of gross rather than net issues is important. A corporation's choice of issuing long-term debt is a discrete one as companies either apply for a mortgage or not, they either issue a bond or they do not. A net measure of debt issued does not contain any information about financing choices, as "pre-determined" repayments of outstanding long-term debt are subtracted from new issues. Gross measures are thus preferable. Secondly, LTB is normalized by the sum of capital expenditures and net assets from acquisitions (NAA) because we need some scaling measure to control for heteroskedasticity. Thirdly, our normalization enables a loose interpretation of the estimation results as a test of the null hypothesis that no relationship 7 We will see below that this is indeed the case in a Tobit model. 8 To deal with the problem, Kremp and Stöss (2001) construct a clever cross-sectional risk measure as the squared relative difference between the firm-specific profit and the average profit of all available firms. Since their study includes a large number of small non-listed companies, which are much more likely to default than their listed counterparts, risk turns out to be important. We do not believe that this would be the case for our sample. 14 exists between issues of debt and capital expenditures. Under such null hypothesis, no pre-determined variable should help predicting the ratio under study. By entering all regressors as lagged variables such hypothesis can be tested 9 . Summary statistics A closer examination of the summary statistics