Document Sample

Corporate Growth: the Role of Financial Structure Silvia Giannangeli Working Paper Series n. 02 ■ May 2010 Statement of Purpose The Working Paper series of the UniCredit & Universities Foundation is designed to disseminate and to provide a platform for discussion of either work of the UniCredit Group economists and researchers or outside contributors (such as the UniCredit & Universities scholars and fellows) on topics which are of special interest to the UniCredit Group. To ensure the high quality of their content, the contributions are subjected to an international refereeing process conducted by the Scientific Committee members of the Foundation. The opinions are strictly those of the authors and do in no way commit the Foundation and UniCredit Group. Scientific Committee Franco Bruni (Chairman), Silvia Giannini, Tullio Jappelli, Catherine Lubochinsky, Giovanna Nicodano, Reinhard H. Schmidt, Josef Zechner Editorial Board Annalisa Aleati Giannantonio de Roni The Working Papers are also available on our website (http://www.unicreditanduniversities.eu) WORKING PAPER SERIES N. 02 - MAY 2010 ■ 1 Contents Abstract 3 1. Introduction 4 2. Data 6 3. The financial structure of Italian firms: 7 definitions and descriptive statistics 4. The relationship between firm 10 financial structure and growth 5. The relation between 22 financial structure and investments 6. Discussion and conclusions 24 WORKING PAPER SERIES N. 02 - MAY 2010 ■ 2 Corporate Growth: the Role of Financial Structure Silvia Giannangeli* Abstract In the present paper I study the relationships between firm financial structure and growth in a large sample of Italian firms (1998-2003). The paper expands upon existing analyses testing whether liquidity constraints affect firm performance by considering among growth determinants also firm debt structure. Panel regression analyses show that more liquid firms tend to grow more. However, firms do not use their capital to expand, but rather to increase debt. I also find that firm growth is highly fragile as it is positively correlated with non-financial liabilities and it not sustained by a long-term debt maturity. The breakdown by industry, adopting Pavitt’s taxonomy, suggests that the growth process is sustained by less fragile financial structure in firms operating in non traditional manufacturing sectors. Finally, quantile regressions suggest that fast-growing firms are characterized by higher growth/cash- flow sensitivities and heavily rely on external debt, but seem to be less bank-backed than the rest of the sample. Results are generally confirmed when firm investment instead of employment growth is considered. Overall, my findings suggest that the link between firms’ investment and expansion decisions is far more complicated than postulated by standard tests of investment/cash-flow sensitivities. Keywords: Firm growth; Financial structure; Cash flow; Financial constraints; Gibrat law; Quantile regressions. JEL Codes: L11, G30, D2 * UniCredit Group - CIB Strategy and Customer Analysis – Milan Italy silvia.giannangeli@unicreditgroup.eu WORKING PAPER SERIES N. 02 - MAY 2010 ■ 3 1. Introduction In the last two decades, a rapidly growing stream of empirical research has investigated the relation between corporate growth and financial structure. In particular, considerable attention has been paid to the hypothesis that firms are constrained in their expansion process by the lack of appropriate financial resources. The “financial constraints hypothesis” proposed by Fazzari, Hubbard and Petersen (1988) states that a wedge between internal and external cost of funds exists, and that capital market failures may limit the ability of firms to raise the financial resources they need in order to undertake the desired level of investments. A rather common way to test this hypothesis has been sought in the significance of the firm investment-cash flow sensitivity. The logic of the test works as follows. If capital markets were perfect, we should not expect cash flow to be relevant for investment and therefore for growth. Whenever firms do not have the necessary liquidity to realize their desired investment, they may simply go “external” and get as much money as they like to sustain their planned growth. Nevertheless, problems of asymmetric information might rise the cost of external finance and may lead to credit rationing. If this is the case, the ability to generate cash flow becomes important for financing investments. Many empirical studies have focused on the impact of cash-flow on firms’ investment decisions and growth (see, as an example, Carpenter and Petersen, 2002; Fagiolo and Luzzi, 2006; Oliveira and Fortunato, 2006). However, a radical scepticism has been expressed about the usefulness of investigating financing constraints through investment-cash flow sensitivity at the end of the ‘90s by Kaplan and Zingales (1997; 2000). The argument against the interpretation of cash flow sensitivity of investments as a proof of financing constraints is based on the scarce theoretical foundations of such a view and, on the contrary, by sound empirical as well as theoretical reasons to expect that investments may positively react on cash flow even in absence of any financial constraint. Indeed, cash flow may contain information about firm investment opportunities, and therefore the relationship between this variable and growth, or investments, may be the product of a virtuous selection process in the market. Kaplan and Zingales (1997) observe that “[t]he most financially successful and least constrained firms in our sample appear to rely primarily on internal cash flow to invest despite they […] exhibit a high investment-cash flow sensitivity”. These authors point out that firms have some degrees of freedom in choosing their preferred way to finance investments. The emphasis given by the literature on the validation of the financial constraints hypothesis through tests on the significance of cash-flow coefficients in investments as well as growth regression equations has diverted the scientists’ attention from the question posed by Kaplan and Zingales (1997; 2000) about the causes of the observed investment cash-flow sensitivity. In their work, the authors propose the conjecture that managerial behaviour and, more generally, the rules of corporate governance have a considerable role in explaining the observed financial strategies of firms. In this context, a growing literature in corporate finance has stressed, in the last decades, the role played by institutional factors such as tax and bankruptcy laws, the role of banks, the rules of WORKING PAPER SERIES N. 02 - MAY 2010 ■ 4 corporate control, in determining the financial structures of firms (Rajan and Zingales, 1995). It is widely accepted that the institutional factors mentioned above may moderate the relationship between some firm specific characteristics, such as size or age, and some dimensions of the firms’ financial structure, such as leverage or debt maturity. However, very little attention has been paid to the hypothesis that something could be learned about firms’ financial choices by looking at the relation among a plurality of dimensions of firm debt structure and firm market real performance. Schiantarelli and Sembenelli (1997) study the relation between debt maturity and firm investments, profitability and growth in UK and Italian firms. They find that positive effect of debt maturity on both variables. Some more recent empirical papers provide mixing evidence on the association between financial structure and corporate real performance (Michaelas et a., 1999; Honjo and Harada, 2006). The present analysis has a few elements of novelty, investigating the relation among a wider set of financial dimensions on one hand and firm employment growth or investments on the other. In section 2 we will discuss the main dimensions of firm capital structure that will be considered in the empirical analysis of corporate growth. In section 3 we will describe the data and provide some descriptive statistics about the financial structure of Italian manufacturing firms. In section 4 we will show the results of estimating a descriptive model of corporate growth: panel regression estimation is adopted to show the relation between firm debt, its sources and maturity structure on one side, and growth on the other, controlling for firm heterogeneity. Quantile regression analysis allows to identify the relation between financial variables and growth at different values of the growth distribution. Estimates are presented for the whole sample of manufacturing firms as well as for distinct Pavitt’s technological categories (Pavitt, 1984). Section 5 extends the analysis to the relation between firm financial structure and investments. Section 6 summarizes the main results and concludes. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 5 2. Data We employ balance sheets data collected by Centrale dei Bilanci from 1998 to 2003. In order to discard from the analysis all the phenomena related with self-employment, we remove firms with less than two employees1. In order to avoid attrition , we use a closed panel of firms that are continuously operating over the period 1998-2003. We drop from, the dataset firms that exhibited a yearly growth rate of employees lower than −200% or larger than 200% in any of the observed years, in order to weaken the problems that misreported data may introduce in the analysis2. Moreover, we removed firms for which the share of trade debt over total debt was less than the 1st percentile or above the 99th percentile. As it will be made clear in the following, trade debt is more subject than other magnitudes to assume extreme values since it is calculated as the residual amount once equity, financial debt and other provisions are removed from the overall liabilities of the firm, and therefore it is affected by the problems of misreported data in any of these variables. The number of available observations in the balanced panel is 9315 per year. Table 1 summarizes the industry composition of the final sample. Table 1. Industry composition of the sample Number NACE code Industry of firms 15 Food and beverages 1029 16 Tobacco 1 17 Textiles 820 18 Wearing apparel and dressing 273 19 Tanning 368 20 Wood products 210 21 Pulp and paper 256 22 Publishing and printing 232 23 Coke petroleum and nuclear fuels 33 24 Chemicals 551 25 Rubber and plastic 602 26 Other non-metallic mineral products 548 27 Basic metals 394 28 Fabricated metal products 1098 29 Machinery and equipment 1351 30 Office machinery and computers 21 31 Electrical machinery 347 32 Radio and TV 123 33 Medical precision and optical instruments 210 34 Motor vehiclestrailers and semi-trailers 194 35 Other transport equipment 90 36 Furniture 564 Total 9315 1 Bottazzi et al. (2006) show that firms with one employee radically differ from firms with two or more employees in terms of production structure. 2 As a result, 150 firms are removed from the sample. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 6 3. The financial structure of Italian firms: definitions and descriptive statistics There are many ways of measuring capital structure, each measure having its pros and cons and, ultimately, its usefulness will depend on the purpose of the investigation. In first place, we consider cash flow as a flow measure of firm internal liquidity. Firms’ cash flow has been calculated by Centrale dei Bilanci, on the basis of detailed information on different flow items3. Since cash flow is highly correlated with any measure of firm size (the correlation coefficient between cash flow and value added, sales or employment equals, respectively, 0.83, 0.60 and 0.54) the ratio between cash flow and sales (SCF) is used throughout the analysis. In addition, we employ the ratio between equity and firm total assets (EQ) as a stock measure of firm propensity toward self-financing or, conversely, reliance on external debt. Firms’ equity is mainly composed by share capital and retained earnings. Hence, the ratio between equity and assets is a proxy for the importance of a firm own resources in financing investments. In order to account for different dimension of the firm debt structure we build several indicators. Indeed, the liability side of a firm’s balance sheet collects very heterogeneous items, being one firm’s “debt” composed by short and long term liabilities, as well as financial and non financial debt. More precisely, firms liabilities are made up by the sum of accounts receivable (the so-called “trade debt”), tax debt, bank debt, bonds, provisions related with pensions or other social obligations as well as other financial resources such as loans from firms belonging to the same group or other minor sources of debt. In this paper we explicitly consider two types of financial resources: trade debt and financial debt. While financial debt has been the object of a large recent literature in law and economics, and its determinants both in terms of leverage and maturity structure have been investigated at length, very little attention has been paid to other sources of finance and, among them, trade debt. Trade debt is generated through the ordinary process of transformation and production carried on by the firm, and is mainly composed by debts toward suppliers or, more generally, toward agents with which the firm establishes trade contacts (debts toward controlled or related firms are also included). Conversely, financial debt is not linked with the “nature of doing business”, but responds to specific financing strategies by firm management, and is mainly composed by bank debt, bonds as well as firm shares owned by other companies of the same group. We define a firm “total debt” to be equal to the sum of all liabilities item except from equity. The relative importance of financial debt, FD, is measured by the ratio between financial debt (defined as the sum of debts toward credit institutions, bonds and other financial debts) and total debt. Similarly, the importance of trade debt, TD, is measured by the ratio between trade debt (defined as the residual amount obtained subtracting from total debt all financial items and other non financial provisions) and total debt. In so doing, the only significant liability item not explicitly included in the analysis (along with 3 The cash flow variables shows a correlation larger than 0.90 with a simpler proxy obtained by summing up firms net profits, depreciation costs and the “Trattamento di Fine Rapporto” (the so-called “TFR”). WORKING PAPER SERIES N. 02 - MAY 2010 ■ 7 minor sources of debt) is the so-called “Trattamento di Fine Rapporto”, or “TFR”, i.e. a fund where firms are legally forced to set aside provisions for their employees and will be used to “compensate” employees at the end of the job contract. These provisions actually represent a “debt” of the firm toward its employees, but their magnitude does not respond to any specific (either financial or commercial) strategy, but just mirrors the age distribution of employees as well as their turnover. For this reason, legally required social (and tax) provisions are omitted in the analysis4. Another interesting aspect of firm debt structure is the choice of financing sources and in particular the extent to which a firm is dependent upon bank debt. The share of bank debt of a firm can be either high or low depending on a number of factors both from the demand and supply side. A firm may choose to have a low (or even zero) degree of indebtedness with banks or may be limited by the supply side, especially if the firm has a low risk ranking. The relation between the amount of bank debt with firm growth in interesting in that it reveals the role played by banks into the dynamics of manufacturing firms. The relative importance of bank debt in firm financial debt is isolated and captured by the share of bank over total financial debt (BD). Finally, we analyze the relation between the maturity structure of debt and firm growth. In spite of a rich literature studying the determinants of debt maturity, very little is known on the link between such dimension of firm financial structure and growth. Firm debt maturity is measured by the share of short-term over total financial debt (SFD)5. Table 2 summarizes the definition of the financial indicators discussed so far. Table .3 reports the mean value and the dispersion (measured through the variation coefficient) of the distribution of SCF, EQ, FD, BD and SFD in different years of the sample period. The evolution over time of the financial indicators shows that the share of financial and bank debt, as well as the share of short term debt, are rather stable, while a weak upward trend characterizes the evolution of firms’ equity to asset ratio. Table 2. Financial indicators adopted in the analysis Variable Construction Cash Flow SCF Sales Equity EQ Total Assets Financial Debt FD Total Debt Trade Debt TD Total Debt Bank Debt BD Financial Debt Short-term Financial Debt SFD Financial Debt 4 It is true, indeed, that TFR provisions represent a relevant and cheap source of finance for firms: they amount, on average, to around 9% of total debt. 5 Short-term financial debt is composed by short-term bank loans and other types of short-term financial resources. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 8 Descriptive statistics confirm some well known features of Italian industrial system. On average, debt accounts for more than twice as much as equity in the financing of firm assets. Equity amounts on average to less than 30% of total assets. International comparisons show that Italian firms have considerably lower equity-to-assets ratios than firms from other European countries (database BACH, European Commission). Second, a very large share of firms’ debt is non financial: on average, around 65% percent of total debt is composed by trade debt and social provisions. Third, Italian firms are largely dependent on bank credit, the share of bank debt in firms total financial debts amounting, on average, to 77%. Table 3. Mean and variation coefficient of financial variables in 1998, 2000 and 2002 Mean Variation Coefficient Variable 1998 2000 2002 1998 2000 2002 SCF 0.069 0.071 0.068 0.922 0.955 1.136 EQ 0.238 0.249 0.262 0.609 0.609 0.621 TD 0.134 0.120 0.072 2.690 2.996 5.600 FD 0.412 0.399 0.401 0.462 0.478 0.490 BD 0.785 0.797 0.786 0.374 0.366 0.392 SFD 0.674 0.683 0.684 0.435 0.423 0.428 Finally, Table 3 shows that the maturity structure of firm financial debt is largely shifted toward short term liabilities. Although a deep investigation of this issue is beyond the scope of our analysis, it seems that possible problems of un-balancedness of the debt structure may affect some Italian firms, inducing long term investments to be financed through short term debt. More generally, a short debt maturity may imply some constraints to the ability of a firm to invest and growth. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 9 4. The relationship between firm financial structure and growth In the vein of the post-Gibrat literature, we investigate the relation between firm financial structure and employment growth by estimating an “augmented” Gibrat-like regression, where the financial indicators discussed in Section 3 are included among regressors. Table 4 shows the correlation matrix among the selected financial variables. Due to the high correlation between the share of trade debt (TD) and the equity-to-assets ratio (EQ), these two variables will not be included in the regression simultaneously. Table 4. Correlation matrix among variable log(EMPi,t-1) log(AGEi,t-1) SCFi,t-1 EQi,t-1 TDi,t-1 FDi,t-1 BDi,t-1 SFDi,t-1 log(EMPi,t-1) 1.000 log(AGEi,t-1) 0.150 1.000 SCFi,t-1 0.154 0.023 1.000 EQi,t-1 0.118 0.133 0.441 1.000 TDi,t-1 -0.136 -0.134 -0.370 -0.835 1.000 FDi,t-1 -0.032 -0.035 -0.149 -0.369 -0.073 1.000 BDi,t-1 -0.076 0.010 -0.140 -0.181 0.158 0.084 1.000 SFDi,t-1 -0.117 -0.060 -0.197 -0.184 0.159 0.026 0.229 1.000 The final model includes a quadratic term on firm size in order to capture possible non-linearities in the size- growth relationship, time dummies to get rid of the trend components (Dtime), sectoral dummies, defined as the first two digit in ATECO classification (Dsector) and firm localization (by adopting a set of dummy variables, Dloc, corresponding to geographical macro areas North–East, North–West, Center and South of Italy). In order to account for possible delayed effects, the model contains lagged values of cash flow. Hence the saturated models have the following expression: G R O W T H i,t = β 1 log(E M Pi,t-1 ) + β 2 log 2 (E M Pi,t-1 ) + β 3 log(A G E i,t-1 ) + β 4 log 2 (A G E i,t-1 ) + + β 5 S C Fi,t-1 + β 6 S C Fi,t-2 + β 7 S C Fi,t-3 + β 8 E Q i,t-1 (or T D i,t-1 ) + β 9 FD i,t-1 + β 1 0 B D i,t-1 (1) + β 11S FD i,t-1 + β 12 D tim e + β 13 D secto r + β 14 D lo c + v i,t where the growth rate of employees has been computed as: EMPi,t − EMPi,t-1 GROWTH i,t = EMPi,t Final model specifications have been selected starting from the saturated model, by dropping variables whose effect is found not to be statically significant on the basis of a Wald test6. 6 Model selection has been performed using pooled OLS estimates of the saturated model. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 10 4.1. Panel regression analysis Table 5 reports the pooled OLS, random and fixed effects estimation results of the final regression model, that takes the following form: GROWTHi,t = β1 log(EMPi,t-1 ) + β 2 log 2 (EMPi,t-1 ) + β 3 log(AGE i,t-1 ) + β 4SCFi,t-1 + + β 5SCFi,t-2 + β 6 EQi,t-1 (or TDi,t-1 ) + β 7 FDi,t-1 + β8 BDi,t-1 + β 9SFDi,t-1 + (2) + β10 D time + β11Dsector + β12 Dloc + vi,t where the error term vi,t may contain both unobservable individual effects, ( ci ), and idiosyncratic error, ( ui,t ), that is: vi,t = ci + ui,t . Pooled OLS estimation is motivated by the weaker exogeneity assumptions made on the idiosyncratic error term: both random and fixed effects estimation make the strong exogeneity assumption that the unobservable component ui,t is in each period uncorrelated with explanatory variables in each other period. However, pooled OLS turn out to be inefficient if the error term in equation 2 does contain unobserved individual components. Indeed, Breusch and Pagan test statistic calculated after random effects estimation does reject the hypothesis of absence of individual unobserved effects. Both random and fixed effects account for the presence of ci in the model. Although Hausman test suggests that fixed effects estimation has to be preferred, random effect results are also reported. Indeed, fixed effect estimation may lead to imprecise estimates due to the low variations over time of the book “stock” variables EQ, FD, BD and SFD (Woolridge, 2002) 7. 7 The first-order autocorrelation of these variables is higher than 0.90. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 11 Table 5. Panel regression results. Dependent variable: GROWTH it. Model 1 Model 2 Pooled Random Fixed Pooled Random Fixed Variable OLS Effects Effects OLS Effects Effects log(EMPi,t-1) -0.051 *** -0.054 *** -0.788 *** -0.050 *** -0.053 *** -0.788 *** (0.006) (0.004) (0.064) (0.006) (0.006) (0.064) log2(EMPi,t-1) 0.004 *** 0.005 *** 0.024 *** 0.004 *** 0.004 *** 0.023 *** (0.001) (0.001) (0.008) (0.001) (0.001) (0.008) log(AGEi,t-1) -0.015 *** -0.015 *** 0.076 *** -0.014 *** -0.014 *** 0.076 *** (0.001) (0.001) (0.019) (0.001) (0.002) (0.019) SCFi,t-1 0.198 *** 0.197 *** 0.141 *** 0.207 *** 0.206 *** 0.148 *** (0.029) (0.016) (0.033) (0.030) (0.032) (0.033) SCFi,t-2 0.063 *** 0.064 *** 0.076 *** 0.081 *** 0.081 *** 0.091 *** (0.020) (0.017) (0.027) (0.020) (0.020) (0.027) EQi,t-1 -0.037 *** -0.036 *** -0.004 (0.008) (0.007) (0.021) TDi,t-1 0.024 *** 0.024 *** 0.017 ** (0.003) (0.003) (0.007) FDi,t-1 -0.028 *** -0.028 *** -0.047 *** -0.013 *** -0.013 *** -0.040 *** (0.005) (0.005) (0.012) (0.005) (0.005) (0.012) BDi,t-1 0.012 *** 0.012 *** 0.024 *** 0.011 *** 0.011 *** 0.023 *** (0.003) (0.003) (0.006) (0.003) (0.003) (0.006) SFDi,t-1 -0.003 -0.003 -0.008 * -0.003 -0.002 -0.008 (0.003) (0.003) (0.005) (0.003) (0.003) (0.005) Dtime Yes Yes Yes Yes Yes Yes Dsector Yes Yes Yes Yes Dloc Yes Yes Yes Yes Number of obs. 37260 37260 37260 37260 37260 37260 F test 22.98 *** 161.99 *** 24.21 *** 163.51 *** Wald test 1222.65 *** 2745.89 *** Note: All estimation procedures account for heteroskedasticity at the firm level and autocorrelation of the error term. The estimation results of Model 1 and Model 2, that differ due to the exclusion of either EQ or TD from the regression, are very similar. We therefore analyze the estimation results jointly. First, we find a negative relationship between firm size and growth. The relation is not monotonic. Rather, the negative relation tends to vanish as the size increases (the coefficient of the quadratic term log2(EMPi,t-1) is significantly different from zero in all the estimated specifications). Second, we don’t find very consistent results on the relationship between firms’ age and growth. While the results WORKING PAPER SERIES N. 02 - MAY 2010 ■ 12 obtained through the pooled and fixed effect estimations point out a negative relation, the within, fixed effect estimation detects a positive one. More interesting for our analysis, we find that the amount of cash flow is positively correlated with firm growth. The positive and significant relationship detected in the data by all estimation procedures might not mirror actual liquidity constraints to firm growth. It is, indeed, true that firms that grow more are those endowed with larger cash flows (both contemporaneous and lagged), but it could well be the case that the causal relation runs the other way round than postulated by standard tests on the presence of financial constraints. The evidence of a positive relation between cash flow and growth can be better interpreted when accounting for other effects, captured through capital structure “stock” variables. Interesting enough, the estimated coefficient of equity-to-assets ratio is negative in the pooled and random effects estimations, suggesting that firms that grow more are less reliant on self financing, and rise more external funds, relative to their assets, than low growth firms do. This result runs against the predictions of models based on asymmetric information and agency costs, that would suggest higher costs of external finance for “good” and high-growth firms. The puzzling negative relation between firms own-capital based financial solidity and growth, although not supported by the within, fixed effect estimation, suggests that firms that decide to grow do so through the creation of new debt. It is interesting to notice that the share of financial debt (FD) is negatively related with growth. This result confirms that non financial debt, such as firms’ provisions for pensions and other social obligations, as well as trade debt, give firms a valuable buffer of resources for firm growth. In particular, the negative sign of the coefficient of FD highlights the relevant correlation between the “cycle” of current liabilities and firms’ expansion. Indeed, as indicated by the results on Model 2, the share of trade debt is positively and significantly associated with firm growth. An interesting result of the empirical analysis is the positive relation between the share of bank debt (BD) and growth. Consistently with the results on EQ and SCF, firms grow through bank debt, possibly using the amount of liquidity as a guarantee of firm solidity. Almost no role is found to be played by debt maturity structure on firm growth: if any, the relation between firms’ debt maturity and growth is negative. These combined results suggest that firms do not use their equity capital to finance their expansion. Rather, firms that decide to grow do so by creating new debt. Our results also emphasize that the growth profile of Italian Manufacturing firms is, on average, highly fragile: on one hand, it is positively correlated with the increase of non financial liabilities, on the other hand, it is not sustained by a long- term debt maturity. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 13 4.2. Quantile regression analysis Panel regression analysis estimates the relation between the mean value of the dependent variable (firm growth) and variations in the explanatory variables. It is possible, however, that marginal effects of changes in some of the variables in (2) are not equal across the whole distribution of firm growth. In other words, the estimated coefficients in Table 3.5 may be a poor estimate of the relation between some of the explanatory variables and firm growth, at different quantiles of its distribution. Quantile regression, introduced by Koenker and Bassett (1978), is a useful way to overcome this problem, by providing estimates of the regression coefficients at different quantiles of the dependent variable. Quantile regression amounts to estimating the following equation: yi = xi ' βτ + uτ i (3) th For the τ quantile of the distribution of y. The distribution of the error term uτi is left unspecified and the only assumption made is Quantτ (uτi | xi) = 0, which allows to write the conditional quantiles of y as a function of explanatory variables and parameters only: Quantτ(yi | xi) = xi’βτ . The estimate βτ of parameters in 3 is found by minimizing with respect to β the quantity: ∑ uτ h i i i , where the function hi is defined as: hi = {2τ −τ ) 2(1 if uτ i > 0 otherwise The estimate of the τth conditional quantile is therefore given by Quantτ ( yi | xi ) = xi ' β τ . Quantile’s coefficient βτ k can be interpreted as the partial derivative of the conditional quantile of y with respect δ Quantτ ( yi | xi ) to one of the kth explanatory variable, δ xk This derivative quantifies the marginal change in the τth conditional quantile due to marginal change in the kth element of x (Buchinsky, 1998). For applications of quantile regression in the medical and economic literature see, respectively. Abrevaya (2001) and Coad and Rao (2006). Table 6 and Table 7 report, respectively, the results of quantile estimation of Model 1 and Model 2. A sequence of quantile regressions was estimated for the 0.5, 0.25, 0.50, 0.75 and 0.95 quantiles of the 8 growth rate distribution and tests for equality of coefficients across quantiles were performed . The estimation results are interesting: first of all, we find that the relation between cash flow and growth is not the same across the whole distribution of growth rates. In particular, the cash flow sensitivity of growth is significantly different for firms growing less or growing more than the median 8 Stata command sqreg was used to perform quantile regression and standard errors were calculated using the bootstrapping method suggested by Gould (1997), with 100 repetitions. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 14 firm in the sample9. Firms growing more than the median value (50th percentile) show a significantly larger sensitivity to cash-flow. This result is consistent with different, but opposites stories: on one hand, one could interpret the result by saying that firms with higher growth opportunities are also riskier from an external investors’ viewpoint, and therefore they may incur in credit rationing with higher probability than low growth firms. This will force high growth firms to use their internal cash flow in order to finance new investments. On the other hand, the result can be interpreted as a support to the view that cash flow contains information about the investments, profit and growth opportunities of a firm: detecting a positive relation between growth and cash flow is therefore not a symptom of the presence of financial constraints to firm decision to expand but, rather, a signal that a virtuous selection mechanism is at play in the market. Table 6. Quantile regression results. Model 1. Numbers in italics represent the value taken by the dependent variable at each of the quantiles shown in table. Model 1 5% 25% 50% 75% 95% Variable (-0.161) (-0.037) (0.000) (0.067) (0.250) log(EMPi,t-1) 0.083 *** 0.007 0.003 ** -0.056 ** -0.275 *** (0.016) (0.044) (0.001) (0.004) (0.036) log2(EMPi,t-1) -0.008 *** -0.001 * -0.001 *** 0.004 *** 0.025 *** (0.002) (0.0004) (0.0001) (0.001) (0.004) log(AGEi,t-1) 0.004 -0.006 *** -0.009 *** -0.020 *** -0.045 *** (0.004) (0.001) (0.001) (0.001) (0.006) SCFi,t-1 0.198 *** 0.167 *** 0.145 *** 0.230 *** 0.297 *** (0.027) (0.010) (0.010) (0.027) (0.068) SCFi,t-2 0.092 *** 0.025 ** 0.022 * 0.035 0.055 (0.025) (0.012) (0.012) (0.027) (0.062) EQi,t-1 0.048 *** -0.007 -0.022 *** -0.066 *** -0.161 *** (0.016) (0.006) (0.003) (0.006) (0.023) FDi,t-1 -0.059 *** -0.029 ** -0.012 *** -0.022 *** -0.020 (0.013) (0.004) (0.002) (0.003) (0.014) BDi,t-1 0.015 ** 0.010 *** 0.008 *** 0.014 *** 0.011 (0.006) (0.002) (0.001) (0.002) (0.008) SFDi,t-1 -0.023 *** -0.005 ** -0.002 -0.003 0.014 (0.005) (0.002) (0.001) (0.002) (0.010) Dtime Yes Yes Yes Yes Yes Dsector Yes Yes Yes Yes Yes Dloc Yes Yes Yes Yes Yes Observations 37260 37260 37260 37260 37260 Pseudo R2 0.039 0.017 0.009 0.034 0.067 9 F tests fail to reject the null hypothesis that the SCF coefficient at the 5% or 25% percentile are equal to the 50% percentile at conventional significance levels (respectively, F test values are equal to 1.31 and 2.82). Similarly, no significant difference is found between coefficients at the 75% and 90% percentile. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 15 Second, the coefficient on EQ is not constant across quantiles of the growth rates distribution: in the case of equity-to-assets ratio, the coefficient at any of the growth rate quantiles are found to be statistically different from any other quantile. In particular, the relationship between firms’ propensity toward self finance and growth is positive and significant for firms in the 5th percentile, meaning that firms that are more reliant on own funds are those that downsize. As the growth rate increases, firms are found to use more and more debt. As for the type of debt the firm is using, we find that (Table 3.6) the negative relation already detected on the whole sample between the share of financial debt and growth is significant at all but the 95th percentiles. A partially different result if shown in Table 3.7, where the share of trade debt is included in the model: in this case FD coefficient is significant only for the lowest part of the support of the growth rates distribution10. Although delivering partially different results, the discrepancy between Model 1 and Model 2 is not radical, both models being consistent with the fact that for firms that are growing less (actually, shrinking), the increase in the amount of financial debt, relative to other types of debt, is associated with a significant decrease in the growth rate11. The share of trade debt (Table 3.7) is instead found to have a positive relation with growth, except from the lowest quantiles of the growth rates distribution. Interesting enough, the relation between the importance of trade debts and growth strengthen as firms' rate of growth increase12. 10 Tests for equality of coefficients at 5th and 25th percentiles suggest that coefficients are statistically different. 11 In Model 1 F test reject the hypothesis that the coefficients of FD at 5th and 25th percentiles are equal, while this is not the case for upper quantiles. 12 F tests reject the hypotheses of equality between coefficients across all quantiles, except when comparing the 25th and 50th ones (F test p-value: 0.143). WORKING PAPER SERIES N. 02 - MAY 2010 ■ 16 Table 7. Quantile regression results. Model 2. Numbers in italics represent the value taken by the dependent variable at each of the quantiles shown in table. Model 2 5% 25% 50% 75% 95% Variable (-0.161) (-0.037) (0.000) (0.067) (0.250) log(EMPi,t-1) 0.082 *** 0.007 ** 0.003 * -0.055 ** -0.273 *** (0.012) (0.003) (0.002) (0.005) (0.023) log2(EMPi,t-1) -0.008 *** -0.001 *** -0.001 *** 0.004 *** 0.025 *** (0.001) (0.0003) (0.0001) (0.001) (0.002) log(AGEi,t-1) 0.006 * -0.005 *** -0.008 *** -0.019 *** -0.043 *** (0.003) (0.001) (0.001) (0.001) (0.006) SCFi,t-1 0.222 *** 0.174 *** 0.159 *** 0.231 *** 0.295 *** (0.054) (0.017) (0.013) (0.024) (0.047) SCFi,t-2 0.119 ** 0.047 ** 0.028 ** 0.053 * 0.056 (0.047) (0.020) (0.009) (0.029) (0.051) TDi,t-1 0.002 0.013 *** 0.015 *** 0.031 *** 0.058 *** (0.006) (0.002) (0.001) (0.002) (0.005) FDi,t-1 -0.069 *** -0.027 *** -0.003 -0.001 0.027 (0.011) (0.004) (0.002) (0.002) (0.020) BDi,t-1 0.014 0.009 *** 0.006 *** 0.013 *** 0.006 (0.008) (0.002) (0.001) (0.002) (0.010) SFDi,t-1 -0.027 *** -0.006 ** -0.002 -0.001 0.017 * (0.008) (0.002) (0.002) (0.002) (0.009) Dtime Yes Yes Yes Yes Yes Dsector Yes Yes Yes Yes Yes Dloc Yes Yes Yes Yes Yes Observations 37260 37260 37260 37260 37260 Pseudo R2 0.038 0.018 0.010 .036 0.069 Finally, some differences are found in the relation between firm growth and the share of bank debt at different quantiles of the growth rate distribution: the most interesting result is that at the 95th percentile of the growth rates distribution there is no association between growth and bank debt. As for the maturity structure of debt, it is interesting to notice that shrinking firms are those for which the more the term structure of debt is short, the less they grow. No significant relation between the share of short- term financial debt, relative to total, and growth is found at upper percentiles. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 17 4.3. Analysis by Pavitt’s categories The results described in Section 4.1 and 4.2 suggest that Italian manufacturing growing firms tend to use debt in order to finance their investments. However, holding fixed the total amount of debt, an increase in the share of financial debt translates into a slowdown in growth, especially for firms that exhibit low (often negative) growth rates. In turn, this is consistent with the fact that growth is associated to an increase in the share of trade debts over the total amount of debt and the relation is significantly stronger as growth rates increase. This may mean that growing firms tend to finance their expansion by simply not paying the cost of growing, i.e. by creating debt toward suppliers. Although not a surprising news, this circumstance highlight the very rudimental financial structure of Italian firms and the weak bases on which they build their investments pattern. We now turn to the analysis of the impact of firms financial structure on growth by taking into account the technological characteristics and the nature of the innovative process in industrial sectors where firms operate. We do so by adopting the Pavitt’s taxonomy that distinguishes between “supplier dominated” (or, following a nomenclature that is often adopted in the Italian debate “traditional”) sectors, “scale-intensive” sectors, “specialized suppliers” and “science-based” sectors (Pavitt, 1984). Figure 1 shows the estimated probability density of SCF, EQ, TD, FD, BD and SFD disaggregated by Pavitt’s category in year 200013. It is quite evident from the probability densities that some of the variables’ distributions show a bi- or even multi-modality. This is the case for the share of bank debt (BD) and the share of short-term financial debt (SFD): both of them present a mode in 1 (all the financial debt being, respectively, bank debt or short-term) and another peak around zero. The differences among Pavitt’s categories in the distribution of financial variables is not sizeable in the case of cash-flow (SCF) and equity-to-assets ratio (EQ). Firms operating in the “specialized suppliers” industrial sectors are characterized by a debt structure where trade debt is more important (and, correspondingly, financial debt is less) as compared with other firms. More interestingly, there are considerable differences in the share of bank debt owned by firms categorized in different technological classes: the share of bank debt held by firms operating in “science-based” sectors is smaller than that in any other class. Similarly, the share of short-term financial debt is lower for these firms. These facts are consistent with the existence of alternative financial sources for firms in high-tech sectors, different from traditional bank credit, like for instance venture capital. Firms in the “traditional” sectors are those where the importance of bank debt relative to total financial debt is the highest and where the maturity structure of debt is the shortest. 13 The estimated probability densities do not differ across years. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 18 SCF EQ 3 10 8 2 6 4 1 2 0 0 -.5 0 .5 -.2 0 .2 .4 .6 .8 Suppl. Scale Trad. Scale Spec. Science Spec Science FD TD 2 1.5 1.5 1 1 .5 .5 0 0 0 .2 .4 .6 .8 1 -3 -2 -1 0 1 Trad. Scale Trad Scale Spec. Science Spec. Science BD SFD 4 2 3 1.5 2 1 1 .5 0 0 0 .5 1 0 .2 .4 .6 .8 1 Trad. Scale Trad. Scale Spec. Science Spec. Science Figure 1 Kernel estimates of the probability density of SCF, EQ, FD, TD, BD and SFD by Pavitt’s taxonomy. Note: “Trad” stands for “traditional sectors”, “Scale” for “scale-intensive” sectors, “Spec.” for “specialized suppliers” and “Science” for “Science-based” sectors. Empirical densities are estimated adopting Epanechnikov kernel with bandwidth h = 0.9(min {σ , IQR /1.34} N −1/ 5 , as suggested by ˆ Silverman (1986). WORKING PAPER SERIES N. 02 - MAY 2010 ■ 19 Table 8 shows the estimation results of Model 1 and Model 2 disaggregated by Pavitt’s category14. We find that growth is positively related with firms’ cash flow in all industrial sectors, although in the “specialized suppliers” the effect appears to be lagged by one year. Interesting enough, the negative relation between equity-to-assets ratio and growth detected on the whole sample is found to apply only in the “traditional” industries, while in scale intensive sectors this relation is found to be positive, although weak. In science-based or specialized suppliers industries the relation between equity ratio and growth is not significant. Although the negative association between FD and growth is found to hold in all sectors, it is worth noting that the claim that firms’ growth is associated with the creation of new trade debts no longer applies in scale intensive and specialized suppliers sectors, while the evidence on science-based industries is very weak. Finally, an increase in the importance of bank debt over total financial debts is not found to be associated with growth in all sectors: in science-based industries no significant relation between the share of bank debts and growth is at play. Summing up, these results highlight a considerable inter-sectoral heterogeneity in the financial structure of firms and in its impact on firm growth. Different from firms in the traditional sectors (for which the results closely mirror what was found on the whole manufacturing), the growth process of firms in scale intensive industries is associated with an increase in the equity-to-assets ratio, while in the other industries there is no significant relation among this ratio and firms’ growth rates. Moreover, the positive dependence between growth and the share of trade debt is limited to the traditional sectors although a much weaker relation is found also in science-based sectors. These combined results are consistent with the hypothesis that the growth pattern of firms outside the traditional sectors is not as fragile as suggested by the results on the whole sample. A final remark: the role of bank debt is important for firm growth in all sectors but science-based industries, suggesting that either demand or supply factors prevent high-tech firms to finance their investments through this type of debt. 14 Fixed effect estimation results are reported. Results using pooled Ordinary Lest Squares or random effects panel estimation are not significantly different. Results are available upon request. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 20 Table 8. Fixed effect panel regression results by Pavitt’s category Model 1 Model 2 Scale Specialized Science Scale Specialized Science Variable Traditional Traditional intensive suppliers based intensive suppliers based log(EMPi,t-1) -0.739 *** -0.826 *** -0.815 *** -0.871 *** -0.739 *** -0.826 *** -0.815 *** -0.869 *** (0.033) (0.038) (0.058) (0.093) (0.033) (0.038) (0.058) (0.094) log2(EMPi,t-1) 0.020 *** 0.027 *** 0.016 ** 0.048 *** 0.020 *** 0.027 *** 0.016 ** 0.048 *** (0.004) (0.005) (0.007) (0.011) (0.004) (0.005) (0.007) (0.011) log(AGEi,t-1) 0.091 *** 0.066 ** 0.057 0.070 0.091 *** 0.067 ** 0.057 0.068 (0.020) (0.027) (0.037) (0.061) (0.020) (0.027) (0.037) (0.061) SCFi,t-1 0.194 *** 0.117 *** 0.041 0.144 ** 0.196 *** 0.137 *** 0.047 0.151 ** (0.028) (0.037) (0.040) (0.073) (0.027) (0.036) (0.040) (0.071) SCFi,t-2 0.087 *** 0.049 0.132 ** 0.200 *** 0.092 *** 0.059 0.147 *** 0.212 *** (0.026) (0.038) (0.056) (0.062) (0.026) (0.038) (0.056) (0.062) EQi,t-1 -0.044 * 0.065 ** 0.016 -0.061 (0.024) (0.032) (0.041) (0.062) TDi,t-1 0.026 *** -0.002 0.015 0.037 * (0.008) (0.010) (0.013) (0.020) FDi,t-1 -0.048 *** -0.022 -0.073 *** -0.093 ** -0.030 ** -0.035 * -0.069 *** -0.070 * (0.014) (0.019) (0.024) (0.039) (0.014) (0.019) (0.024) (0.036) BDi,t-1 0.015 * 0.027 ** 0.038 *** 0.030 0.014 * 0.027 ** 0.038 *** 0.029 (0.008) (0.011) (0.013) (0.021) (0.008) (0.011) (0.013) (0.021) SFDi,t-1 -0.013 * -0.002 -0.011 0.001 -0.013 * -0.002 -0.011 -0.001 (0.007) (0.009) (0.007) (0.020) (0.007) (0.009) (0.007) (0.020) Dtime Yes Yes Yes Yes Yes Yes Yes Yes Dsector Yes Yes Yes Yes Yes Yes Yes Yes Dloc Yes Yes Yes Yes Yes Yes Yes Yes Observations 18904 10852 5868 1636 18904 10852 5868 1636 R2 0.01 0.01 0.01 0.01 0.01 0.01 0.01 0.01 F test 435.01 *** 286.85 *** 179.13 *** 36.21 *** 435.94 *** 286.37 *** 179.27 *** 36.5 *** WORKING PAPER SERIES N. 02 - MAY 2010 ■ 21 5. The relation between financial structure and investments Let us now turn to the last piece of analysis, by studying the relation between firms’ financial structure and investments. It is possible indeed that the fluctuations in the financial items that compose the liabilities structure of a firm do not have sizeable impact on firms’ growth process as measured in previous section, and that firms’ employment growth is not an appropriate indicator to capture any effect of financial variables on firms’ real performance. We therefore focus on a “intermediate” indicator, i.e. investment rate, that is in principle more likely to respond to variations in the financial structure of the firm. We do so by estimating equation (2), where firms investment rate in tangible assets15 is included as the dependent variable instead of employment growth. The regression equation is therefore the following: INVi,t = β1 log(EMPi,t-1 ) + β 2 log 2 (EMPi,t-1 ) + β 3 log(AGE i,t-1 ) + β 4SCFi,t-1 + + β 5SCFi,t-2 + β 6 EQi,t-1 + β 7 FDi,t-1 + β8 BDi,t-1 + β 9SFDi,t-1 + (3 + β10 D time + β11Dsector + β12 Dloc + vi,t where investment rate is defined as: fixed tangible assetsi,t − fixed tangible assetsi,t-1 INVi,t = fixed tangible assetsi,t Results reported in Table 9 do not differ significantly from those shown in Table 5 as for the sign and significance of coefficients on cash-flow, equity-to-assets ratio, the share of financial and trade debt. 15 We disregard investments in immaterial assets due to the noisy measurement of the highly heterogeneous components of this type of assets. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 22 Table 9. Panel regression results on investment rate. Fixed effect estimation. Model 1 Model 2 Pooled Random Fixed Pooled Random Fixed Variable OLS Effects Effects OLS Effects Effects log(EMPi,t-1) -0.041 -0.041 -0.015 -0.041 -0.041 -0.017 (0.032) (0.014) (0.246) (0.032) (0.031) (0.247) log2(EMPi,t-1) 0.004 0.004 -0.019 0.004 0.003 -0.019 (0.003) (0.003) (0.027) (0.003) (0.003) (0.027) log(AGEi,t-1) -0.032 *** -0.032 *** -0.087 -0.032 *** -0.032 *** -0.098 (0.005) (0.005) (0.072) (0.005) (0.005) (0.072) SCFi,t-1 0.185 *** 0.177 ** -0.063 0.177 *** 0.188 *** -0.129 * (0.055) (0.054) (0.071) (0.054) (0.054) (0.072) SCFi,t-2 0.127 ** 0.129 *** 0.036 0.129 ** 0.126 ** 0.006 (0.051) (0.051) (0.074) (0.051) (0.053) (0.072) EQi,t-1 -0.121 *** -0.117 *** -0.481 *** (0.023) (0.022) (0.095) TDi,t-1 0.046 *** 0.045 *** 0.132 *** (0.008) (0.007) (0.024) FDi,t-1 -0.092 *** -0.088 *** -0.302 *** -0.092 *** -0.047 *** -0.155 *** (0.017) (0.015) (0.017) (0.015) (0.015) (0.041) BDi,t-1 0.006 0.006 -0.006 0.006 0.006 -0.008 (0.012) (0.010) (0.028) (0.012) (0.010) (0.028) SFDi,t-1 0.004 0.004 0.021 0.005 0.005 0.021 (0.011) (0.010) (0.024) (0.011) (0.010) (0.025) Dtime Yes Yes Yes Yes Yes Yes Dsector Yes Yes Yes Yes Yes Yes Dloc Yes Yes Yes Yes Yes Yes Observations 24748 24748 24748 24748 24748 24748 F test 19.70 *** 50.56 *** 24.21 *** 51.02 *** Wald test 749.69 *** 720.41 *** Firms’ investments are financed through debt. However, as already discussed, it seems that investments are supported by a particularly rudimental and fragile financial structure, with investments being positively correlated with the growth of non-financial debt items. Differently from the results on firms’ employment growth, the share of bank debt over total financial debt held by the firm is not found to affect investments. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 23 6. Discussion and conclusions This paper is an attempt to extend the analysis of the links between firm financial structure and performance, beyond the traditional tests on financial constraints based on estimated investment cash-flow sensitivities. In particular, the purpose of the analysis is to shed some light on several aspects that obtained very little attention in the literature, namely the relation between the liability structure, the sources of debt and the debt maturity on one side and firm growth on the other. Our results suggest that Italian manufacturing firms, on average, do not use their own capital to finance their expansion. Rather, firms that decide to grow do so by creating new debt. Our results also emphasize that the growth profile of Italian manufacturing firms is, on average, highly fragile: on one hand, it is positively correlated with the increase of non financial liabilities, on the other hand, it is not sustained by a long-term debt maturity. The relation among financial variables and growth is not constant across the distribution of growth rates: firms that grow more are characterized by higher growth cash-flow sensitivities and heavily rely on external debt, especially non financial debt. The very rudimental and fragile characterization of Italian manufacturing firms that emerges from the aggregate estimation hides a considerable inter-sectoral heterogeneity. As a general result from the sectoral disaggregation, we find that the growth pattern of firms outside the traditional sectors is not as fragile as suggested by the results on the whole sample. The results leave the door open to many possible and different interpretations. Although it is very difficult to identify the channels through which the inter-sectoral differences in the competitive and technological environments impact on the demand (and supply) of debt and ultimately on the firms’ financial structure-performance relationship, a conjecture might be proposed. In sectors dominated by low innovation and technological competition, firm growth can be conditioned by the costs of growth, hereby including coordination costs and the fear of dependence on external investors. The prevalence of short-term debt, as well as the role played by non financial debt in allowing firm growth confirm this evidence. This is less the case for more technologically advanced industries, where firms use a plurality of financial sources in order to sustain their investments, and where - with the exception of science-based sectors - the role of bank debt is crucial. A last remark is in order: all the relationships detected through the empirical analysis can not be interpreted as causal ones. Each single dimension of firm financial structure is undoubtedly dependent upon the real side of firm activities, and vice versa. Equations (2) and (3) should thus be conceived as part of a structural model where all financial and real variables are endogenously determined. This reverse causality among financial indicators and firm growth prevent us to give our results any causal interpretation and rather imposes to accept them as descriptive evidence, calling for future research both at an empirical and theoretical level. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 24 References Abrevaya, J., 2001, The effects of demographics and maternal behavior on the distribution of birth outcomes, Empirical Economics, 26, 247-257. Bottazzi, G., Secchi, A, and F. Tamagni, 2006, Financial Fragility and Growth Dynamics of Italian Business Firms, LEM Working Paper 2006/07 Sant’Anna School of Advanced Studies, Pisa, Italy. Buchinsky, M., 1998. Recent Advances in Quantile Regression Models: A Practical Guide for Empirical Research, Journal of Human Resources, 33, 1, 88-126. Carpenter, R. E. and B. C. Petersen, 2002, Is the growth of small firms constrained by internal finance?, The Review of Economics and Statistics, 84(2), 298-309. Coad, A. and R. Rao, 2006, Innovation and Firm Growth in High-Tech Sectors: A Quantile Regression Approach, Sant’Anna School of Advanced Studies, LEMWorking Paper 2006/18. Fagiolo, G. and A. Luzzi, 2006, Do liquidity constraints matter in explaining firm size and growth? Some evidence form the Italian manufacturing industry, Industrial and Corporate Change, 15(1), 1-39. Fazzari, S. M., R. G. Hubbard and B. C. Petersen, 1988, financial constraints and corporate investment, Brooking Papers on Economic Activity, 1, 141-206. Gibrat, R., 1931, Les Inegalités Economiques, Librairie du Recueil Sirey, Paris. Gould, W. W., 1997, sg70: Quantile regression with bootstrapped standard errors, Stata Technical Bullettin, 9, 19-21. Hall, B., 1992, Investment and research and development at the firm level: does the source of financing matter?, NBER Working Paper 4096. Honjo, Y., Harada, N., 2006, SME policy, financial structure and firms growth: evidence from Japan, Small Business Economics, 27, 289-300. Kaplan, S. N. and L. Zingales, 1997, Do investment-cash flow sensitivities provide useful measures of financing constraints?, The Quarterly Journal of Economics, 169-225. Kaplan, S. N. and L. Zingales, 2000, Investment-cash flow sensitivities are not valid measures of financing constraints, The Quarterly Journal of Economics, 707-712. Koenker, R. and G. Bassett, 1978, Regression quintiles, Econometrica, 46, 1, 33-50. Koenker, R. and K. F. Hallock, 2001, Quantile regression, Journal of Economic Perspectives, 15, 4, 143-156. Michaelas, N., F. Chittenden and P. Poutziouris, Financial policy and capital structure choice in the U. K: SMEs: Empirical evidence from company panel data, Small Business Economics, 12, 113-130. Oliveira, B. and A. Fortunato, 2006, Firm growth and liquidity constraints. A dynamic analysis, Small Business Economics, 27(2), 139-156. Pavitt K., 1984, Sectoral patterns of technical change: towards a taxonomy and a theory, Research Policy, 13, 343-373. Rajan, R. G. and L. Zingales, 1995, What do we know about capital structure? Some evidence from international data, The Journal of Finance, 1421-1460. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 25 Silverman, B. W. (1986) Density Estimation for Statistics and Data Analysis. London: Chapman and Hall. Schiantarelli, F. and A. Sembenelli, 1997, The maturity structure of debt. Determinants and Effects on firms’ performance: evidence from the United Kingdom and Italy, The World Bank Policy Research Working Paper 1699. Schiantarelli, F., 1996, Financial constraints and investment: methodological issues and international evidence, Oxford Review of Economic Policy, 12, 70-89. Woolridge, J. M., 2002, Econometric analysis of cross section and panel data, Cambridge: The MIT Press. WORKING PAPER SERIES N. 02 - MAY 2010 ■ 26 UniCredit & Universities Knight of Labor Ugo Foscolo Foundation Via Broletto, 16 20121 Milan Italy Giannantonio De Roni – Secretary General Tel. +39 02 8862 8039 e-mail: giannantonio.deroni@unicreditgroup.eu Annalisa Aleati - Scientific Responsible Tel. +39 02 8862 2668 e-mail: annalisa.aleati@unicreditgroup.eu Info at: unicreditanduniversities@unicreditgroup.eu www.unicreditanduniversities.eu WORKING PAPER SERIES N. 02 - MAY 2010 ■ 27 1

DOCUMENT INFO

Shared By:

Categories:

Tags:

Stats:

views: | 2 |

posted: | 12/28/2011 |

language: | |

pages: | 29 |

OTHER DOCS BY yurtgc548

Docstoc is the premier online destination to start and grow small businesses. It hosts the best quality and widest selection of professional documents (over 20 million) and resources including expert videos, articles and productivity tools to make every small business better.

Search or Browse for any specific document or resource you need for your business. Or explore our curated resources for Starting a Business, Growing a Business or for Professional Development.

Feel free to Contact Us with any questions you might have.