Sample and data resources

The information used was obtained from SABI by downloading the Balance Sheets and the Income Statements as of 31 December in the financial years 2008 to 2017 of companies engaged in the agri-food sector. SABI is a database of Bureau van Dijk that contains the accounting information of all Spanish firms that deposit their annual statement in the registry according several business laws.

We downloaded information from 159 CNAE tree-digit-codes and grouped them into four activities according to ICEX (2017). As can be seen in Fig. 1, the agri-food system comprises the activities of the primary sectors (agriculture and livestock) and processing industry as the first links of the agri-food chain, and there is a strong relationship with suppliers and services. The distribution companies (transport and wholesale) and those responsible for the retail marketing of products and the provision of catering services are the final link in the chain, establishing the direct link with the final consumer.

The time period was selected according to the standard of uniformity of the accounting information contained in the financial statements used in the analysis. Thus, because the current accounting regulations, the PGC (Spanish General Accounting Plan), were implemented in 2008, this date determines the beginning of the study period as the annual accounts of previous years are not homogeneous in terms of masses heritage. The completion year of 2017 stems from it being the last year with information available as of the date of data download.

The population corresponds to 2,600,000 Spanish companies for which financial data are available in the SABI. In the first phase of selection, the 224,050 companies in the agri-food sector that appeared in the database during the period 2008–2017 were identified. Subsequently, information was downloaded from active agri-food firms, companies that had a total asset and net amount of sales figure different from 0 in 1 of the years analysed. The observations for which some measures had missing values were then deleted, and those with extreme values for the variables of interest (i.e., outliers) were also removed. In addition, companies for which their CNAE code was not available were deleted. After this process, the sample comprised 490,856 observations from 45,196 Spanish companies in the agri-food sector for the period 2008–2017.

The information for each company was required to refer to a minimum of three consecutive years to have a data panel with reduced analysis bias resulting from the business turbulence associated with the onset and extinction of companies in the period under investigation. In addition, the requirement of three years is associated with the need for a minimum period of time to control for unobservable heterogeneity and correct endogeneity problems in the analyses. Thus, the final sample contains 41,109 companies (364,020 observations), forming a panel of companies that is not complete (unbalanced panel) in the sense that the information is not available for all the companies in all the years analysed.

Research hypothesis

Innovation promotes greater productivity and competitiveness, but making such investments requires an adequate financing policy. Previous scholars posit that R&D investment is negatively (positively) associated with leverage (the capital structure) because innovation creates primarily intangible assets that are not good collateral (Simerly & Li, 2000). In the agri-food sector, the effect of the financial structure on innovation is realised by taking an indirect approximation of the indebtedness effect associated with business size. Thus, some authors find that the larger the business size, the greater the degree of innovation due to better access to human capital and financial resources (Karantininis et al., 2010; Bayona et al., 2013; Hirch et al., 2013). However, other work highlights the innovation capacity of SMEs thanks to their greater flexibility and adaptability to changes (Bayona et al., 2001; Maravelakis et al., 2006; Alarcón & Sánchez, 2013).

In this sense, our proposal aims to analyse the advantages and disadvantages associated with the financial and capital structure for the innovation strategies of agri-food companies. In this regard, various theories about capital structure are proposed in the literature and numerous studies try to analyse the factors that determine companies’ level of indebtedness (Pindado et al., 2011, 2017). However, there is insufficient evidence regarding the effects that indebtedness has on other business dimensions, such as the degree of innovation.

Among the main corporate capital structure theories, the pecking order theory and trade-off theory are the most common. According to the first theory (Myers & Majluf, 1984), companies that need to undertake new investments are turning firstly to self-financing, secondly to debt and thirdly to common stock. Financial preferences have been observed in several papers (i.e., O’ Brien, 2003; Capizzi et al., 2011; Bartolini, 2013), although other authors (i.e., Hall, 1992; Vicente-Lorente, 2001; Hsu et al., 2014) evidenced a negative impact of debt in R&D investment given the high risk associated to an intangible project, what causes that creditors apply higher borrowing costs (Capizzi et al., 2011; Bartoloni, 2013).

In addition, the second of the theories indicates that, to understand the capital structure of companies, other issues must be considered. For example, on the one hand, indebtedness has tax advantages in reducing the payment of taxes. On the other hand, a high level of indebtedness can lead to financial insolvency proceedings. For these reasons, the static balancing theory proposes that there is an optimal level of indebtedness to which companies aspire (de Miguel & Pindado, 2001).

For SMEs, which are normally managed by their owners, according to Hutchinson’s arguments (1995), the adequacy of the financial and capital structure in relation to investment policies is more complex. On one hand, f irms avoid using debt contracting due to this option involving the monitoring of the company by financial institutions. On the other hand, these corporations show an aversion to increasing the current common stock that may lead to the entry of new owners, as it dilutes the power of the current owners. Therefore, taking into account the fact that SMEs could present important limitations on increasing equity from funds provided by the current owners, they can lead to the self-generation of internal funds becoming the policy for financing investment projects that are more suited to their interests (De Massis et al., 2018).

Empirically, previous papers note that corporate investment in physical capital depends to a large extent on the capacity to generate internal funds or self-financing (Pindado & de la Torre, 2009; Pindado et al., 2011) and the self-financing capacity by SMEs is usually not sufficient to meet the necessary investment funds, so it is fundamental to enjoy a good level of current common stock to undertake innovation projects due to (i) the discretion arising from the use of owners’ funds is greater in the face of the reissuing of accounts to lenders and (ii) the intangible nature and high risk of investments in innovation that provoke restrictions in the access to finance. In this regard, we establish the first hypothesis of the paper:

H1: Spanish agri-food companies will use equity ‒ current common stocks and self-financing ‒ to undertake innovation projects.

In addition, following García-Sánchez & García-Meca (2018), it is necessary to take into account the efficiency of innovations investment as less able managers could overinvest (implement projects with a negative NET) or underinvest (reject projects with a positive NET). These decisions, according to agency theory, could be corrected by the presence of debt in the capital structure as banks are specialised institutions in the supervision of borrowers (Petersen & Rajan, 1994). Therefore, debt could be a discipline or control mechanism aimed at a more efficient use of financial resources as a consequence of the effects associated with the risk of bankruptcy in highly leveraged firms and the higher control over the discretionary use of cash flow that guides companies to only finance investment opportunities with positive net present value (Singh & Davidson, 2003).

Spanish businesses[1] are characterised by presenting financial leverage close to 60% (Cutillas & Sánchez, 2014) and 76% of these external funds expire in the short term, so the quality of the indebtedness is quite low (CES-CyL, 2018). However, in line with the previous agency arguments, the shortening of the debt maturity allows for more comprehensive control over management decisions, as it requires continuous renegotiation and may lead to changes in the initial conditions of debt contracts. In other words, short-term debt use will help to control underinvestment and overinvestment, as it allows greater financial flexibility for borrowers and greater control by lenders (Cutillas & Sánchez, 2014). Therefore, we posit our second hypothesis:

H2: Spanish agri-food companies with a financial structure characterised by a larger volume of short-term debt will undertake more efficient innova-tion projects.

Moreover, we have mentioned that Spanish businesses present serious liquidity problems arising from the need for funds of operation (NOF) being higher than the wor-king capital (Baños et al., 2016) and resort to the use of short-term debt as a source of current asset financing due to the difficulty in obtaining long-term financing (Peters et al., 2013). This is because the payment terms for suppliers and creditors are longer than the expiration of receivables. These NOFs will adversely affect innovation strategies in the agri-food sector, leading to underinvestment because funds cannot be allocated to profitable projects when used in the financing of current assets. In addition, underinvest-ment is a severe problem for small companies (MacMa-hon, 2003) due to the concentration of decision making and monitoring of the owner figure, a duality that makes it difficult to consider different points of view and involves less specialisation (Danielson & Scott, 2007). Therefore, we establish the last hypothesis of the paper:

H3: Spanish agri-food companies with liquidity pro-blems underinvest in innovation.

Figure 1.  Phases of the agri-food system

Models and econometric procedures

To test the proposed hypotheses, the following analy-sis models were proposed in which the independent va-riables are ‘Leverage’ in Model 1 and ‘CommonStock’, ‘Self-financing’, ‘Debt_Maturity’ and ‘Liquidity_Pro-blem’ in Model 2. Model 1 was designed to evaluate the first hypothesis, H1, concerning the effect of the capital structure on the investment in innovation projects and the investment efficiency. Model 2 aimed to test the validity of H2 and H3, deepening the analysis of H1.

where the notational term INV is expressed by two diffe-rent dependent variables, ‘Investment’ and ‘Optimal_In-vestment’, i identifies company 1 to company 41,109, t takes values for the years 2008 to 2017, β represents the coefficients to estimate, η, represents unobservable hete-rogeneity, and μ the error term. The definitions of varia-bles are listed in Table S1 [suppl.].

Regression models were proposed to take into account dependency techniques for panel data and methodolo-gies that strengthen the power and explanatory capacity as well as the consistency, which are all associated with the period analysed; to control the specific characteristics of each company that remains unchanged in the period analysed or unobservable heterogeneity; and to facilitate a dynamic study of companies and reduce the aggrega-tion bias that arises when time series analyses are used to characterise the behaviour of individuals. Specifically, for the variable ‘Investment’, we used the two-stage dynamic estimator proposed by Arellano & Bond (1991), based on the generalised method of moments. For the ‘Opti-mal_Investment’ variable, due to its dichotomist nature, we employed a logit regression. In both regressions, we employed two lags for the independent variables to avoid endogeneity problems as previous papers have evidenced in the management and business field. These lags refer to the instrumental variables introduced in the Generalised Method of Moments (GMM) procedure and the inclusion of independent variables in t-2 in logistic regression. To avoid any pattern of multicollinearity, heteroskedasticity and autocorrelation within panel regression models, the robust command is used for the GMM estimator. For lo-gistic regression, we required that the eigenvalues would be close to 0, indicating that the predictors are highly intercorrelated.

The goodness of fit of the logistic regressions was determined by the Log-Likelihood and the Wald test of the joint significance of the coefficients obtained, distri-buted asymptotically as χ2 under the null hypothesis of no relation. In addition to the Wald test for GMM, we estimate two serial correlation tests of order i that use first-difference residuals, distributed asymptotically as N (0.1) under the null hypothesis of no serial correlation and the Hansen test of over-identification restrictions, distri-buted asymptotically as χ2 under the null hypothesis of no correlation between the instruments and the error term.

The first variable of analysis is ‘Investment’, which, following Biddle et al. (2009), is defined as the sum of investment in innovation in capital goods and other in-tangible assets minus cash from sales of similar elements, multiplied by 100 and scaled by the total assets. Second, we proposed an abnormal investment proxy, Optimal_Investment, which is obtained from the residuals of a firm-specific deviation from optimal investment, calcula-ted by estimating an industry–year model of investments as a function of growth opportunities (as measured by the percentage change in sales from year t-1 to year t). The model is described as follows:

However, the previous measure could not be considered as a dependent variable due to this residual identifying different deviations (negative and positive) from the optimal investment, so we ranked this measure in quartiles and rescaled the quartile rankings from 0 to 4. Firm–year observations in the middle quartile are considered to be the optimal investment. Firm–year observations in the bottom quartile (e.g., the most negative residuals) are classified as underinvestment. Observations in the top quartile (i.e., the most positive residuals) are classified as overinvestment.

The independent variable ‘Leverage’ of Model 1 represents the capital structure and is measured using the ratio total debt to total assets. It is recommended that it should be between 0.5 and 0.6, reflecting a balance in the use of equity and liabilities in the financing of the economic structure. The breakdown of this variable in Model 2 entails the incorporation of the independent variable ‘Debt_Maturity’, determined as the degree of short-term enforceability of the debt and measured using the current to total liabilities ratio. Its value must be between 0 and 0.5, indicating that the maturity of most of the debt is long term, while, if the value is greater than 0.5 and close to 1, all debts are enforceable in less than a year. To represent the volume of equity, we incorporated the variables ‘CommonStock’ and ‘Self-financing’, measured as the logarithm of common stock and the accumulated retained earnings, respectively, relativised by the total assets. ‘Liquidity_Problem’ is a dummy variable that takes the value 1 if the company has liquidity problems, 0 otherwise. It is determined by the negative difference between the working capital (current assets minus current liabilities) and the need for funds of operations or NOF (Stocks + Receivables + Operating Treasury – Spontaneous Liabilities). If the working capital is lower than the NOF, it implies that the company does not have sufficient liquidity to meet its operational needs and has to resort to external financing to meet the payment of obligations for commercial operations. In the other case, the company is meeting its operational needs without resorting to other financing as well as enjoying financing through operating debts that do not usually involve explicit costs.

Finally, a series of control variables were included to avoid biased results based on previous studies (García-Sánchez & García-Meca, 2018): ‘Size’, the logarithm of total assets; ‘ROA’ or economic profitability; ‘Growth_Opportunities’ determined as the year-on-year variation in sales over the period; ‘Age’, the logarithm of the age of the company; ‘Cash’, the standard deviation of the cash flows from t-2 to t; ‘Loss’, a dummy variable that takes the value 1 if the ordinary result before extraordinary income and expenses is negative and 0 otherwise; and ‘Inventory’, the duration of the operating cycle. In addition, we control the subsectors, the time period and the time of crisis. For the first two variables, we used categorical measures that represent the four activities in the agri-food industry and each year in the period 2008–2017. The crisis variable is a dummy that takes the value of 1 if t=2008 to t=2012, 0 otherwise.

NotesTop