1 Introduction

A stylized fact nowadays, though not sufficiently understood, is the absence of debt in some firms’ balance sheets. Indeed, previous literature shows evidence that a significant percentage of firms operate with almost zero or even zero leverage (Boustanifar & Verriest, 2022; Saona et al., 2023). Firms, because they cannot obtain (or abstain from) debt financing, do not rely on this financing instrument to sustain growth. The “zero‐leverage puzzle” phenomenon characterizes large quoted firms in the US (e.g., Strebulaev & Yang, 2013), United Kingdom (e.g., Dang, 2013), and other developed and emerging markets (e.g., El Ghoul et al., 2018; Huang et al., 2017; Saona et al., 2020), as well as unquoted firms, namely Small and Medium Sized Enterprises (SMEs) (e.g., Bigelli et al., 2014; Lefebvre, 2021).

The literature offers several explanations for why firms choose a zero-leverage capital structure. Both firm-level characteristics and business environment effects help explain empirically the decision to eschew debt (Saona et al., 2020). At the firm level, relevant determinants include size, tangibility, non-debt tax shields, growth opportunities, internal governance mechanisms, liquidity, and profitability (e.g., Morais et al., 2020; Saona et al., 2020). Moreover, country-level quality of governance indicators and regulatory quality indicators influence the likelihood of being a zero-debt firm (e.g., Morais et al., 2020, 2022b; Saona et al., 2020). Regarding the business environment, one of the determinants of this phenomenon, already identified in the literature, is the level of information asymmetries between lenders and borrowers (e.g., Lai, 2011).

Our paper contributes to the literature on the conservative financing phenomenon by exploring a specific characteristic of the business environment. Does local bank branch density matter for the decision to eschew debt? The local banking development (LBD) literature offers relevant insights. First, a higher density of cooperative banks at the county level favors debt and lowers debt costs (Hasan et al., 2017). Similarly, LBD positively relates to SMEs’ leverage ratios in several European countries (Fasano & Cappa, 2022; Fasano & La Rocca, 2023a). Second, LBD positively influences new firms’ use of debt (Deloof et al., 2019). Third, LBD negatively affects the cash holdings of indebted SMEs (Fasano & Deloof, 2021), corroborating the hypothesis that LBD reduces the need to hold precautionary cash by facilitating access to bank credit. Fourth, the physical presence of banks in Italy contributes to a redistribution of loans via trade credit (Deloof & La Rocca, 2015).

We make several contributions to the literature on the zero-leverage phenomenon and LBD. First, this is the first study to address the impact of LBD on the likelihood of being an already established zero-leverage SME firm (ZLF). The accessibility of debt, signaled by a higher bank branch density, may contribute to firms’ decision to avoid debt. In addition, we focus on ZLF with established business models, complementing the existing literature on the use of debt by new firms (e.g., Deloof et al., 2019). Second, it addresses a sample of privately held SMEs understudied in the literature on the zero-leverage phenomenon (Lefebvre, 2021). Third, it studies a small, open, developed economy, Portugal, where public equity and debt markets are much less prevalent than in other developed countries, thus rendering increased importance to bank credit as a source of financing. As in many countries, the economic importance of SMEs is noticeable. For example, in 2021, SMEs represent, in Portugal, about 70% of the value added and 80% of the employees, and their Access to Finance Index was the seventh worst in Europe (Kraemer-eis et al., 2022). Fourth, we control for local banking structure to account for the heterogeneous relationship banking models of local branches of national banks versus branches of local cooperative banks (e.g., Hasan et al., 2017).

Empirically, our paper relies on a sample of 7,448 SMEs located in mainland Portugal, observed between 2016 and 2021. It uses multivariate panel logistic models to estimate the relationship between local bank branch density and a binary variable that identifies conservatively financed firms, controlling for firm-level and municipality-level characteristics that relate to the debt financing by firms.

First, our results show that the increase in the lagged density of bank branches in a given municipality is associated with an increase in the odds of being a conservatively financed firm. Second, the relative importance of local cooperative bank branches negatively affects the likelihood of being a conservatively financed SME. Third, the municipality’s socio-economic development (crime rate, manufacturing employment) influences the likelihood of a zero-leverage financing policy. Fourth, we show that firm age, liquidity, and profitability contribute positively to eschewing debt, while growth and tangibility contribute negatively to the likelihood of eschewing debt. Fifth, the effect of bank branch density on the likelihood of eschewing debt is economically more important for long-term than short-term debt.

The remainder of this paper is organized as follows. Section 2 describes the Portuguese banking sector. Section 3 reviews recent literature on the importance of LBD to SMEs. Section 4 describes the data and the methodology. Section 5 presents and discusses the results. Finally, Section 6 concludes.

2 Overview of the banking structure in Portugal

Since the mid-80s, the banking sector in Portugal has undergone a process of liberalization, accompanied by privatizations and the emergence of new private banks. Competition increased, and a period of consolidation, with mergers and acquisitions, with a pike in the year 2000, took place in the Portuguese banking sector (Bonfim et al., 2011). Over time, the share of the five largest banks increased, and the bank branches per capita decreased because Portuguese banks needed to rationalize their costs and increase profitability (Bonfim et al., 2011). In 2018, Portugal was still characterized by high branch density, ranked in the 14th position on a World Bank list of 230 countries (Bonfim et al., 2021). However, in 2021, the number of inhabitants per branch was similar to the Euro Area average, and the productivity (measured by assets per branch) was lower than the Euro Area average (Associação Portuguesa de Bancos, 2022).

In addition to national banks, one important network of local cooperative banks is present nationwide: Caixa de Crédito Agrícola Mútuo (CCAM). According to the Banco de Portugal (2023), in April 2023, out of 3,921 bank branches in Portugal, 656 were CCAM branches. The CCAM network was established with similar objectives to those of cooperative banking across Europe (see Minetti et al., 2021). Although cooperative banking models are different in each European country, they share characteristics that are fundamental pillars such as membership, social commitment, benefits to the local community, and the commitment to territorial cohesion (Castelló et al., 2018). The regulation of the Portuguese cooperative banks reflects the Community Directives in this field and the network of CCAMs is a member of the European Association of Cooperative Banks (EACB).

Figure 1 illustrates the evolution of the Portuguese banking sector in terms of the total number of bank branches and employees (Panel A) and CCAM branches and CCAM employees (Panel B). Between 2013 and 2021, the average annual growth rate of the CCAM branches and CCAM employees is, respectively, -1,1% and -0,5%. For the same period, Germany and Italy, two countries with a significant presence of this bank type, show, an average annual growth rate of the cooperative bank branches of -5,8% and -0,9%, respectively. The average annual growth rate of the number of employees in cooperative banks in Germany and Italy is -1,4% and -1,2%, respectively. The trend in the number of branches and employees in the cooperative banking sector is therefore similar in these countries.Footnote 1

Fig. 1
figure 1

The evolution of the Portuguese banking sector

Full size image

The increase in competition, fostered by liberalization and the European Union membership, created the conditions for a robust expansion of credit (data from Banco de Portugal shows that between 1995 and 2011, banks’ credit to non-financial corporations increased 362% (10% annual average rate). However, the deterioration of government finances led to the sovereign debt crisis in 2011, which, together with the international financial crisis, led to high unemployment levels and negative growth rates. In 2011, the country received a 3-year Economic and Financial Assistance Program. As a result, Portuguese banks experienced an increase in non-performing loans (NPL), the devaluation of their financial assets, the increasing cost of funding, and the imposition of deleveraging and higher capital ratios. Until the first quarter of 2019, the growth rate of loans to non-financial companies was always negative (Banco de Portugal, 2018, 2022). In 2020, bank loans to SMEs grew, and in 2021, due to the COVID-19 pandemic, they also grew but at a lower rate (Banco de Portugal, 2022).

Concerning the CCAM network, in the period after the sovereign debt crisis and until 2018, it presents the lowest loan-to-deposit ratio among the larger commercial banks operating in Portugal (representing more than 90% of the banking activity), more robust capital ratios and a ratio of NPL (return on assets) lower (higher) than the average (Faria & Pacheco, 2022). The ratios were maintained until 2021 (Grupo Crédito Agrícola, 2022).

Over the years, Portuguese banks adopted new technologies that transformed their relationship with clients. According to the European Banking Federation, digital transformation is a priority for Portuguese banks, and substantial progress has been achieved.Footnote 2

3 Conservative financing policy and local banking development

The absence of debt in some firms’ balance sheets (e.g., Strebulaev & Yang, 2013) challenges the assumptions of the trade-off theory (Kraus & Litzenberger, 1973), which posits that firms choose an optimal leverage ratio that maximizes firm value by combining the benefits (tax gains) and costs (financial distress costs) of debt. Haddad and Lotfaliei (2019) challenged this view grounded on the static trade-off theory, claiming that ZL can be optimal if the ideal time for issuing debt is considered. It also challenges the pecking-order theory, which argues that firms have a hierarchy of financing policies according to their costs (Myers, 1984; Myers & Majluf, 1984). Indeed, Saona et al. (2023) emphasize that theoretical frameworks cannot explain ZL policy. However, previous literature provides evidence that many firms operate with almost zero or even zero leverage (e.g., Boustanifar & Verriest, 2022; Saona et al., 2023).

There are several explanations for firms to opt for a zero-leverage capital structure, such as the country’s financial development (Rajan & Zingales, 1998), the cultural characteristics, namely ethnic and religious beliefs and values (Zheng et al., 2012) and corporate governance mechanisms and managerial entrenchment (Strebulaev & Yang, 2013). In addition, according to the flexibility hypothesis, firms can desire to keep financial flexibility (Dang, 2013; Huang et al., 2017; Morais et al., 2020, 2021). Financial constraints may render debt financing inaccessible, forcing firms to eschew debt by restriction rather than choice (Bessler et al., 2013; Dang, 2013; Fitzgerald & Ryan, 2019). Finally, looking at the business environment, the credit favorableness of the legal environment, the industry, the level of information asymmetries between lenders and borrowers, and credit market momentum also affect the amount of credit offered, and thus the likelihood of being debt-free (Miglo, 2020; Saona et al., 2020; Villarón-Peramato et al., 2018).

LBD is important for entrepreneurs and firms. It contributes to growth (Guiso et al., 2004), employment (Petach et al., 2021), SME performance (Hossain et al., 2021), and reduces the probability of bankruptcy for young SMEs (Arcuri & Levratto, 2020).

In bank-based economies, the local development of financial services relies on bank branches, which play a crucial role in access to credit. While technological development has facilitated financial transactions, credit operations remain highly information-intensive and prone to information asymmetry problems (Agarwal & Hauswald, 2010). This is particularly relevant for SMEs for whom information asymmetry problems are more severe since they are more informationally opaque (Agarwal & Hauswald, 2010). For example, SME managers motivated by trust-related factors prefer to choose a small local bank and maintain a close and long relationship (Jackowicz et al., 2020).

Geographical proximity matters because it reduces the cost of transmitting information and facilitates relationships (Nguyen, 2019; Zou & Wang, 2022). Borrower proximity eases the collection of firm intelligence and soft information (Agarwal & Hauswald, 2010), and positive shocks on bank liquidity are transmitted to nearby borrowing SMEs (Petkov, 2023). Banks enjoy a local informational advantage that distance erodes (Agarwal & Hauswald, 2010). Using branch-firm distance as a proxy for asymmetric information, Zou and Wang (2022) report that physical distance impairs the transmission of soft information and decreases bank lending volume. Fasano and La Rocca (2023a) and Khan (2023), analyzing six European markets and the market of Bangladesh, respectively, report that the geographical proximity of local bank branches helps SMEs obtain financial resources. However, the national banking development has a moderate effect on SME debt since the positive impact of local banks tends to decrease as the level of national banking development rises (Fasano & La Rocca, 2023a).

Local cooperative banks favor the creation of new firms, improve SMEs’ access to financing, and provide lower financing costs (Hasan et al., 2017). A high share of cooperative bank branches is negatively associated with the level of financial costs of small businesses (Hasan et al., 2017). The level of provincial banking development positively influences the use of debt by Italian SMEs (La Rocca et al., 2010), and the results are similar in the Spanish market (González and González (2008); Palacín-Sánchez & Di Pietro, 2016). This evidence suggests that LBD simplifies information acquisition, reducing the information asymmetry between banks and SMEs (Cowling et al., 2020; Fasano & La Rocca, 2023a).

LBD contributes to the financing of new firms (Deloof et al., 2019). New firms in Italian provinces with higher branch density are more likely to use bank debt. The positive effect on the use of debt by new firms is lower for local banks than for national banks, yet the credit provided by local banks is higher than that assured by national banks (Deloof et al., 2019). Finally, the authors found an adverse effect on foreign bank density, which they interpreted as resulting from an informational asymmetry problem.

LBD also affects trade credit. According to Cassia and Vismara (2009), firms tend to obtain funds from suppliers when their chances of obtaining loans from banks are not easy, suggesting that suppliers and banks can be considered substitutes in short-term financing. SMEs with better access to credit can more easily channel funds to financially weaker customers (Deloof & La Rocca, 2015). However, a higher share of cooperative banks significantly reduces the positive effect of branch density on trade credit: loan decisions of cooperative banks are based on soft information and are, therefore, a substitute for trade credit. Furthermore, Fasano and Deloof (2021) show that the cash balances of Italian SMEs are negatively related to LBD, as better access to bank financing reduces the need for cash buffers.

Technological development can reduce the geographical barriers between SMEs and banks and alleviate asymmetrical information (Fasano & La Rocca, 2023b; Lu et al., 2022; Lv et al., 2021). Galardo et al. (2021) use broadband diffusion at the municipality level as a proxy for the availability of remote banking channels. They find that the probability of closing a branch is lower for those cities where the broadband is less diffused: banks are better at substituting brick-and-mortar branches where remote channels are available. According to Fasano and La Rocca (2023b), new technologies reduce the positive influence of geographical proximity and have a moderate effect on bank-firm proximity. The authors also report that this effect is not relevant in the case of cooperative bank branches where soft information is available. Internet banking can also impair the transmission of soft information, with negative consequences for SME financing. The growth of Internet banking in Italy from 2013 to 2019 is negatively related to the level of SME bank debt (Fasano & Cappa, 2022).

According to the literature on the firms’ conservative financing policy, the credit-friendliness of the legal environment and the degree of information asymmetry between lenders and borrowers are determinants of the firms’ leverage. The LBD literature emphasizes the role of LBD in reducing the information asymmetry between banks and SMEs, thereby facilitating SME financing. Therefore, we expect to find a (negative) relationship between the SMEs’ probability of being a ZLF and LBD.

However, we do not discard an opposite sign effect. Firms may not make use of debt because they wish to maintain their financial flexibility and assume they can obtain it in the future without access barriers. Indeed, several authors find evidence that the financial flexibility hypothesis is one of the main factors that lead firms to choose a ZL policy (e.g., Bessler et al., 2013; Dang, 2013; Huang et al., 2017; Iliasov & Kokoreva, 2018). They tend to hold cash funds to address contingency situations (Bessler et al., 2013). Therefore, it is possible to find a positive relationship between the SMEs’ probability of being a ZLF and LBD, as a more favorable banking framework can contribute to postponing debt. Furthermore, technological development can facilitate the transmission of information between SMEs and banks, complementing physical closeness (Fasano & La Rocca, 2023b). If a technological link becomes important, LBD loses importance in SME funding.

4 Research design

4.1 Data and sample

We combine data from two data sources. At the municipality level, we use INE, the Portuguese official statistical office, for bank branches, macroeconomic, and social data. At the firm level, we rely on the SABI database from Bureau van Dijk, a Moody’s Analytics Company.

In Portugal, municipalities correspond to a territorial grouping between the NUTS3 Eurostat’s classification and the lower-level Eurostat’s LAU grouping. We prefer the municipality classification because it remains stable over the sample period, and most data is unavailable at the LAU level. Specifically, for each municipality, from INE, we collect the number of national and local cooperative bank branches, the resident population, and the area (in square kilometers). We also collect data on each municipality’s non-financial firms’ Gross Value Added, the number of reported crimes, and the number of personnel employed by enterprises by sector of economic activity, as these variables may be correlated with local financial development (e.g., Deloof et al., 2019).

Our sample period spans from 2016 to 2021. NUTS regional classification and Portuguese accounting standards changed in 2015, and most of the municipality-level data available ends in 2021. This period comes after the financial turmoil resulting from the sovereign debt crisis that affected Portugal in the recent past. In addition, it covers the COVID-19 pandemic period, whose impact is captured by introducing year dummy variables.

For building our sample of unlisted non-financial SMEs from the SABI database, we started to select all active firms with accounting information available for each year between 2015 and 2021 (we specify 2015 to allow for growth rate calculation) located in mainland Portugal and incorporated before 2011. By imposing at least five years of operation in the initial year of our sample period, we aim to abstract from analyzing the impact of LBD on the use of debt by new firms (see Deloof et al., 2019) and focus on already established SMEs.

Next, we selected SMEs, following the European Commission definition,Footnote 3 with primary activity classified in NACE Rev2 sections A to C and F to J. Therefore, we exclude financial, research and development, public administration, education, arts, human health, social work, sports, and membership organizations. The result is a sample of 8,084 Portuguese non-financial SMEs. Finally, we discarded firms with input errors (e.g., negative total assets or other necessarily positive accounting information), with negative equity (a sign of financially distressed firms), zero sales, or with extreme year-to-year variation in sales or assets to account for extraordinary operations like a business restructuration.Footnote 4 The final sample consists of 7,448 Portuguese SMEs from 261 municipalities in mainland Portugal.

4.2 Variables

To determine the importance of banking development to the debt avoidance policy of firms (e.g., Strebulaev & Yang, 2013), we consider several dependent binary variables that identify firms following a conservative financing policy. First, we identify firms that have no debt in their financing structure. A ZLF in a given year is a firm that has zero long-term and short-term debt outstanding.Footnote 5 Second, we identify a firm as an almost-zero debt firm (AZL) in a given year if its total debt-to-assets ratio is below 5%. Finally, as cash can be considered negative debt, we identify firms with nonpositive net debt (NPND), measured by a nonpositive difference between total debt and cash holdings.

Table 1 describes our sample coverage. About 13% of observations correspond to ZLF, in line with the published evidence (between 10 and 25%, e.g., Lefebvre, 2021). The industry coverage of our sample is dominated by the “Manufacturing” and the “Wholesale and retail trade” sectors. However, the ZLF seems more prevalent in the “Information and telecommunication sector.”

Table 1 Sample coverage and conservatively financed SMEs
Full size table

Our independent variable of interest is the total number of bank branches per capita (TBB), including national and local cooperative bank branches, in the municipality where the firm is domiciled (e.g., Deloof et al., 2019; Fasano & La Rocca, 2023a) since we aim to determine the influence of physical banking services on the decision to avoid debt.

Moreover, we include controls to account for the municipalities’ economic and social characteristics. First, we control for differences in the municipalities’ banking structure with the ratio of local cooperative bank branches to total bank branches (Bankstructure) to account for different personal proximity to clients (e.g., Jackowicz et al., 2020). Second, local year-by-year gross value-added growth (GVA) proxies for increased demand for credit resulting from improved local economic development (GDP data is unavailable at the municipality level). Third, we use the ratio of employment in manufacturing to total employment (IndWork) as a measure of industrial density because the close relationship between similar firms may facilitate bank information collecting (e.g., Deloof et al., 2019). Fourth, we use the local crime rate (Crime) to proxy for local trust, a key ingredient to facilitate bank lending (e.g., Deloof et al., 2019). Fifht, to proxy for historical differences in financial development between municipalities, we add to all regressions a binary variable (RDCI) that takes the value 1 for those municipalities integrated into NUTS3 regions that score above 100 in the INE’s Regional Development Composite Index in a given year between 2016 and 2020 (the last year available).

We also control for firm characteristics, using SABI data on factors that characterize SMEs (e.g., Bigelli et al., 2014; Lefebvre, 2021). Our control variables include the natural logarithm of years since incorporation to quantify firm age (AGE), the natural logarithm of total assets to measure the firm size (SIZE), and the ratio of tangible assets to total assets to measure firm tangibility (TANG). These variables relate to the easiness of access to debt financing (La Rocca et al., 2010). Additionally, we control for cash preferences, considering the ratio of cash to assets (LIQUIDITY) as higher internal financing or cash accumulation may render external debt unnecessary, for non-debt tax shields (NDTS) using the ratio of depreciation and amortization costs to total assets since firms may be more incentivized to eschew debt if the non-debt tax shields are high,Footnote 6 and for growth opportunities (GROWTH) using the year-by-year growth in sales. Firms may abstain from incurring debt to preserve borrowing capacity in the case of future investment opportunities. The ratio of earnings before interest and taxes to total assets controls profitability (PROF). Finally, we include industry (by NACE Rev 2 section) and year dummy variables to account for unobservable heteroscedasticity.

Table 2 presents descriptive statistics for the independent variables. The average branch density (0.47) is slightly lower than in other European countries (Fasano & La Rocca, 2023a). Concerning the municipal bank structure (Bankstructure), the local cooperative bank branches represent, on average, 14.79% of the total bank branches, slightly less than in Poland (e.g., Jackowicz et al., 2020).Footnote 7

Table 2 Independent variables descriptive statistics and correlation matrix
Full size table

4.3 The model

For estimation, we pool the data for all firms in all years. Because of the binary nature of our dependent variables, we use maximum likelihood methods to estimate the following logit baseline model:

$$\mathrm{Pr }({{\text{DV}}}_{{\text{i}},{\text{m}},{\text{t}}} = 1) = F({\upbeta }_{0}+{\upbeta }_{1} {{\text{TBB}}}_{{\text{m}},{\text{t}}-1} + {{\text{B}}}_{2} {{\text{FLC}}}_{{\text{i}},{\text{t}}-1} + {{\text{B}}}_{3} {{\text{MLC}}}_{{\text{m}},{\text{t}}-1})$$
(1)

where DVi,m,t is the binary dependent variable for firm i, on municipality m, at time t (either a ZLF, an AZL, or NPND), and TBBm,t-1 is the bank branch density in municipality m, at time t-1. FLCi,t-1 is a vector of firm-level controls at time t-1, and MLCm,t-1 is a vector of municipality-level controls at time t-1. F is the cumulative logistic distribution function.

In all regressions, the time-dependent explanatory variables are lagged one year to minimize reverse causality bias. In addition, we add industry (by NACE Rev 2 section) and time (years) indicator variables. Moreover, all standard errors are heteroscedasticity robust.

5 Results

5.1 Baseline results

Table 3 presents our baseline results. Looking at the three columns (that refer to ZLF, AZL, and NPND firms, respectively), we show that an increase in the lagged density of bank branches in a given municipality is associated with an increase in the odds of being a conservatively financed firm. This result suggests that SMEs avoid and postpone debt in a favorable LBD framework. In a given municipality, a more favorable credit environment reduces asymmetrical information, rendering debt financing easier for firms, thus enabling them to postpone its use.Footnote 8

Table 3 Bank branch density and the conservative financing phenomenon
Full size table
Table 4 Robustness test—small sample
Full size table
Table 5 Robustness test—linear probability model IV-GMM
Full size table
Table 6 Branches density, debt maturity, and conservative financing policy
Full size table

The evidence that banks' branch density at the municipality level is positively related to the SMEs’ decision to eschew debt is not aligned with the idea of reduction of information asymmetry between banks and SMEs (Cowling et al., 2020; Fasano & La Rocca, 2023a). However, it supports the empirical evidence that firms are more likely to follow a ZL policy for financial flexibility reasons (Bessler et al., 2013; Dang, 2013; Huang et al., 2017; Iliasov & Kokoreva, 2018).

Another contribution to explaining the result is the possibility of transmitting hard information through technological channels, reducing information asymmetry and rendering physical closeness less important (Fasano & La Rocca, 2023b). According to Eurostat, in 2022, 30.8% of the companies in Portugal have high digital intensity, a percentage higher than Europe27’s average and Italy (where LBD is deeply studied), respectively, 28.1% and 25%.Footnote 9

Furthermore, Portugal is one of the European countries with the lowest percentage of the financially literate population (Klapper & Lusardi, 2020), and the financial literacy of SME executives helps to provide access to credit and to reduce financial constraints (García-Pérez-de-Lema et al., 2021).

Table 3 also shows that the estimated coefficients for the municipal banking structure are negative and statistically significant. Therefore, firms in municipalities with a higher proportion of local cooperative bank branches are less likely to be conservatively financed. This result supports the hypothesis of Hasan et al. (2017) that local cooperative banks provide more loans to SMEs than other banks. Note that better liquidity conditions of the CCAMs, as these banks have lower loan-to-deposit ratios, as mentioned above, could also contribute to this result.

Concerning our other lagged control variables at the municipality level, we see that firms located in high-developed areas of Portugal tend to have a higher likelihood of being conservatively financed. On the contrary, firms in municipalities with higher crime rates and a higher percentage of employees in the manufacturing sector are less likely to adopt conservative financing practices.

Regarding our lagged control variables at the firm level, we see that older, more liquid, and profitable firms are more likely to be conservatively financed. Also, faster-growing firms and firms with more tangible assets are less likely to eschew debt. These findings corroborate most empirical evidence on the determinants of the zero-leverage phenomenon, supporting the financial flexibility view (e.g., Boustanifar & Verriest, 2022; Morais et al., 2022a).

5.2 Robustness and additional tests

Table 4 presents the results of a robustness test on sample selection. First, we discard firms from the “Manufacturing” and “Wholesale and retail trade” industries because they have a disproportional higher representation in our sample. Then, we estimate model (1) on the remaining small sample. The new results corroborate our baseline results, both for the positive influence of bank branch density on the likelihood of eschewing debt and for the control variables, including the BankStructure variable.Footnote 10

Table 5 presents an additional robustness test that more thoroughly addresses endogeneity concerns motivated by omitted variable bias. Specifically, based on Deloof et al. (2019), we use the total bank branch density in each municipality in 2001 as an instrument for municipal branch density (as this is the oldest data available and the banking institutional framework has not changed since then). We complemented it with the exogenous determined municipality area (in Km2), to allow for the overidentification test of the selected instruments. According to Galardo et al. (2021), physical distance matters for branch network decisions. We also use a linear probability model and rely on the more efficient IV-GMM estimator. The insignificant Hansen J-statistic corroborates the statistical validity of the instruments.

The results show that our baseline findings are robust to the estimation method and endogeneity concerns. Municipality’s bank branches density is positively associated with the likelihood of being a conservatively financed firm. Moreover, the control variables at the firm and municipality levels retain their statistical significance and sign.

Finally, Table 6 presents an additional test focusing on the debt maturity structure. First, we create four new binary variables that take the value one for firms with zero long-term debt (ZLF_LT), zero short-term debt (ZLF_ST), almost zero (less than 5% relative to assets) long-term debt (AZL_LT), and nearly zero (less than 5% relative to assets) short-term debt (AZL_ST), and zeroes otherwise. Then, we estimate the baseline model (1) using each one as the dependent variable.

We show that branch density positively affects both long-term and short-term decisions of not using debt. However, columns ZLF_LT and ZLF_ST show that the effect seems economically more important for the long-term decision (a unit increase in branch density implies an increase in the odds ratio of 2.1 = e0.7646) than for the short-term decision (an increase in the odds ratio of 1.5 = e0.4216). The almost zero-leverage firms’ results corroborate this result.Footnote 11

6 Conclusion

In a small, open, and bank-based economy like Portugal, in recent times, local bank branch density relates positively to non-financial unquoted SMEs’ likelihood of being conservatively financed. We reach this conclusion controlling for several firm-level and local-level factors related to firms' debt rejection. In particular, we show that the effect of branch density is more noticed for long-term debt-avoiding decisions than for short-term debt-avoiding ones. However, the weight of local cooperative banks is negatively related to the odds of being a conservatively financed SME. Therefore, results suggest that national bank branches influence the debt initiation decisions of SME managers differently from the local cooperative bank branches.

We offer two potential explanations. First, firms may see higher local bank branch density as making debt more accessible, thus postponing the debt financing decision for future needs. This corroborates the financial flexibility hypothesis for eschewing debt. Second, the digital transition processes banks are implementing, together with enhanced digital communications channels, may render banks’ physical proximity with SMEs less important for debt financing initiation, particularly for national banks.

Our study is not exempt from limitations. Unfortunately, the data we have access to does not provide the means to test the alternative explanations we offer. Also left for future research is one question motivated by our results. Are local cooperative banks more debt-friendly than national banks? Our results highlight the impact of experiencing a higher presence of local cooperative banks’ branches relative to national banks’ branches on SMEs’ debt financing policy. Still, they are silent on the potential procedures implemented for credit evaluation in each market segment.

One may identify several implications of our results for SMEs, financial institutions, and policymakers. First, SMEs valuing their financial financial flexibility, should consider the potential benefits of debt to their business strategy and not take for granted the potencially higher credit availability suggested by the higher bank branch density in their municipalities. SMEs also need to recognize and incorporate banks’ digital information channels as a feature of the business environment, particularly anticipating artificial intelligence techniques to support banks’ credit decisions so as not to be ignored in the credit market and to help reduce information asymmetries. Despite the increasing significance of digital channels, this study emphasizes the role of local cooperative banks in SMEs’ debt financing, aligning with Fasano and La Rocca (2023b). Hence, SMEs should not ignore the benefits of a close relationship with banks fostering the exchange of soft information and favoring financing alternatives.

Second, lower bank branch density does not seem to have contributed to reducing the number of potential SME borrowers. However, specific SME segments may easily become left out. Banks may have to consider this limitation of relying solely on centralized hard information-based decision processes and consider the potential benefits of integrating human soft information-based mechanisms and building personalized customer relationships.

Third, policymakers, also contributing to a sustainable future, could promote and implement measures supporting SME digital transformation and fostering SME financial literacy, particularly concerning digital business information reporting, thus supporting SME participation in the already digitalized bank credit market, benefiting from debt without delaying projects. Policymakers are frequently concerned with the consolidation of the local banking system and its influence on the SMEs’ credit. Our results are challenging, as they show that the weight of local cooperative bank branches is associated with firms less likely to eschew debt. This suggests that regulations that strengthen and promote the cooperative banking sector may favor SMEs’ access to external resources.

Finally, the evidence that local banks' branch density at the municipality level increases the likelihood of SMEs following a conservative financing policy is important for researchers. This result may be explained by reasons of financial flexibility (e.g., Dang, 2013; Huang et al., 2017; Morais et al., 2020, 2021) but not by the reduction of information asymmetry between banks and SMEs (e.g., Cowling et al., 2020; Fasano & La Rocca, 2023a). Therefore, more research is needed to disentangle the two hypotheses.