A. Data

The data set used in the analysis combines several data sources. Information on minibond issuances comes from the Italian Security Register Database (Bank of Italy), which covers both listed and unlisted bonds issued by Italian companies and institutions and is integrated with data from Borsa Italiana (the Italian Stock Exchange in Milan). In line with the reform’s requirements, we define minibonds as debt securities issued by unlisted firms and traded on regulated markets or multilateral trading facilities. Commercial paper with maturity below 1 year is excluded.

We focus on securities issued from 2013 to 2019, thereby excluding minibonds issued during the pandemic period. Although issuances remained robust at that time, many firms issued minibonds primarily for liquidity purposes, supported by public guarantees applicable to both loans and bonds, which may have altered the profile of issuers relative to previous years. We gather information on 540 minibonds issued by 275 firms from 2013 to 2019, for a total issuance amount of EUR 50.2 billion (Table 1). In the population of issuer firms (Table 1, Panel A), most bonds pay a fixed rate coupon (76.7%) of around 5%.

TABLE 1Characteristics of Minibond Issuances

TABLE 1 Long description

The table is divided into two sections.

Panel A, Population of Issuer Firms:

* Total issuances: 540 bonds from 275 firms, valued at 50,228 million E U R, with a 76.7 percent fixed rate share and 5.0 percent average coupon.

* First issuances: 308 bonds from 275 firms, valued at 33,976 million E U R, with a 74.4 percent fixed rate share and 5.1 percent average coupon.

* Annual trends: Issuance volume peaked in 2017 with 129 bonds (18,964 million E U R). Average coupon rates fluctuated between a high of 6.2 percent in 2018 and a low of 3.3 percent in 2017.

Panel B, Sample Used in the Empirical Analysis:

* Total issuances: 321 bonds from 173 firms, valued at 23,515 million E U R, with an 81 percent fixed rate share and 4.5 percent average coupon.

* First issuances: 187 bonds from 173 firms, valued at 14,060 million E U R, with an 80.2 percent fixed rate share and 4.8 percent average coupon.

* Annual trends (total): The highest number of bonds in the sample occurred in 2017 (80 bonds), while the highest value was in 2017 (5,361.7 million E U R). The average coupon rate in the sample reached its lowest point in 2017 at 2.3 percent.

We obtain firms’ annual balance sheet data, Z-scores (Altman (Reference Altman1968)), industry classification, and headquarters’ location from the Cerved database.Footnote ⁸ By merging the bond issuance data with Cerved, we have balance sheet information on 232 issuers for the 2012–2021 period.

We use quarterly data on bank loans, including both granted and used loan amounts and the interest rates charged by banks in individual firm–bank relationships. Data on bank loans come from the Bank of Italy’s Central Credit Register (Centrale dei Rischi, CR). The CR covers all loans above EUR 30,000 and reports lender and borrower identities, as well as loan type (credit lines, advances, long-term loans). These loan contracts differ in design, collateral, and conditions.Footnote ⁹ Credit lines allow the borrower to draw funds up to a contractually predetermined limit and generally lack real collateral. Advances provide short-term lending against commercial credit claims held by the borrower toward third parties; repayment occurs as intermediaries collect payments on the underlying claims, which serve as collateral. Long-term loans, such as mortgages, have contractually determined maturities and are typically collateralized (mostly real, related to buildings, and sometimes personal collateral). Due to these collateral features, advances and long-term loans are senior relative to corporate bonds.

Data on bank lending rates come from the Bank of Italy’s Interest Rate Database, which reports the interest rates charged by a sample of over 200 banks operating in Italy, along with lender and borrower identities and loan type. Short-term interest rates (credit lines and advances) refer to all outstanding positions at a given date, while interest rates on long-term loans refer to newly granted loans in each quarter.

By merging data on bond issuances, firm balance sheets, and credit register information on loan amounts and rates, we obtain complete records for 216 issuer firms (prior to matching). We exclude financial firms (typically holding companies of industrial groups), as well as firms without balance sheet data or without bank relationships, around the time of issuance.

B. Identification Strategy

The Italian minibond reform provides a notable deregulation experiment to study the effects of access to debt capital markets on corporate financing conditions and performance. This applies primarily to unlisted firms, the direct beneficiaries of the reform, and particularly to SMEs. We exploit this reform to analyze the impact of the first bond issuance for issuer firms relative to ex ante similar non-issuer firms.

Among all unlisted firms potentially eligible under the new regulatory framework, only a limited number actually issued and listed minibonds on a regulated market or a multilateral trading facility. Therefore, when defining the control sample, we cannot rely on the universe of eligible firms but must select a sample of ex ante comparable corporations.

We allow the selection of issuer versus non-issuer firms to be driven by firm-level characteristics rather than by a purely random assignment. Firms decide whether to issue minibonds based on their financial conditions and forward-looking growth prospects. Such self-selection could bias our impact evaluation if these firm features—affecting issuance probability—are correlated with the outcome variables. We address this concern by implementing an exact matching procedure based on ex ante firm characteristics, restricting the analysis to comparable groups of issuers and non-issuers in a way consistent with the unconfoundedness assumption. This methodology extends the empirical approach developed by Ioannidou and Ongena (Reference Ioannidou and Ongena2010) to study switching behavior across banks.Footnote ¹⁰

We identify the matched control sample by considering observable firm-level features—prior to the first issuance—that are relevant for bank lending conditions and volumes. These include balance sheet variables, proxies for investment opportunities and creditworthiness, the type of main bank, and industry and location dummies (see the next subsection for details). By controlling for these ex ante characteristics, and conditional on them, treatment and outcomes can be assumed independent, allowing us to attribute post-issuance changes in loan conditions to minibond issuances. The large and heterogeneous population of eligible firms allows us to fulfill the overlap assumption: for almost 80% of treated firms with complete data, we can identify units in both the treatment and control groups for any setting of covariates.

We define ex ante and ex post outcomes considering only the first minibond issuance: even if some firms undertake multiple issuances over time, the first one conveys the initial signal to banks that a given firm can also obtain finance from debt capital markets. The analysis compares the 2 quarters before issuance (ex ante) with the issuance quarter and the following 2 quarters (ex post).

C. Treated and Control Samples

The pool of potential control firms includes all Italian firms with turnover and total assets of at least EUR 2 million (as for issuer firms) at the beginning of the sample period, and reporting balance sheet information in the Amadeus-Bureau van Dijk database (around 35,000 firms). Within this sample, we define the control group by matching firms’ characteristics in the year before the first issuance.

We apply coarsened exact matching (CEM), which requires matching on all characteristics relevant for the control sample. By design, CEM removes the imbalance between treated and control firms for a given level of coarsening chosen for relevant observable firm characteristics. In contrast, propensity score matching (PSM) identifies matched entities based on the expected probability of treatment, computed over an overall set of characteristics, without guaranteeing exact matching on all criteria.

CEM has been shown to outperform alternative matching methods in reducing imbalance between treated and control groups (Iacus, King, and Porro (Reference Iacus, King and Porro2011)). This advantage makes it preferable for our study, as it helps address the potential endogeneity of firms’ issuance decisions. CEM is applied to both categorical and continuous variables; continuous variables are grouped into discrete intervals, determined either by economically relevant categories (e.g., firm size class) or by the distribution of covariates in the sample (e.g., investment measures or interest rates). For robustness, we also run PSM using the same covariates as in CEM.Footnote ¹¹ The results based on PSM confirm the findings obtained with CEM (see Section VII.B and Supplementary Material Appendix F).

Since the treatment (i.e., the first minibond issuance) occurs at different times across firms, we apply the CEM procedure separately for each year using firm characteristics observed in the year preceding the first issuance. Each control firm is then assigned the same event time as its matched treated firm.Footnote ¹²

Among the pre-treatment characteristics, we consider:

1. 1. At least one bank loan in the pre-treatment year and no bond issuance.

2. 2. Total assets (three classes): less than EUR 10 million; EUR 10–43 million; greater than EUR 43 million (following the EU classification of firms as small, medium, and large).

3. 3. Economic activity (five classes): i) agriculture and fishing; ii) manufacturing industries; iii) non-manufacturing industries (mining, electricity, gas); iv) construction; v) services.

4. 4. Firm location (three areas): north, center, and south.

5. 5. Firm credit risk (two classes): low risk (Z-score between 1 and 4); medium and high risk (Z-score between 5 and 9).

6. 6. Leverage ratio (two classes): up to 50%; above 50%.

7. 7. Ratio of total fixed assets to total assets (average over the previous 3 years, five classes).

8. 8. Investment rate (average growth rate of total fixed assets over the previous 3 years, 10 classes).

9. 9. Size of the main bank (two size classes): bank total assets above or below EUR 26 billion; the main lender bank is defined as the bank from which the firm receives the largest share of loans (out of total used bank loans) in the year before issuance.

10. 10. Average firm-level interest rate on bank loans (weighted average across all loan contracts in the year before issuance, eight classes).

The matching criteria, particularly 7 and 8, account for firms’ investment opportunities prior to treatment. Firms with higher growth trajectories are more likely to have higher financing needs—and hence issue bonds—and could benefit from more favorable bank lending conditions. To address potential self-selection based on investment opportunities, we include indicators of both the level and growth rate of investment in the pre-treatment period.Footnote ¹³

The matching variables also capture credit risk: in addition to balance sheet indicators (leverage, Z-score), we consider a market-based measure—the weighted average interest rate on existing bank loans at the firm level—which reflects banks’ ex ante pricing of borrower risk. Moreover, information barriers may play some role in firms’ financing decisions, by limiting the diffusion of information about new funding opportunities like minibonds. Given the limited contacts of unlisted firms with investment banks, their main relationship bank may influence their awareness of, or advice regarding, minibond issuance.Footnote ¹⁴ We therefore include the size of the main bank among the matching criteria.Footnote ¹⁵

We are able to match 173 treated firms (more than 80% of all treated firms with complete data) with 6,053 control firms (Table 2). On average, each treated firm has 34 control firms, though the distribution is positively skewed, with a median of 13 controls per treated firm (see Tables D2 and D3 for details).Footnote ¹⁶ For robustness, we also run propensity score matching using the same selection variables, with a varying number of matched control firms.

TABLE 2Characteristics of Minibond Issuers and Control Firms: Firms’ Composition and Balance Sheet Data

TABLE 2 Long description

The table contains four columns: Category, Treated Firms, Matched Control Firms, and All Firms. The number of observations is 173 for Treated, 6,053 for Matched Control, and 30,313 for All Firms.

* Economic Activity (percentage): Agriculture/fishing is 1.2, 1.2, 1.3. Manufacturing is 40.5, 40.5, 38.2. Non-manufacturing is 15.6, 15.6, 5.3. Construction is 5.2, 5.2, 6.0. Service industries are 37.6, 37.6, 49.0.

* Firm Size (total assets percentage): Small is 7.5, 7.5, 31.6. Medium is 30.1, 30.1, 48.0. Large is 62.4, 62.4, 20.5.

* Firm Location (percentage): North is 76.3, 76.3, 67.8. Centre is 16.2, 16.2, 18.0. South and isles is 7.5, 7.5, 14.2.

* Firm Z-Score (percentage): Low risk is 47.4, 47.4, 61.4. Medium risk is 33.5, 32.4, 27.1. High risk is 19.1, 20.2, 11.4.

* Firm Balance Sheet (mean values): Leverage is 59.19, 59.13, 47.4. Investment rate is 2.3, 2.25, 1.9. Fixed assets over assets is 31.36, 35.92, 25.1.

* Main Bank Size (percentage): Small-medium is 19.7, 19.7, 23.2. Large is 80.3, 80.3, 76.8.

By construction, treated and matched control firms display the same distribution across categorical matching variables (economic activity, location, and size class) and negligible differences across coarsened variables (leverage ratios, Z-score, investment rates). Issuer and matched non-issuer firms are concentrated in manufacturing and services, based in Northern Italy, and are mostly medium- or large-sized firms by total assets; they primarily rely on a large bank as the main lender (Table 2). Compared with the full population of firms (excluding treated firms), matching substantially reduces the imbalance between treated and matched control groups on all criteria. Among matched issuer firms, 80% issued fixed coupon bonds, with an average coupon rate of 4.8% for the first minibonds (Table 1, Panel B).

Table 3 presents descriptive statistics on interest rates, bank loans, and total financial debt. The analysis defines the ex ante period as the 2 quarters before issuance and the ex post period as the issuance quarter and the following 2 quarters. Before issuance, issuer firms display an average bank lending rate of 3.55% on advances, 4.25% on long-term loans, and 4.29% on all existing bank loans (including credit lines, advances, and long-term loans) (Table 3, Panel A, column 1).Footnote ¹⁷ Small differences in ex ante bank loan rates between treated and control firms before issuance are not statistically significant for advances, long-term loans, or total bank loans (column 3), confirming the comparability of the two groups after exact matching. The univariate DID test (column 4) further shows that, after the first minibond, issuer firms experience a larger decline in bank lending rates for advances, long-term loans, and total loans, with differences in post-issuance changes across the two groups significant at the 1% level.

TABLE 3Characteristics of Minibond Issuers and Control Firms: Interest Rates, Bank Loans and Financial Debt

TABLE 3 Long description

The table is divided into three sections.

Panel A. Interest Rates:

* Outstanding credit lines: Treated 6.262, Control 6.961, Difference -0.698, Final Diff-in-Diff 0.718.

* Outstanding advances: Treated 3.546, Control 3.362, Difference 0.184, Final Diff-in-Diff -0.223.

* New long-term loans: Treated 4.246, Control 4.073, Difference 0.172, Final Diff-in-Diff -0.291.

* Total bank loans: Treated 4.290, Control 4.063, Difference 0.227, Final Diff-in-Diff -0.226.

* Total financial debt: Treated 4.290, Control 4.063, Difference 0.227, Final Diff-in-Diff 0.170.

Panel B. Bank Loans Composition (as percent of total assets):

* Used loans/granted loans: Treated 70.625, Control 67.020, Difference 3.606, Final Diff-in-Diff -9.301.

* Composition of granted loans for Credit lines: Treated 3.621, Control 6.005, Difference -2.383, Final Diff-in-Diff 0.497.

* Composition of granted loans for Long-term loans: Treated 22.581, Control 21.376, Difference 1.205, Final Diff-in-Diff -1.916.

* Composition of used loans for Advances: Treated 8.250, Control 7.654, Difference 0.595, Final Diff-in-Diff -1.304.

* Composition of used loans for Long-term loans: Treated 18.473, Control 18.770, Difference -0.298, Final Diff-in-Diff -2.271.

Panel C. Financial Debt (as percent of total assets):

* Used loans (banks): Treated 28.535, Control 29.828, Difference -1.293, Final Diff-in-Diff -3.726.

* Minibonds: Treated 0, Control 0, Difference 0, Final Diff-in-Diff 34.625.

* Total financial debt: Treated 32.440, Control 33.620, Difference -1.180, Final Diff-in-Diff 29.883.

In terms of bank loan composition, long-term loans and advances account for the vast majority of used bank loans. After issuance, firms experience an easing of credit constraints: the ratio of used to granted loans decreases substantially for issuer firms relative to control firms (Table 3, Panel B). Issuer firms also reduce the amount of used bank loans while significantly increasing total financial debt through minibond issuances (Panel C).

Finally, we test the common trend assumption underlying the DID approach. We check graphically whether the outcomes of treated and matched control firms follow parallel trends in the absence of treatment. We find evidence supporting a common trend before the first issuance for lending rates on advances and long-term loans, as well as for the ratio of used to granted loans, though not for lending rates on credit lines.Footnote ¹⁸ After issuance, the reduction in interest rates is evident for advances and long-term loans and strengthens over time, while the impact on the ratio of used to granted loans occurs immediately and then stabilizes, suggesting partial repayment of outstanding loans.

D. Difference-in-Differences Specification and Hypothesis Testing

Based on exact matching between issuer firms and ex ante comparable non-issuer firms, we apply a DID model to estimate the impact of the first minibond issuance on bank loan rates and volumes. In the firm–bank level analysis, the main dependent variables are: i) quarterly bank loan rates, classified by loan type—credit lines, advances, long-term loans—and for overall loans; ii) quarterly amounts of granted and used bank loans, the ratio of used to granted loans for overall loans, and amounts of used loans by loan type.

The analysis of the first set of dependent variables investigates whether minibond issuances allow firms to obtain lower bank loan rates than non-issuer firms. We examine three potential economic channels: i) the release of public information by firms issuing bonds; ii) the change in the seniority structure of corporate debt, which reduces the risk of (senior) bank loans relative to (junior) corporate bonds; iii) an increase in the bargaining power of issuer firms in their lending relationships with banks. The empirical analysis explores the contribution of each channel in explaining the observed effects of minibond issuances on firms’ financing conditions (theoretical background and empirical results are presented in Sections IV.B and V.B).

The second set of dependent variables examines whether, and to what extent, minibond issuances affect the volume of loans granted by banks or the amount of loans used by firms. The impact on used bank loans depends on the purpose of the issuance: firms may issue minibonds to replace bank loans with market funding or to finance new investment projects requiring additional resources. The impact on granted bank loans may instead reflect supply decisions, which can be either bank- or firm specific: a bank may adjust the amount of loans granted due to changes in its general lending policy or for firm-specific reasons (e.g., the firm’s creditworthiness or profitability). Since generalized changes in granted amounts across all firms are captured by bank and time fixed effects, the effect of minibond issuance on loans granted to a firm reflects the bank’s assessment of firm-specific factors.

Our main specification exploits panel data for individual firm–bank relationships, with outcomes defined as interest rates and loan amounts for each quarter and each firm–bank pair. The treatment is defined as whether firm i has issued minibonds (Minibond = 1). The dummy variable “Minibond” is interacted with the dummy “Post,” which equals 1 after the first issuance:

(1) $$ {Y}_{i,j,t}={\alpha}_i+{\delta}_j+{\gamma}_t+{\beta}_1{\mathrm{Minibond}}_i\ast {\mathrm{Post}}_{i,t}+Z\;{\mathrm{Controls}}_{i,T-1}+{\varepsilon}_{i,j,\mathrm{t}} $$

where i indicates the firm, j the bank, t the quarter, and T the year. We include firm ( $ {\alpha}_i $ ), bank ( $ {\delta}_j $ ), and quarter ( $ {\gamma}_t $ ) fixed effects. The coefficient on the interaction between the first minibond issuance (Minibond) and the post-issuance dummy (Post) captures the average treatment effect of the first issuance on bank lending rates and volumes.Footnote ¹⁹

As control variables, we include annual firm-level characteristics: total assets (log), leverage, and Z-score, all measured with a 1-year lag. Regressions are weighted using CEM-strata weights, and standard errors are clustered at the firm level. The effect of minibond issuance is measured in the quarter of the first issuance and the following 2 quarters, relative to the 2 quarters before the event. As a robustness check, we also estimate the model using a longer sample—including 6 quarters after issuance—and the results are confirmed.Footnote ²⁰

A. Empirical Results: Post-Issuance Bank Lending Rates and Volumes

The main specification, as described in Section III.D, uses firm–bank level data on loan volumes and rates for each quarter, allowing us to control for firm, bank, and quarter fixed effects. Table 4, columns 1–4, presents the estimates of the impact on bank lending rates. The coefficient of interest ( $ {\beta}_1 $ ) is negative and significant for advances and long-term loans, as well as for overall firm–bank loans.Footnote ²¹ Lending rates decrease by 11 bps on advances, 15 bps on long-term loans, and 14 bps on overall loans; the effect on credit lines is not significant.

TABLE 4Firm–Bank Level Analysis: Effects of the First Minibond Issuance on Bank Lending Rates and Volumes

TABLE 4 Long description

The table is organized into three main categories across 10 columns. Columns 1 through 4 cover Bank Lending Rates (Credit Lines, Advances, Long-Term, Overall Loans). Columns 5 through 7 cover Bank Lending Volumes (Used Loans, Granted Loans, Used/Granted). Columns 8 through 10 cover Used Bank Loans by Loan Type (Credit Lines, Advances, Long-Term).

Key findings for the primary variable Minibond * post include:

* Bank Lending Rates: A non-significant increase in Credit Lines (0.054) and significant decreases in Advances (-0.114*), Long-Term (-0.146**), and Overall Loans (-0.136***).

* Bank Lending Volumes: Significant decreases across Used Loans (-0.635***), Granted Loans (-0.049*), and the Used/Granted ratio (-0.101***).

* Used Bank Loans by Type: Significant decreases in Credit Lines (-0.376***), Advances (-0.369***), and Long-Term loans (-0.528***).

Control variables include ln(total assets sub T minus 1), Leverage sub T minus 1, and Z-score sub T minus 1. Summary statistics at the bottom indicate the number of observations ranges from 78,774 to 236,222, R-squared values range from 0.359 to 0.538, and the number of issuers is consistently between 129 and 167. Standard errors are provided in brackets below the coefficients.

Table 4, columns 5–7, reports the estimated effects on granted and used bank loan amounts. After the first minibond, issuer firms reduce used bank loans at the firm–bank level by 64%,Footnote ²² while the decline in granted bank loans is much smaller, only 5%. This suggests that even when firms repay bank loans with bond proceeds, banks remain willing to extend loans, reducing granted amounts only marginally. As a result, the ratio of used to granted loans falls by 10 pp.

Some heterogeneity may exist across firms: while some reduce used bank loans to substitute across funding sources, others may use minibond proceeds to finance new investment opportunities. In particular, during the early years of our sample, the sovereign debt crisis strongly constrained bank lending, making bond finance an important alternative source of external funding (Becker and Ivashina (Reference Becker and Ivashina2018)).Footnote ²³ We further explore this point using firm-level data in Section V.A. The effects on loan rates and volumes are persistent over time and are confirmed in analyses over a longer ex post period of 6 quarters after issuance (Table E1 in the Supplementary Material).

B. The Decrease in Bank Loan Rates: The Mechanisms at the Firm–Bank Level

The decrease in bank loan rates following minibond issuances can be driven by several potential mechanisms, as discussed in Section III.D. In this section, we examine the hypothesis that the first bond issuance entails some public information disclosure: if so, outside banks could benefit from information previously accessible only to inside banks and might have an incentive to offer issuer firms better lending conditions than incumbents (Hale and Santos (Reference Hale and Santos2009), Schenone (Reference Schenone2010)). However, the strength of this information release effect likely depends on the extent of information actually disclosed through the issuance.Footnote ²⁴

We investigate this hypothesis by exploiting heterogeneity between inside and outside lenders. Firm–bank loan-level data allow us to compare the lending behavior of different banks toward the same firm, depending on whether they have (inside) or lack (outside) a pre-existing lending relationship. Using a DID analysis with heterogeneous treatment, we study whether outside banks apply a larger post-issuance reduction in loan rates than inside banks. While the post-issuance decrease in bank loan rates is confirmed for advances and long-term loans, we do not find a statistically significant difference between outside and inside banks (Table E2 in the Supplementary Material).

Consistent with the limited information disclosure associated with minibond issuances, this evidence suggests that information release does not play a major role in explaining the decline in bank loan rates after the first minibond. The minibond initiative was designed to incentivize bond issuance by unlisted firms by minimizing fixed issuance costs; since only professional investors are allowed to purchase minibonds, disclosure requirements are less stringent than those for corporate bonds listed on regulated markets. These features indicate a lower informational value of the issuance compared with a bond IPO (Hale and Santos (Reference Hale and Santos2009)). Thus, the first minibond issuance may have a relatively limited information release effect, while still signaling publicly that a firm can access capital markets.Footnote ²⁵ We further explore the economic channels behind the reduction in bank loan rates in Section V.B.

A. Bank Loans and Total Debt: Rates and Volumes

Based on the described matching procedure, we conduct a DID analysis using the following specification at the firm level:

(2) $$ \hskip0.24em {Y}_{i,t}={\alpha}_i+{\gamma}_t+{\beta}_1\;{\mathrm{Minibond}}_i\hskip0.42em \ast \hskip0.42em {\mathrm{Post}}_{i,t}+Z\;{\mathrm{Controls}}_{i,T-1}+{\varepsilon}_{i,t} $$

where i indicates the firm, t the quarter, and T the year. We control for firm $ \left({\alpha}_i\right) $ and quarter $ \left({\gamma}_t\right) $ fixed effects. Table 5 presents the firm-level results, confirming the effects observed at the firm–bank level. Columns 1–4 show the results for bank lending rates: after the first minibond, issuer firms experience a decrease of 13 bps on advances and 27 bps on long-term loans. The average interest rate on overall bank loans falls by 20 bps.

TABLE 5Firm-Level Analysis: Effects of the First Minibond Issuance on Bank Loan and Total Debt Rates and Volumes

TABLE 5 Long description

The table is organized into 10 columns categorized under four main headers: Bank Lending Rates (columns 1 to 4), Debt Rates (column 5), Bank Lending Volumes (columns 6 to 8), and Debt Volumes (columns 9 to 10).

* Column headers: 1. Credit Lines, 2. Advances, 3. Long-Term, 4. Overall Loans, 5. Total Financial Debt, 6. Used Loans, 7. Granted Loans, 8. Used/Granted, 9. Total Financial Debt, 10. Total Fin. Debt/Total Assets.

* Row 1: Minibond * post. Shows significant negative coefficients for Advances (-0.133), Long-Term (-0.272), Overall Loans (-0.202), Used Loans (-0.423), and Used/Granted (-0.100). It shows a significant positive coefficient for Total Financial Debt volume (0.621).

* Row 2: Standard errors for Minibond * post are provided in brackets below each coefficient, ranging from 0.018 to 0.187.

* Row 3: ln(total assets sub T minus 1). Shows significant positive coefficients for Used Loans (0.416), Granted Loans (0.258), Used/Granted (0.093), and Total Financial Debt volume (0.406).

* Row 4: Leverage sub T minus 1. Shows near-zero coefficients across all columns.

* Row 5: Z-score sub T minus 1. Shows a significant positive coefficient for Used/Granted (0.012).

* Summary Statistics:

- No. of obs. ranges from 22,017 to 30,803.

- R super 2 ranges from 0.463 to 0.939.

- No. of firms ranges from 4,956 to 6,190.

- No. of issuers ranges from 121 to 168.

We also examine interest payments on total financial debt (column 5). Despite the increase in financial debt due to minibonds, the average interest rate on total financial debt does not show a statistically significant change, with the coefficient pointing to a non-significant decrease of 6 bps. This suggests that the reduction in bank loan rates for issuer firms partially offsets the higher coupons on minibonds. It is worth noting that bank loans and minibonds differ in cash flow structure, seniority, and maturity; thus, costs based on par values may not be directly comparable.

Columns 6–8 report estimates for bank loan volumes at the firm level. Consistent with the firm–bank level analysis (Section IV.A), issuer firms reduce used bank loans by 42% after the first minibond, whereas granted bank loans show a smaller and non-significant decrease (around 11%). Consequently, the ratio of used to granted loans falls by approximately 10 pp.

We also analyze the overall debt position of issuer firms by including other financial debts and minibond issuances alongside used bank loans (Table 5, columns 9–10). Issuer firms increase total financial debt by 62%. The ratio of total financial debt to total assets displays a non-significant increase by 25 pp. The sizeable expansion in total financial debt—enabled by minibonds—occurs without a statistically significant change in its average interest rate; it is doubtful whether these firms could have reached the same result only relying on bank loans.

It is important to note that, while our analysis matches issuer and control firms before issuance, it remains an open question why some firms issued minibonds while similar firms did not. Firms may differ in investment opportunities, which could be recognized by lenders or investors. Our matching strategy addresses this concern by including indicators of pre-issuance investment opportunities for unlisted firms, capturing firms’ investment patterns prior to the minibond issuance.

B. The Decrease in Bank Loan Rates: The Mechanisms at the Firm Level

We examine two key potential channels at the firm level: i) changes in the debt seniority structure of issuer firms, and ii) increases in the bargaining power of issuer firms with their relationship banks. To study these mechanisms, we exploit heterogeneity across firms along two dimensions: the increase in total financial debt following bond issuance and the strength of pre-existing hold-up by banks.

1. The Seniority Structure of Corporate Debt

The issuance of bonds changes the seniority structure of corporate debt: as bank loans are senior to bonds, banks are repaid before bondholders (on the seniority structure of corporate debt, see Park (Reference Park2000), Hackbarth et al. (Reference Hackbarth, Hennessy and Leland2007), and Donaldson, Gromb, and Piacentino (Reference Donaldson, Gromb and Piacentino2025); for empirical studies, see Rauh and Sufi (Reference Rauh and Sufi2010), Cerqueiro, Ongena, and Roszbach (Reference Cerqueiro, Ongena and Roszbach2016, Reference Cerqueiro, Ongena and Roszbach2020)). In our setting, both long-term loans and advances are senior to bonds because they benefit from collateral: for long-term loans, mostly real collateral and occasionally personal; for advances, the credit claim from the underlying commercial transaction (see Section III.A and Table D1 on loan types).Footnote ²⁶ This seniority is also reflected in interest rates: despite the limited comparability of bonds and loans on several features, in our sample, at the time of issuance, the coupons on first minibonds are on average higher than the lending rates on pre-existing bank loans.

In structural models of corporate debt (Merton (Reference Merton1974), Leland (Reference Leland1994), and Hackbarth et al. (Reference Hackbarth, Hennessy and Leland2007) for corporate debt with non-zero coupon), firms that issue bonds in addition to bank loans raise total financial debt and, in turn, expand the total assets that can serve as collateral to banks as senior creditors. This larger collateral can reduce the risk on bank loans, and consequently, lower bank loan rates. Accordingly, issuer firms with a larger increase in asset value due to bond issuances would be expected to benefit from a larger ex post decrease in bank loan rates.

To test the debt seniority argument, we examine the change in total financial debt following minibond issuance. We construct this variable as a ratio, with the minibond issuance amount (net of changes in used bank loans) as the numerator and total existing used bank loans as the denominator.Footnote ²⁷ This captures the extent to which issuer firms increase total financial debt—and consequently total assets available as collateral—after the bond issuance. We split the sample at the median of this ratio (54.9% in our sample) and report the results in Table 6.

TABLE 6Firm-Level Analysis: Effects of the First Minibond: Firm Heterogeneity in Increase of Total Financial Debt

TABLE 6 Long description

The table is divided into four panels (A, B, C, and D) analyzing the effects of minibonds on bank lending rates, bank lending volumes, and total financial debt.

Columns are numbered 1 through 7:

1. Bank Lending Rates: Advances

2. Bank Lending Rates: Long-Term

3. Bank Lending Rates: Overall Loans

4. Bank Lending Volumes: Used Loans

5. Bank Lending Volumes: Granted Loans

6. Bank Lending Volumes: Used/Granted

7. Debt Volumes: Total Financial Debt

Panel A: Firms with [(Minibond minus Change in Used Loans) all over Total Existing Used Loans] greater than Median.

- Minibond * post for Total Financial Debt (Column 7) is 0.950***.

- Bank lending rates show significant decreases: -0.264** (Advances) and -0.314** (Overall Loans).

- Bank lending volumes show significant decreases: -0.842*** (Used Loans) and -0.282* (Granted Loans).

Panel B: Firms with [(Minibond minus Change in Used Loans) all over Total Existing Used Loans] less than or equal to Median.

- Minibond * post for Total Financial Debt is 0.233***.

- Bank lending rates show a significant decrease only in Long-Term rates (-0.237*).

- Bank lending volumes show a significant increase in Granted Loans (0.059**).

Panel C: Firms with (Minibond all over Total Existing Used Loans) greater than Median.

- Minibond * post for Total Financial Debt is 1.084***.

- Bank lending rates show significant decreases across all three categories (approx -0.224* to -0.297*).

- Bank lending volumes show significant decreases: -0.830*** (Used Loans) and -0.259* (Granted Loans).

Panel D: Firms with (Minibond all over Total Existing Used Loans) less than or equal to Median.

- Minibond * post for Total Financial Debt is 0.179***.

- Bank lending rates show significant decreases in Long-Term (-0.340**) and Overall Loans (-0.150*).

- Bank lending volumes show a decrease in Used Loans (-0.051*) but an increase in Granted Loans (0.032*).

Statistical significance is indicated by asterisks: * p < 0.1, ** p < 0.05, *** p < 0.01. Standard errors are provided in brackets below the coefficients.

In the subsample above the median (Panel A), issuer firms with a larger balance sheet expansion experience a stronger reduction in bank loan rates: 26 bps on advances, 28 bps on long-term loans (significant at the 10.1% level), and 31 bps on overall loans. These firms also reduce used bank loans by 84% and granted bank loans by 28%, easing credit constraints (the ratio of used to granted loans decreases by 17 pp). In the subsample below the median (Panel B), issuer firms with a smaller balance sheet increase show no significant change in bank loan rates on advances and overall loans, while observing a 24-bps decline on long-term loans. These firms do not reduce used bank loans, increase granted loans by 6%, and experience a modest decrease in the ratio of used to granted loans of 3 pp.

This evidence indicates that issuer firms with a larger increase in total financial debt due to minibonds experience a stronger reduction in bank loan rates, as the higher amount of collateral assets reduces more the credit risk for banks as senior creditors. Overall, this supports the debt seniority structure argument in explaining the ex post decrease in bank loan rates.

In a complementary analysis, we examine the potential role of credit substitution resulting from minibond issuances. Across the overall sample, issuer firms reduce used bank loans substantially (42% at the firm level). However, heterogeneity in minibond amounts relative to existing used bank loans may affect the extent of this substitution. We study whether the effects of minibonds vary in relation to credit substitution. To explore this, we compute the ratio of the minibond issuance amount to total existing used loans (averaged over the 4 quarters preceding the issuance) and split the sample of first-time issuers at the median of this ratio (43% in our sample), with results reported in Table 6.

In the subsample above the median (Panel C), firms with larger minibond issuances exhibit stronger credit substitution, with a reduction in used bank loans of 83% and a decrease in the ratio of used to granted loans by 15 pp, indicating a pronounced easing of credit constraints. These firms also experience a decline in bank loan rates of 22 bps on advances, 28 bps on long-term loans, and 30 bps on overall loans. In the subsample below the median (Panel D), firms with smaller minibond issuances display limited credit substitution, reducing used bank loans by only 5%. Nevertheless, they still observe decreases in long-term loan rates of 34 bps and overall loan rates of 15 bps. These findings suggest that credit substitution plays a limited role in explaining the post-issuance reduction in bank loan rates.Footnote ²⁸ The observed decline in lending rates can occur even when issuer firms do not reduce bank loans, provided that the increase in collateral value lowers the credit risk for lender banks as senior creditors.

2. The Bargaining Power of Firms with Lender Banks

We investigate whether access to bond markets strengthens the bargaining power of issuer firms with their relationship banks. This hypothesis is supported by the theoretical literature on hold-up by relationship banks due to informational advantages (Sharpe (Reference Sharpe1990), Rajan (Reference Rajan1992), and Santos and Winton (Reference Santos and Winton2008)). If firms are subject to hold-up, access to bond markets could relax funding constraints and increase their bargaining power with relationship banks, resulting in lower bank loan rates. Importantly, structural models like Hackbarth et al. (Reference Hackbarth, Hennessy and Leland2007) also predict that firms with high bargaining power issue both (senior) bank loans and (junior) bonds, whereas firms with little or no bargaining power rely solely on bank loans. In this context, a reform enabling access to capital markets could improve firms’ bargaining power with banks.

In our setting, minibond issuance may enhance bargaining power because it creates a credible threat that firms could partially replace loans from their relationship banks with minibonds. Banks might then have an incentive to retain creditworthy borrowers by offering more favorable loan conditions. To examine this channel, we exploit heterogeneity in the intensity of hold-up, proxied by the share of loans provided by the main bank relative to total used bank loans prior to the first issuance. The main bank supplies the largest share of a firm’s bank loans, and it is more difficult to replace its loans entirely with bonds; given its informational advantage, the main bank can exert the strongest hold-up power. Firms that are more reliant on their main bank could, after issuing minibonds, improve their bargaining position and secure better lending terms. Conversely, firms less dependent on their main bank are likely to benefit less from minibond issuance, as their bargaining power is already relatively strong and the hold-up effect on their loan spread is limited.

We compute the ratio of loans from the main bank to total existing used loans, averaged over the 4 quarters preceding the first issuance, and split the sample into two subsamples: below and above the median share of main bank loans (28.6% in our sample). The results are reported in Table 7 (Panels A and B). Issuer firms that were ex ante more reliant on the main bank (Panel A) experience a larger reduction in bank loan rates: 37 bps on long-term loans and 27 bps on overall loans. These firms reduce used bank loans by 34% and increase total financial debt by 89%. In contrast, firms less dependent on the main bank (Panel B) do not experience statistically significant changes in bank loan rates; they reduce used bank loans by 50% and raise total financial debt by 35%. These findings provide some evidence that firms more exposed to hold-up may benefit more from minibond issuance via larger reductions in bank loan rates. Notably, this effect emerges for more recent issuances, coinciding with a more developed minibond market.

TABLE 7Firm-Level Analysis: Effects of the First Minibond Issuance: Firm Heterogeneity in Reliance on Main Bank

TABLE 7 Long description

The table is divided into four panels (A, B, C, and D) across seven numbered columns. Columns 1 to 3 cover Bank Lending Rates (Advances, Long-Term, Overall Loans). Columns 4 to 6 cover Bank Lending Volumes (Used Loans, Granted Loans, Used/Granted ratio). Column 7 covers Debt Volumes (Total Financial Debt).

* Panel A: Firms with high reliance on main bank (above median). The interaction ‘Minibond * post’ shows significant negative coefficients for Long-Term rates (-0.371), Overall Loans rates (-0.273), Used Loans volume (-0.344), and Used/Granted ratio (-0.073). Total Financial Debt shows a significant positive coefficient of 0.887.

* Panel B: Firms with low reliance on main bank (below or equal to median). Significant negative effects are seen in Used Loans volume (-0.497) and Used/Granted ratio (-0.128). Total Financial Debt increases by 0.353.

* Panel C: High reliance firms with intensity of treatment. Significant negative effects for Long-Term rates (-0.202) and Used Loans volume (-0.246). Total Financial Debt shows a positive coefficient of 0.758.

* Panel D: Low reliance firms with intensity of treatment. Significant negative effects across all lending rates (Advances -0.154, Long-Term -0.232, Overall -0.199) and volumes (Used -0.851, Granted -0.467, Used/Granted -0.134). Total Financial Debt increases by 0.524.

Standard errors are provided in brackets below each coefficient. Statistical significance is indicated by asterisks and section symbols.

We also observe that firms ex ante more reliant on the main bank increase their total financial debt to a greater extent through minibonds. As discussed in Section V.B.1, a larger rise in total financial debt can increase total assets, and thus the collateral available to lender banks as senior creditors. Consequently, the greater reduction in bank loan rates for firms more dependent on their main bank may reflect a lower credit risk due to larger minibond issuances rather than an increase in bargaining power.

To disentangle these channels, we need to control for the size of minibond issuances. We replace the binary issuance variable with a continuous measure of relative issuance size, defined as the ratio of the minibond amount to total used bank loans, and estimate regressions for the same subsamples (Table 7, Panels C and D). For firms more reliant on their main bank (Panel C), a minibond issuance equal in size to existing used bank loans reduces long-term loan rates by 20 bps. For firms less dependent on their main bank (Panel D), an issuance of the same relative size reduces lending rates by 15 bps on advances, 23 bps on long-term loans, and 20 bps on overall bank loans. These results indicate that, after controlling for relative issuance size, there is only limited support for the bargaining power channel: ex ante more hold-up-exposed firms do not systematically experience larger reductions in bank loan rates. Overall, the evidence suggests that the debt seniority channel dominates the bargaining power mechanism in explaining the post-issuance decrease in bank loan rates for issuer firms.

C. Ex Post Outcomes: Firm Performance and Balance Sheet Indicators

The firm-level analysis suggests that minibond issuance improves firms’ financing conditions, enabling an expansion of overall financial debt while reducing reliance on bank loans. We now investigate whether these improved financing conditions translate into enhanced firm performance.

We estimate a DID model using firm-level data, where the dependent variables capture asset and liability composition, turnover, and profitability. On the asset side, we consider the log of total assets, fixed assets, and intangible fixed assets, as well as the ratio of total fixed assets to total assets. On the liability side, we analyze the ratio of bank loans to total financial debt and the leverage ratio. Additionally, we examine the log of turnover and return on assets (ROA) as a measure of profitability. Using annual financial statement data, we compare firm performance in the year before the first issuance and the two subsequent years,Footnote ²⁹ employing the following specification:

(3) $$ \hskip0.24em {Y}_{i,T}={\alpha}_i+{\alpha}_T+{\beta}_1\;{\mathrm{Minibond}}_i\hskip0.42em \ast \hskip0.42em {\mathrm{Post}}_{i,T}+{\varepsilon}_{i,T} $$

where i indicates the firm and T the year, and $ {\mathrm{Post}}_{i,T} $ is equal to 1 in the 2 years following the first issuance and 0 in the year before. We include firm $ \left({\alpha}_i\right) $ and year $ \left({\alpha}_T\right) $ fixed effects.

The results are reported in Table 8. Following the first minibond issuance, issuer firms experience a 21% increase in total assets (column 1), partly reflecting the rise in overall financial debt. These additional resources appear to be primarily allocated to investment: total fixed assets increase by 16% (column 2) and intangible fixed assets by 22% (column 3). While these increases could also reflect greater investment opportunities, we have designed our matching strategy to explicitly control for pre-issuance investment patterns in unlisted firms, including the ratio of fixed assets to total assets and the 3-year growth rate of fixed assets (Section III.C). Hence, we interpret these results as evidence that issuer firms utilize minibond proceeds to support internal growth and raise investments in fixed assets, consistent with theoretical and empirical findings on bond market access and firm investment (Crouzet (Reference Crouzet2018), Harford and Uysal (Reference Harford and Uysal2014)).Footnote ³⁰

TABLE 8Firm-Level Analysis: Effects of the First Minibond Issuance on Firm Ex Post Outcomes

TABLE 8 Long description

The table presents regression results for eight dependent variables across three categories.

1. Assets Category:

- Column 1: ln(Total Assets) shows a coefficient of 0.208 with a standard error of [0.030].

- Column 2: ln(Total Fixed Assets) shows 0.159 [0.062].

- Column 3: ln(Total Fixed Intangible Assets) shows 0.220 [0.112].

- Column 4: Total Fixed Assets / Total Assets shows -0.003 [0.009].

2. Liabilities Category:

- Column 5: Bank Loans / Total Financial Debt shows -0.219 [0.020].

- Column 6: Leverage shows 0.073 [0.014].

3. Turnover and Profitability Category:

- Column 7: ln(Turnover) shows 0.141 [0.040].

- Column 8: R o A shows -1.390 [0.519].

Summary Statistics:

- Number of observations ranges from 16,983 to 17,954.

- R-squared values range from 0.012 to 0.293.

- Number of firms is approximately 6,214.

- Number of issuers is 157 across all columns.

On the liability side, the increase in total financial debt is reflected in a 7-pp rise in leverage and a 22-pp reduction in the share of bank loans in total financial debt. We also find a significant 14% increase in turnover over the 2 years following the first minibond, while profitability shows a 1.39-pp ex post decrease in RoA.Footnote ³¹

A. The Intensity of Treatment

Beyond the issuance decision, the size of the issuance may influence the effects of minibonds. To capture this, we examine the impact of minibond issuances allowing for heterogeneous treatment intensity by replacing the binary variable for the first issuance with a continuous measure of the issuance amount, normalized by total existing used bank loans.

The results are presented in Table 11, at both the firm–bank level (Panel A) and the firm level (Panel B). At the firm level, a minibond issuance equal in size to existing used bank loans (i.e., a ratio of 1) reduces bank loan rates by 14 bps on advances and 19 bps on long-term loans, decreases used bank loans by 38%, and lowers the ratio of used to granted loans by 4 pp. It also increases total financial debt by 70%, while raising the average interest rate on total financial debt by 23 bps. This evidence suggests that larger minibond issuances relative to existing bank loans can substantially increase leverage, potentially raising the minibond coupon rate and the average interest rate on total financial debt.

TABLE 11Firm–Bank and Firm-Level Analysis: Effects of the First Minibond Issuance: Relative Issuance Amount as Treatment Variable

TABLE 11 Long description

The table is divided into two main sections.

Panel A: Firm-Bank Level Analysis.

It includes 10 columns. Columns 1 to 4 cover Bank Lending Rates (Credit Lines, Advances, Long-Term, Overall Loans). Columns 5 to 7 cover Bank Lending Volumes (Used Loans, Granted Loans, Used/Granted). Columns 8 to 10 cover Used Bank Loans by type (Credit Lines, Advances, Long-Term).

The treatment variable (Minibond/tot. exist. used loans) * post shows significant negative coefficients for Used Loans (-0.064), Used/Granted (-0.008), and Long-Term used loans (-0.067), all significant at the 1 percent level.

Panel B: Firm-Level Analysis.

It includes 10 columns. Columns 1 to 4 cover Bank Lending Rates. Column 5 covers Debt Rates (Total Financial Debt). Columns 6 to 8 cover Bank Lending Volumes. Columns 9 to 10 cover Debt Volumes (Total Financial Debt and Financial Debt/Tot Assets).

The treatment variable shows a significant negative effect on Used Loans (-0.380) and a significant positive effect on Total Financial Debt volume (0.697), both at the 1 percent level.

Both panels include rows for Number of observations, R-squared, Number of issuers, and Number of firms. The number of firms analyzed is approximately 6,190 across both panels.

At the firm–bank level, a comparable minibond issuance (equal to existing used loans) reduces used bank loans by 6% and the ratio of used to granted loans by 1 pp, and lowers rates on advances by 2 bps, but without statistically significant effects on long-term or overall loan rates.

For further robustness, we also consider the minibond amount in absolute terms (log of minibond issuances plus 1, to account for non-issuance). The results, reported in Tables E6 and E7 (Supplementary Material Appendix E), confirm that larger minibond issuances lead to greater reductions in bank loan rates, stronger decreases in used bank loans, and larger increases in total financial debt, both at the firm–bank and firm levels.

B. Propensity Score Matching

As an alternative matching strategy, we perform a propensity score matching (PSM) (see also Hale and Santos (Reference Hale and Santos2009)) to construct a sample of comparable control firms, with results reported in Supplementary Material Appendix F (Tables F1–F5). The propensity score is estimated based on the same firm characteristics used in the coarsened exact matching (CEM). The flexibility of PSM allows us to further refine the matching criteria, for example, using more granular classifications for firm location and industry. Control groups are identified via nearest neighbor or radius matching algorithms. After matching, we re-estimate the DID regressions on these alternative samples. The results based on PSM confirm the estimates obtained with CEM and remain robust to stricter matching criteria, different matching algorithms, and to the selection of a smaller number of matched control firms.

C. Further Robustness on Subsamples

We further test the robustness of our findings to heterogeneity in minibond characteristics and firm types. The firm-level analysis is reported in Table E8 in the Supplementary Material .

First, we examine bond maturity. A longer maturity profile of total debt, including long-term bonds, may signal lower firms’ risk to banks, potentially improving financing conditions; bonds with longer maturities, however, may also carry higher interest rates due to greater uncertainty. Results for minibonds with maturities of at least 5 years confirm the post-issuance reduction in long-term bank loan rates and the increase in total financial debt (Table E8, Panel A).

Second, we investigate heterogeneity by firm size, focusing on SMEs (defined by turnover below or equal to EUR 50 million), the primary target of the reform. SMEs exhibit an increase in total financial debt and a notable decrease in bank loan rates, comparable to the main findings. Importantly, the reduction in loan rates occurs even without credit substitution between minibonds and bank loans, since SMEs show no significant decrease in used bank loans, suggesting that SMEs mainly use minibonds to obtain additional funding for growth, in line with the purposes of the minibond reform (Table E8, Panel B).

Finally, we explore geographic heterogeneity, considering that the proximity to financial centers may foster information about market financing sources (Saunders and Steffen (Reference Saunders and Steffen2011)).Footnote ³² We consider a sample of firms located outside Lombardy (the region of Milan, the Italian financial center); our main results are confirmed for these firms as well (Table E8, Panel C).