A. Sample Collection

Our sample consists of the universe of actively managed domestic equity mutual funds in the U.S. covered in the CRSP Mutual Fund database from 1982 to 2016. We obtain mutual fund holdings from the Thomson Reuters Mutual Fund Holdings (s12) database and merge the two data sets using the Mutual Fund Links (MFLINKS) obtained from Wharton Research Data Services (WRDS). Stock return and accounting information are sourced from CRSP and Compustat. We obtain business and geographic segment data from the Compustat Historical Segment database. Data on bank loan contracts come from the LPC’s Dealscan database, while bond yield spreads and CDS spreads are obtained from the Bank of America-Merrill Lynch Corporate and High Yield Master Index Compositions database and the Markit CDS database. We collect data on mergers and acquisitions (M&As) from the Thomson Financial SDC Platinum M&A database.Footnote ¹⁴

B. Fund Leverage Constraints

To quantify the degree of leverage constraints among mutual funds, we use the BAB factor in Novy-Marx and Velikov (Reference Novy-Marx and Velikov2022), which is constructed as the value-weighted, market-neutral portfolio return from longing leveraged low-beta stocks and shorting high-beta stocks. For convenience, we use the negative BAB factor to construct our fund leverage constraint measure. When leverage-constrained investors are forced into high-beta stocks, the high-beta stock portfolio underperforms the leveraged low-beta stock portfolio. In such cases, the negative BAB factor generates a negative net-of-risk return (as per CAPM). Therefore, we estimate fund leverage constraints as the regression loadings of fund returns on the negative BAB factor (i.e., a higher loading on the negative BAB factor implies that the fund’s leverage constraint is more binding). This measure quantifies the extent to which a fund is constrained in taking on leverage by being compelled to overweight high-beta stocks to achieve its intended gross-of-risk performance.

Specifically, we perform rolling regressions at the fund level, detailed as follows: For each month t and each fund i, we estimate

(1) $$ {R}_{i,\tau }-{r}_{f,\tau }={\alpha}_{i,t}+{\beta}_{i,t}^{MKT}\left({R}_{Mkt,\tau }-{r}_{f,\tau}\right)+{\gamma}_{i,t}\left(- BA{B}_{\tau}\right)+{\varepsilon}_{i,\tau },\tau \in \left[\;t-23,t\right], $$

where R _i,τ is the monthly return of fund i in month $ \tau $ , $ {R}_{Mkt,\tau } $ and $ {r}_{f,\tau } $ are the monthly market return and the risk-free rate in month τ. We require each fund to have nonmissing returns for at least 12 months within the 24-month estimation period. We obtain the loading $ {\gamma}_{i,t} $ as our proxy for the tightness of fund leverage constraints.

Notably, the loading $ {\gamma}_{i,t} $ is not driven by the general risk-taking incentives of mutual funds. This is because the negative BAB factor is, by construction, market-neutral, and we also include the market factor in the rolling regressions. In later analyses, we always control for fund beta $ {\beta}_{i,t}^{MKT} $ to ensure that our results are not influenced by the general risk preferences of mutual funds.Footnote ¹⁵ We then aggregate fund leverage constraints at the stock level. Fich, Harford, and Tran (Reference Fich, Harford and Tran2015) show that both the fraction of a stock held by a fund and the fraction of the fund’s portfolio represented by that stock matter for the fund’s incentive to actively influence the firm. Therefore, we consider both an equally weighted measure (EW leverage constraint) and a value-weighted measure (VW leverage constraint), where the weight is the fund’s stake in the firm.

C. Variable Definitions

To examine the effects of aggregated fund leverage constraints on the cost of borrowing, corporate risk-taking, and firm capital structure, we consider two sets of variables.

The first set of variables pertains to the cost of borrowing, loan contract features, and firm capital structure. We consider two market-based measures of the cost of debt: the CDS spread and the bond yield spread. The CDS spread is the U.S. dollar-denominated spread of 5-year CDS contracts with the modified restructuring clause. For CDS contracts, we control for the recovery rate in the event of default. Yield spread is estimated as the issue amount-weighted bond yield spread aggregated at the issuer level. The bond yield spread is measured by the number of percentage points by which the fair value of the Treasury spot curve is shifted to match the present value of discounted cash flows to the bond’s price.Footnote ¹⁶ For bond characteristics, we control for bond maturity, defined as the issue amount-weighted bond maturity at the issuer-year level, and the issue amount outstanding, defined as the logarithm of the total bond face value aggregated at the issuer level.

In terms of the characteristics of bank loan contracts, the loan spread is the all-in-drawn spread that the borrower pays over LIBOR for the drawn portion of the loan. The collateral requirement is an indicator variable that equals 1 if the loan is secured by collateral, and 0 otherwise. The number of covenants is the total number of covenants included in a loan agreement. The maximum leverage slack is defined as the maximum leverage ratio set by the bank minus the most recent book leverage ratio. We control for loan maturity, defined as the maturity of a bank loan expressed in years, and loan size, defined as the logarithm of the loan amount outstanding. To measure a firm’s capital structure, we calculate book leverage as the ratio of long-term debt to total book assets.

The second set of variables are related to corporate risk-taking in business operations. We consider business risk-taking measures including operating cash flow volatility and operating leverage. For each firm-year, we calculate operating cash flow volatility as the standard deviation of quarterly operating cash flow in the next 3 years, including the current year. Operating cash flows are defined as the difference between income before extraordinary items and total accruals using the Compustat quarterly data (IB – ΔACT + ΔCHE + ΔLCT – ΔDLC + DP). Operating leverage arises from the proportion of fixed costs in a firm’s cost structure. We follow Aboody, Levi, and Weiss (Reference Aboody, Levi and Weiss2018) to estimate a time-series regression for each firm i and quarter t with a rolling window of next 12 quarters: OC _i,k = α + β_i,t REV _i,k + ε _i,k, k = t + 1, …, t + 12, where OC is the logarithm of operating costs (XOPRQ), and REV is the logarithm of sales revenue (REVTQ). If the costs are primarily fixed, then the estimated 1 – β _i,t is high, indicating a high ratio of fixed to variable costs, which results in a high operating leverage.

We also consider several risk-taking business activities, including the number of business and geographic segments, as well as the likelihood of pursuing industry- and geography-diversifying M&As. The number of business segments is the number of segments reported by a company as industry or product line breakouts. The number of geographic segments is the number of segments reported as country-by-country breakouts. An industry-diversifying merger is an indicator that equals 1 if the acquirer and target operate in different industries, as identified by their 2-digit SIC codes, and 0 otherwise. A geography-diversifying merger is an indicator that equals 1 if the acquirer and target are located in different states, and 0 otherwise. For M&A deals, we control for the relative size of the bidder and target, as well as deal value. Footnote ¹⁷

D. Summary Statistics

Table 1 presents summary statistics for the main variables. For each variable, we report the sample size, mean, median, standard deviation, 5th percentile, and 95th percentile. The complete sample comprises 103,347 firm-year observations with nonmissing information on aggregated fund leverage constraints and major firm characteristics. The average value-weighted (equally weighted) fund leverage constraint is 0.085 (0.076), with a standard deviation of 0.203 (0.187). The average fund ownership is 14.3% of total shares outstanding, and the average ownership-weighted fund beta is 1.065. Tangible assets and free cash flow account for 26.5% and −4.0% of total assets, respectively. The sample firms have an average book-to-market ratio of 0.709 and an average profitability (return on assets) of 0.069.

Table 1Summary Statistics

Table 1 Long description

The table contains 30 rows of data across 6 statistical columns. The columns are Sample Size, Mean, Median, Std. Dev., 5th Percentile, and 95th Percentile. Key variables include:

* Value-weighted leverage constraint: Sample 103,347; Mean 0.085; Median 0.098; Std. Dev. 0.203; 5th Percentile minus 0.287; 95th Percentile 0.395.

* Equal-weighted leverage constraint: Sample 103,347; Mean 0.076; Median 0.091; Std. Dev. 0.187; 5th Percentile minus 0.290; 95th Percentile 0.344.

* Ownership-weighted fund beta: Sample 103,347; Mean 1.065; Median 1.061; Std. Dev. 0.184; 5th Percentile 0.772; 95th Percentile 1.378.

* Book leverage: Sample 103,347; Mean 0.163; Median 0.113; Std. Dev. 0.172; 5th Percentile 0.000; 95th Percentile 0.499.

* Firm size: Sample 103,347; Mean 6.171; Median 6.007; Std. Dev. 2.173; 5th Percentile 2.961; 95th Percentile 9.998.

* Profitability: Sample 103,347; Mean 0.069; Median 0.103; Std. Dev. 0.232; 5th Percentile minus 0.283; 95th Percentile 0.274.

* Yield spread percentage: Sample 9,509; Mean 2.713; Median 1.942; Std. Dev. 2.230; 5th Percentile 0.655; 95th Percentile 7.200.

* C D S spread percentage: Sample 5,909; Mean 1.492; Median 0.872; Std. Dev. 1.692; 5th Percentile 0.212; 95th Percentile 5.047.

* Loan spread percentage: Sample 20,928; Mean 1.667; Median 1.500; Std. Dev. 1.241; 5th Percentile 0.250; 95th Percentile 3.850.

* Deal value: Sample 19,113; Mean 3.955; Median 3.784; Std. Dev. 1.769; 5th Percentile 1.335; 95th Percentile 7.148.

On average, the sample firms have 4.784 business segments and 4.220 geographic segments. Among the M&A deals, 52.5% are industry-diversifying mergers, and 68.0% are geography-diversifying mergers. The sample firms have an average bond yield spread of 2.713% and an average CDS spread of 1.492% for bond issuers. Among bank borrowers, the average loan spread is 1.667%, 43.2% of bank loan contracts have a collateral requirement, and the average number of covenants per contract is 2.202.

E. Validating the Measure of Fund Leverage Constraint

The theoretical link between mutual funds’ leverage constraints and their exposure to the negative BAB factor was introduced earlier. We now provide empirical validation of the leverage constraint measure at the fund level.

We first examine an explicit leverage constraint: a fund’s ability to engage in margin trading as part of its investment practices. We expect funds with margin purchase constraints in their investment policies to exhibit higher exposure to the negative BAB factor. We rely on the SEC requirement for registered investment companies to disclose whether their investment policies permit margin purchases. We obtain data from mutual funds’ N-SAR filings from 2006 to 2017, including both N-SAR A (semi-annual reports) and N-SAR B (annual reports), to identify constraints on margin purchases.Footnote ¹⁸ Specifically, we extract information from Item 70.G (“Margin purchases: permitted by investment policies? Y/N”) to determine whether a fund’s investment policy imposes margin purchase constraints. We then test whether funds with such constraints have higher exposure to the negative BAB factor. We report our findings in Table 2, specifications 1 and 2, where the dependent variable is the fund-level leverage constraint measure, and the independent variable is an indicator for the presence of margin purchase constraints. Given the mean fund leverage constraint of 0.061 at the fund-year level, we find that funds with margin purchase constraints in their investment policy are associated with a 24.0% higher exposure to the negative BAB factor.Footnote ¹⁹

Table 2Validation of Fund Leverage Constraints

Table 2 Long description

A table titled Validation of Fund Leverage Constraints. The dependent variable is Fund Leverage Constraint, divided into Evidence: Margin Purchase (Models 1 and 2) and Evidence: Bank Lending (Models 3 and 4).

* Margin purchase constraint indicator: Model 1 shows 0.016 with a t-stat of 2.74; Model 2 shows 0.015 with a t-stat of 2.48.

* New neighboring bank indicator: Model 3 shows minus 0.004 with a t-stat of minus 1.89; Model 4 shows minus 0.004 with a t-stat of minus 2.07.

* Controls (Model 2 and Model 4 respectively):

- Fund return: minus 0.213 (t-stat minus 1.55) and minus 0.646 (t-stat minus 7.01).

- Expense ratio: minus 4.144 (t-stat minus 3.23) and 0.641 (t-stat 0.71).

- Fund T N A: minus 0.014 (t-stat minus 6.01) and 0.010 (t-stat 5.46).

- Fund turnover: minus 0.007 (t-stat minus 1.82) and 0.001 (t-stat 0.26).

* Fixed Effects: All models include Year F E and Fund F E.

* Summary Statistics:

- Number of observations: 13,511 for Models 1 and 2; 33,864 for Models 3 and 4.

- R-squared: 0.624 (Model 1), 0.626 (Model 2), 0.515 (Model 3), and 0.518 (Model 4).

The second constraint we consider is access to bank loans, as the ability to borrow from banks is a typical leverage constraint. Fund regulations stipulate that mutual funds can only borrow externally from banks.Footnote ²⁰ A large body of banking literature has documented that proximity lending alleviates information asymmetry, such that the cost of borrowing increases and the availability of borrowing decreases with greater distance between the borrower and the lender.Footnote ²¹ It is reasonable to expect that a higher number of nearby banks increases access to bank credit for local funds. To examine this, we analyze changes in mutual fund behavior when a new bank begins lending in the local area. We define an indicator variable, new neighboring bank, which equals 1 when a new bank starts lending within 50 miles of the fund’s headquarters, and 0 otherwise. We identify a bank’s starting year using the first facility start date recorded in the LPC’s DealScan database. Table 2, specifications 3 and 4, presents the results. In specification 4, which includes all fund-level controls as well as fund and year fixed effects, we find that the event of a new neighboring bank starting to lend reduces the fund leverage constraint by 6.5% relative to the sample average.

Taken together, these results suggest that a fund’s exposure to the negative BAB factor serves as a valid empirical proxy for overall fund leverage constraints.

A. Credit Spread

We start by examining how fund leverage constraints are perceived by debtholders and test whether they exacerbate the cost of borrowing and the agency cost of debt. If leverage-constrained funds are associated with high risk-taking by firms, this should be particularly concerning for debtholders, who may impose a higher cost of borrowing.

To assess whether this is the case, we focus on two widely used market-based proxies for the cost of borrowing: CDS spreads and bond yield spreads. We report the results in Table 3. In Panel A, we examine CDS spreads, focusing on the most commonly observed CDS contracts (i.e., those with a 5-year maturity and a modified restructuring clause). We find a strong positive relationship between shareholder leverage constraints and CDS spreads, a result that holds across all specifications. Notably, this finding is robust to the inclusion of ownership-weighted fund beta as an additional control, demonstrating that the generic risk preferences of mutual funds do not drive the result. In the fully specified model, a 1-standard–deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 14.4% (21.1%) higher CDS spread relative to the sample average.

Table 3Shareholder Leverage Constraints and Credit Spread

Table 3 Long description

The table is divided into two main sections.

Panel A: C D S Spread.

This panel examines V W leverage constraint and E W leverage constraint. V W leverage constraint shows positive coefficients in models 1 through 3, ranging from 2.817 to 1.056, all significant at the 1 percent level. E W leverage constraint shows positive coefficients in models 4 through 6, ranging from 3.916 to 1.687, also significant at the 1 percent level. Controls include Weighted fund beta, Fund ownership, Firm size, Profitability, Free cash flow, Tangibility, Book-to-market, Book leverage, Amihud illiquidity, and C D S recovery rate. R-squared values for these models range from 0.554 to 0.793.

Panel B: Bond Yield Spread.

This panel mirrors the structure of Panel A. V W leverage constraint shows positive coefficients in models 1 through 3, ranging from 2.206 to 1.415. E W leverage constraint shows positive coefficients in models 4 through 6, ranging from 3.042 to 2.130. All leverage constraint coefficients are significant at the 1 percent level. Additional controls specific to this panel include Issue amount outstanding and Bond maturity. R-squared values range from 0.674 to 0.830.

Both panels include Year F E and Industry F E for models 1, 2, 4, and 5, while models 3 and 6 utilize Firm F E. The number of observations is 5,909 for Panel A and 9,509 for Panel B.

In Panel B of Table 3, the dependent variable is the issue amount-weighted bond yield spread, aggregated at the bond issuer level. We include total issue amount and issue amount-weighted bond maturity as additional controls. Again, we find a strong positive relationship between shareholder leverage constraints and bond yield spreads across specifications. In the fully specified model, a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 10.6% (15.9%) higher yield spread relative to the sample average.

B. Loan Contracts

The preceding results suggest that debtholders perceive the presence of leverage-constrained investors as increasing credit risk and aggravating concerns about risk-shifting and wealth transfer. Do lenders also design debt contracts in response to such heightened concerns about debtholder–shareholder conflicts? To address this question, we examine the features of bank loan contracts. We expect banks to respond by increasing loan spreads and imposing more stringent borrowing conditions, such as collateral requirements and covenant restrictions.

We begin by analyzing loan spreads and collateral requirements, with results reported in Panel A of Table 4. We estimate contract-level specifications with industry, credit rating, and year fixed effects, clustering standard errors at the borrower level. In specifications 1 and 2, we document a strong positive relationship between shareholder leverage constraints and loan spreads. In the fully specified model, a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 5.3% (5.8%) higher loan spread relative to the sample average.

Table 4Shareholder Leverage Constraints and Bank Loan Contracts

Table 4 Long description

The table is divided into two panels.

Panel A, Loan Spread and Collateral Requirement, contains four columns. Columns 1 and 2 use Loan Spread as the dependent variable, while columns 3 and 4 use Collateral Requirement.

- V W leverage constraint has a coefficient of 0.434 in column 1 and 0.734 in column 3, both significant at the 1 percent level.

- E W leverage constraint has a coefficient of 0.513 in column 2 and 0.828 in column 4, both significant at the 1 percent level.

- Control variables include Weighted fund beta, Fund ownership, Firm size, Profitability, Free cash flow, Tangibility, Book-to-market, Book leverage, Amihud illiquidity, Loan maturity, and Loan size.

- Fixed effects for Year, Industry, and Credit rating are included.

- The number of observations is 20,928 for all columns. R-squared is 0.421 for Loan Spread and 0.248 for Collateral Requirement.

Panel B, Number of Covenants and Covenant Slack, contains four columns. Columns 1 and 2 use Number of Covenants as the dependent variable, while columns 3 and 4 use Maximum Leverage Slack.

- V W leverage constraint has a coefficient of 0.353 in column 1 and minus 0.094 in column 3, both significant at the 1 percent level.

- E W leverage constraint has a coefficient of 0.411 in column 2 and minus 0.092 in column 4, significant at the 1 percent and 5 percent levels respectively.

- The same set of control variables and fixed effects from Panel A are applied.

- For Number of Covenants, the number of observations is 11,469 with an R-squared of approximately 0.207.

- For Maximum Leverage Slack, the number of observations is 2,116 with an R-squared of approximately 0.281.

In specifications 3 and 4, we conduct Probit regressions in which the dependent variable is an indicator equal to 1 if the loan facility is secured by collateral, and 0 otherwise. The results show that a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 10.1% (10.5%) higher probability of requiring collateral.

Next, we examine covenant restrictions. As a proxy for covenant tightness, we consider the number of financial covenants in a loan package. We also focus on covenants specifically designed to limit a borrower’s ability to raise additional debt (i.e., the maximum leverage covenant (Robert and Sufi (Reference Roberts and Sufi2009))). We calculate maximum leverage slack as the maximum leverage ratio set by the bank minus the most recent book leverage ratio. A higher covenant slack represents a lower degree of covenant strictness. We expect shareholder leverage constraints to be associated with stricter covenants and, consequently, lower covenant slack.

Panel B of Table 4 reports the results. A 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 3.3% (3.8%) increase in the number of financial covenants and a 6.0% (5.4%) decrease in maximum leverage slack relative to the sample average.

Overall, these results suggest that debtholders perceive the presence of leverage-constrained funds as potentially detrimental and respond by increasing the cost of borrowing and imposing tighter borrowing conditions on underlying firms.

A. Operating Cash Flow Volatility and Operating Leverage

We begin by analyzing the overall riskiness of firms’ business operations, measured by the volatility of operating cash flow. We use operating cash flow rather than operating income, as cash flow analysis is more critical in lenders’ credit-related decisions (Bhojraj and Swaminathan (Reference Bhojraj and Swaminathan2009)). In addition, we consider operating leverage as a proxy for the flexibility of a firm’s cost structure. A less elastic cost structure (i.e., one characterized by a higher proportion of fixed costs and a lower proportion of variable costs) amplifies a firm’s exposure to business cycle fluctuations. Accordingly, operating leverage is considered a real-asset risk-taking component that drives the systematic risk of common stocks (Gahlon and Gentry (Reference Gahlon and Gentry1982), Mandelker and Rhee (Reference Mandelker and Rhee1984), and Aboody, Levi, and Weiss (Reference Aboody, Levi and Weiss2018)).

Table 5 presents the results of the regression analysis for corporate risk-taking. In specifications 1 and 2, we use operating cash flow volatility as a proxy for corporate risk-taking. In specifications 3 and 4, we use operating leverage as an alternative proxy. We find that shareholder leverage constraints are positively related to firms’ degree of risk-taking. In the fully specified model, which includes all control variables as well as firm and year fixed effects, a 1-standard–deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 5.6% (4.0%) increase in operating cash flow volatility and a 5.1% (3.6%) increase in operating leverage relative to the sample average. These results support the hypothesized risk-taking channel, suggesting that leverage-constrained funds increase the cost of debt due to higher corporate risk-taking in business activities.

Table 5Channel Test: Shareholder Leverage Constraints and Corporate Risk-taking

Table 5 Long description

The table presents regression results for two dependent variables: Operating Cash Flow Volatility (Models 1 and 2) and Operating Leverage (Models 3 and 4).

Key Independent Variables:

* V W leverage constraint: Model 1 shows a coefficient of 0.014 (t-stat 2.78); Model 3 shows 0.050 (t-stat 3.98).

* E W leverage constraint: Model 2 shows a coefficient of 0.011 (t-stat 2.78); Model 4 shows 0.038 (t-stat 2.44).

Control Variables (consistent across models):

* Weighted fund beta: 0.005 in Models 1-2; 0.008 to 0.011 in Models 3-4.

* Fund ownership: -0.003 in Models 1-2; -0.079 to -0.078 in Models 3-4.

* Firm size: -0.009 in Models 1-2; -0.022 in Models 3-4.

* Profitability: 0.054 in Models 1-2; -0.225 in Models 3-4.

* Free cash flow: -0.081 in Models 1-2; -0.004 in Models 3-4.

* Tangibility: -0.041 in Models 1-2; 0.071 to 0.070 in Models 3-4.

* Book-to-market: 0.012 in Models 1-2; 0.027 in Models 3-4.

* Book leverage: -0.006 in Models 1-2; -0.052 in Models 3-4.

* Amihud illiquidity: 0.001 in Models 1-2; -0.016 in Models 3-4.

Fixed Effects and Statistics:

* Year F E and Firm F E are included (Yes) for all models.

* Number of observations: 87,481 for Models 1-2; 88,640 for Models 3-4.

* R super 2: 0.710 for Models 1-2; 0.382 for Models 3-4.

B. Business Activities

Next, we identify the specific activities through which firms may increase the riskiness of their operating cash flow. We consider three types of corporate actions: product diversification, geographic diversification, and engagement in diversification enhancing M&As. The underlying intuition is that firms can reduce the riskiness of their operating cash flow by achieving greater internal diversification across business and geographic segments, as well as by engaging in diversifying M&As. We therefore examine the relationship between shareholder leverage constraints and these policies.

We report the results in Table 6. Panel A focuses on the degree of internal diversification in business and geographic segments, while Panel B examines firms’ engagement in M&As. In Panel A, the dependent variable in specifications 1 and 2 is the number of business segments, reported as industry or product line segment breakouts. The dependent variable in specifications 3 and 4 is the number of geographic segments, reported as country-by-country segment breakouts. We find a robust negative relationship between shareholder leverage constraints and various measures of internal diversification. In the fully specified model, a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 5.8% (7.7%) decrease in product diversification and a 2.7% (3.8%) decrease in geographic diversification.

Table 6Channel Test: Shareholder Leverage Constraints and Business Activities

Table 6 Long description

Panel A: Number of Business Segments and Geographic Segments.

* V W leverage constraint: -1.359 (t-stat -4.72) for business segments and -0.565 (t-stat -3.29) for geographic segments.

* E W leverage constraint: -1.826 (t-stat -4.64) for business segments and -0.852 (t-stat -3.50) for geographic segments.

* Controls: Weighted fund beta, Fund ownership, Firm size, Profitability, Free cash flow, Tangibility, Book-to-market, Book leverage, and Amihud illiquidity all show statistically significant coefficients across most models.

* Model Stats: Year and Firm Fixed Effects (F E) are included. Number of observations is 101,077. R-squared is 0.761 for business segments and 0.746 for geographic segments.

Panel B: Likelihood of Industry-Diversifying Merger and Geography-Diversifying Merger.

* V W leverage constraint: -0.465 (t-stat -4.33) for industry and -0.358 (t-stat -3.36) for geography.

* E W leverage constraint: -0.567 (t-stat -4.31) for industry and -0.456 (t-stat -3.55) for geography.

* Controls: Includes the same variables as Panel A plus Relative size and Deal value. Firm size, Tangibility, and Deal value are highly significant.

* Model Stats: Year and Industry Fixed Effects (F E) are included. Number of observations is 19,113. R-squared is 0.094 for industry and 0.065 for geography.

Next, in Panel B, we analyze firms’ decisions to engage in diversification enhancing mergers, considering both industry-diversified and geography-diversified mergers. In specifications 1 and 2, the dependent variable is an indicator equal to 1 if the acquirer and target operate in different industries (as identified by the 2-digit SIC code), and 0 otherwise. In specifications 3 and 4, the dependent variable is an indicator equal to 1 if the acquirer and target are located in different, nonadjacent states, and 0 otherwise. We find a strong negative relationship between shareholder leverage constraints and the likelihood of engaging in diversifying M&As. In the fully specified model, a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 6.5% (7.2%) lower probability of industry-diversifying M&As and a 2.8% (3.6%) lower probability of geography-diversifying M&As. Taken together, these results support the risk-taking channel, indicating that firms held by funds with more binding leverage constraints engage in riskier business activities.

C. Financial Leverage

A natural question is whether firms also take on additional financial risk by increasing leverage. While increasing leverage may be appealing from a risk-taking perspective, it is directly constrained by rising borrowing costs and stricter borrowing conditions, such as tighter maximum leverage covenants. Moreover, firms may seek to balance operating and financial leverage. As noted by Chen, Harford, and Kamara (Reference Chen, Harford and Kamara2019), higher operating leverage can reduce a firm’s ability to meet debt obligations, prompting firms to adopt lower financial leverage. To investigate this issue, we analyze firms’ capital structure decisions.

We report the results in Table 7, where the dependent variable is book leverage. We find a robust negative relationship between shareholder leverage constraints and book leverage. Specifically, in the fully specified model, a 1-standard-deviation increase in value-weighted (equally weighted) fund leverage constraints is associated with a 3.6% (4.0%) decrease in book leverage relative to the sample average. This suggests that, although firms might be incentivized to increase risk through higher financial leverage, this tendency is offset by higher agency costs of debt and the trade-off between operating and financial leverage.

Table 7Shareholder Leverage Constraints and Capital Structure

Table 7 Long description

A table with four numbered columns representing different regression models.

Key variables and their coefficients (with t-statistics in brackets):

* V W leverage constraint: -0.029 (t = -9.47 in model 1, -9.37 in model 2), significant at the 1 percent level.

* E W leverage constraint: -0.035 (t = -9.72 in model 3, -9.64 in model 4), significant at the 1 percent level.

Controls:

* Weighted fund beta: -0.001 (model 2) and -0.002 (model 4), both statistically insignificant.

* Fund ownership: -0.028 across all models, significant at the 1 percent level.

* Firm size: 0.037 (models 1-3) and 0.036 (model 4), significant at the 1 percent level.

* Profitability: -0.033 across all models, significant at the 1 percent level.

* Free cash flow: -0.002 across all models, significant at the 10 percent level.

* Tangibility: 0.088 (model 1) and 0.087 (models 2-4), significant at the 1 percent level.

* Book-to-market: -0.007 across all models, significant at the 1 percent level.

* Amihud illiquidity: 0.011 across all models, significant at the 1 percent level.

Model Statistics:

* Year F E and Firm F E: Included (Yes) for all models.

* Number of observations: 103,347 for all models.

* R-squared: 0.738 for all models.

D. Moderation Tests

Moreover, if shareholder leverage constraints are associated with heightened concerns about risk-shifting and wealth transfer impacting debtholders, we would expect our findings to be more pronounced in firms facing significant conflicts between debtholders and shareholders. Prior research indicates that the benefits for shareholders from wealth transfers are especially notable in financially distressed firms (Eisdorfer (Reference Eisdorfer2008), Gilje (Reference Gilje2016)). Conversely, the shareholder–debtholder conflict tends to be mitigated in companies where shareholders also have considerable debt holdings (Jiang, Li, and Shao (Reference Jiang, Li and Shao2010), Chen, Zhang, and Zhu (Reference Chen, Zhang and Zhu2023)). Consequently, we perform moderation tests to explore how shareholder leverage constraints interact with the level of financial distress and the dual holdings of mutual fund families.

We first examine the moderation effects of dual holdings of mutual fund families. For the specific features of loan contracts, such as loan spread, we follow Jiang, Li, and Shao (Reference Jiang, Li and Shao2010) and construct a loan-based dual-holder indicator variable. Specifically, we match lender names from the LPC’s DealScan database with those of mutual fund management companies in the Thomson Reuters Mutual Fund Holdings database. We manually verify the matching results to ensure their accuracy. The dual-holder indicator is equal to 1 if one of the lenders in the syndicated loan package is part of the same financial institution as the mutual fund families that possess the largest mutual fund stake in the firm, and 0 otherwise. For bond-specific features, such as the bond yield spread, we adopt the approach of Chen, Zhang, and Zhu (Reference Chen, Zhang and Zhu2023) to define a dual-holder indicator based on bond holdings of institutional investors. Specifically, we gather data on institutional bond holdings from the Thomson Reuters eMAXX (formerly Lipper eMAXX) database and match bondholder names with those of mutual fund management companies in the Thomson Reuters Mutual Fund Holdings database. We manually verify these matches to ensure their accuracy. The dual-holder indicator is set to 1 if the same financial institution owns at least 1% of the firm’s outstanding shares and at least 1% of one of its outstanding bonds. For other firm-level dependent variables, such as operating leverage, we calculate a weighted average of the dual-holder indicator, using the loan amount and the face value of outstanding bonds as weights.

We present the results in Panel A of Table 8. The dependent variable in specifications 1–9 includes the CDS spread, bond yield spread, loan spread, collateral requirement, number of covenants, maximum leverage slack, operating cash flow volatility, operating leverage, and book leverage, respectively. Our focus is on the interaction term between shareholder leverage constraints and the dual-holder indicator. As expected, we find that the effects of shareholder leverage constraints are generally mitigated when the severity of debtholder–shareholder conflict is weaker, particularly when fund families hold significant debt stakes in the company.

Table 8Moderation Tests

Table 8 Long description

The table consists of nine columns representing financial metrics: 1 C D S Spread, 2 Yield Spread, 3 Loan Spread, 4 Collateral Requirement, 5 Number of Covenants, 6 Max. Lev. Slack, 7 Operating Volatility, 8 Operating Leverage, and 9 Book Leverage.

Panel A: Interaction with Dual-Holding Indicator

* V W leverage constraint: Shows positive significant coefficients for columns 1 through 5 (ranging from 0.373 to 2.186) and column 7 (0.095), and a negative significant coefficient for column 9 (-0.064).

* V W leverage constraint times Dual-Holding Indicator: Shows negative significant coefficients for columns 1, 2, 3, 5, and 7 (ranging from -0.188 to -1.153) and a positive significant coefficient for column 9 (0.031).

* Dual-Holding Indicator: Shows a negative significant coefficient for column 3 (-0.044) and column 6 (-0.037), and a positive significant coefficient for column 9 (0.013).

Panel B: Interaction with Expected Default Frequency (E D F)

* V W leverage constraint: Shows positive significant coefficients for columns 1 through 5 and column 8 (ranging from 0.464 to 1.581) and a negative significant coefficient for column 9 (-0.012).

* V W leverage constraint times E D F: Shows positive significant coefficients for columns 1, 2, 3, 7, and 8 (ranging from 0.034 to 61.877) and negative significant coefficients for column 6 (-0.536) and column 9 (-0.034).

* E D F: Shows positive significant coefficients for columns 1 through 4 and column 9 (ranging from 0.062 to 3.109).

Both panels include ‘Other controls’ (Yes) and specific references to previous tables for each column's specification.

Next, we examine the moderation effects of financial distress. We expect that the effects of shareholder leverage constraints should be stronger for more financially distressed companies. This is because shareholders of levered firms have incentives to take on higher levels of operating risk as they approach distress, reaping the upside rewards in good states of the world and transferring losses to debtholders in bad states (Jensen and Meckling (Reference Jensen and Meckling1976), Eisdorfer (Reference Eisdorfer2008), and Gilje (Reference Gilje2016)). To capture the degree of financial distress, we use Bharath and Shumway (Reference Bharath and Shumway2008) measure of expected default frequency (EDF), which is estimated based on a simplified version of Merton (Reference Merton1974) structural distance-to-default model. Firms with a higher EDF are considered more financially distressed than those with a lower EDF.

We present the results in Panel B of Table 8. The dependent variable in specifications 1–9 includes the CDS spread, bond yield spread, loan spread, collateral requirement, number of covenants, maximum leverage slack, operating cash flow volatility, operating leverage, and book leverage, respectively. Our focus is on the interaction term between shareholder leverage constraints and the EDF measure. Consistent with our expectation, the results show that the effects of shareholder leverage constraints are more pronounced for more financially distressed companies.

E. Robustness Checks

Next, we investigate the robustness of our results using alternative measures of fund leverage constraints. In the preceding analyses, we identified fund leverage constraints based on rolling regressions of fund returns on the market factor and the negative BAB factor. To fully rule out the concern that fund leverage constraints may be mechanically correlated with the generic risk preferences of mutual funds, we construct a residual leverage constraint measure that is orthogonal to fund beta. Specifically, for each year quarter at the fund level, we regress fund leverage constraints (i.e., fund loading on the negative BAB factor) on fund beta (i.e., fund loading on the market factor) and use the regression residual as our measure of leverage constraints. We then compute ownership-weighted residual fund leverage constraints at the stock level and re-estimate our tests using this alternative measure.

In addition, we consider a multifactor model, incorporating the Fama–French–Carhart 4-factor model, which include market (MKT), value (HML), size (SMB), and momentum (UMD) factors. We estimate:

(2) $$ {\displaystyle \begin{array}{l}{R}_{i\tau}={\alpha}_{i t}+{\beta^{MKT}}_{i t}{R}_{MKT,\tau }+{\beta^{HML}}_{i t}{R}_{HML,\tau }+{\beta^{SMB}}_{i t}{R}_{SMB,\tau}\\ {}\hskip3.5em +\hskip2px {\beta^{UMD}}_{i t}{R}_{UMD,\tau }+\left(-{BAB}_{\tau}\right)+{\varepsilon}_{i\tau}.\\ {}\hskip3.5em \tau \left[t-23,t\right].\end{array}} $$

We then use $ {\gamma}_{i,t} $ as leverage constraint at the fund level and calculate the ownership-weighted leverage constraint at the stock level. For the Fama–French 4-factor model, the factors are market (MKT), value (HML), size (SMB), profitability (RMW), and investment (CMA). We estimate:

(3) $$ {\displaystyle \begin{array}{l}{R}_{i\tau}={\alpha}_{i t}+{\beta^{MKT}}_{i t}{R}_{MKT,\tau }+{\beta^{HML}}_{i t}{R}_{HML,\tau }+{\beta^{SMB}}_{i t}{R}_{SMB,\tau}\\ {}\hskip3.5em +\hskip2px {{\beta^{RMW}}_i}_{,t}{R}_{RMW,\tau }+{{\beta^{CMA}}_i}_{,t}{R}_{CMA\tau}+\left(-{BAB}_{\tau}\right)+{\varepsilon}_{i\tau},\\ {}\hskip3.5em \tau \left[t-23,t\right].\end{array}} $$

We then use $ {\gamma}_{i,t} $ as leverage constraint at the fund level and calculate the ownership-weighted leverage constraint at the stock level.

The results based on these alternative measures are consistent with our main findings, alleviating concerns that when leverage constraints compel funds to take on more systematic risk in their portfolios, they may load more heavily on other dimensions of priced risk rather than market risk.Footnote ²²

A. Degree of Fund Influence

First, we examine whether the impact is stronger when funds exert greater influence on the company. We consider several measures of fund influence: geographical distance, investment horizon, fund ownership, and ownership concentration. Geographically closer shareholders are more likely to communicate directly with management and have their views heard (Ayers, Ramalingegowda, and Yeung (Reference Ayers, Ramalingegowda and Yeung2011)). Moreover, proximity facilitates information collection and reduces the travel costs associated with attending meetings or serving on the board (Chhaochharia, Kumar, and Niessen-Ruenzi (Reference Chhaochharia, Kumar and Niessen-Ruenzi2012)). In addition, the literature documents that long-term shareholders are more likely to influence management through direct engagement (Chen, Harford, and Li (Reference Chen, Harford and Li2007)), whereas short-term investors prioritize information acquisition for trading (Yan and Zhang (Reference Yan and Zhang2009)) and have lower incentives to exert influence on corporate policies. Furthermore, survey evidence suggests that large institutional investors frequently influence management through private, behind-the-scenes engagement (McCahery, Sautner, and Starks (Reference McCahery, Sautner and Starks2016)). For each stock, we reconstruct shareholder leverage constraints based on the largest fund, large funds (top five funds), and small funds (bottom five funds). The larger a fund’s stake, the stronger its incentives to monitor the firm and influence management. We expect our previously documented results to be driven by funds that are geographically closer, have long-term investment horizons, hold larger stakes, and exhibit higher ownership concentration.Footnote ²³

We report the results in Table 9. Panels A–D present regressions in which the ownership-weighted fund leverage constraints are interacted with weighted fund distance, weighted fund turnover, fund ownership, and ownership concentration, respectively. As before, we repeat all prior analyses using the fully specified models and, for brevity, only report the coefficients of the key variables. As expected, we find that the previously documented results are primarily driven by the leverage constraints of funds with long-term horizons, located nearby, holding larger stakes, and exhibiting higher ownership concentration.

Table 9Interaction with Fund Influence

Table 9 Long description

A complex regression table with nine numbered columns representing different dependent variables: 1 C D S Spread, 2 Yield Spread, 3 Loan Spread, 4 Collateral Requirement, 5 Number of Covenants, 6 Max. Lev. Slack, 7 Operating Volatility, 8 Operating Leverage, and 9 Book Leverage.

* Panel A: Interaction with Weighted Fund Distance. Shows coefficients for V W leverage constraint, its interaction with weighted fund distance, and the distance itself. Significant positive effects are seen in columns 1 through 4 for the constraint, while the interaction term is generally negative and significant.

* Panel B: Interaction with Weighted Fund Turnover. Features a strong negative interaction between V W leverage constraint and weighted fund turnover, particularly in columns 1 and 2 with coefficients of minus 10.485 and minus 12.642.

* Panel C: Interaction with Fund Ownership. Shows positive interaction coefficients for columns 1, 3, 4, 5, 7, and 8, with a significant negative coefficient for book leverage in column 9.

* Panel D: Interaction with Ownership Concentration. Similar to Panel C, showing positive interaction effects across most spread and covenant metrics, and a significant negative interaction for Max. Lev. Slack at minus 0.157.

* Panel E: Analysis Using Mutual Fund with the Largest Holding. Focuses on the V W leverage constraint for the top 1 fund, showing highly significant positive coefficients for spreads and negative for book leverage.

* Panel F: Analysis Using Top Five Mutual Funds. Shows consistent significant results for the top 5 funds across all columns except column 6.

* Panel G: Analysis Using Bottom Five Mutual Funds. Shows mostly statistically insignificant results across all nine columns.

All panels include rows for Other controls (Yes) and Specification details referencing other tables in the study.

We also reaggregate fund leverage constraints based solely on the largest fund (Panel E), large funds (top five funds, Panel F), and small funds (bottom five funds, Panel G). In most specifications of Panels E and F, the coefficients of fund leverage constraints are larger in absolute terms, suggesting that leverage constraints become more influential when funds hold larger ownership stakes in the company. In contrast, in Panel G, where we examine funds with the bottom five ownership stakes, the insignificance of the fund leverage constraint coefficients across all regressions suggests that funds with minimal influence do not explain our findings.

B. Merger of Asset Management Companies

Second, we explore a dynamic specification that leverages time-series variation in shareholder leverage constraints resulting from mergers of asset management companies. Portfolio repositioning often occurs following a merger between two institutions, typically with the acquired fund adjusting its investment policies or strategies to align with those of the acquiring fund. This process leads to time-series changes in fund leverage constraints for stocks primarily held by the acquired funds. As fund mergers are largely driven by deregulation, industry consolidation, or mutual funds’ pursuit of economies of scale (Jayaraman, Khorana, and Nelling (Reference Jayaraman, Khorana and Nelling2002), He and Huang (Reference He and Huang2017)), these time-series changes in fund leverage constraints are relatively exogenous to other unobservable firm-specific characteristics. This allows us to analyze firms’ responses in terms of corporate risk-taking and borrowing costs.

We obtain data on financial company mergers from the SDC Mergers and Acquisitions database during our sample period. We manually match the names of acquirers and targets with the management company names from the CRSP Mutual Fund database to ensure that the mergers were completed. The initial sample consists of 347 investment company mergers. We then integrate this sample with the Thomson Reuters Mutual Fund Database to obtain information on mutual fund holdings.

Our analysis focuses on fund family mergers involving the largest mutual fund investors in our sample firms, as these influential investors are most likely to significantly impact shareholder leverage constraints before and after the merger. This approach yields 5692 firm-year observations in which the largest mutual fund investors experienced merger events among asset management institutions. For each firm year, we define the event year as year t, when the firm’s largest mutual fund investor is acquired by another asset management company. We then calculate the change in leverage constraint Δ Fund leverage constraint for the firm’s largest mutual fund holder between year t + 1 and year t. Footnote ²⁴ Next, we regress the changes in our primary focus variables between year t + 1 and year t—specifically, changes in CDS spread, bond yield spread, loan spread, likelihood of collateral requirements, covenant slack, number of covenants, operating volatility, operating leverage, and book leverage—on the change in fund leverage constraints. For nonpanel variables such as loan spread, collateral requirements, maximum leverage covenant slack, and the number of covenants, we compute yearly averages and then take the differences.Footnote ²⁵

We present the results in Table 10. Our analysis reveals a positive relationship between changes in fund leverage constraints and our key dependent variables: the cost of borrowing, the strictness of loan contracts, and corporate risk-taking. In addition, we continue to observe a negative relationship between changes in fund leverage constraints and changes in book leverage. Most regression analyses indicate that these effects are statistically significant at least at the 10% level. These results, derived from time-series changes in shareholder leverage constraints following mergers of asset management companies, reinforce the causal relationships identified in our main findings.Footnote ²⁶

Table 10Mutual Fund Family Merger Analysis

Table 10 Long description

The table presents regression coefficients and t-statistics in parentheses for nine dependent variables, each represented by a column numbered 1 through 9.

Column headers (Dependent Variables):

1. Delta C D S Spread

2. Delta Yield Spread

3. Delta Loan Spread

4. Delta Collateral Requirement

5. Delta Number of Covenants

6. Delta Max. Lev Slack

7. Delta Operating Volatility

8. Delta Operating Leverage

9. Delta Book Leverage

Key Independent Variable:

Delta Fund leverage constraint: Shows positive significant correlations with Delta C D S Spread (0.457), Delta Yield Spread (1.696), Delta Loan Spread (0.420), Delta Collateral Requirement (0.124), Delta Operating Volatility (0.007), and Delta Operating Leverage (0.070). It shows a negative significant correlation with Delta Book Leverage (-0.013).

Control Variables:

Includes Delta Weighted fund beta, Delta Fund ownership, Delta Firm size, Delta Profitability, Delta Free cash flow, Delta Tangibility, Delta Book-to-market, Delta Book leverage, and Delta Amihud illiquidity. Delta Amihud illiquidity shows a strong positive correlation with Delta C D S Spread (10.003) and Delta Loan Spread (1.964).

Bottom Summary Statistics:

- Delta Other relevant controls: Yes for all columns.

- Specification: References to Tables 3, 4, 5, and 7.

- Number of observations: Ranges from 50 (Column 6) to 5,792 (Column 9).

- R-squared: Ranges from 0.009 (Column 8) to 0.227 (Column 6).

C. Instrumental Variables Regression

Third, we provide a direct test of causality using an instrumental variable specification. The instrument is based on Edmans, Goldstein, and Jiang (Reference Edmans, Goldstein and Jiang2012), which identifies an exogenous supply-driven shock using the occurrence of extreme fund outflows. Such outflows affect a fund’s need to borrow. The Investment Company Act allows mutual funds, subject to formal regulations and reporting requirements, to “borrow for liquidity” in order to meet redemption requests. By borrowing to meet large redemptions, a fund avoids the immediate sale of portfolio assets. As stated by the SEC, “it is relatively common for funds to establish lines of credit to manage liquidity risk” (https://www.federalregister.gov/documents/2015/10/15/2015-24507/open-end-fund-liquidity-risk-management-programs-swing-pricing-re-opening-of-comment-period-for).

Using a new data set—specifically, recent N-CEN filings reported by open-end mutual funds after 2019—on credit line (and other types of loan) extensions by bank holding companies,Footnote ²⁷ Cai and Shin (Reference Cai and Shin2021) show that bank credit, particularly in the form of credit lines, has provided open-end funds with an important liquidity source from which they can draw cash in times of excessive fund outflows. These findings suggest that leverage constraints become more binding for funds facing extreme outflows. Therefore, we expect that during surges in redemptions caused by extreme fund outflows, mutual funds will be subject to tighter leverage constraints. We define a fund as facing an “extreme outflow” if it experiences redemptions of at least 5% of total fund assets. We expect that large outflows will exacerbate the binding nature of fund leverage constraints.Footnote ²⁸

Moreover, to mitigate the concern that extreme outflows may be driven by deteriorating firm conditions in the underlying portfolio companies, we identify, for each fund, a neighboring extreme fund outflow indicator. This indicator reflects whether another fund headquartered within 50 miles experiences extreme outflows. The intuition behind this approach is that neighboring funds tend to share a similar investor base and are subject to similar supply-driven shocks.

To further address the concern that investors with correlated liquidity shocks may invest in funds with correlated holdings, we construct firm-specific measures that proxy for the degree of correlation in liquidity shocks and fund holdings across funds holding a given stock. For correlated liquidity shocks, we follow Gong, Lin, and Zwinkels (Reference Gong, Lin and Zwinkels2017) to construct a measure of flow-driven skewness, which captures the correlation of liquidity-driven trading among mutual funds.Footnote ²⁹ This measure sums the skewness in the flows of a stock’s mutual fund investors and the co-skewness between their flows. A higher value indicates that the flows of a stock’s mutual fund investors are more negatively skewed and display greater co-movement. To proxy for the degree of correlated holdings among a firm’s mutual fund investors, we construct, for each stock quarter, the portfolio correlation across mutual funds.Footnote ³⁰ By controlling for these two measures, we mitigate the concern that the effect of the instrument is driven by correlated liquidity shocks.

We report the results of the instrumental variables analysis in Table 11. We find that the proportion of funds with neighboring funds experiencing extreme outflows in the previous year is strongly positively related to investor leverage constraints in the first-stage regression. The second-stage regression results confirm our previous findings: the more constrained mutual funds are in terms of leverage, the more they induce firm risk-taking in business activities. In addition, debtholders respond by charging a higher cost of borrowing, imposing tighter covenants, and leading to a reduction in firm leverage. A 1-standard-deviation increase in investor leverage constraints is associated with an 11.4% higher CDS spread, a 9.5% higher bond yield spread, a 3.2% higher loan spread, an 11.2% higher likelihood of collateral requirements, a 2.8% increase in operating cash flow volatility, a 4.8% increase in operating leverage, and a 1.5% decrease in book leverage, relative to the unconditional sample average. Overall, these results provide further evidence of causality.

Table 11Instrumental Variables Regression

Table 11 Long description

The table presents Instrumental Variables Regression results.

Header Structure:

Column 1 is the First-Stage Regression for V W Leverage Constraint.

Columns 2 through 10 are Second-Stage Regressions for various dependent variables: C D S Spread, Yield Spread, Loan Spread, Collateral Requirement, Number of Covenants, Max. Lev. Slack, Operating Volatility, Operating Leverage, and Book Leverage.

Key Row Data:

- Neighboring extreme fund outflows (Instrument): In the first-stage regression (Column 1), the coefficient is 0.069 with a t-statistic of 11.67, significant at the 1 percent level.

- V W leverage constraint (Instrumented): In the second-stage regressions, it shows positive significant coefficients for C D S Spread (1.068), Yield Spread (1.653), Loan Spread (0.695), Collateral Requirement (0.795), and Number of Covenants (1.072). It has a negative significant coefficient for Book Leverage (-0.056).

- Controls: Includes Weighted fund beta, Fund ownership, Firm size, Profitability, Free cash flow, Tangibility, Book-to-market, Book leverage, Amihud illiquidity, Correlation in liquidity shocks, and Correlation in fund holdings.

- Fixed Effects: All models include Year F E. Columns 1 to 3 and 8 to 10 include Firm F E. Columns 4 to 7 include Industry F E and Credit rating F E.

- Observations: Range from 2,079 (Column 7) to 59,176 (Columns 1 and 10).

Taken together, all previous results suggest a causal relationship between shareholder leverage constraints and debtholder-related outcomes. Furthermore, we address concerns regarding reverse causality by performing Granger causality tests. Specifically, we regress the yearly change in the value-weighted fund leverage constraint on the yearly change in key variables identified as being affected by leverage constraints (i.e., CDS spread, bond yield spread, loan spread, likelihood of collateral requirements, number of covenants, maximum leverage covenant slack, operating cash flow volatility, operating leverage, and book leverage). All independent variables are lagged by 1 year.Footnote ³¹ We also include lagged changes in all relevant control variables and lagged shareholder leverage constraints in each regression. The results show no significant evidence of reverse causality from debtholder-related variables to shareholder leverage constraints.Footnote ³²

A. Implications on Market Risk of Corporate Bonds

As previously discussed, even though leverage-constrained investors are willing to accept a lower return–risk tradeoff than their unconstrained counterparts, this does not imply that they become risk-loving in all respects. In particular, leverage-constrained investors increase firms’ exposure to priced systematic risk. Our results confirm this, and the economic channel operates through business activities such as higher operating cash flow volatility and higher operating leverage. We now explore the implications of such systematic business risk exposure for debtholders. Extant studies have documented that firms with greater business exposure to macroeconomic shocks, as measured by operating leverage, face a higher cost of borrowing (Chen, Collin-Dufresne, and Goldstein (Reference Chen, Collin-Dufresne and Goldstein2009), Kuehn and Schmid (Reference Kuehn and Schmid2014), and Bhojraj et al. (Reference Bhojraj, Bloomfield, Jang and Yehuda2021)). This suggests that the systematic risk profile of a firm’s underlying business is an important consideration for debtholders. In addition, recent research has shown that exposure to the overall corporate bond market is a key factor in explaining cross-sectional bond returns (Dickerson, Mueller, and Robotti (Reference Dickerson, Mueller and Robotti2023)).

Therefore, we expect that the higher systematic risk embedded in a firm’s business should manifest in the sensitivity of bond returns to the overall bond market. This suggests a positive relationship between shareholder leverage constraints and the sensitivity of bond returns to the corporate bond market (“bond beta”). To test this, we estimate the bond-specific beta as the regression coefficient of monthly excess bond returns (over the risk-free rate) on the bond market’s excess return (over the risk-free rate), using monthly bond returns from the current year and the following 2 years. We then calculate the issue-amount-weighted bond beta at the issuer level and examine its relationship with the measure of shareholder leverage constraints.

The results reveal that a firm’s shareholder leverage constraint is positively related to its bond beta. This suggests that higher systematic risk-taking in the firm’s business operations increases the sensitivity of bond returns to the overall bond market, prompting bondholders to demand a higher cost of borrowing for firms held by leverage-constrained shareholders.Footnote ³³

B. Implications on Cost of Equity

Next, we consider the implications on cost of equity. All else equal, leverage-constrained investors are willing to accept a lower net-of-risk return in exchange for higher gross-of-risk performance. They achieve this by influencing corporate managers to increase business exposure to systematic risks, thereby raising the cost of equity. To quantify this effect, we match each firm with a comparable firm that is closest in terms of market size and belongs to the same industry (using either the 3-digit SIC code or the Hoberg–Phillips industry classification). We then calculate the excess cost of equity as the excess market beta (i.e., the beta of the stock minus the beta of the matched firm) multiplied by an assumed market risk premium of 6%. Finally, we relate the excess cost of equity to shareholder leverage constraints.Footnote ³⁴ The intuition is that the comparable firm has the “correct” beta given its characteristics, whereas the studied firm’s beta may be distorted by the influence of leverage-constrained funds. The excess cost of equity serves as a proxy for the higher cost of equity associated with leverage-constrained investors. We find that a 1-standard-deviation increase in shareholder leverage constraints is associated with a 5.6% to 6.7% standard deviation higher excess cost of equity.Footnote ³⁵

C. Discussions on Equity Value

Finally, we discuss the implications of shareholder leverage constraints on equity value. We argue that leverage constrained mutual funds want to influence their portfolio companies to take higher risk in their business operations, which would elevate the stockholder–bondholder conflict and cause a wealth transfer from debtholders to equity holders. In other words, when facing mutually exclusive projects with similar NPVs, we expect that firms influenced by leverage constrained mutual funds would favor the high-beta projects at the expense of debtholders.Footnote ³⁶ This is due to the nature of equity as a call option on the firm’s underlying asset (Merton (Reference Merton1974)). Greater operational risks enhance the call option’s value (equity) due to the asymmetric nature of option value: equity holders gain from upside potential while facing limited downside risk. More severely, firms may even choose negative NPV projects if the wealth transfer effect is large enough to compensate the loss in firm value in choosing negative NPV projects (Damodaran (Reference Damodaran1999)).

These discussions align with our previous findings that shareholder leverage constraints have a greater influence on corporate risk-taking and debt costs for firms experiencing heightened debtholder-shareholder conflict. In this scenario, the impact on equity value resulting from the transfer of wealth from debtholders to equity holders will be more pronounced in financially distressed firms, while this effect tends to be mitigated for those where shareholders hold both equity and debt stakes in the company.Footnote ³⁷