The agency theory admits conflicts between different individuals who, in the context of debentures, correspond to disagreements between bondholders, shareholders, and managers. Thus, covenants are effective instruments in mitigating agency problems linked to debt contracts (Smith & Warner, 1979).

According to Jensen and Meckling (1976), individuals’ rights are determined by signed contracts, and the agency relationship originates from established contracts between one or more people. That is, the principal hires an agent in a competitive market and can choose the way of this agent will perform, but principal cannot force this agent to act in accordance with what he expects (Christensen & Feltham, 2005). Hence, due to property rights’ reduction, a manager's effort for maximizing enterprises’ value can also be diminished (Jensen & Meckling, 1976).

Since both contract parties, principal and agent, aim to maximize their benefits, it cannot be expected that an agent will always act in accordance with a principal’s interests, but the principal may limit the agent's behavior with appropriate incentives. Therefore, it is impossible for the agent to perform at an optimal level, from principal's point of view, without any involved cost. Thus, the main agency costs can be classified as follows: (i) monitoring costs, (ii) contractual guarantees cost; and (iii) residual cost (Jensen & Meckling, 1976).

It is not usual for large companies to be financed almost entirely by outside capital because, in the face of a financial structure mostly composed of creditors' capital, the owner-manager can be encouraged to engage in higher risk activities. Hence, if decisions are successful, they can generate higher personal compensation, but if they are not, creditors will assume most of the costs (Jensen & Meckling, 1976).

For Sheng (2005) "all corporate debt issuances provokes conflicts between bondholders and shareholders, also known as moral hazard." (p. 61) Moral hazard, as presented by Beaver (1998), results from information asymmetry, which enables agents to have inside information for acting in accordance with their own interests, thus expropriating the principal.

The moral hazard associated with conflicts between bondholders and shareholders is related to shareholders’ detrimental behavior in regards to dividends payment, new debtsm or even changes in company's policy investment (Smith & Warner, 1979).

For bondholders, agency costs correspond to monitoring and the possibility of debt contract renegotiation. On one hand, the lack of monitoring for bondholders can encourage managers to reduce creditors’ wealth, which leads to credit quality deterioration. Hence, bondholders may require a higher interest rate to offset these risks (Saito, Sheng, & Bandeira, 2007; Ghouma, 2017).

On the other hand, if there is a credit quality improvement, issuing companies may choose to renegotiate the debt agreement, or even modify new issued debentures’ indentures. Thus, this new contract can involve additional costs, which are feasible if benefits to one of the parties were sufficient for justifying them (Saito, Sheng, & Bandeira, 2007).

Restrictive clauses in debentures’ contracts are significantly influenced by managerial performance, and indentures are efficient for evaluating different risks associated to managerial fraud. Hence, covenants are relevant, since they correspond to an opportunity for reducing bondholders’ agency cost and financing cost (Chava, Kumar, & Warga, 2010).

Moreover, covenants are common features in debt contracts and they antecipate potential managers and shareholders’ opportunistic performance by imposing limitations on bond issues which inhibit bondholders’s wealth transfer (Bradley & Roberts, 2015; Devos, Rahman, & Tsang, 2017).

Thus, in the Brazilian reality, one way to mitigate these conflicts is the adoption of standardized indentures, which are easier to understand, because they have simpler clauses. So, bondholders have their rights protected by fiduciary agents who monitor companies' activities (Sheng, 2005).

The literature about covenants show situations in which these clauses grant shareholders and bondholders conflicts mitigation (Smith & Warner, 1979; Billet, King, & Mauer, 2007). There must be a balance between these clauses, since, on the creditors side, there are those that correspond to monitoring and deb contract renegotiation while, on the issuing companies side, those that can influence decision making and investment policy (Saito, Sheng, & Bandeira, 2007).

Both bondholders and managers are incentivized to maintain covenants, and it is feasible for creditors to support the cost of drafting clauses and monitoring managers’ actions until the moment the marginal cost of these activities will be the same as marginal benefits obtained with this contract elaboration (Jensen & Meckling, 1976).

Covenants are not uncommon in debt issues (Jensen & Meckling, 1976), they have been a typical feature for over 100 years, although standardized contracts were few at the time (Smith & Warner, 1979). These clauses need to be detailed enough to cover companies’ operational aspects, including projects’ risk. Therefore, standardized debt securities contracts are suggested, especially in incomplete markets (Sheng, 2005).

Smith and Warner (1979) considered 87 public debt issues from January 1974 to December 1975 and noted that opportunity costs associated with restrictive clauses were substantial because, although they involved additional drafting costs, they reduced agency conflict’s costs.

Covenants reduce debt cost and increase firms value. However, this does not mean that these clauses, considered optimal at the time of issuance, will remain relevant over time. Since these clauses can be amended, with the approval of creditors, bondholders coordinate their actions to modify or disapprove disadvantageous changes, in view of seeking gains with these indentures’ changes (Kahan & Tuchman, 1993).

Debentures are flexible and adaptable in accordance with bondholders’ well-being. In Brazil, there are more debentures that relate to floating interest rates rather than to inflation, there are less restrictive (or unrestricted) clauses associated with financing, whereas the most restrictive clauses concern changes in companies control, and the lack of guarantees (Saito, Sheng, & Bandeira, 2007).

Companies adopt restrictive clauses and short-term debt for mitigating shareholder and bondholder disputes. An analysis of more than 50 covenants for 15,000 issued debentures by non-financial corporations, from 1960 to 2003, revealed that these clauses have grown over the years, especially in environments with greater leverage and growth opportunity. These clauses decreased with the increment of short-term debt (Billet, King, & Mauer, 2007).

Concerning 159 issued debentures in Brazil from 2000 to 2009 by 82 different companies, it is possible to corroborate this evidence, since the main results of this study show that restrictive clauses and short-term financing are alternative tools for minimizing agency conflict between shareholders and creditors. Moreover, due to these covenants, companies with growth possibilities are more likely to switch from short to long-term financing. In this way, covenants tend not to restrict growth opportunities (Silva, Saito, & Barbi, 2013).

According to the aforementioned studies, debentures in Brazil are more flexible than other financing alternatives, and can offer different compensation rates, re-contracting clauses with lower transaction costs, and their deeds can involve call and put options (Saito, Sheng, & Flag, 2007). Covenants number and types depend on the level of agency conflicts, and on costs and benefits associated with restrictive clauses (Qi, Roth, & Wald, 2011). In addition, bondholders generally choose to protect themselves against the negligent performance of managers by means of debentures’ restrictive clauses (Nash, Nette & Poulsen, 2003). Based on this reasoning, our hypothesis corresponds to:

H₁: bondholder’s protection index is positively related to the percentage that multimarket funds allocate in corporate bonds.

The sample period starts in 2009. The Brazilian investment funds industry has increased, since 2009 at an average rate of increase of about 9.1% until 2017 (ANBIMA, Consolidated Historical of Investment Funds, 2018). Moreover, we were able to collect information related to portfolio composition of investment funds starting in 2009, and in such year there was the publication of ICVM 476, that allowed an expansion in the emission of corporate bontds and operations in the capital market (Carvalho, 2017).

The sample period ends in 2017 since it is the last year in which data was available to download when we started this research. The year of 2017 also showed a large inflow of resources through corporate bonds, about BR$ 88 billion and the investment funds industry showed an amout of total net assets close to BR$ 4.1 billion.

The analysis of corporate bonds was annual and each debentures’ series was evaluated considering its issue date too its due date. The data collection involved several different databases (ANBIMA website, CVM website and Economatica® Database) and comprises 23,480 observations.

The dependent variable of this study is the multimarket funds interest for corporate bonds. This variable was estimated based on the value that each multimarket funds invested in each corporate bond over the sample period, as described in Chart 1. The value invested by the fund was weighted by its respective total net assets.

Chart 1.

Variable		Description	Measurement	Source
Dependent	%FUND	Percentage of each multimartket fund portfolio invested in each corporate bond	%FUNDijt= Value (in BR$) that the fund i invested in the bond j in the period tTotal Net Asset (in BR$) of the fund i in the period t	(1); (2)
Independent	BPI	Bondholders’ Protection Index	BPIi= Sum of 15 Covenants15	(3)
	BPI * AM	Interaction between BPI and the number of “automatic maturity” clauses	This is the interaction between BPI and AM, in which AM is a dummy variable that receives 1 for debentures' indentures with the large number of “automatic maturity” clauses. We used two measures of position to define the “large number” of clauses: the median and the third quartile of the number of automatic maturity clauses.	(3)

Note. Data sources: (1) Economatica®; (2) CVM; (3) Debentures ANBIMA. Source: research database.

As indicated in Chart 1, the independent variable is BPI. This index was calculated based on Billet, King and Mauer (2007), and Silva, Saito and Barbi (2013) and it also involves restrictive clauses and clauses that indicate what companies need to follow. The advancement of this research is to consider the interaction of BPI with the number of “automatic maturity” clauses available in the debentures’ indentures. This interaction was done through the creation of a dummy variable that receives 1 for debentures’ indentures that have a large number of “automatic maturity” clauses when compared to the other indentures. As the cutpoint, we used two measures of position: the median and the third quartile of the number of automatic maturity clauses. The third quartile was considered as an evaluation of the BPI behavior in the presence of a large number of automatic maturity clauses.

Therefore, the BPI comprises the following restrictive clauses: dividends restrictions; capital reduction; liquidation, dissolution and bankruptcy; modification of business essence; modification in the company’s structure; modification in the control; assets sale or transference; negligence; legal obligations and environmental permits; financial indexes; investments; leverage; stocks emission or amortization; rating classification; and merger, split and incorporation. Each one of these 15 covenants is a dummy variable that receives 1 when the indenture has the respective clause and 0 if it does not. The bondholders’ protection index considers the sum of the 15 covenants, divided by 15, which is the maximum score that each document can achieve. Therefore, the BPI varies from 0 to 1, and values next to 1 indicate indentures with higher number of clauses for bondholder’s protection.

Regarding control variables, we considered the characteristics of corporate bonds, the features of companies, and the characteristics of multimarket funds, following evidence from previous research on corporate bonds and investment funds. Chart 2 contains the description of such variables.

Chart 2.

Variable		Description	Measurement	Reference	Source
Corporate Bonds	REST	Corporate Bonds issued with restrict efforts	Dummy, 1 for indentures issued with restrict efforts following the instruction CVM 476 and 0 for the other cases	Konraht and Vicente (2017) Bragança, Pessoa and Souza (2015) Billet, King and Mauer (2007) ANBIMA (2018)	(3)
	INCENT	Incentivized Corporate Bonds	Dummy, 1 for incentivized corporate bonds and 0 for the other cases
	REAL	Kind of Corporate Bonds Guarantee	Dummy, 1 for indentures with the guarantee is real and 0 for the other cases
	MAT	Corporate Bonds Maturity	Dummy, 1 for indentures with maturity equals or lower than three years and 0 for the other cases
	RATE	Rate of Remuneration of Corporate Bonds	Bonds Indexers: Dummy, 1 for bonds indexed as a percentage of ID (interbank deposits) rate and 0 for the other cases (IPCA and other rates)
Companies	SIZE	Company Size	SIZEit= NL(Market Value) where: “nl” represents the natural logarithmic and Market Value representes the product between the stock’s price and the number of stocks of the company	Fama and French (2015)* Sobrinho (2016)	(1)
	BTM	Book-to-market	BTMit= Book ValueMarket Value
	PROF	Return over the Net Worth	PROFitNet ProfitNet Worth
	INV	Investment	INVit= Total Asset t-1-Total Asset t-2Total Asset t-2
	NM	B3’s New Market	Companies listed at the New Market of B3	Ripamonti and Kayo (2016)	(4)
Multimarket Funds	NL(TNA)	Fund’s Size	NL(TNA) = Natural Logarithmic of Total Net Assets	Amin and Kat (2003) Rochman and Eid Jr. (2006) Milani and Ceretta (2013) Malaquias and Eid Jr. (2014) Bono Milan and Eid Jr. (2014) Malaquias and Mamede (2015) Malaquias and Pontes (2018)	(1); (2)
	FoF	Funds of Funds	Dummy, 1 for funds of funds and 0 for the other cases
	MGMFee	Management Fee, charged as a remuneration by the management services	Represents the maximum management fee charged by the fund in the year (in %)
	PERFFee	Performance Fee, charged when the profitability of the fund is higher than a benchmark previously established	Dummy, 1 for funds that charge performance fees and 0 for the other cases
	AGE	Fund’s Age	Fund’s Age, in years, until December 31, 2017

Note. Data Souces: (1) Economatica®; (2) CVM; (3) Debentures ANBIMA; (4) B3 Website. * In the study of Fama and French (2015), the size was measured by the difference between the average return of stocks from small firms and the average return of stocks from large firms. However, as we do not construct factors in this study, we consider the natural logarithmic of the market value of each firm as a proxy for its size. Source: research database.

Our analysis also includes a variable for the crisis period, which is a dummy variable that receives 1 for the years 2015 and 2016 and 0 for the other years. In the years of 2015 and 2016, Brazil experienced its worst accumulated recession since 1948, with GDP falling by 7.2% (Saraiva & Sales, 2017). In this time period, we also observed a reduction in the number of domestic emissions in the Brazilian capital market and, particularly in 2015, a negative net inflow in the investment funds (fixed income, stocks and multimarkets).

The database of this study comprises cross sectional data (i) related to the acquisition of corporate bonds by different multimarket funds over a time series (t), from January 2009 to December 2017. Therefore, we employed a regression model with panel data (Greene, 2002) as described in Equation 1.

%FUNDijt= β0+ β1BPIj+ β2BPIj*AM + β3CVBj+β4CVCet+ β5CVFit+β6CRISISt+ uejit  (1)

Where:

e = refers to the company “e”;

j = refers to the corporate bond “j”;

i = refers to the multimarket fund “i”;

t = refers to time;

β = coefficients of the variables;

%FUND = this is the dependent variable, it represents the percentage of each multimartket fund portfolio invested in each corporate bond;

BPI = independent variable and it represents the Bondholders’ Protection Index;

BPI*AM = independent variable and it represents the Interaction between BPI and the number of “automatic maturity” clauses;

CVB = control variables related to the corporate bonds characteristics;

CVC = control variables related to the companies features;

CVF = control variables related to the funds characteristics;

VCFM = variáveis de controle correspondentes às características dos fundos multimercados;

CRISIS = dummy variable for periods of crisis.

u = error term.

The variable %FUND indicates the percentage invested in each corporate bond by each fund in the sample. In other words, it represents the multimarket funds interest for corporate bonds.

We can observe in Table 1 that many years had the minimum value equal to zero. These cases correspond to situations in which the value invested in corporate bonds is less expressive when compared to the total net assets of the respective fund. Therefore, the percentage is next to zero. An opposite situation can be observed in the year 2016, in which one fund of the sample invested more than 99% of its total net assets in a specific corporate bond. This bond was issued by the company Ceb Distribuição S. A. in 2015 (due date is 2020).

Table 1.

Year	Nº	Average	SD	Max.	Min.
2009	137	1.48	1.92	9.30	0.07
2010	712	1.62	1.62	12.33	0.01
2011	1171	1.56	1.53	12.75	0.00
2012	2489	1.31	1.42	24.67	0.00
2013	2917	1.27	1.47	23.14	0.00
2014	3608	1.36	1.54	27.94	0.00
2015	3660	1.41	1.80	43.00	0.00
2016	3977	1.45	2.72	99.58	0.00
2017	4809	1.24	1.70	47.47	0.00
Total	23480	1.36	1.85	99.58	0.00

Note. We considered the information of portfolio holdings of each multimarket fund at the end of December of each year. Source: research database.

Despite the representative case commented on in the previous paragraph, the multimarket funds of the sample allocate on average 1.36% of their TNA in debentures. In this way, there is a relatively small participation of Brazilian multimarket funds in the acquisition of corporate bonds, even with the significant participation of institutional investors as debentures subscribers in Brazil.

Regarding the independent variable: first, we analyzed the frequency of restricted clauses related to the index (BPI). The results in Table 2 indicated that, on one hand, the most frequently cited clauses are: (i) negligence by the issue companies; (ii) liquidation, dissolution and bankruptcy; and (iii) restrictions related to the modification in the company’s structure. All of these three clauses presented a relative frequency higher than 94%. On the other hand, the clauses with the lower frequency are: (i) leverage policy; (ii) stocks emission or amortization; and (iii) lowering in the rating classification. These items had a relative frequency equal or lower to 27%. Table 2 also highlights information about clauses with automatic maturity. Therefore, we can observe that some of the most frequently cited clauses are not necessarily expressed as items that guarantee an immediate payment to the bondholders. For example, clauses related to financial indexes had a relative frequency of 76%, but they were identified as automatic maturity clauses in only 4% of the cases.

Table 2.

Clauses Description	Non Identified Clauses		Identified Clauses						Obs.
			Total		Automatic Maturity?
					No		Yes
	AF	RF	AF	RF	AF	RF	AF	RF
Dividends Restrictions	176	19.01%	750	80.99%	367	39.63%	383	41.36%	926
Capital Reduction	132	14.25%	794	85.75%	397	42.87%	397	42.87%
Liquidation, Dissolution and Bankruptcy	2	0.22%	924	99.78%	35	3.78%	889	96.00%
Modification of Business Essence	257	27.75%	669	72.25%	439	47.41%	230	24.84%
Modification in the Company’s Structure	50	5.40%	876	94.60%	150	16.20%	726	78.40%
Modification in the Control	99	10.69%	827	89.31%	508	54.86%	319	34.45%
Assets Sale or Transference	321	34.67%	605	65.33%	416	44.92%	189	20.41%
Negligence	1	0.11%	925	99.89%	37	4.00%	888	95.90%
Legal Oblig. and Environmental Permits	389	42.01%	537	57.99%	355	38.34%	182	19.65%
Financial Indexes	217	23.43%	709	76.57%	669	72.25%	40	4.32%
Investments	515	55.62%	411	44.38%	192	20.73%	219	23.65%
Leverage	787	84.99%	139	15.01%	83	8.96%	56	6.05%
Stocks	685	73.97%	241	26.03%	110	11.88%	131	14.15%
Rating Classification	676	73.00%	250	27.00%	184	19.87%	66	7.13%
Merger, Split and Incorporation	135	14.58%	791	85.42%	392	42.33%	399	43.09%

Note. AF: Absolute Frequency; RF: Relative Frequency. Source: research database.

The bondholders’ protection index (BPI) reveals the level of protection offered to bondholders, following each bond issued. These clauses represent a way to reduce debt costs and increase companies’ value (Smith & Warner, 1979; Kahan & Tuchman, 1993), as well as allowing to minimize bankruptcy and mitigate agency conflicts (Jerzemowska, 2006).

The results in Table 3suggest that, on average, the value of BPI was 0.6731. The maximum value was 1.000 and it indicates that the sample includes at least one debenture indenture that has all the clauses. Regarding the minimum value of 0.0667, it indicates that all debenture indentures of the sample have at least one of the clauses that comprise the BPI.

Table 3.

Year	Nº	Average	SD	Max.	Min.
2009	53	0.6352	0.1369	0.9333	0.3333
2010	99	0.6290	0.1337	0.9333	0.3333
2011	90	0.6696	0.1374	0.8667	0.2000
2012	164	0.6134	0.1686	1.0000	0.2667
2013	107	0.6430	0.1374	0.9333	0.2667
2014	105	0.6984	0.1542	0.9333	0.1333
2015	82	0.7252	0.1399	0.9333	0.2667
2016	60	0.6878	0.1640	0.9333	0.2000
2017	166	0.7450	0.1353	0.9333	0.0667
Total	926	0.6731	0.1536	1.0000	0.0667

Source: research database.

Regarding the characteristics of corporate bonds, Table 4 indicates that they usually have a due date of up to three years (88.84%), have restrict efforts (73.47%), and that their remuneration is indexed to the ID rate (73.04%). Moreover, the frequency of incentivized corporate bonds is low (7.23%) and they generally do not have real guarantee.

Table 4.

Characteristics	Yes		No		Total
	AF	RF	AF	RF	AF	RF
Restrict Efforts	2752	73.47%	994	26.53%	3746	100%
Incentivized Corporate Bonds	271	7.23%	3475	92.77%
Corporate Bonds Guarantee (Real)	614	16.39%	3132	83.61%
Maturity <= 3 years	3328	88.84%	418	11.16%
Rate of Remuneration (ID)	2736	73.04%	1010	26.96%

Note. AF: Absolute Frequency; RF: Relative Frequency. Source: research database.

As presented in Table 5, on average, the companies of the sample have a profitability index of 13% and 8% of investments expansion. The average size, measured by the natural logarithm of companies’ market value, is 15.84 on average. Based on the results for the variable book-to-market, we can observe that firms tend to present a market value higher than their book value, since the average index was 0.8260.

Table 5.

Variables	Nº	Average	SD	Max.	Min.
Profitability	23480	0.1351	0.2139	0.5959	-1.9294
Investiment	23480	0.0786	0.2767	2.1586	-0.3087
Size (NL of Market Value)	23480	15.8370	0.9793	19.1910	12.8096
Book-to-market	23480	0.8260	0.6545	3.3099	0.0805

Note. Source: research database.

The sample of this study comprises 1753 multimarket funds. The size of the funds included in the sample, measured by the natural logarithm of their TNA, corresponds to 18.03, as described in Table 6.

Table 6.

Variables	Nº	Average	SD	Max.	Min.
NL(TNA)	23480	18.0331	1.2798	21.9142	15.4690
Management Fee	23480	0.4267	0.6199	5.0000	0.0000
Fund’s Age	23480	9.0631	4.6806	22.4917	0.7750
Funds of Funds	23480	0.0341	0.1814	1.0000	0.0000
Performance Fee	23480	0.1695	0.3752	1.0000	0.0000

Note. Source: research database.

The management fees are 0.43% by year, on average, and the age of the funds is 9 years (on average). Only 3.41% of the sample are funds of funds and about 16.95% of the samples correspond to funds that have performance fees.