2.1. Trade-off

This theory proposes that the issue securities of a company move its capital structure towards an optimal leverage that is determined by a trade-off between the marginal costs and benefits of the debt. Debt costs consist of potential bankruptcy and agency costs arising from conflicts between shareholders and bondholders; while the benefits of debt include, for example, the debt tax shield and the overinvestment problem control (^{Jensen, 1986}; ^{Fama and French, 2005}).

In relation to the evidence, the debate regularity has been focused on that companies’ issue stocks when their prices are high (^{Dittmar and Thakor, 2007}). According to the theory, the stock prices increase reduces the firm’s debt ratio, which should lead to the issuance of debt toward the optimal leverage. However, ^{Baker, and Wurgler (2002)}, among other authors, find that US companies issue equity instead of debt when stocks prices are high (^{Dittmar and Thakor, 2007}, p. 1). On the other hand, ^{Mongrut, Fuenzalida, Pezo, and Zdenko (2010)} analyze the pecking order and trade-off theories for companies of Argentina, Brazil, Chile, Mexico and Peru and find that firms behave according to the trade-off theory. The empirical evidence provided by the US market is inconsistent with the trade-off theory, while the capital structure of Latin American firms would be according to it.

2.2. Signaling

The signaling theory is related to the existence of information asymmetries and to the assumption that managers often have better information about the firm value than external investors.

In markets with asymmetric information, it is difficult to distinguish good quality from bad quality and this difficulty is inherent to the business world (^{Akerlof, 1970}, p. 500). In this context, the work of ^{Spence (1972)} on signaling, whose fundamental idea is that individuals can perform actions to provide information to others, despite the fact that these actions have no effects on productivity or what the buyer wants (^{Lazear, 2008}).

The information asymmetries between management and investors have led to two theories of funding decisions: 1) pecking order; and 2) signaling. According to the signaling and the idea of ^{Spence (1972)}, the firm management can make financing decisions to provide information to investors and cause a market reaction, even when that decision has no effect on firm productivity. In this regard, there are authors who argue that managers who make financial decisions are concerned with the signaling effects of such decisions (^{Barclay and Smith, 1999}).

With regard to financing decisions and the bond issuances, there are different theories in the financial literature about the announcements of the bond issuances and their impact on stock prices reaction. For this reason, ^{Castillo (2004}) classifies these theories into three groups: 1) those that state that the advertisement does not generate abnormal returns; 2) those that argue that the announcement of the risky debt issuance has a negative impact on the firm value and negative abnormal returns; and 3) those that state that generate positive abnormal returns.

In relation to the last group, and from an information asymmetry context about firms’ activities, an increase in the debt level communicates to the market that the firm is of good quality. This fact constitutes an effective signaling strategy (^{Ross, 1977}; ^{Barclay and Smith, 1999}). Debt contracts oblige the company to make fixed payments, and its failure to comply has serious consequences including bankruptcy of the firm (^{Barclay and Smith, 1999}). Therefore, the increase in indebtedness can be a credible (and costly to imitate) sign that the company expects higher cash flows in the future (^{Ross, 1977}; ^{Barclay and Smith, 1999}, p. 12). On the other hand, ^{Barclay and Smith (1999)} point out that stock prices are more sensitive than bond prices to information about future expectations. Therefore, signaling theory suggests that the management will choose to issue debt if it believes the firm is undervalued.

The empirical evidence for the US market is diverse and shows that the announcements of debt issuances generate stock price increases or does not affect them (^{Castillo, 2004}). Regarding the relationship between bond issuance and information asymmetries, there is evidence that firms with greater information asymmetries issue more short-term debt (^{Barclay and Smith, 1995}). There is also evidence about negative and significant reactions in bond and stock prices when the bond issue is motivated by an unexpected cash flow deficit (^{Akhigbe, Easterwood, and Pettit, 1997}). In other markets the results are also diverse. The firms’ characteristics such as the dividend payments can affect the market reaction. The market would react positively and significantly when the announcements are made by companies with low dividend payments, but such reaction would not be relevant if the announcements are made by firms with high dividend payments. This fact supports the substitute relationship about the information on debt and dividends (^{Verona and García, 2006}).

In Chile, the authorization of bond issues granted by the market regulator does not produce significant abnormal returns (^{Castillo, 2004}). In other emerging markets such as Malaysia, there is evidence about positive abnormal returns after bond issuance, indicating that the market considers bond offerings as favorable news. Even the signaling effect of bond issuance announcements is not influenced by firm characteristics (^{Chin and Abdullah, 2013}). In Malaysia, Singapore and Thailand also find positive and significant abnormal returns (^{M’ng, Rahman, and Kit, 2020}).

The empirical results are ambiguous and vary among markets. In this regard, ^{Fama and French (2005)} point out that event studies that calculate abnormal returns cannot resolve the issue unambiguously, since price responses to announcements may not be due solely to asymmetric information problems.

2.3. Pecking order

This theory was formulated by ^{Myers and Majluf (1984)} and suggests that corporate financing depends on the firm’ preference for internal sources of funds, and debt and equity as external financing. Information asymmetries are assumed to exist because management knows more about the firm value than potential investors, and therefore investors interpret the firm’s decisions rationally. The authors suggest that capital structures are the result of individual financing decisions in which the management follows a hierarchical order (pecking order). Furthermore, the management does not follow an optimal long-term capital structure, but simply chooses the financing alternative with the lowest cost and information asymmetries (^{Barclay and Smith, 1999}). The pecking order theory assumes that management is better informed than investors, which generates adverse selection costs that would affect the costs and benefits related to the trade-off theory. Therefore, companies will finance new investments with retained earnings, then with risk-free debt, risky debt, and finally with equity (^{Dittmar and Thakor, 2007}).

The empirical evidence indicates that the US companies’ issue or withdraw capital each year and rejects the pecking order theory main predictions about the frequency and circumstances in which firms issue and repurchase stocks (^{Fama and French, 2005}). For Latin American companies, the evidence indicates that the pecking order theory does not explain the financing policy, since Latin American firms prefer to contract debt rather than finance themselves with generated funds (^{Mongrut et al., 2010}). In addition, ^{Goodell and Goyal (2018)} studied 21 emerging markets and showed that firms prefer the issuance of bonds over bank loans. This would occur when there is less corporate opacity, lower foreign access restrictions, an environment with lower transaction costs and limits to legal protection.

The studies by ^{Mongrut et al. (2010)} and ^{Goodell and Goyal (2018)} show the importance of bonds as a financing source in Latin America and emerging markets, which suggests an interest in analyzing the effects of the bond issuance announcements on the stock returns.

2.4. Hypotheses

Based on signaling theory, this research raises hypotheses H₁ and H₂:

H₁: Chilean companies experience average positive abnormal returns on the announcement day (t = 0) of corporate bond issuance.

H₂: Chilean companies experience average positive abnormal returns on the day after the announcement (t = + 1) of corporate bond issuance.

According to ^{Ross (1977)} and ^{Barclay and Smith (1999)}, if management has information that investors do not know, then the financial structure choice conveys information to the market and relates the debt increase to its perception of value. For this reason, the announcement of a bond issue is expected to generate a positive average abnormal return at t=0 or at t=+ 1 if the market reacts late.

The average effect is expected to be a positive abnormal return. However, there is evidence that companies with low growth opportunities, large or regulated, have more long-term debt in their capital structure (^{Barclay and Smith, 1995}). For this reason, the firms’ characteristics could influence the signaling effect of the announcements. Companies can present positive or negative abnormal returns because the information conveyed by the announcement depends on market expectations about firm profitability and growth that are essential to evaluate their financing decisions (^{Fama and French, 2005}). Therefore, we proposed the third hypothesis (H3) that expect to find significant differences in the profitability and growth indicators between both groups of companies:

H₃: There are differences between the financial performance of companies with positive abnormal returns and those with negative abnormal returns.

3.1. Data and sample

The corporate bond issuance announ­cements made by Chilean firms listed on the Santiago de Chile Stock Exchange (BCS) between January 1, 2010 and December 31, 2017 are analyzed.

The study sample is defined as follows. First, the companies that announced bond issues between 2010 and 2017 are identified. Next, the firms that have a stock market presence between 240 days before the announcement date and 40 days after the announcement are selected¹. In accordance with the above, the study sample is made up of 29 companies that corresponds to 87 bond issuance announcements during the period. Table 1 shows the list of 29 firms that announced bond issues during the 2010-2017 period and the date of each bond issuance announcement.

3.2. Event study

To measure the effect of the announcement of the bond issue on the stock returns, the event study methodology is used. The events study makes it possible to analyze the impact of an event on securities prices of the companies (^{Brown and Warner, 1980}). According to this methodology, the stock abnormal returns are calculated on the day of the event (the announcement day) and around that date.

First, the event day (t=0) is identified as the date that firm announces the issue, and the Chilean Financial Market Commission (CMF²) publishes it as “essential event”. Then, a period of 240 days is defined before the announcement (^{Weston, Siu, and Johnson, 2000}). The period of the event is constituted by 3 days, one day before (t=-1) and one day after (t=+1) around the announcement date (t=0). The 39 days immediately prior to the event period (from t=-40 to t=-2) are excluded since they could be contaminated by information leaks (^{Aktas, De Bodt, and Roll, 2004}); while the previous 200 days (t=-241 to t=-41) constitute the period without event. The purpose is isolate the announcement effect. The event period starts from day t=-1 to analyze if there was an anticipated effect on returns.

Second, the normal or expected return of a stock i is calculated for each day t of the event period (), assuming normal conditions or absence of the event under study. ^{Brown and Warner (1980}, ¹⁹⁸⁵) proposed three models to calculate : 1) mean adjusted return; 2) market model; and 3) market adjusted return.

The mean adjusted return calculates as the average daily return of the stock for the 200-day non-event period (^{Weston et al., 2000}). Thus, is calculated using model (1), where R _it is the observed return of stock i on day t of the period without event of T=200 days:

According to model (2), the market model calculates through a linear regression between the daily observed returns of a stock i (denoted as R _it ) and the daily observed returns of a representative stock market index (R _Mt ), measured through General Stock Price Index (IGPA) of the BCS on day t of the period without an event. Both time series come from the non-event period. In model (2), α _i measures the average return; β _i measures the sensitivity of the return of stock i to market fluctuations; and ε _it is an error term. Model (2) estimates the values of α _i and β _i ( and , respectively). Then, these estimates are used in model (3) to calculate from the observed return of the IGPA for the event period. Therefore:

Finally, the market adjusted return assumes that is equal to the market return on each day t of the event period, as indicated by model (4):

These models appear frequently in the empirical literature about events study and each one has advantages and disadvantages. The mean adjusted return is perhaps the simplest model. However, ^{Brown, and Warner (1980}, ¹⁹⁸⁵) find that it often produces results similar to those of more sophisticated models. The market model represents a potential improvement over the mean adjusted return, since it eliminates the part of the return that is related to the market return variation. This fact reduces the abnormal returns variance and increases the model capability to detect effects around events (^{MacKinlay, 1997}). Regarding the market adjusted return, ^{MacKinlay (1997)} points out that this model has the greatest restrictions³, but it is useful when the data are limited (when it is not possible to have an estimation period prior to the event, for example, in a stock public offer).

It should be noted that the general properties of events study tests come from large-scale simulations by Brown and Warner (^{Khotari and Warner, 2007}). ^{Brown and Warner (1980)}argue that more complex methodologies can be worse, both compared to the market model and with the mean adjusted return. ^{Brown and Warner (1985)} confirm these conclusions using daily returns instead of monthly returns; and ^{Aktas et al. (2004)} point out that the event study method is robust to the choice of a specific return generation process. However, all three methods use the variance of the no-event returns to evaluate the returns of the event period. This is a limitation because adjustments are necessary when there are event-induced increases in variance or when the non-event period event is contaminated by unrelated events (^{Savickas, 2003}; ^{Aktas, De Bodt, and Cousin, 2007}).

Third, the abnormal return of a stock i is calculated for each day t of the event period. The abnormal return (r _it ) is the difference between the observed return of stock i on day t of the event period (R _it ) and the expected return in the absence of the event (). Then the average abnormal return (AR _t ) is calculated for each day t of the event period, according to model (5). The AR _t measures the average effect of the event on the stock returns of the n firms into the sample:

Fourth, the accumulated average abnormal return on each day τ of the event period (CAR _τ) is calculated using model (6), with τ=-1, 0, +1 days. The CAR _τ measures the cumulative average effect of the announcement during the event period.

3.3. Univariate model

The previous models have advantages but also disadvantages. For this reason, a fourth model is used to strengthen the analysis. According to ^{Li, Pincus and Rego (2008)}, a univariate analysis is performed according to model (7) in order to measure the average abnormal return related to the announcement.

Where RA _it is the abnormal return of stock i on day t. It is calculated as the difference between the return of stock i on day t (R _it ) and the market return on day t (R _Mt ), and over a period of 41 days of stock transactions around the announcement date (t =-20, -19,…, +19, +20)⁴. The market return corresponds to the return of the General Stock Price Index (IGPA). In addition, D _j is a dummy variable that takes the value 1 for the days of events j and 0 otherwise. Three dummy variables are used in the study: D ₁ is equal to 1 on day t=0 and 0 otherwise; D ₂ is equal to 1 on days t=0 and t=+1, and 0 otherwise; and D ₃ is equal to 1 on days t=-1, t=0 and t=+1, and 0 otherwise. In this way, the analysis focuses on day t=0 and on the windows closest to the day of the event: [0; +1] and [-1; +1].

In model (7), the constant α ₀ is the estimated average daily abnormal return for the days wi thout events. The coefficient β _j of each variable D _j is an estimate of the average daily abnormal return related to the announcement. To apply model (7), the sample was divided into two parts: a sample composed of the announcements that register positive abnormal returns on day t=0, and another sample composed of the announcements that register negative abnormal returns on day t=0. The sample is divided because positive abnormal returns can offset negative abnormal returns and dilute the effect of the announcement on stock returns. The division of the sample allows to observe more clearly the average effect of the announcement on each group.

3.4. Characterization of financial performance

The financial situation of the issuing companies is characterized by the following financial performance indicators: 1) return on assets (ROA), which measures the company’s corporate performance; 2) the price-to-book (PtB) ratio, which measures investment opportunities; 3) the firm size (Size); 4) the debt ratio (Leverage); and 5) beta (β), which measures the firm systematic risk. The purpose of this is to analyze whether there are differences in financial indicators, mainly in corporate performance, between firms that experienced positive abnormal returns and those that figured negative abnormal returns on announcement date.

^{Fama and French (2005)} argue that the profitability and growth opportunities of the firms are essential to evaluate their financing decisions. This fact justifies the use of ROA and PtB to characterize the financial situation of companies. Furthermore, the choice of financial performance indicators is based on the studies made by ^{Maquieira, Olavarrieta, and Zutta (2007)} and ^{Castañeda and Contreras (2017}) for Chilean context. ^{Maquieira et al. (2007)} analyzed the capital structure determinants. Their results show that the profitability had the sign predicted by the theory (the higher the profitability of the company, the lower the debt). On the other hand, ^{Castañeda and Contreras (2017)} analyze the debt maturity determinants. These authors conclude that the growth opportunities had the expected negative effect; and that the firm size had a positive impact. These studies justify the use of Size and Leverage. Finally, a widely used risk measure such as β is incorporated.

4.1. Event study results

Table 2 shows the average abnormal return (AR _t ) on days t=-1, t=0, and t=+1. In addition, Table 2 shows the accumulated average abnormal return between days t=0 and t=+1 (CAR _[0,1]) and between days t=-1 and t=+1 (CAR _[-1,1]) around the event. The AR_t=-1 of the day before the announcement (t=-1) is analyzed due to the possibility that the market anticipates the news or that it leaks. The AR _t=0 of the announcement date is published (t=0) measures the average effect of this event on the stock returns. The analysis also includes the AR _t=+1 of the day after the announcement (t=+1) due to the possibility that the Chilean stock market has some degree of inefficiency and stock prices react to the announcement with one-day delay. AR _t and CAR _t are calculated using the mean adjusted return, market model and market adjusted return models. Given the characteristics of these models, the most robust and relevant results correspond to the mean adjusted return and market model.

Table 2 shows that AR _t=0 is negative (between -0.09% and -0.03%), but it has no statistical significance. On the other hand, AR _t=+1 is positive and varies between 0.27% (market model) and 0.32% (mean adjusted return). In both models, AR _t=+1 is significant. Furthermore, CAR _[0;+1] and CAR _[-1;+1] are positive, but not statistically significant. The results present evidence in favor of hypothesis H₂, but they do not support hypothesis H₁.

4.2. Univariate model results

Table 3 (Panel A) shows the results of model (7) for firms that experienced positive abnormal returns at t=0. The average daily abnormal return of the days without events (α ₀ ) is negative and not statistically significant. The coefficients β₁, β₂ and β₃ show that the average daily abnormal return related to the issuance announcement is positive and significant: 0.83% at t=0; 0.43% between t=0 and t=+1; and 0.36% between t=-1 and t=+1.

Table 3 (Panel B) shows the results of model (7) for firms with negative abnormal returns at t=0. The average daily abnormal return of the days without events (α ₀ ) is positive and not statistically significant. On the other hand, the coefficients β₁, β₂ and β₃ show that the daily average abnormal return related to the issuance announcement is negative and significant: -1.21% at t=0; -0.49% between t=0 and t =+1; and -0.34% between t=-1 and t=+1.

4.3. Financial performance characterization

Table 4 shows the descriptive statistics of the financial indicators of the companies that announced corporate bond issuances. This analysis considers only non-financial firms, so the number of announcements decreases from 87 to 82 and the number of companies decreases from 29 to 27. The results did not vary when we included financial firms. Panel A shows the firm’s financial indicators that registered positive abnormal returns at t=0, according to the market model. The abnormal return calculated by means of the market model is used. The market model is used since it is a less restrictive and more robust model than the market adjusted model. Panel B shows the firm’s financial indicators that reported a negative abnormal return at t=0, according to the market model.

Panels A and B indicate large differences between the minimum and maximum values of the financial indicators. This shows the presence of outliers that can influence the mean of the indicators. For this reason, the analysis of financial indicators is conducted based on the median. Therefore, Panel B reports differences between the means and medians of the indicators.

On the median, the ROA (6.48%) of the companies with positive abnormal returns is higher than the ROA (5.96%) of the companies with negative abnormal returns. The size (19.21) of the firms with positive abnormal returns is smaller than the size (19.43) of the companies with negative figures. The indebtedness (0.66) related to the firms with positive abnormal returns is lower than the leverage (0.67) of the companies with negative abnormal returns. The investment opportunities (1.58) of companies with positive rates are lower than those related with firms with negatives abnormal returns (1.63). Finally, the systematic risk (1.00) is also lower for firms with positive abnormal returns.

However, according to the t-student test, the differences between the financial indicators of both samples of companies are not statistically significant. The results of the nonparametric Welch test, which measures the statistical significance of the differences between medians, are similar to those of the t-student test. Based on these results, there is no evidence that supports the hypothesis H₃. Consequently, there are no differences between the financial performance of companies that register positive abnormal returns, compared to those that present negative abnormal returns.