Discussions on investment and financing extend from the seminal work of Modigliani and Miller (1958), whose idea is that, assuming perfect markets, investments are independent of financial characteristics such as liquidity, leverage or dividend payments. Subsequently, the development of new theories and models led to the questioning of Modigliani and Miller's (1958) position. At the same time, market imperfections were considered as factors that could interfere with corporate decisions. These imperfections, or financial frictions, are linked to the existence of information asymmetries.

In the credit rationing literature regarding information asymmetries, Jaffee and Russell (1976) pointed out that credit rationing emerges as a market response to adverse selection, where lenders are unable to distinguish honest borrowers, who pay debts, and the dishonest, who do not pay debts whenever default costs are low. Whereas in Jaffee and Russell (1976) credit is rationed to the extent that borrowers receive less than desired loans, Stiglitz and Weiss (1981) explored the idea that not all businesses that need credit are able to do so.

Facing sustained approaches to informational asymmetry theories, Fazzari and Athey (1987) pointed out that, if the firm has sufficient operating cash flow to finance its investments, they may end up avoiding the foreign capital market, where credit rationing may exist. The connections between informational asymmetry, financial constraints, and investments, proposed by Fazzari and Athey (1987), were the precursor to the seminal contributions by Fazzari et al. (1988) in the field of financial constraints involving the financing of corporate investments.

Fazzari et al. (1988), when studying the influence of financial characteristics on corporate investments, argued that, unlike perfect markets, market imperfections exist, making external resources and internal business resources into imperfect substitutes. This idea arises from differences in the costs of such financing, which occur due to the presence of informational asymmetries. In other words, the costs of external financing are higher, i.e., the costs of issuing new debt and equity differ from the opportunity cost of financing generated by cash flows and retained earnings, as they involve, among other factors, transaction costs and asymmetric information.

Exploring the impacts of resource friction on capital structure decisions, Leary (2009) argued that the expected response to loan shocks depends on firms’ access to different segments of the capital markets. Thus, when faced with a contraction in lending, companies without access to public debt markets would need to find alternative sources to avoid capital constraints, including, for example, internal resources. This substitution would result in lower (higher) leverage following a contraction (expansion) in the loan offer.

In turn, Lemmon and Roberts (2010) analyzed how shocks in the supply of credit not only affect the financing of companies, but also investment. The evidence by Lemmon and Roberts (2010) pointed out that the substitution between bank debt and alternative sources of capital, such as cash balances, was limited in times of credit supply shock, almost a one-on-one decline in net investment with a decline in net debt issues.

In order to investigate and compare the investment plans, financial policies, and corporate spending of constrained and unconstrained firms during the 2008 crisis, Campello et al. (2010) interviewed CFOs from companies in various countries. The results indicated that constrained firms planned larger investment cuts compared to unconstrained firms, and the 2008 financial crisis did not affect the cash level of unconstrained firms. Further, the evidence is consistent with the view that constrained firms build cash reserves as a way to prepare against potential credit supply shocks.

Interested in the effects of credit supply during the 2008 crisis, Duchin et al. (2010) stressed that negative shocks in the supply of external financing, coupled with the presence of financial frictions, can hinder investments if firms have no financial slack, with these effects being particularly severe on firms that face higher costs to raise external resources. In this sense, when investigating the relationship between financial constraints, cash reserves, and investments, before and after the crisis, the results showed that investments fell for both firms with and without financial constraint after the crisis, although the decline was greater for constrained firms. Moreover, cash withholdings are cited as a precaution when credit is lower and companies are constrained.

Focusing on the impact of credit supply shocks during the 2008 crisis on UK private investment and financing, Akbar et al.(2013) explain that these companies are characterized by high information asymmetry and face higher external financing costs, and may worsen in times of economic downturn. Thus, these companies are naturally more constrained and tend to prefer, in times of slowdown, internal financing.

That said, the authors analyzed how private companies minimized the effects of bank credit contraction through alternative sources of financing such as domestic resources and issuance of shares. Evidence showed that, for these companies, short-term financing was most affected by the 2008 financial crisis, as they issued more shares and retained more cash as a precaution. In addition, the inability of private companies to obtain external credit caused them to cut their investments.

On the other hand, Driver and Muñoz-Bugarin (2019) investigated the effects of financial constraints on investments by UK companies, considering the size of the companies, seeking to reflect the effects of the 2008 crisis. The authors obtained evidence that the effects of credit constraints on large and small companies were different during the crisis, and for larger (less constrained), the disadvantages were greater. Larger firms’ sensitivity to financial constraints emerged during the crisis period, while, for smaller companies, the effect was perennial and did not increase during the crisis.

In Brazil, Oliveira and Cunha (2012) pointed out that companies with greater financial constraints rely more on their own resources than less constrained ones. Regarding the supply of resources during times of financial crisis, Paula et al. (2013) analyzed the evolution of the credit market in Brazil, emphasizing the role of public banks, especially that of Banco Nacional de Desenvolvimento Econômico e Social(BNDES), during periods of financial instability and credit contraction. The authors observed the expansion of companies’ productive capacity and investment decisions from mid-2007, increasing demand for corporate credit, served by the retail banking sector (working capital) and BNDES, with long-term financing.

Given the discussions that constrained firms investments are more affected by crises compared to unconstrained company investments (Campello et al.,2010; Duchin et al.,2010), and especially the constrained firms, which supposedly face greater difficulties in accessing external credit, rely more on cash to mitigate the effects of credit shocks and finance their investments (Duchin et al.,2010). We raise the following hypotheses to be tested in this study:

H₁: Financial crises impact the investment of constrained firms more strictly than unconstrained firms;

H₂: The relationship between investments and cash in crisis is more relevant for constrained firms than for unconstrained firms.

Concerning the impacts of crises on corporate financing, authors such as Leary (2009) and Akbar et al. (2013) mention that the leverage of narrower firms would be more negatively impacted by negative credit supply shocks. Moreover, Paula et al. (2013) argue that, in economically unstable periods, the trend is for banks to seek to reduce their risk by shortening the maturity of their loans and the reduction of longer-term credit supply, riskier than short-term credit supply, which would have a greater impact on short-term debt, especially from unconstrained firms.

That said, we raise the following hypotheses:

H₃: Financial crises have more significant negative effects on the leverage of constrained firms compared to unconstrained firms;

H₄: Crises cause an increase in the proportion of the short-term debt of constrained firms more sharply than that of unconstrained firms.

The data from this study was collected in the Economatica® software, quarterly, between the first quarter of 2007 and the third quarter of 2016.

Initially, companies with no required quarterly information were excluded, as well as companies whose capital stock was less than 10 million BRL, in 2007 values, at the beginning of the period, to eliminate very small firms, as we find in Duchin et al. (2010). The final overall investment model sample is composed of 203 publicly traded non-financial Brazilian companies, actively registered with the Securities Commission (CVM) in 2016. In turn, the general sample of the financing model is composed of 192 companies, respecting the same characteristics mentioned above.

To analyze the impacts of financial crises on investment and financing decisions of segregated Brazilian companies, according to their financial constraint, we use the criteria of the existence or non-existence of ratings for the separation of firms, whose advantage is evaluating the indication of the market, by an external agent, as for the credit quality.

Just as in Duchin et al.(2010), in this study the financial constraint of firms was determined based on observations from the years prior to the 2008 and 2015 crises, avoiding selection problems endogenous to company decisions. Therefore, for the classification of financial constraints for companies, we adopted information from 2007 (year immediately prior to the 2008 crisis) and 2013 (years immediately preceding the crisis of 2014/2015/2016).

The existence or otherwise of a rating was used by Lemmon and Roberts (2010) and Duchin et al. (2010). The argument is that companies without ratings by a risk assessment agency, such as Standard & Poor’s (S&P), Moody’s and Fitch, have a greater financial constraint than firms whose debt was valued, as the assessment indicates the credit quality of these companies, minimizing informational asymmetries.

For the classification of the companies, the rating data of the companies was collected from the “Rating of Listed and Closed Companies” 1994-2017” made available on Professor Tatiana Albanez’s Finance Research Portal. From this data, we listed companies that were rated by at least one of the risk agencies in both 2007 and 2013. Companies that necessarily had a rating in both 2007 and 2013, were classified as unconstrained. The other firms -both unrated companies in both years and companies rated in only one year - were classified as constrained.

It is worth mentioning that, in addition to the criterion of whether or not a rating was in play, additional analyzes were performed classifying the companies by their investment grade (non-investment grade and investment grade). This information is also found in the previously mentioned database. In this case, comparisons were made between investment grade firms and non-investment grade firms with no rating - the last two groups were analyzed together. The results using both criteria led to the same conclusions. Additionally, we made estimates by fixed double effect panel, also leading to the same conclusions. Supplementary material containing the robustness tests mentioned can be found in the Figshare scientific repository or can be provided by the authors.

The dependent variable for the study of the impact of crises on investments is represented by the following indicator, based on Duchin et al. (2010):

Investt = CAPEXt/TAt

where Invest: Investment; CAPEX: Capital expenditures, involving net acquisitions of property, plant and equipment, quarterly; TA: Total Assets.

The dependent variables for the study of the impact of crises on financing are as follows:

Levert = Total Debtst/TAtDebtSTt = Debts STt/Total Debtst

where Lever: Leverage; Total Debt: sum between short and long term debt; TA: Total Assets; DebtST: Short Term Debts. The leverage variable was built based on Leary (2009), Lemmon and Roberts (2010) and Akbar et al. (2013).

The main independent variables in this paper are the crises, specifically the crises of 2008 and 2015. Crisis variables are represented by dummies in the models, with a value of 1 for quarters during crisis (between the fourth quarter of 2008 and the second quarter of 2009 and between the second quarter of 2014 and the third quarter of 2016) and zero for the other quarters between 2007 and 2016.

a) Investments

In the literature studying the impacts of crises on corporate investments, there is evidence that crises adversely affect corporate investments. However, these effects may differ between companies: impacts may be stronger on constrained firms, as the credit constraint for these firms is further intensified in periods of credit contraction (Campello et al., 2010; Duchin et al., 2010).

Still, in the financial constraint and investment literature, it is common to include the variable Cash, representing the internal resources to the companies. In this sense, we added this variable as an independent variable in the investment model, calculated as the ratio of cash to total assets over a period. For constrained firms, the relationship between investments and cash in the crisis is expected to be positive and stronger compared to unconstrained firms. The argument is that, for constrained firms, cash would mitigate the effects of the credit supply shock, linked to the impact on investments.

b) Financing

Regarding impacts on leverage, there is evidence that crises negatively impact the leverage of constrained firms, while unconstrained firms are not as impacted (Leary, 2009; Akbar et al., 2013). The argument is that access to resources by firms with greater financial constraints in times of negative credit supply shocks becomes even more difficult.

Regarding the impact of crises on the participation of short and long term debt in relation to total debt, there are ambiguous relationships pointed out by the literature: positive, with the argument that banks seek to mitigate risk and reduce the maturity of their loans and financing, thus increasing the proportion of short-term debt (Paula et al., 2013), or negative, on the grounds that the short-term debt tends to be reduced in times of crisis, impacted by the drop in bank credit supply, mainly from constrained firms, whose debts are mostly from bank resources due to limited access to alternative sources (Akbar et al., 2013; Bremus, 2015).

a) Investments

In the investment model, the proxy used to control investment opportunities (or growth opportunities) is defined by the Market-to-Book variable (asset at market value by asset at book value). We expect a positive relationship between investment opportunities and investments, so the greater the investment opportunities, the greater the investments for both constrained firms and unconstrained firms.

b) Financing

In the financing model, as well as in Leary (2009), we use control variables commonly used in capital structure literature in order to control the companies’ demand for financing. These variables were added according to arguments by Rajan and Zingales (1995) and Frank and Goyal (2009), with variables being profitability, tangibility, size and growth opportunities.

The profitability variable was calculated by the ratio between EBITDA and total assets, and the expected ratio is ambiguous: a negative relationship indicates corporate preference for domestic financing over debt and a positive relationship indicates that resource providers would prefer to lend to more profitable companies.

The tangibility variable was calculated as the ratio of fixed assets to total assets, and the ambiguous relationship between tangibility and leverage was expected: a positive relationship indicates the role of collateral in minimizing agency costs and a negative one indicates lower share issuance costs.

Size, in turn, was calculated based on the natural logarithm of corporate assets, expecting an ambiguous relationship between size and leverage: positive relationship indicates that larger companies tend to be more diverse and less likely to go bankrupt, and negative indicates investor preference for stocks due to access to information from larger companies.

Finally, as in the investment model, growth opportunities were measured by the market-to-book index, with a negative relationship being expected, as companies with future growth prospects would use a larger share of equity financing.

Table 1 shows the construction of investment and financing model variables:

Table 1.

Investments
Variables	Acronym	Proxy
Dependent Variables
Investment	Invest	CAPEX/Total Assets
Independent Variables
Crisis	Crisis 2008 Crisis 2015	Dummy 1 for Crisis and 0 for other periods
Cash	Cash	Cash and cash equivalent/Total Assets
Crisis*Cash	Crisis*Cash	Crisis*(Cash and cash equivalent/Total Assets)
Control Variable
Investment Opportunities	M/B	Assets at Market Value/Assets at book value
Financing
Variables	Acronym	Proxy
Dependent Variables
Leverage	Lever	Total Debts /Total Assets
Short Term Debts	DebtST	Debt ST/Total Debt
Independent Variables
Crisis	Crisis 2008 Crisis 2015	Dummy 1 for Crisis and 0 for other periods
Control Variable
Profitability	Profit.	EBITDA/Total Assets
Tangibility	Tangib	Fixed assets /Total Assets
Size	Size	Ln Assets
Growth Opportunities	M/B	Assets at Market Value/Assets at book value

The general forms of Investment models (Invest) and Financing (Fin) to be studied are presented below for each company i in each quarter t:

Investi,t=fCrisis+Crisis*Cashi,t-1+M/Bit

it+M/BitCrisis+Profitabilityit+Tangibilityit+Fini,t=f

The models of the study of the impacts of crises on corporate investments are as follows:

Invest=Crisis+Crisis*Rest+Cash+Cash*Rest+M/B (1)

Invest=Crisis2008+Crisis2008*Rest+Crisis2015+Crisis2015*Rest+Crisis2008*Rest*Cash+Crisis2015*Rest*Cash+M/B (2)

where Invest represents the Investments, Crisis is the crisis dummy, Crisis*Cons, Crisis2008*Cons and Crisis2015*Cons represent dummies from the 2008 and 2015 crisis for the constrained firms dummy. The variables of each crisis and the financial constraint dummy were interacted with the cash (in t-1). M/B represents the investment opportunities in the period t. As for the study of models of financing of firms on crisis impacts are as follows:

Lever=Crisis+Crisis*Rest+Profit+Tangib+B (3)

Lever=Crisis2008+Crisis2008*Rest+Crisis2015+Crisis2015*Rest+Profit+Tangib+B (4)

DebtST=Crisis+Crisis*Rest+Profit+Tangib+B (5)

DebtST=Crisis2008+Crisis2008*Rest+Crisis2015+Crisis2015*Rest+Profit+Tangib+B (6)

where Lever represents leverage, DebtST represents the proportion of short-term debt, Crisis is the crisis dummy, Crisis*Cons, Crisis2008*Cons and Crisis2015*Cons represent dummies from the 2008 and 2015 crisis for the constrained firms dummy, Profit is profitability, Tangib is tangibility, Size is size and M/B represents growth opportunities (all variables in period t).

We adopted descriptive analysis and panel data analysis as our methodological tools. In the models, dummies were included for the constrained group of firms, and unconstrained firms were taken as the basis for the regression model, as adopted by Aldrighi and Bisinha (2010). In other words, the companies included in the regression showed their differential in relation to unconstrained firms, which are those that were unconstrained in the 2008 crisis that remained unconstrained in the 2015 crisis.

To ensure more consistent estimators, this study used the fixed effects estimation method, just as it was in Duchin et al.(2010), with estimations with robust standard errors grouped by firm being made.

Table 5 presents the statistical results of the impacts of crises on investments.

Table 5.

Dependent Variable: Invest = CAPEX/Total Assets
	Equation (1)		Equation (2)
	Coef.	p-val FE, RC	Coef.	p-val FE, RC
Crisis	-0.002	0.103
Crisis*Cons	-0.001	0.472
Crisis 2008			0.002	0.593
Crisis 2008*Cons			-0.002	0.581
Crisis 2015			-0.004	0.034
Crisis 2015*Cons			-0.005	0.064
Cash	0.019	0.650
Cash*Cons	0.045	0.34
Crisis2008*Cons*Cash			0.014	0.242
Crisis2015*Cons*Cash			0.036	0.005
M/B	0.003	0.128	0.004	0.083
Constant	0.009	0.002	0.014	0.000
Nº observations		6097		6097
Nº companies		203		203
Prob > F		0.000		0.000
R² within		0.024		0.013
R² between		0.013		0.041
R² overall		0.015		0.017
Estimation	Fixed Effects		Fixed Effects
Breusch and Pagan
Chi2(1):		280.91		282.42
Prob > Chi2:		0.000		0.000
Chow
Statistic F		3.11		2.96
Prob > F		0.000		0.000
Hausman
Chi2(5):		37.55		12.15
Prob > Chi2:		0.000		0.096

Notes. Invest: CAPEX divided by total assets; crisis: 2008 and 2015 crisis dummies; cons: constrained firms; crisis*cons: interaction between crises and constrained firms; crisis 2008: dummy 1 between the fourth quarter of 2008 and the second quarter of 2009, and zero otherwise; crisis 2015: dummy 1 between the second quarter of 2014 and the third quarter of 2016, and zero otherwise; crisis2008*cons and crisis2015*cons: interaction between the crises of 2008 and 2015 and constrained firms; cash: cash and cash equivalents divided by total assets; cash*cons: interaction between cash and constrained firms; crisis2008*cons*cash and crisis2015*cons*cash: interaction between crises of 2008 and 2015, constrained firms and cash; M/B: market-to-book- asset at market value divided by asset at book value; coef.: coefficients; p-val FE, RC: coefficient significance level for fixed effects regression with clustered robust standard errors; nº observations: number of observations; Prob > F: significance level of the model; Breusch and Pagan: p-value of the test LM of Breusch-Pagan; Chow: p-value of the test F Chow; Hausman: p-value of the Hausman’s test; R² within: coefficient of explanation of the effects of variation over time for a given individual; R² between: coefficient of explanation of the effects of variation between individuals; R² overall: coefficient of general explanation of the model.

From the statistical results, we noted that the impact of the 2008 crisis on constrained firms was not statistically different from the impact on unconstrained firms, possibly due to the fact that it has its initiation in a country other than Brazil, and the companies were investing relatively more at that time, as shown in table 4.

The crisis of 2015 negatively and significantly affected the investments of companies in general (as can be seen the coefficients of the variable Crisis 2015), with the negative impact being even greater for constrained firms compared to unconstrained firms (reference group in regression). Major declines in the investments of constrained firms in times of crisis are in line with evidence in the literature (Campello et al., 2010; Duchin et al., 2010), as access to resources by these companies becomes even more difficult.

Considering the importance of cash in mitigating the effects of crises on corporate investments, the relationship between cash and investments was significant, positive, and only statistically higher for constrained firms compared to unconstrained firms in the 2015 crisis. This result is consistent with Duchinet al. (2010), who got evidence that the relationship between cash and investment in the crisis is stronger for constrained firms, following the precautionary idea of cash, since these companies would have access to external resources made difficult especially in times of crisis and would rely more on their internal resources.

Thus, the results allow us to conclude that, at the same time as the 2015 crisis caused an observable fall in investments of constrained firms, the investments from these companies were more sensitive to cash at that time because of the greater difficulty in accessing external resources in crisis.

Table 6 presents the results of the impact of crises on corporate leverage:

Table 6.

Dependent Variable: Lever = Total Debt/Total Asset
	Equation (3)		Equation (4)
	Coef.	p-val FE, RC	Coef.	p-val FE, RC
Crisis	0.006	0.566
Crisis*Cons	0.008	0.478
Crisis 2008			0.011	0.337
Crisis 2008*Cons			-0.026	0.078
Crisis 2015			0.006	0.644
Crisis 2015*Cons			0.015	0.299
Profit	-0.139	0.031	-0.134	0.031
Tangib	-0.015	0.615	-0.012	0.701
Size	0.030	0.009	0.026	0.026
M/B	-0.047	0.000	-0.048	0.000
Constant	-0.107	0.532	-0.035	0.836
Nº of observations		5726		5726
Nº of companies		192		192
Prob > F		0.000		0.000
R² within		0.140		0.146
R² between		0.040		0.041
R² overall		0.061		0.064
Estimation	Fixed Effects		Fixed Effects
Breusch and Pagan
Chi2(1):		44762.19		44974.68
Prob > Chi2:		0.000		0.000
Chow
F Statistics		96.17		96.77
Prob > F		0.000		0.000
Hausman
Chi2(5):		50.56		14.67
Prob > Chi2:		0.000		0.066

Notes. Lever: leverage, - total debt divided by total assets; crisis: 2008 and 2015 crisis dummies; cons: constrained firms; crisis*cons: interaction between crises and restricted companies; crisis 2008: dummy 1 between the fourth quarter of 2008 and the second quarter of 2009, and zero otherwise; crisis 2015: dummy 1 between the second quarter of 2014 and the third quarter of 2016, and zero otherwise; crisis2008*cons and crisis2015*cons: interaction between the crises of 2008 and 2015 and constrained firms; profit: profitability- relationship between EBITDA and total assets; tangib: tangibility - relationship between fixed assets and total assets; size: size - natural log of total assets; M/B: market-to-book- asset at market value divided by asset at book value; coef.: coefficients; p-val FE, RC: coefficient significance level for fixed effects regression with clustered robust standard errors; nº of observations: number of observations; Prob > F: significance level of the model; Breusch and Pagan: p-value of the LM testing by Breusch-Pagan; Chow: p-value if the Chow F test; Hausman: p-value of Hausman’s test; R² within: coefficient of explanation of the effects of variation over time for a given individual; R² between: coefficient of explanation of the effects of variation between individuals; R² overall: coefficient of general explanation of the model.

The results showed that the 2008 crisis had a more negative and significant impact on the leverage of constrained firms in relation to unconstrained firms, in line with Leary (2009) and Akbar et al.(2013). The impacts of the 2015 crisis on leverage were not significantly different between constrained and unconstrained firms.

The non-significant impacts of the 2015 crisis on corporate leverage raise some possibilities for interpretation. On the one hand, as shown in Figure 1, the supply of credit to Brazilian companies was considerably reduced in the 2015 crisis, which, following studies like Leary (2009), would lead to an expected negative relationship between crises and leverage. On the other hand, at least two particularities could counteract this expected negative relationship with Brazilian companies: the countercyclical performance of BNDES and the increase in foreign currency debt.

As observed by Sant'anna, Junior and Araujo (2009), in the crisis of 2008 the countercyclical performance of public banks, especially BNDES, offset the slowdown in private bank credit operations, avoiding further reductions in the volume of credit granted to firms. Additionally, according to Oliveira and Cunha (2012), access to BNDES financing is easier for less financially constrained firms, which would explain the outcome of this study for the 2008 crisis.

Furthermore, as disclosed by CEMEC (2016), the indebtedness of publicly traded Brazilian companies increased between 2010 and 2016, evidencing the increase in the participation of foreign currency debts in the composition of total corporate indebtedness. In this context, in times of crisis, whose exchange rate tends to increase, it is expected that the balance of foreign currency debt will become even larger. Therefore, it is possible to hypothesize that the absence of significant effects of crises on corporate leverage can be explained by the combination of compensatory factors, such as the drop in credit supply on the one hand and, on the other, the effects of countercyclical performance of public banks and the increase in foreign currency debt.

Analyzing the control variables profitability, tangibility, size, and market-to-book, we observed a negative sign for the profitability and market-to-book variables and a positive sign for the size variable, with tangibility being the only non-significant variable to explain firms’ leverage.

On the impacts of crises on the proportion of short-term debt, Table 7 shows their results:

Table 7.

Dependent Variable: DebtST - Debt ST /Total Debt
	Equation (5)		Equation (6)
	Coef.	p-val FE, RC	Coef.	p-val FE, RC
Crisis	0.029	0.007
Crisis*Cons	0.000	0.983
Crisis 2008			-0.005	0.776
Crisis 2008*Cons			0.055	0.019
Crisis 2015			0.038	0.010
Crisis 2015*Cons			-0.014	0.504
Profit	-0.124	0.070	-0.125	0.067
Tangib	-0.072	0.071	-0.067	0.094
Size	-0.063	0.000	-0.063	0.000
M/B	0.006	0.649	0.006	0.641
Constant	1.333	0.000	1.332	0.000
Nº of observations		5726		5726
Nº of firms		192		192
Prob > F		0.000		0.000
R² within		0.029		0.032
R² between		0.439		0.436
R² overall		0.301		0.302
Estimation	Fixed Effects		Fixed Effects
Breusch and Pagan
Chi2(1):		20372.51		20424.44
Prob > Chi2:		0.000		0.000
Chow
F statistics		32.31		32.26
Prob > F		0.000		0.000
Hausman
Chi2(5):		24.00		24.95
Prob > Chi2:		0.001		0.0016

Notes. DebtST: short term debt - short-term debt divided by total debt; crisis: 2008 and 2015 crisis dummies; cons: constrained firms; crisis*cons: interaction between crises and constrained firms; crisis 2008: dummy 1 between the fourth quarter of 2008 and the second quarter of 2009, and zero otherwise; crisis 2015: dummy 1 between the second quarter of 2014 and the third quarter of 2016, and zero otherwise; crisis2008*cons and crisis2015*cons: interaction between 2008 and 2015 crises and constraiend firms; profit: profitability - relationship between EBITDA and total assets; tangib: tangibility - relationship between fixed assets and total assets; size: size- natural log of total assets; M / B: market-to-book - asset at market value divided by asset at book value; coef.: coefficients; p-val FE, RC: coefficient significance level for fixed effects regression with clustered robust standard errors; nº of observations: number of observations; Prob > F: significance level of the model; Breusch and Pagan: p-value of the LM testing by Breusch-Pagan; Chow: p-value if the Chow F test; Hausman: p-value of Hausman’s test; R² within: coefficient of explanation of the effects of variation over time for a given individual; R² between: coefficient of explanation of the effects of variation between individuals; R² overall: coefficient of general explanation of the model.

The results presented in Table 7 show that the 2008 crisis had a more significant positive impact on the proportion of short-term debt of constrained firms compared to unconstrained firms, corroborating the idea by Paula et al. (2013) that, in times of crisis, banks seek to reduce the maturity of corporate lending to reduce risk, such that if the debts of restricted companies come mainly from bank resources, given the greater difficulty of access to alternative sources - as mentioned by Leary (2009) and Akbar et al.(2013) - the sensitivity of these firms’ short-term debts to crises marked by credit supply shocks would be greater.

Moreover, the 2015 crisis impacted the proportion of short-term debt held by both constrained and unconstrained firms, not being statistically different between both groups of firms.

It is worth drawing attention to the interpretations of the 2008 crisis results analysis. At that moment, facing the process of convergence of national and international accounting standards brought with the Law nº 11.638, 2007, the transition toward which began to occur in 2008, certain adjustments are now reflected in the financial statements of that year. As accounting parameters are used in this study to construct the variables used in the analyzes, there may be confounding factors impacting the estimates.