Credit supply responses to reserve requirement: loan-level evidence from macroprudential policy

BIS Working Papers

No 674


Credit supply responses to reserve requirement: loan-level evidence from macroprudential policy

by João Barata R. B. Barroso, Rodrigo Barbone Gonzalez and Bernardus F. Nazar Van Doornik


Monetary and Economic Department

November 2017


Paper produced as part of the BIS Consultative Council for the Americas (CCA) research project on “The impact of macroprudential policies: an empirical analysis using credit registry data” implemented by a Working Group of the CCA Consultative Group of Directors of Financial Stability (CGDFS).

JEL classification: E51, E52, E58, G21, G28.

Keywords: Reserve requirement, credit supply, capital ratio, liquidity ratio, macroprudentialpolicy


BIS Working Papers are written by members of the Monetary and Economic Department of the Bank for International Settlements, and from time to time by other economists, and are published by the Bank. The papers are on subjects of topical interest and are technical in character. The views expressed in them are those of their authors and not necessarily the views of the BIS.

This publication is available on the BIS website (


© Bank for International Settlements 2017. All rights reserved. Brief excerpts may be reproduced or translated provided the source is stated.

ISSN 1020-0959 (print)

ISSN 1682-7678 (online)


Credit Supply Responses to Reserve Requirement: Loan-level evidence from macroprudential policy

 João Barata R. B. Barroso Rodrigo Barbone Gonzalez Bernardus F. Nazar Van Doornik



This paper estimates the impact of reserve requirements (RR) on credit supply in Brazil, exploring a large loan-level dataset. We use a difference-in-difference strategy, first in a long panel, then in a cross-section. In the first case, we estimate the average effect on credit supply of several changes in RR from 2008 to 2015 using a macroprudential policy index. In the second, we use the bank-specific regulatory change to estimate credit supply responses from (1) a countercyclical easing policy implemented to alleviate a credit crunch in the aftermath of the 2008 global crisis; and (2) from its related tightening. We find evidence of a lending channel where more liquid banks mitigate RR policy. Exploring the two phases of countercyclical policy, we find that the easing impacted the lending channel on average two times more than the tightening. Foreign and small banks mitigate these effects. Finally, banks are prone to lend less to riskier firms.

JEL Codes: E51, E52, E58, G21, G28.

Keywords: Reserve requirement, credit supply, capital ratio, liquidity ratio, macroprudential policy.

Paper produced as part of the BIS Consultative Council for the Americas (CCA) research project on "The impact of macroprudential policies: an empirical analysis using credit registry data" implemented by a Working Group of the CCA Consultative Group of Directors of Financial Stability (CGDFS). We are grateful to seminar participants at the 2016 BIS Meeting in Mexico, and at the XI Seminário de Riscos e Estabilidade Financeira in Sao Paulo (Brazil) for many helpful comments and suggestions. Additionally, we would like to thank the Department of Banking Operations and Payment System of the Central Bank of Brazil for providing us with data and permanent support on reserve requirements. We also appreciate the help of Carlos Leonardo Kulnig Cinelli. The views expressed in this paper are those of the authors and not necessarily reflect those of the Central Bank of Brazil or the Bank for International Settlements.

1. Introduction

Reserve requirements (RR) operate directly on the narrow credit channel defined by the supply reaction of bank credit to a change in funding composition (Calomiris and Khan, 1991, Stein, 1998, Diamond and Rajan, 2011, Calomiris et al., 2015). This reaction may depend on the state of the macroeconomy, and on bank characteristics, such as liquidity or capital (Kashyap and Stein (2000), Holmstrom and Tirole (1997), Mora). It has also implications for the composition of credit along the riskiness of the borrowers (Camors et al. (2016)). In this paper, we estimate the impact of RR on credit supply in Brazil.

Quantitative estimates of the effect of RR in the supply of credit, as well as its complementarity or substitution relations with other variables, are important for emerging markets that traditionally use RR policy to smooth the credit cycle (Montoro and Moreno (2011), Cordella et al. (2014)). Yet, with the exception of Camors et al. (2016), there is no loan-level evidence of the impact of such policies in these markets. We build on their work, but exploring a larger and longer dataset with both tightening and loosening cycles. Additionally, we provide our analysis using a long panel to capture macroeconomic effects along with bank heterogeneity effects.

We use quarterly data from 2008Q1 to 2015Q2 from “Sistema de Informaçoes de Crédito” (SCR), Central Bank of Brazil (BCB) credit registry dataset covering virtually all loans to private non-financial firms1. During this time span, BCB made several macroprudential interventions using RR including a major countercyclical one in the aftermath of the global crisis. The intervention consisted of: (1) an easing, i.e. releasing RR in November 2008 in response to a credit crunch following the global financial crisis; and (2) a tightening, i.e. reversing the easing policy on March 2010, when credit growth was overheated.

BCB made other interventions though. For instance, a tightening in December 2010, in the context of high capital inflows and credit growth2; and several easing innovations starting with the reversal of this policy in 2012, but also along 2013 and to 2015 during an economic downturn. Before and after this period, RR ratios were mostly flat and revolving around the long-term average of 23% of liabilities subject to reserve requirements (LRR).

The measurement of reserve requirement innovation and sample selection is a central piece in the identification strategy. We evaluate two different approaches. In the first approach, we build an index, adding or subtracting one unit upon tightening or easing of RR policy, respectively, interact this with a treatment indicator for banks with non­zero reservable liabilities, and use a long panel with controls for macroeconomic confounding factors. Notice that the identification comes from both variation in the index and bank treatment status, and is improved with the inclusion of macro controls. In the second approach, we define bank level treatment variables based on RR counterfactuals. Specifically, we define a bank-level treatment as the excess variation in RR over the counterfactual variation one would observe in RR under the old regulation. Notice that the counterfactual filters out determinants of reserve requirement other than the regulatory changes. The counterfactuals are separately calculated to capture the regulatory changes of November 2008 (easing - following “bad times”), and of March 2010 (tightening - following “good times”). Relative to the first approach, our identification benefits from the continuous nature of the treatment variable and the filtering of other determinants of reserve requirements directly in the treatment variable..

We identify the complementarity or substitution relations with RR policy by introducing interaction terms in our models. We explore interactions with bank control variables such as size, liquidity, capital ratio and risk proxies.

Following Khwaja and Mian (2008) and Jimenez et al. (2014)3, we focus on firms with multiple bank relationships and firm (or firm*bank) fixed effects to partially control for credit demand. In order to explore interactions of the treatment variable with firm or firm-bank characteristics such as credit risk of a particular firm, we also include bank fixed effects.

This paper contributes to the scarce literature estimating the effects of changes in RR on credit supply. It also addresses synergies between macroprudential and bank and firm heterogeneity, covering a very large dataset of firm loans. The dynamics of the Brazilian case allow the study of both macroprudential loosening and tightening separately.

We find in the long panel that RR policy impacts credit in the expected direction, which is RR easing increases credit, while RR tightening decreases credit on the treated banks relatively to the non-treated banks. The exact quantitative impact depends on the specification, and it is sensibly higher in the long-run (one-year ahead cumulative effect) than in the short-run (one-quarter ahead). In the cross-section approach, we find that the tightening phase of countercyclical RR events affected the lending channel on average less than the easing one, suggesting that bank credit supply is more reactive to the easing than to the tightening.

We also find bank and firm heterogeneity in the composition of the events of interest. Foreign and small banks mitigate the policy effects. On the risk-taking channel, we find that banks more affected by countercyclical RR policy avoid riskier firms. These results are of great concern to policymakers in charge of financial stability, because riskier firms are the ones more affected by credit crunches and more prone to leverage during credit booms.

1 Up to December 2011 it covered all loans greater than BRL 5,000 (USD 3,000 in 2011), and, after that, all loans greater than BRL 1,000 (USD 425 in 2014).

2 See Barroso (2015) for evidence on the link between capital inflows and credit growth.

3 In contrast with Jiménez el al (2014), we can study the risk-taking channel without the triple interaction proposed in that paper. That is, the capital ratio is not a source of identification.

2. Literature review

The rationale for reserve requirements effects on credit supply follows Stein (1998), and Kashyap and Stein (2000). They explore imperfect substitution between insured and reservable bank liabilities on one side, and noninsured and non-reservable bank liabilities on the other. The risk-taking channel on macroprudential policy follows mostly Adrian and Shin (2009) and Dell’Ariccia et al. (2009). They show that changing the cost of liabilities affects banks’ leverage and therefore the incentives for banks to monitor. The interaction with banks’ liquidity and capital are presented in Kashyap and Stein (2000), and Holmstrom and Tirole (1997), respectively.

Tovar et al. (2012), Montoro and Moreno (2011), and Bustamante and Hamman highlight the use of reserve requirements with macroprudential purposes, especially to foster financial stability. First, it can serve as a countercyclical tool to manage the credit cycle in a broad context, limiting the excessive leverage of borrowers in the upswing and operating as a liquidity buffer in the downswing. Second, it can help to contain systemic risk accumulation by improving the liquidity of the banking system. Third, RR can target specific sectors to ease (or impose) liquidity constraints. Fourth, it can be a complementary tool for capital requirements.

Cerutti et al. (2015) document that macroprudential policies are more effective and used more broadly in less developed and more closed economies, with effectiveness measured by the correlation with credit aggregates. Cordella et al. (2014) argue that developing countries use reserve requirements for stabilizing capital flows and the credit cycle when there are severe limits on the typical monetary policy ability to smooth the level of credit and/or economic activity. According to these authors, the financial stability and business cycle-driven uses of reserve requirements cannot be separated one from the other. When reserve requirements are used to prevent financial instability, they can contribute to macroeconomic stabilization, whereas when they are used to smooth activity, they also smooth the credit cycle and promote financial stability.

There is a growing empirical literature exploring the risk-taking channel of monetary policy. Jimenez et al (2014) find that banks extend more credit to riskier firms during monetary policy easing cycles. Altumbas (2012) show solvency problems during the crisis were more severe for banks in jurisdictions with low interest rates for a long time and for banks with less capital. Maddaloni and Peydro (2011) show a deterioration in lending standards across several jurisdictions in response to lower short­term interest rates. Lee (2015) use syndicated loan data to show that, before the crisis, lenders invest in riskier loans in response to a decline in short-term US rates while, after it, in response to a decline in long-term US interest rates.

In passing it through, the effect of typical monetary policy on credit supply and risk taking could be, in theory, similar to reserve requirements, although operating through other channels. Cerutti et al. (2012) document with macro data that RR affects credit growth, but has no implications for risk-taking. However, recent loan-level evidence from Camors et al. (2016) and Jimenes et al. (2017) is the opposite of what one would expect from the analogy with monetary policy. While suggesting a similar bank lending channel, these authors find a positive risk-taking channel. In other words, they find a “search-for-yield” or positive risk-taking response to the tightening of RR and countercyclical dynamic provisions respectively.

Camors et al. (2016) is the closest paper to ours in the literature. Using loan level data and an identification strategy equal to ours, they show that an increase in RR in Uruguay implies a contraction of credit supply. However, the macroprudential tightening event they explore is different in nature. While we explore countercyclical RR events motivated by a credit crunch and later by a credit boom, the tightening RR they explore is motivated by intense foreign cash inflows (bypassing monetary policy). Their results for the lending channel is of similar magnitude to ours, i.e. a RR increase of 1 percentage point (pp) translates into a credit supply contraction of 0.66% for the most affected bank relatively to the same firm. The authors also find that the most affected banks mitigate this tightening contracting less credit to the riskier firms. We find a similar risk-taking channel to the easing, but not to the tightening.

3. Background

The ratio of reserve requirement to deposits in Brazil is large by international standards. It averages 23% of total liabilities subject to RR (LRR) from 2008 to 2015, while Montoro and Moreno (2015) report emerging market ratios below 15% and developed market ratios below 5%. The ratio in Brazil is mostly flat before the global financial crisis. During the crisis, in face of a liquidity squeeze in the interbank and credit market, BCB reduces RR to the historical low levels of 18% in November 2008. In March 2010, RR is rebuilt to its prior levels, in the first countercyclical policy use of this kind. The easing policy was highly relevant with an immediate release of cash into the financial system worth 3.27% of total banks" assets (or 15% of banks’ liquid assets).

In response to an increase in capital flows and high credit growth, a major tightening cycle starts in December 2010. Relative to other local macroprudential policies implemented during the same period, RR is arguably the macroprudential tool with broadest scope and biggest impact4. Along 2012, with growing external uncertainties, decrease of international capital flows and reduced credit supply from private banks, RR is eased again to pre-crisis levels (this latest tightening cycle is complete). See Figures 1 and 2.

BCB manages mainly four RR components; RR on demand deposits (unremunerated), savings (remunerated according to savings accounts), time and term deposits (remunerated at the overnight funds rate, SELIC), and an additional component comprised of three subcomponents, one for each of the previous components, (all remunerated at the daily prime rate, SELIC). BCB also manages RR deductibles, conditional deductibles, exemption thresholds, eligible liabilities and remuneration. The details of the regulatory changes in the period considered in the paper are complex. We only summarize the most relevant measures in the following subsections and present more details in Chart 1.

Main measures

The global financial crisis led to a liquidity squeeze that affected mostly small financial institutions. Moreover, banks’ risk aversion (stemming from both bigger and smaller institutions) substantially affected domestic credit growth. In response, BCB eased reserve requirements, increased deductions, and created conditional deductibles to stimulate larger banks to provide liquidity support to small and medium-sized ones.

It is worth noticing a “fly-to-quality” movement, with depositors from smaller banks running to bigger ones (perceived as safer). Smaller financial institutions were mostly weaved from RR because of a minimum capital threshold to start computing LRR. Consequently, RR easing mostly affected bigger banks (also more representative in term of credit provided to firms), because smaller institutions use the cash release to recompose liquidity (Schiozer et al., 2016, Oliveira et al., 2015).

Around 75% of the bank institutions are unaffected by RR, the remaining ones receive smaller or larger shocks pending on their ex-ante exposure to the more affected liabilities. Figure 3 illustrates the average impact on these two groups (5%).

The countercyclical measures adopted in November 20085 are the following:

  1.  Reduction in RR ratios for demand deposits, term deposits and the additional component;
  2.  Higher deductions, lower remuneration and changes in eligible liabilities for time and term deposits and in the additional component that released some small banks from RR and reduced significantly RR on big banks.
  3.  Conditional deductibles on certain exposures (from mostly big banks) to small-and-medium sized financial institutions.

Measure “(i)” releases close to BRL 26 billion and the two remaining ones combined, BRL 40 billion. In March 2010, BCB reverses the policy adopted during the crisis6 (Figure 4)

Counterfactual RR

The Central Bank of Brazil routinely computes counterfactual RR to monitor the implementation of its policies. In light of these constant changes in RR, comparing current and counterfactual RR is useful to summarize these changes in one figure. The counterfactual is straightforward to calculate. The liabilities subject to RR (LRR) are the same7, but RR ratios, deductibles, conditional deductions and exemptions are calculated for every bank based on the pre-changes’ rule.

In this paper, we take the pre-crisis state counterfactual for November 2008. In particular, the counterfactual rules available until October 2008 were:

  • 15%    on term deposits;
  • 45%     for demand deposits;
  • 20%      for savings deposits;
  • In the additional components, (8% on demand and term deposits; and 10% on savings).

In the cross-section strategy, we compute the difference between the counterfactual RR and the current new rules for each bank as a treatment variable to study the event of November 2008. Similarly, we also build the counterfactual to capture the event of March 2010.

4 During the post-crisis environment of large global liquidity, the Central Bank of Brazil issued many with­in sector regulations focusing on financial stability, such as loan-to-value caps on housing loans (Araujo et al., 2016) and higher capital requirements on auto-loans (Martins and Schetchman, 2013). However, RR is arguably the more representative measure. See Pereira da Silva and Harris (2012).

5 Two announcements are worth mentioning. The first announcement happens at the end of October, and the most relevant one at the beginning of November, where banks had only 15 days to comply.

6 In March 2010, BCB also creates a deductible on Term Deposits and on the Additional component conditional on the capital of banks, virtually exempting small institutions from RR (Circular 3,485/2010).

7 Eligible liabilities changed in 2010 for six months and comprehend the inclusion of a bond called “letra financeira” with maturities over 2 years in the eligibility list. Tracing these effects is a limitation of this study. Other changes are also untraceable. For instance, changes in remuneration of RR components (Chart 1).

8 Ratings go from “AA” (highest quality) to “H” (lowest quality), and provisioning increases nonlinearly with each step. Measured as the required provision, the ratings relate on average to expected losses and from “AA” to “H” are 0.005, 0.01, 0.03, 0.1, 0.3, 0.5, 0.7 and 1, respectively. There is a close correspondence between such provisions and the following scale of days overdue, 0, 15-30, 31-60, 61-90, 91-120, 121-150, 151-180, >180.

4. Data and Methodology

The main dataset of the paper is the Brazilian Credit Register (SCR), which encompasses virtually all corporate loans in the domestic financial system. Data is quarterly from 2008Q1 to 2015Q2. The dependent variable of interest is the log change in the credit granted to a firm (f), by a bank (b) in a quarter (t), winsorized at the 2nd and 98th percentile. We restrict our sample to firms with loans granted from more than one
bank. This sample has over 36 million data points (27 periods, 132 banks and 478 thousand firms). See Tables 1 and 2 for summary data and variables" definition.

The firm risk indicator is the loan level provision to non-performing loans (PNL) weighted across all banks to which the firm has a credit exposure8 (Firm Risk), or simply the PNL given by the bank to a particular firm (Firm-Bank Risk).

Reserve Requirements

We measure reserve requirement innovations in two ways. In one measure, we build a simple index, adding or subtracting a unit respectively, on a tightening or easing policy event. In order to do so, we use the events from Chart 1. The change in the index is the policy innovation.

For the second measure, we use the counterfactual treatment variable described above and represented in equation (1).

where b refers to a bank and t to quarter.

In equation (1), we use the variation in counterfactual reserves to filter out the determinants of reserve requirements other than the regulatory changes. Additionally, using equation (1) as a treatment variable implies that total liabilities are not endogenously changing in response to RR shocks. This may look as a strong assumption,
especially because changes are not homogenous across components and may leave room to changes towards unaffected liabilities.

We take that regulatory changes are unexpected and substitution is gradual and lags behind the regulatory innovations. Notice that we measure the treatment variable in in the announcement quarter t. In principle, making substantial changes in the liabilities mix is costly and takes time, but assuming no substitution during the implementation quarter seems reasonable. Camors et al. (2016) use the same treatment variable and identification strategy. We follow them for greater comparability.

Identification Strategy

We present our results in two sections. The first section comprises the long panel estimates using the RR index. The second section presents the cross-sectional estimates around the two countercyclical changes in RR.

Long Panel

The long panel models considered in this paper are special cases of the following linear regression. For simplicity, we omitted the coefficients:

This time index reflects the number of RR interventions in place and AReservReq becomes a (+1) or (-1) indicator depending if the RR are tightening or easing in the quarter t-1. There are several policy events happening in different periods. Since the index makes no distinction over the intensity of the shock for different periods or different banks, there is also a presumption that no single event dominates the sample. In our data, this assumption is about right, since the regulation authority implements and later reverses the policy experiments, so that effects are balanced.

Treat is a dummy variable for the banks belonging to conglomerates that are affected by the policies, and zero otherwise. We are also interested in interaction terms of the policy innovation with Treat and a vector of control variables denoted by X in the equation. In this interaction, we consider macro variables, bank, firm and firm-bank controls. The term af*r represents the fixed effects introduced in the model. We introduce firm*bank and time fixed effects across our regressions. The last term in the equation refers to an idiosyncratic error term. We cluster standard errors at the bank and quarter level. Additionally, we use a distributed lag model, as well as a model with a simple lag structure.


This identification strategy fully replicates Camors et al. (2016). In this monthly diff-in-diff, the dependent variable is the change in the log of credit between t-1 and t+2. The treatment variable is the same presented in equation (1) and measured in t, the announcement month. We take all controls from t-1 to alleviate endogeneity concerns. We precisely estimate equation (3) on our most saturated regression. We measure the results relatively to t+2, because t+1 is still part of the implementation lag (that can take up to two weeks pending on RR subcomponents that are affected by the regulation):

We start estimating the lending channel, then bank interactions and firm interactions (risk-taking) progressively, and introducing firm fixed effects, bank controls and bank fixed effects in the risk-taking channel.

5. Results

We present two sets of results. The first set uses the long panel from 2008Q2 to 2015Q2 and the second set analyses the two shocks of the countercyclical RR policy. While the long panel measures the average shock across different events, the cross-section studies independently the easing and the tightening of countercyclical policies.

Long Panel

In Table 3, we present the single lag regressions.

The average effect of a (positive) change in RR in the treatment group is a credit contraction lying in the range of 0.73% to 1.16% (Table 3) in the following quarter. The exact absolute value of the elasticity is sensitive to the set of interactions included in the model. In the last column for example, this short run effect (of -1.16) is statistically and economically significant when considering both bank and firm heterogeneity interactions.

In Table 4, we use distributed lag model to estimate the one-year accumulated average effect of the same change in RR. Since there is no feedback from credit growth into the model, we assume complete transmission after one year. In this case, the average effect on the treatment group of a positive shock is a credit contraction lying in the range of 1.08% to 1.64%.

Some interactions are also noteworthy. First, banks" ex-ante liquidity ratio mitigates the effects of changes in RR, particularly one-year after the policy. Moreover, importance (i.e., total banks" ex-ante exposure to the firm relatively to its total capital) seems to reinforce the impact from RR policy. In other words, banks contract more credit to the firms that are more representative in their portfolio; or, increase diversification. These results are statistically and economically significant.

In the Appendix, we present a different strategy for the long panel, where we do not incorporate a treatment group and time fixed effects. This identification strategy allows us to assess synergies with macroeconomic conditions or monetary policy stances. Particularly, we use the following linear regression:

We run equation 4 to assess the average effect of a shock on one-quarter ahead credit (Appendix 1) and one-year ahead (Appendix 2). Results are consistent with the ones we find in Tables 3 and 4, but the magnitudes are a bit higher and only partially significant. We find weak evidence of synergies between monetary policy (measured as Selic) and RR shocks.


In Table 5, we present the results of the bank lending channel of countercyclical RR policy from the least to the most saturated regressions. We use identical identification strategies to both the easing (November, 2008) and the tightening (March, 2010) phases of the RR policy.

The results of our bank lending channel are statistically and economically significant. During the easing, we find that a 1pp decrease in RR, increases credit supply on the range of 1.30% to 1.43% on the most affected bank relatively to same firm. Similarly, a 1 pp increase in RR decreases credit supply on the range of 0.45% to 0.66%. These results suggest that the tightening phase of countercyclical policy affects the bank lending channel of RR on average less than the easing one. In other words, bank credit supply could be more reactive to the easing than to the tightening of countercyclical policy.

We also find compositional effects in credit supply related to banks" ex-ante observable characteristics. In Tables 6 and 7, we explore several bank interactions of the easing and tightening of countercyclical policy respectively.

During the easing phase, we find that foreign and small banks mitigate the effects of the policy by extending less credit to firms. Relatively to the same firm, a 1pp decrease in RR stimulates big, private and domestic banks to expand credit on average by 3% (Table 6). During tightening, big domestic banks respond contracting credit by 0.93% and big domestic private banks by 1.7% (Table 7). These results suggest that foreign banks respond primarily to the state of the global financial cycle (Moraes et al., 2017). In 2008, when global liquidity is short, foreign banks rebuild liquidity buffers, but do not extend credit in response to the easing policy. During tightening, they more than offset the policy, importing global liquidity, bypassing local macroprudential policy, and (contrary to domestic banks) extending credit.

As we mentioned, smaller banks suffered a liquidity squeeze because of a “fly-to- quality” movement from depositors (Oliveira et al., 2015). These banks fully mitigate the policy, because they are rebuilding liquidity buffers using the cash release. On the other hand, during tightening (Table 7), the small domestic banks expand credit.

In Tables 8 and 9, we present results for firm heterogeneity and the risk-taking channel of the easing and tightening of countercyclical policy, including firm and bank fixed effects.

We use two risk proxies. Firm Riskt-1 is the weighted average provision against the same firm across the banking sector and Future Defaultt+12 is a dummy variable that takes the value of 1 if the firm defaults in any of the 12 months in the future We also control for the number of employees of each firm.

During the easing phase, we find that banks extend less credit to firms considered riskier and, particularly, to firms that defaulted more in the future. These results suggest bank risk aversion during the easing phase of countercyclical policy In other words, credit extensions provided during (and empowered by the resources of) the easing policy are more carefully assessed by banks. Similarly, a “reach-for-yield” response is put in motion to compensate profitability losses during tightening. These results are both statistically and economically significant. Firms that end-up defaulting on their bank lending relationships 12 months into the future receive on average 40% less credit than the other firms during the easing (Table 8) and 36% more during tightening (Table 9). This result corroborates to the hypothesis of a positive risk-taking channel (or reach-for-yield response) of the macroprudential policy. It is also in line with Camors et al. (2016) and Jimenez et al. (2017).

In Table 10, we collapse our sample to the firm level to assess real effects. We find that the average firm ends up with 0.93% more credit in response to a decrease in

RR of 1pp. We also find significant and lower results for the tightening (0.6%). These results are not as strong as the ones of the loan-level sample, suggesting that firms (more likely related to small and foreign banks) end-up with less credit during easing


As a robustness check, we estimate the lending channel in placebo periods for both counterfactuals independently, 12 months after the tightening (when RR levels are relatively stable - Figure 2). Results are insignificant (Table 11).

6. Conclusion

We address the effects of reserve requirement (RR) changes on credit supply using different identification strategies applied to a large panel with several episodes of both loosing and tightening episodes, and two cross-sections focusing on the major countercyclical RR policies in Brazil.

The evidence is suggestive that RR policy impacts credit in the expected direction, i.e. RR easing increases credit, while RR tightening decreases credit. The exact quantitative impact depends on the specification, and it is sensibly higher in the long-run than in the short-run. We find suggestive evidence that higher ex-ante bank liquidity appears to reduce the impact of change in reserve requirements.

Exploring cross-section results, we find economically and statistically significant estimates of a bank lending channel of macroprudential policy using RR as an instrument. We find that during countercyclical easing, the more affected banks increase credit supply to the same firm on average by 1.3 to 1.4% in response to a 1pp RR reduction. During tightening, banks were less responsive and decrease credit supply to the same firm on average by 0.45% to 0.66% in response to a 1pp increase in RR.

We also find compositional bank effects. Foreign banks mitigate the easing policy and bypass the tightening more in line with the global financial cycle. We also find suggestive evidence that smaller banks caught in a liquidity trap during a “fly-to-quality” episode are less likely to extend credit during easing.

Similarly to Jimenez et al. (2017) and Camors et al. (2016), we find a positive risk-taking channel on countercyclical RR policy. We find this channel to be economically and statistically significant during the easing and tightening of countercyclical RR policy. This has direct implications for policy-makers in charge of financial stability.

7. References

Adrian, Tobias, and Hyun Song Shin (2009). "Money, liquidity, and monetary policy.” The American Economic Review, Papers and Proceedings of the One Hundred Twenty-First Meeting of the American Economic Association. Vol. 99, No. 2, pp. 600­605.

Altumbas, Yener, Leonardo Gambacorta, and David Marques-Ibanez (2014). "Does Monetary Policy Affect Bank Risk?" International Journal of Central Banking.

Araujo, Douglas K., Joao Barata R. Barroso, and Rodrigo B. Gonzalez (2016); “Loan- To-Value Policy and Housing Loans: effects on constrained borrower”. Central Bank of Brazil. Working Paper, WP445.

Bonomo, Marco, Ricardo Brito and Bruno Martins (2015). “The after crisis government driven credit expansion in Brazil: A firm level analysis”. Journal of International Money and Finance 55: 111-134.

Borio, Claudio and Halbin Zhu (2012). “Capital Regulation, risk-taking and monetary policy: a missing link in the transmission mechanism?” Journal of Financial Stability 8(4):236-251

Bustamante, Christian and Franz Hamann (2015). “Countercyclical reserve requirements in a heterogeneous-agent and incomplete financial markets economy”. Journal of Macroeconomics 46:45-70.

Camors, Cecilia D., Jose-Luis Peydro, and Francesc R. Tous (2016). “Macroprudential and Monetary Policy: Loan-Level evidence from Reserve Requirements”. Proceeding of XI Risk and Financial Stability Meeting, Central Bank of Brazil. Sao Paulo.

Cerutti, Eugenio, Stijn Claessens, and Luc Laeven. "The use and effectiveness of macroprudential policies: new evidence." Journal of Financial Stability (2015).

Claessens, Stijn, Swati R. Ghosh, and Roxana Mihet. "Macro-prudential policies to mitigate financial system vulnerabilities." Journal of International Money and Finance 39 (2013): 153-185.

Coleman, Nicholas and Leo Feler (2015). “Bank ownership, lending, and local economic performance during the 2008-2009 financial crisis”, Journal of Monetary Economics 71; 50-66.

Cordella, Tito, Pablo Federico, Carlos Vegh and Guillermo Vuletin (2014). “Reserve requirements in a brave new world”. The World Bank. Policy research paper, WPS6793.

Dell’Ariccia, Giovanni, Robert Marquez, and Luc Laeven (2017). “Bank leverage and monetary policy's risk-taking channel: evidence from the United States”. Journal of Finance 72: 613-654.

Glocker, Chirstian and Pascal Towbin (2012). “The Macroeconomic Effects of Reserve Requirements”. WIFO, Working Papers n. 420.

Jimenez,  Gabriel, Steven   Ongena, Jose-Luis        Peydro and    Jesus    Saurina    (2017). “Macroprudencial Policy, Countercyclical Bank Capital Buffers and Credit Supply:

Evidence from the Spanish Dynamic Provisioning Experiments”. Journal of Political Economy (forthcoming).

Hahm, Joon-Ho, Shin, Huyn Song, and Shin, Kwanho (2013). “Non-core bank liabilities and financial vulnerability”. Journal of Money, Credit and Banking, Blackwell Publishing, vol. 45:3-36.

Holmstrom, Bengt and Jean Tirole (1997). “Financial Intermediation, loanable funds and real sector”. The Quarterly Journal of Economics 112:668-691

Kashyap, Anil K. and Jeremy C. Stein. "What do a million observations on banks say about the transmission of monetary policy?" American Economic Review (2000): 407-428.

Kwajha, Asim, and Atif Mian (2008). “Tracing the Impact of Bank Liquidity Shocks: Evidence from an Emerging Market”. American Economic Review (2008): 1413-42.

Maddaloni, Angela, and Jose-Luis Peydro. "Bank risk-taking, securitization, supervision, and low interest rates: Evidence from the Euro-area and the US lending standards." Review of Financial Studies 24.6 (2011): 2121-2165.

Martins, Bruno, and Ricardo Schechtman (2013). “Loan Pricing Following a Macro Prudential Within-Sector Capital Measure”. Central Bank of Brazil, Working Paper, WP323.

Montoro, Carlos and Ramon Moreno (2011). “The use of reserve requirements as a policy instrument in Latin America” BIS Quarterly Review, March 2011:53-65.

Moraes, Bernardo, Jose-Luis Peydro, Jessica Roldan-Pena and Claudia Ruiz (2017). "The international bank lending channel of monetary policy rates and QE: credit supply, reach-for-yield, and real effects”. Banxico Working Paper Series, September n.15.

Oliveira, Raquel. D.F., Rafael Schiozer, Lucas Barros (2015). Depositors’ Perception of ’Too-Big-to-Fail’. Review of Finance. 19: 191-227

Pereira da Silva, Luiz A., and Ricardo E. Harris (2012). “Sailing through the Global Financial Storm: Brazil’s recent experience with monetary and macroprudential policies to lean against the financial cycle and deal with systemic risks”. Central Bank of Brazil, Working Paper, WP290.

Schiozer, R. F.; Oliveira, Raquel de Freitas (2016). “Asymmetric Transmission of a Bank Liquidity Shock”. Journal of Financial Stability, v. 25: 234-246.

Stein, Jeremy C. “An Adverse Selection Model of Bank Asset and Liability Management with Implications for the Transmission of Monetary Policy.” Rand Journal of Economics, Autumn 1998, 29(3), pp. 466-86.