traditional and matter-of-fact financial frictions in a
TRANSCRIPT
Traditional and Matter-of-fact Financial Frictions in a DSGE Model for Brazil: the role of macroprudential
instruments and monetary policy
Fabia A. de Carvalho, Marcos R. Castro and Silvio M. A. Costa
November, 2013
336
ISSN 1518-3548 CGC 00.038.166/0001-05
Working Paper Series Brasília n. 336 November 2013 p. 1-84
Working Paper Series
Edited by Research Department (Depep) – E-mail: [email protected]
Editor: Benjamin Miranda Tabak – E-mail: [email protected]
Editorial Assistant: Jane Sofia Moita – E-mail: [email protected]
Head of Research Department: Eduardo José Araújo Lima – E-mail: [email protected]
The Banco Central do Brasil Working Papers are all evaluated in double blind referee process.
Reproduction is permitted only if source is stated as follows: Working Paper n. 336.
Authorized by Carlos Hamilton Vasconcelos Araújo, Deputy Governor for Economic Policy.
General Control of Publications
Banco Central do Brasil
Comun/Dipiv/Coivi
SBS – Quadra 3 – Bloco B – Edifício-Sede – 14º andar
Caixa Postal 8.670
70074-900 Brasília – DF – Brazil
Phones: +55 (61) 3414-3710 and 3414-3565
Fax: +55 (61) 3414-1898
E-mail: [email protected]
The views expressed in this work are those of the authors and do not necessarily reflect those of the Banco Central or
its members.
Although these Working Papers often represent preliminary work, citation of source is required when used or reproduced.
As opiniões expressas neste trabalho são exclusivamente do(s) autor(es) e não refletem, necessariamente, a visão do Banco
Central do Brasil.
Ainda que este artigo represente trabalho preliminar, é requerida a citação da fonte, mesmo quando reproduzido parcialmente.
Citizen Service Division
Banco Central do Brasil
Deati/Diate
SBS – Quadra 3 – Bloco B – Edifício-Sede – 2º subsolo
70074-900 Brasília – DF – Brazil
Toll Free: 0800 9792345
Fax: +55 (61) 3414-2553
Internet: <http//www.bcb.gov.br/?CONTACTUS>
Traditional and Matter-of-fact Financial Frictions
in a DSGE Model for Brazil: the role of
macroprudential instruments and monetary policy
Fabia A. de Carvalho ∗
Marcos R. Castro †
Silvio M. A. Costa ‡
The Working Papers should not be reported as representing the views of theBanco Central do Brasil. The views expressed in the papers are those of the
author(s) and do not necessarily reflect those of the Banco Central do Brasil.
Abstract
This paper investigates the transmission channel of macroprudential instru-ments in a closed-economy DSGE model with a rich set of financial frictions.Banks are exposed to default risk in consumer, housing and investment loans.Consumer loans are extended based on expected borrowers’ capacity to pay offdebt with labor income. Housing loan extensions are subject to both debt-to-income and loan-to-value constraints. Banks optimally choose balance sheetallocation facing regulatory constraints, risk and frictions. The model is esti-mated with Brazilian data using Bayesian techniques, and is able to reproducenot only price effects from macroprudential policies, but also quantity effects.The effects of macroprudential instruments on banks’ balance sheets can spanthrough the stock and composition of loans, holdings of liquid assets, the shareof distributed dividends on total profits, or the composition of bank liabilities.The final impact of shocks to macroprudential instruments on capital buffersis tightly related to the type of regulatory instrument being adjusted. Unan-ticipated changes in reserve requirements have important quantitative effects,especially on banks’ liquid assets and on the choice of funding. This result alsoholds when required reserves deposited at the central bank are remunerated atthe base rate. The magnitude of the impact depends on the size of the incidencebase. Changes in required capital substantially impact the real economy andbanks’ balance sheets. Announcements of changes in capital requirement thatprecede actual implementation trigger immediate responses from banks, mostnoticeably with respect to dividend distribution, smoothing out the impact oncredit. Capital adequacy ratios improve immediately after the announcement.Countercyclical capital buffers have an important role to mitigate excessivecredit growth and to attenuate the recessionary impact of capital impairment.
Keywords: DSGE models, Bayesian estimation, financial regulation, mon-etary policy, macroprudential policyJEL classification: E4, E5, E6.
∗Research Department, Central Bank of Brazil, [email protected]†Research Department, Central Bank of Brazil, [email protected]‡Research Department, Central Bank of Brazil, [email protected]
3
Acknowledgements
We are grateful to Werner Roger, Enrique Mendoza, Matteo Iacoviello,Waldyr Areosa, the participants in the BIS CCA Research Network Confer-ence, in the XV Annual Inflation Targeting Seminar of the Banco Central doBrasil, at Ecomod 2013, at LuBraMacro 2013, at the IFABS 2013, the mem-bers of Central Bank of Brazil Working Group on Countercyclical Buffers andan anonymous referee for insightful comments and suggestions. Errors andomissions in the paper are of course the authors’ sole responsibility. The viewsexpressed in this work are those of the authors and do not necessarily reflectthose of the Central Bank of Brazil or its members.
4
1 Introduction
The literature on DSGE models with credit frictions has been built under a strongassumption with respect to collateral constraints: that loan extensions are tightlyrelated to the value of physical collateral that backs up the operation. The mainstrands of this literature incorporate agency problems in loan extensions backed upby physical capital (Bernanke, Gertler & Gilchrist (1999), Fiore & Tristani (2013),Glocker & Towbin (2012)), or binding credit constraints based on the value of house-holds’ assets, most usually housing (Iacoviello (2005), Gerali et al. (2010), Dib (2010),Andres, Arce & Thomas (2010)) or a mix of both (Paries, Sørensen & Rodriguez-Palenzuela (2011), Roger & Vlcek (2011), among others). Brzoza-Brzezina, Kolasa& Makarski (2013) provide an extensive comparison of the economic implications ofboth modeling assumptions.
Secured bank loans might be representative of the credit market in advanced economies,but other types of bank loans delinked from physical collateral have been gainingground1. At the beginning of 2013, for instance, the rating agency Moody’s down-graded Canadian banks mostly because of an important exposure of the financialsystem to unsecured consumer loans, whose performance is tightly related to house-holds’ disposable income.
In countries with strong impediments to the execution of collateral warranties, cred-itors find creative ways to devise contracts that mitigate default risk. In Brazil, forinstance, it is common practice for banks to require that borrowers prove to havefinancial capacity to settle debt installments with labor income. In such cases, debt-to-income or debt service ratios are more relevant than loan-to-value to determinelending rates and debt limits. As a matter of fact, about half the total volume ofconsumer loans in Brazil are not collateralized by physical assets, and are grantedwith no constraints on the final destination of borrowed funds. Credit lines linked topurchases of vehicles represent another third part of consumer loans, and althoughthere are constraints with respect to the destination of funds, a share of them are notcollateralized.
The way financial frictions are incorporated in the models can substantially affect themodel prescriptions. As a matter of fact, important conclusions in the DSGE liter-ature are model-dependent2. In BGG-type financial accelerators, fluctuations in theprice of physical collateral determine the occurrence of default, generating a strongconnection between the external finance premium and borrowers’ leverage. In thisenvironment, financial frictions operate mainly through their impact on investmentdecisions. On the other hand, when loans are extended based on the expected streamof labor income, other sources of bank vulnerability arise. These types of financialfrictions might also generate stronger procyclicality in the economy given their feed-back effect from labor conditions to credit risk and credit conditions, and hence from
1Mendoza (2002) mention cases in which variants of debt-to-income ratios were determinant toestablish loan contracts in the US.
2Brzoza-Brzezina, Kolasa & Makarski (2013) provide an extensive analysis of model-implieddifferences in responses of the main economic variables by examining credit constraint and externalfinance premium financial accelerators vis-a-vis a standard New Keynesian model.
5
consumption decisions funded by loans to the demand for goods, and back to laborconditions.
In this paper, we build a DSGE model with financial frictions that are suited toreproduce the dynamics of an economy where: 1) households take consumer loansbased on their expected future stream of labor income, and can default on theseloans if they face bad shocks to their income; 2) housing loans are extended based onloan-to-value and debt-to-income constraints; 3) consumer loans compete against in-vestment loans in bank asset allocation; 4) the perception of significant risk in lendingoperations makes liquid assets, such as public bonds, an attractive investment choicethat competes against bank loans; 5) banks’ external funding faces competition fromother investment opportunities easily available to bank clients; 4) banks make opti-mal decisions with respect to their balance sheet allocations and dividend distributionby making non-trivial choices between low-risk-low-return and high-risk-high-returnassets, subject to a number of regulatory constraints and macroprudential policiesthat distort their optimal balance sheet allocation. We use the model to assess thetransmission channels of macroprudential policies in Brazil, where all those featuresare very representative of the commercial and retail banking industry. To this end,we also add a few other frictions that make the model a better fit for Brazil’s tightlyregulated savings and mortgage market.
The possibility of credit default when labor income is a key factor to extend consumerloans is an assumption with important implications for the model responses. Ananalogous model where the debt-to-income constraint binds and where there is nodefault, such as in Mendoza (2002), underestimates the impact of macroprudentialinstruments on the economy and overestimates the impact of monetary policy onbanks’ capital adequacy ratios. In contrast, another analogous model where loans tohouseholds are fully collateralized by housing, such as in Iacoviello (2005) and Geraliet al. (2010), makes the contraction in consumer loans stronger.
The set of macroprudential instruments analyzed in this paper comprises: simplifiedBasle-1 and Basle-2 core capital requirements, with or without previous announce-ments of changes in regulation; risk-weights on bank assets to compute capital ad-equacy ratios; and remunerated and non-remunerated reserve requirements on dis-tinct bank funding sources. We also conduct experiments with countercyclical capitalbuffers to assess the potential stabilizing effect of these types of macroprudential in-struments.
The model is estimated with Bayesian techniques using Brazilian time series duringthe inflation targeting regime (1999Q3 to 2013Q2). Bayesian IRFs are computed, andcounterfactual exercises are reported to help understand the transmission channelsof macroprudential instruments and refine the assessment of their economic effects.
We find that monetary policy has a substantially larger impact on the real economycompared to macroprudential instruments3. On the other hand, reserve requirements
3This result is conditional on the chosen magnitude of policy shocks. Since the estimated varianceof reserve requirements is low, given the fact that the required capital ratio has not changed alongthe period we investigate, we chose the following values for the shocks: 1 p.p. for the monetarypolicy shock, 1 p.p. for capital requirement, 0.1 p.p. for risk weights on CAR, and 10 p.p. for
6
on time deposits and capital requirements have a potentially stronger impact oncredit. Bank balance sheet adjustments differ according to the macroprudential in-strument being shocked. Increased reserve requirements on time deposits trigger asizable adjustment of bank liquidity, whereas increased capital requirements lead tostrong earnings retention. In contrast, when the risk weight of a particular bank assetincreases in the computation of capital adequacy ratios, banks show a clear prefer-ence for cutting off excess capital instead of building up new capital. The oppositeholds for reserve and capital requirements, and for monetary policy shocks. For theseinstruments, actual Basel ratios improve when policy becomes stricter.
Bank liquidity, which in the model is represented by holdings of riskless public bonds,has an important role in attenuating the impact of reserve requirements on the realeconomy. The estimated economic impact of non-remunerated reserve requirementson demand deposits is small, but we show that this is a consequence of the smallincidence base.
Shocks to core capital requirement have strong effects on banks’ funding costs. Whenthe shock hits, banks try to improve their capital adequacy ratios by reshuffling theirassets and by change the rate of retained earnings over total profits. If announcementsof changes in capital requirements precede actual implementation, banks immediatelyincrease the ratio of retained earnings over total profits thus improving their capitaladequacy ratios over the impact period. Previous announcements are more effectivein reducing the risk exposure of the economy even after the shock hits.
We simulated scenarios in which bank capital is severely impaired or in which banksreduce their liquidity targets, creating conditions for excessive credit expansion. Inthese two cases, where shocks originate from the banking system, countercyclicalcapital buffers can substantially mitigate the impact of the shocks on credit and onthe real economy.
Our paper relates to the literature that analyzes the impact of macroprudential poli-cies in a DSGE framework (Glocker & Towbin (2012), Paries, Sørensen & Rodriguez-Palenzuela (2011), Roger & Vlcek (2011), Montoro & Tovar (2010), Areosa & Coelho(2013)). However, in most of these references housing or capital have a leading rolein credit extensions. Our paper also relates to the literature of endogenous banklending (Andres, Arce & Thomas (2010), Gerali et al. (2010)). Notwithstanding,our model goes beyond introducing monopolistic competition in bank lending. Thefrictions introduced in banks’ optimization problem are particularly suited to endoge-nously map the main determinants of lending spreads in Brazil: markup, default risk,administrative costs, direct and indirect taxes, and regulatory costs.
The paper is presented as follows. Section 2 describes the theoretical model. Section 3discusses the stationarization of the model and the computation of the steady state.Section 4 discusses the estimation conducted with Bayesian techniques. Section 4analyzes the transmission channel of macroprudential policies. Section 5 examinescounterfactual exercises and discusses some policy issues. The final section concludes.
reserve requirements.
7
2 The theoretical model
The economy is composed of households, entrepreneurs, producing firms and a finan-cial sector. Households are distributed in two groups: savers and borrowers. Theydiffer with respect to their intertemporal discount factors, to their access to invest-ment opportunities, and to their ownership of business activities. Both of them supplylabor to a labor union. Entrepreneurs engage in risky projects that are financed withtheir own net worth and with bank debt. Intermediate firms combine labor suppliedby unions and capital rented from entrepreneurs to produce inputs that will be as-sembled and distributed to final goods producing firms. These firms specialize inthe production of private and public consumption and investment goods, capital andhousing.
The financial sector is composed of a bank conglomerate and a retail money fund.The retail money fund represents an investment opportunity that dominates in returnall other financial options available to the household4. The fund’s portfolio is com-posed of government bonds and time deposits issued by the bank conglomerate. Thebank conglomerate comprises deposit branches, lending branches and a continuum ofbanks that make optimal choices with respect to dividend distribution and to balancesheet composition, constrained by regulatory requirements on mandatory reserves atthe central bank, capital adequacy ratio based on risk-weighted assets, and subject toregulatory constraints on housing loans and on the remuneration of savings accounts.External funding to the bank conglomerate comes from time, savings and demand de-posits. The conglomerate can also augment its net worth by retaining profits. Loansare risky since entrepreneurs’ projects and households’ labor income are subject toidiosyncratic shocks that might reduce their capacity to settle their debt obligations.Banks have preferences over balance sheet components, such as liquid assets and timedeposits. We also introduce rigidity in time deposit balances and lending rates, andfurther assume that bank activities generate administrative costs and are subject totax incidence.
In this session, we describe the main features of the theoretical model, emphasizingour contributions to existing models and adjustments to Brazilian particularities.The complete derivation of the theoretical model can be obtained from the authors.
2.1 Households
The economy is inhabited by two groups of households: net creditors and net debtorsof the financial system. Net creditors, henceforth ”savers”, have a range of availablefinancial investment opportunities, namely demand and savings deposits issued by thebank conglomerate and quotas of a retail money fund5. In addition, savers have right
4Notwithstanding, households have preferences over other financial investment opportunities withlower nominal return. This allows the model to find a non-negligible role for assets that are domi-nated in return and that are important to understand the transmission channel of reserve require-ments.
5The yield on savings accounts is regulated by the government as a markdown on the base rateof the economy, according to Brazilian practice.
8
to after-tax-profits from all business activities. Savers derive utility from consumptiongoods, housing, and liquid financial balances6.
Net debtors, henceforth ”borrowers”, also derive utility from consumption goods,housing, and demand deposits. Consumer and housing loans add resources to theirbudgets so as to finance their purchases of goods and housing. Loans are grantedby the bank conglomerate based on expectations with respect to borrowers’ capacityto settle debt payments with labor income. Household loans are risky since laborincome is subject to idiosyncratic shocks that realize only after loan contracts aremade.
In this respect, the model differs from the mainstream macroeconomic literature thatintroduces financial frictions in consumer loans. Although housing collateral has beenthe preferred choice in this literature, weakly collateralized or unsecured bank loansare gaining ground in the real world, bringing along concerns about the building upof vulnerabilities in financial systems7.
Non-corporate loans in Brazil amount to 43% of total bank loans (as of 2013Q2).About half the stock of unsecured consumer loans that are not collateralized withphysical capital and are not associated with the purchase of a particular good. Creditlines financing purchases of vehicles represent another third part of consumer loans,but the underlying goods are not necessarily put as collateral. Moreover, regardless ofcollateral requirements, banks decisions on consumer loans strongly rely on borrowers’capacity to settle their debt obligations with labor income.
Housing loans are about 12% of total bank loans. Although banks establish minimumLTV ratios in these operations, banks attribute great importance to the analysis ofdebt service-to-income ratios to make their final decisions on whether or not to extendthe loan. Interest rates on housing loans that are not regulated by the governmentare set according to the client’s basket of bank products and services, and less so onclient’s leverage or LTV ratios of those operations.
In this environment, events that affect the labor market potentially spillover to banks’risk taking, with feedback effects to the rest of the economy.
2.1.1 The Saver’s program
Savers are uniformly distributed in the continuum ∈ (0, ωS) and choose a stream{CS,t, HS,t, NS,t, D
SS,t, D
DS,t, D
FS,t
}of consumption, housing, labor supply, savings de-
6Since savings accounts are return-dominated by investment fund quotas during most of theanalyzed period in Brazil, we let depositors have preferences for savings.
7In 2013, Moody’s downgraded Canadian banks strongly based on an important exposure tounsecured consumer loans, whose performance is tightly related to households’ disposable income.The Canadian Quarterly Financial Report of the First Quarter 2013 highlights the risks of highdebt-service ratios that built up as a result of a prolonged period of low interest rates in Canada.The stress simulation points to a significant increase in loans in arrears should unemployment rise.
9
posits, demand deposits, and quotas of the retail money fund, to maximize
E0
∑t≥0
βtS
11−σX
(XS,t)1−σX − εLt LS1+σL
(NS,t)1+σL
+ψS,S1−σS
εS,St
(DSS,t
PC,tCS,t
)1−σS+
ψD,S1−σD
εD,St
(DDS,t
PC,tCS,t
)1−σD
εβt (1)
subject to the budget constraint
(1 + τC,t)PC,tCS,t + PH,t (HS,t − (1− δH)HS,t−1) +DFS,t +DS
S,t +DDS,t
= RF,t−1DFS,t−1 +RS,t−1D
SS,t−1 +DD
S,t−1 + (1− τw,t)(WNt NS,t
)+ TTS,t + ΠLU
S,t + ΠS,t + TTΓ,S,t + TGNS,t (2)
where
XS,t =
[(1− εHt ωH,S
) 1ηH
(CS,t − hSCS,t−1
) ηH−1
ηH +(εHt ωH,S
) 1ηD (HS,t)
ηH−1
ηH
] ηHηH−1
,
εβt , εLt , and εHt are preference shocks, LS, ψS,S, and ψS,D are scaling parameters, ωH,S
is a bias for housing in the consumption basket, hS is group-specific consumptionhabit, δH is housing depreciation, and τC,t and τw,t are tax rates on consumption andlabor income, respectively. Housing is priced at PH,t.
Labor is competitively supplied to labor unions at a nominal wage WNt . Labor unions
distribute their net-of-tax profits ΠLUS,t from monopolistic competition back to house-
holds as lump-sum transfers. Savers also receive lump sum transfers TTS,t from thegovernment, in addition to net-of-tax profits ΠS,t from firms, entrepreneurs, and thebank conglomerate. TTΓ,S,t are costs from capital utilization, which we assume to bedistributed as lump-sum transfers to savers and TGNS,t are transfers from entrepreneursthat quit their projects at each period. One-period returns on savings accounts andon retail money fund quotas are RS,t and RF,t, respectively.
2.1.2 The Borrower’s program
Borrowers are distributed in the continuum (0, ωB) . They take bank loans againsta proportion γB,Ct of expected future labor income. Borrower i’s total income fromlabor is subject to idiosyncratic shocks, $B,i,t ∼ lognormal (1, σB), a short-cut foridiosyncratic income shocks that do not affect firms’ aggregate production but thathave an impact on borrowers’ capacity to pay their debt installments. After therealization of the shock, borrower i’s net-of-tax nominal labor income is
$B,i,t [(1− τw,t)NB,i,tWt] (3)
where Wt is wage negotiated between firms and unions8.
8It can be shown that the borrower’s net-of-tax income from labor (1− τω,t)NB,i,tWt equalsthe net-of-tax labor income obtained from unions (1− τω,t)NB,i,tW
Nt plus her share on unions’
net-of-tax profits ΠLUS,t .
10
At period t, household i gets two types of credit: a consumer loan, with nominalvalue BC
B,i,t, and a housing loan, BHB,i,t. Both loans redeem in the subsequent period
and are negotiated at fixed interest rates, RL,CB,i,t and RL,H
B,t , respectively. The interestrate on housing loans is exogenously set by the government and does not depend onborrowers’ leverage. This assumption accords with the tightly regulated market ofBrazilian housing loans to low-priced real estate, which represents the bulk of thehousing loans market9. 10
In case of an adverse shock to the borrower’s labor income that leads to default onbank loans, the bank seizes a fraction γB,Ct of household’s net-of-tax labor income,after incurring proportional monitoring costs µB,C , in case default is on consumerloans, and µB,H , in case default is on housing loans11.
Housing loans are senior to consumer loans with respect to income commitment12.After labor decisions are made, the shock $B,i,t+1 realizes, and the borrower choosesto default if the amount of labor income previously committed to the loan is lessthan the total debt redeeming13. This threshold, $B,i,t+1, is such that the borroweris indifferent between settling debt obligations or letting the bank seize the committedshare of her labor income:
γB,Ct $B,i,t+1 (1− τw,t+1)NB,i,t+1Wt+1 = RL,CB,i,tB
CB,i,t +RL,H
B,t BHB,i,t (4a)
For convenience, we define another threshold $HB,i,t+1 with respect to the housing
loan:γB,Ct $H
B,i,t+1 (1− τw,t+1)NB,i,t+1Wt+1 = RL,HB,t B
HB,i,t (5)
Since lending branches are risk neutral and operate under perfect competition, for
9As of June 2013, the upper bound for the price of houses that qualify for these cheaper creditlines was BRL 500 thousand (approx. USD 250 thousand). Housing loans to low-priced real estateamounted to 76% of total housing loans financed through savings deposits in Brazil. Apart from theloans funded from savings deposits, an important segment of housing loans is funded with resourcesfrom the Severance Indemnity Fund (FGTS), managed by Caixa Economica Federal, a state-ownedbank. These housing loans represent 36% of total housing loans in Brazil and are granted at lowrates that bear little correspondence with the borrower’s leverage or collateral.
10Housing loans to low-priced real state are subject to an interest rate cap of 12% p.a.. However,the market of loans to low-priced real estate is by far dominated by Caixa Economica Federal (CEF),and the rates charged on these loans are not intimately associated with leverage or LTVs, althoughminimum LTV applies to the decision of whether or not to extend the loan to each particularproponent. The fact that banks are required to channel a certain share of their savings deposits tomortgage loans makes them closely track CEF’s lending rates in order to attract enough demand fortheir housing loans. Several other regulatory requirements apply to the market of housing loans andsavings deposits in Brazil. Our model addresses only the main aspects of this regulatory framework.
11These monitoring costs can be regarded as the cost of bankruptcy (including auditing, legal andenforcement costs).
12This assumption guarantees that expected default in housing markets is lower than in the marketfor consumer loans, which conforms with Brazilian empirical evidence.
13We rule out strategic default by assuming an implicit clause in the debt contract that if theborrower deviates from the optimal labor supply plan under commitment, she will be ruled out ofthe debt market in every subsequent period of the model.
11
each borrower the expected return from loans (left side of the following equation)must equal the funding costs associated with these operations (right side):
Et (1− µB,C)
∫ $B,i,t+1
$HB,i,t+1
[γB,Ct $B,i,t (1− τw,t+1)NB,i,t+1Wt+1 −RL,H
B,t BHB,i,t
]dF ($B,i,t)
(6)
+ Et
∫ ∞$B,i,t+1
RL,CB,i,tB
CB,tdF ($B,i,t) = RC
B,tBCB,i,t
orγB,Ct
[Et (1− τw,t+1)NB,i,t+1Wt+1GB,C
($B,i,t+1, $
HB,i,t+1
)]= RC
B,tBCB,i,t (7)
where
GB,C ($1, $2) = (1− µB,C)
[∫ $2
$1
$dF ($)−$1 [F ($2)− F ($1)]
](8)
+ ($2 −$1) (1− F ($2))
The household’s expected repayment on the consumer loan is given by
γB,Ct Et[(1− τw,t+1)NB,i,t+1Wt+1H
($B,t+1, $
HB,i,t+1
)]where
H($B, $
HB
)=
∫ $B
$HB
$dF ($)−$HB
(F ($B)− F
($HB
))+($B −$H
B
)(1− F ($B))
(9)
Similarly, the household’s expected settlement of the housing loan is
γB,CEt[(1− τw,t+1)NB,i,t+1Wt+1H
($HB,i,t+1, 0
)]and the expected settlement of bank loans is
γB,CEt [(1− τw,t+1)NB,i,t+1Wt+1H ($B,i,t+1, 0)]
Although housing loan rates for low-priced real estate in Brazil are not associatedwith borrowers’ leverage or collateral, banks abide by minimum LTV ratios to meetthe demand for housing loans. For this reason, we impose a collateral constraint onthis credit segment such that the nominal value of housing loans cannot exceed afraction γB,Ht of borrower’s housing stock.
BHB,i,t ≤ γB,Ht PH
t HBi,t (10)
This helps the model link fluctuations in housing prices to changes in the demand formortgage loans.
12
The LTV ratio γB,Ht is time varying, and allows the model to accommodate therecent increase in household indebtedness in Brazil, a trend that seems to be morerelated to the financial deepening of the economy than to a possible bubble in housingprices.
The representative borrower14 chooses the stream{CB,t, NB,t, HB,t,XB,t, DD
B,t, $B,t, $HB,t, B
CB,t, B
HB,t
}to maximize the utility function
E0
∑t≥0
βtB
1
1− σX(XB,t)1−σX − εLt LB
1 + σL(NB,t)
1+σL +ψD,B
1− σDεD,Bt
(DDB,t
PC,tCB,t
)1−σD εβt
(11)
subject to the budget constraint
(1 + τC,t)PC,tCB,t + PH,t (HB,t − (1− δH)HB,t−1) + γB,Ct (1− τw,t)NB,tWNt H ($B,t, 0) +DD
B,t
≤ BCB,t +BH
B,t +DDB,t−1 + (1− τw,t)
(WNt NB,t
)+ TTB,t + ΠLU
B,t
and the constraints from the optimal contract
γB,Ct Et (1− τw,t+1)NB,t+1WNt+1GB,C
($B,t+1, $
HB,t+1
)= RC
B,tBCB,t (12)
γB,Ct $HB,t (1− τw,t)NB,tW
Nt = RL,H
B,t−1BHB,t−1
BHB,t ≤ γB,Ht PC,tQH,tH
Bt
where
XB,t =
[(1− εHt ωH,B
) 1ηH
(CB,t − hBCB,t−1
) ηH−1
ηH +(εHt ωH,B
) 1ηH (HB,t)
ηH−1
ηH
] ηHηH−1
,
(13)and the auxiliary variables $B,t and $H
B,t are defined by
γB,Ct
($B,t −$H
B,t
)(1− τw,t)NB,tWt = RL,C
B,t−1BCB,t−1 (14)
2.2 Entrepreneurs
Investment loans are modeled as in Christiano, Motto and Rostagno (2010), exceptthat we introduce time varying LTV ratios to account for the fact that the share ofcapital stock financed through non-earmarked bank loans in Brazil is small and toaccommodate changes in leverage that are dissociated from innovations in the value ofcollateral. The recent financial deepening of the Brazilian economy can be captured
14In order to avoid heterogeneity issues that might arise if each household, faced with an idiosyn-cratic shock to her labor income, is allowed to freely choose her allocations, we assume that there isan insurance contract that evens out any income discrepancy among borrowers. We should imposethat every single household follow the same allocation plan that maximizes households’ averageutility.
13
through this variable.
At the end of period t, each entrepreneur i purchases capital KE,i,t+1 from capitalgoods producers and, at t + 1, rents it to the producers of intermediate goods atthe rental rate RK
t+1. Capital purchases are financed with entrepreneur’s net worthNEi,t+1 and commercial loans BE,i,t+1:
PK,tKE,i,t+1 = NEi,t +BE,i,t (15)
At the beginning of period t + 1, before capital is rented to domestic goods pro-ducers, it is subject to an idiosyncratic shock ωi,t+1˜ log normal (µE,t+1, σE,t+1), withEtωi,t+1 = 1, which represents the riskiness of business activity. We assume thatσE,t+1 follows an AR(1) process and that its realization is known at the end of periodt, prior to the entrepreneur’s investment decision.
After ωi,t+1 realizes, physical capital becomes ωi,t+1KE,i,t+1. After depreciation at therate δK , capital is sold back to capital goods producers at the market price PK,t+1.Therefore, the average nominal return of entrepreneur’s capital at period t+ 1 is
RTKt+1 ≡
∫ ∞0
ω[RKt+1 + PK,t+1 (1− δK)
]dF (ω, σE,t+1) (16)
= RKt+1 + PK,t+1 (1− δK)
The nominal amount BE,i,t is borrowed at the fixed rate RL,Ei,t . Loans are collateralized
by a fraction γEt of entrepreneurs’ stock of capital. We define the threshold value$i,t+1 such that
RLE,i,tBE,i,t = $i,t+1γ
Et R
TKt+1KE,i,t (17)
Whenever ωi,t+1 < $i,t+1, the entrepreneur goes bankrupt and the bank seizes thecollateral by incurring monitoring costs that correspond to a fraction µE of recoveredassets. Let RE,t be the proportional funding cost of the commercial lending branch.Since the idiosyncratic risk is diversifiable, the interest rate on commercial loans issuch that the expected profit of the financial intermediary is zero:
RE,tBE,i,t = γEt EtRTKt+1KE,i,tG ($i,t+1, σE,t+1) (18)
where
G ($t+1, σE,t+1) = (1− µE)
∫ $t+1
0
ωdF (ω, σE,t+1) + (1− F ($i,t+1, σE,t+1))$t+1
(19)
The entrepreneur’s expected cash flow is:
EtRTKt+1KE,i,t
[1− γEt H ($i,t+1, σE,t+1)
](20)
14
where
H ($t+1, σE,t+1) =
∫ $t+1
0
ωdF (ω, σE,t+1) + (1− F ($i,t+1, σE,t+1))$t+1 (21)
The entrepreneur’s problem amounts to choosing a sequence of {$i,t+1, BE,i,t, KE,i,t} tomaximize (20) constrained by (15), (17), (18) and BE,i,t ≥ 0.
At the end of each period, only a fraction γNt of the entrepreneurs survive. The onesthat leave the market have their capital sold and the proceeds are distributed to thehouseholds as lump-sum transfers TGNt . The average nominal value of entrepreneurs’own resources NE
t at the end of period t is
NEt = γNt R
TKt Kt−1
[1− γEt H ($E,t, σE,t)
](22)
where the survival rate γNt is given by
γNt =1
1 + e−γN−γNt
(23)
γNt = ρNγ γNt−1 + σNγ ε
Nγ,t
andTGNt =
(1− γNt
) (RKTt Kt−1
[1− γEH ($E,t, σE,t)
])(24)
2.3 Goods producers
Goods producers are modeled according to the standard DSGE literature. There isa continuum j ∈ (0, 1) of competitive intermediate goods producers that combinelabor and capital to produce homogeneous goods. The production function is
Zdj,t = A.εAt [utKj,t−1]α (εtLj,t)
1−α (25)
where εAt is a temporary shock to total factor productivity, A is a scaling constant,and εt is a permanent common shock to labor productivity whose growth rate gε,tfollows
gε,t = ρε.gε,t−1 + (1− ρε) .gε + εZt (26)
Cost minimization is subject to capital utilization adjustment costs Γu(ut) ≡ φu,1 (ut − 1)+φu,2/2 (ut − 1)2 .
Intermediate goods producers sell their output to retailers, who operate under mo-nopolistic competition setting prices on a staggered basis a la Calvo. Retailers whoare not chosen to optimize set their prices according to the indexation rule:
P dt (k) = πd,γdt−1 π
1−γdP dt−1 (k) (27)
15
where π is steady-state inflation. Retailers differentiate the homogeneous goodsand sell them to competitive distribution sectors. These, in turn, reassemble thedifferentiated goods using a CES production function:
Y dt =
[∫ 1
0
Zdt (k)
1µd dk
]µd(28)
Distributers sell their output to final goods firms, which specialize in the production ofgoods for government consumption G, private consumption C, capital investment IK ,and housing investment IH . Final goods producers are competitive and face nofrictions. Therefore, the zero profit condition yields
Y d,Jt = {G,C, IK,IH} (29)
P Jt = P d
t (30)
Perfectly competitive firms produce the stock of housing and fixed capital. At thebeginning of each period, they buy back the depreciated capital stock from en-trepreneurs as well as the depreciated housing stock from households. These firmsaugment their capital and housing stocks using final goods and facing adjustmentcosts to investment. At the end of the period, the augmented stocks are sold back toentrepreneurs and households at the same prices.
2.4 Investment Fund
A recent strand of the literature has introduced imperfect competition in the bankdeposits market15. This has implications for the dynamic responses of changes inreserve requirements. Under this assumption, the impact of shocks to reserve re-quirements on credit is partially buffered by adjustments in the cost of funding tobanks.
In Brazil, banks’ time deposits face fierce competition from retail money funds andfrom domestic federal bonds. About half the outstanding balance of domestic federalbonds are held by bank’s non-financial clients, either through holdings of securitiesor through quotas in mutual funds. Money market funds hold about 30% of domesticfederal bonds. In addition, the National Treasury allows private individuals to holdclaims to federal bonds negotiated at ”Tesouro Direto”16.
Such competition results in very narrow markdowns of time deposit rates on the baserate of the economy. For instance, in the period analyzed in this paper, the base ratewas merely 0.2 p.p higher on average than the effective 90-day time deposits (CDB)rate.
15Dib (2010), Paries, Sørensen & Rodriguez-Palenzuela (2011), Andres, Arce & Thomas (2010),Roger & Vlcek (2011).
16The stock of outstanding debt negotiated at Tesouro Direto is about 1% of the stock outstandingof domestic federal bonds.
16
We therefore assume that the interest rate on time deposits, RTt , and on domestic
public bonds, Rt, are equal at every point in time. This assumption has implicationsto the response of credit conditions given changes in reserve requirements.
Without loss of generality, we let the group of savers in the model hold quotas ofa retail money fund, whose portfolio is composed of time deposits DT
t issued bybanks and government bonds BF
t . Transactions with the retail money funds are freeof administrative costs. Since RT
t = Rt, the retail money fund is indifferent withrespect to its portfolio composition.
2.5 Banking sector
Our modeling strategy for the banking sector is adequate to assess the impact ofmacroprudential policy instruments not only on bank rates (prices) but also on quan-tities, through shifts in the composition of banks’ balance sheets.
The bank conglomerate is composed of a continuum [0, 1] of competitive banks thatget funding from deposit branches and extend credit to households and entrepreneursthrough their lending branches. Banks are the financial vessel of the conglomerate:they channel money market funds to the lending branches while making all importantdecisions with respect to the composition of the conglomerate’s balance sheet. Theconglomerate is subject to regulatory requirements and can only accumulate capitalthrough retained earnings, but the conglomerate can also choose to distribute divi-dends and accommodate changes in the regulatory environment by adjusting otherbalance sheet components. Our adopted segmentation of the bank conglomerate al-lows the model to endogenously reproduce the most relevant determinants of lendingspreads and the effects of regulatory requirements on bank rates and volumes.
2.5.1 Deposit branches
There is one representative deposit branch for each type of deposit. The demanddeposit branch costlessly takes unremunerated demand deposits, DD
S,t and DDB,t,
which are determined from households’ optimization problems. It then costlesslydistributes the resources to each bank j ∈ [0, 1] . In the following period, banksreturn the unremunerated funds to the deposit branch, which, in turn, returns themto households:
ωSDDS,t + ωBD
DB,t = DD
t =
1∫0
ωb,jDDj,tdj (31)
The savings deposit branch operates analogously, except that savings deposits accrueinterest RS
t that is regulated by the government according to:
RSt = 1 + ϕSR,t (Rt − 1) (32)
where ϕSR,t follows an AR(1) process around the steady state markdown.
17
The time deposit branch issues deposit certificates to the retail money fund, at interestrates equal to the base rate
(RTt = Rt
). The resources are also costlessly distributed
to the banks, and, in the following period, returned to the retail money fund withaccrued interest .
2.5.2 Lending branches
Lending branches get funding from banks and extend commercial and consumer loansto entrepreneurs and to borrowers, respectively. Without loss of generality, we assumeone representative lending branch for commercial loans and another for consumerloans.
The representative commercial lending branch is competitive and seeks to diversifyits funding sources. It borrows Bb
E,j,t from bank j at the interest rate RE,j,t. Total
loans BLB,EE,t , which are extended to entrepreneurs at the fixed-rate RL
E,t, are a CESaggregate of funding resources:
BLB,EE,t =
[∫ 1
0
ωb,j(BbE,j,t
) 1
µRE dj
]µRE(33)
whereBLB,EE,t = ωEBE,t (34)
In the following period, the lending branch chooses the amount to borrow from eachbank BE,j,t so as to maximize
RLE,tB
LB,EE,t −
∫ 1
0
ωb,jRE,j,tBbE,j,tdj (35)
subject to (33).
The FOC, together with the zero-profit condition, results in a demand function forcommercial loans funding from bank j :
BbE,j,t =
(RE,j,t
RE,t
) µRE1−µR
EBLB,EE,t (36)
As a result, each bank j has some market power in the allocation of available funds,and is free to choose the interest rate that it will charge the lending branches, con-strained by Calvo-type interest rate rigidities.
Aggregate funding to the investment lending branch has the following representation:
BbE,t ≡
∫ 1
0
ωb,jBbE,j,tdj (37)
= BLB,EE,t ∆R
E,t
18
where
∆RE,t =
∫ 1
0
ωb,j
(RE,j,t
RE,T
) µRE1−µR
Edj (38)
From Jensen’s inequality, ∆RE,t > 1. The net cash flow ΠE,LB
t from the investmentlending branch is
ΠE,LBt =
∫ 1
0
ωb,jBE,j,tdj −BLB,EE,t (39)
= BLB,EE,t
(∆RE,t − 1
)> 0
which is distributed to banks as lump-sum transfers:
ΠE,LBt =
∫ 1
0
ωb,jΠE,LBj,t dj (40)
The decisions of the representative investment lending branch are analogous to thoseof the representative consumer lending branch. The demand curve for funding is:
BC,bB,j,t =
(RCB,j,t
RCB,t
)− µRB,C
µRB,C
−1
BCB,t (41)
2.5.3 Housing loan branch
The Brazilian housing loans market is heavily regulated by the government. Theregulatory authority requires that a fraction τH,S,t of savings deposits be channeledto housing loans, and regulates on the lending rates17. We therefore assume that thefinal lending rate RL,H
B,t is set by the government as a markdown on the base rate:
RLB,H,t
Rt
=
(RLB,H,t−1
Rt−1
)ρRH(RLB,H
R
)1−ρRH
exp(εRH ,t
)(42)
Consequently, the only role played by the mortgage loan branch is to channel fundsfrom savings deposits to housing loans, not influencing either interest rates or vol-umes. It follows that
ωBBHB,t = BH,wb
B,t
Since mortgage loans are risky, the actual cash flow received by the mortgage branchis
ΠHt = ωBγ
B,Ct (1− τw,t)NB,tWtGB,H
($HB,t, 0
)−RH,wb
B,t−1BH,wbB,t−1 (43)
17Banks have discretion to extend loans to finance more expensive real estate. However, the bulkof housing loans in Brazil finance low-priced real estate, which is subject to such regulation.
19
where
GB,H ($1, $2) = (1− µB,H)
[∫ $2
$1
$dF ($)−$1 [F ($2)− F ($1)]
](44)
+ ($2 −$1) (1− F ($2))
The bank conglomerate absorbs the cost of default of housing loans as a loss thatcannot be passed through to volumes or rates in this market.
2.5.4 Banks
Banks act like treasury departments with a mandate on strategic decisions aboutdividend distribution, bound by regulatory constraints. Each bank j takes demanddeposits DD
j,t, time deposits DTj,t and savings deposits DS
j,t from households. Aftercomplying with regulation and making strategic decisions on capital accumulationand balance sheet composition, the bank channels the available resources to lendingand mortgage branches.
Banks have to comply with a number of regulatory requirements. Although the choiceof regulation introduced in the model was made to reflect the regulatory frameworkfaced by banks operating in Brazil, most of them are common place in the world.First, funding in the money market is subject to reserve requirements. In additionto unremunerated reserve requirements, which are commonly addressed in the liter-ature, we introduce remunerated requirements on savings and time deposits, and an”additional” reserve requirement detailed below 18. Second, the benchmark modelintroduces a simplified version of Basle 1 and Basle 2-type capital requirement, whichis based on the computation of capital adequacy ratios after weighting bank assetsaccording to their risk factors. Third, we introduce tight regulation on housing loansand savings accounts. Finally, we introduce tax collection on specific credit operationsin addition to an expense-deductible income tax on conglomerate’s activities.
Banks have preferences over some balance sheet components, particularly liquid assetsand time deposits. These preferences are introduced in banks’ optimization problemas targets to be attained in the balanced growth path. We let the data determinethe power of each of these assumptions by estimating cost-elasticity parameters.
Bank j′s balance sheet is:
BbE,j,t+B
C,bB,j,t+B
H,bB,j,t+BOM,j,t+RR
Tj,t+RR
Sj,t+RR
Dj,t+RR
addj,t −RR
S,Hj,t = DT
j,t+DSj,t+D
Dj,t+Bankcapj,t
(45)where Bankcapj,t is net worth, BOM,j,t are liquid assets (i.e., public bonds holdings),RRT
j,t, RRSj,t, and RRD
j,t are balances of required reserves on time, savings and demand
18Reserve requirements in Brazil have been used for a number of reasons: general financial stabilityconcerns, disruptions in specific segments of the credit or bank liquidity market, overall economicstability, or, outside the sample considered for estimation in this paper, for income distribution(Carvalho & Azevedo (2008), Montoro & Moreno (2011), Mesquita & Toros (2011), Tovar, Garcia-Escribano & Martin (2012))
20
deposits, respectively, and RRaddj,t are additional required reserves19.
Reserve requirements are determined as:
RRkj,t = τRR,k,tD
kj,t (46)
RRaddj,t = τRR,add,t
(DDj,t +DT
j,t +DSj,t
)(47)
where τRR,k,t are required ratios on deposits of type k = [Demand, T ime, Savings, Additional]set by the government and follow AR(1) processes around the steady state. Reserverequirements deposited at the monetary authority accrue the same rate as their inci-dence base.
Banks that take savings deposits in Brazil have to extend the equivalent of a fractionof their savings deposits as loans to low-priced housing. However, the publicly-ownedbank Caixa Economica Federal (CEF) also uses the resources from the SeveranceIndemninity Fund (FGTS) to fund mortgage loans. For this reason, we introducethe variable RRS,H
j,t to represent funding for housing loans obtained from the FGTS.For simplicity, we assume that FGTS funds fill the gap between required and actualdestination of savings deposits to housing loans.
RRS,Hj,t =
(τH,S,tD
Sj,t −BH
B,j,t
)(48)
Banks make no strategic decisions with respect to housing loans or interest rates onsavings deposits. On the other hand, the balance of time deposits is chosen by thebank, subject to adjustment costs that introduced to reproduce the strong persistencein the data:
ΓT
(DTj,t
DTj,t−1
)≡ φT/2
(DTj,t
DTj,t−1
εDTt − gε,tπC,t
)2
(49)
Bank capital can change with net operational results, FCbj,t and dividend distribution,
PC,tCB,j,t. It is also subject to shocks εbankcapt that can capture changes in market’sperception with respect to the quality of bank capital or any other shocks that changethe marked-to-market value of banks’ net worth20. The capital accumulation rule is:
Bankcapj,t = Bankcapj,t−1 + FCbj,t − PC,tCB,j,t +Bankcapj,tε
bankcapt (50)
Banks are constrained by a minimum capital requirement, γBankKt , modeled as anAR(1) with very high persistence (0.999). The capital adequacy ratio CARb
j,t mea-
19In addition to traditional reserve requirements on the main types of bank deposits, the CentralBank of Brazil has often used so called ”additional reserve requirements”, whose incidence baseis the same as standard required reserves. However, these additional reserve requirements can beremunerated differently from their standard counterparts or have a different form of compliance. Forsimplicity, we assume in our model that they have a homogeneous incidence rate upon the simpleaverage of all deposits. Other types of reserve requirements have been eventually introduced inBrazil, such as requirements on marginal changes in deposits, among others.
20Our modeling choice dispenses with the need to artificially introduce depreciation to bank capital
21
sures how much of total risk-weighted assets can be backed up by bank’s net worth:
BIj,t =Bankcapj,tCARb
j,t
(51)
where CARbj,t is
CARbj,t = τχ1B
C,bB,j,t + τχ2B
bE,j,t + τχ3B
H,bB,j,t + τχ4BOM,j,t + εCARj,t (52)
and where τχ • are risk weights modeled as AR(1) processes and εCARt is an AR(1)process centered on the value of risk-weighted assets that compose actual CAR’s inBrazil but that are not included in the model.
The Brazilian financial system operates with high excess of capital (5.4 p.p. abovethe minimum required as of 4Q2013, and 5.7 p.p. on average since 2000). After thebreak of the financial crisis in 2008, banks raised excess capital even further (7 p.p.above the minimum required in 2009). Although internal financing is usually costlierthan external financing, the capital buffer has a potential signaling effect of banks’soundness, with positive effects on wholesale funding costs and on the probability ofsudden stops in funding facilities. In addition, capital buffers can also prevent banksfrom falling short of the required minimum, an event that could result in undesiredsupervisory intervention.
We introduce precautionary capital buffer by letting banks face an appropriate costfunction when deviating from the minimum capital requirement. Since the modelsolution is linearized around the balanced-growth path, it suffices to introduce a costfunction that fulfills Γ′bankK < 0, Γ′′bankK > 0, and, at the balanced growth path,
ΓbankK
(BI
γBankK
)= 0, where BI
γBankK> 1 . For convenience, and w.l.g. since the cost
parameters that affect the model dynamics are estimated, we choose the followingrepresentation:
ΓbankK
(BIj,tγBankKt
)=χbankK,2
2
(BIj,tγBankKt
)2
+ χbankK,1
(BIj,tγBankKt
)+ χbankK,0 (53)
22
The one-period cash flow from bank j’s operations is:
FCbj,t =
(RE,j,t−1 − τB,E,t−1 − sadm,Et−1
)BbE,j,t−1 −Bb
E,j,t (54)
+(RCB,j,t−1 − τB,B,t−1 − sadm,Bt−1
)BC,bB,j,t−1
−BC,bB,j,t +RH
B,t−1BH,bB,j,t−1 −B
H,bB,j,t
+Rt−1BOM,j,t−1 −BOM,j,t −RTt−1D
Tj,t−1 +DT
j,t − ΓT
(DTj,t
DTj,t−1
)DTj,t
−RSt−1D
Sj,t−1 +DS
j,t −DDj,t−1 +DD
j,t
+RTRR,t−1RR
Tj,t−1 +RS
RR,t−1RRSj,t−1 +RD
RR,t−1RRDj,t−1
+RaddRR,t−1RR
addj,t−1 −R
S,HRR,t−1RR
S,Hj,t−1
−RRTj,t −RRS
j,t −RRDj,t −RRadd
j,t +RRS,Hj,t
− ΓbankK
(Bankcapj,t
γBankKt CARbj,t
)Bankcapj,t
− χOM2
(BOM,j,t
Lbbj,t− νOMt
)2
Lbt
− χd T2
(DTj,t
Lbbj,t− νd Tt
)2
Lbt
+ ΠLj,t + Ξb
j,t
where sadmt are administrative costs, which we assume to be proportional to therespective loan portfolio, τB,t are tax rates on credit, RRR,t is the interest rate paidby the monetary authority on required reserves, χOM and χd T are cost parametersassociated with the deviation of bank holdings of liquid assets and time deposits fromtheir targeted paths. We introduce a lump-sum transfer Ξb
j,t to insure against cashflow variations originating from interest rate rigidity:
Ξbj,t = (RE,t−1 −RE,j,t−1)Bb
E,j,t−1 +(RCB,t−1 −RC
B,j,t−1
)BC,bB,j,t−1 (55)
and ΠLj,t are lump sum transfers from conglomerate branches to bank j, introduced
to make aggregation straightforward:
ΠLj,t = ΠE,LB
j,t + ΠC,LBj,t + ΠL,B,C
j,t + ΠL,B,Hj,t + ΠL,E
j,t (56)
ΠE,LBj,t = BE,j,t − ωEBE,t (57)
ΠC,LBj,t = BC
B,j,t − ωBBCB,t (58)
ΠL,B,Cj,t = γB,C (1− τω,t)ωBNB,tW
Nt GB,C
($B,t, $
HB,t
)−RC
B,t−1BCj,B,t−1 (59)
ΠL,B,Hj,t = γB,C (1− τω,t)ωBNB,tW
Nt GB,H
($HB,t, 0
)−RH
B,t−1BHj,B,t−1 (60)
23
ΠL,Ej,t =
[γE(RKt + PK,t (1− δK)
)Kt−1G ($Et, σE,t)−RE,t−1Bj,E,t−1
](61)
Banks choose the stream of real dividend distribution {CB,j,t} to maximize
E0
{∑t≥0
βtBank
[1
1− σB
(CB,j,tεt
)1−σB]εβ,Bt
}(62)
subject to (36), (41), and (45) to (55) , where CB,j,t = divbj,t/PC,t , and divbj,t arebanks’ nominal dividends. We assume that banks’ intertemporal discount factor,βBank, is lower that of bank stockholders. This is a short-cut to modeling savers’risky investment choices, since, in practice, bank shareholders demand a return ontheir risky investment in bank operations that is higher than the risk-free opportunitycost Rt.
Let ΛBankjt be the Lagrange multiplier associated with the capital accumulation con-
straint (50), νBankj,t be the Lagrange multiplier of the balance sheet constraint, and
ηBank,•j,t be the Lagrange multiplier of the lending branches’ demand for loans. Firstorder conditions to bank j’s optimization problem are:
1
εt
(CB,j,tεt
)−σBεβ,Bt = ΛB
j,t (63)
ΛBankj,t
(1− εbankcapt
)= (64)
βBankEtΛBankj,t+1
πC,t+1
+ νBankj,t
−ΛBankj,t ΓbankK
(Bankcapj,t
γBankKt CARbj,t
)− ΛBank
j,t Γ′bankK
(Bankcapj,t
γBankKt CARbj,t
)Bankcapj,t
γBankKt CARbj,t
+ ΛBankj,t
χOM2
(BOM,j,t
Lbbj,t− νOMt
)(BOM,j,t
Lbbj,t+ νOMt
)
+ ΛBankj,t
χd T2
(DTj,t
Lbbj,t− νd Tt
)(νd Tt +
DTj,t
Lbbj,t
)
24
βBankEtΛBankj,t+1
πC,t+1
(RTt −RT
RR,tτRR,T,t −RaddRR,tτRR,add,t
)(65)
= βBankEtΛBankj,t+1
πC,t+1
Γ′T
(DTj,t+1
DTj,t
)(DTj,t+1
DTj,t
)2
+ νBankj,t (1− τRR,T,t − τRR,add,t)
+ ΛBankj,t
(1− τRR,T,t − τRR,add,t)− ΓT
(DTj,tDTj,t−1
)− Γ′T
(DTj,tDTj,t−1
)DTj,tDTj,t−1
+χOM2
(BOM,j,tLbbj.t
− νOMt)(
BOM,j,tLbbj,t
+ νOMt
)+χd T
2
(DTj,tLbbj,t− νd Tt
)(DTj,tLbbj,t
+ νd Tt − 2)
ΛBankj,t = βBankEt
ΛBankj,t+1
πC,t+1
Rt − νBankj,t (66)
+ γBankKt τχ4ΛBankj,t Γ′bankK
(Bankcapj,t
γBankKt CARbt
)(Bankcapj,t
γBankKt CARbt
)2
− ΛBankj,t χOM
(BOM,j,t
Lbt− νOMt
)
ΛBankj,t = βBEt
ΛBankj,t+1
πC,t+1
(RE,j,t − τB,E,t − sadm,Et
)− νBankj,t − ηBank,Ej,t (67)
+ ΛBankj,t τχ2γ
BankKt Γ′bankK
(Bankcapj,t
γBankKt CARbt
)(Bankcapj,t
γBankKt CARbt
)2
ΛBankj,t = βBankEt
ΛBankj,t+1
πC,t+1
(RCB,j,t − τB,B,t − s
adm,Bt
)− νBankj,t − ηBank,BCj,t (68)
+ ΛBankj,t γBankKt τχ1Γ′bankK
(Bankcapj,t
γBankKt CARbj,t
)(Bankcapj,t
γBankKt CARbj,t
)2
ROE,j,tβBEt
∑i≥0
ξiEβiBank
ΛBankj,t+i+1
πC,t+i+1
BbE,j,t+i
PC,t+i(69)
=µRE
µRE − 1Et∑i≥0
ξiEβiBankη
Bank,Ej,t+i
BbE,j,t+i
PC,t+i
25
RC,OB,j,tβBankEt
∑i≥0
ξiB,CβiBank
ΛBankj,t+i+1
πC,t+i+1
BC,bB,j,t+i
PC,t+i(70)
=µRB,C
µRB,C − 1Et∑i≥0
ξiB,CβiBankη
Bank,BCj,t+i
BC,bB,j,t+i
PC,t+i
First order conditions for consumer lending rates can be recursively represented as:
RC,OB,j,tβBankℵ
R,BC,2j,t =
µRB,CµRB,C − 1
ℵR,BC,1j,t (71)
ℵR,BC,1j,t = Et∑i≥0
ξiB,CβiBankη
Bank,BCj,t+i
BC,bB,j,t+i
PC,t+i
= ηBank,BCj,t
BC,bB,j,t
PC,t+ ξB,CβBankEtℵR,BC,1j,t+1
ℵR,BC,2j,t = Et∑i≥0
ξiB,CβiBank
ΛBankj,t+i+1
πC,t+i+1
BC,bB,j,t+i
PC,t+i
=BC,bB,j,t
PC,tEt
ΛBankj,t+1
πC,t+1
+ ξB,CβBankℵR,BC,2j,t+1
Commercial lending rates have analogous representations.
The FOCs show that the relevant opportunity cost for the bank is not just the baserate. Holding fixed the impact in the following period, higher capital buffers anddeviations from optimal time deposit balances increase banks’ opportunity cost. Forsmall deviations of the liquidity buffer from the target, greater liquidity decreases theopportunity cost so that loans can have more appealing rates to banks’ clients. Onthe other hand, when there is shortage of liquidity, the opportunity cost increasesand loans get more expensive, which will generate some asset reshuffling. SinceβBank < βS, the shadow price of one additional unit of bank capital in the balanced-growth path is higher than one unit of external funds.
26
2.5.5 Aggregating the bank conglomerate
The insurance Ξbj,t eliminates the heterogeneity that arises from interest rate rigidity,
and allows for a uniform representation of banks’ decisions. Aggregate variables are:
Bankcapt =
1∫0
ωb,jBankcapj,tdj (72)
BbE,t =
∫ 1
0
ωb,jBbE,j,tdj; BC,b
B,t =
∫ 1
0
ωb,jBC,bB,j,tdj (73)
DTt =
∫ 1
0
ωb,jDTj,tdj; DD
t =
∫ 1
0
ωb,jDDj,tdj; DS
t =
∫ 1
0
ωb,jDSj,tdj; (74)
1
εt
(CB,tεt
)−σBεβ,Bt = ΛB
t (75)
ΛBankt
(1− εbankcapt
)= βBankEt
ΛBankt+1
πC,t+1
+ νBankt (76)
− ΛBankt ΓbankK
(Bankcapt
γBankKt CARbt
)− ΛBank
t Γ′bankK
(Bankcapt
γBankKt CARbt
)Bankcapt
γBankKt CARbt
+ ΛBankt
χOM2
(BOM,t
Lbbt− νOMt
)(BOM,t
Lbbt+ νOMt
)+ ΛBank
t
χd T2
(DTt
Lbbt− νd Tt
)(νd Tt +
DTt
Lbbt
)
βBankEtΛBankt+1
πC,t+1
(RTt −RT
RR,tτRR,T,t −RaddRR,tτRR,add,t
)(77)
= βBankEtΛBankt+1
πC,t+1
Γ′T
(DTt+1
DTt
)(DTt+1
DTt
)2
+ νBankt (1− τRR,T,t − τRR,add,t) (78)
+ ΛBankj,t
(1− τRR,T,t − τRR,add,t)− ΓT
(DTtDTt−1
)− Γ′T
(DTtDTt−1
)DTtDTt−1
+χOM2
(BOM,tLbbt− νOMt
)(BOM,tLbbt
+ νOMt
)+χd T
2
(DTtLbbt− νd Tt
)(DTtLbbt
+ νd Tt − 2)
ΓT
(DTt
DTt−1
)≡ φT/2
(DTt
DTt−1
εDTt − gε,tπC,t)2
(79)
27
Γ′T
(DTt
DTt−1
)≡ φT
(DTt
DTt−1
εDTt − gε,tπC,t)εDTt (80)
ΛBankt = βBankEt
ΛBankt+1
πC,t+1
Rt − νBankt (81)
+ γBankKt τχ4ΛBankt Γ′bankK
(Bankcapt
γBankKt CARbt
)(Bankcapt
γBankKt CARbt
)2
− ΛBankt χOM
(BOM,t
Lbbt− νOMt
)
ΛBankt = βBEt
ΛBankt+1
πC,t+1
(RE,j,t − τB,E,t − sadm,Et
)− νBankt − ηBank,Et (82)
+ ΛBankt τχ2γ
BankKt Γ′bankK
(Bankcapt
γBankKt CARbt
)(Bankcapt
γBankKt CARbt
)2
ΛBankt = βBankEt
ΛBankt+1
πC,t+1
(RCB,j,t − τB,B,t − s
adm,Bt
)− νBankt − ηBank,BCt (83)
+ ΛBankt γBankKt τχ1Γ′bankK
(Bankcapt
γBankKt CARbt
)(Bankcapt
γBankKt CARbt
)2
ROE,tβBankℵ
R,E,2t =
µREµRE − 1
ℵR,E,1t (84)
ℵR,E,1t = ηBank,Et
BbE,t
PC,t+ ξEβBankEtℵR,E,1t+1
ℵR,E,2t =BbE,t
PC,tEt
ΛBankt+1
πC,t+1
+ ξEβBankEtℵR,E,2t+1
RC,OB,t βBankℵ
R,BC,2t =
µRB,CµRB,C − 1
ℵR,BC,1t (85)
ℵR,BC,1t = ηBank,BCt
BC,bB,t
PC,t+ ξB,CβBankEtℵR,BC,1t+1
ℵR,BC,2t =BC,bB,t
PC,tEt
ΛBankt+1
πC,t+1
+ ξB,CβBankEtℵR,BC,2t+1
28
ΠLt =
1∫0
ωb,jΠLj,tdj (86)
Interest rates on commercial loans can be recursively represented as
1 =(1− ξRE
)(ROE,t
RE,t
) 1
1−µRE
+ ξRE
(RE,t−1
RE,t
) 1
1−µRE
Similarly, for consumer loans:
1 =(1− ξRB,C
)(RC,OB,t
RCB,t
) 1
1−µRB,C
+ ξRB,C
(RCB,t−1
RCB,t
) 1
1−µRB,C
In addition,
∆RE,t =
(1− ξRE
)(ROE,t
RE,t
)− µREµRE
−1
+ ξRE
(RE,t−1
RE,t
)− µREµRE
−1
∆RE,t−1
∆RB,C,t =
(1− ξRB,C
)(RC,OB,t
RCB,t
)− µRB,C
µRB,C
−1
+ ξRB,C
(RCB,t−1
RCB,t
)− µRB,C
µRB,C
−1
∆RB,C,t−1
BbE,t = ωEBE,t∆
RE,t
BC,bB,t = ωBB
CB,t∆
RB,C,t
ΠE,LBt = Bb
E,t − ωEBE,t
ΠC,LBt = BC,b
B,t − ωBBCB,t
Banks’ balance sheets and dividends are aggregated as:
BbE,t +BC,b
B,t +BH,bB,t +BOM,t +RRT
t +RRSt +RRD
t +RRaddt +RRS,H
t (87)
= DTt +DS
t +DDt +Bankcapt
29
PC,tCB,j,t = Bankcapt−1 −(
1− εbankcapt
)Bankcapt (88)
−RTt−1D
Tt−1 −RS
t−1DSt−1 −DD
t−1 − ΓT
(DTt
DTt−1
)DTt
+RTRR,t−1RR
Tt−1 +RS
RR,t−1RRSt−1 +RRD
t−1 +RaddRR,t−1RR
addt−1 −R
S,HRR,t−1RR
S,Ht−1
+DTt +DS
t +DDt +Rt−1BOM,t−1
+ γB,C (1− τω,t)ωBNB,tWNt GB,C
($B,t, $
HB,t
)+ γB,C (1− τω,t)ωBNB,tW
Nt GB,H
($HB,t, 0
)+ γE
(RKt + PK,t (1− δK)
)Kt−1G ($Et, σE,t)
−RRTt −RRS
t −RRDt −RRadd
t +RRS,Ht
− ΓbankK
(Bankcapt
γBankKt CARt
)Bankcapt
− χOM2
(BOM,t
Lbt− νOMt
)2
Lbt −χd T
2
(DTt
Lbt− νd Tt
)2
Lbt
−(τB,E,t−1 + sadm,Et−1
)BbE,t−1 −
(τB,B,t−1 + sadm,Bt−1
)BC,bB,t−1
−BOM,t − ωBBCB,t − ωBBH
B,t − ωEBE,t
Finally, aggregation of reserve requirements results in
RRDt = τRR,D,tD
Dt
RRTt = τRR,T,tD
Tt
RRSt = τRR,S,tD
St
RRaddt = τRR,add,t
(DDt +DT
t +DSt
)RRS,H
t =(τH,S,tD
St −BH
B,t
)2.6 Government
The government is composed of a monetary and a fiscal authority. The monetaryauthority makes the following decisions: 1) the base rate of the economy; 2) ratiosand the remuneration of reserve requirements; 3) capital requirement ratio; 4) housingloans lending rate. The fiscal authority purchases goods, issues public bonds, leviestaxes, and makes lump sum transfers to households.
2.6.1 The monetary authority
The base interest rate is set by the monetary authority according to:
30
R4t =
(R4t−1
)ρR [(EtPC,t+4
PC,t
1
π4t
)γπ (gdptgdp
)γYR4
]1−ρ
εRt (89)
where unsubscribed R is the equilibrium nominal interest rate of the economy giventhe steady state inflation π, π4
t is a time-varying inflation target, and gdpt = GDPtεt
isthe stationary level of output that excludes banking costs:
GDPt = Yt − Tbank,t (90)
Tbank,t = sadm,Et−1 ωEBE,t−1 + sadm,Bt−1 ωBBCB,t−1 (91)
+χOM,t−1
2
(BOM,t−1
Lbt−1
− νOMt−1
)2
× Lbt−1
+ ΓT(DTt−1
)DT,wbt−1
+χd T
2
(DTt−1
Lbt−1
− νd Tt−1
)2
Lbt−1
+ γEt µE(RKt + PK,t (1− δK)
)ωEKE,t−1J ($E,t)
+ γB,Ct (1− τw,t)ωBNB,tWt µB,C[J ($B,t)− J
($HB,t
)−$H
B,t
(F ($B,t)− F
($HB,t
))]+ γB,Ct (1− τw,t)ωBNB,tWt µB,H J
($HB,t
)where
J ($B,t) = ℵcdf(
log ($B,t)
σB− σB
2
)
J($HB,t
)= ℵcdf
(log($HB,t
)σB
− σB2
)
($E,t) = ℵcdf(
log ($E,t)
σE− σE
2
)The time varying inflation target follows
π4t =
(π4t−1
)ρπ4 (π4)1−ρπ4 επ
4
t (92)
The rules for the other instruments set by the monetary authority are described in(32), (42), (46), (47), (48), and (52).
31
2.6.2 The fiscal authority
The fiscal authority consumes final goods according to the rule:
Gt
εt(93)
= ρg
(Gt−1
εt−1
)+ (1− ρg)
(g − µB,G
(Bt−1 +RRD
t−1 +RRTt−1 +RRS
t−1 +RRaddt−1
PC,t−1εt−1
−(b+ rrD + rrT + rrS + rradd
)))+ εGt
where lower-case variables denote stationary variables, and g is the steady statevalue of stationary government consumption. Government consumption has a role instabilizing gross public sector debt, which incorporates central bank’s liabilities.
Public debt issued by the government meets the demand from the retail money fundand the wholesale bank:
Bt = BOM,t +BF,t (94)
Tax rates τC,t, τw,t, τΠ,t,and τB,B,t follow AR(1) processes around their steady states21.
The joint public sector budget constraint is
PG,tGt + TTt (95)
−RS,Ht−1RR
S,Ht−1
+RRDt−1 +RRR,T
t−1 RRTt−1 +RRR,S
t−1 RRSt−1 +RRR,add
t−1 RRaddt−1 +Rt−1Bt−1
= τw,tΠLUt + τΠ,tΠt + τΠ,tν
bΠbt + τw,tW
Nt Nt + τC,tPC,tCt
+ ωEτB,E,t−1BE,t−1 + ωBτB,B,t−1BCB,t−1
+RRDt +RRT
t +RRSt +RRadd
t −RRS,Ht +Bt
2.7 Market clearing
Market clearing requires that the following equalities hold:
Y dt = Y C,d
t + Y G,dt + Y IK ,d
t + Y IH,dt (96)
Y G,dt = Gt (97)
Y IH,dt = IH,t (98)
Y IK,dt = IK,t (99)
21Due to lack of time series of tax levied on financial intermediation, disaggregated into privateindividuals and firms, we assume that τB,E,t is a fixed proportion of τB,B,t.
32
Y Ct = Ct (100)
3 The steady state and calibrated parameters
The model variables were stationarized by dividing real variables by the technologyshock εt and nominal variables by both the technology shock and the consumer pricelevel, PC
t .
Pinning down the steady state of the Brazilian economy is an exercise that involves agreat amount of judgement. Most series have trends, and long series are the exception,not the rule. In addition, some markets have been deepening over the past years,adding uncertainty about what is trend, what is transition, or what is structuralchange. The prescription of using filtered series when trends are an issue does notapply indistinctly to Brazilian data. Filtered series in many cases give the wrong ideaof where economic variables are in the business cycle.
With that in mind, we followed two different strategies to calibrate the steady state.The main economic ratios were fixed according to their average during the inflationtargeting period (Table 1)22. GDP growth and the base rate were also fixed accordingto the average in this period.
We chose to calibrate credit and deposits as a share of GDP, as well as lending ratesand the markdown of savings rates, according to the most recent observations in thedata. The reasons for this choice are as follows. Bank series in Brazil show serioustrends and possible issues related to transition (Figure 1). Over the past decade,credit-to-GDP ratios have accelerated. Much of this acceleration is attributed tothe financial deepening of the economy, a consequence of greater social inclusion anda prolonged period of macroeconomic stability with low inflation. Yet, the currentlevels of credit-to-GDP are still low compared to the rest of the world, and creditgrowth has been strongly financed stable funding. In fact, credit-to-time deposits hasdeclined during most part of the credit acceleration period. We expect the financialdeepening of the economy to continue allowing for further rounds of sustainable creditexpansion.
The ex-ante default ratios in the steady state were set at 2.9% for investment loansand 7% for consumer loans, in line with recent available data on actual default. Wefixed steady state lending rates and balances as shares of GDP, in addition to bankingspread components. We set the variance of the idiosyncratic shock to entrepreneur’scollateral value (σE) to 0.58 to calibrate capital depreciation at 1.5% per quarter.The variance of the idiosyncratic shock to borrower’s committed income (σB) wasfixed at 0.223. From these assumptions, all the remaining variables related to finan-
22In this table, GDP ratios are expressed in terms of yearly GDP. In the implementation of themodel, the ratios were all computed in terms of quarterly GDP.
23This parameter has an important effect on the model’s impulse responses. Higher values drivethe responses of consumer loans to monetary policy rate shocks to a very unlikely region.
33
cial accelerators, including threshold levels of idiosyncratic shocks, LTV-ratios, andmonitoring costs are obtained after evaluating the model at the steady state. Thestock of capital is then determined from the entrepreneur’s financial accelerator.
The capital adequacy ratio was fixed at the actual average of the Brazilian FinancialSystem24 from the beginning of the series on December 2000. Required capital wasset at 11%, the regulatory rate for tier-1 capital since the implementation of Basle1. Risk weights on bank assets were set at the actual values reported by Brazilianbanks on portfolios that are related to the ones included in our model (i.e., 1.5 forconsumer loans, 1 for investment loans, 0.9 for housing loans, and 0 for governmentbonds). Given the capital adequacy ratio and banks’ intertemporal discount factor,we calibrated the intercept and the slope parameter of the cost function associatedwith deviations from the capital requirement. Hence, the curvature parameter couldbe estimated.
We assumed a log-linear utility function for banks’ optimization problem, and setbanks’ intertemporal discount factor at 0.98 which would represent a 17.5% nominalreturn on banks’ dividends25.
Reserve requirement ratios were fixed at the average of their effective ratios, whichwere calculated as the share of reserves deposited at the central bank to the volumeof deposits in the economy. For time deposits, the average ratio was taken fromDecember 2001, when this requirement was last reintroduced, to December 2013.Average additional reserves were calculated from the series starting on December2002, when they were introduced. Requirements on savings accounts and demanddeposits are averages of the entire inflation targeting period. The minimum requiredallocation of funds from savings deposits in housing loans was set according to actualcompliance26.
The tax on financial transactions was calibrated to match the share of indirect tax onbanking spreads, as reported by the Central Bank of Brazil in its Banking Reports27.
The participation of each group of households in labor, consumption goods and hous-ing has important implications for the model dynamics. As a result, we attempted tofind out-of-the-model relations that could help pin down this participation. We fixedthe share of housing consumed by borrowers in the steady state as the ratio betweenthe approximate value of collateral put up in housing loans and the model’s impliedvalue of real estate in the economy28. We also assumed that the government does notmake transfers to borrowers29.
24The reported capital adequacy ratio does not include development banks, such as the NationalDevelopment Bank (BNDES).
25The impulse responses are not sensitive to this parameterization as long as 0.9 << βBank < βS .Values near the lower bound generate unlikely responses to monetary policy shocks.
26The actual compliance does not include compliance in the form of securitized debt (FCVS) orother instruments that alleviate the burden of the requirement.
27www.bcb.gov.br/?spread28Since the LTV ratio in housing loans was 0.6 in 2012, we assumed that the value of the collateral
in this market was twice the stock of loans divided by the LTV ratio.29By the time we finished this version of the paper, we had not had access to data on debt
34
From the assumed ratios of banks’ balance sheet components, we obtained the steadystate balance of public bonds at banks’ assets, and consequently pinned down banks’liquidity target. From the assumed ratio of public debt, we calibrated the total stockof public bonds in the economy and at the retail money fund’s portfolio.
4 Estimation
The model was linearized by Dynare around the steady state and estimated usingBayesian techniques. We used the following time series as observables, sampled fromthe inflation targeting regime in Brazil (1999:Q3 to 2013:Q4)30:
• Consumer inflation(πobsC,t
): inflation index used to assess compliance with the
inflation target (IPCA - Indice de Precos ao Consumidor Amplo – IBGE).
• Inflation target(πobsC,t
): 4-quarter-ahead actual inflation target.
• Nominal interest rate(Robst
): quarterly effective nominal base rate (Selic).
• Aggregate private consumption(cobst)
: share of seasonally adjusted privateconsumption in nominal values to the seasonally adjusted proxy for a closedeconomy nominal GDP. The proxy for a closed economy GDP was calculatedas the sum of the nominal values of private and public consumption and fixedcapital formation.
• Government consumption(gobst)
: share of seasonally adjusted governmentconsumption in nominal values to the seasonally adjusted proxy for a closedeconomy nominal GDP.
• Unemployment(U obst
): Brazilian National Statistics Institute (IBGE)’s new un-
employment series with missing values filled up by an interpolation of a serieseconometrically built from IBGE’s discontinued series of unemployment. Theresulting series was detrended by its mean from 1999Q1 to 2012Q1.
• Real wage change(∆wobst
): quarterly change in IBGE’s seasonally adjusted
real wage series with missing values filled up by an interpolation of a serieseconometrically built from IBGE’s discontinued series of real wages.
• GDP(gdp
obs
t
): HP cycle of the log of the proxy for the real GDP of the closed
economy. This proxy was constructed by deflating the proxy for the closedeconomy nominal GDP by a composite of consumer and producer price inflation,to proxy for the quarterly GDP deflator.
• Installed capacity utilization(uobst): quarterly capacity utilization published by
Fundacao Getulio Vargas, demeaned by the average from 1999Q1 to 2012Q2.
commitment by indebted households. We thus fixed borrowers’ participation in the labor marketunder the arbitrary assumption that indebted households in Brazil have a debt commitment of 50%of their annual labor income.
30Missing variables were filled up with standard Dynare routines.
35
• Bank capital(bankcapobst
): Brazilian financial system’s core capital as defined
by the Central Bank of Brazil, as a share of quarterly nominal GDP. Both seriesare seasonally adjusted.
• Capital adequacy ratio(CARobs
t
): actual average capital adequacy ratio of the
Brazilian financial system
• Commercial loans(bobsE,t): stock outstanding of investment loans granted by
banks with freely allocated funds as a share of quarterly nominal GDP. Bothseries are seasonally adjusted.
• Consumer loans(bC,obsB,t
): stock outstanding of consumer loans granted by banks
with freely allocated funds as a share of quarterly nominal GDP. Both seriesare seasonally adjusted.
• Housing loans(bH,obsB,t
): stock outstanding of regulated mortgage loans to house-
holds as a share of quarterly nominal GDP. Both series are seasonally adjusted.
• Lending spread for commercial loans(RL,obsE,t
): Ratio between the quarterly
effective nominal interest rate on investment loans granted with freely allocatedfunds and the base rate. The lending rates on each type of loan are weightedby their respective stock outstanding. Missing observations at the beginningof the series were filled up by an interpolation of the series of lending rates onconsumer loans.
• Lending spread for consumer loans(RL,obsB,C,t
): Ratio between the quarterly ef-
fective nominal interest rate on consumer loans granted with freely allocatedfunds and the base rate. The lending rates on each type of loan are weightedby their respective stock outstanding.
• Lending spread for housing loans(RL,obsB,H,t
): Ratio between the quarterly effec-
tive nominal interest rate on housing loans granted with freely allocated banks’funds and the base rate. The lending rates on each type of loan are weightedby their respective stock outstanding. Although the bulk of housing loans inBrazil are granted with mandatorily allocated funds, the series for lending rateson these loans is only available from September 2000 onwards.
• Default rate on commercial loans(defaultobsE,t
): investment loans in arrears for
over 90 days as a share of total outstanding investment loans.
• Default rate on consumer loans(defaultobsB,t
): consumer loans in arrears for over
90 days as a share of total outstanding consumer loans.
• Time deposits(dT,obst
): quarterly average of the total stock of non-financial in-
stitutions’ and households’ time deposits held by the Brazilian financial systemas a share of nominal quarterly GDP. Both series are seasonally adjusted.
• Demand deposits(dD,obst
): quarterly average of the total stock of non-financial
institutions’ and households’ demand deposits held by the Brazilian financialsystem as a share of nominal quarterly GDP. Both series are seasonally adjusted.
36
• Savings deposits(dS,obst
): quarterly average of the total stock of non-financial
institutions’ and households’ savings accounts in the Brazilian financial systemas a share of nominal quarterly GDP. Both series are seasonally adjusted.
• Markdown on savings rates(µR
S ,obst
): Ratio between the quarterly effective
nominal interest rate on savings accounts and the base rate.
• Required reserve ratio on time deposits(rrT,obst
): quarterly average balance of
required reserves on time deposits held at the central bank as a share of thetotal balance of non-financial institutions’ and households’ time deposits heldby the Brazilian financial system.
• Required reserve ratio on demand deposits(rrD,obst
): quarterly average balance
of non-remunerated required reserves on demand deposits held at the centralbank as a share of the total balance of non-financial institutions’ and households’demand deposits held by the Brazilian financial system.
• Required reserve ratio on savings deposits(rrS,obst
): quarterly average balance
of required reserves on savings accounts held at the central bank as a share ofthe total balance of non-financial institutions’ and households’ savings depositsheld by the Brazilian financial system.
• Additional required reserves ratio(rradd,obst
): quarterly average balance of sup-
plementary required reserves on demand, time and savings deposits held atthe central bank as a share of the total balance of demand, time and savingsdeposits held by the Brazilian financial system on behalf of non-financial in-stitutions and households. Although the incidence base of additional requiredreserves singles out each type of deposit, we choose a simplified approach tocalculate the aggregate effective required reserve ratio.
• Civil construction(constobst
): quarterly change in IBGE’s seasonally adjusted
index of civil construction.
Employment in the model was mapped into the unemployment series according to:
(1 + βS
)Et = βSEt+1 + Et−1 +
(1− βSξE
) (1− ξE)
ξE
(Nt − Et)
∆wobst =Wt/P
Ct εt
Wt−1/PCt−1εt−1
/∆n (101)
where ∆n is the steady state growth of the employed population.
For the choice of prior means, we used information from Brazilian-specific empiricalevidence, whenever available, or followed the related literature. We tried to com-pensate for the arbitrariness in the choice of some priors by setting large confidence
37
intervals. Table 2 shows the results of the estimation, including prior and posteriormoments31.
5 The Transmission Channel of Macroprudential
Policies
5.1 Reserve requirement shocks
To analyze the transmission channel of reserve requirements (RR), their ratios wereshocked at 10 p.p., a reasonable magnitude considering Brazil’s recent history. Thisimplies that RR on demand deposits rise on impact to 59.6%, from the steady statelevel of 49.6%, RR on time deposits rise to 21% from 11%, RR on savings accountsrise to 28% from 18%, and the additional RR rises to 17.7% from 7.7%.
Figure 6 shows the 10 p.p. shock to (unremunerated) RR on demand deposits((τDRR,t
)) 32. This instrument has a small contractionist impact on the real economy.
Although this might seem at odds with the literature, we argue below that the smallbase of incidence has an important contribution to this result. The most importanteffects are restricted to banks’ balance sheets, with some spillover to capital and hous-ing investment decisions. On impact, banks immediately unleash liquidity and cutdown on dividend distribution to alleviate the burden of strained liquidity. Fundingfrom time deposits increases only gradually due to nominal rigidities. The liquiditystrain causes an important increase in banks’ funding cost, which is only partiallypassed through to final lending rates, since leverage is not put under great pressurein the real economy. Higher lending rates of commercial loans reduce the demand forinvestment goods, which drives down the price of capital, further constraining creditconditions in the commercial segment. The overall impact of this shock on banks’balance sheet slightly improves the capital adequacy ratio.
A shock to (remunerated) RR on time deposits (Figure 7) has a much stronger impacton the economy. The transmission channel differs with respect to banks’ dividenddistribution. Since this reserve is remunerated at the base rate, the loss of revenuesfrom interest rate accrued on bank assets is not as in the case of an increase inunremunerated RR. As a result, banks choose not to retain dividends. A shock to(remunerated) RR on savings accounts (Figure 8) is qualitatively similar, yet theamplitude of the responses is lower given the smaller incidence base.
The total balance of time deposits in Brazil is almost eight times as large as thatof demand deposits. A fair comparison of the potential impact of RR needs to takeinto account the size of their incidence base. After scaling the shocks to generate anequivalent impact in terms of the amount of funds seized by the central bank, we
31We ran 2 chains of 180,000 draws of the Metropolis Hastings to estimate the posterior.32We computed Bayesian impulse responses to the shocks in the model using the standard Dynare
toolkit. 95% confidence intervals are plotted in Bayesian IRFs alongside the estimated mean re-sponse.
38
obtain the traditional prediction that reserve requirements on demand deposits havestronger marginal impact on the economy mostly through the direct impact on banks’profits and less so on banks’ balance sheet allocations. In particular, we applied a 50p.p. shock to RR on demand deposits, a 7 p.p. shock to RR on time deposits, and a15 p.p. shock to RR on savings deposits. Figure 12 compares the impulse responses33. In all cases, monetary policy was kept unresponsive so that we could evaluate thefull impact of RR.
In Brazil, reserve requirements on time deposits have been the instrument of choicewhen the central bank needs to drain liquidity from the economy. There is an implicitassumption that this would be the least distortionary instrument for this purpose.However, the frictions introduced in the optimal bank balance sheet allocation in ourmodel, and that are validated by the estimation, imply that an exogenously imposedasset allocation is costly to the bank, and thus higher funding costs translate intohigher lending rates. This has important policy implications34.
The literature interprets the modest degree of the real impact of reserve requirementsas a consequence of a responsive monetary policy. Glocker & Towbin (2012), forinstance, argue that if interest rate setting is dissociated from decisions on reserverequirements, the former may neutralize the impact of the latter. We conduct acounterfactual exercise in which monetary policy remains nonresponsive to economicconditions while we stress the model with a shock to reserve requirements. Figure11 shows the responses35. When monetary policy is unresponsive, the impact ofchanges in reserve requirements on GDP is stronger and more prolonged. In thatsituation, when banks increase lending rates to accommodate the increase in fundingcosts, households’ consumption and investment demand are no longer stimulated bymonetary policy. As such, the impact on the demand for goods is not alleviated,and consequently the drop in the demand for labor curtails borrowers’ capacity totake loans. As a result, borrowers’ consumption is more severely affected. The shockis further reinforced since the cost of funding rises. Banks reduce their positions inrisky assets (i.e., loans) and the capital adequacy ratio further improves. In sum, ourresults are in line with the literature. When monetary policy does not relieve thecontractionist impact of shocks to reserve requirements, the economy faces a moresignificant downturn.
The analysis of the responses to changes in the ratios of remunerated RR, either ontime deposits or savings accounts, when monetary policy is kept unchanged, yieldsthe same conclusions outlined above (Figures 9 and 10).
33All counterfactual exercises use the mean of the posterior estimation to produce impulse re-sponses
34Montoro & Moreno (2011) claim that if RR are partially remunerated, the distortionary taxeffect is reduced, but their overall impact on the banking system is also lessened. In our model, theestimated impulse responses of changes in remunerated reserve requirements on time deposits canhave non-negligible effects on the real economy notwithstanding the fact that there is no mismatchbetween the interest rate paid on bank deposits and that accrued on required reserves.
35To do this exercise, we perturbed the model with unexpected shocks to the interest rate rulesuch that the nominal base rate would remain at the steady state level over the response period.
39
5.1.1 A brief note on the literature
The international literature finds evidence of a moderate degree of the impact of non-remunerated RR on the economy. The assumptions underlying these conclusions aremanifold. Tovar, Garcia-Escribano & Martin (2012) use event study and dynamicpanel VAR on a number of Latin American countries to find that RR have a moderateand transitory effect on private banking growth, playing a complementary role tomonetary policy. Montoro & Moreno (2011) argue that RR have smaller impactsif the amount of deposits subject to RR relative to domestic bank credit is small.Glocker & Towbin (2012) find that RR have a role in supporting price stability if,among other conditions that are to some extent addressed in our model, debt isdenominated in foreign currency.
Few studies analyze the aggregate impact of RR in Brazil. Souza-Rodrigues & Takeda(2004) find empirical evidence that higher unremunerated reserve requirements inBrazil increase the mean of lending rates. Areosa & Coelho (2013) build a DSGEmodel with agency problems in banks’ funding and find that RR have qualitativelyequivalent (yet weaker) impact on the economy compared to the monetary policyinstrument. Our model differs from Areosa & Coelho (2013) in several importantways. Apart from a more comprehensive description of the financial sector, our modelfeatures default in loans to the real sector, whereas Areosa & Coelho (2013) introducedefault in bank deposits. An immediate consequence of their assumption is that therewill be a wedge between banks’ cost of funding from deposits and the base rate, drivenby solvency concerns. We choose not to adhere to this assumption since the spreadbetween 90-day bank certificates of deposits (CDB) and the effective base rate (Selic)has been negligible after the implementation of the inflation targeting regime (0.2 p.p.from a nominal quarterly base rate of 3.6% on average), despite strong movementsin volumes. This evidence also challenges the assumption extensively used in theliterature36 that banks have monopolistic power in setting deposit rates. In thisrespect, there are a number of investment opportunities that compete with demanddeposits in Brazil. Another important difference from Areosa & Coelho (2013) is thatin their model RR can only affect the economy through price effects, since they aredominated in return by public bonds. Instead, if RR were fully remunerated, as isthe case with time deposits in Brazil, reserve requirements would be neutral to theeconomy. In our model, we find important quantitative effects of these instruments.
5.2 Capital requirement shocks
An unanticipated 1 p.p. increase in required capital, from 11% to 12% (Figure13), has important real effects that are triggered by changes in banks’ balance sheetcomposition and in their decisions with respect to the share of profits to be distributedto bank owners.
The transmission mechanism is as follows. Given that deviations from the requiredcapital are costly to the banks, the shadow cost of banks’ operations increases, andis passed through to lending rates. More expensive consumer loans reduce income
36Some examples are Roger & Vlcek (2011), Gerali et al. (2010) and Dib (2010).
40
available for consumption. The drop in consumption and capital investment thatfollows from consumer and commercial credit contraction is enough to drag downGDP. The impact on the labor market is such that even though housing lendingrates fall – and that results from our assumption that these rates are decided by thegovernment and are tightly linked to the base rate of the economy – the demand forhousing loans also falls.
Since banks can decide how much of their earnings will be distributed and how muchwill be retained, they also accommodate part of the cost implied by higher capitalrequirements by retaining profits. This, together with the drop in risky assets, allowsthem to improve their net worth position, and liquid assets increase. However, sincebanks hold a large amount of excess capital in the steady state, final compliance withthe capital requirement comes mostly from reducing this capital buffer, and not somuch from increasing the actual Basel ratio.
With respect to the impact of changes in capital requirement on different types ofloans, we find that the demand for collateralized loans is more sensitive to changes inlending rates. This, together with the fact that the risk weight of commercial loansin the CAR is lower than that of unsecured loans, causes the increase in lending ratesof commercial loans to be less than that of consumer loans and also to show lesspersistence.
These conclusions were obtained from the baseline model, where monetary policy isresponding to economic conditions by lowering the base rate. However, this responseis not strong enough to offset the impact of tighter capital regulation on the shadowprice of banks’ operations. Hence, even if monetary policy is kept unchanged after ashock to capital requirement ratios (Figure 14), bank funding costs, capital accumu-lation, and liquidity are not substantially changed compared to the baseline scenario.However, since monetary policy cannot alleviate the burden of tighter credit condi-tions on the real economy, a more pronounced drop in collateral value and in laborincome causes lending rates to increase further. The final drop in GDP is thereforemuch more severe as the impact of the shock builds up.
5.2.1 Anticipated vs. unanticipated announcements of changes in capital
requirements
The baseline model assumes that changes in capital requirements are not anticipated.However, regulatory changes of this nature are usually announced with a substantiallag to the implementation. To investigate whether announcements made in advanceof the implementation trigger any anticipatory behavior, we compare the impulseresponses of the model in two alternative scenarios: one in which the macroprudentialauthority announces a 1 p.p. increase in required capital to be implemented only 4quarters after the announcement, and the other in which the announcement is madetogether with the implementation. Figure 15 shows the results.
Announcements trigger an anticipatory behavior of banks: they immediately startto retain earnings and improve their capital adequacy ratios over the entire period.As a matter of fact, the announcement is more effective in reducing the banks’ riskexposure even after the shock hits. Economic agents also anticipate the impact of
41
the shock and the demand for loans becomes more sensitive to lending rates. As aresult, lending rates do not need to rise as much to curtail credit as when the shockis unanticipated. Real variables, such as GDP and inflation are impacted from start,but show smoother trajectories.
5.2.2 Countercyclical capital buffer
In the baseline model, capital requirement decisions are represented by an autore-gressive process with very strong persistence (0.999). This is a fair representation ofthe Brazilian regulatory framework during our sampled period, in which Brazil hadadhered to either Basel I or Basel II accords, and when capital requirement ratiosremained practically invariable. 37
However, the Basel III regulatory framework introduces a countercyclical capitalbuffer that has been more commonly interpreted as a rule that requires banks tobuild up capital during expansionary credit cycles and loosens up in downturns. Thepurpose of this buffer is to dampen excessive oscillations in credit supply, which arecommonly associated with financial system distress.
We compare the impact of countercyclical capital buffers and traditional capital re-quirement under two different scenarios. In the first, we assume that banks reducetheir liquidity target to simulate a supply-driven credit expansion associated withloosened risk-taking standards. In the second, we simulate a severe bank capitalimpairment that can potentially depress the economy. The capital buffers can reacteither to contemporaneous (χE = 0) or expected (χE = 1) deviations of credit-to-GDP from its stationary trend:
γBankKt,cc = (1− ρBankK,cc)
(1− χE)(ωBBankK,cc
(BC,wbt +BC,wbt +BC,wbt
bC,wb+bC,wb+bC,wbgdpGDPt
))+χE
(ωBBankK,ccEt
(BC,wbt+4 +BC,wbt+4 +BC,wbt+4
bC,wb+bC,wb+bC,wbgdp
GDPt+4
)) (102)
+ρBankK,ccγBankKt−1,cc + εBankKt,cc
and γBankK,totalt = γBankKt + γBankKt,cc
Figures 16 and 17 show the model responses to a drop in bank’s liquidity target(νOMt ) such that total credit-to-GDP rises on impact by 1% from its stationary trendwhen traditional capital requirements are in effect38. To better isolate the effectof macroprudential policies, we shut down monetary policy by assuming that thebase rate remains constant. Compared with traditional capital requirements, if thecountercyclical capital buffer can be immediately adjusted to react to credit expansion
37Although Basel II includes operational and market risk in the computation of capital adequacyratios, we believe credit risk was the most preponderant factor in capital requirement rules, so thatour rule is a reasonable approximation when associated with the CAR equation (52).
38For this exercise, we set ρBankK,cc = 0.8 and ωBBankK,cc = 20, which implies a 1 p.p. rise in
total capital requirement (γBankK,totalt ) on impact.
42
(Figure 16), it drives down the variance of GDP by over 80% when the regulatoryauthority responds to contemporaneous credit conditions and by about 75% whenthe rule is forward looking, reacting to credit conditions four-periods ahead. Thecountercyclical capital buffers also mitigate the expansionist impact of the liquidityshock on the real economy. We performed the same exercise but now assuming thatthe countercyclical buffer can only be altered 4 periods after the shock (Figure 17).This simulates the framework to be adopted in Brazil, which requires the regulatoryauthority to announce changes in the instrument one year in advance. In this case,the drop in the variance of credit-to-GDP is a little smaller but the proportionalitybetween the contemporaneous and the forward-looking rule is maintained.
The second exercise makes a strong case in favor of the countercyclical capital buffer.Figure 18 shows the model responses to a shock to dividend distribution that severelyimpairs bank capital, such that credit-to-GDP falls by 1% from its stationary trendunder traditional capital requirements. Using the same rules as in the previous ex-ercise, the introduction of countercyclical capital buffers drop the variance of credit-to-GDP by over 90%, and the difference in the types of countercyclical rules is notso relevant with respect to credit stabilization. If the implementation of the counter-cyclical buffer is lagged (Figure 19), the forward looking rule is a little more efficientin driving down credit variance. The countercyclical capital buffers also substantiallymitigate the effect of the negative shock to bank capital on the real economy.
In the particular case of Brazil, required capital buffers cannot exceed 2.5 p.p. abovethe standard capital requirement. Assuming the same smoothing coefficient, if weconstrained the maximum hike in required capital to 2.5 p.p., the variance of credit-to-GDP would drop by about 75%.
5.3 Risk weight shocks
Figures 20 and 21 show the impact of 10 a p.p. hike in the risk weight of consumerand commercial loans, respectively. The shocks have an immediate impact on thelending rate of their specific credit segment, reducing credit. Banks choose to retaindividends so as to avoid further deterioration in capital adequacy ratios. Net fundsreleased from the drop in loans are redirected to bank liquidity. Altogether, banksaccommodate the overall impact of risk weight shocks on their balance sheet byreleasing part of the capital buffer, which implies that the Basle ratio remains belowpre-shock values for a prolonged period of time. Tighter credit conditions impactconsumption and capital investment, depressing output.
With respect to risk weights of housing loans (Figure 22), the over regulation of thelending rate in this market shifts the main burden of the adjustment to the othercredit segments. Hence, banks increase lending rates of commercial and consumerloans and cut dividends so as to improve capital adequacy ratios. The contractionistimpact that follows worsens labor market conditions in such a way that the demandfor housing loans also drops, notwithstanding the fact that the lending rate falls bytracking the base rate.
43
5.4 Monetary policy under financial frictions
The estimated model features traditional shapes of the responses of the key macroe-conomic variables to a monetary policy shock (Figure 23)39. Notwithstanding, thefinancial frictions of the model imply more elaborate transmission channels. A 100bp shock to the nominal base rate reduces consumption, hours worked and outputthrough the traditional channels. Financial frictions reinforce the responses. Thereduction in labor income puts pressure on the delinquency rates of consumer loans,increasing final lending rates. Hence, the demand for consumer loans falls, and bor-rower’s consumption further adjusts to accommodate tighter funding conditions.
Worsened demand conditions reduce prices. In particular, the fall in the price ofcapital reduces the value of collateral put up for commercial loans, putting pressureon default rates and, consequently, on lending rates. This reduces the demand forinvestment loans, further depressing investment.
The monetary policy shock has important implications for the composition of banks’balance sheets. The increase in the base rate puts pressure on external and internalbank funding costs. The reduction in the stock of loans resulting from higher fundingcosts is accommodated through an expansion in bank liquidity and an increase in theshare of retained earnings. The recomposition of banks’ balance sheet towards saferassets and greater capital accumulation improve the capital adequacy ratio. Theprice of housing falls with depressed demand conditions, therefore lower collateralvalues reduce housing loans.
6 Comparative Analysis
6.1 Macroprudential vs. Monetary Policy Shocks
To better understand the differences in the responses of macroprudential instrumentscompared with monetary policy, we simulated scenarios in which the regulatory andmonetary authorities tighten macroprudential and monetary policy in the first fourquarters, phasing them out in the subsequent 4 periods. The magnitude of each shockon impact was the same used to plot the impulse response functions in the previoussession40. Figure 27 shows the results, and each response corresponds to a full-blownsimulation of the model given one particular shock.
Monetary policy has a stronger impact on the labor market and, consequently, onconsumption. Here, among the many policies considered in this exercise, only themonetary policy shock affects housing lending rates. This causes a strong drop in
39We present the IRFs of temporary technology and price markup shocks in the appendix (Figures24 and 25)
40Although this choice is arbitrary, we believe it is better than using the estimated varianceof macroprudential shocks since over the sampled period, there have not been changes to capitalrequirements and we do not have aggregate series of risk weights applied to compute capital adequacyratios that would conform to the credit segmentation that we used in this paper. Our choice seemsreasonable considering our understanding of policy choices in Brazil
44
housing investment, further depressing the demand for intermediate goods, and hencethe labor market. The final impact on GDP is stronger than any of the macropru-dential policies considered.
Capital requirements have the strongest impact on non-regulated credit markets (i.e.,consumer and commercial), followed by reserve requirements on time deposits41. In-creased reserve requirements on time deposits trigger a sizable adjustment of bankliquidity, whereas increased capital requirements bring about a strong cut in bankdividend distribution. Moreover, when the weight of a particular risky bank asset in-creases in the computation of capital adequacy ratios, banks show a clear preferencefor cutting off excess capital instead of building up new capital; hence the capitaladequacy ratio deteriorates. The opposite holds for reserve and capital requirements,and for monetary policy shocks. For these instruments, banks prefer to accumulatecapital, and actual Basel ratios improve.
6.2 Reserve requirements and the role of bank liquidity pref-
erences
The model features two transmission channels through which reserve requirementsimpact banking costs and the rest of the economy: the cost channel and the liquiditypreference (or targeted liquidity) channel. The first one emerges from the gap betweenthe remuneration of required reserves at the central bank and the opportunity cost ofbanks’ internal funds. If remuneration is lower than the opportunity cost, any increasein reserve requirement ratios will produce higher funding costs to banks, which will bepartially passed on to firms and households through higher lending rates. This reserverequirement transmission channel is traditionally mentioned in the literature (e.g.,Areosa & Coelho (2013)). However, if reserve requirement remuneration matchesthe bank’s opportunity costs, this channel might be muted, and the impact of theinstrument through this channel becomes negligible.
Nonetheless, some categories of reserve requirements in Brazil are remunerated at theshort term policy rate, reducing banks’ opportunity cost of keeping reserves at thecentral bank. However, other types of liquid assets have analogous remuneration butgive banks more flexibility to manage their portfolios. This is introduced in our modelthrough a liquidity target in the form of public bonds. By introducing liquidity targetsin bank’s optimization problem, the model is able to produce relevant responses topolicy changes in reserve requirements, even when they are remunerated. This can beseen in Figures 28 and 29, which compare the impact of a permanent 10 p.p. increasein reserve requirements on time and demand deposits in the baseline model and in analternative version of the model with no role for liquidity targets (i.e., by imposingχOM = 0).
In the absence of liquidity targets, the shock to reserve requirements on time depositscauses no impact on the economy or on banks’ balance sheets, except for the fact thatliquidity falls so as to cope with the new requirements without creating any further
41We did not include RR on demand and savings deposits in this exercise, since we have shownthat in the baseline model, RR on time deposits have the strongest impact
45
costs to the bank. With respect to reserve requirements on demand deposits, sincethey do not accrue interest, only the cost channel is present in the alternative model.The responses of the alternative model are considerably smaller than those of thebaseline model. Hence, we can conclude that most of the impact on lending ratesand on the real economy comes from the liquidity target channel.
6.3 Alternative borrowing constraints
Our model incorporates debt-to-income borrowing constraints associated with thepossibility of delinquency in consumer loans, which, in fact, exists in equilibrium.In this respect, the model follows Bernanke, Gertler & Gilchrist (1999) except thatwage income replaces physical assets – capital or housing – as collateral for loans tohouseholds. We single out the differences in the responses of our model, compared tothe main strands of the macroeconomic literature that incorporates financial frictionsin household decisions by simulating two variants of our model: one in which onlyhousing can serve as collateral for consumer loans (HM) and another in which debt-to-income constraints bind (WM)42.
In the version where housing is put up as collateral for consumer loans (HM), theform of borrowing constraints follows Iacoviello (2005) and Gerali et al. (2010) 43:
RL,CB,t B
CB,t +RL,H
B,t BHB,t ≤ γB,Ct EtPH,t+1 (1− δH)HB,t
In the second alternative version of the model, (WM), debt constraints are strictlytied to current wage income. This formulation follows Mendoza (2002):
RL,CB,t B
CB,t +RL,H
B,t BHB,t ≤ γB,Ct (1− τw,t)NB,tW
Nt
There is no default on household loans in Iacoviello (2005), Gerali et al. (2010) andMendoza (2002); hence, we rule out the possibility of default in the alternative ver-sions of our model presented in this exercise.
Figures 30, 31, and 32 show each model’s impulse responses of shocks to capitalrequirement, reserve requirement on demand deposits, and monetary policy.
Compared to WM, the possibility of default on consumer loans in the benchmarkmodel produces a stronger impact of macroprudential instruments on the real econ-omy. It also reflects on bank capital and dividend distribution. The assumptionabout the existence of default also has implications for the responses to a monetary
42To perform this exercise, we set the model parameters at the mean of the estimated poste-rior distribution of our benchmark model. The modifications in the model were restricted to theequations that we present in this subsection and to the consequent changes that these equationsimply in first order conditions of borrower’s optimization program. The financial frictions incorpo-rated in commercial loans (which are a tight variant of Bernanke, Gertler & Gilchrist (1999)) weremaintained in all exercises.
43In the original works, γB,Ct is set to 1. In this exercise, we need to relax this constraint so as to
pin down a reasonable value for the steady state of housing in the economy.
46
policy shock. In WM, the absence of default allows banks to both accumulate capitaland increase dividend distribution. This overestimates the impact of a contractionistmonetary policy stance on actual Basel ratios.
HM generates much stronger impact of macroprudential policy on housing investment.This leads to a sharper decline in credit-to-GDP, most especially in the segment ofconsumer credit. Notwithstanding, since the labor market is now only weakly con-nected to credit decisions and since we did not introduce habit in housing decisions–which has an important impact on the responses of aggregate consumption – the im-pact of macroprudential policies on consumption, on the labor market and on GDP issubstantially less contractionist in HM. Although HM implies a greater sensitivity ofthe value of housing collateral to lending rates, and that results in stronger responsesof credit to macroprudential policy, the spillover effect of strained credit conditionson the economy is much more subdued given the low linkages with the labor market.
7 Conclusion
This paper builds a DSGE model with matter-of-fact financial frictions to assess thetransmission channel of a set of selected macroprudential policy instruments. Banks’decisions on consumer loans are based on expectations with respect to borrowers’labor income. These loans are risky, and default exists even in equilibrium. Thisentails stronger impact of macroprudential policies on the real economy as comparedto traditional models of collateral constraints in consumer loans that do not considerthe possibility of default in this credit segment.
The model also features frictions in the optimal composition of banks’ balance sheet.Banks are assumed to have liquidity targets, and this amplifies the impact of macro-prudential policies in the credit market and in the real sector. Banks can also accom-modate the impact of changes in the regulatory framework by deciding on the shareof profits to be distributed or retained.
The model is estimated with Bayesian techniques using Brazilian data from the in-flation targeting regime. We analyze the transmission channel of the following policyinstruments: 1) traditional (Basle 1 and 2) core capital requirements, with anticipatedor unanticipated implementation, 2) Basle 3 capital buffers, 3) reserve requirementson demand deposits, savings deposits, time deposits, and ”additional” deposits, and4) risk-weights that are used in the computation of capital adequacy ratios.
We find that monetary policy has a substantially larger impact on the real economythan macroprudential instruments. On the other hand, reserve requirements on timedeposits and capital requirements have a potentially stronger impact on bank credit-to-GDP. Adjustment in bank balance sheets are very different depending on themacroprudential instrument being shocked. Increased reserve requirements on timedeposits trigger a sizable adjustment of banks’ liquid assets, whereas increased capitalrequirements bring about a strong retention of bank earnings. When the weight of aparticular risky bank asset increases in the computation of capital adequacy ratios,banks show a clear preference for cutting off excess capital instead of building up new
47
capital. The opposite holds for reserve and capital requirements, and for monetarypolicy shocks. For these instruments, actual Basel ratios improve.
Shocks to reserve requirements can have important effects on the real economy. Weshow that even when required reserves at the central bank are remunerated, theyhave a non-neutral effect on bank aggregates and on the economy. In particular, thesize of deposits in the economy is a key variable to determine the magnitude of theimpact of the shock to the financial sector and to the real economy. If incidence basesare equal in size, the opportunity cost of maintaining reserve requirements that donot accrue interest rate dominates the magnitude of the impact.
When the implementation of new capital requirements is preceded by an announce-ment, banks anticipate the impact of the new regulation by immediately improvingcapital adequacy. As a result, the economic effects of the shock can be seen longbefore the shock hits.
We simulated scenarios in which bank capital is severely impaired or in which banksreduce their liquidity targets, creating conditions for excessive credit expansion. Inthese two cases, countercyclical capital buffers can substantially mitigate the impactof the shocks on credit and on the real economy. We also analyze cases in whichannouncements precede the actual implementation of countercyclical capital buffers.
LATEX
48
References
ANDRES, Javier; ARCE Oscar; THOMAS, Carlos. Banking Competition, Collateral Con-straints and Optimal Monetary Policy. [S.l.], 2010. (Banco de Espana Working Papers,1001). Disponıvel em: 〈http://ideas.repec.org/p/bde/wpaper/1001.html〉.
AREOSA, Waldyr Dutra; COELHO, Christiano Arrigoni. Using a DSGE Model to Assessthe Macroeconomic Effects of Reserve Requirements in Brazil. [S.l.], 2013. Disponıvel em:〈http://ideas.repec.org/p/bcb/wpaper/303.html〉.
BERNANKE, Ben S.; GERTLER, Mark; GILCHRIST, Simon. The financial acceleratorin a quantitative business cycle framework. In: TAYLOR, J. B.; WOODFORD, M. (Ed.).Handbook of Macroeconomics. Elsevier, 1999, (Handbook of Macroeconomics, v. 1). cap. 21,p. 1341–1393. Disponıvel em: 〈http://ideas.repec.org/h/eee/macchp/1-21.html〉.
BRZOZA-BRZEZINA, Michal; KOLASA, Marcin; MAKARSKI, Krzysztof. The anatomyof standard dsge models with financial frictions. Journal of Economic Dynamics and Con-trol, v. 37, n. 1, p. 32–51, 2013. Disponıvel em: 〈http://ideas.repec.org/a/eee/dyncon/v37y2013i1p32-51.html〉.
CARVALHO, Fabia Aparecida de; AZEVEDO, Cyntia F. The incidence of reserve re-quirements in brazil: Do bank stockholders share the burden? Journal of Applied Eco-nomics, v. 0, p. 61–90, May 2008. Disponıvel em: 〈http://ideas.repec.org/a/cem/jaecon/v11y2008n1p61-90.html〉.
CHRISTIANO, Lawrence; ROSTAGNO, Massimo; MOTTO, Roberto. Financial Factorsin Economic Fluctuations. [S.l.], 2010. (Working Paper Series, 1192). Disponıvel em: 〈http://ideas.repec.org/p/ecb/ecbwps/20101192.html〉.
DIB, Ali. Capital Requirement and Financial Frictions in Banking: Macroeconomic Im-plications. [S.l.], 2010. (Bank of Canada Working Papers, 2010-26). Disponıvel em: 〈http://EconPapers.repec.org/RePEc:bca:bocawp:10-26〉.
FIORE, Fiorella de; TRISTANI, Oreste. Optimal monetary policy in a model of the creditchannel. The Economic Journal, Blackwell Publishing Ltd, v. 123, n. 571, p. 906–931, 2013.ISSN 1468-0297. Disponıvel em: 〈http://dx.doi.org/10.1111/j.1468-0297.2012.02558.x〉.
GERALI, Andrea; NERI, Stefano; SESSA, Luca; SIGNORETTI, Federico M. Creditand banking in a dsge model of the euro area. Journal of Money, Credit and Banking,v. 42, n. s1, p. 107–141, 09 2010. Disponıvel em: 〈http://ideas.repec.org/a/mcb/jmoncb/v42y2010is1p107-141.html〉.
GLOCKER, Christian; TOWBIN, Pascal. Reserve requirements for price and financialstability: When are they effective? International Journal of Central Banking, v. 8, n. 1, p.65–114, March 2012. Disponıvel em: 〈http://ideas.repec.org/a/ijc/ijcjou/y2012q1a4.html〉.
IACOVIELLO, Matteo. House prices, borrowing constraints, and monetary policy in thebusiness cycle. American Economic Review, v. 95, n. 3, p. 739–764, June 2005. Disponıvelem: 〈http://ideas.repec.org/a/aea/aecrev/v95y2005i3p739-764.html〉.
JIMENEZ, Gabriel; ONGENA, Steven; PEYDRO, Jose-Luis; SAURINA, Jesus. Macro-prudential policy, countercyclical bank capital buffers and credit supply: Evidence from theSpanish dynamic provisioning experiments. [S.l.], 2012. Disponıvel em: 〈http://ideas.repec.org/p/nbb/reswpp/201210-231.html〉.
49
MENDOZA, Enrique G. Credit, Prices, and Crashes: Business Cycles with a Sudden Stop.[S.l.], 2002. 335-392 p. (NBER Chapters). Disponıvel em: 〈http://ideas.repec.org/h/nbr/nberch/10639.html〉.
MESQUITA, Mario; TOROS, Mario. Brazil and the 2008 panic. In: BANK FOR IN-TERNATIONAL SETTLEMENTS. The Global Crisis and Financial Intermediation inEmerging Market Economies. 2011, (BIS Papers, 54). p. 113–120. Disponıvel em: 〈http://EconPapers.repec.org/RePEc:bis:bisbpc:54-06〉.
MONTORO, Carlos; MORENO, Ramon. The use of reserve requirements as a policyinstrument in latin america. BIS Quarterly Review, v. 1, March 2011. Disponıvel em:〈http://ideas.repec.org/a/bis/bisqtr/1103g.html〉.
MONTORO, Carlos; TOVAR, Camilo. Macroprudential tools: Assessing the implicationsof reserve requirements in a DSGE model. Incomplete Draft. 2010.
PARIES, Matthieu Darracq; SØRENSEN, Christoffer Kok; RODRIGUEZ-PALENZUELA,Diego. Macroeconomic propagation under different regulatory regimes: Evidence from anestimated dsge model for the euro area. International Journal of Central Banking, v. 7, n. 4,p. 49–113, December 2011. Disponıvel em: 〈http://ideas.repec.org/a/ijc/ijcjou/y2011q4a3.html〉.
QUEIROZ, Mardilson Fernandes. Comportamento diario do mercado brasileiro de reservasbancarias – nıvel e volatilidade – implicacoes na polıtica monetaria. In: ANPEC [BRAZIL-IAN ASSOCIATION OF GRADUATE PROGRAMS IN ECONOMICS]. Proceedings ofthe 32th Brazilian Economics Meeting. 2004. Disponıvel em: 〈http://ideas.repec.org/p/anp/en2004/096.html〉.
ROGER, Scott; VLCEK, Jan. Macroeconomic Costs of Higher Bank Capital and LiquidityRequirements. [S.l.], 2011. (IMF Working Papers, 11/103). Disponıvel em: 〈http://ideas.repec.org/p/imf/imfwpa/11-103.html〉.
SOUZA-RODRIGUES, Eduardo; TAKEDA, Tony. Reserve requirements and bank interestrate distribution in Brazil. In: ANPEC [BRAZILIAN ASSOCIATION OF GRADUATEPROGRAMS IN ECONOMICS]. Proceedings of the 32th Brazilian Economics Meeting.2004. Disponıvel em: 〈http://ideas.repec.org/p/anp/en2004/096.html〉.
TOVAR, Camilo; GARCIA-ESCRIBANO, M.; MARTIN, M. Credit Growth and the Effec-tiveness of Reserve Requirements and Other Macroprudential Instruments in Latin America.[S.l.], 2012. (IMF Working Papers, 12/142). Disponıvel em: 〈http://ideas.repec.org/p/imf/imfwpa/12-142.html〉.
50
1em
A Tables
51
Table 1: Steady state calibrations
Description Value
Values
gε GDP growth (% per annum) 3.4πC CPI Inflation (% per annum) 4.5iH Investment in housing (% of GDP) 3.0iK Investment in capital (% of GDP) 14.4g Government spending (% of GDP) 20.4DD Demand deposits (% of GDP) 3.4DT Time deposits (% of GDP) 20.9DS Savings deposits (% of GDP) 10.7BB,C Retail loans outstanding(% of GDP) 12.5BB,H Housing loans outstanding (% of GDP) 5.5
BBE Commercial loans outstanding (% of GDP) 13.8RL,B,c Nominal retail lending rate (% per annum) 34.3RL,B,h Nominal housing lending rate (% per annum) 7.4RL,E Nominal commercial lending rate (% per annum) 21.1τC Consumptio tax(%) 16.2τW Tax on labor income (%) 15.0τπ Tax on profits (%) 15.0
bankcap Bank capital (% of GDP) 13.0τRR,T Reserve requirement ratio on time deposits (%) 11.0τRR,S Reserve requirement ratio on savings deposits (%) 18.1τRR,D Reserve requirement ratio on demand deposits (%) 49.6τadic Additional reserve requirement on total deposits (%) 7.7τH Mininum required allocation of savings deposits in housing loans (%) 34.0
γbankK Capital requirement (%) 11.0
Parameters
σbank Bank’s inverse elasticity of intertemporal substitution 1.00βbank Bank’s discount factor 0.98δH Housing depreciation (% per annum) 4.00τχ1 Risk weight of retail loans 1.50τχ2 Risk weight of commercial loans 1.00τχ3 Risk weight of housing loans 0.90τχ4 Risk weight of bank’s liquid assets 0.00τB Tax on retail loans (%) 0.3τE Tax on commercial loans (%) 0.1µB,H Monitoring cost of housing loans 0.00µb,G Productivity growth rate 0.06µW Wage markup 1.1µD Price markup 1.1
sadm,B Adm. costs in retail lending (%) 0.6sadm,E Adm. costs in commercial lending (%) 0.3σB Std of the shock to borrowers’ labor income 0.2σE Std of the shock to entrepreneur’s collateral 0.6µB,g Response to debt in government spending rule 0.06CAR Basle index (%) 16.8
52
Tab
le2:
Est
imat
edP
aram
eter
san
dS
hock
s
Des
crip
tion
Pri
orD
istr
ibu
tion
Post
erio
rD
istr
ibu
tion
Dis
trib
uti
onM
ean
Std
Dev
Mea
nC
red
ible
set
Pre
fere
nce
an
dT
ech
nolo
gy
hS
Hab
itp
ersi
sten
ceB
eta
0.75
0.05
0.6
26
0.5
61
0.6
94
σL
Inver
seF
risc
hel
asti
city
ofla
bor
Gam
ma
1.50
0.10
1.4
45
1.2
78
1.6
08
σS
EoS
ofsa
vin
gsd
epos
its
Gam
ma
4.00
3.00
23.5
09
16.2
54
30.3
31
σD
EoS
ofdem
and
dep
osit
sG
amm
a4.
003.
00
8.9
68
4.7
80
12.9
28
η HE
oSb
etw
een
con
sum
pti
onan
dh
ousi
ng
Gam
ma
1.00
0.25
2.1
13
1.3
71
2.9
03
φu,2
Cap
ital
uti
liza
tion
cost
Gam
ma
0.20
0.15
0.0
99
0.0
40
0.1
54
ξ EA
dj.
cost
ofem
plo
ym
ent
toh
ours
Bet
a0.
750.
10
0.6
21
0.5
69
0.6
75
φK
Ad
j.co
stof
cap
ital
inves
tmen
tG
amm
a3.
000.
50
3.8
70
2.9
81
4.7
30
φH
Ad
j.co
stof
hou
sin
gin
ves
tmen
tG
amm
a3.
000.
50
3.9
41
3.0
65
4.8
20
Nom
inal
Rig
idit
ies
ξ DC
alvo
-p
rice
sB
eta
0.75
0.05
0.8
18
0.7
56
0.8
81
αW
Cal
vo-
wag
esB
eta
0.75
0.05
0.8
37
0.7
89
0.8
86
γD
Pri
cein
dex
atio
nB
eta
0.50
0.20
0.2
38
0.0
51
0.4
05
γW
Wag
ein
dex
atio
nB
eta
0.50
0.20
0.2
65
0.0
55
0.4
66
ξRE
Cal
vo-
com
mer
cial
cred
itin
tere
stra
teB
eta
0.50
0.20
0.2
76
0.0
65
0.4
70
ξRB,c
Cal
v o-
reta
ilcr
edit
inte
rest
rate
Bet
a0.
500.
20
0.3
61
0.1
24
0.5
88
Policy
rule
s
ρR
Inte
rest
rate
smoot
hin
gB
eta
0.70
0.03
0.8
07
0.7
79
0.8
37
γπ
Infl
atio
nco
effici
ent
Gam
ma
2.00
0.05
1.9
63
1.8
81
2.0
44
γY
Ou
tpu
tga
pco
effici
ent
Gam
ma
0.20
0.10
0.0
89
0.0
22
0.1
52
ρg
Gov
ern
men
tsp
end
ing
smoot
hin
gB
eta
0.70
0.20
0.7
09
0.5
71
0.8
43
Fin
an
cia
lFri
cti
on
s
χbankK,2
Cos
tw
.r.t
.ca
pit
alb
uff
erG
amm
a0.
060.
010.0
64
0.0
48
0.0
80
χb OM
Cos
tw
.r.t
.li
qu
idit
yb
uff
erG
amm
a0.
100.
05
0.0
54
0.0
30
0.0
76
χd,T
Cos
tw
.r.t
.ti
me
dep
osit
s/b
ank
liab
ilit
ies
Gam
ma
0.10
0.05
0.0
87
0.0
32
0.1
38
Con
tin
ued
on
nex
tpa
ge
53
Tab
le2
–(c
ont.
)
Des
crip
tion
Pri
orD
istr
ibu
tion
Post
erio
rD
istr
ibu
tion
Dis
trib
uti
onM
ean
Std
Dev
Mea
nC
red
ible
set
φT
Ad
just
men
tco
stof
tim
ed
epos
its
Gam
ma
0.20
0.10
0.2
92
0.1
35
0.4
48
Au
tore
gre
ssiv
esh
ock
s
ρεIK
Ad
j.co
stof
cap
ital
inves
tmen
tB
eta
0.50
0.10
0.4
82
0.3
26
0.6
34
ρεIH
Ad
j.co
stof
hou
sin
gin
ves
tmen
tB
eta
0.50
0.10
0.4
43
0.3
00
0.5
95
ρεB,S
Sav
ers’
pre
fere
nce
Bet
a0.
500.
100.7
57
0.6
39
0.8
68
ρεB,B
Bor
row
ers’
pre
fere
nce
Bet
a0.
500.
100.5
30
0.3
43
0.7
12
ρεA
Tem
por
ary
tech
nol
ogy
Bet
a0.
500.
100.7
79
0.7
03
0.8
52
ρεu
Cap
ital
uti
liza
tion
Bet
a0.
500.
10
0.7
33
0.6
35
0.8
28
ρµD
Pri
cem
arku
pB
eta
0.50
0.10
0.4
73
0.3
16
0.6
30
ρµW
Wag
em
arku
pB
eta
0.50
0.10
0.3
83
0.2
64
0.5
00
ρε
Per
man
ent
tech
nol
ogy
Bet
a0.
950.
030.9
68
0.9
41
0.9
95
Au
tore
gre
ssiv
efi
nan
cia
lsh
ock
s
ρεS,S
Pre
fere
nce
for
savin
gsd
epos
its
Bet
a0.
500.
200.9
58
0.9
30
0.9
88
ρRH
Hou
sin
glo
ans
IRsm
oot
hin
gB
eta
0.50
0.20
0.8
78
0.8
07
0.9
60
ρµR E
Mar
ku
pon
com
mer
cial
loan
sB
eta
0.50
0.10
0.6
09
0.5
02
0.7
20
ρµR B,C
Mar
ku
pon
reta
illo
ans
Bet
a0.
500.
100.8
91
0.8
46
0.9
44
ρεbankcap
Div
iden
dd
istr
ibu
tion
Bet
a0.
500.
20
0.6
79
0.4
97
0.8
64
ρσB
S.D
.of
risk
inre
tail
loan
sB
eta
0.50
0.20
0.5
36
0.2
57
0.8
07
ρσE
S.D
.of
risk
inco
mm
erci
allo
ans
Bet
a0.
500.
200.6
60
0.5
78
0.7
44
ρd,D
Pre
fere
nce
for
dem
and
dep
osit
sB
eta
0.70
0.20
0.9
14
0.8
50
0.9
83
ρd,T
Ad
j.co
stin
tim
ed
epos
its
Bet
a0.
700.
200.7
99
0.6
93
0.9
04
ργB,H
Hou
sin
glo
ans
deb
t-to
-in
com
eB
eta
0.90
0.05
0.9
74
0.9
59
0.9
92
ργE
LT
Vof
com
mer
cial
loan
sB
eta
0.90
0.05
0.9
88
0.9
80
0.9
97
ργB,C
Ret
ail
loan
sd
ebt-
to-i
nco
me
Bet
a0.
900.
050.9
73
0.9
59
0.9
86
ρIBrem
CA
Rex
ogen
ous
com
pon
ent
Bet
a0.
500.
200.9
57
0.9
29
0.9
89
ρφRS
Mar
kd
own
onsa
vin
gsIR
Bet
a0.
500.
200.9
61
0.9
35
0.9
90
ρεbankcap
Div
iden
dd
istr
ibu
tion
Bet
a0.
500.
20
0.6
79
0.4
97
0.8
64
Con
tin
ued
on
nex
tpa
ge
54
Tab
le2
–(c
ont.
)
Des
crip
tion
Pri
orD
istr
ibu
tion
Post
erio
rD
istr
ibu
tion
Dis
trib
uti
onM
ean
Std
Dev
Mea
nC
red
ible
set
Tra
dit
ion
al
shock
s
εRM
onet
ary
pol
icy
shock
Inv.
Gam
ma
0.01
Inf
0.0
16
0.0
13
0.0
19
εGG
over
nm
ent
spen
din
gsh
ock
Inv.
Gam
ma
0.01
Inf
0.0
07
0.0
06
0.0
08
εIK
Cap
ital
inves
t.ad
j.co
stIn
v.
Gam
ma
0.05
Inf
0.0
81
0.0
63
0.0
98
εIH
Hou
sin
gin
ves
t.ad
j.co
stIn
v.
Gam
ma
0.05
Inf
0.0
35
0.0
27
0.0
42
εβS
Sav
ers’
pre
fere
nce
shock
Inv.
Gam
ma
0.05
Inf
0.0
63
0.0
40
0.0
85
εβB
Bor
row
ers’
pre
fere
nce
shock
Inv.
Gam
ma
0.05
Inf
0.0
89
0.0
46
0.1
34
εAT
emp
orar
yte
chn
olog
ysh
ock
Inv.
Gam
ma
0.02
Inf
0.0
17
0.0
14
0.0
20
εuC
apit
alu
tili
zati
onsh
ock
Inv.
Gam
ma
0.02
Inf
0.0
18
0.0
14
0.0
22
εµD
Pri
cem
arku
psh
ock
Inv.
Gam
ma
0.03
Inf
0.0
46
0.0
29
0.0
62
εµW
Wag
em
arku
psh
ock
Inv.
Gam
ma
0.03
Inf
0.1
02
0.0
68
0.1
35
εZP
erm
anen
tte
chn
olog
ysh
ock
Inv.
Gam
ma
0.00
Inf
0.0
01
0.0
01
0.0
02
επIn
flat
ion
targ
etIn
v.
Gam
ma
0.01
Inf
0.0
05
0.0
04
0.0
06
Fin
an
cia
lsh
ock
s
εS,S
Pre
fere
nce
for
savin
gsd
epos
its
Inv.
Gam
ma
0.10
Inf
0.6
72
0.4
72
0.8
75
εRH
Mar
ku
pon
hou
sin
glo
ans
Inv.
Gam
ma
0.01
Inf
0.0
05
0.0
04
0.0
06
εµRE
Mar
ku
pon
com
mer
cial
loan
sIn
v.
Gam
ma
0.02
Inf
0.0
05
0.0
04
0.0
06
εµRB,C
Mar
ku
pon
reta
illo
ans
Inv.
Gam
ma
0.02
Inf
0.0
08
0.0
06
0.0
09
εbankK
Div
iden
dd
istr
ibu
tion
Inv.
Gam
ma
0.02
Inf
0.0
36
0.0
30
0.0
42
εσB
S.D
.of
risk
inre
tail
loan
sIn
v.
Gam
ma
0.10
Inf
0.0
71
0.0
53
0.0
90
εσE
S.D
.of
risk
inco
mm
erci
allo
ans
Inv.
Gam
ma
0.10
Inf
0.1
60
0.1
28
0.1
89
εD,S
Pre
fere
nce
for
dem
and
dep
osit
sIn
v.
Gam
ma
0.10
Inf
0.3
28
0.1
77
0.4
66
εd,T
Ad
j.co
stin
tim
ed
epos
its
Inv.
Gam
ma
0.05
Inf
0.0
43
0.0
26
0.0
60
εγB,H
Hou
sin
glo
ans
deb
t-to
-in
com
eIn
v.
Gam
ma
0.05
Inf
0.0
43
0.0
33
0.0
52
εγE
LT
Vof
com
mer
cial
loan
sIn
v.
Gam
ma
0.05
Inf
0.2
12
0.1
68
0.2
54
εγB,C
Ret
ail
loan
sd
ebt-
to-i
nco
me
Inv.
Gam
ma
0.05
Inf
0.0
41
0.0
33
0.0
50
εIB,rem
CA
Rex
ogen
ous
com
pon
ent
Inv.
Gam
ma
0.10
Inf
0.0
99
0.0
82
0.1
15
Con
tin
ued
on
nex
tpa
ge
55
Tab
le2
–(c
ont.
)
Des
crip
tion
Pri
orD
istr
ibu
tion
Post
erio
rD
istr
ibu
tion
Dis
trib
uti
onM
ean
Std
Dev
Mea
nC
red
ible
set
εφRS
Mar
kd
own
onsa
vin
gsIR
Inv.
Gam
ma
0.01
Inf
0.0
66
0.0
55
0.0
77
ετRR,T
RR
onti
me
dep
osit
sIn
v.
Gam
ma
0.02
Inf
0.0
29
0.0
24
0.0
33
ετRR,add
RR
onad
dit
ion
ald
epos
its
Inv.
Gam
ma
0.02
Inf
0.0
12
0.0
10
0.0
14
ετRR,S
RR
onsa
vin
gsd
epos
its
Inv.
Gam
ma
0.02
Inf
0.0
07
0.0
06
0.0
08
ετRR,D
RR
ond
eman
dd
epos
its
Inv.
Gam
ma
0.02
Inf
0.0
40
0.0
34
0.0
46
56
B Figures
Figure 1: Consumer and Commercial Loans as a share of GDP and Time Deposits
Note: Retail and Commercial Loans in this graph are outstanding balances of non-mandatoryloans in the Brazilian financial system, excluding BNDES.
57
C Observable Variables
Figure 2: Observable Variables
58
Figure 3: Observable Variables
59
D Priors and Posteriors
Figure 4: Priors and Posteriors
60
Figure 5: Priors and Posteriors
61
E The Transmission of Macroprudential and Mon-
etary Policy
0 5 10−0.04
−0.02
0
GDP(% ss dev)
0 5 10−0.04
−0.02
0
Inflation(4−Q % ss dev)
0 5 10−4
−2
0
Interest rate(bp, yearly)
0 5 10−0.04
−0.02
0
0.02
Consumption(% ss dev)
0 5 10−0.1
0
0.1
Government spending(% ss dev)
0 5 10−0.4
−0.2
0
Capital investment(% ss dev)
0 5 100
0.2
0.4
Housing investment(% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Hours(% ss dev)
0 5 10−0.04
−0.02
0
Employment(% ss dev)
0 5 10−0.03
−0.02
−0.01
0
Real wage(% ss dev)
0 5 10−0.02
0
0.02
0.04
Demand deposits(% ss dev)
0 5 100
0.5
1
1.5
Time deposits(% ss dev)
0 5 10−0.05
0
0.05
Savings deposits(% ss dev)
0 5 10−0.2
−0.1
0
Retail loans(% ss dev)
0 5 10−0.1
−0.05
0
Housing loans (% ss dev)
0 5 10−0.4
−0.2
0
Commercial loans(% ss dev)
0 5 100
10
20
Retail lending rate(bp, yearly)
0 5 10−4
−2
0
Housing lending rate(bp, yearly)
0 5 100
10
20
Commercial lending rate(bp, yearly)
0 5 10−0.04
−0.02
0
0.02
Delinq. in retail loans(pp)
0 5 10−0.04
−0.02
0
0.02
Delinq. in commercial loans(pp)
0 5 10−4
−2
0
Liquidity buffer(% ss dev)
0 5 100
0.1
0.2
Bank capital(% ss dev)
0 5 100
0.02
0.04
Basel ratio(pp)
0 5 10−1
−0.5
0
Bank dividend distr.(% ss dev)
0 5 100
5
10
15
Shock to RR on Demand Deposits(pp)
Figure 6: Shock to Reserve Requirement Ratio on Demand Deposits
62
0 5 10−0.2
−0.1
0
GDP(% ss dev)
0 5 10−0.2
−0.1
0
Inflation(4−Q % ss dev)
0 5 10−20
−10
0
Interest rate(bp, yearly)
0 5 10−0.1
0
0.1
Consumption(% ss dev)
0 5 10−0.05
0
0.05
Government spending(% ss dev)
0 5 10−1.5
−1
−0.5
0
Capital investment(% ss dev)
0 5 100
0.5
1
1.5
Housing investment(% ss dev)
0 5 10−0.4
−0.2
0
Hours(% ss dev)
0 5 10−0.2
−0.1
0
Employment(% ss dev)
0 5 10−0.2
−0.1
0
Real wage(% ss dev)
0 5 10−0.2
0
0.2
Demand deposits(% ss dev)
0 5 100
5
10
Time deposits(% ss dev)
0 5 100
0.2
0.4
Savings deposits(% ss dev)
0 5 10−1
−0.5
0
Retail loans(% ss dev)
0 5 10−1
−0.5
0
Housing loans (% ss dev)
0 5 10−3
−2
−1
0
Commercial loans(% ss dev)
0 5 100
50
100
Retail lending rate(bp, yearly)
0 5 10−20
−10
0
Housing lending rate(bp, yearly)
0 5 100
50
100
150
Commercial lending rate(bp, yearly)
0 5 10−0.2
0
0.2
Delinq. in retail loans(pp)
0 5 10−0.5
0
0.5
Delinq. in commercial loans(pp)
0 5 10−30
−20
−10
0
Liquidity buffer(% ss dev)
0 5 10−1
0
1
2
Bank capital(% ss dev)
0 5 100
0.2
0.4
Basel ratio(pp)
0 5 100
1
2
3
Bank dividend distr.(% ss dev)
0 5 100
5
10
15
Shock to RR on Time Deposits(pp)
Figure 7: Shock to Reserve Requirement Ratio on Time Deposits
63
0 5 10−0.2
−0.1
0
GDP(% ss dev)
0 5 10−0.1
−0.05
0
Inflation(4−Q % ss dev)
0 5 10−15
−10
−5
0
Interest rate(bp, yearly)
0 5 10−0.04
−0.02
0
0.02
Consumption(% ss dev)
0 5 10−0.1
0
0.1
Government spending(% ss dev)
0 5 10−1
−0.5
0
Capital investment(% ss dev)
0 5 100
0.5
1
Housing investment(% ss dev)
0 5 10−0.2
−0.1
0
Hours(% ss dev)
0 5 10−0.1
−0.05
0
Employment(% ss dev)
0 5 10−0.1
−0.05
0
Real wage(% ss dev)
0 5 10−0.1
0
0.1
Demand deposits(% ss dev)
0 5 100
2
4
Time deposits(% ss dev)
0 5 100
0.1
0.2
Savings deposits(% ss dev)
0 5 10−0.4
−0.2
0
Retail loans(% ss dev)
0 5 10−0.4
−0.2
0
Housing loans (% ss dev)
0 5 10−1.5
−1
−0.5
0
Commercial loans(% ss dev)
0 5 100
20
40
60
Retail lending rate(bp, yearly)
0 5 10−15
−10
−5
0
Housing lending rate(bp, yearly)
0 5 100
20
40
60
Commercial lending rate(bp, yearly)
0 5 10−0.1
0
0.1
Delinq. in retail loans(pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans(pp)
0 5 10−15
−10
−5
0
Liquidity buffer(% ss dev)
0 5 10−0.5
0
0.5
1
Bank capital(% ss dev)
0 5 100
0.1
0.2
Basel ratio(pp)
0 5 10−0.5
0
0.5
1
Bank dividend distr.(% ss dev)
0 5 100
5
10
15
Shock to RR on Savings Deposits(pp)
Figure 8: Shock to Reserve Requirement Ratio on Savings Deposits
64
0 5 10−0.4
−0.2
0
GDP (% ss dev)
0 5 10−0.4
−0.2
0
Inflation (4−Q % ss dev)
0 5 10−15
−10
−5
0
Interest rate (bp, yearly)
0 5 10−0.1
0
0.1
Consumption (% ss dev)
0 5 10−0.5
0
0.5
Government spending (% ss dev)
0 5 10−1.5
−1
−0.5
0
Capital investment (% ss dev)
0 5 100
0.5
1
Housing investment (% ss dev)
0 5 10−0.4
−0.2
0
Hours (% ss dev)
0 5 10−0.2
−0.1
0
Employment (% ss dev)
0 5 10−0.2
−0.1
0
Real wage (% ss dev)
0 5 10−0.2
0
0.2
Demand deposits (% ss dev)
0 5 100
2
4
6
Time deposits (% ss dev)
0 5 100
0.2
0.4
Savings deposits (% ss dev)
0 5 10−1
−0.5
0
Retail loans (% ss dev)
0 5 10−1
−0.5
0
Housing loans (% ss dev)
No MP reactionTaylor rule
0 5 10−2
−1.5
−1
−0.5
Commercial loans (% ss dev)
0 5 1020
40
60
80
Retail lending rate (bp, yearly)
0 5 10−20
−10
0
10
Housing lending rate (bp, yearly)
0 5 100
50
100
Commercial lending rate (bp, yearly)
0 5 10−0.1
−0.05
0
Delinq. in retail loans (pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans (pp)
0 5 10−20
−15
−10
−5
Liquidity buffer (% ss dev)
0 5 10−1
0
1
2
Bank capital (% ss dev)
0 5 100
0.2
0.4
Basel ratio (pp)
0 5 101
1.5
2
2.5
Bank dividend distr. (% ss dev)
0 5 10−1
0
1
Capital requirement (pp)
0 5 100
0.5
1x 10
−34RR on demand dep.
(pp)
0 5 10−2
−1
0
1x 10
−12RR on savings dep.
(pp)
0 5 1010
10
10
10
RR on time dep. (pp)
No MP reactionTaylor rule
Figure 9: The role of Monetary Policy behavior on the transmission mechanisms ofa shock to Reserve Requirement Ratio on Time Deposits
65
0 5 10−0.2
−0.1
0
GDP (% ss dev)
0 5 10−0.2
−0.1
0
Inflation (4−Q % ss dev)
0 5 10−10
−5
0
Interest rate (bp, yearly)
0 5 10−0.1
−0.05
0
Consumption (% ss dev)
0 5 10−0.2
0
0.2
Government spending (% ss dev)
0 5 10−1
−0.5
0
Capital investment (% ss dev)
0 5 100
0.5
1
Housing investment (% ss dev)
0 5 10−0.2
−0.1
0
Hours (% ss dev)
0 5 10−0.2
−0.1
0
Employment (% ss dev)
0 5 10−0.1
−0.05
0
Real wage (% ss dev)
0 5 10−0.1
0
0.1
Demand deposits (% ss dev)
0 5 100
1
2
3
Time deposits (% ss dev)
0 5 100
0.1
0.2
Savings deposits (% ss dev)
0 5 10−0.4
−0.2
0
Retail loans (% ss dev)
0 5 10−0.4
−0.2
0
Housing loans (% ss dev)
No MP reactionTaylor rule
0 5 10−1.5
−1
−0.5
0
Commercial loans (% ss dev)
0 5 1010
20
30
40
Retail lending rate (bp, yearly)
0 5 10−10
−5
0
5
Housing lending rate (bp, yearly)
0 5 100
20
40
60
Commercial lending rate (bp, yearly)
0 5 10−0.1
0
0.1
Delinq. in retail loans (pp)
0 5 10−0.1
0
0.1
Delinq. in commercial loans (pp)
0 5 10−10
−5
0
Liquidity buffer (% ss dev)
0 5 100
0.5
1
Bank capital (% ss dev)
0 5 100
0.1
0.2
Basel ratio (pp)
0 5 100
0.5
1
Bank dividend distr. (% ss dev)
0 5 10−1
0
1
Capital requirement (pp)
0 5 10−1
−0.5
0x 10
−34RR on demand dep.
(pp)
0 5 1010
10
10
10
RR on savings dep. (pp)
0 5 10−1
−0.5
0x 10
−12RR on time dep.
(pp)
No MP reactionTaylor rule
Figure 10: The role of Monetary Policy behavior on the transmission mechanisms ofa shock to Reserve Requirement Ratio on Saving Deposits
66
0 5 10−0.06
−0.04
−0.02
0
GDP (% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Inflation (4−Q % ss dev)
0 5 10−4
−2
0
Interest rate (bp, yearly)
0 5 10−0.04
−0.02
0
Consumption (% ss dev)
0 5 10−0.1
0
0.1
Government spending (% ss dev)
0 5 10−0.4
−0.2
0
Capital investment (% ss dev)
0 5 100
0.1
0.2
Housing investment (% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Hours (% ss dev)
0 5 10−0.04
−0.02
0
Employment (% ss dev)
0 5 10−0.04
−0.02
0
Real wage (% ss dev)
0 5 10−0.04
−0.02
0
0.02
Demand deposits (% ss dev)
0 5 100
0.5
1
Time deposits (% ss dev)
0 5 100
0.02
0.04
Savings deposits (% ss dev)
0 5 10−0.1
−0.05
0
Retail loans (% ss dev)
0 5 10−0.2
−0.1
0
Housing loans (% ss dev)
No MP reactionTaylor rule
0 5 10−0.4
−0.2
0
Commercial loans (% ss dev)
0 5 100
5
10
15
Retail lending rate (bp, yearly)
0 5 10−4
−2
0
Housing lending rate (bp, yearly)
0 5 100
5
10
15
Commercial lending rate (bp, yearly)
0 5 10−0.02
−0.01
0
0.01
Delinq. in retail loans (pp)
0 5 10−0.04
−0.02
0
0.02
Delinq. in commercial loans (pp)
0 5 10−3
−2
−1
Liquidity buffer (% ss dev)
0 5 100
0.1
0.2
Bank capital (% ss dev)
0 5 100
0.02
0.04
Basel ratio (pp)
0 5 10−0.7
−0.6
−0.5
Bank dividend distr. (% ss dev)
0 5 10−1
0
1
Capital requirement (pp)
0 5 109
10
11
RR on demand dep. (pp)
0 5 100
0.5
1x 10
−12RR on savings dep.
(pp)
0 5 10−1
−0.5
0x 10
−12RR on time dep.
(pp)
No MP reactionTaylor rule
Figure 11: The role of Monetary Policy behavior on the transmission mechanisms ofa shock to Reserve Requirement Ratio on Demand Deposits
67
0 5 10−0.4
−0.2
0
GDP (% ss dev)
0 5 10−0.4
−0.2
0
Inflation (4−Q % ss dev)
0 5 10−1
0
1
Interest rate (bp, yearly)
0 5 10−0.2
−0.1
0
Consumption (% ss dev)
0 5 10−0.2
0
0.2
Government spending (% ss dev)
0 5 10−1.5
−1
−0.5
0
Capital investment (% ss dev)
0 5 100
0.5
1
Housing investment (% ss dev)
0 5 10−0.4
−0.2
0
Hours (% ss dev)
0 5 10−0.2
−0.1
0
Employment (% ss dev)
0 5 10−0.2
−0.1
0
Real wage (% ss dev)
0 5 10−0.2
−0.1
0
Demand deposits (% ss dev)
0 5 100
2
4
6
Time deposits (% ss dev)
0 5 100
0.1
0.2
Savings deposits (% ss dev)
0 5 10−1
−0.5
0
Retail loans (% ss dev)
0 5 10−1
−0.5
0
Housing loans (% ss dev)
Shock to RR on time depositsShock to RR on savings depositsShock to RR on demand deposits
0 5 10−2
−1
0
Commercial loans (% ss dev)
0 5 1020
40
60
Retail lending rate (bp, yearly)
0 5 10−2
−1
0
1x 10
−10Housing lending rate
(bp, yearly)
0 5 100
50
100
Commercial lending rate (bp, yearly)
0 5 10−0.1
0
0.1
Delinq. in retail loans (pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans (pp)
0 5 10−15
−10
−5
Liquidity buffer (% ss dev)
0 5 10−0.5
0
0.5
1
Bank capital (% ss dev)
0 5 100
0.1
0.2
Basel ratio (pp)
0 5 10−4
−2
0
2
Bank dividend distr. (% ss dev)
0 5 10−1
0
1
Capital requirement (pp)
0 5 10−50
0
50
RR on demand dep. (pp)
0 5 10−10
0
10
20
RR on savings dep. (pp)
0 5 10−5
0
5
10
RR on time dep. (pp)
Shock to RR on time depositsShock to RR on savings depositsShock to RR on demand deposits
Figure 12: Comparing same scale shocks to Reserve Requirement Ratios
68
0 5 10−0.2
−0.1
0
GDP(% ss dev)
0 5 10−0.2
−0.1
0
Inflation(4−Q % ss dev)
0 5 10−20
−10
0
Interest rate(bp, yearly)
0 5 10−0.1
0
0.1
Consumption(% ss dev)
0 5 10−0.2
−0.1
0
Government spending(% ss dev)
0 5 10−1.5
−1
−0.5
0
Capital investment(% ss dev)
0 5 100
0.5
1
1.5
Housing investment(% ss dev)
0 5 10−0.4
−0.2
0
Hours(% ss dev)
0 5 10−0.2
−0.1
0
Employment(% ss dev)
0 5 10−0.1
−0.05
0
Real wage(% ss dev)
0 5 10−0.2
0
0.2
Demand deposits(% ss dev)
0 5 10−2
−1
0
Time deposits(% ss dev)
0 5 100
0.5
1
Savings deposits(% ss dev)
0 5 10−1
−0.5
0
Retail loans(% ss dev)
0 5 10−0.5
0
0.5
Housing loans (% ss dev)
0 5 10−3
−2
−1
0
Commercial loans(% ss dev)
0 5 100
50
100
Retail lending rate(bp, yearly)
0 5 10−20
−10
0
Housing lending rate(bp, yearly)
0 5 100
50
100
Commercial lending rate(bp, yearly)
0 5 10−0.2
−0.1
0
Delinq. in retail loans(pp)
0 5 10−0.5
0
0.5
Delinq. in commercial loans(pp)
0 5 100
2
4
6
Liquidity buffer(% ss dev)
0 5 10−2
0
2
4
Bank capital(% ss dev)
0 5 100
0.5
1
Basel ratio(pp)
0 5 10−15
−10
−5
0
Bank dividend distr.(% ss dev)
0 5 100
0.5
1
Capital Requirement Shock(pp)
Figure 13: Capital Requirement Shock
69
0 5 10−0.4
−0.2
0
GDP (% ss dev)
0 5 10−0.2
−0.1
0
Inflation (4−Q % ss dev)
0 5 10−15
−10
−5
0
Interest rate (bp, yearly)
0 5 10−0.1
0
0.1
Consumption (% ss dev)
0 5 10−0.4
−0.2
0
Government spending (% ss dev)
0 5 10−1.5
−1
−0.5
0
Capital investment (% ss dev)
0 5 100
0.5
1
1.5
Housing investment (% ss dev)
0 5 10−0.4
−0.2
0
Hours (% ss dev)
0 5 10−0.2
−0.1
0
Employment (% ss dev)
0 5 10−0.2
−0.1
0
Real wage (% ss dev)
0 5 10−0.1
0
0.1
Demand deposits (% ss dev)
0 5 10−1
−0.5
0
Time deposits (% ss dev)
0 5 100
0.2
0.4
Savings deposits (% ss dev)
0 5 10−0.5
−0.4
−0.3
Retail loans (% ss dev)
0 5 10−1
−0.5
0
Housing loans (% ss dev)
No MP reactionTaylor rule
0 5 10−2
−1
0
Commercial loans (% ss dev)
0 5 1040
60
80
Retail lending rate (bp, yearly)
0 5 10−20
−10
0
10
Housing lending rate (bp, yearly)
0 5 100
20
40
60
Commercial lending rate (bp, yearly)
0 5 10−0.2
−0.1
0
Delinq. in retail loans (pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans (pp)
0 5 101
2
3
4
Liquidity buffer (% ss dev)
0 5 100
1
2
Bank capital (% ss dev)
0 5 100
0.2
0.4
Basel ratio (pp)
0 5 10−10
−9
−8
−7
Bank dividend distr. (% ss dev)
0 5 100.98
0.99
1
Capital requirement (pp)
0 5 10−1.5
−1
−0.5
0x 10
−16RR on demand dep.
(pp)
0 5 100
2
4x 10
−12RR on savings dep.
(pp)
0 5 10−4
−2
0
2x 10
−12RR on time dep.
(pp)
No MP reactionTaylor rule
Figure 14: The role of Monetary Policy behavior on the transmission mechanisms ofa Capital Requirement Shock
70
2 4 6 8 10 12
−0.1
−0.05
0
GDP (% ss dev) ()
2 4 6 8 10 12
−0.1
−0.05
0
Inflation (4−Q % ss dev) ()
2 4 6 8 10 12
−0.05
0
0.05
Consumption (% ss dev) ()
2 4 6 8 10 12
−1
−0.5
0
Capital investment (% ss dev) ()
2 4 6 8 10 12−0.15
−0.1
−0.05
0
Employment (% ss dev)
2 4 6 8 10 12−0.08−0.06−0.04−0.02
0
Real wage (% ss dev)
2 4 6 8 10 12−1
−0.5
0
Credit to GDP ratio (% ss dev)
2 4 6 8 10 12
−200
20406080
Retail IR (bp, yearly)
2 4 6 8 10 12
−200
204060
Commercial IR (bp, yearly)
2 4 6 8 10 12
02468
Liquidity (% ss dev)
2 4 6 8 10 12
−10
−5
0
Dividend distribution (% ss dev)
2 4 6 8 10 12
0
2
4
6
Bank capital (% ss dev) ()
2 4 6 8 10 12
0
0.5
1
Basel ratio (pp)
2 4 6 8 10 12
0
0.5
1
Capital requirement (pp)
Unanticipated shock to capital requirementAnticipated shock to capital requirement
Figure 15: Anticipated x Non-anticipated capital requirement shocks
71
5 10 15 200
0.5
1GDP (% ss dev)
5 10 15 200
0.5
1Inflation (4−Q % ss dev)
5 10 15 200
0.2
0.4Consumption (% ss dev)
5 10 15 20−1
0
1
2Capital investment (% ss dev)
5 10 15 20−0.5
0
0.5Employment % ss dev
5 10 15 200
0.2
0.4Real wages (% ss dev)
5 10 15 20−1
0
1
2Credit to GDP ratio % ss dev
5 10 15 20−400
−200
0
200Retail lending rate bp, yearly
5 10 15 20−400
−200
0
200Commercial lending rate bp, yearly
5 10 15 20−30
−20
−10
0Bank liquidity (% ss dev)
5 10 15 20−4
−3
−2
−1Bank dividends % ss dev
5 10 15 20−4
−2
0
2Bank capital (% ss dev)
5 10 15 20−1
−0.5
0Basel ratio pp
5 10 15 20−1
0
1
2Capital requirement pp
BenchmarkCC capital bufferCC capital buffer: forward looking rule
Figure 16: Comparing Countercyclical Capital Buffers after a Shock to Bank Liquid-ity Target
5 10 15 200
0.5
1GDP (% ss dev)
5 10 15 200
0.5
1Inflation (4−Q % ss dev)
5 10 15 200
0.2
0.4Consumption (% ss dev)
5 10 15 20−1
0
1
2Capital investment (% ss dev)
5 10 15 200
0.2
0.4Employment % ss dev
5 10 15 200
0.2
0.4Real wages (% ss dev)
5 10 15 20−1
0
1
2Credit to GDP ratio % ss dev
5 10 15 20−400
−200
0
200Retail lending rate bp, yearly
5 10 15 20−400
−200
0
200Commercial lending rate bp, yearly
5 10 15 20−30
−20
−10
0Bank liquidity (% ss dev)
5 10 15 20−4
−3
−2
−1Bank dividends % ss dev
5 10 15 20−4
−2
0
2Bank capital (% ss dev)
5 10 15 20−1
−0.5
0Basel ratio pp
5 10 15 20−1
0
1
2Capital requirement pp
BenchmarkCC capital bufferCC capital buffer: forward looking rule
Figure 17: Delayed Implementation of Countercyclical Capital Buffers after a Shockto Bank Liquidity Target
72
5 10 15 20−1.5
−1
−0.5
0GDP (% ss dev)
5 10 15 20−1
−0.5
0Inflation (4−Q % ss dev)
5 10 15 20−0.5
0
0.5Consumption (% ss dev)
5 10 15 20−10
−5
0Capital investment (% ss dev)
5 10 15 20−1
−0.5
0Employment % ss dev
5 10 15 20−1
−0.5
0Real wages (% ss dev)
5 10 15 20−5
0
5Credit to GDP ratio % ss dev
5 10 15 20−200
0
200
400Retail lending rate bp, yearly
5 10 15 20−100
0
100
200Commercial lending rate bp, yearly
5 10 15 20−60
−40
−20
0Bank liquidity (% ss dev)
5 10 15 20−50
0
50Bank dividends % ss dev
5 10 15 20−60
−40
−20
0Bank capital (% ss dev)
5 10 15 20−8
−6
−4
−2Basel ratio pp
5 10 15 20−4
−2
0Capital requirement pp
BenchmarkCC capital bufferCC capital buffer: forward looking rule
Figure 18: Comparing Countercyclical Capital Buffers after a Negative Shock to BankCapital
5 10 15 20−1.5
−1
−0.5
0GDP (% ss dev)
5 10 15 20−1
−0.5
0Inflation (4−Q % ss dev)
5 10 15 20−0.5
0
0.5Consumption (% ss dev)
5 10 15 20−10
−5
0Capital investment (% ss dev)
5 10 15 20−1
−0.5
0Employment % ss dev
5 10 15 20−1
−0.5
0Real wages (% ss dev)
5 10 15 20−6
−4
−2
0Credit to GDP ratio % ss dev
5 10 15 200
100
200
300Retail lending rate bp, yearly
5 10 15 200
100
200
300Commercial lending rate bp, yearly
5 10 15 20−60
−40
−20
0Bank liquidity (% ss dev)
5 10 15 20−50
0
50Bank dividends % ss dev
5 10 15 20−60
−40
−20
0Bank capital (% ss dev)
5 10 15 20−8
−6
−4
−2Basel ratio pp
5 10 15 20−4
−2
0Capital requirement pp
BenchmarkCC capital bufferCC capital buffer: forward looking rule
Figure 19: Delayed Implementation of Countercyclical Capital Buffers after a Nega-tive Shock to Bank Capital
73
0 5 10−0.04
−0.02
0
GDP(% ss dev)
0 5 10−0.04
−0.02
0
Inflation(4−Q % ss dev)
0 5 10−4
−2
0
Interest rate(bp, yearly)
0 5 10−0.06
−0.04
−0.02
0
Consumption(% ss dev)
0 5 10−0.02
−0.01
0
0.01
Government spending(% ss dev)
0 5 10−0.2
−0.1
0
Capital investment(% ss dev)
0 5 100
0.5
1
Housing investment(% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Hours(% ss dev)
0 5 10−0.04
−0.02
0
Employment(% ss dev)
0 5 10−0.02
−0.01
0
0.01
Real wage(% ss dev)
0 5 10−0.04
−0.02
0
Demand deposits(% ss dev)
0 5 10−0.2
−0.1
0
Time deposits(% ss dev)
0 5 100
0.1
0.2
Savings deposits(% ss dev)
0 5 10−0.4
−0.2
0
Retail loans(% ss dev)
0 5 10−0.2
0
0.2
Housing loans (% ss dev)
0 5 10−0.2
0
0.2
Commercial loans(% ss dev)
0 5 100
20
40
60
Retail lending rate(bp, yearly)
0 5 10−4
−2
0
Housing lending rate(bp, yearly)
0 5 10−5
0
5
10
Commercial lending rate(bp, yearly)
0 5 10−0.1
−0.05
0
Delinq. in retail loans(pp)
0 5 10−0.02
0
0.02
Delinq. in commercial loans(pp)
0 5 100
0.5
1
Liquidity buffer(% ss dev)
0 5 10−0.5
0
0.5
Bank capital(% ss dev)
0 5 10−0.4
−0.2
0
Basel ratio(pp)
0 5 10−1
−0.5
0
Bank dividend distr.(% ss dev)
0 5 100
5
10
Shock to Risk Weight of Retail Loans in CAR(% ss dev)
Figure 20: Sectoral Risk Weight Shock to Credit for Consumption
74
0 5 10−0.1
−0.05
0
GDP(% ss dev)
0 5 10−0.1
−0.05
0
Inflation(4−Q % ss dev)
0 5 10−10
−5
0
Interest rate(bp, yearly)
0 5 10−0.05
0
0.05
0.1
Consumption(% ss dev)
0 5 10−0.04
−0.02
0
0.02
Government spending(% ss dev)
0 5 10−1
−0.5
0
Capital investment(% ss dev)
0 5 100
0.5
1
Housing investment(% ss dev)
0 5 10−0.2
−0.1
0
Hours(% ss dev)
0 5 10−0.1
−0.05
0
Employment(% ss dev)
0 5 10−0.1
−0.05
0
Real wage(% ss dev)
0 5 100
0.1
0.2
Demand deposits(% ss dev)
0 5 10−1
−0.5
0
Time deposits(% ss dev)
0 5 100
0.1
0.2
Savings deposits(% ss dev)
0 5 10−0.2
0
0.2
Retail loans(% ss dev)
0 5 10−0.2
−0.1
0
Housing loans (% ss dev)
0 5 10−2
−1
0
Commercial loans(% ss dev)
0 5 10−10
0
10
Retail lending rate(bp, yearly)
0 5 10−10
−5
0
Housing lending rate(bp, yearly)
0 5 100
20
40
60
Commercial lending rate(bp, yearly)
0 5 10−0.02
0
0.02
Delinq. in retail loans(pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans(pp)
0 5 100
1
2
Liquidity buffer(% ss dev)
0 5 10−0.5
0
0.5
Bank capital(% ss dev)
0 5 10−0.4
−0.2
0
Basel ratio(pp)
0 5 10−1.5
−1
−0.5
0
Bank dividend distr.(% ss dev)
0 5 100
5
10
Shock to Risk Weight of Commercial Loans in CAR(% ss dev)
Figure 21: Sectoral Risk Weight Shock to Credit for Investment
75
0 5 10−0.01
−0.005
0
GDP(% ss dev)
0 5 10−10
−5
0
5x 10
−3Inflation
(4−Q % ss dev)
0 5 10−1
−0.5
0
0.5
Interest rate(bp, yearly)
0 5 10−4
−2
0
2x 10
−3Consumption
(% ss dev)
0 5 10−0.01
−0.005
0
Government spending(% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Capital investment(% ss dev)
0 5 100
0.02
0.04
0.06
Housing investment(% ss dev)
0 5 10−10
−5
0
5x 10
−3Hours
(% ss dev)
0 5 10−0.01
−0.005
0
Employment(% ss dev)
0 5 10−4
−2
0
2x 10
−3Real wage(% ss dev)
0 5 10−5
0
5
10x 10
−3Demand deposits
(% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Time deposits(% ss dev)
0 5 100
0.01
0.02
Savings deposits(% ss dev)
0 5 10−0.04
−0.02
0
Retail loans(% ss dev)
0 5 10−0.02
−0.01
0
0.01
Housing loans (% ss dev)
0 5 10−0.1
−0.05
0
Commercial loans(% ss dev)
0 5 100
2
4
6
Retail lending rate(bp, yearly)
0 5 10−1
−0.5
0
0.5
Housing lending rate(bp, yearly)
0 5 100
2
4
Commercial lending rate(bp, yearly)
0 5 10−0.01
−0.005
0
Delinq. in retail loans(pp)
0 5 10−10
−5
0
5x 10
−3Delinq. in commercial loans
(pp)
0 5 10−0.5
0
0.5
Liquidity buffer(% ss dev)
0 5 10−0.2
0
0.2
Bank capital(% ss dev)
0 5 10−0.2
−0.1
0
Basel ratio(pp)
0 5 10−1
−0.5
0
Bank dividend distr.(% ss dev)
0 5 100
5
10
15
Shock to Risk Weight of Housing Loans in CAR(% ss dev)
Figure 22: Sectoral Risk Weight Shock to Credit for Housing
76
0 5 10−1
−0.5
0
GDP(% ss dev)
0 5 10−1
−0.5
0
Inflation(4−Q % ss dev)
0 5 10−100
0
100
200
Interest rate(bp, yearly)
0 5 10−1
−0.5
0
0.5
Consumption(% ss dev)
0 5 10−1
−0.5
0
0.5
Government spending(% ss dev)
0 5 10−2
−1
0
1
Capital investment(% ss dev)
0 5 10−1
0
1
Housing investment(% ss dev)
0 5 10−1
−0.5
0
0.5
Hours(% ss dev)
0 5 10−0.5
0
0.5
Employment(% ss dev)
0 5 10−0.4
−0.2
0
Real wage(% ss dev)
0 5 10−2
−1
0
1
Demand deposits(% ss dev)
0 5 10−0.5
0
0.5
Time deposits(% ss dev)
0 5 10−1
−0.5
0
0.5
Savings deposits(% ss dev)
0 5 10−1
−0.5
0
0.5
Retail loans(% ss dev)
0 5 10−2
−1
0
Housing loans (% ss dev)
0 5 10−1
−0.5
0
0.5
Commercial loans(% ss dev)
0 5 10−100
0
100
200
Retail lending rate(bp, yearly)
0 5 10−100
0
100
200
Housing lending rate(bp, yearly)
0 5 10−100
0
100
200
Commercial lending rate(bp, yearly)
0 5 100
0.2
0.4
Delinq. in retail loans(pp)
0 5 10−0.2
0
0.2
Delinq. in commercial loans(pp)
0 5 100
0.5
1
1.5
Liquidity buffer(% ss dev)
0 5 10−0.5
0
0.5
1
Bank capital(% ss dev)
0 5 100
0.1
0.2
Basel ratio(pp)
0 5 10−1
−0.5
0
0.5
Bank dividend distr.(% ss dev)
0 5 10−100
0
100
200
Monetary Policy Shock(bp, yearly)
Figure 23: Monetary Policy Shock
77
0 5 10−0.5
0
0.5
GDP(% ss dev)
0 5 10−1
−0.5
0
0.5
Inflation(4−Q % ss dev)
0 5 10−100
−50
0
Interest rate(bp, yearly)
0 5 10−0.5
0
0.5
Consumption(% ss dev)
0 5 10−0.2
0
0.2
Government spending(% ss dev)
0 5 10−1
−0.5
0
0.5
Capital investment(% ss dev)
0 5 100
1
2
3
Housing investment(% ss dev)
0 5 10−2
−1
0
1
Hours(% ss dev)
0 5 10−1
−0.5
0
0.5
Employment(% ss dev)
0 5 10−0.5
0
0.5
Real wage(% ss dev)
0 5 100
0.5
1
1.5
Demand deposits(% ss dev)
0 5 100
0.5
1
Time deposits(% ss dev)
0 5 100
0.5
1
Savings deposits(% ss dev)
0 5 10−1
0
1
Retail loans(% ss dev)
0 5 10−1.5
−1
−0.5
0
Housing loans (% ss dev)
0 5 100
0.5
1
1.5
Commercial loans(% ss dev)
0 5 10−60
−40
−20
0
Retail lending rate(bp, yearly)
0 5 10−100
−50
0
Housing lending rate(bp, yearly)
0 5 10−50
0
50
Commercial lending rate(bp, yearly)
0 5 100
0.5
1
Delinq. in retail loans(pp)
0 5 100
0.1
0.2
Delinq. in commercial loans(pp)
0 5 10−2
−1
0
1
Liquidity buffer(% ss dev)
0 5 10−0.5
0
0.5
Bank capital(% ss dev)
0 5 10−0.1
0
0.1
Basel ratio(pp)
0 5 10−1
−0.5
0
Bank dividend distr.(% ss dev)
0 5 100
0.5
1
1.5
Temporary Technology Shock(% ss dev)
Figure 24: Temporary Technology Shock
78
0 5 10−0.2
0
0.2
GDP(% ss dev)
0 5 10−0.5
0
0.5
Inflation(4−Q % ss dev)
0 5 10−10
0
10
20
Interest rate(bp, yearly)
0 5 10−0.2
−0.1
0
Consumption(% ss dev)
0 5 100
0.1
0.2
Government spending(% ss dev)
0 5 10−0.5
0
0.5
Capital investment(% ss dev)
0 5 10−0.5
0
0.5
Housing investment(% ss dev)
0 5 10−0.2
0
0.2
Hours(% ss dev)
0 5 10−0.1
0
0.1
Employment(% ss dev)
0 5 10−0.4
−0.2
0
Real wage(% ss dev)
0 5 10−0.5
0
0.5
Demand deposits(% ss dev)
0 5 10−0.2
0
0.2
Time deposits(% ss dev)
0 5 10−0.2
0
0.2
Savings deposits(% ss dev)
0 5 10−0.5
0
0.5
Retail loans(% ss dev)
0 5 10−0.4
−0.2
0
Housing loans (% ss dev)
0 5 10−0.5
0
0.5
Commercial loans(% ss dev)
0 5 10−10
0
10
20
Retail lending rate(bp, yearly)
0 5 10−10
0
10
20
Housing lending rate(bp, yearly)
0 5 10−10
0
10
20
Commercial lending rate(bp, yearly)
0 5 10−0.02
0
0.02
Delinq. in retail loans(pp)
0 5 10−0.04
−0.02
0
0.02
Delinq. in commercial loans(pp)
0 5 10−0.5
0
0.5
Liquidity buffer(% ss dev)
0 5 10−0.1
−0.05
0
Bank capital(% ss dev)
0 5 10−0.01
0
0.01
0.02
Basel ratio(pp)
0 5 10−0.2
0
0.2
Bank dividend distr.(% ss dev)
0 5 100
0.5
1
1.5
Price Markup Shock(% ss dev)
Figure 25: Price Markup Shock
79
0 5 10−0.04
−0.02
0
GDP(% ss dev)
0 5 10−0.2
0
0.2
Inflation(4−Q % ss dev)
0 5 100
5
10
Interest rate(bp, yearly)
0 5 10−0.06
−0.04
−0.02
0
Consumption(% ss dev)
0 5 10−0.04
−0.02
0
0.02
Government spending(% ss dev)
0 5 10−0.05
0
0.05
Capital investment(% ss dev)
0 5 10−0.4
−0.2
0
Housing investment(% ss dev)
0 5 10−0.1
−0.05
0
Hours(% ss dev)
0 5 10−0.06
−0.04
−0.02
0
Employment(% ss dev)
0 5 100
0.1
0.2
Real wage(% ss dev)
0 5 10−0.2
−0.1
0
Demand deposits(% ss dev)
0 5 10−0.1
−0.05
0
Time deposits(% ss dev)
0 5 10−0.2
−0.1
0
Savings deposits(% ss dev)
0 5 10−0.2
0
0.2
Retail loans(% ss dev)
0 5 100
0.1
0.2
Housing loans (% ss dev)
0 5 10−0.2
−0.1
0
Commercial loans(% ss dev)
0 5 100
5
10
Retail lending rate(bp, yearly)
0 5 100
5
10
Housing lending rate(bp, yearly)
0 5 10−5
0
5
10
Commercial lending rate(bp, yearly)
0 5 10−0.06
−0.04
−0.02
0
Delinq. in retail loans(pp)
0 5 10−0.02
−0.01
0
Delinq. in commercial loans(pp)
0 5 10−0.2
0
0.2
Liquidity buffer(% ss dev)
0 5 10−0.02
0
0.02
0.04
Bank capital(% ss dev)
0 5 10−0.01
0
0.01
0.02
Basel ratio(pp)
0 5 100
0.05
0.1
Bank dividend distr.(% ss dev)
0 5 100
0.5
1
1.5
Wage Markup Shock(% ss dev)
Figure 26: Wage Markup Shock
80
F Macroprudential Policy vs Monetary Policy Shocks
;
2 4 6 8 10 12−0.8−0.6−0.4−0.2
0
GDP (% ss dev) ()
2 4 6 8 10 12−1
−0.5
0
Inflation (4−Q % ss dev) ()
2 4 6 8 10 12−0.8−0.6−0.4−0.2
00.2
Consumption (% ss dev) ()
2 4 6 8 10 12−1.5
−1−0.5
00.5
Capital investment (% ss dev) ()
2 4 6 8 10 12−1
0
1
2
Housing investment (% ss dev) ()
2 4 6 8 10 12−0.6
−0.4
−0.2
0
Employment (% ss dev)
2 4 6 8 10 12
−0.4
−0.2
0
Real wages (% ss dev) ()
2 4 6 8 10 12−0.6−0.4−0.2
00.20.4
Credit to GDP ratio (% ss dev)
2 4 6 8 10 12
0
50
100
Retail lending rate (bp, yearly)
2 4 6 8 10 12−50
0
50
100
Commercial lending rate (bp, yearly)
2 4 6 8 10 12−20
−10
0
Bank liquidity (% ss dev) ()
2 4 6 8 10 12−10
−5
0
Bank dividends (% ss dev)
2 4 6 8 10 12
0
0.5
1
Bank capital (% ss dev) ()
2 4 6 8 10 12
−0.2
0
0.2
Basel ratio (pp)
Monetary policy shockCapital requirement shock, no MPShock to the risk weight of commercial loans, no MPShock to risk weight of retail loans, no MPShock to RR on time deposits, no MP
2 4 6 8 10 12
0
50
100
Interest rate (bp, yearly)
2 4 6 8 10 12
0
5
10
Time dep. RR rate (pp)
2 4 6 8 10 12
0
0.5
1
Capital requirement (pp)
2 4 6 8 10 12
0
2
4
6
Risk Weight of Retail Loans in CAR (% ss dev)
2 4 6 8 10 12
0
5
10
Risk Weight of Commercial Loans in CAR (% ss dev)
Monetary policy shockCapital requirement shock, no MPShock to the risk weight of commercial loans, no MPShock to risk weight of retail loans, no MPShock to RR on time deposits, no MP
Figure 27: Comparing the Impact of Macroprudential Instruments and MonetaryPolicy
81
G Reserve Requirements and the Liquidity Target
5 10 15 20
−0.04
−0.02
0GDP (% ss dev)
5 10 15 20−0.1
−0.05
0Inflation (4−Q % ss dev)
5 10 15 20−0.04
−0.02
0Consumption (% ss dev)
5 10 15 20−0.4
−0.2
0Capital investment (% ss dev)
5 10 15 200
0.1
0.2Housing investment (% ss dev)
5 10 15 20−0.04
−0.02
0Employment % ss dev
5 10 15 20
−0.04
−0.02
0Real wages (% ss dev)
5 10 15 20−0.2
0
0.2Credit to GDP ratio % ss dev
5 10 15 20−20
0
20Retail lending rate bp, yearly
5 10 15 200
10
20Commercial lending rate bp, yearly
5 10 15 20−4
−2
0Bank liquidity (% ss dev)
5 10 15 20−1.5
−1
−0.5Bank dividends % ss dev
5 10 15 20−0.2
0
0.2Bank capital (% ss dev)
5 10 15 20−0.05
0
0.05Basel ratio pp
5 10 15 20−2
0
2x 10
−13Capital requirement pp
BenchmarkNo preference for liquidity
Figure 28: Removing the target for liquidity: permanent shock to reserve require-ments on time deposits
5 10 15 20−0.5
0
0.5GDP (% ss dev)
5 10 15 20−0.5
0
0.5Inflation (4−Q % ss dev)
5 10 15 20−0.1
0
0.1Consumption (% ss dev)
5 10 15 20−2
0
2Capital investment (% ss dev)
5 10 15 20−1
0
1Housing investment (% ss dev)
5 10 15 20−0.2
0
0.2Employment % ss dev
5 10 15 20−0.2
0
0.2Real wages (% ss dev)
5 10 15 20−1
0
1Credit to GDP ratio % ss dev
5 10 15 20−100
0
100Retail lending rate bp, yearly
5 10 15 20−100
0
100Commercial lending rate bp, yearly
5 10 15 20−40
−20
0Bank liquidity (% ss dev)
5 10 15 200
2
4Bank dividends % ss dev
5 10 15 20−2
0
2Bank capital (% ss dev)
5 10 15 20−0.5
0
0.5Basel ratio pp
5 10 15 20−1
0
1x 10
−12Capital requirement pp
BenchmarkNo preference for liquidity
Figure 29: Removing the target for liquidity: permanent shock to reserve require-ments on demand deposits
82
H Comparing Collateral Constraints
5 10 15 20
−0.2
−0.1
0
GDP (% ss dev)
5 10 15 20−0.2
−0.1
0
Inflation (4−Q % ss dev)
5 10 15 20
−0.05
0
0.05
Consumption (% ss dev)
5 10 15 20
−1
−0.5
0
Capital investment (% ss dev)
5 10 15 20
−4
−2
0
Housing investment (% ss dev)
5 10 15 20−0.2
−0.1
0
Employment % ss dev
5 10 15 20−0.15
−0.1
−0.05
0
Real wages (% ss dev)
5 10 15 20−3
−2
−1
0
Credit to GDP ratio % ss dev
5 10 15 20
020406080
Retail lending rate bp, yearly
5 10 15 20
0
20
40
60
Commercial lending rate bp, yearly
5 10 15 20
02468
Bank liquidity (% ss dev)
5 10 15 20−20
−10
0
Bank dividends % ss dev
5 10 15 20
0
2
4
Bank capital (% ss dev)
5 10 15 20
0
0.5
1Basel ratio pp
BenchmarkBinding debt−to−income constraintHousing collateral, no delinquency
Figure 30: Alternative Collateral Constraints: Capital Requirement Shock, Unre-sponsive Monetary Policy
5 10 15 20
−0.04
−0.02
0
GDP (% ss dev)
5 10 15 20
−0.04
−0.02
0
Inflation (4−Q % ss dev)
5 10 15 20
−0.03
−0.02
−0.01
0
Consumption (% ss dev)
5 10 15 20
−0.2
−0.1
0
Capital investment (% ss dev)
5 10 15 20
−0.2
0
0.2Housing investment (% ss dev)
5 10 15 20−0.04
−0.02
0
Employment % ss dev
5 10 15 20−0.04
−0.02
0
Real wages (% ss dev)
5 10 15 20−0.4
−0.2
0
Credit to GDP ratio % ss dev
5 10 15 20
0
5
10
15Retail lending rate bp, yearly
5 10 15 20
0
5
10
15Commercial lending rate bp, yearly
5 10 15 20
−3
−2
−1
0
Bank liquidity (% ss dev)
5 10 15 20−1.5
−1
−0.5
0
Bank dividends % ss dev
5 10 15 20−0.1
0
0.1
Bank capital (% ss dev)
5 10 15 20
0
0.02
0.04
Basel ratio pp
BenchmarkBinding debt−to−income constraintHousing collateral, no delinquency
Figure 31: Alternative Collateral Constraints: Shock to Reserve Requirement onDemand Deposits, Unresponsive Monetary Policy
83
5 10 15 20
−0.4
−0.2
0
GDP (% ss dev)
5 10 15 20
−0.4
−0.2
0
Inflation (4−Q % ss dev)
5 10 15 20−0.6
−0.4
−0.2
0
Consumption (% ss dev)
5 10 15 20−0.8−0.6−0.4−0.2
0
Capital investment (% ss dev)
5 10 15 20
−0.5
0
0.5
Housing investment (% ss dev)
5 10 15 20
−0.2
−0.1
0
Employment % ss dev
5 10 15 20−0.2
−0.1
0
Real wages (% ss dev)
5 10 15 20−0.6−0.4−0.2
00.20.4
Credit to GDP ratio % ss dev
5 10 15 20
0
50
100
Retail lending rate bp, yearly
5 10 15 20
0
50
100
Commercial lending rate bp, yearly
5 10 15 20
0
0.5
1
1.5
Bank liquidity (% ss dev)
5 10 15 20−1
0
1
Bank dividends % ss dev
5 10 15 20
00.20.40.60.8
Bank capital (% ss dev)
5 10 15 20
0
0.05
0.1
0.15
Basel ratio pp
BenchmarkBinding debt−to−income constraintHousing collateral, no delinquency
Figure 32: Alternative Collateral Constraints: Monetary Policy Shock
84