INTERBANK MARKET WITH DSGE BANK

WP/12/2014 WORKING PAPER INTERBANK MARKET WITH DSGE BANK Harmanta Aditya Rachmanto Fajar Oktiyanto Idham December, 2014 Conclusions, opinions, and views expressed by the authors in this paper are personal conclusions, opinions, and views of the authors and are not official conclusions, opinions, and views of Bank Indonesia.

INTERBANK MARKET WITH DSGE BANK Harmanta, Aditya Rachmanto, Fajar Oktiyanto, Idham Abstract In this research DSGE model is built for Indonesian small open economy which has been completed with interbank market mechanism to illustrate financial frictions from bank s supply side. Within the supply there exists portfolio optimization mechanism by banks, which is optimization in channeling loans or depositing in risk free asset. Meanwhile, financial friction from demand side is modelled by collateral constraint and financial accelerator. Banking sector in the model is also designed to conduct simulation of monetary policy mix (BI rate and exchange rate) and macroprudential policy (CAR requirement and LTV ratio requirement). Simulation result shows that shock in interbank market will affect bank conditions in general, especially in bank capital, CAR, and loan to deposit ratio (LDR). Bank s balance sheet conditions will affect the real sector. This model can also capture procylicality and financial accelerator happening due to financial frictions in the economy. GDP will become higher during expansion phase if compared to condition without financial frictions, and vice versa, GDP will be lower during contraction phase. Contraction in the economy will be responded by banks by reducing loan disbursement level, caused by high risk faced by banks, which will also increase lending rate so that entrepreneur will have more difficulties in securing loans. This condition causes banks to push down loan disbursement to avoid its capital shrinking. Simulation result also shows that shock in form of monetary and macroprudential policy mix will result in more stable GDP and inflation dynamics compared to using only one policy instrument. Key words : monetary policy, DSGE with banking sector, macroprudential policy JEL Classification : E32, E44, E52, E58 1

I. PREFACE Global financial crisis happening currently underlines the need to develop DSGE model which has explicit relations between real and financial sectors as well as an active banking sector. Model with such capacity will allow empiric evaluations on the role and behavior of banks in transmitting shock coming from supply or demand sides. However, literatures on DSGE modelling used in making policy formulation mostly ignore the banking sector. The global financial crisis that happened provides lessons on the importance of relations between real and financial sectors in DSGE model as the focus of attention. In Indonesia empirical research finds that procylicality from financial sector in Indonesia is considerably high. A research from Agung (2010) shows that real credit growth is faster than GDP in expansion period. On the other hand, bigger real credit decline than GDP decline happens in contracting period. High procylicality in banking sector in Indonesia shows the need to synergize monetary policy and macroprudential policy to mitigate economic fluctuation (business cycle) and excessive financial cycle. Harmanta et al. (2013) research has modelled financial sector procyclicality using financial accelerator in accordance with BGG (1999) in entrepreneur agents and collateral contraints in household agents. The research has modelled sector designed according to Indonesian conditions and can conduct monetary policy (BI rate) and exchange rate as well as macroprudential policy simulation in financial institutions, in this case banking, in form of simulations on CAR requirement changes and LTV ratio requirement for household. However, the research still uses homogeneous agent to represent the banking sector so that financial frictions in new model happened in one side of credit market, which is the demand side as modelled by financial accelerator and collateral constraints mechanism. Meanwhile, financial friction from supply side of credit market is not yet modelled. Several researches in latest literatures emphasize on the importance of modelling supply side of credit market which can provide vital information on interbank transmissions as well as the relations with financial authority and central bank. Moreover, in crisis, supply side of credit market has important role in spreading the crisis taking place. This paper follows up the research of Harmanta et al. (2013) with main development on supply side of credit market. By considering the structure of 2

interbank market in Indonesia, banking sector development can be done by following Ali Dib (2009): there are two heterogeneous agents in the banking sector which offer different banking services and interacting in a market called interbank market. The aim of this research is to develop DSGE model for banking sector with financial friction, both collateral constraint and financial accelerator, as well as adding interbank market mechanism for monetary or macroprudential policies simulation needs. The benefit of this research is as follows; a. As one of the tools in making formulation of monetary and macroprudential policy mix which will be stipulated by Bank Indonesia. b. As a step in competence building in developing DSGE model with simulation feature of monetary and macroprudential policy mix for FPAS core model development in the future (according to best practice from advanced countries which currently have adopted DSGE-based core model). This research is outlined as follows. Section 1 describes preface, aim, and benefit of the research. Section 2 is a short review on related literatures. Section 3 explains the detail of developed DSGE model. Section 4 is the detail of estimate and simulation. Section 5 is conclusion and further development plan. 3

II. LITERATURE REVIEW 2.1 Financial Friction Modelling in DSGE Model In literature the efforts to modelling financial system procyclicality among others is by introducing financial friction in DSGE model. To model friction which happens in demand side of credit market, there are two major approaches that can be widely accepted: collateral constraint and financial accelerator. Basic assumption of financial accelerator approach was first introduced by Bernanke, Gertler, and Gilchrist in 1999 (BGG), in which there is information asymmetry between borrowers and lenders that results in external finance premium which illustrates cost difference if borrowing compared to using own funding. External finance premium is determined by the amount of net worth from borrowers and will determine the amount of borrowings that can be accepted. Meanwhile, collateral constraint approach, as introduced by Kiyotaki and Moore (1997), is the movement of asset price which interacts with market imperfection (there is information asymmetry between creditors and debtors, such as debtors payment ability) causes a process which enlarges response from shocks. However, contrary to financial accelerator approach, borrower assets directly will affect the amount of borrowings that can accepted and is not through the influence to external finance premium. The drawback of financial frictions which only use financial accelerator or collateral constraint is both only model one side of the credit market, which is the demand side. Gertler and Kiyotaki (2009) develop banking sector framework which finds that liquidity shock in banks can cause interbank money market segmentation which eventually will cause spillover effect to the real sector. Based on that finding, they argue that interbank market should be included in DSGE model financial block because during financial crisis, interbank market has important role in spreading the crisis. Ali Dib (2009) develops fully micro founded closed economy DSGE model which incorporates explicit relations between real sector and financial sector as well as having active banking sector. It can be achieved by modelling optimization of banks and both sides of credit market (supply and demand) explicitly. Demand side of credit is modelled using financial accelerator in accordance with BGG (1999), while supply side of credit market is modelled by introducing assumption that there are 4

two types of heterogenous banks: savings bank and lending bank which offer different banking services and both interact in a market called interbank market. Figure 1. Financial Intermediaries Scheme of Ali Dib (2009) As in the scheme pictured above, Ali Dib models monopolistically competitive savings banks as deposit recipients from household workers and pay deposit interest rate, R D j,t. In allocating the portfolio, savings bank determines the optimal composition between lending through interbank market to lending banks which give interest rate R t IB and investing in risk-free government bond assets which can give interest of R t. In every period there is probability lending banks suffer default and cannot pay back their interbank borrowing. On the other hand, when making investment on risk free assets, savings bank must pay premium insurance (cost of using risk free assets). Monopolistically competitive lending banks are modelled borrowing from savings bank through interbank market and request bank capital from bankers by paying bank capital price Q t Z. Each lending bank can also receive liquidity injection from the central bank, m j,t, as well as financial intermediaries Γ t. Carrera and Vega (2012) use different approach in modelling interbank market. In the paper they assume there are two types of bank: retail bank and narrow bank. In this approach only primary dealers are allowed to have direct contact with the central bank in accordance with financial intermediaries condition in the United States. 5

Figure 2. Financial Intermediaries Scheme of Carrera and Vega (2012) Mechanism in the model illustrates retail bank collects deposit from households and borrows from interbank market to be channeled to entrepreneur in the form of loans, while narrow bank makes liquidity placement in interbank market which the source of funds come from equity issues. 2.2 Characteristic of Indonesian Economy and Banking Sector Indonesian economy shows constant growth in the past decade, with average GDP of 5.45% in the 2001 2013 period. The economy kept growing reaching its peak in 2011 when economic growth reached 6.49% year on year. The achievement is quite impressive compared to neighboring countries which suffer the 2007/2008 global crisis. On demand side, Indonesian economy is supported by private consumption which has share of 62.58% to total GDP, followed by investment of 26.80%, see Table 2. Persistent domestic consumption and high export share due to high demand of main export destinations, such as China and India especially for commodities and mining products provide important contributions for economic growth. Increasing share of investment from year to year brings economic progress by creating employment and revenue so as to maintain the level of private consumption. On production side, Indonesian economy is supported by the manufacturing industry which still has the largest share to Indonesian GDP and followed by trade, hotel, and restaurants. Increasing domestic private consumption and commodity export demand by exporting partner countries have boosted economic growth in various sectors. 6

Table 1. Indonesian GDP Component Growth Increasing revenue as well as manageable inflation and low interest rate also boost the growth of production sectors, such as construction and transportation. Other sectors that also rapidly develop are the financial, rental and service sector, as well as the service sector. Growth in those sectors impacted a total GDP increase of 6.23% in 2012. Tabel 2. Indonesian GDP Component Share One assumption implemented in banking sector in DSGE model by several central banks is that there is market power from banks in collecting or channeling 7

funds market so that banks have power in determining TPF deposit rate or interest rate. Several empirical researches in Indonesia show the same conclusion. One of them is Purwanto (2009) who concludes that the dynamic of bank interest rate spread (defined as the difference between interest rate from fund disbursement subtracted by interest rate from fund collection) is mainly affected by the dynamic of concentration level of banking industry in Indonesia. The research uses Herfindahl- Hirschman Index as a measure of concentration level from the banking industry. Based on empirical model estimates, monthly data of individual bank (panel) from January 2002 to April 2009 is used. The conclusion is that the decline of interest rate spread during estimated period is caused by increasing competitions in the banking sector due to the increase in market share from majority banks followed by the decline in market share from banks with big assets. This is in line with research that uses structure-conduct-performance approach which connect market concentration with market power and behavior of interest rate determination (Berger et al., 2004). Moreover, the DSGE model developed by various central banks also assumes there is stickiness in retail interest rate of banking if associated with dynamic from policy rate. From theoretical point of view, banks can view that it is optimal to not often change interest rate if consumer demand is inelastic in short term because of the high switching cost (Calem et al., 2006) or because of fixed cost (menu cost) in making interest rate changes (Berger and Hannan, 1991). Another theoretical reason that is also stated by economic experts is that there is banking interest to maintain relationship with customers so that banks conduct interest rate smoothing to protect customers from interest rate fluctuations (policy). This allows banks to determine high interest rate when policy rate is low (Berger and Udell, 1992). In simple, a rigid short term response from bank retail interest rate on policy rate dynamic has been discussed in previous research by Harmanta et al. (2012). Impulse response analysis is conducted to bivariate VAR system 1 which shows that short term response from bank retail interest rate to changes in BI rate is pretty limited, especially for consumer loan interest rate. Deposit rate and lending rate for corporates have similar response. Even though the value is not as small as consumer lending rate, a high stickiness level remains. 1 Each VAR system also consists of exogenous variables: the amount of reserve ratio for VAR from deposit rate; and the amount of capital, the weight of risky assets (RWA divided by total loans), and the amount of loans channeled for lending rate VAR. 8

With evidence that financial sector in Indonesia shows high procyclicality, such as reflected by real credit growth which follows GDP growth, and experience of 2008 global financial crisis, that maintaining inflation only is not enough to achieve macroeconomic stability, Bank Indonesia adopts flexible inflation targeting framework to be able to better maintain financial system stability. One of the efforts made is by taking an approach of mixing monetary policy (conventional policy) and macroprudential policy (unconventional policy). Instruments from macroprudential policy used among others are loan to value ratio (LTV) and reserve requirement (GWM). For LTV requirement, Bank Indonesia, considering that mortgage loans (KPR) and auto loans (KKB) grew above loan average, through SE No. 14/10/DPNP dated March 15 2012, effective June 15 2012, imposed LTV for KPR at maximum 70%, tighter than actual average of 82.5%. However, for KKB the rules are split as follows: KKB for two-wheel vehicle, with minimum DP of 25% (or LTV of 75%); KKB for fourwheel vehicle, minimum DP of 30% (vehicle for non-productive means) and 20% (vehicle for productive means); and KKB for public transports, minimum DP of 20%. Therefore, prior to the Bank Indonesia regulation, average LTV (KPR and KKB) was around 85% and became tighter to around 72.5% after the regulation was implemented (Figure 3). 40.0 % 35.0 30.0 25.0 20.0 15.0 10.0 5.0 0.0 % 90 85 80 75 70 65 Pert. Nominal Kredit Credit Nominal Growth GWM Primer LTV (average, rhs) Figure 3. Historical Data of Nominal Credit Growth, GWM and LTV Meanwhile for GWM in Indonesia, in the context of relationship with economic cycle, when the world s economy and domestic economy contracted in 2008 2009, primary GWM was set at 5%. Furthermore, with improving economic conditions, 9

followed by increasing inflation (also inflation expectations), Bank Indonesia decided to increase primary GWM from 5% to 8%. With tightening in both of these macroprudential instruments, especially on LTV ratio which was reduced to average 72.5% (KKB and KPR) in semester II of 2013, it can be seen that nominal credit growth in Indonesia started to experience a gradual decline from its historical pattern. Meanwhile for interbank market in Indonesia, with existing condition of excess banking liquidity, overnight PUAB interest rate mainly hovered in the lower corridor of Bank Indonesia bordered by deposit facility (DF). Figure 4. RED December 2014 Overnight Interbank Rates 10

III. DSGE MODEL WITH BANKING SECTOR 3.1 Households and Entrepreneurs Patient household maximizes its utility function based on consumption level choice c P t, resting time (out of working time n P P t ), and housing asset ownership χ t with discount factorβ p. max c t P,χ t P,n t P (i) (β P ) t [ (c t P (i) ξc P t 1 χ + ε P t (i) 1 σ χ n χ,t ε P t (i) 1+σ n t=0 n,t... (1) 1 σ c ) 1 σ c 1 σ χ 1+σ n ] Parameter ξ is the level of external habit formation and ε χ,t, ε n,t is shock intertemporal, housing preference, and labour preference which have dynamic AR(1) with i.i.d. error. Patient household has revenue coming from labor supply to business W t n t P, deposit revenue (1 + r D t 1 )d t 1, dividend from the company they own Π P t, and revenue from government as well as external bond interests. These revenues are used to pay taxes T t P, finance consumer spending, purchase housing asset, and save the remaining in form of deposits d t, government bonds B P g,t, and external bonds B t. Therefore, budget constraint faced by patient household is: P t c P t (i) + P χ,t (χ P t (i) (1 δ χ )χ P t 1 (i)) + d t (i) + B P g,t T P t (i) + Π P t (i) P + (1 + r t 1 )B g,t 1 + e t B t = W t n P t (i) + (1 + r D t 1 )d t 1 (i) + ρ t 1 (1 + r t 1 )e t B t 1...(2) In budget constraint, consumer spending and housing asset variables are each multiplied by the price value to get the nominal form. Parameter δ χ is the depreciation level of housing asset owned by households. From the objective function and budget constraint of patient household mentioned above, an equation solution is obtained which can explain the amount of patient household consumption, affected by the amount of deposit rate, taxes on deposit interest, as well as inflation which can be written: P ξc P t ) σ c (β P ( 1 + r t D (1 α TD ) )) = (c t P ξc P t 1 (c t+1 P t+1 P t ) σ c λ t = (c t P ξc P t 1 ) σ c βp ξ(c P t+1 P t ξc t P ) σ c...(3) while the housing ownership accumulation from patient household is also obtained by finding solution from the objective function and budget constraint which are 11

affected by taxes on deposit rate, inflation, housing price, as well as housing price expectation in the future which can be formulated as follows. P (i) ξc P t ) σ c ( β PP χ,t+1 (1 δ χ ) ) + ε χ,t (χ P t ) σ χ = (c P t (i) ξc P t 1 ) σ c ( P χ,t ) P t (c t+1 P t+1 ε χ,t (χ t P ) σ χ = λ t P χ,t β P λ t+1 (1 δ χ )P χ,t.. (4) The amount of deposits saved by patient household in banks is affected by the profit received, return of deposit in the previous period, wage resulted from labor supply, consumption, housing investment made, return of bonds in the previous period, as well as return of external bonds in the previous period which can be formulated as follows. d t = d t 1 (1 + r D t 1 α TD r D t 1 ) (α TW W t n P t + α TΠ Π P t ) + Π P t P χ,t (χ P t (1 δ χ )χ P P t 1 ) P t c t + W t n P t + (1 + r t 1 )B g,t 1 + ρ t 1 (1 + r t 1 )e t B t 1 B g,t e t B t λ t = β P λ t+1 (1 + (1 α TD )r t D ). (5) while impatient household also has utility function which has indifferent variables to patient household: max c t I (i),χ t I (i),n t I (i),b t I (i) (β I ) t [ (c t I (i) ξc I t 1 χ + ε I t (i) 1 σ χ n χ,t ε I t (i) 1+σ n t=0 n,t...(6) 1 σ c ) 1 σ c 1 σ χ 1+σ n ] In financing spending, other than the result of labor supply W t n t I, impatient household also borrows from banks worth b t I (i). Therefore, impatient household also has liabilities to pay for borrowings made in the previous period amounting at (1 + r BI I t 1 )b t 1 in the expenditure column. Budget constraint of impatient household is P t c I t (i) + P χ,t (χ I t (i) I (1 δ χ )χ t 1 (i)) + (1 + r BI I t 1 )b t 1 (i) = W t n I t (i) + b I t (i) T I t (i)...(7) In making borrowings to finance consumption, total borrowings which can be obtained by impatient household are limited by the price of building asset they own multiplied by the applicable loan-to-value ratio, m t I requirement. (1 + r t BI )b t I (i) m t I E t [P χ,t+1 (1 δ χ )χ t I (i)]......(8) From microeconomics side, value (1-m t I ) can be defined as proportional cost of collateral repossession for banks if there is default. From macroeconomics side, value m t I determined the amount of borrowings offered by banks to households for certain housing asset value they own. It is assumed that variations of this LTV ratio 12

do not depend on the choice of each bank but is an exogenous stochastic process which allows us to learn about credit-supply restriction on the real sector of the economy. From the aforementioned objective function and budget constraint of impatient household, an equation solution is obtained which can explain the amount of impatient household consumption that affected by the size of labor supply wage, borrowings from banks, inflation, consumer loan rate, housing price, as well as housing stock which can be formulated as follows. c t I = W tn t I P t λ t = (C t I I ξc t 1 + b I t α I TWW t n t P t P t ) σ c I βi ξ(c t+1 P t (1 + r t 1 ξc t I ) σ c BI ) ( b I t 1 P t I ) ) P χ,t(χ I t (1 δ χ )χ t 1 P t.(9) while housing ownership accumulation from impatient household is also obtained by finding solution from the objective function and budget constraint, affected by the size of LTV ratio, housing price, consumer loan rate, as well as inflation which can be formulated as follows. (c I t (i) I ξc t 1 ) σ c ( m t I E t [P χ,t+1 (1 δ χ )] P t (1 + r BI t ) P χ,t P t ) I + (c t+1 (i) ξc I t ) σ c ( β IP χ,t+1 (1 δ χ ) P t+1 β I ( m t I E t [P χ,t+1 (1 δ χ )] )) P t+1 + ε χ,t (χ t I (i)) σ χ = 0 ε χ,t (χ I t ) σ χ + β I λ t+1 [(1 δ χ ) m t I E t [P χ,t+1 (1 δ χ )] (1 + r BI t ) 2 ] = λ t [P χ,t m t I E t [P χ,t+1 (1 δ χ )] (1 + r BI ] t ) (10) The size of borrowings by impatient household from banks is affected by the size of LTV ratio, housing price expectation, housing stock, inflation expectation, as well as consumer loan rate which can be formulated as follows. b I t = m t I E t [P χ,t+1 (1 δ χ )χ I t ].(11) (1+r BI t ) Utility function from businesses is based on the return on capital which determines the size of income and loan repayment to banks or foreign creditors so the amount of realized entrepreneur profit can be formulated as follows. 13

K V t+1 = ωr t+1 P k,t K i t f(ω)dω (1 F(ω ti )) (1 + r be E t )b ω t (12) ti There is variable ω which is an idiosyncratic shock to entrepreneur and ω ti is a threshold which determines if the entrepreneur goes default (if ω < ω ti ) or makes payment (if ω > ω ti ) with probability default F(ω ti ) normal log. Financial contract between banks and entrepreneurs will happen when banks in minimum will get the same expected return as the opportunity cost. Because in this model what relates to loan on entrepreneur is loan unit, which already has minimal target of loan rate from saving unit, the size of bank s opportunity cost is the same as funding rate set by saving unit, which is the size of interbank borrowing rate R t. The determination of base lending rate by the wholesale unit already includes mark up which takes into account the stickiness as well as the probability of entrepreneur to default, F(ω ti ), based on bank expectation on return on capital entrepreneur. If entrepreneur cannot pay for its liabilities according to contract and suffers default, bank pays for monitoring cost and confiscates the entrepreneur s asset, can be written as (1 μ m )ωr K t+1 P k,t K i t, while the entrepreneur who defaults does not get anything. Financial contract between bank and entrepreneur must follow this relation. K max V t+1 = ωe t (1 + R t+1 )P k,t K i t f(ω)dω (1 F(ω ti )) (1 + r be E t )b ω t.. (13) ti Subject to: ω ti 0 k (1 F(ω ti )) (1 + r be t )b E t + (1 μ m ) ωe t R t+1 P k,t K i t f(ω)dω = (1 + R b E t )b t (14) The left side from the equation shows the expected gross of return from borrowings to entrepreneur and the right side is the bank s opportunity cost. Parameter μ m is the bank s monitoring cost if there is default, whose value will increase along with verification by bank to monitor remaining project if there is default. The default probability F(ω ti ) of entrepreneur is a cumulative distribution function, while f(ω) is a probability distribution function. ω ti is the expected threshold. Solution from the aforementioned issues is relational equation between corporate leverage k t = P k,tk t N t K ) and external finance premium s Ei t = E t(1+r t+1. (1+R t b ) s Ei t = E t(1+r K t+1 ) = f(k (1+R b t ) t ) = f ( P k,tk t ), f (. ) > 0 (15) N t 14

An increase in expected discounted return to capital will reduce expected probability of default so that entrepreneur can get more debt and expand the company. The mechanism is called financial accelerator because if given positive shock that will increase the company s net worth, with better balance sheet, the company will increase investment to expand the business with smaller external finance premium. 3.2 Producers Intermediate good producers work in a perfectly competitive market and have objective function to maximize profit which is the difference between sold products and cost of capital and labour, as follows. max E t s=0(β P θ F ) s {P w,t+s (j)y w,t+s (j) (w p,t+s (j)n p,t+s (j) + w I,t+s (j)n I,t+s (j) + p t (j) z t+s (j)k t+s (j))}.. (17) P w,t is the price of produced products and y w,t is the produced intermediate homogeneous product using production function as follows: y W,t (i) = A t [u t (i)k t (i)] α ((n P,t (i)) μ l (ni,t (i)) 1 μ l ) 1 α.... (18) A t is the total factor productivity, u t ε[0, ) is the level of capital utilization, k t is capital stock, n P,t is labour input from patient household, and n I,t is labour input from impatient household. There are three other types of producers in the model: capital good producers, housing producers, and final (consumption) goods producers. Capital good producers operate in a perfectly competitive market and use consumer goods to produce capital goods. Moreover, capital good producers also use old capital goods which do not depreciate, (1 δ k )k t 1, to be sold to entrepreneur, and can be written as follows. k t = (1 δ)k t 1 + ε i,t (1 1 κ 2 k ( i 2 k,t 1) ) i i k,t... (19) k,t 1 ε i,t is shock variable which has AR(1) dynamic with i.i.d. error. Old capital goods from entrepreneur are directly transformed to new capital goods, while transformation of consumer goods into capital goods will be subjected to adjustment cost function S k = ( i k,t i k,t 1 ) which has characteristics as follows. 15

S k (1) = S k (1) = 0; S k (1) = κ K > 0... (20) It means that in a steady state condition, there will be no adjustment cost and the further the utilization level of consumer goods from steady, adjustment cost will continue to increase. Objective function from capital good producers is to maximize profit as follows. max (β p ) s s=0 (P k,t+s k t+s (P k,t+s (1 δ)k t+s 1 + P t+s i k,t+s )). (21) k t Housing producers act in similar behavior as capital good producer, which is χ t = (1 δ χ )χ t 1 + ε iχ,t (1 1 κ 2 χ ( i 2 χ,t 1) ) i i χ,t..... (22) χ,t 1 The adjustment cost function also has the same characteristic as capital good producer, which is S χ (1) = S χ (1) = 0; S χ (1) = κ χ > 0......(23) The objective function is to maximize profit as follows. max (β p ) s s=0 (P χ,t χ t (P χ,t (1 δ χ )χ t 1 + P t i χ,t )) (24) χ t Final good producers are agents who combine goods from domestic retailer y H,t (j H ) and imported goods retailer y F,t (j F ) with CES model and make them into one final products which then are sold in a perfectly competitive market. Objective function from final good producers is as follows. y t = [η μ 1 1+μy H,t 1+μ + (1 η) μ 1 1+μ 1+μy 1+μ F,t ].... (25) η is home bias parameter and μ is parameter which determines the elasticity of substitution between domestic and foreign goods. Optimization of objective function from final good producers will result in equation of domestic goods demand (y H,t ), imported goods demand (y F,t ), and (final) price of consumer goods (P t ) as follows. y H,t = η ( P H,t ) 1+μ μ y P t.. (26)y F,t = t (1 η) ( P F,t ) 1+μ μ y P t......(27) t P t 1 μ = η(p H,t ) 1 μ + (1 η)(p F,t ) 1 μ.... (28) 16

Demand of imported goods (y F,t ) is affected by relative import price to final goods price, and the size of domestic goods demand (y F,t ) is affected by relative domestic price to final goods price, while the price of final goods itself (P t ) is formed by domestic price and import price. 3.3 Retailers Retailers in the model consist of domestic retailers, exporting retailers, and importing retailers who all of them are in a market condition with monopolistic competition, meaning retailers have market power in making price setting. Domestic retailers buy undifferentiated intermediate goods from businesses, change them into differentiated goods, then sell them to final good producers. Exporting retailers buy undifferentiated intermediate goods from businesses, change them into differentiated goods, then sell them in international market. Importing retailers buy undifferentiated goods from international market, change them into differentiated goods, then sell them to final goods producers. Price setting in the three retailers are based on sticky price model in accordance with Calvo, which in every period, only some of the retailers who can conduct price re-optimization, while the others adjust price based on inflation level happened in the previous period (backward looking). For domestic retailers who do not perform re-optimization, price will be set using function P H,t = P H,t 1 π t 1. Therefore, aggregate price during t is obtained with function: P H,t = (θ H (P H,t 1 π H,t 1 ) 1 εh + (1 θ H ) (P H,t (i)) 1 ε H ) 1 1 ε H. (29) Final log-linearization result from first order condition (FOC) objective function from domestic retailers shows inflation New Keynesian Phillips Curve (NKPC) equation, that domestic price is affected by self expectations, both backward or forward, other than affected by the price of intermediate goods, which is formulated as follows: π H,t = 1 (1+β P ) π H,t 1 + β P (1+β P ) (π H,t+1) + (1 β Pθ H )(1 θ H ) (1+β P )θ H (P W,t ). (30) For importing retailers who do not perform re-optimization, price will be set with function P F,t = P F,t 1 π t 1. Therefore, aggregate price during t is obtained with function as follows. 17

P F,t = (θ F (P F,t 1 π F,t 1 ) 1 εf + (1 θ F ) (P F,t (i)) 1 ε F ) 1 1 ε F (31) Final log-linearization result from FOC objective function of importing retailers is NKPC as follows. π F,t = 1 (1+β P ) π F,t 1 + β P (1+β P ) (π F,t+1) + (1 β Pθ F )(1 θ F ) (1+β P )θ F (ŝ t + P F,t ). (32) From the equation above it can be seen that import price inflation is also affected by self-expectation, both backward and forward expectation, as well as overseas price. Exporting retailers buy domestic undifferentiated goods, place brand and sell them overseas with price of P H,t, which is stated in foreign currency notes. It is assumed that price is sticky in foreign currencies. Demand equation for export goods is as follows: y H,t (1+μ H ) = ( P H,t μ P ) H y H,t... (33) H,t y H shows output from retailer which is defined as y H,t 1 0 = ( y H,t (j 1+μ H ) H dj H ) 1 1+μ H... (34) and P H,t as P H,t 1 0 = ( P H,t 1 μ H (j μ H ) H dj H ).... (35) Moreover, it is assumed that overseas demand is given by: y H,t (1+μ H ) = (1 η ) ( P H,t μ P ) H y t..... (36) t As with other retailers in the model, price setting from exporting retailers refers to Calvo standard scheme. The opportunity to change price is (1 θ) and the opportunity not to make price re-optimization is θ. For exporting retailers who do not perform re-optimization, price will be set with function P H,t aggregate price during t is obtained with function as follows. = P H,t 1 π t 1. Therefore, P H,t = (θ H (P H,t 1 π H,t 1 ) 1 ε H + (1 θ H ) (P H,t (i)) 1 ε H ) 1 1 ε H... (37) 18

Final log-linearization result from FOC objective function of exporting retailers shows that export price inflation is affected by self-expectation, both backward and forward, as well as by the price of intermediate goods and exchange rate, formulated as follows. π H,t = 1 (1+β P ) π H,t 1 + β P (1+β P ) (π H,t+1 ) + (1 β Pθ H )(1 θ H ) (1+β P )θ (P W,t ŝ t )... (38) H 3.4 Bank In this model banks are built upon two units: saving unit, which collects deposit from patient household as well as becoming supplier to interbank market; the other unit is lending unit which channels loans to entrepreneurs and impatient household as well as making purchase of government bond. In interbank market mechanism there is friction in the money market caused by probability of default from lending bank which cannot repay to saving bank. Figure 5. Bank's Financial Intermediation Process Patient household conducts deposit saving to saving unit which already conducts deposit rate markdown from interbank rate. Part of deposit fund is channeled to the purchase of risk free assets and the remaining is channeled in interbank market. Lending unit will guarantee available funds in interbank market to finance loan to impatient household and entrepreneurs with markup loan rate from its cost of fund (interbank rate). Lending unit can also make purchase of risk free assets. 19

Figure 6. Banking Block Scheme 3.4.1 Savings Unit Savings unit operates in a monopolistic competitive market condition and collects deposit, D t, from household workers. Deposit is assumed entirely to have no default or is entirely guaranteed. Bank sets deposit interest rate, R D j,t, which is paid to depositors and sets optimal portfolio allocation between placement in interbank market of D j,t = s j,t D j,t, or placement in risk free assets (government bond) of B sb j,t = (1 s j,t )D j,t. In every period there is a probability of default from interbank market placement with probability of default of δ t D which must be borne by savings banks. There are insurance premiums which must be paid by savings banks when conducting placement in risk free assets (cost of holding risk free assets), of χ s ((1 s 2 j,t)d j,t ) 2 Balance sheet savings unit: Assets Table 3. Balance Sheet's Savings Unit Liabilities Interbank lending: D j,t sb Government bonds: B j,t Deposits: D j,t 20

In a monopolistic competition condition and imperfect substitution among deposits, every savings bank faces supply deposit function as follows. D j,t = ( R D υ D j,t RD ) D t... (39) t From the function above it can be seen that supply deposit will increase along with the change in relative deposit interest rate throughout period. Variable D j,t is supply deposit in bank j, while D t is total deposit in the economy. There is quadratic adjustment cost on the change in deposit interest rate which creates the price rigidity and in the end generates varied interest rate spread in every period: R Ad D j,t = φ R D ( R D j,t 2 R j,t 1 D 1) 2 D t....(40) so that objective function from this saving unit is: max E 0 β t P λ b t {[(1 δ D t s j,t )R t R D j,t ]D j,t χ s ((1 s {s j,t, R D 2 j,t)d j,t ) 2 t=0 Ad j,t j,t } Subject to (39) and (40). R D }. (41) In symmetry equilibrium assumption, first order condition from the optimization is as follows. s j,t = 1 δ t D R t χ s D t..... (42) ( 1+υ D ) (R D υ t 1) = (1 s t δ D t )(R t 1) χ s (1 s t ) 2 D t φ R D D υ D ( R D t RD t 1 1) R t D RD t 1 + β pφ R D υ D 1) ( R D t+1 RD )... (43) t ( Rt+1 D RD t Equation on (42) explains the placement allocation made by savings banks to interbank market. Allocation placed in interbank market, s j,t,will decline along with the increase in probability of default, δ t D and increase in total deposit will increase the amount of allocation in interbank market. Meanwhile, equation on (43) above explains deposit rate, R t D, which is markdown from interbank rate, R t. The increase of risk in interbank market, δ t D, will make saving banks reducing placement allocation in interbank market and adding placement allocation in risk free assets. The increase in interbank rate or the return rate on risk free assets will make savings banks also reducing funding supply in 21

interbank market. Likewise if there is an increase in total deposit, interbank lending will increase so there will be expansion of credit supply. The framework developed from two equations above show two channels of credit supply behavior transmission from savings banks affect the real economy. First, by determining deposit return rate, which is in nominal rigidity condition, savings banks affect inter-temporal substitution of consumption throughout periods and cause a smooth consumer behavior. Second, by dividing portfolio optimally, savings banks affect credit supply condition by developing and tightening credit market condition. 3.4.2 Lending Unit Lending unit also operates in a monopolistic competitive condition to provide loans to entrepreneurs. To provide loans to entrepreneurs, lending unit j uses interbank borrowing D j,t added by liquidity injection from the central bank (quantitative monetary easing) M j,t, and total market value from bank capital itself Q Z t Z j,t added by liquidity from the central bank x j,t. Here it is assumed that banks use Leontief technology to produce loans as follows. b j,t = min{d j,t + M j,t ; K j,t (Q Z t Z j,t + x j,t )} τ t.. (44) The use of Leontief technology to produce loans implies that there is perfect complementary effect between interbank borrowing and bank capital. Furthermore, marginal cost to produce loans is the amount of marginal cost from interbank borrowing and cost to produce capital. Lending bank s balance sheet in period t: Table 4. Balance Sheet's Lending Unit Assets Liabilities Loans:b j,t - x j,t Government bonds: B lb j,t = Q Z t Z j,t + x j,t Interbank borrowing: D j,t Bank capital: Q t Z Z j,t Central bank s money injection: m j,t Other terms: (τ t 1)(D j,t + M j,t ) 22

As in Gerali et al. (2009), adjustment cost is related to the change in prime lending rates, R L j,t R Ad L j,t = φ R L ( R L j,t 2 R j,t 1 L 1) modelled in accordance with Rotemberg (1982), as follows. 2 L t.. (45) Problem of optimization from lending bank is choosing R L j,t, K j,t, δ D j,t, δ Z j,t so that maximization problem of lending banks can be elaborated as follows. max {R L j,t,k j,t,δ D j,t,δ j,t D 2 χ δ D 2 (δ j,t 1D j,t ) π t t=0 Z } E 0 β b t λ t b { R j,t χ δ Z L D 2 (δ j,t 1Q Z 2 t Z j,t ) π t (1, δ D j,t )R t D j,t R t m j,t [(1 δ Z L j,t )R t+1 R t ]Q Z t Z j,t + χ k (K K j,t 2 K Q Z t Z j,t ) 2 (R L j,t R R t )x j,t Ad L j,t }... (46) with b j,t = min{ D j,t + m j,t ; κ j,t (Q K t K b j,t + x j,t )} Γ t... (47) R Ad be j,t = φ R be 2 R Ad bi j,t = φ R bi ( R be 2 j,t be 1) b E R t.. (48) j,t 1 ( R bi j,t 2 R j,t 1 bi 1) 2 b t I... (49) b E j,t = ( R be υ LE j,t R t be) b E t......(50) H = ( R bi j,t b j,t R t bi) υ LH b t I... (51) b t = b t E + b t I. (52) K b t = (1 δ b b )K t 1 + w b j b t 1.... (53) In symmetric equilibrium assumption, all banks take the same decision, first order condition from this optimization, among others, result in stickiness equation in the lending rate of entrepreneur,r t LE, and the lending rate of impatient household, R t LI, as follows. R t LE = 1 + R t LI = 1 + υ LE (υ LE 1) (ζ t 1) υ LI (υ LI 1) (ζ t 1) κ LE ( R LE t (υ LE 1) RLE t 1 κ LI ( R LI t (υ LI 1) RLI t 1 LE (υ LE 1) (R t+1 1) R t LE RLE + β pκ LE t 1 LI (υ LI 1) (R t+1 RLI t 1) R t LI RLI + β pκ LI t 1 LE RLE 1) R t+1 t RLE. (54) t 1) R LI t+1 RLI... (55) t Equation on (54) explains the relation between entrepreneur prime lending rate and marginal cost from loan and future profit from the adjustment of entrepreneur lending rate. Similar thing for equation (55) explains the relation 23

between impatient household prime lending rate and marginal cost from loans and future profit from the adjustment of impatient household lending rate. 3.5 Government and Central Bank The government and central bank in this model can be illustrated as follows. Banks Impatient Households Domestic Loan Tax Government Consumption Final Goods Producers Patient Households Foreign Loan ROW Figure 7. Government and Central Bank Model Scheme The government collects taxes and borrows from domestic market (via banking) and overseas market to finance expenditure. Budget constraint of the government in the economy is: P t g t + (1 + r B,t 1 )e t b G,t 1 + (1 + r t 1 )b G,t 1.. (60) = (T P t + T I t ) + e t b G,t + b G,t With explanation that g t is government spending which is modelled by dynamic AR(1),b G,t is government foreign debt which is also modelled as AR(1), and T P as well as T t I are taxes collected from patient and impatient households. The determination of policy rate (r t ) by central bank is modelled in the Taylor rule equation as follows. (1 + r t ) = ( 1 + r φ R t 1 ) (( π φ φ 1 φ t π y R y t ) ( 1 + r π t y ) ) ε r,t. (61) 24

With explanation that φ π and φ y are weights imposed to inflation and output, r is the steady state nominal interest rate, and ε t r is shock i.i.d. on monetary policy with normal distribution and standard deviation σ r. 3.6. Market Clearing Condition To close the model, needs a market clearing condition equation for goods produced by final goods producers, goods produced by intermediate good producers (intermediate homogeneous goods), housing market, balance of payment, and GDP definition in the model. Moreover, since the modelled economy is open economy, there needs an equation specification from risk premium which is the function of total foreign debt to GDP ratio (in line with Schmitt-Grohe and Uribe, 2003). Final Goods Producers Output π t = η(p H ) 1μ (π H,t + p H,t 1) + (1 η)(p F ) 1μ (π F,t + p F,t 1).....(62) c y c t = γi c I y c t I + γp c P c t P + RnY N y......(63) Intermediate Homogenous Goods Market y H,t (j)dj 0 1 + y H,t (j)dj 0 1 = y W,t. (64 ) Housing Market γ P χ t P + γ I χ t I = χ t..... (65) Balance of Payment P F,t y F,t + e t (1 + r t 1 )ρ t 1 b tot,t 1 = e t P H,t y H,t + e t b tot,t. (66) With explanation b tot,t = b G,t.. (67) GDP 25

P t y t = P t y t + e t P H,t y H,t P F,t y F,t...... (68 Risk Premium ) (1 + ρ t ) = exp ( ρ e tb tot,t ) ε P ρ,t t y t.. (69) 26

IV. ESTIMATION AND SIMULATION For estimation needs, quarterly data is used from quarter I of 2001 to quarter IV of 2012. Real sector data used for estimation are private consumption, private investment, government expenditure, exports, import, CPI inflation, import deflator, export deflator, and exchange rate. For disaggregated GDP, export deflator, and import deflator, the data used comes from GDP publications based on output from BPS. Exchange rate and CPI inflation data are obtained from ARIMBI/SOFIE model database. For external sector variable, the data used is also used by ARIMBI and SOFIE model, which are world GDP, US inflation, and LIBOR, while transaction volume of each bank is used for interbank market transaction data. For banking sector, the data used is policy rate (BI rate), interest rate and the amount of third party funds (TPF) collection, bank capital, interest rate, as well as household loan (consumer loan) disbursement, interest rate, and the amount of loan disbursement to corporates (investment and working capital loans), the amount of SBI (and other monetary operations) owned by banks, the amount of bank claims to the central government (SBN), the amount of bank reserve (including cash in vault), and non-performing loan (NPL). For bank balance sheet composition, the data used comes from analytical balance sheet of commercial banks. NPL data is obtained from SOFIE model database. 4.1 Estimation In determining steady state value of real sector variable, realized data during estimated period (quarter I 2001 to quarter IV 2012) is used as main benchmark. However, we also consider steady state value used in DSGE model of developed countries or developing countries as comparison. For disaggregated GDP variable, based on processed data during estimated period, HP filter is used and gets a result as seen in Figure 7. 27

.7.6.5.4 Consumption Gov Investment Export Import Mean 0.589 0.077309 0.220452 0.43901 0.340948 Median 0.590 0.078438 0.220191 0.448076 0.356237 Maximum 0.620 0.083718 0.247197 0.48653 0.374467 Minimum 0.553 0.06537 0.196657 0.391218 0.286572 Std. Dev. 0.021 0.005743 0.01582 0.034335 0.032558.3.2.1.0 00 01 02 03 04 05 06 07 08 09 10 11 12 CRL_SS GRL_SS PMTBRL_SS XRL_SS MRL_SS Figure 7. Steady State of Disaggregated GDP Variable Based on Data Different from disaggregation conducted by BPS for investment variables (business investment and building investment), in the model investment is divided into two: housing investment and investment for capital goods. To get steady state value from housing investment to total GDP ratio, we multiply the ratio of building completion value for building category (0,4) with building investment from total investment average ratio (0,83), then we multiply it again with investment to GDP ratio (0,22). By using that approach (and rounding off), we set steady state value for housing investment from total GDP ratio of 0,08. Figure 8. Building Completion Value Ratio per Category and Building Investment Ratio Using the same approach, we can also get steady state value for component variable of bank balance sheet. However, as seen on Figure 9, HP filter result for bank balance sheet component to total assets ratio variable does not show stability 28

in certain value. Besides using HP filter result which is presented in the figure, research result of Gunadi and Budiman (2011) on the optimization of bank portfolio composition in Indonesia is used to determine steady state value of bank balance sheet variables which are completely presented on Table 5. Figure 9. HP Filter Result from Bank Balance Sheet Component Variable to Total Asset Ratio Table 5. Steady State Value of Bank Balance Sheet Variables Assets Liabilities Total Loan 0.7 Deposit 0.9 SBI 0.12 Capital 0.1 Loan to Government (SBN) 0.08 Reserve 0.1 Steady state value of policy rate variable (BI rate) variable is using the same value as the one used by ARIMBI model. If we see Figure 10 which shows HP filter result from various interest rate variables in the model, it can be seen that spread between BI rate and TPF rate is not stable. When BI rate is high, spread on TPF rate is also high, while when BI rate is low, spread on TPF rate is also low. Because we are using a considerably low steady state value of BI rate, for data consistency, we use low spread in calculating the steady rate of TPF rate. Using this method, we set the steady state value of TPF rate at 4.5%. To determine the steady state value of interest rate in consumer and investment loans, we add average difference between both interest rates and BI rate in the estimation periods so that we obtain steady 29

state value of consumer loan rate at 13.65% and steady state value of corporate credit rate (working capital and investment) at 11.4%. For LIBOR rate which becomes proxy from overseas interest rate, we use the same number used in ARIMBI, of 3%. 20 16 12 8 4 0 2004 2005 2006 2007 2008 2009 2010 2011 BI_RATE_TREND R_DEP_TREND R_KE_TREND LIBOR_TREND R_KK_TREND Figure 10. HP Filter Result from Various Interest Rate Variables in Model In complete, steady state value for all variables used in the model is on Table 6. Table 6. Steady State Value of All Variables Variables Values Consumption to GDP ratio 0.59 Capital investment to GDP ratio 0.19 Housing investment to GDP ratio 0.08 Government expenditure to GDP ratio 0.09 Import to absorption ratio 0.38 Export to output ratio 0.44 Loan to HH to GDP ratio 0.31 Loan to entrepreneur to GDP ratio 0.71 Deposit to GDP ratio 1.28 Importer s profit margin 0.03 Exporter s profit margin 0.026 Domestic retailer s profit margin 0.18 30

Table 6. (continued) Variables Values Rate on loan to HH* 14.98% Rate on loan to entrepreneur* 12.9% Rate on deposit* 4.5% Foreign interest rate* 3% CAR 0.14 Bank s profit to total asset ratio 0.025 Deposit to bank s total asset ratio 0.9 Bank s capital to total asset ratio 0.1 Loan to bank s total asset ratio 0.7 Risk free asset to bank s total asset ratio** 0.2 Reserve to total asset ratio 0.1 Interbank volume to total asset 0.5267 Part of the parameters used in the model are calibrated using the value used by the model which has been developed by Bank Indonesia and related empirical research results. Capital share in the production function is set at 0.54 according to estimation result from the 2012 MODBI model. Value of home bias parameter is determined according to HP filter value from Indonesia s import to absorption ratio during estimation period. Parameters which determine elasticity of substitution between domestic and foreign goods and elasticity of substitution for export goods use value derived from the research of Zhang and Verikios (2006) 2. Parameter value for risk premium and which sets cost to manage bank capital is obtained from steady state relations among various variables in the model. Calvo parameter for labor follows estimation result from BISMA model (2009). Parameter of ad hoc equation which determines the dynamic of risky assets weight (equation 36) and reserve owned by bank (equation 37 39) uses estimation result of partial equation based on data during estimation period. 2 Parameter calculation is used based on CES based estimation in accordance with the assumptions used in the model developed in this research. 31

Table 7. Parameter Value of Calibration Result Parameters Values Mark-up parameter in labor market ε w 11 Depreciation rate of capital δ k 0.025 Depreciation rate of housing asset δ χ 0.0125 Cost to managing bank s capital δ b 0.1 Risk premium parameter ρ b 0.11 Capital share in production function α 0.54 Home bias parameter η 0.62 Elasticity of subtitution between domestic and foreign goods μ 0.63 Elasticity of subtitution for export goods μ H 0.45 Labour income share of unconstrained household μ L 0.67 The probability of given labor (from patient and impatient HH) is selected not to reoptimize its wage θ wp dan θ wi 0.65 Reserve equation s parameter ρ Γ 0.197 Excess reserve equation s parameter ρ ε 0.632 Determination of prior for estimated parameter uses the same approach as determination of calibrated parameter: using value from model which has been developed before or from related empirical research. For parameter κ d,κ be, and κ bi, prior is determined by setting the response of bank retail interest rate to shock of policy rate according to estimation result from immediate pass-through which was conducted by Harmanta and Purwanto (2012). For Taylor rule parameter (φ r,φ π, and φ y ), value of prior is set according to the value used in ARIMBI model. Prior for parameter which regulates habit persistence in household consumption activities using estimation result of BISMA model (2009). In complete, prior distribution, type of distribution and posterior distribution from parameter of estimation result are in Table 8. 32

Table 8. Parameter Value of Estimation Result Parameters Distributions Prior Distribution Posterior Distribution Mean Std. Dev. Mean Inverse of intertemporal elasticity of substitution for housing Inverse of intertemporal elasticity of substitution for consumption Inverse of Frisch elasticity of labour supply Adjustment cost paremeter for deposit rate Adjustment cost paremeter for entrepreneur loan rate Adjustment cost paremeter for household loan rate Adjustment cost paremeter for capital investment Adjustment cost paremeter for housing investment Adjustment cost paremeter for bank s CAR Calvo paremeter for import goods Calvo paremeter for domestic goods Calvo parameter for export goods σ χ normal 4 0.2 4.1670 σ c normal 2 0.2 2.1274 σ n normal 2 0.2 4.1417 κ d gamma 3.25 0.2 3.2675 κ be normal 3.5 0.2 3.7420 κ bi normal 8 0.2 8.1676 κ k gamma 5 0.5 5.1631 κ χ normal 50 0.5 49.3372 κ kb beta 1 0.05 0.9684 θ f beta 0.7 0.05 0.6254 θ h beta 0.4 0.05 0.3948 θ h beta 0.6 0.05 0.7898 4.2 Simulation In this part dynamic from impulse response produced by model will be studied. Discussion will be focused on simulation from monetary policy in form of 33

shock in BI rate and simulation from macroprudential policy. Since the developed model assumes a small open economy, there will also be discussion on transmission from exchange rate shock. Furthermore, according to model development design, this part will also focus on discussion of financial accelerator mechanism simulation and shock coming from interbank market. 4.2.1 BI Rate s Shock Figure 11. Impulse Response Shock of BI Rate In literature transmission of monetary policy rate starts from policy rate (BI rate) which affects deposit rate and lending rate. The effect starts from short-term interest rate and continues to long-term interest rate. With the price stickiness, the change in policy rate will affect real interest rate of working capital, investment, and consumer loans which eventually will affect real variables (income statement and balance sheet of banks, corporates, and households accounts). By watching impulse response function model, as seen in Figure 11, an increase of BI rate of 1% will be transmitted to various interest rates in the banking sector, either deposit rate or lending rate. The size of interest rate increase is adjusted with the size of mark-up and the level of stickiness from each interest rate. Response of BI rate increase is the soonest transmitted to deposit rate which directly increases in the same period as the BI rate increase and has the same pattern as BI rate if compared to the increase in lending rate. This is caused by deposit rate which has a smaller level of stickiness compared to lending rate. The increase in lending rate will 34

decrease total loans of household which will impact to the decrease of total consumption in the economy. The decrease of private demand will cause producers to reduce goods production, which is seen in the decrease of final good output and eventually will reduce GDP. The decrease of output production by producer also causes the decreasing needs for labor so there is decreasing labor supply, both from patient household or impatient household. The decrease in employment opportunity will cause the decline in household income so that household consumption will continue to reduce. The decline in private demand will pressure inflation downward. The increase in BI rate will also cause exchange rate to appreciate which will cause export decline due to reduced competitiveness which eventually will cut GDP. From simulation result above, it can be seen that shock propagation of policy rate affects intermediate variables and real variables with behavior in line with economic theory. Therefore, the developed DSGE model can capture dynamic of policy rate transmission properly. Then, in accordance with one of the aims of DSGE model development, there will be simulation on the impact of financial accelerator existence by comparing it if the model is not completed with financial accelerator. 2.00E-03 3.00E-04 1.00E-03 1.50E-03 1.00E-03 5.00E-04 0.00E+00-5.00E-04 BI Rate 1 6 11 16 21 26 31 36 2.50E-04 2.00E-04 1.50E-04 1.00E-04 5.00E-05 0.00E+00-5.00E-05-1.00E-04 Loan Rate to Household 1 6 11 16 21 26 31 36 8.00E-04 6.00E-04 4.00E-04 2.00E-04 0.00E+00-2.00E-04 Loan Rate to Firm 1 6 11 16 21 26 31 36 9.00E-04 8.00E-04 7.00E-04 6.00E-04 5.00E-04 4.00E-04 3.00E-04 2.00E-04 1.00E-04 0.00E+00-1.00E-04-2.00E-04 Deposit Rate 1 6 11 16 21 26 31 36 4.00E-03 2.00E-03 0.00E+00-2.00E-03-4.00E-03-6.00E-03-8.00E-03 pi_4 1 6 11 16 21 26 31 36 1.00E-03 5.00E-04 0.00E+00-5.00E-04-1.00E-03-1.50E-03-2.00E-03-2.50E-03-3.00E-03-3.50E-03-4.00E-03 1 6 11 16 21 26 31 36 GDP Black line without financial accelerator Red Line with financial accelerator Figure 12. Impulse Response of Shock BI Rate with Financial Accelerator and Without Financial Accelerator Effect of financial accelerator mechanism will cause GDP growth to become lower during economic contraction, also when the economy is in expansion phase, 35

financial accelerator mechanism will cause GDP to grow bigger, as seen in the figure above. GDP which accelerates due to financial accelerator mechanism also gives impact to a more volatile inflation if compared to the condition without financial accelerator. High volatility in GDP and inflation variables arise in the economy will cause policy rate (BI rate) to become higher during contraction and lower during expansion, followed by movement in other bank interest rates. The policy implemented is assumed not only use BI rate, but combined with macroprudential countercyclical policy to hold lending rate by reducing LTV ratio (red line). Simulation result shows that shock in form of policy mix will pressure lending growth downward if compared to the condition without LTV shock. GDP and inflation decline, but do not change much if compared the condition of using BI rate only. In policy mix use, the decline in consumption is covered by the decline in import so that GDP tends to be stable. Simulation result also shows that policy mix, besides generating stable GDP growth and inflation, can also control consumption so that import demand drops. With a stable export, the decline in import will provide positive impact to current account. 4.2.2 Households LTV Ratio Requirement s Shock Figure 13. Impulse Response Shock of Household's LTV The increase in expansionary loan to value ratio according to economic theory will increase total loan disbursed by banks and increase leverage of borrowers 36

(companies and households). It will increase consumption which eventually will increase GDP. However, GDP increase due to rising consumption will trigger import and worsen current account. Simulation above (Figure 13) shows that the increase in LTV ratio requirement for household borrowing (consumer loans) causes increasing volume of household loans due to incentive in high amount of loans which can be given by banks on collateral owned by households. With the increase in LTV, with the same asset value, households get more loans from banks. The increase of household loans volume pushes banks to manage their asset portfolio by reducing entrepreneurs loan volume and shift them to household loans so in the figure there is a decrease in entrepreneur loans which is followed by an increase in household loans. Increasing loans to households will increase household consumption which boost producers to increase their final good output. A high increase of final good output needs an increase in production factor, which is the increase in labor number, both from patient household or impatient household which in the end will increase household income. Rising consumption will eventually increase GDP. However, rising GDP causes import to increase and export to decrease which lead to a worsening current account (CA). This shows that result of model simulation is in line with economic theory and can capture LTV shock propagation through main real and financial variables which become focus of attention from policy makers. 37

4.2.3 Interbank Market s Shock Figure 14. Impulse Response Shock of Interbank Market Interbank market has a significant role in propagating the recent financial crisis, which is well documented in literature, increasing risk in interbank market can cause resource reallocation from interbank lending to risk-free government bond. As the main source of liquidity provider for banking in creating new loans, interbank market shock causes the downfall of credit supply for firm and household which can create recession. Several empirical study findings such as in Socio et al. (2011) confirm that shock in interbank market is a significant factor in financial crisis. Simulation on model during shock of interbank market volume decrease (Figure 14) shows that the number of loans, either to household or entrepreneur experiences a decrease so that in total banks will have a reduction in loan to deposit ratio (LDR). The decrease in total loan will cause banks to suffer a profit decline in which bank capital will also decrease because bank capital is an accumulation of capital from the previous period and retained earnings. Bank CAR will also decline 38

along with the decrease in bank capital. The reduction of loan for household and firm will lead to the decline in total consumption. The impact of weakening consumption then causes a drop in GDP. Then, with the GDP drop monetary authority will respond by cutting policy rate (BI rate) which eventually will impact to the depreciation of exchange rate. By comparing propagation behavior of interbank market shock which is illustrated by model simulation and literature, it can be concluded that shock transmission in interbank market can also be captured by the model comprehensively. Main phenomenon such as resources reallocation between interbank lending and risk-free government bond in interbank market which has important role in propagating crisis can also be well simulated. 4.2.4 Exchange Rate s Shock Figure 15. Impulse Response Shock Exchange Rate Simulation result (Figure 15) above shows depreciation in rupiah exchange rate will increase competitiveness of export products which increase export volume and increase the production of intermediate goods. The increase in intermediate goods production will cause an increase in production factor needs such as labor, both from patient household or impatient household. The increase in intermediate goods production will push the increase in final goods so that GDP will also rise. 39