404Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

I. INTRODUCTION

The period after the financial crisis is always followed by the introduction of new set of regulations that tighten the activity in the financial sector, particularly banks. In the international sphere this can be seen from the introduction and the implementation period of the Basel Rules by the Basel Committee on Banking Supervision (BCBS), which is a committee of central banks from various countries around the world. Basel I was officially introduced in 1988, triggered by Latin America debt crises in early 1980s (BIS Website, accessed 2017) and US Saving and Loan (S&L) crisis in the late 1980s and early 1990s (FDIC Website, accessed 2017). In the 1997-98 many Asian countries were hit by financial crises. This was followed by the proposal for Basel

II.After a long process, it was launched in 2004 and was called the Revised Capital Framework. Likewise, the Basel III, which was introduced in 2010, was a reaction to 2008 financial crisis.

	- Early 1980s -	- Late 1980s -	- 1988 -
	Latin America	US Saving &
			Basel I
	Debt Crisis	Loans Crisis
	- 2004 -	- Early 2000s -	- 1997-1998 -
			Asian Financial
	Basel II	DotCom Crisis
			Crisis
	- 2008 -	- 2010 -
	US Subprime
		Basel III
	Mortgage Crisis

Source: Collaborated from many sources

Figure 1.

Sequence of Financial Crisis and Basel Accord

Each Basel regulation is not introduced to replace the previous one, but rather to revise or to complement with more detailed and tighter regulations on the banking system. Basel I, which is the first attempt to regulate bank’s capital ratio took focus only on the application of minimum ratio of capital to risk weighted assets. This rule is then revised to be more detailed and stringent by Basel II which governs: (i) the application of more extensive minimum capital requirements, (ii) the strict process of monitoring and assessment of capital adequacy, and (iii)

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 407

Figure 4 describes the main focus of this study in stylized way. Each arrow represent hypothesis to be tested in this study. This study examines the cyclicality relationship between the business cycle toward the credit cycle and the risk cycle. By comparing the relationship of those Tri-Cycles, it can be determined which cycle should receive more attention from the regulators. So that the implementation of the countercyclical regulation could be better targeted.

Furthermore, this study examines the relationship of the credit cycle and the risk cycle toward the performance of individual banks. This phase of analysis focus on the impact of the Tri-Cycles to the bank’s performance. Especially, to measure the cost borne by the banks resulted from controlled of credit cycle and risk cycle. The results of this study might provide insight for the regulator to estimate the impact of regulating lending and risk-taking to the performance of individual banks.

II. THEORY

Since 1980s, the business cycle literatures has been largely driven by Standard Real Business Cycle (RBC) developed by Kydland and Prescott (1982) which assume no financial frictions in the economy. The theory was revised by Bernanke, Gertler and Gilchrist / BGG (1999) which revealed important role of Financial Accelerator – which is refer to credit market friction and financial cycle – in determining the business cycle dynamics. Then, 2007/8 financial crisis stimulated a new strand of literatures of business cycle which not only accommodate financial cycle, but also risk dynamics or risk cycle.

Burns and Mitchell (1946); in Jacobs (1998), one of the first among others, defined business cycle as:

“. . . a type of fluctuations found in the aggregate economic activity of nations that organize their work mainly in business enterprises: a cycle consists of expansions occurring at about the same time in many economic activities, followed by similarly general recessions, contractions, and revivals which merge into the expansion phase of the next cycle . . .”

The definition above embodies some notable features of business cycle, which becomes focus of many literature on this field. The first one is expansion, the period of surging business activity and gross domestic product expands until the period reach its peak. This period is also known as an economic recovery. The highest point of a cycle, the peak, is a key period in which the economic bubble get burst and economy the economy turns into contraction period. This period is a phase in which ecoomy as a whole is in decline. The lowest point of this phase, which signals the reversal or revival, is called recession.

408Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

2.1.Development of Business Cycle Theory

2.1.1. Econometric Business Cycle Research (EBCR)

Econometric Business Cycle Research (EBCR) is a term referring to strand of literatures combining theoretical and statistical approach in studying business cycle. Term EBCR has been popularized for the first time by Tinbergen (1940). The EBCR approach was developed in some way as a critic toward the then-previous approach that did not combine theoretical framework with data specification (Jacobs, 1998).

Theory		Facts
Model		Data
Economic		Refined
Model		Data

Combination of

Theory and Data

Policy Evaluation

Description

Forecasting

Figure 5. The EBCR Framework

2.1.2. Real Business Cycle (RBC) Theory

Standard Real Business Cycle (RBC) model is based on seminal work of Kydland and Prescott (1982). In the model, a competitive market creates resource allocation that maximizes the household utility with limited budget constraint of each resource (Kiyotaki, 2011). On the standard model, the most prominent determinant of the business cycle are shock by the government budget and technological development (Romer, 2012).

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 409

An ongoing debate is still going on about this theory ability to explain the heterogeneity of households and companies in the real world. This assumption is considered too strict. Even so, it encourages the further development/refinement of the RBC theory and not rather neglected its reliability.

In general, RBC theory explains how factors of production, namely labor, capital and other factors (such as technology and government budget) affect productivity (output) in the aggregate sense (Kiyotaki (2011) and Romer (2012)). In conditions where there is positive shock to productivity, the marginal product of labor will increase which will lead to a rise in real wage rates and labor supply quantity. The combined effect of rising wages and employment will drive the output to rise. However, because of the increase in productivity is only temporary, then the growth of output in the future will increase by lower pace than the present growth of output, and the growth of income and consumption will not rise as much as the output growth in the present (Kiyotaki, 2011). This condition will then encourage increase in investment and capital stock in the foreseeable future. This process then creates a new expansion phase. It applies in the opposite direction for the contraction phase.

One of the most criticized aspect of RBC model is its ignorance of the frictions in financial market. This stylized feature of the model departed from the strong assumption of efficient financial market. The hypothesis, which is very strict, assumes that in time of business fluctuation, every agent in the economy will instantly do recalculation of its economic behavior and decision to adapt with the change. Consequently, there is no such frictions in the financial market. Latter, in the late 1990s, BGG (1999) promoted a model to revised this view in examining business cycle.

2.1.3. Financial Accelerator

BGG (1999) are the first to develop a framework which they called as “Financial Accelarator”. In their seminal work, they developed a dynamic general equilibrium model to reveal the frictions in the credit market which play prominent role in determining business fluctuations. The term Financial Accelerator refers to endogenous developments in the credit markets which amplify and propagate shocks to the macroeconomy.

They materialized financial frictions in their model in three aspects. First, they internalized money and price stickiness to examine role of the friction in the transmission of monetary policy. Second, they relax the efficient financial market assumption by introducing lags in investment. Third, they relax the assumption of firms homogeneity in order to describe the condition in which every borrower has different access to capital markets. The main contribution of the model is how the financial accelerator give significant influence on business cycle dynamics.

410Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

Departing from the standard RBC model, and influenced by Financial Accelerator model, Kiyotaki (2011) describes the effects of the business cycle toward the credit cycle. In general, relation between the two cycles are influenced by asset quality at every phase of the business cycle. In the period of expansion, there is substantial increase in the value of assets. The increase, including the value of the assets of the firm, make the firm has higher collateral to be utilized to get credit. Moreover, the business boom condition make firm’s balance sheet substantially sound, which is a sign of growth. These conditions make the firm could obtain more credit to finance new investments and expand further. In line with this, the good economic conditions make the firm able to repay their credit and then it has good credit rating. The level of non- performing loan (NPL) in the banking system is then generally low.

Capital stock, Output

No Friction

Financial Friction

Time

Source: Kiyotaki (2011)

Figure 6.

Response toward Shock in Asset Quality

The reverse condition happens when the economy is in contraction phase. Impairment of assets (the burst of the boom) and deterioration of economic conditions will generally make the firm experience a decrease in performance and asset values. As a result, loan repayments begin to deteriorate (NPL increase). On the other hand, the banks tend to have lower credit growth. This happens due to (1) deterioration of the financial condition of the firm/debtor; (2) the impairment of the value of assets/collateral; and (3) the bank’s internal condition is getting worse by the rising of NPL rate.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 411

Expansion Phase

Increase in		Higher		Sound debtor		Credit
		collateral		balance
asset value						expansion
		value		sheet

Contraction Phase

Shrinks in		Plummeting		Deterioration		Credit crash &
		collateral		of debtor
asset value		value		balance sheet		crunch

Figure 7.

Expansion and Contraction Sequence in Business Cycle

2.2. Empirical Studies

2.2.1. Relationship between Cycles

Bertay, et.al (2015) in their study analyzed the cyclicality of individual bank lending toward business cycle. They segregated the sample based on the ownership, state-owned banks and private banks. The result shows the state-owned bank proved to be less cyclical than the private banks. The finding applies particularly in countries with higher governance index. In case for developed country, the result even shows counter-cyclical lending behavior by state-owned banks. The study came to the conclusion that state-owned banks can effectively play counter- cyclical role toward the country business cycle. State-owned banks could play a stabilizing role of the business cycle and financial cycle. However, Bertay, et.al. (2015) also found that loans allocation made by state-owned banks tend to be bad so that from business point of view, the behavior of the state-owned bank is not economically optimal because of its role to support government policy. The study therefore concluded that implementation of micro- and macro- prudential banking regulations such as monetary policy and statutory reserves are better tools than altering behavior of the state-owned banks. Empirical model applied in the study are the first-difference GMM of Arellano and Bond (1991) and the system GMM of Blundell and Bond (1998) enhanced by Windmeijer (2005). The study was conducted with a sample of 1,633 banks from 111 countries in the 1999-2010 time period.

Ferri, et.al. (2014) conducted an analysis of influence between bank ownership and lending behavior of individual banks and its cyclicality over the business cycle. The sample of the study is banks in Europe in the period 1999-2011. Segregation of the sample conducted between

412Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

profit-oriented bank (conventional bank) and not-for-profit bank (cooperative banks and saving banks). In his study Ferri, et.al. (2014) used first difference GMM by Arellano and Bond (1991). The result showed that there was no significant difference between the two groups of banks based on profit orientation. The main factors that can explain the behavior of bank lending is the monetary policy of the European Central Bank (ECB).

Ibrahim (2016) conducted a study to compare the lending cyclicality between conventional banks and Islamic banks in Malaysia. The sample used in the study includes 21 conventional banks and 16 Islamic banks in Malaysia during the period 2001-2013. The data used is unbalanced panel data. The results showed that generally the behavior of bank lending is pro-cyclical to the business cycle. However, by segregating the samples it is observed different cyclicality behavior between conventional banks and Islamic banks. Pro-cyclical behavior only observed at samples of conventional banks. While for Islamic banks, business cycle appears to have no effect on its lending behavior. In fact, the estimation results obtained show a negative value indicating a counter-cyclical lending behavior of Islamic banks. Similar with Bertay, et.al. (2016), the model estimation used in the study is the first-difference GMM and GMM system.

As for the case of Indonesia, Pramono, et.al. (2015) in their study examined the influence of Countercyclical Capital Buffer (CCB) policy on the growth of bank-lending in Indonesia. Estimation sample period is 2005Q1 to 2015Q2. Just like Bertay, et.al. (2015) and Ibrahim, et.al. (2016), the study used both the First Difference GMM and System GMM for estimation.

2.2.2. Individual Bank Performance

Glen & Mondragon-Velez, (2011) conducted a study of the effects of business cycles on the performance of the credit portfolio of commercial banks in developing countries. The study period is 1996-2008. The results obtained indicate that economic growth is the main determinant of the performance of the loan portfolio. While the interest rate is the second strongest determinant. The estimation results also showed that the relationship between loan loss provision and economic growth is non-linear in conditions where the economy is in a state of stress.

Guidara, Lai, Soumare, & Tchana (2013) conducted a study of co-movement between the level of capital buffer and business cycle for the six largest banks in Canada. The results found positive relationship between the two varaibels. The data used are quarterly data for the period of 1982-2010. The study results also showed that the implementation of the Basel regulatory framework does not affect cyclicality behavior of the banking industry in Canada. This suggests that banks in Canada are mostly well-capitalized.

Vithessonthi & Tongurai (2016) in his study analyze the effect of business cycles on the development of financial markets and the risk of individual banks. The samples used were 37 bank went public in seven countries in South America. The result shows that the business cycle

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 413

is significantly affecting the banking risks. Besides, the development of financial markets also improve capital ratios and reduce the level of risk exposure of banks, indicating that financial market developments have lowered the banking risk.

Psillaki & Mamatzakis (2017) analyzing the impact of financial regulation and structural reforms and their effects on the efficiency of the banking industry. The sample used was a bank from 10 countries in Eastern and Central Europe in the period 2004-2009. Scores of cost efficiency is estimated using stochastic frontier analysis. The results obtained show that structural reforms in the labor market and businesses have a positive impact on the bank’s performance. It was also found that credit regulations raise the cost efficiency of the banks. As well, it appears that banks with stronger capital has a higher cost efficiency.

As for Indonesia case, Winata & Viverita (2013) analyzed the effect of the bank’s income structure to market-based performance for listed-banks in Indonesia. The study was conducted using panel data for the period 2004-2012. The result obtained suggest that income diversification does not have a significant effect on the performance of the banks in Indonesia. Meanwhile, other variables such as total assets, asset-to-equity ratio, NPL and ROA have significant influence on the bank’s performance, while the variable cost-to-income and loan growth has no significant effect on the performance of the bank.

Table 1.

Empirical Literature Review: Business Cycle and Credit Cycle

No	Author		Analysis		Variables		Sample		Result
1	Ibrahim (2016)	•	Business cycle	•	GDP	•	Conventional Bank (Malaysia)	•	Pro-cyclical
		•	Bank lending	•	Loan	•	Sharia Bank (Malaysia)	•	Counter-cyclical
		•	Type of Banks
2	Bertay, et. al. (2015)	•	Business cycle	•	GDP	•	111 countries	•	Pro-cyclical
		•	Bank lending	•	Loan	•	State-Owned Bank	•	Counter-cyclical
		•	Ownership			•	Private Bank
3	Ferri, et.al. (2014)	•	Business cycle	•	GDP	•	Conventional Bank (Europe)	•	Bank lending significantly
		•	Bank lending	•	Loan	•	Cooperative Bank (Europe)		correlated with monetary
		•	Monetary policy	•	Policy Rate				policy
4	Pramono, et. al. (2015)	•	Business cycle	•	GDP	•	Conventional Bank (Indonesia)	•	Pro-cyclical
		•	Bank lending	•	Loan
		•	Counter-Cyclical Buffer	•	Counter-
					Cyclical
					Buffer

414Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

Table 2.

Empirical Literature Review: Business Cycle and Bank Performance

No	Author		Analysis		Variables		Sample		Result
1	Glen & Mondragon-	•	Business cycle	•	GDP	•	Conventional Bank	•	Non-linearity
	Velez (2011)	•	Credit Performance	•	Loan Loss		(Developing Countries)
					Provision
2	Guidara et. al. (2013)	•	Business Cycle	•	GDP	• Six largest banks (Canada)		•	Pro-cyclical
		•	Capital Buffer	•	Capital Buffer			• Basel Framework does not
		•	Basel Framework	•	Basel Dummy				affect cyclicality
3	Vinthessonthi &	•	Business Cycle	•	GDP	• 37 listed banks (South		•	Bank risk is procyclical
	Tongurai (2016)	•	Financial Development	•	Risk Exposure		America)
		•	Bank Risk	•	Capital Ratio
4	Psillaki &	•	Financial Regulation	•	Credit regulation	• 10 countries (Eastern &		•	Regulation increase
	Mamatzakis (2017)	•	Banking industry		dummy		Central Europe)		efficiency
			Efficiency	•	Cost efficiency			•	Stronger capital raise
				•	Capital ratio				efficiency
5	Winata & Viverita	•	Bank’s income structure	•	Income	•	Listed banks (Indonesia)	•	Income diversification
	(2013)	•	Market based		diversification				does not affect
			performance		ratio				performance
				•	Stock return

2.3. Conceptual Framework

This conceptual framework refers to model developed by Acharya and Naqvi (2012). The overall economy in this model consists of several sectors, namely, the banking sector, savers, borrowers (both savers and borrowers are referred to as households, for simplicity), and the entrepreneurial sector (corporation).

2.3.1. Bank Lending: Base Case

The framework is based on three-date model of a bank, in which at t = 0, the bank receives deposits D from risk-neutral investors (savers of the economy) with reservation utility . Depositors are compensated with rD, the (gross) rate of return on deposits – deposit rate. In t

=1, the bank makes investments in projects (loans) L, while holding a fraction of a deposits as liquid reserves r. In t = 2, the bank-funded projects either success or fail, with the probability

of success is given by θ. The bank observes θ after receiving deposits and sets rL, the (gross) rate of return on loans – lending rate.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 415

		t = 0		t = 1			t = 2
-	Bank raises deposits			- Bank suffers early		-	Bank projects either
-	Bank observes success			withdrawals, xD			succeed with probability
	probability θ and seta			- Bank incurs a penalty cost			θ of fail
	lending rate rL and			if xD > R		-	Payoffs divided
	borrowing rate rD						among parties

-Investments made and bank sets aside resserves R

Figure 8.

Three-date Model Framework

Bank reserves R are residual after the bank meets the loan demand:

(1)

The bank could experience withdrawals at, which is represented by random variable

given by, where. Thus, the total amount of with drawals atis given by. If , then the bank faces a liquidity shortage, and it incurs penalty, given by ,

which is proportional to the liquidity shortage, where. The bank owners’ problem is then summarized by:

(2)

subject to

(3)

and

(4)

where is the expectations operator over the distribution ofand profit,, is given by:

(5)

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 417

The manager earns income, , which can be interpreted as bonuses, which increases as the manager sell more loans. But the manager also faces a penalty, , if the principal conducts an audit and it is revealed that the manager had acted over-aggressively to increase loan volume by setting a loan rate lower than the one that maximizes the principal’s expected profits.

The managerial penalty is some proportion, , of the penalty cost incurred by the bank due to liquidity shortfalls. However, there is maximum penalty level received by the manager, , so that the managerial penalty is given by, where

represents the liquidity shortfall, if any, and . Thus, the net wage earned by the

manager is given by .

2.3.3. Bank Liquidity Flush

Acharya and Naqvi (2012) assume an economy in which entrepreneurs have access to projects that yield a terminal cash flow if it succeeds and zero otherwise. Macroeconomic risk is

given by .

The probability of success depends partly on the realization of the state variable, , and partly on the entrepreneurs’ effort decision, ε, which identifies whether the entrepreneur is

diligent () or shirks () in which case entrepreneurs extract a private benefit .

If the entrepreneur is diligent, the probability of success as before is given by , but in the

presence of shirking the probability of success is , where . Entrepreneurs promise to pay the risk-neutral investors who invest directly in their projects a face value of . The entrepreneur’s problem as maximizing the expected return is then:

(9)

Subject to the constraint:

(10)

and

(11)

Constraint (10) is the investor’s rationality constraint that says that the expected return to the investor must at least equal the investor’s reservation utility. Meanwhile constraint (11) means the expected entrepreneurial return conditional on the entrepreneur being diligent exceeds his expected return from shirking.

418Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

Then, there is such that, for , the entrepreneur’s offer to the investors are not enough to satisfy the investors’ reservation utility. Intuitively, if macroeconomic risk is sufficiently high, the probability of success is low, and thus, the entrepreneur has little incentive to exert effort and is better off by shirking and consuming his private benefit.

However, if the entrepreneurial projects are financed by banks instead of dispersed investors, then such moral hazard can be alleviated via monitoring. Formally, in the presence

of bank borrowing, entrepreneurs suffer a cost from shirking, say, . As long as , the entrepreneur has no incentive to shirk.

Because investors earn on average from depositing money in the bank, in the presence of entrepreneurial moral hazard, investors are better off by depositing their endowments in

banks. However, if , entrepreneurs can attract investors by offering them an expected return slightly above . In summary, if investors observe identically, then all investments are

channeled directly into entrepreneurial projects if and into banks if .

2.3.4. Theoretical Framework: Modified

This study employs conceptual framework developed by Acharya and Naqvi (2012) with some modification. The main framework implemented in this study is based on equation (5), which is:

First modification, we distinguish the risk in the model,, into two risk. First, macroeconomic

risk, , which plays significant role in determining deposit flush received by the banks.

Second, individual bank risk, , which is represented as the share of the performing loan compared to total loan made by the bank. By doing so, the conceptual framework in this study becomes as follow:

(12)

III. METHODOLOGY

3.1. Characteristic of the Data and Some Related Issues

This study employs quarterly unbalanced panel data from balance sheet of all conventional banks in Indonesia in the period 2005q1-2014q4. The main goal of this study is to measure magnitude of dynamic cyclicality between business cycles macro risk, credit cycles and risk cycle toward the performance of individual bank. Therefore, it is necessary for the empirical model to be able describe the dynamic relationship of the Tri-Cycles and its influence on the performance of the bank.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 419

One of the benefits of the use of panel data is able to provide interpretation that can meet these objectives, in term of variation between individual and over-time (Baltagi, 2005). However, the utilization of unbalanced panel data, although not problematic, requires special attention. First, from the point of view of classical assumption, it makes the OLS estimators remain consistent and unbiased, but the standard error will be biased. Second, ANOVA methods cannot be perfectly applied for unbalanced data thus makes the property of unbiased in the optimal model of standard ANOVA are not met. Third, the asymptotic distribution of critical values of ​unbalanced panel data becomes unbalanced. Nonetheless, the use of unbalanced panel data remains possible and common statistical software such STATA has automatically accommodated this type of (Cameron & Trivedi, 2009). This study uses STATA 13 MP to perform estimation and data processing.

One of the most popular models for cycle analysis is dynamic panel model as used by Ferri, et.al. (2014), Bertay, et.al. (2015), Pramono, et.al. (2015) and Ibrahim (2016) in a study somewhat similar to this study. This model was chosen because of its ability to capture the dynamics between the two variables. So that it is very suitable for use in analyzing the relationship between two variables cycle.

The dynamic model in practical is implemented by entering the lag of the dependent variable as one of the independent variables (Baltagi, 2005). Dynamic panel model is characterized by the presence of more than one source of time persistence. Such conditions create dynamic panel model cannot be separated from the autocorrelation lag because of the presence of one of the dependent variables as independent variables. In addition, this model also cannot be separated from the heterogeneity among individuals who became observation.

Because of that, dynamic panel model is automatically not met the BLUE (Best Linear Unbiased Estimator) assumptions as owned OLS model standard and standard assumptions of GLS models (Baltagi, 2005). This happens because in dynamic panel model there is correlation between the independent variables with the error term. Because of that, the OLS estimators will be biased and inconsistent. Moreover, the standard GLS estimator cannot be used because there is a correlation between the variables predetermined by the error term. Furthermore, GLS models with instrument variable (IV) also cannot be used because of although it produces a consistent model, it did not provide an efficient parameter. This is due to the model GLS-IV does not use all the conditions of moment conditions.

Therefore, the use of dynamic panel model is not necessary to test classic assumptions on the data samples used. Dynamic panel model was essentially developed to be able to accept the conditions in which the classical assumptions are not met.

Specifically, dynamic panel model that will be used in this research is Panel Vector Autoregressive (PVAR) developed by Abrigo and Love (2015) based on GMM style estimation from Arellano and Bond (1991). Sprecifically, the GMM model is used in combination with a

420Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

robust estimator of Windmeijer (2005) to avoid problems of overidentification and downward bias (Baltagi, 2005).

3.2. Overview of Panel Vector Autoregressive (PVAR) Model

In line with the aim of this study, cyclical relationship between the Tri-Cycles is the object to be observed. Thus, this study needs such method that can observe potential bi-directional relationship between each cycle. Because of that, this study employs Vector Autoregressive (VAR) approach. VAR methodology is categorized as an extension of autoregressive (AR) model in term of its multi-variate characteristic. Furthermore, it also resembles simultaneous-equation modeling (SEM) style estimation (Gujarati and Porter, 2009). The main difference is, each variable in the model is treated as endogenous variable and lag of every variable in the system is considered as independent variable.

This approach is then very suitable to fulfill the aim of this study. By placing every variable as dependent variable, it can observe bi-directional relationship of each cycle and also shows how each cycle affects bank performance. In the context of business cycle analysis, the approach might give insight of the dynamics between business cycle and financial cycle, which attracts big concern in the development of the literatures.

The other advantage is it has rich of features. VAR estimation has features of Impulse Response Function (IRF), Forecast-Error Variance Decomposition (FEVD), and also Granger- Causality Test. These features make VAR estimation very popular for policy simulation. IRF describe describes response of each variable due to one standard deviation in other variable. Meanwhile FEVD decomposes degree of impact of each variable to the dependent variable. Granger-Causality test is useful in examining the potential bi-directional relationship between variables.

However, VAR approach has several disadvantages. First, it can be applied even as an a-theory approach. Researcher does not need basic theory and can simply put any variable into VAR system. This disadvantage then has been avoided in this study since the selection process of the variables in this study is all based on the theoretical framework as explained in Section

3.Second, the coefficient result from VAR estimation cannot be interpreted directly. The result is interpreted only it term of its direction (positive or negative) and significance.

Essentially, VAR estimation is developed for time-series process. However, as the availability of cross-individual data is increasing, VAR is getting popular to be implemented in panel estimation. However, standard built-in statistical package is not yet available to estimate Panel VAR. This study then employs PVAR user-based package developed by Abrigo and Love (2015) which uses GMM estimation to estimate Panel VAR.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 421

Theoretically, Panel VAR have the same structure as time-series VAR model. Each variable is treated as endogenous variable and also interdependent. The main difference is Panel VAR also observe cross-sectional dimension of the data. Panel VAR is also noticeable for several advantages, as revealed by Canova and Cicarelli (2013) that Panel VAR: (i) captures both static and dynamic interdependencies; (ii) in unrestrictive style treats individual variation; (iii) incorporates coefficient variations and the variance of the shocks in term of time series; and (iv) examine cross-sectional dynamic heterogeneities.

Think of is stacked version of variable , while the is vector of variables for unit

of i.e., . Individual is represented by . Meanwhile represents time. Common time-series VAR empirical form is:

Meanwhile Panel VAR empirical form is then:

where is random disturbances in the form of vector and and depend on the unit of observation, and is the lag operator.

3.3. Empirical Model

Recalling equation (13) in Chapter 3, the bank profit is function is as follow:

(13)

So that the bank profit is function of:

(14)

Due to data characteristic, rather than using Loan Rate () and Deposit Rate (), this study use data of Net Interest Margin, , which is difference of the and . So equation

(14) becomes:

(15)

Based on equation (15), an empirical model might be constructed as follow:

(16)

422Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

where is intercept, and is parameter. Meanwhile is the error-term. Both and are lag operator.

= Bank profit

= Share of performing loans to total loans made by the bank

= Macroeconomic risk, Indonesia CDS 1Y – Spread

= Total bank loans

= Total bank deposits, represented by total Third Party Funds = Net interest margin (NIM)

3.4. List of Variables

Table 3 gives complete list of variables employed in the empirical estimation. In total, the data set contain unbalanced panel data of 150 individual conventional bank balance sheet in Indonesia.

Focus variables on this study are comprised of four variables. First, the business cycle macro risk, which is represented by CDS spread. In line with theoretical framework on this study which is based on Acharya and Naqvi (2012), this study then uses CDS spread rather than GDP which is very common to represent business cycle. On their paper, Acharya and Naqvi (2012) specifically recommend to use commercial paper spread as the measure of business cycle fluctuations. However, due to limitation and irregularity of commercial paper spread data in Indonesia, this study then employs credit default swap (CDS) spread data, which also represents macroeconomic risk.

The second focus variable of this study is bank lending, which represents credit cycle. The data come from individual bank balance sheet. All bank balance sheet data are acquired from website of Bank Indonesia and Otoritas Jasa Keuangan, which are based on monthly and quarterly report of bank balance sheet. Bank lending data employed in this study uses outstanding credit data on the balance sheet. The data is then transformed into natural logarithm and then extracted to its cycle component using Hodrick-Presscott Filter. The third focus variable

is performing loan (), which represents the risk cycle. The data acquired from Net Non-

Performing Loan (NPL) data, specifically = 100 - NPL. The data is then transformed into natural logarithm and the cycle component is extracted. Lastly, the fourth focus variable is bank performance, which is represented by the bank quarter profit. The data is also transformed into natural logarithm and then extracted to its cycle component.

Besides of the focus variables, there are also two other variables which are theoretically important, the deposit and net interest margin. The deposit data is sum of the total of third party fund, comprised of giro, savings and time-deposit. The data is transformed into natural log and its cycle component is extracted. Meanwhile the Net Interest Margin (NIM) data is employed as substitute of loan rate and deposit rate. The use of this substitute is due to bank

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 423

do not report their loan rate and deposit rate in their balance sheet, but they report their net interest margin, which is the difference of both. The data is extracted into its cycle component by using HP filter.

Table 3.

List of Variables

Variables	Symbol	Unit	Source	Treatment
Business Cycle Macro Risk
Credit Default Swap (CDS) –		Percentage	Bloomberg	Cycle
Spread
Credit Cycle
Bank Lending		Million Rp	OJK	Cycle of Natural Log
Risk Cycle
Performing Loan		Percentage	OJK	Cycle
Bank Performance
Profit		Million Rp	OJK	Cycle of Natural Log
Bank Specific–Theoretically
Important
Net Interest Margin (NIM)		Percentage	OJK	Cycle
Deposit–Third Party Fund		Million Rp	OJK	Cycle of Natural Log

*OJK stands for Otoritas Jasa Keuangan or Financial Service Authority

IV. RESULT AND ANALYSIS 4.1. Estimation Result

The suitability of the theoretical framework and rich feature of VAR estimation method has made it possible to answer all research questions in only one estimation process. VAR estimation is conducted with variables lag of 3 and instrument lag of 1/9. The complete result of tests and estimations are in the Appendix section.

Table 4 presented unit-root stationarity test based on Panel ADF-Fisher tests. All variables are stationary at level. Thus, there is no need to conduct cointegration. This condition means VAR model is eligible to be applied to analyze the data. This study applies VAR estimation based on GMM developed by Abrigo and Love. If the variables are found to be non-stationary at level, then VEC Model need to be considered.

Table 4.

ADF-Fisher Panel Unit-Root Test

Variable	ADF-Fisher Prob		Variable	ADF-Fisher Prob
business cycle macro risk	0.000		nim	0.000
credit cycle	0.000		deposit	0.000
risk cycle	0.000		profit	0.000

424Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

With selected lag specification, VAR stability check indicates stability of the model (Figure 10). It can be concluded as all eigenvalue lie inside the unit circle. It means that all eigenvalue has value equal to or less than 1. This stability condition is necessary for VAR estimation otherwise the estimation result become unreliable for analysis and the VAR system need to be recalibrated.

Roots of the companion matrix

1
.5
Imaginary 0
-.5
-1
-1	-.5	0	.5	1

Real

Figure 10.

VAR Stabtility Check Result

Results presented on this section are arranged following 8 research questions of this study. Each sub-section (from 5.1.1. to 5.1.8.) address one research question. The results presented are VAR estimation result, Granger Causality wald-exogeneity test result, and Cholesky IRF result. From those result each research question can be comprehensively addressed so that the aim of this study can be fulfilled.

4.1.1. Tri-Cycles Dynamics

Estimation result from VAR and Granger Causality test reveal bi-directional cyclicality between business cycle macro risk and credit cycle. VAR estimation result indicate that shock in the credit cycle has lagged impact toward the business cycle macro risk (Table 5). On the other hand, shock in business cycle macro risk also has significant impact toward credit cycle. Granger Causality test results resembles the VAR estimation result. It reveals bi-direction relationship between business cycle macro risk and credit cycle. From the result can be inferred that the credit cycle granger-cause the business cycle macro risk. While on the opposite, the business cycle macro risk granger-cause the credit cycle. Cholesky Impulse Response Function shows that shock in the credit cycle has lagged impact over the business cycle macro risk and the pattern is stable. Meanwhile business cycle macro risk has considerably unstable response toward shock in credit cycle (Figure 11).

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 425

Table 5.

VAR Estimation and Granger Causality Result

VAR Estimation Result

Granger-Causality

Variable		Independent	Coefficient	Prob > |z|		Prob >
	Dependent				Variable
						chi2
Business			7.2628	0.002		0.000
Cycle			-6.9603	0.015
Macro			2.5197	0.035
Risk	)		-0.0063	0.000		0.000
(
			0.0044	0.002
Credit			-0.0026	0.046
Cycle (	)
Business			-0.0054	0.665		0.024
Cycle			0.0044	0.766
Macro			-0.0485	0.004
Risk
	)		-0.1132	0.004		0.000
(
Risk Cycle			-0.0261	0.628
			-0.0030	0.953
(	)
Credit			0.0006	0.297		0.000
	)		0.0009	0.102
Cycle (			0.0018	0.002
Risk Cycle			5.6197	0.000		0.000
(	)		1.5556	0.067
			1.4232	0.047

Estimation result from VAR and Granger Causality test also reveal bi-directional cyclicality between business cycle and risk cycle. VAR estimation result indicate that shock in the risk cycle has lagged impact toward the business cycle macro risk (Table 5). On the other hand, shock in business cycle macro risk also has significant indirect impact toward risk cycle. Granger Causality test results resembles the VAR estimation result. It reveals bi-directional relationship between business cycle macro risk and risk cycle. From the result can be inferred that the risk cycle granger-cause the business cycle macro risk. While on the opposite, the business cycle macro risk also granger-cause the risk cycle. Based on IRF pattern, risk cycle exhibit unstable response toward shock in business cycle macro risk (Figure 11).

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 427

it reveals two-way relationship between credit cycle and risk cycle. From the result can be inferred that the credit cycle granger-causes the risk cycle. While on the opposite, the risk cycle also granger-causes the credit cycle. The IRF pattern reveals considerably unstable response of risk cycle toward shock in credit cycle. Meanwhile at the opposite, credit cycle has stable and increasing response toward shock in risk cycle.

4.1.2. Tri-Cycles and Bank Performance

Estimation result from VAR and Granger Causality test reveal one-way cyclicality between business cycle macro risk and bank performance (profit). VAR estimation result indicate that shock in business cycle macro risk has impact toward bank performance (Table 6). On the other hand, as shown by the estimation result, shock in bank performance does not seem to have significant impact toward business cycle macro risk. Granger Causality test results confirm the VAR estimation result. It reveals one-way relationship between business cycle macro risk and bank performance. Business cycle macro risk granger-cause bank performance. While on

Table 6.

Estimation Result: Business Cycle and Bank Performance

VAR Estimation Result

Granger-Causality

Variable		Independent	Coefficient	Prob > |z|		Prob >
	Dependent				Variable
						chi2
Business			-0.0117	0.312		0.246
Cycle			0.0193	0.077
Macro	Risk		-0.0006	0.957
(	)
			0.1404	0.227		0.006
Bank			0.0275	0.827
			-0.2195	0.014
Performance
(	)
Credit Cycle			-0.0003	0.473		0.530
			0.0004	0.345
( )
			-0.0005	0.232
Bank			-13.8393	0.063		0.011
Performance			29.0063	0.003
(	)		-13.4835	0.010
Bank			-0.0203	0.223		0.430
			-0.0074	0.666
Performance
			0.0152	0.307
(	)
			0.0994	0.036		0.075
Risk	Cycle		-0.0363	0.524
(	)		0.0493	0.284

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 429

risk_cycle : profit

profit : risk_cycle

.2	95% CI	IRF	.05
.1
			0
0
-.1				95% CI	IRF
			-.05
0	5	10	0	5	10
	step			step

impulse : response

impulse : response

Figure 12.

Impulse Response Function: Business Cycle Macro Risk and Bank Performance

Shock in the the risk cycle has weak impact toward bank performance (Table 6). On the oppposite, shock in bank performance does not has impact toward the risk cycle. Granger Causality test results resembles the VAR estimation result. It can be inferred that the risk cycle weakly granger-causes bank performance. While on the opposite, bank performance does not granger-causes the risk cycle. The Cholesky IRF of the risk cycle and bank performance reveals unstable response pattern of bank performance toward shock in risk cycle.

4.1.3. Business Cycle Macro Risk Dynamics: Credit Cycle and Risk Cycle

Figure 13 present Cholesky IRF of response of the credit cycle and risk cycle due to shock from the business cycle macro risk. The left side is the IRF of the credit cycle. Meanwhile the right side is the IRF of the risk cycle. By comparing the left picture and the right picture, it can be seen different response of the credit cycle and the risk cycle toward shock of business cycle macro risk. Credit cycle reveals stable response pattern toward shock in business cycle macro risk. Meanwhile risk cycle exhibit cyclical response toward shock in business cycle.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 433

macroeconomics. Financial accelerator phenomenon addresses the dynamics of financial cycle which propagates to output fluctuation. However, the formal model presented by them is not suitable enough for study at individual bank level. The framework is too macro for such study.

This study unexpectedly exhibits the financial accelerator phenomenon in Indonesia banking system in the period of 2005q1 to 2014q4. VAR estimation and Granger-Causality yield result that support the existence of financial accelerator phenomenon. However, this claim only applies in financial market context, because this study employs macroeconomic risk – rather than GDP – as representation of business cycle. VAR estimation result give a sign of significant relationship between credit cycle to business cycle macro risk and between risk cycle to business cycle macro risk.

As in BGG (1999), such phenomenon is caused by the existence of credit market frictions. In such ideal world, financial market is perfectly efficient so that every agent in economy can recalculate their decision instantly if any shock happens. Three features of credit market frictions in the BGG model are (i) money and price stickiness; (ii) lags in investment; and (iii) heterogeneity among firms. This study, even though is not specifically intended to examine BGG framework, accommodates those three features in some way. First, money and price stickiness are accounted in the estimation with net interest margin as representative of cost of fund. Second, lags in investment, which is represented by credit cycle, is accounted in the estimation which the result reveales lagged / indirect / non-contemporaneus effect between credit cycle and business cycle macro risk. Third, the heterogeneity of firms is presented by the span of the data which cover the whole conventional bank population in Indonesia. In total, the observation covers up to 119 banks which differs significantly in term of size and market specialization. However, since BGG model is not the fundamental of this study, this study cannot infer any conclusion based on the model. There is a room for future study to examine this phenomenon with BGG model in Indonesia banking industry.

4.2.2. Inside the Financial Cycle: Credit Cycle and Risk Cycle

As Acharya and Naqvi (2012) published their work about the “Seeds of Crisis”, scope of the business cycle once again became more comprehensive. Financial cycle, which was solely represented by credit cycle, started to account the importance of risk dynamics. Their model talks about the process of building-up of of a bubble in the economy, which then turns into crisis when it bursts.

This study only implements the beginning phase of the model presented by Acharya and Naqvi (2012). This study focuses on the propagation of shock in business cycle macro risk toward balance sheet of individual bank. The whole study is conducted in the business cycle context, in which all variables are extracted to its cycle component.

434Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

Estimation result exhibits pro-cyclicality behavior of the credit cycle toward business cycle macro risk. Deterioration of business condition, represented by increase in macroeconomic risk, will further decrease bank lending. When this context is about to happen, such counter-cyclical policy by the regulator is very important to prevent deeper deterioration of the economy.

Meanwhile risk cycle is shown to be unstable along the fluctuation in business cycle macro risk. The relationship is also pro-cyclical. Deterioration of business condition, represented by increase in macroeconomic risk, will further decrease the rate of performing loan. Therefore, such counter-cyclical policy which focuses on risk cycle need to be enhanced to prevent deeper contraction.

4.2.3. Fund Allocation and Performance Risk

As the bank get liquidity flush resulted from business cycle swing, it has two option to increase risk; it can either simply lend more money to borrower with similar risk profile or conduct risk- shifting activity by giving credit to other borrower with worse risk profile. In other words, the bank can simply switch into riskier assets. Unfortunately the data and capability of the model applied in this study cannot observe this phenomenon closer. The only available channel to explain risk-taking activity in the model is represented by the bank distribute more credit.

This feature then make the model assume that risk in the model are solely caused from increasing amount of the credit. Meanwhile the risk profile of the projects to invested by the banks does not vary. This disadvantage gives the room for further improvement of the model. Separating low-risk project and high-risk project will make the model able to explain the risk- taking behaviour more resourceful. The dynamics of risk cycle is then no more solely depend on the growth of credit, but also structure of asset of the bank.

4.2.4. Liquidity and Risk-Taking

As the framework predict, we get a sign of pro-cyclicality between deposit and credit cycle. In the model, risk-taking activity is proxied by bank-lending. When the bank lend more money, that means the bank take more risk and this is the exact moment of the emergence of the seeds of a crisis. Flush of liquidity into the bank will induce risk-taking behavior, which is represented by the bank channeling more credit into the economy.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 435

deposit : credit_cycle

1.5	95% CI	Cumulative IRF
1
.5
0
	0	5	10
		step

impulse : response

Figure 17.

IRF of Credit Cycle and Deposit

In the next episode of the model, this risk-appetite will stimulate excessive credit volume channeled into the market, which then result in asset price bubbles. However, that episode is beyond the scope of this study and is subject for further research. From bank’s internal perspective, the model tells that this behavior is supposed to result in punishment if the manager of the bank get “caught” for practicing excessive lending. Another concern of the bank manager behavior in the moment of liquidity flush is that the manager might underprice of risk of projects or a credit. This will result on asset prices bubble in the long run.

Further the model also addresses a more detail aspect of excessive lending activity. The model tells that flush of liquidity into the bank will trigger excessive lending through the bank manager will set lower rate of lending. This logic is strongly supported by the estimation result. Net Interest Margin (NIM), which represents rate of lending and deposit react negatively to shock in deposit. This counter-cyclical behavior indicate that banks tend to lower its lending rate, as shown by low NIM. The result also confirm counter-cyclical relationship of credit cycle to shock in NIM. Lower NIM will trigger over-lending and then will result on bigger compensation received by the bank manager.

436Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

deposit : nim

nim : credit_cycle

0	95% CI	Cumulative IRF	0
-10			-.01
-20			-.02
-30			95% CI	Cumulative IRF
			-.03
0	5	10	0	5	10
	step			step
impulse : response			impulse : response

Figure 18. NIM, Deposit, and Credit Cycle

At practical level, the story is more interesting. Many banks in Indonesia are outsourcing some of its activity to the third-party service provider. This behavior makes Otoritas Jasa Keuangan (OJK) or Indonesia Financial Service Authority impose a regulation, even though very loose, to remind banks to pay close attention about this outsourcing practice. One of the most popular practice is on marketing division, both at deposit side and credit side. The regulation states that the bank may outsource some only its low-risk activities to the third-party. Some activities on this classification for instance are call center services and telemarketing services, as clearly stated by the regulation (OJK, 2017). This regulation is a clear example that macroprudential regulation has thouched deeply into the bank daily activity. On the one hand, this regulation is a good example of comprehensive microprudential regulation in Indonesia. On the other hand, as much concerned by this study, every regulation has tendency to overstring bank activity, which may impact the bank performance.

4.2.5. Performance Dynamics and Reponse of Banker and Regulator

This study further contributes by accommodating bank performance dynamics in the discussion on context of the business cycle dynamics. For the bankers, profit is one of the most important barometer of their success. This notion applies since bank as a financial institution is basically similar with other business entity, in sense that their objective is to maximize profit.

The model presented by Acharya and Naqvi (2012) is very suitable not only to discuss policy-making context, but also profit maximization behavior of the bank. The existance of relationship between Tri-Cycles and bank performance in the model make it possible to explain the sensitivity of bank performance along fluctuation of business cycle.

Business cycle, through credit cycle and risk cycle, is shown to play important role in determining bank performance (profitability). The result obtained in this study also shows that

438Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

Figure 19 wraps up the conclusion of this study. First, business cycle macro risk and credit cycle exhibit two-way strong relationship, in which shock in credit cycle has significant impact toward business cycle macro risk and vice versa. Second, business cycle macro risk and risk cycle shows bi-directional strong relationship. Shock in business cycle macro risk has significant impact toward risk cycle. However, shock in risk cycle only has weak impact to business cycle macro risk. Third, credit cycle and risk cycle both are inter-dependent which means shock in each cycle has significant impact toward its counterpart.

Fourth, business cycle macro risk and bank performance indicates one-way relationship. Shock in business cycle macro risk has weak impact to bank performance. Fifth, credit cycle and bank performance has one-way relationship in which credit cycle has significant impact toward bank performance. Sixth, risk cycle and bank performance has one-way relationship in which risk cycle has weak impact toward bank performance.

Seventh, when comparing impact of shock of business cycle macro risk toward credit cycle and risk cycle, it can be inferred that risk cycle is more sensitive. The Cholesky IRF reveals response of response of risk cycle toward shock in business cycle macro risk. Last, eighth, bank performance seems to be sensitive toward both shock of risk cycle and credit cycle.

5.2. Notable Contributions

This study exhibits the initial finding of the existence of financial accelerator phenomenon in Indonesia. The result show dynamics of financial cycle – in the form of credit cycle and risk cycle - preceded the business cycle macro risk. This study then has contributed to the business cycle literature by revealing this phenomenon especially in emerging country. This finding is very important in order to deeply understand the financial cycle characteristic in Indonesia. Further research need to extend the analysis by examining real output fluctuation, which was not addressed in this study.

This study is also one of the first to employ CDS spread as alternative representative of the business cycle fluctuation in Indonesia. Especially since previous studies mostly employed GDP as representative of business cycle fluctuation. In fact, Indonesia did not experience GDP contraction in 2007/8 financial crisis. So GDP can explain almost nothing when examining business cycle fluctuation in the episode of 2007/8 financial crisis in Indonesia.

Furthermore, most studies addressed financial cycle issue by only focusing on credit cycle. This study is then one of the first to examine financial cycle in the form of credit cycle and risk cycle at individual bank level in Indonesia. Risk cycle – besides of credit cycle – has attracted many attentions after 2007/8 crisis. Credit cycle can no more solely explain the dynamics of financial cycle and its relationship with business cycle.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 439

In term of econometrical method, this study attempted to employ a newly developed statistical package – Panel VAR / PVAR – from Abrigo and Love (2015). VAR approach which had long historical implementation in time series context is now possible to be implemented in panel context because of this package.

5.3. Recommendations

From the stance of the policy maker, specifically banking regulator, the result obtained in this study reveal the cyclical and unstable response of the financial cycle (credit and risk) due to shock in business cycle macro risk. The result then gives important insight for the regulator in the implementation of counter-cyclical policy to maintain the bank balance sheet stability.

For market participants, especially the bankers, this study has revealed unstable response of profit due to the shock in business cycle through the risk and credit channel. This result somewhat strengthens the existence of the notion of “high risk – high return” in the banking business. The bankers then need to give special attention in the internal bank risk cycle, along with the internal bank credit cycle. As modeled by theoretical framework used in this study, the bankers are assumed to get more bonuses by selling more credit. The bankers might need to find the alternative scheme of incentive in lending system. Separation of the marketing department with the credit approval analyst is one of good example. The bankers might also consider the effectiveness of flat incentive system in credit selling. However, this study does not intend to examine these alternatives. Further research is needed to provide comprehensive discussion toward this topic.

5.4. Disadvantages and Suggestion for Further Research

From the point of view of econometrical approach, Panel VAR approach employed in this study gives satisfying solution. If it seems possible, further study might explore a more complex statistical setting such as Panel VECM approach to conduct more comprehensive examination.

For further research, it might be very fruitful if the design of this research can be uplifted to cover cross-country experience, such as ASEAN countries. By doing so, the study will reveal cross-country variation of the Tri-Cycles dynamics. Such setting is very important regarding the fact that Indonesia as single country did not experience contraction in GDP cycle in financial crisis 2007/8. While other countries in ASEAN such as Singapore and Malaysia did.

Finally, recalling the complete formal model presented by Acharya and Naqvi (2012), the design employed in this study has not yet captured the whole feature of the model. Their model essentially was designed to explain the full story of the birth of crisis, which they called as “The Seeds of Crisis”. So, the story presented by this study is only the beginning phase of the complete tale covered by the model. Further research certainly need to address the model in complete setting.

440Buletin Ekonomi Moneter dan Perbankan, Volume 19, Nomor 4, April 2017

REFERENCES

Abrigo, Michael and Love, Inessa. (2015). Estimation of Panel Vector Autoregression in Stata: a Package of Programs. University of Hawaii Working Paper.

Acharya, V., & Naqvi, H. (2012). The Seeds of a Crisis: A Theory of Bank Liquidity and Risk Taking Over the Business Cycle. Journal of Financial Economics, 106, 349-366.

Arellano, M., & Bond, S. (1992). Some Tests of Specification for Panel Data: Monte Carlo Evidence and Application to Employment Equations. Review of Economics Studies, 58, 277-297.

Baltagi, B. H. (n.d.). Econometric Analysis of Panel Data (Vol. 3). West Sussex, England: John Wiley & Sons Ltd.

Bank for International Settlement. (2017, January 3). Basel Committee for Banking Supervision. Retrieved from BIS Website: http://www.bis.org/bcbs/history.htm.

Bernanke, S. B., Gertler, M., & Gilchrist, S. (1999). The Financial Accelerator in a Quantitative Business Cycle Framework. In J. B. Taylor, & M. Woodford, Handbook of Macroeconomics (Vol. 1, pp. 1341-1393). Elsevier Science B.V.

Bertay, A. C., Demirguc-Kunt, A., & Huizinga, H. (2015). Bank Ownership and Credit Over the Business Cycle: Is Lending by State Banks Less Procyclical? Journal of Banking and Finance, 50, 326-339.

Blundell, R., & Bond, S. (1998). Initial Conditions and Moment Restrictions in Dynamic Panel Data Models. Journal of Econometrics, 87, 115-143.

Burns, A. F., & Mithcell, W. C. (1946). Measuring Business Cycle. In N. B. (NBER), Studies in Business Cycles. New York.

Cameron, A. Colin and Trivedi, Pravin K. (2009). Microeconometrics Using Stata. Texas: StataCorp LP.

Canova, Fabio and Ciccarelli, Matteo. (2013). Panel Vector Autoregressive Models: A Survey. European Central Bank Working Paper Series No 1507 / January 2013.

Federal Deposit Insurance Company. (2017, January 3). FDIC. Retrieved from FDIC: https:// www.fdic.gov/bank/historical/sandl/

Ferri, G., Kalmi, P., & Kerola, E. (2014). Does Bank Ownership Affect Lending Behavior? Evidence from the Euro Area. Journal of Banking and Finance, 48.

Glen, J., & Mondragon-Velez, C. (2011). Business Cycle Effecs on Commercial Bank Loan Portfolio Performance in Developing Economies. Review of Development Finance, 150-165.

Tri-Cycles Analysis on Bank Performance: Panel Var Approach 441

Guidara, A., Lai, V. S., Soumare, I., & Tchana, T. F. (2013). Banks’ Capital Buffer, Risk and Performance in the Canadian Banking System: Impact of Business Cycles an Regulatory Changes. Journal of Banking and Finance, 3373-3387.

Gujarati, D. N., & Porter, D. C. (2009). Basic Econometrics (Vol. 5). Boston: McGraw-Hill Irwin.

Ibrahim, M. H. (2016). Business Cycle and Bank Lending Procyclicality in a Dual Banking System. Economic Modelling, 55.

Jacobs, J. (1998, February). Econometric Business Cycle Research: An Assessment of Method. University of Groningen Working Paper.

Kiyotaki, N. (2011, September 1). A Perspective on Modern Bussiness Cycle Theory. FRB Richmond Economic Quarterly, 97(3), pp. 195-208.

Kydland, F. E., & Prescott, E. C. (1982). Time to Build and Aggregate Fluctuations. Econometrica, 50, 1345-1370.

Minetti, R., & Peng, T. (2013). Lending Constraints, Real Estate Prices and Business Cycle in Emerging Economies. Journal of Economic Dynamics and Control, 37.

Otoritas Jasa Keuangan. (2017). Surat Edaran OJK Nomor 11/SEOJK.03/2017 Tentang Prinsip Kehati-hatian bagi Bank Umum yang Melakukan Penyerahan Sebagian Pelaksanaan Pekerjaan kepada Pihak Lain.

Pramono, B., Hafidz, J., Adamanti, J., Muhajir, M. H., & Alim, M. S. (2015). Dampak Kebijakan Countercyclical Buffer Terhadap Pertumbuhan Kredit di Indonesia. Working Paper Bank Indonesia.

Psillaki, M., & Mamatzakis, E. (2017). What Drives Bank Performance in Transitions Economies? The Impact of Reforms and Regulations. Research in International Business and Finance, 578-594.

Romer, D. (2012). Advanced Macroeconomics. New York: McGraw-Hill.

Tinbergen, J. (1940, February). Econometric Business Cycle Research. The Review of Economic Studies, 7, 73-90.

Vithessonthi, C., & Tongurai, J. (2016). Financial Market Development, Business Cycle, and Bank Risk in South America. Research in International Business and Finance, 472-484.

Winata, A., & Viverita. (2013). Analisis Pengaruh Struktur Pendapatan Bank terhadap Market- Based Performance: Studi Empiris Bank yang Terdaftar pada Bursa Efek Indonesia Periode 2004-2012. Working Paper Fakultas Ekonomi dan Bisnis, Universitas Indonesia.

Windmeijer, F. (2005). A Finite Sample Correction for the Variance of Linear Efficient Two-Step GMM Estimators. Journal of Econometrics, 126, 25-51.