Bulletin of Monetary Economics and Banking, Vol. 22, No. 2 (2019), pp. 137 - 162

p-ISSN: 1410 8046, e-ISSN: 2460 9196

MONETARY POLICY TRANSMISSION AND CREDIT CARDS:

EVIDENCE FROM INDONESIA

K.P. Prabheesh1, R. Eki Rahman2

1Indian Institute of Technology Hyderabad, Hyderabad, India. Email: prabheeshkp@gmail.com

2Bank Indonesia Institute, Bank Indonesia. Email: eki.r@bi.go.id

ABSTRACT

This paper empirically tests the dynamics of credit cards and monetary policy in the context of Indonesia. Using monthly data from 2006 to 2018 and a structural vector autoregressive model, our findings indicate that credit card usage is mainly driven by Indonesia’s fast economic growth over the last decade, which indeed reflects the role of credit cards in consumption smoothing. The study also finds that monetary policy transmission through the lending channel is weak, with a more prevalent role for exchange rates and global oil prices in the transmission process.

Keywords: Monetary policy; Structural vector autoregression; Credit cards.

JEL Classification: E44; E50.

Article history:
Received	: February 20, 2019
Revised	: March 30, 2019
Accepted	: June 13, 2019

Available online : July 30, 2019

https://doi.org/10.21098/bemp.v22i2.1039

140	Bulletin of Monetary Economics and Banking, Volume 22, Number 2, 2019

Figure 3. Usage of Credit Card for Purchase and Cash Withdrawal

The figure represents the proportion of credit card usage (value of transactions) in terms of purchase and cash withdrawal. The values are represented in percentages for two years, i.e., 2006 and 2018, and data come from CEIC and Bank Indonesia website.

2006

2018

3%

6%

94%					97%
Cash Withdrawal		Purchase		Cash Withdrawal			Purchase

II.LITERATURE REVIEW, HYPOTHESES, AND METHODOLOGY

According to Fulford and Schuh (2018), the availability of credit cards helps individuals make three important decisions: First, credit cards help smooth their consumption when their income drops; hence, it is considered as an instrument to meet the precautionary and liquidity needs. Second, they can be used to revolve debt over the short and long term, and credit cards are thus a way of allocating life cycle consumption. Finally, as a means of payment, the amount spent with credit cards comprises part of consumer expenditures. Therefore, the economic implications of credit cards become more broader as their usage increases.

The literature on credit cards, especially related to monetary policy, generally focuses on the following: 1) the role of central banks in digital money and their independence in terms of monetary policy and 2) the implications of credit cards on monetary policy transmission. Regarding the first focus, studies suggest that the substitution of money with any alternative payment option, such as credit cards, debit cards, and digital currencies, reduces the overall demand for money in the economy (Akhand and Milbourne, 1986; Yilmazkuday and Yazgan, 2011). The usage of credit cards reduces the demand for money for transaction purposes, since they can be used as a medium of exchange in the transactions (Mandell, 1972). Hence, the traditional approach of implementing a monetary policy based on changing the monetary base might not be effective; for instance, the possibilities of raising seigniorage (income from printing money) using monetary policy would be limited. Therefore, central banks could be forced to depend on governmental financial support for their operational needs, thus affecting the independence of their monetary policy (Friedman, 1999; Freedman, 2000; Goodhart, 2000; Woodford, 2000).

Regarding the second focus, credit cards and monetary transmission, various challenges are posed by credit cards in the implementation of monetary policy. The usage of credit cards doubles the velocity of money, since the cash proceeds

Monetary Policy Transmission and Credit Cards: Evidence from Indonesia	141

from the sale of goods can be reused to pay the debt on credit card purchases. Theoretically, the velocity of money is inversely proportional to the demand for money, and hence lower demand for money due to credit card usage increases the velocity of money (Geanakoplos and Dubey, 2010). A high level of money velocity increases inflation, which can deviate from the central bank’s inflation target. Similarly, the lower demand for money reduces the demand for central bank’s reserves, which in turn shrinks the central bank’s balance sheet. The lower demand thus reduces the central bank’s ability to influence the short-term interest rate through open market operations (Friedman, 1999, 2000). Moreover, the evidence suggests that the central bank’s balance sheet is potentially reduced through the increased use of electronic money, with serious implications on the bank’s ability to effectively manage monetary policy and carry out its functions as lender of last resort (Bank for International Settlements, 2015).

It is also argued that inflation in the United States in the 1970s and early 1980s coincided with the introduction of credit cards (Geanakoplos and Dubey, 2010). Credit card usage can stimulate spending, since consumers underestimate or forget credit card purchases, because the act of paying by credit card is less painful than paying by cash or check (Soman, 2001). Moreover, the interest rates charged on credit cards are sticky and do not change with monetary policy, which complicates the implications of monetary policy through the credit card channel (Calem and Mester, 1995). If credit card interest rates were elastic in response to changes in the policy rate, monetary policy would have a multiplier effect on the consumption level through the availability of credit card funds. Further, Yilmazkuday (2011) argues that a contractionary monetary policy forces commercial banks to restrict lending through credit cards, and the credit (or lending) channel of monetary policy transmission would therefore be more effective in the presence of credit cards, compared to other channels of monetary transmission, such as the interest and exchange rates. Moreover, external risk factors, such as fluctuations in oil price and exchange rates, further complicate the mechanism of monetary transmission. Hence, it would be interesting to analyze the role of credit cards in monetary policy transmission, along with other global risk factors, such as exchange rates and oil price shocks.

Given this background, this paper examines the following questions in the context of Indonesia: 1) Does credit card usage play any role in the transmission of monetary policy? 2) Is the prevalence of the lending channel in the transmission of monetary policy due to credit card usage, as compared to other channels? 3) Does credit card play a consumption-smoothing role in monetary policy dynamics? Since studies related to the impact of credit cards on monetary transmission are relatively scarce, our paper marks an essential contribution to the literature in the following ways: 1) It is the first study that examines monetary policy transmission by taking into account global risk factors such as oil price shocks in the presence of credit card usage; 2) it is one of the first attempts to examine the empirical relation between credit card usage and monetary policy transmission, using a structural vector autoregressive (SVAR) approach; and 3) it is the first attempt to understand the dynamics of monetary policy in the presence of credit card usage in the context of Indonesia.

142	Bulletin of Monetary Economics and Banking, Volume 22, Number 2, 2019

In this paper, we hypothesize the following: 1) Credit card usage reduces the effectiveness of monetary policy transmission because it raises price levels due to the high velocity of money (Friedman, 1999, 2000), and 2) credit card usage is positively affected by income levels, since higher income or output encourages consumers to spend more on consumption with credit cards and hence supports the consumption-smoothing role of credit cards Yazgan and Yilmazkuday (2011) Fulford and Schuh, (2017).2 Similarly, credit card usage is negatively affected by interest rates, since they represent the opportunity cost of credit card usage, especially when credit card debt is not paid on time Yilmazkuday (2011). In addition, exchange rates and global oil prices affect domestic price levels through international trade, as we assume Indonesia to be a small open economy (Basnet and Upadhyaya, 2015). Finally, the central bank alters the policy rate to curb inflation, since BI has officially adopted an inflation-targeting framework.

Our approach to testing the above interlinkage is as follows. We use monthly data from 2006 to 2018 and an SVAR model. We use impulse response function analysis and forecast error variance to analyze monetary transmission. Further, the analysis includes commercial bank lending, to account for the lending channel of monetary policy. Accordingly, monetary policy actions affect the lending capacity of commercial banks and thus affect monetary policy targets. Similarly, we also incorporate exchange rates and global oil prices into the analysis to account for external shocks. Our empirical findings suggest the following. First, credit card usage is significantly explained by output, indicating the consumption-smoothing role of credit cards. Second, credit card usage is not affected by the interest rates, indicating the stickiness of interest rate charged on credit card. Fourth, variations in policy rates are significantly determined by variations in inflation, supporting the inflation-targeting objective of BI’s monetary policy. Fifth, global oil prices play a significant role in explaining domestic inflation, indicating that external shocks pass through to domestic inflation. Finally, the impact of policy rates on inflation through the lending channel is not strong; however, the role of exchange rates in the transmission process is more prevalent, because exchange rate variations affect the liquidity conditions of commercial banks.

III. ECONOMETRIC FRAMEWORK

This paper employs an SVAR model to analyze the dynamics of credit card usage and monetary policy.3 The SVAR model is an alternative to the simultaneous equation models originally proposed by Sims (1980). A standard SVAR model can be written as

(1)

2The level of income is an important indicator of consumers’ repayment patterns. Households with highly liquid assets or income are more likely to use credit cards for transactions and pay their credit card debt on time (Canner and Cyrnak, 1985; Zhang and DeVaney, 1999).

3Most of the studies that analyze the monetary policy transmission mechanism use SVAR models, due to their dynamic nature, compared to other econometric techniques. For a survey on the use of SVAR models of the monetary transmission mechanism, see Christiano et al. (1999).

144	Bulletin of Monetary Economics and Banking, Volume 22, Number 2, 2019

In model 2, we expand the above model by including information related to commercial banks’ lending and exchange rates.5 The inclusion of these variables will help us to identify the interaction of the credit card effect on monetary policy transmission through lending and the exchange rate channel.

The identification strategies for model 2 are as follows:

Xt = (Output, Inflation, Credit, Lending, Interest rate, Exchange rate)

(4)

where we assume commercial bank lending does not respond contemporaneously to the interest rate, since they do not change lending rates quickly in response to monetary policy changes, and lending thus changes with a lag. However, exchange rate is contemporaneously endogenous to all the other variables in the system, since they respond quickly to changes in real variables as well as monetary variables in the system.

Finally, in model 3, we incorporate global oil prices to account for global supply shocks, since increases in oil prices can increase the cost of production and thus lead to higher price levels (Narayan et al., 2014; Besnet and Upadhyaya, 2015). Hence, the central bank is assumed to react to global oil price movements by changing the policy rate. In this case, we assume oil prices are contemporaneously exogenous to all the other factors in the system. The identification strategies for model 3 are as follows:

Xt = (Oil, Output, Inflation, Credit, Interest rate, Exchange rate)6

(5)

5The incorporation of inflation and exchange rate also helps address well-known price and exchange rate puzzles whereby inflation increases and the exchange rate depreciates, given a contractionary monetary policy (Sims, 1992; Cushman and Zha, 1997). A similar approach is adopted by Juhro and

Iyke (2019) to address the effect of monetary transmission on financial conditions in Indonesia.

6To maintain degrees of freedom, we do not include commercial bank lending in model 3.

146	Bulletin of Monetary Economics and Banking, Volume 22, Number 2, 2019

V. EMPIRICAL RESULTS

As a first step, we examined the stationarity properties of the variables. Since standard unit root tests, such as the augmented Dickey–Fuller test and Phillip– Peron tests, do not take into account possible structural breaks in the data series, this study employs the unit root test proposed by Narayan and Popp (2010; hereafter NP) that accounts for two endogenous structural breaks. The main highlight of this test, compared to other structural break tests, such as those of Lumsdaine and Papell (1997) and Lee and Strazicich (2003), is that it uses a Dickey–Fuller-type test approach and the break date is determined by maximizing the significance of the break date coefficient. The NP test has good size and stable power and identifies the structural breaks accurately in finite samples. The NP test suggests two models, M1 and M2, that permit two breaks in levels and two breaks in level and trend, respectively. The test statistics reported in Table 2 shows that the null of the unit root can be rejected in all cases except the exchange rate, as for model M1 (intercept). Moreover, the break dates are found to be in various months during 2009–2011 for most of the variables. For instance, credit card growth experienced a break in the first month of 2010 and in the sixth month of 2011. Our empirical findings for the structural break test are in line with those of Sharma et al. (2018), who find that most of the macroeconomic data of Indonesia suffer from structural breaks.

Table 2.

Structural Break Unit Root Test

This table shows the Narayan and Popp (2010) unit root test results for monthly data. We refer to Table 3 of Narayan and Popp (2010) for critical values for unknown break dates. Models 1 and 2 are two models for testing unit root. Model 1 (see Column 2, denoted M1) allows for two breaks in intercept and the Model 2 allows for two breaks in intercept as well as trend (see Column 2, denoted M2). The true break dates are denoted by TB1 and TB2; k represents the optimal lag length; and ***, **, and * indicate that the unit root null hypothesis is rejected at the 1%, 5%, and 10% levels of significance, respectively. In in the break date, for example 2011M8 denotes month. Similarly, output, credit, lending and oil denote growth rate of the index of industrial production, credit card transactions, commercial banks’ lending, and oil prices, respectively.

	M1: Two Breaks in Intercept					M2: Two Breaks in Intercept and Trend
	Variables	k	t-stat	TB1	TB2	Lag	t-stat	TB1	TB2
	Output	0	-0.579	2011M8	2012M1	0	-0.537	2011M8	2012M2
			(-6.704)*				(-6.156)*
	Inflation	4	-0.122	2013M6	2014M12	4	-0.120	2013M6	2014M12
			(-4.202)**				(-4.290)
	Credit	2	-0.400	2010M1	2011M6	3	-0.487	2010M1	2010M6
			(-4.210)**				(-4.523)***
	Lending	4	-0.083	2009M8	2011M11	2	-0.030	2011M11	2010M6
			(-3.859)***				(-1.153)
	Interest Rate	4	-0.783	2009M9	2009M11	4	-0.810	2009M9	2009M11
			(-6.741)*				(-6.821)*
	Exchange Rate	4	-0.072	2013M10	2015M9	4	-0.132	2013M10	2015M9
			(-1.622)				(-2.805)
	Oil	4	-0.780	2009M9	2010M9	4	-0.810	2009M9	2010M9
			(-6.741)*				(-6.821)*
	Critical Values for Unit root test				1%		5%		10%
	Model M1 (Break in Intercept only)				-4.731		-4.136		- 3.825
	Model M2 (Break in Intercept and				-5.318		-4.741		-4.430
	Trend)

Monetary Policy Transmission and Credit Cards: Evidence from Indonesia	147

To perform the SVAR test, we convert the exchange rate into first differences to ensure stationary variable levels. The optional lag length for vector autoregression is then determined through the Schwarz–Bayesian and likelihood ratio criteria, which, respectively, find three to be the optimal number of lags for model 1 and two for models 2 and 3. The models fulfill the criteria for diagnostic tests of, for example, autocorrelation, normality, and heteroskedasticity. Apart from that, the intercept dummy variables are included, are considered exogenous in the SVAR system, and correspond to the break dates suggested by the NP test.

A. Estimation of Model 1

The impulse response function for the SVAR system is depicted in Figure 4. The dashed lines correspond to plus or minus two standard errors around the impulse responses. It is evident from Figure 4 that the interest rate significantly and positively responds to a positive shock on inflation, and the response is statistically significant for up to four months. This finding is not surprising, since BI adheres to an inflation-targeting policy and thus changes the policy rate when inflation deviates from its target, a monetary policy strategy popularly known as the Taylor rule. It is also important to note that in interest rate does not respond to output and credit card transactions, which further underlines BI’s preference for inflation over output. However, the response of inflation to changes in the interest rate is found to be negative, but the effect is not statistically significant, signaling weak monetary policy transmission through the lending channel.

It is also interesting to note that inflation responds positively to increase in credit card transactions. The rise in inflation can be attributed to the higher velocity of money with the use of credit cards. Similarly, the response of credit cards to output is positive and significant. The significant effect of output on credit card usage indicates that consumers use credit cards for transactions or purchases when their incomes increase. This finding underscores the role of credit cards in consumption smoothing among high-income consumers. It is also evident that the policy rate does not have any significant effect on credit card usage. This finding could be attributed to the sticky interest rates charged on credit cards. This inelasticity of credit card usage to the policy rate could weaken monetary transmission through the credit card channel.

Table 3 reports the variance decomposition results. It reveals that variations in output are largely driven by variation in the output itself (94% in the 10th month). Similarly, variations in inflation are largely explained by inflation’s own variations, that is, around 84% in the 10th month, whereas the contributions of output and the interest rate to inflation are found to be 0.8% and 9.6%, respectively. Importantly, output explains around 38% of the variation in credit card transactions in the 10th month, further shedding light on the role of the rapid economic growth in Indonesia over the last decade on credit card usage. Similarly, around 15% of the variation in interest rates is explained by inflation.

The key findings of the analysis can be summarized as follows:

1.Inflation significantly explains the variations in interest rate, reflecting the central bank’s monetary policy response to price stability.

2.Credit card transactions marginally explain variations in inflation.

Monetary Policy Transmission and Credit Cards: Evidence from Indonesia	149

		Figure 4. Impulse Response Function (Model 1) (Continued)
6			Response of Output to Credit							6			Response of Output to Inflation
4										4
2										2
0										0
-2										-2
-4										-4
-6	2	3	4	5	6	7	8	9	10	-6	2	3	4	5	6	7	8	9	10
1										1
.8		Response of Inflation to Interest Rate								.8			Response of Inflation to Output
.6										.6
.4										.4
.2										.2
.0										.0
-.2										-.2
1	2	3	4	5	6	7	8	9	10	1	2	3	4	5	6	7	8	9	10
.8			Response of Inflation to Credit							.8			Response of Inflation to Inflation
.6										.6
.4										.4
.2										.2
.0										.0
-.2										-.2
1	2	3	4	5	6	7	8	9	10	1	2	3	4	5	6	7	8	9	10
8		Response of Credit to Interest Rate								8			Response of Credit to Output
4										4
0										0
-4										-4
-8										-8
1	2	3	4	5	6	7	8	9	10	1	2	3	4	5	6	7	8	9	10
8			Response of Credit to Credit							8			Response of Credit to Inflation
4										4
0										0
-4										-4
-8										-8
1	2	3	4	5	6	7	8	9	10	1	2	3	4	5	6	7	8	9	10

154	Bulletin of Monetary Economics and Banking, Volume 22, Number 2, 2019

									Figure 5.
				Impulse Response Function (Model 2) (Continued)
.2		Response of Interest Rate to Exchange Rate							400			Response of Exchange Rate to Ouput
									300
.1									200
									100
.0									0
									-100
-.1									-200	1	2	3	4	5	6	7	8	9	10
1	2	3	4	5	6	7	8	9	10
400		Response of Exchange Rate to Inflation							400			Response of Exchange Rate to Credit
300									300
200									200
100									100
0									0
-100									-100
-200									-200
1	2	3	4	5	6	7	8	9	10	1	2	3	4	5	6	7	8	9	10
400		Response of Exchange Rate to Interest Rate							400			Response of Exchange Rate to Lending
300									300
200									200
100									100
0									0
-100									-100
-200									-200	1	2	3	4	5	6	7	8	9	10
1	2	3	4	5	6	7	8	9	10