An analysis of the transmission mechanisms of
monetary policy in South Korea
University of Groningen
Name: Yu Shen
Student Number: 1349082
To my family members in Jiaozuo and Groningen
Abstract...3
I Introduction ...4
II Monetary and exchange rate policies in Korea ...5
II-1 Monetary policy...5
II-2 Exchange rate policy ...7
III Literature review...7
IV Methodology...11
V Data, the choice of variables, and stability test ...13
VI Results...19
VI-1 Pre-crisis...20
VI-2 Post-crisis ...21
VI-3 Monetary policy shocks as the source of fluctuations ...23
Abstract
In the paper, a simple VAR model is employed to examine the monetary transmission mechanisms in Korea. The Likelihood-Ratio test results suggest separating analysis into two subsamples. The model cannot solve the exchange rate and price puzzles in the pre-crisis time, but it shows that all results are generally in line with theory in the post-crisis time. This paper also makes some other specifications to compare with the basic VAR model.
I Introduction
South Korea has been regarded as a country with rapid economic growth since its introduction of the First Five-year Development Plan in the early 1960s. South Korea, together with Taiwan, Hong Kong and Singapore, challenges the dominance of Japan in Asian economy. Besides, South Korea, like Thailand and Indonesia, was severely affected by the Asian financial crisis in October 1997. At that moment, the Korean currency (Won) depreciated dramatically against the U.S dollar. Foreign reserves were almost drained. As a consequence, there was a significant economic breakdown in South Korea in the early 1998. The main negative effects of this crisis on South Korea are embodied in the banking sector and stock markets. Only after the IMF-supported program was implemented, the Korean economy recovered.
The fragile Korean financial system during the Asian financial crisis needs a deep inspection of monetary policy, since there is a close connection between them. Besides, monetary policy can also have effects on the whole economy. The Asian financial crisis raises a number of questions with respect to monetary policy. For instance, dose monetary policy really affect output and inflation? If so, how important it has been in the Korea economy? Was the Asian financial crisis associated with a swinging of the transmission of monetary policy? If so, which policy instrument is more appropriate in the current Korean economy? The main task of this paper is to systematically analyze the monetary transmission mechanisms in South Korea, and to test whether there is a structural change of the system. After all, it is recognized that the monetary policy regime changed from monetary targeting to inflation targeting in Korea in 1998. All the answers to the questions will be based on an empirical econometric model of the transmission mechanisms and timing to which the Bank of Korea’s monetary policy impulses affect the Korea economy. The model applied in this paper is a simple structural VAR model.
There are few doubts that forecasting of macroeconomic variables is important for the monetary policy decisions of any central bank. However, it is generally accepted that central banks cannot have perfect control on their targeted variables. Nevertheless, monetarists generally agree that in the short run, a change of monetary policy induced by the central bank triggers a number of actions and mechanisms by economic agents, eventually having effects on developments in economic variables such as output and prices. A better understanding of the monetary transmission mechanism is essential for the Bank of Korea, since it helps determine the correct trajectory of monetary policy to achieve its monetary objectives like inflation targeting and so on.
The structure of this paper is constructed as follows: In order to have a good knowledge on monetary exchange rate policies in South Korea, the brief background of monetary and exchange rate policies is given in section II. After gaining the basic knowledge on the Korea economy, section III provides theoretical explanations on how monetary policy shocks affect the economy. In other words, section III gives details on how monetary transmission mechanisms work. Then, Section IV brings in an econometric model for the practical analysis of monetary policy shocks. Section V describes data and additionally tests if the economic structure is consistent over sample periods. In other words, we will conduct the stability test for the whole sample. Based on the stability test results from section V, Section VI examines the effects of monetary policy shocks in Korea. Besides, this section also addresses shortcomings of the basic VAR model, offers explanations and extends experiments. Finally, section VII ends the paper with conclusions.
II Monetary and exchange rate policies in Korea
II-1 Monetary policy
economic situation. From 1979 to 1998, the two announced objectives of monetary policy were to maintain a stable position of the Won and to ensure strength of the financial system. Although the objectives of monetary policy vary over time, price stability is in general addressed as the most important objective of monetary policy in Korea. This derives from the belief that sustainable economic growth, which is the ultimate objective of monetary policy, can only be realized if price stability is maintained.
At the beginning of the 1980s, M21 was regarded as the leading monetary indicator in
South Korea and the monetary policy framework was based on a monetary targeting regime. The modified Fisher equation is employed to estimate the target range of money. In other words, the Bank of Korea took into account of economic growth, inflation rates, and the money velocity to determine its money supply.
The Bank of Korea maintained the monetary targeting regime until the mid-1990s when the monetary aggregates could no more predict inflation movements. The leading monetary indicator, M2, became unreliable due to reconstruction of the trust account system in 1996. Under this circumstance, the Bank of Korea started to emphasize the importance of MCT, which consists of M2, certificates of deposit (CDs) and monetary-in-trust. In 1997, the Bank of Korea adopted a double monetary targeting system, setting M2 and MCT as the target indicators.
The roles of the target indicators deteriorated during the time of the Asian financial crisis. At that time, the Bank of Korea took a serious consideration on changing its monetary policy. Inflation targeting became a potential substitute for monetary targeting. One reason why inflation targeting was considered is because it has been successful in many industrial economies such as New Zealand, Austria, etc, and it was then a safer choice for the crisis-hit Korean economy. The other but important reason
6
1
was that the IMF-supported program suggests the Korean authority to adopt inflation targeting. The revision of the Bank of Korea Act that came into force in April 1998 then adopts inflation targeting as the primary objective of monetary policy. The Act specifies that the central bank should determine an annual inflation target and be devoted to achieve it. The Bank of Korea’s operating variable is the overnight call rate, the target for which can be maneuvered on a monthly basis in response to movements in expected inflation and other economic variables (Bank of Korea, 2003)
II-2 Exchange rate policy
Korea applied a managed floating exchange rate regime for approximately two decades. In 1980 a peg of a basket of currencies replaced a single currency peg to the US dollar. In other words, a controlled, floating effective rate was established. The effective rate was a combination of currencies of Korea’s main trading partners, such as the U.S., Canada and Japan. In 1990, the market average rate took the position of the effective Rate. Therefore, Korea had a managed floating exchange rate regime and the exchange rate was determined by market forces in the inter-bank market. However, there was actually no much room for the daily exchange rate of Won-US dollar to fluctuate in the spot market, because the exchange rate was restricted to a narrow band. When the Asian financial crisis spread out into Korea, the Korean government broadened exchange rate bands in order to defend the position of Won in November 1997, but failed. As a consequence, reserves in the Bank of Korea were almost wiped out and the Korea Won was allowed to float freely in December, 1997. (Ponomareva, 2006)
III Literature review
This section aims to present views of the transmission mechanisms through which monetary policy affects the economy. The monetary transmission mechanism,
explained by Taylor (1995), refers to “the process through which monetary policy
decisions are transmitted into changes in real GDP and inflation”. According to the
monetary literature, transmissions of monetary policy often take place via a number of transmission channels. They include the interest rate channel, other asset prices channel, the credit channel, and the exchange rate channel.
Interest rate channel: It is traditionally agreed that the interest rate channel is the
major channel of monetary transmission mechanisms. This idea is based on the conviction that monetary policy affects both nominal and real interest rates, which in turn have an impact on aggregate consumption and investments, aggregate demand, output and prices (Mishkin, 1996). The relationship between nominal and real interest rate is guided by the assumptions of price rigidities and rational expectations. A rise in nominal interest rate, under the assumptions of price rigidities and rational
expectations, causes the rise of the real interest rate too, at least for a short period2. It
is assumed that business investments are affected by the cost of capital that positively correlates with real interest rates. Consumption is affected because movements in real interest rates affect the cost of future consumption. For this channel, a contractionary monetary policy causes the rise of the nominal interest rate and consequently the real interest rate. The higher real interest rate discourages business investments. Meanwhile, consumers that notice the higher real interest rate, start to increase their savings. Therefore, a contractionary policy shock eventually produces the decline in aggregate output.
Other asset prices channel: The interest rate can only be considered as one of asset
prices. Apart from interest rates, the other asset prices, namely the prices of equity, bonds, real estate, can lead to the identification of alternative mechanism channel- other asset prices channel. Tobin’s q theory offers a mechanism through which
8
2
monetary policy affects the economy, that is, it is through its effects on the value of equities. According to Tobin (1969), q is defined as the ratio of the market value of a firm to the replacement cost of capital. When q is high, the market value of firms is high relative to the market value of the average industry level. Taking the advantage of the inequality, firms having higher market value are willing to issue new equities and offer them in a higher price, so that they are able to purchase relatively cheaper plants or equipments. An expansionary monetary policy results in higher equity prices, which, through Tobin’s q, make investments in firms more attractive. The investment spending hence increases. On the other hand, according to Ando and Moligliani (1963), higher equity prices also raise financial wealth of consumers and consequently their consumption. Together, the increase of investments and consumption brings about higher aggregate demand.
Credit channel: The credit channel is a relatively novel view in monetary transmission
theory. The basic idea is that monetary policy can have effects on prices and output via the credit rationing, which stems from information imperfection in credit markets. Discussions of the credit channel often distinguish between the banking lending channel and the balance sheet channel.
The bank lending channel stresses the special nature of bank credit and the role of banks in the financial system. This channel is based on the view that monetary policy may affect the external finance premium by shifting the supply of bank credit. Banks supply loans to small and medium-sized firms. Banks and small and medium-sized firms have established systems that are well-matched to cope with information imperfection problems, which are prominent for these particular groups of borrowers. If the supply of bank loans is disrupted, bank dependent borrowers may not be excluded from credit. However, they may incur costs associated with finding a new lender. This is likely to increase the external finance premium and to reduce real activity (Bernanke and Blinder, 1988).
The balance sheet channel is based on the theoretical prediction that the external finance premium encountered by a borrower depends on borrower’s financial position. That is, a borrower who has sound financial position and hence a high quality balance sheet can face a lower external finance premium. The lower external premium therefore will affect the borrower’s overall investment and consumption decisions. The balance sheet channel exists because shifts in the monetary policy affect not only market interest rates, but also indirectly the financial position of the borrower (Walsh, 1998).
Exchange rate channel: This channel examines the relationship between net private
capital inflows and monetary policy, especially after financial liberalization is implemented. Under the flexible exchange rate regime, when domestic real interest rates rise, domestic currency deposits and other forms of financial assets appear more attractive than the ones in foreign countries. This results in domestic currency appreciation. Yet, appreciation of the domestic currency makes domestic goods more expensive than foreign goods. As a consequence, net exports and aggregate output start to fall. Hence, the monetary authority, under the floating exchange rate regime, is capable of implementing its monetary control. Usually, the monetary authority exchanges volatility of the exchange rate for minimal fluctuations of real GDP and price stability. However, under a fixed exchange rate regime, a central bank is constrained to conduct discretionary monetary policy, because its each attempt to supply money will be offset by losses of foreign exchange reserves (Obstfeld and Rogoff, 1995).
IV Methodology
In this paper a VAR model is employed to examine the monetary transmission mechanisms in Korea. Although there is no consensus about how the monetary transmission mechanisms look like in Korea, this methodology nevertheless places minimal restrictions on how monetary policy shocks affect the economy. Besides, the VAR model recognizes explicitly the simultaneity between monetary policy and macroeconomic developments. The choice of a VAR approach is also motivated by the existence of literature using VARs to examine monetary transmission mechanisms.
This section briefly describes the basic VAR model used to analyze the effects of monetary policy in the Korea economy. We assume that the Korean economy is described by a structural form VAR as:
11 t + 1 ( ) t t t Ay =B L y− +Dx e (1)
Where y is a vector of endogenous variables, t xt is a vector of exogenous variables,
and denotes a structural disturbances vector. Here, is assumed to be serially
uncorrelated and , where
t
e et
var( )et = Λ Λis a diagonal matrix. The diagonal elements
are the variances of structural disturbances. It then confirms that structural
disturbances are assumed to be mutually uncorrelated.B L( )is a p-th degree matrix
polynomial in the lag operator L; that is B L( )=B0+B L1 +B L2 2+ ⋅⋅⋅+B Lp P, where p
is the number of lagged periods used in the model. With n standing for the number of
endogenous variables, the squared n n× matrix A contains the structural parameters
of the contemporaneous endogenous variables. The squared n n× matrix D is the
coefficient matrix of the exogenous variables.
12 1 t − t 1 1 1 ( ) t t t y =A B L y− − +A Dx− +A e (2)
Equation (2) can also be rewritten as
1 ( ) t t t y =C L y− +Ex +u u = A e− ut (3)
Where , and . is the reduced form residual
vector and 1 ( ) ( ) C L =A B L− 1 E=A D− 1 t t var( )ut = Σ .
It is apparent now that the relationship between the structural disturbances and the reduced form residuals is expressed as
t e = Aut 1 A− A− Σ = Λ and 1 ' (4) ( ) ( )
providing restrictions on contemporaneous structural parameters (see Blanchard and Watson, 1986). In this paper, the recursive VAR model is employed and the order of the endogenous variables is carefully made based on relevant assumptions and theory.
V Data, the choice of variables, and stability test
To estimate the model, the variables need to be specified. In our basic VAR model, the endogenous variables include IP, CPI, CR and ER, where IP is industrial production, representing real GDP, CPI is the consumer price index, CR is a measure of the short-term interest rate, e.g. the overnight call rate in Korea, and ER is the exchange rate expressed as units of the Korean won for one unit of the US dollar. To capture the close connection between the Korean economy and the world economy, two exogenous variables are included. Two exogenous variables are the world oil price (OP), which is expressed as units of the US dollar per barrel, and the federal funds rate (FFR).
The ordering of endogenous variables in the basic VAR model is as follows: IP, CPI, CR and ER. This order is based on some assumptions. In this VAR specification, industrial production is contemporaneously not affected by monetary policy, while the growth of consumer price index (i.e. inflation) contemporaneously depends only on current production. One motivation for these assumptions is that firms do not reset their production and prices in reaction to unexpected changes of financial signals and monetary policy because of “inertia, adjustment costs and planning delays” (Kim, 2000). In this specification, the exchange rate represents financial signals and the change of the overnight call rate represents monetary policy. The overnight call rate equation in the VAR system is a simple version of the Taylor rule (Taylor, 1993). The central bank models and sets the interest rate based on changes in prices and output according to the Taylor rule. As recognized, the overnight call rate is treated as the
policy instrument, that is, the Central Bank of Korea changes the interest rate in response to movements in output and the price level. The exchange rate equation portrays the financial market equilibrium. We assume that all variables contemporaneously affect the exchange rate because it is a forward-looking asset price.
Clearly, this VAR specification describes an economy in a rather single way but nevertheless it embraces the minimal set of variables that are necessary for monetary policy analysis. The first three endogenous variables are commonly used in the monetary business cycle literature. Since Korea under the analysis is a small open economy, it is necessary to include the exchange rate in this specification. After all, the exchange rate channel is one of the key channels of monetary transmission in open economies. The inclusion of the oil price and the foreign variable, namely the federal funds rate, is to isolate exogenous monetary policy changes. Also, the introduction of the oil price and the federal funds rate in the specification is to capture the close link between the world economy and the Korean economy.
Economists like Kim and Roubini (2000) treat the world oil price and foreign variables, such as the federal funds rate as the other endogenous variables. In their specification, although both variables are assumed not to be contemporaneously affected by the domestic monetary policy, they are affected by Korean monetary policy in subsequent periods. In our opinion, Korea, regarded as a small open economy, could not affect the world economy via its monetary policy, neither contemporaneously nor consequently. Therefore, it is more appropriate to label them as exogenous variables. In addition, since there are some contemporaneous restrictions imposed in the basic VAR model, monthly data are more suitable than quarterly data. Data in this study are all obtained from the International Financial Statistics (IFS) issued by the International monetary Funds and all variables are in level forms. The range of time series is from 1980:01 till 2005:12, consisting of 312 observations.
The four domestic time series are plotted in Figure 1. By looking at the figure, we cast a doubt on the stationarity of the variables. Following the argument made by Bagliano
and Favero (1998), we use all variables in levels and with three time lags3, regardless
of the possible cointegration relationships in the system. Bagliano and Favero state that in doing so, the tougher problem of long-run identification can be avoided. In general, this specification does not lose the information on the long run properties and merits of consistency of estimators.
Before we proceed to analyze monetary policy identification issues, we should be aware of the famous Lucas critique (Lucas, 1976). When this critique is applied to monetary policy, it signifies that the parameters of monetary policy models rely on the monetary policy regime. Therefore, a model estimated under the specific monetary
Figure 1 Endogenous variables used in the VAR model (1980:01-2005:12)
0 20 40 60 80 100 120 140 160 80 82 84 86 88 90 92 94 96 98 00 02 04 Industrial production 20 40 60 80 100 120 80 82 84 86 88 90 92 94 96 98 00 02 04
Consumer price index
0 4 8 12 16 20 24 28 80 82 84 86 88 90 92 94 96 98 00 02 04
Overnight call rate
400 600 800 1000 1200 1400 1600 1800 80 82 84 86 88 90 92 94 96 98 00 02 04
Exchange rate (per U.S dollar)
Source: International Financial Statistics (IMF)
policy regime could not be applied to investigate the effects in a different monetary
15
3
policy regime. In the case of Korea, its economy experienced the change of the policy regime from monetary targeting to inflation targeting in 1998, so we have the reason to doubt parameter stability in our basic model.
Hereafter, we follow the methods employed by Bagliano and Favero (1998) to test whether there are structural changes in the Korea economy. The starting point of this test is the argument that in the empirical analysis, the residuals of reduced VAR models must be homoskedastic and non-autocorrelated. Only when residuals are stationary and non-autocorrelated, the reduced VAR models are valid. All the
Figure 2 Residuals from the reduced-form VAR (Whole sample)
-20 -16 -12 -8 -4 0 4 8 12 16 80 82 84 86 88 90 92 94 96 98 00 02 04 IP -3 -2 -1 0 1 2 80 82 84 86 88 90 92 94 96 98 00 02 04 CPI -4 -2 0 2 4 6 8 80 82 84 86 88 90 92 94 96 98 00 02 04 CR -200 -100 0 100 200 300 400 80 82 84 86 88 90 92 94 96 98 00 02 04 ER
residuals from estimation of the system with four endogenous variables over the whole sample are depicted in Figure 2. Residuals from all equations exceeding the ± 2 standard error bands show that features of normality and homoskedasticity are
dubious. In particular, larger residuals from the exchange rate and overnight call rate equations are observed during the period of the Asian financial crisis.
However, a visual recognition from graphs needs the backup of more formal tests. A warning from Rudebusch (1996) says that VAR systems, covering a large range of sample size, normally display structural instability in at least some equations. Therefore, it is necessary to form a stability test for the sample period of the Korea economy. The most frequently employed stability test is the Chow test. However, as recognized by Andrews (1993), the information provided by the Chow test can be valuable only if the structural break is one-off and the break date is well-documented. In reality, the existence of known break dates is questionable. Especially, in some cases, the structural change may be even more complex than a one-off change. For example, in the case of Korea, its economy had used monetary aggregates as the policy instrument for decades until the time the Asian financial crisis stroke in October, 1997. From that time on, monetary aggregates could no longer correctly serve as the target indicators. Until April, 1998, the revised Act of the Bank of Korea came into effect. In provisions of that Act, it states that the overnight call rate replaces monetary aggregate to serve as the policy instrument. From then on, the Bank of Korea began to control the call interest rate to achieve price stability. Therefore, there is a transitional period in between the two different monetary regimes.
For the stability test, we modify the method provided by Andrews (1993), introduce an uncertainty of one year around the point estimate of the break dates, and compute the likelihood-ratio (LR) statistic of structural stability for each break date. The largest statistic suggests a stability test for an unknown break point. For each equation of the reduced-form VAR, we test jointly for the stability of the equation using the Wald version of the Quandt (1960) LR test (that is, the Andrews sup-Wald test) for the test of each equation. We use a heteroskedasticity consistency covariance matrix, in which the residuals are calculated under the null rather than each of the alternative hypotheses for the sake of convenience.
We apply the LR test to each equation of the VAR system. Given 15 regressors in each equation, if the break point is known, the critical values of test statistics are
around 34 and 39 at 5% and 1% significance level, respectively.4 The examined
period is selected from the time when Thailand was first hit by the financial crisis till one year later. As we can read from Table 1, there is strong evidence of instability in 1997-1998. Except for few beginning dates in the overnight call rate equation, the rest
Table 1 Stability test on the Reduced-form VAR
Dependent Variables Industrial production Consumer price index
Call rate Exchange rate
1997:06 52.90 51.39 29.57 52.87 1997:07 52.89 51.91 35.60 65.38 1997:08 52.86 53.72 35.69 81.64 1997:09 52.60 52.58 33.52 101.10 1997:10 52.57 53.00 32.81 131.38 1997:11 52.70 57.21 37.13 197.89 1997:12 52.92 64.98 47.44 244.21 1998:01 53.53 74.09 61.19 228.61 1998:02 53.93 56.79 41.97 67.01 1998:03 55.28 56.82 43.69 72.07 1998:04 54.33 56.79 35.18 73.47 1998:05 54.54 56.05 36.24 62.81
Source: Author’s calculations
Notes: the figures are the LR statistic with χ2distribution introduced by Andrews (1993). Under the null of the test, the coefficients of each equation are time invariant.
observed statistics are all significant, ranging from a minimum of 52.70 for the
18
4
industrial production equation, to a maximum of 244.21 for the exchange rate equation. In sum, during the Asian financial crisis, the Korea economy in a whole experienced a structural change. Thus, we interpret these findings as strong evidence of structural changes in the VAR model.
In principle, one could estimate the break dates for different subsamples as a by-product of the LR test. However, for our case, the estimated break dates could not supply a consistent picture of the timing of the observed instability. In effect, Korea suffered from the Asian financial crisis for months. There was actually a turmoil period before the monetary policy regime was changed. Boivin and Giannoni (2002) suggest splitting the sample on the basis of anecdotal events and then to analyze monetary policy shocks in different subsamples. They compare subsamples based on the date on which Federal Reserve Chairman Paul Volcker announced a policy change. In this study, it is even more difficult to identify the break date based on anecdotal events, because the influence of the Asian financial crisis last for months in Korea. We here, adopting the idea from Fung (2002), ignore the most volatile period from 1997:10 to 1998:03. The subsamples are then called pre-crisis and post-crisis. It can be identified that Korea was hit by the Asian financial crisis in October, 1997, and in April, 1998, the revised Act of the Bank of Korea came into effect.
VI Results
Figure 3 and 4 show the responses to a contractionary monetary shock before the crisis and after the crisis, respectively. Each graph displays the response for a one-standard-deviation contractionary monetary shock over 3 years. The corresponding variable is named at the top of each graph. The upper and lower dashed lines plotted in each graph are one-standard error bands. In the following sub-sections, we analyze impulse responses of monetary policy before and after the crisis. The third sub-section invesitigates variance decomposition. By doing so, we are likely to
understand how important the policy shock accounts for movements of the economic variables, such as industrial production and the consumer price index. The fourth sub-section extends our experiments. The other specifications rather than this basic VAR model are introduced.
VI-1 Pre-crisis
Focusing on the subsample of pre-crisis, in response to a contractionary monetary shock, the overnight call rate initially rises significantly. The increase in the interest rate continues for nearly one and a half years. Industrial production exhibits a hump-shaped pattern in response to a contractionary policy shock. This pattern is statistically insignificant and lasts only for two months. After that, it starts to fall
Figure 3 Impulse responses to monetary policy shocks (Contractions, 3 years, pre-crisis)
-.6 -.4 -.2 .0 .2 .4 .6 5 10 15 20 25 30 35 IP -.3 -.2 -.1 .0 .1 .2 .3 5 10 15 20 25 30 35 C PI -0.4 0.0 0.4 0.8 1.2 5 10 15 20 25 30 35 CR -4 -2 0 2 4 6 8 5 10 15 20 25 30 35 ER
commonly observed in previous studies (see, for example, Sims, 1992). The Price
puzzle5 is seen in the upper-right Figure 3. A contractionary policy shock results in an
insignificant fall in the price level, which only lasts for first five months. In the following time, the price rises persistently above the base line for over two years and then fall below the baseline again. We now turns to the effects of the policy shock on the level of the exchange rate. Contrary to the study by Kim (2000), the exchange rate
puzzle6 is found in the pre-crisis Korea economy. The exchange rate persistently
depreciates with no trend to fall in response to a policy shock.
VI-2 Post-crisis
Figure 4 reports the results of impulse responses after the crisis. The responses of all variables to the monetary policy shock are in general in line with theory, but all responses are all statistically insignificant. The response of output shows no hump-shaped pattern during the post-crisis period. Output falls and reaches its trough in the seventh month. In addition, output approaches the base line after approximately 2 years. This pattern is in line with theory. Economic theory states that in response to a policy shock, output falls in the short run. In the long run, monetary policy shock does not have effects on the real variable. The interest rate initially rises and dies out hump-shaped pattern during the post-crisis period. Output falls and reaches its trough in the seventh month. In addition, output approaches the base line after approximately 2 years. This pattern is in line with theory. Economic theory states that in response to a policy shock, output falls in the short run. In the long run, monetary policy shock does not have effects on the real variable. The interest rate initially rises and dies out over time, which is the similar as the result from the pre-crisis period. Although the price level decreases in response to the policy shock, this phenomenon does not last for 5 months. In next 5 months, the price level jumps over the base line. Then, the
21
5
The price puzzle is defined as an increase in the price level under monetary contraction, in contrast to the most theoretical prediction that a monetary contraction results in a fall in the price level.
6
price level again falls below the baseline in the long run. Therefore, price puzzle is insignificant but still exists for a short period. The response of the exchange rate after
Figure 4 Impulse responses to monetary policy shocks (Contractions, 3 years, post-crisis)
-1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 5 10 15 20 25 30 35 IP -.6 -.4 -.2 .0 .2 .4 .6 5 10 15 20 25 30 35 C PI -.1 .0 .1 .2 .3 .4 5 10 15 20 25 30 35 CR -12 -8 -4 0 4 8 12 5 10 15 20 25 30 35 ER
the crisis is in line with the Dornbusch overshooting hypothesis (Dornbush, 1976) and Uncovered Interest Parity (UIP). In other words, a contractionary monetary policy shock causes the exchange rate to appreciate almost on impact. The maximum effect occurs immediately, and the exchange rate gradually depreciates.
In sum, before the financial crisis occurred, the responses of output and the interest rate are consistent with monetary theory. The price and exchange rate puzzles bring problems for the pre-crisis study. After the crisis, impulse responses improve, but the scales of responses are statistically insignificant. Except that the price level shows irregular pattern in the first ten months, the rest responses are all in line with theoretical predictions. Fung (2002) provides some reasonable explanations for puzzles in VAR studies. First, the existent price puzzle could result from problems of
missing variables, for instance, other commodity prices that capture the underlying inflation that the central bank reacts Second, the policy instrument employed in the analysis, in our case, the overnight call rate, may not fully reflect central bank reactions. Third, the identification restrictions in the model are not fit for disentangling monetary policy shocks from other shocks.
Apart from Fung’s explanation, we provide another and even more convincing explanation for why the exchange rate and price puzzles are found in the pre-crisis period. Before the crisis, the Korean authority used the managed floating exchange rate regime in attempts to protect its export industries, since they take a big share of GDP (Dooley, Dornbusch and Park, 2002). However, under this exchange rate regime, the Korea government’s attempt to reduce the interest rate would fail because its acquisitions of domestic bonds are offset by its losses of exchange reserves. This results in a constraint in conducting discretionary monetary policy. In addition, if the pass-through effect of the exchange rate on the consumer price is large, the depreciation of the exchange rate will be accompanied by the rise of the consumer price. In other words, the exchange rate puzzle most likely goes with the price puzzle if the pass-through effect of the exchange rate on the commodity price is strong. It can be observed from Figure 3 and Figure 4 that, in our case, the exchange rate and the consumer price move in a similar pattern, especially in the post-crisis period.
VI-3 Monetary policy shocks as the source of fluctuations
which says that monetary policy shocks merely explain the price and output fluctuations at any horizon. Moreover, policy shocks can only explain a small part of exchange rate fluctuations.
Table 2 Variance Decomposition due to Monetary policy shocks (Pre-crisis) 6 months 12 months 24 months 36months
IP 1.35 1.31 1.09 0.91
CPI 0.32 0.41 0.38 0.22
CR 92.52 89.49 74.60 62.98
ER 2.54 3.57 4.73 5.37
Source: Author’s calculation
Table 3 Variance Decomposition due to Monetary policy shocks (Post-crisis)
6 months 12 months 24 months 36 months
IP 0.94 1.26 1.10 0.97
CPI 0.53 0.31 0.17 0.15
CR 79.79 65.68 61.51 59.01
ER 2.04 1.93 1.69 1.51
Source: Author’s calculation
VI-4 Extensions
The unsatisfactory results of our study, especially the result of the analysis of pre-crisis might lead to the conclusion of our weak specification. After all, there are many empirical analyses in this study field that have solved a variety of puzzles. Therefore, in this sub-section, we extend our analysis by modifying the basic model presented above.
First, since the monetary aggregate is the main target indicator, we add M2 into the basis model and study the model with 5 endogenous variables and 2 exogenous variables. We still use the recursive structural model, but the ordering and the form of the endogenous variables change a bit. The ordering of the endogenous variables are IP, CPI, CR, M2, and ER, where the fourth equation is the traditional money demand function. All variables, endogenous and exogenous, are used in logarithmic forms except for interest rates. The money demand function in theory says that the demand for real money balances depend on the real output and the opportunity cost of holding money, i.e. the nominal interest rate. So, the money aggregate, price, real output and nominal interest rate are included in this equation. The results show that there is not much difference between the basic model and the specification with 5 endogenous variables in the pre-crisis period. That is, all the puzzles are unsolved. For the case of post-crisis, the result is also quite similar with the one in the basic model. Hence, incorporating the money aggregate into the basic model dose not help improve the results (See Appendix 1).
Second, as mentioned above, Kim and Roubini (2000) treat the world oil price and foreign variables, such as the federal funds rate as endogenous variables, and find no evidence of price, or exchange rate puzzles. Although we have argued the implausibility of this treatment, we cannot yet tell whose assumptions are most accurate. For that reason, we attempt to follow and to modify a bit the methods
employed by Kim and Roubini.7 The results from the pre-crisis and the post-crisis
periods are still similar with the ones from the basic model. Although Kim and Roubini solve the puzzles in their application for non-US G-7 countries, their model cannot be applied in the Korean economy (See appendix 3).
Third, as for choosing the break point, we followed the method employed by Fung (2002). In another experiment, we choose the single break point generated by the
25
7
statistical result. From Table 1, we identify that both the price and the interest rate equations reach their maximum in January, 1998. So, we select the 1998:01 as the break point date. It is found that before the crisis, the effects of a contractionary monetary policy shock on other non-policy variables become unrecognizable. For the case after the crisis, the price and exchange rate puzzles still exist.
In short, although our basic model in the case of Korea is quite simple, we still can notice that this model is sufficient enough to explain the monetary transmission mechanisms in Korea in comparisons with other specifications. Other extended experiments examined in this section cannot do better than the basic model.
VII Conclusion
There is a large amount of literature on VARs for developed countries. However, the literature on VARs for Korea is rare. In this paper, we analyzed the monetary transmission in Korea using the recursive VAR model. By using the VAR model, we can see how monetary policy has effects on the economy. To avoid ignorance of the possible structural change, we tested the structural stability of our model. The results from the LR test released that there was indeed some structural instability during the Asia financial crisis. Although we could determine the break date of structural changes using the byproduct of the LR test, it is recommended to use anecdotal evidence to choose break dates. In the case of Korea, we recognized that the Asian financial crisis last for several months, so following the method employed by Fung (2002). We chose two subsamples and took no consideration on the most economically severe period. The econometric results demonstrate that the effects of monetary policy shocks on variables are quite different between pre-crisis and post-crisis. In the pre-crisis period, the price and exchange rate puzzles are found. However, in the post-crisis period, the responses of economic variables are generally in line with theoretical predictions, but all the responses are statistically insignificant.
Further investigation into variance decomposition shows that monetary policy shocks are not important for explaining the movements of non-policy variables. Future studies on Korean monetary transmission might take into account of explanations of exchange rate and price puzzles given in the paper. Finally, in order to show the robustness of the basic VAR model, we also made some extended experiments. The results demonstrated that three other specifications were not as good as our basic VAR model. None of them can solve the problem of puzzles.
References:
Ando, A. and F. Modigliani, 1963, The “Life Cycle” Hypothesis of Saving: Aggregate Implications and Tests, American Economic Review, 53(1), 55-84.
Andrews, D.W., 1993. Tests for Parameter Instability and Structural Change with Unknown Change Point, Econometrica, 61(4): 821-856.
Bagliano, F.C. and C.A. Favero, 1998, Measuring Monetary Policy with VAR Models: An Evaluation, European economic review, 42(6): 1069-1112.
Bank of Korea, 2003, Monetary Policy in Korea.
Bernanke, B.S. and A. S. Blinder,1988, Credit, Money, and Aggregate Demand,
The American Economic Review, 78(2): 435-439.
Boivin, J. and Giannoni, M., 2002, Assessing Changes in the Monetary Transmission Mechanism: VAR Approach, Federal Reserve Bank of New York Economic Policy
Review, 8(1): 97-111.
Dooley, M., R. Dornbusch, and Y.C. Park, 2002, A Framework for Exchange Rate Policy in Korea, Finance Working Papers 125, East Asia Bureau of Economic Research.
Dornbusch, R., 1976, Expectations and Exchange Rate Dynamics, Journal of Political
Economy, 84(6): 1161-1176.
Fung, B.S.C., 2002, A VAR Analysis of the Effects of Monetary policy in East Asia,
BIS Working paper No 119.
Kim. S., 2000, Effects of Monetary Policy Shocks in a Small Open Economy: The Case of Korea, The Bank of Korea Economic Papers, 3(1): 91-108.
Kim, S. and N. Roubini, 2000, Exchange Rate Anomalies in the Industrial Countries: A Solution with a Structural VAR Approach, Journal of Monetary Economics, 45(3): 561-586
Leeper, E. M., C.A. Sims, and T. Zha, 1996, What Does Monetary Policy Do?,
Brookings papers on economic activity, 2:1-63.
Lucas, R.,1976, Econometric Policy Evaluation: A Critique. Carnegie-Rochester
Conference Series on Public Policy 1: 19–46.
Mishkin, S.F., 1996, The Channels of Monetary Transmission: Lessons for Monetary Policy, NBER Working Paper Series No. 5464.
Obstfeld, M., and K. Rogoff, 1995, Exchange Rate Dynamics Redux, Journal of
Political Economy, 103(3): 624-660.
Ponomareva, N., 2006, Forecastability of Inflation and Nominal Income Growth: The Case of Korea, Working paper, Report Number: ECMT2006-3, Sydney University, School of Economics and Political Science.
Quandt, R.E., 1960, Tests of the Hypothesis that a Linear Regression System Obeys Two Separate Regimes, Journal of the American Statistical Association 55: 324-330.
Rudebusch, G..D., 1998, Do Measures of Monetary Policy in a VAR Make Sense?,
International Economic Review, 39(4): 907-931.
Sims, C.A., 1980, Macroeconomics and Reality, Econoetrica 48(1): 1-48
Sims, C.A., 1992, Interpreting the Macroeconomic Time Series Facts: The Effects of Monetary Policy, European Economic Review, 36(5): 975-100
Taylor, J. B.,1993.,Discretion Versus Policy Rules in Practice, Carnegie-Rochester
ConferenceSeries on Public Policy, 39, pp.195–214.
Taylor, J. B., 1995, The Monetary Transmission Mechanism: An Empirical Framework, Journal of Economic Perspectives, 9(4): 11-26.
Tobin, J., 1969, A General Equilibrium Approach to Monetary Theory, Journal of
Money, Credit, and Banking 1(1): 15-29.
Walsh, C. E, 1998, Monetary Theory and Policy, the MIT Press.
Appendices
Appendix 1
Recursive VAR model with 5 endogenous variables
(1) Pre-crisis -.016 -.012 -.008 -.004 .000 .004 .008 .012 5 10 15 20 25 30 35 I P -.002 -.001 .000 .001 .002 .003 .004 .005 5 10 15 20 25 30 35 C PI -.002 .000 .002 .004 .006 .008 .010 .012 5 10 15 20 25 30 35 M2 -0.4 0.0 0.4 0.8 1.2 5 10 15 20 25 30 35 CR -.004 .000 .004 .008 .012 5 10 15 20 25 30 35 ER Pre-crisis (5 endogenous variables, recursive)
(2) Post-crisis -.012 -.008 -.004 .000 .004 .008 .012 .016 5 10 15 20 25 30 35 IP -.002 -.001 .000 .001 5 10 15 20 25 30 35 CPI -.006 -.004 -.002 .000 .002 5 10 15 20 25 30 35 M2 -.1 .0 .1 .2 .3 .4 5 10 15 20 25 30 35 CR -.010 -.005 .000 .005 .010 5 10 15 20 25 30 35 ER Post-crisis (5 endogenous varialbes, recursive)
Appendix 2
The setup of the modified Kim and Roubini’s model
32 ⎞ ⎟ ⎟ ⎠ 21 31 41 43 2 53 54 56 57 2 63 64 67 (/ $) 71 72 73 74 75 76 (/ $) 1 0 0 0 0 0 0 1 0 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 0 0 1 0 0 0 1 1 OP OP FFR FFR IP IP CPI CPI M M CR CR ER ER e u e g u e g u e g g u e g g g g u e g g g u e g g g g g g u ⎛ ⎞ ⎛ ⎞⎛ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ = ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎜ ⎟⎜ ⎜ ⎟ ⎝ ⎠⎜ ⎝ ⎠ ⎝ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟ ⎟⎟ Notes:
Appendix 3
Modified Kim and Roubini’s model (7 endogenous variables)
(1) Pre-crisis -.02 -.01 .00 .01 .02 .03 5 10 15 20 25 30 35 OP -.5 -.4 -.3 -.2 -.1 .0 .1 .2 5 10 15 20 25 30 35 FFR -.008 -.004 .000 .004 .008 .012 .016 5 10 15 20 25 30 35 I P -.003 -.002 -.001 .000 .001 .002 .003 .004 .005 5 10 15 20 25 30 35 CPI -.008 -.004 .000 .004 .008 .012 5 10 15 20 25 30 35 M2 -0.4 0.0 0.4 0.8 1.2 1.6 5 10 15 20 25 30 35 CR -.012 -.008 -.004 .000 .004 .008 .012 5 10 15 20 25 30 35 ER Pre-crisis
(7 endogenous variables, modified Kim & Roubini)
(2) Post-crisis -.04 -.02 .00 .02 .04 .06 5 10 15 20 25 30 35 OP -.10 -.05 .00 .05 .10 .15 .20 5 10 15 20 25 30 35 FFR -.015 -.010 -.005 .000 .005 .010 .015 .020 5 10 15 20 25 30 35 I P -.002 -.001 .000 .001 .002 .003 5 10 15 20 25 30 35 CPI -.010 -.008 -.006 -.004 -.002 .000 .002 5 10 15 20 25 30 35 M2 -.1 .0 .1 .2 .3 .4 .5 .6 5 10 15 20 25 30 35 CR -.03 -.02 -.01 .00 .01 .02 5 10 15 20 25 30 35 ER Post-crisis
(7 endogenous variables, modified Kim & Roubini)
Appendix 4
Model with Break point selected by the test statistic
