The monetary policy stance index in vietnam

Nội dung text: The monetary policy stance index in vietnam

1. THE MONETARY POLICY STANCE INDEX IN VIETNAM * Doan Ngoc Thang 1 ABSTRACT: Analyzing monetary policy in Vietnam is not straightforward because the State Bank of Vietnam (SBV) uses more than one instrument in conducting monetary policy. This paper looks for a policy stance measure that captures most of the important changes in the stance of the SBV’s monetary policy and has a closed link with the state of the economy. We apply the stance index method proposed by He and Pauwels(2008) to construct various monetary policy indexes based on different perspectives. We then apply the ordered probit model to estimate the monetary policy reaction functions. Our finding is that consolidating the growth rates of money supply and domestic credit is adequate for constructing Vietnam’s monetary policy stance index through period 1999-2015. The empirical results show that monetary policy reacts to the index of industrial production, inflation, and nominal effective exchange rate. Keywords: Monetary policy, Monetary stance, Ordered probit 1. INTRODUCTION In general, measuring accurately monetary is necessary not only to policymakers but also to scientists for both practical and analytical reasons. In particular, understanding Vietnamese monetary policy is more critical than ever before as Vietnam becomes a member of the ASEAN Economic Community (AEC) established in 2015. AEC’s goal of monetary and financial integration gives rise to the monetary cooperation problem among member states. However, interpreting the monetary policy in Vietnam is far from easy for the reason that the SBV employs more than one instrument to carry out the monetary policy. In detail, the SBV frequently uses a combination of price instruments–required reserve ratio, refinancing rate, discount rate, and prime rate, and quantity tools–broad money supply M2 and credit to set policy. It is inadequate if using one or part of the whole instrument set to capture the monetary policy stance. There is a need to develop an indicator which can characterize the Vietnamese monetary policy stance. The quality of policy stance index depends on its ability to reveal whether monetary policy becomes easing, unchanged, or tightening, and on its connect to macroeconomic variables. Most present papers concentrate on the causal effects of macroeconomic variable and adjustment of a single policy tool in investigating monetary policy stance. For instance, the growth rates of M2 and domestic credit are often employed in many papers to capture the SBV’s monetary policy stance. The levels of M2 and domestic credit in 2015 are about sixty and seventy times higher than in 1998, with an average nominal annual growth rate over the period of 28% and 77%, respectively. Based on the high growth of M2 or of credit, we can not simply conclude that Vietnam has implemented an easing monetary policy over the past fifteen years. It is because in 2008, Vietnam’s government officially announced a seven-measure toolkit to fight against high inflation and highlighted the tightening monetary policy as the most important one. * Banking Academy of Vietnam, Hanoi, 10000, Vietnam. Corresponding author. Tel.: +84989 14 2988. E-mail address: ngocthangdoan@hvnh.edu.vn
2. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 299 Concerning the policy stance measurement, Bernanke and Mihov (1998) distinguish their studies from others by accounting for more than one variable instrument. He and Pauwels (2008) develops the ”index-based method” that picks the discrete changes in the volume of various monetary policy instruments to measure the monetary policy stance in China. This approach is precisely appropriate for the analysis of the Vietnam situation. Corresponding to the high speed of economic transition and of economic integration, the conduct of Vietnam monetary policy has been changed drastically through time. When there are more than one policy instruments, it is natural to construct an indicator that covers as much as them. This approach makes use of different sources of monetary policy information which are complementary because central bank changes its favorite set of instruments over time, and of the indicator method that allows offsetting the opposite policy tools in order to produce a single signal at a point in time. Multiple sources of information, however, may cause a policy inconsistency as the indicator consecutively flips the sign. In brief, both single and multiple instruments-based stance indexes have their own merits, this calls for a thorough concern of entire instrument set to have an accurate measurement of policy stance. We aim at filling the gap by constructing a policy stance indicator which covers only two ingredients. This construction not only utilizes the advantage of the indicator- based method but also minimizes the drawback of calculation based on a variety of information. The conduct of monetary policy in Vietnam after 1998 showed signs of change from the use of direct to indirect tools. Since 1998 SBV has decided not to use credit limitation as a regular tool in managing monetary policy and to increase the use of indirect instruments such as rediscount rates, refinancing rates, base rates, and reserve requirement ratio. In 2000, the open market operations (OMOs) was officially launched and became one of the tools regularly used by the SBV. Despite these optimistic signals, throughout the period after 1998, the SBV has not yet separated from the direct and administrative tools. It occasionally used administrative instruments to curb escalating inflation, such as promulgating the interest rate cap in 2008. Since 2011, the SBV began to reuse credit limitation in implementing the monetary policy. As an agency under the government and on behalf of the government to administer monetary policy, the SBV must annually meet the targets of inflation, money supply growth and credit growth approved by the National Assembly. The SBV still relies on money supply control and credit limitation formally or informally to run monetary policy. For that reason, in this paper, we use money supply growth and domestic credit growth in building a monetary policy stance index. We then employ an ordered probit model to estimate the monetary policy function in order to examine the linkage between this index and macroeconomic variables. The remainder of this paper is organized as follows. Section 2 review the literature on the monetary policy stance index and the policy reaction function obtained by using an ordered probit technique. Section 3 describes the construction of Vietnam’s monetary policy stance indicator. Section 4 introduces the ordered probit model and the data set, and then discusses the main results of estimation. Section 5 concludes. 2. LITERATURE REVIEW Our work is related to two strands of literature: First, we build on literature on multivariate- based monetary policy stance indicator. The wave of criticism on a single measure of policy stance at the end of 20th century paves a way to the birth of multiple one. Romer and Romer (1989) and Boschen and Mills (1995) start this movement by applying the traditional narrative-based method to study monetary stance. They collect information on a variety of documents issued by the monetary authorities and then convey them into integer numbers in the interval [−2, 2] to represent a range from strongly easing to strongly tightening respectively. However, the problems of this method are the subjectivity and the divergence between intention and actual policy actions of the monetary authorities. While the former severely affects the results’ accuracy, the latter misleads the monetary policy analysis. Bernanke and Mihov (1998) sort the
3. 300 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA nonborrowed reserves, total reserves, and federal funds rate in a structural vector autoregression (SVAR) model and define the policy stance as a linear combination of the policy shock. This SVAR has become the workhorse model for monetary policy analysis in the U.S. and other developed countries. The use of SVAR’s disturbance to capture monetary shocks causes discrepancies among different estimations, this questions the application of this model in policy analysis. Moreover, in practice the monetary policy is predicted, i.e., the Taylor rule, that is in contrast to the unpredicted focus of the SVAR. Therefore, the feedback rule derived from the SVAR-based measure could understate the role of monetary policy. Second, the response of the monetary policy stance to macroeconomic variables is estimated to assess the appropriate index to characterize a central bank’s behavior. This response is contingent on a monetary policy rule. Taylor (1993) proposes a quantitative study of this framework by regressing the federal funds rate on inflation and the output gap. The result is known as the Taylor rule which becomes a standard in the analysis of the monetary authorities’ reaction in a variety of countries, for example, Peersman and Smets (1999) for the Euro area, Gerlach and Schnabel (2000) for the European monetary union, and Taylor et al. (2006) for the United Kingdom. Because of the discrete quality of the dependent variable-the monetary policy stance index, the conventional ordinary least square regression causes a biased estimation. This problem can be solved with the ordered choice method, such as probit model whose the disturbance is assumed to follow a normal distribution. Several papers apply this method to analyze the monetary policy reaction function. Eichengreen et al. (1985) study the Bank of England’s discount policy under the interwar gold standard. The weekly decision whether to increase, cut or leave the discount rate unchanged is formulated as a nonlinear function of reserve position, difference between domestic and foreign interest rates, level of economic activity, and level of discount rate whose value takes the discount rate during the previous period if this rate was higher than 4% or zero otherwise. Gerlach (2007) uses the European Central Bank’s Monthly Bulletin to create the choice variable. The empirical results indicate that monetary policy responds to M3 growth, real economy status, and changes in exchange rate, except inflation. Kim et al. (2016) employ three types of constrained ordered choice model and showed the essential role of output gap and exchange rate in studying the Bank of Korea’s interest rate decision-making process. Our paper is closely related to He and Pauwels (2008) and Xiong (2012) which measure the monetary policy stance index of the People’s Bank of China and then use the ordered probit technique to estimate the policy reaction function. However, our paper is different from them in several aspects. First, unlike He and Pauwels (2008), we do not fix the nonstationary problem because we do not find any firm evidence of nonstationarity in our model regressors. Second, instead of covering as many instruments as possible when constructing policy stance, we rely on Vietnam’s characteristics and comparison of estimation results from regressing the benchmark stance index and the proposed one to examine whether the latter can capture the SBV’s behavior. 3 VIETNAM’S MONETARY POLICY STANCE INDEX Table 1: Main monetary policy instruments: 1999-2015 Period Main monetary policy instruments Before 2000 Credit limit/Refinancing Various interest rate Required reserves ratio 2000-2010 Refinancing/OMOs Various interest rate Required reserves ratio
4. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 301 2011-2015 Credit limit/Refinancing/OMOs Various interest rate Required reserves ratio The objective of the SBV’s monetary policy is promulgated by law as stabilizing the value of the currency and thereby promoting economic growth. The former is interpreted as control of the inflation rate together with stabilization of the exchange rate. In order to reach these multiple objectives, the SBV sets multiple instruments, and it rather combines different monetary instruments through time. For example, the credit limit and refinancing facility used as the key monetary policy tool during nearly the whole 1990s. Although credit limit was revoked in 1998, the SBV has conducted it again since March 2011. During 2008, the open market operations, reserved requirement ratio, and various interest rate were simultaneously used to reduce inflation. Because of the inadequacy of each monetary policy tool solely as a policy stance measure and the SBV’s habit of implementing the monetary policy, it could be possible that the more information on monetary policy conducts, the better indicator we have. 3.1 Key Policy Instruments It goes back as far as the early stage of Vietnam economic reform when reserved require- ment ratio (RRR) was rarely used to control for money supply. As the banking system scale becomes more massive since 1999, the SBV uses RRR with increasing frequency, especially during the time fighting high inflation. Refinancing has been used in Vietnam since 1991 in order to facilitate liquidity. It includes two forms: re- discount and refinance. In 2003, SBV created an interest rate corridor on the interbank market by regulating the rediscount rate as floor rate and refinance rate as cap rate. In July 2000, open market operations (OMOs), a more market-oriented policy tool, was introduced. SBV frequently transacts the central bank bill to sterilize operations derived from foreign capital interventions. Both refinancing and OMOs reflect the liquidity facility supplied by the SBV, but these data are not available. Therefore, we look at the net change by using the net claims on deposit banks as a proxy. 3.2 Constructing the Monetary Policy Stance Index The construction of the policy index requires us to categorize the change in each monetary series. We follow most of the literature on monetary policy stance index to assign value in a set of three choices (1, 0, −1) to represent a tightening change, no change, and an easing change in the policy reactions, respectively (Bernanke and Blinder, 1992; He and Pauwels, 2008; Xiong, 2012). For rate-based policy tools, we regard the direction of the change as policy stance. For example, a decrease in the interest rate or the reserved requirement ratio are treated as easing change. The rest of the variables need thresholds to identify their changes. We filter the M2 and domestic credit growths by employing the Hodrick-Prescott method and keep the cyclical components. All the growth rate is calculated as the per- centage change during the past twelve months unless otherwise stated. The one standard deviation of each series is used as a threshold to classify the fluctuations. We mark tight- ening (easing) change for any cyclical component which decreases (increases) more than one standard deviation. The rest is treated as an unchanged one. Applying the above method to net claims on deposit banks seems unreasonable since it fluctuates wildly in terms of percentage changes. We rather use the average of absolute month-on-month change as a criterion. If a net claims increase (decrease) more than 15,116 billion dong, the change is viewed as a big one, and the corresponding monetary stance is then defined as easing (tightening). This criterion is approximately equal to 80% of the compulsory sterilization bond issued by the SBV to absorb the banks’ excess reserves from
5. 302 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA commercial banks in early 2008. This event reflects the strong reaction of the SBV to curb surging inflation and also shows a clear signal of monetary stance to the economy. Hence, this threshold balances the risk of under-identifying and over-identifying the right signal of changes to the policy stance. After obtaining the policy change of all variables, we sum up them before constructing the overall monetary policy stance change index, denoted by St. Similar to the single index of each policy tool, we assign value to St in a set of three choices (1, 0, −1) to represent a tightening change, no change, and an easing change, respectively, as follows: It is noteworthy that we follow He and Pauwels (2008) to assume equal importance to all the instruments when calculating the sum. This assumption deals with the opposite directions that happen 24 times in our sample period. The overall stance can be no change if the different signs cancel out each other or determined by the majority of changes. For example, in June 2007 the SBV increased reserved requirement ratios while rising the net claims on deposit banks by 16, 655.7 billion dong. The former stands for the tightening and the latter represents the easing, leading no change in the overall monetary policy stance. In February 2008, the policy stance index was specified by easing in both domestic credit and net claims on deposit banks together with tightening in various interest rate and reserved requirement ratio.
6. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 303 3.3 An Overview of the Policy Stance Index The policy stance index plotted in Figure 1, in general, captures most of the fun- damental changes in the stance of the SBV’s monetary policy. The consequence of the Asian crisis in 1997 reduced the world price and demand and in turn, dampened the domestic and foreign demand for Vietnam products. The year 2000 first time witnessed a slight deflation after persistent inflation. To boost the economy, SBV implemented an expansionary monetary policy. For example, SBV cut the required reserves ratio from 10% to 7% in March 1999, then to 5% in July 1999. It also cut the refinance rate in June, September, and November of 1999. In April and August 2000, SBV reduced the discount rate by 0.6% each time. The average growths of money and credit in the period 1997-2003 were about 30 − 40% and the Vietnam dong was devalued about 36%. Inflation came back in the year 2004 with 9.5% which was much higher than the Vietnam government’s targeted rate, 6%. To curb the inflation, SBV started to tight the monetary policy by increasing interest rate and fixing the exchange rate since 2004. The Vietnam Ministry of Finance and SBV used the administrative tools to adjust interest rate rather than indirect monetary policy (Camen, 2006). Inflation, after a slight drop in 2006, had risen sharply to 12.6% in 2007 and to 20% in 2008. Several reasons have been put forward to explain the sharp rise in inflation in 2007-2008. They are a sharp increase in the minimum wage, an increase in international commodity prices, loose and inflexible monetary policy, rigid and flexible exchange rate management policies. Moreover, Vietnam’s opening up to the world since Vietnam’s accession to the WTO at the end of 2006 had led to the influx of foreign indirect investment into Vietnam, driving up stock prices and asset prices. In order to stabilize the exchange rate, the SBV had injected a significant amount of Vietnam dong into the economy, contributing to the exacerbation of the inflation. The 2008-2009 world economic crisis has contributed to lowering inflation in Vietnam since late 2009. The fall in international prices together with the decrease in aggregate demand has helped Vietnam reverse the dreaded rise in inflation in 2008. As the government’s stimulus packages began to increase from the second quarter of 2009, money supply also began to increase sharply, and credit also showed similar signs. Commercial banks were becoming less cash-strapped and trying to raise the interest rates to attract deposits. As a result, the interest rate competition had begun to push up lending rates. As a result, SBV changed its stance to contractionary policy since February 2011. The policy change index computed in this part seems to represent the crucial changes in the monetary policy of Vietnam. However, the index cannot capture the pressure of tightening or easing in a given period. For instance, SBV stops tightening or changing to easing for one month. This temporary signal may confuse those who use this index. Two reasons are used to explain this drawback. First, the construction of this index are based on multiple variables that may cancel out each other, taking October 2008 as an example. SBV increases the refinance by 1% to 14% while decreases the base rate by 1% to 13%. Second, the SBV usually adjusts the interest rates and the required reserve ratio with a broad margin. Therefore, the current index only records the policy change one time for a long period. For example, SBV raises the discount rate from 6% in February 2008 to 10% in November 2008 and raises the refinance rate from 7.5% in February 2008 by 5.5% in May 2008. The margins of required reserve ratio are 1% − 5% which is much higher than the one, 0.5%, frequently employed by the Bank of China. We propose a new way to construct the Vietnam monetary policy stance index. Instead of using all of the policy instruments, we only use the growth rates of money supply (M2) and domestic credit. The policy changes of each variable and the overall stance are defined as in subsection 3.2. Figure 2 describes the
7. 304 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA proposed policy stance index which seems to well capture the pressure of policy stance in a given period. Table 2 shows a comparison between the benchmark and proposed stance indices. Fit measures the number of times that both indices reflect the same stance. The proposed index captures 50% stance of the benchmark index. No change ranks first with 59.1% which is higher than the average. The fit of tightening stance is much higher than that of easing with 48.3% and 29.3%, respectively. The numbers of tightening stance in both indices are roughly equal while the amount of easing stance in the proposed index is lower than that in the benchmark case. 4 AN ORDERED PROBIT APPROACH 4.1 The Model * We assume that the SBV sets its policy stance St by the following linear function at time t. * where St is a latent variable capturing a preferred monetary policy change by the SBV with regard to the policy change index. Xt−1 is a k × 1 vector of macroeconomic characteristics covariates of the economy. β is a k × 1 parameter vector and εt is a residual term. The equation (2) is analogous to the Taylor monetary reaction function where the policy rate responds to the fluctuations of one-period lag of macroeconomic variables. The explanatory variables include the growth rate of index of industrial production (IIP) and the inflation. In Vietnam’s monetary framework, the SBV actively intervenes in the foreign market to stabilize the dong value; therefore the growth rate of the nominal effective exchange rate enters as an external force. The growth of the Vietnam stock index is also used to explain the SBV’s stance. This paper uses equation (2) as the baseline model. Some papers study the reaction of policy stance to the deviation of a set of macroeconomic variables from its targets. When the targets are set by policymakers, this type of model is inappropriate in Vietnam because of the missing targeted values of inflation before 2005 and of the low credibility. The targeted inflation had not been announced until 2005 while our sample starts from the year 1999. Camen (2006) pointed out that in Vietnam the economic growth has been the de facto primary goal, it is usual to observe the targeted inflation being overshot. On the other hand, we run the regression with the proposed index for comparison. As the aforementioned part shows, there are three policy action: easing, no change, or tightening, * which indicates three regimes for the sake of St . These three regimes imply that there are two unknown cut points τ1 < τ2 and we can define the following policy index measure St Given the standard normal assumption for εt , the conditional probabilities of the observed policy changes are followed
8. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 305 where Φ is a normal distribution cumulative function. The econometric structure of equation (4) reveals the movement of the monetary policy stance. It should be noted that the parameter β itself is of limited interest because it is inadequate to determine the sign of the marginal effect. The sign also depends on the level of covariates Xt−1. The direction effect of Xk on the probabilities P(St = −1|Xt−1) and P(St = 1|Xt−1) is unambiguously determined by the sign of βk, whereas the sign of βk does not always determine the direction effect for the outcome 0 (no change). The parameters β, τ1, and τ2 can be estimated by maximum likelihood method. We employ the Stata 14 software. 4.2 Data In Table 3, we briefly describe the variables and the sources of the data. The monthly observations for all variables from 2001M7 to 2015M12 are used. The year-on-year percentage change in real IIP growth, CPI inflation, NEER and Vnindex growth are incorporated into the model to capture real output, inflation, the SBV’s foreign exchange policy, and the transaction of Hochiminh stock exchange. We employ the ADF test for all variables and see that all of them are stationary. We provide graphs for these covariates in Figure 3. Table 3: Definition and data sources of variable Variable Definition Source IIP y-o-y change in real IIP General Statistics Office, Vietnam Inflation y-o-y change in consumer price index International Financial Statistics, IMF y-o-y change in nominal effective exchange rate Author’s calculation NEER NFA y-o-y change in net foreign asset General Statistics Office, Vietnam Vnindex y-o-y change in Vnindex Hochiminh Stock Exchange The multicollinearity among variable in Xt−1 is one of the potential problem. However, Table 4 shows the small correlation among explanatory variables that implies a weak evidence of multicollinearity. Hence, we abstract this issue in this paper.
9. 306 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA 4.3 Results The estimation results are shown in Table 5. Regression 1 uses the benchmark stance index as a dependent variable while regression 2 uses the proposed index. All the explana- tory variables are lagged by one month. Most of the coefficients are statistically significant except for Vnindex. The coefficient of NFA becomes insignificant when we run regression 2. Although the thresholds τ1 and τ2 are used for computations, they are not our key interest. The pseudo R2 is interpreted as a goodness of fit. Regression 2 has a higher pseudo R2 as compared to regression 1. However, both of them are relatively low that is common in the ordered probit method. The estimation results reveal several things about the SBV’s policy reaction. Firstly, the IIP growth, inflation, and NEER play a crucial role in the SBV’s objective function which corresponds withthe conventional wisdom about the SBV’s monetary policymaking. According to the State Bank of Vietnam’s Law in 1992 and adjusted in 2010, the ultimate goals of SBV are to promote economic growth and to stabilize the value of the currency. The latter is interpreted as the stability of the exchange rate. Secondly, the SBV does not respond to Vnindex. The monetary policy may affect the stock price, but this is not our interest in this paper. Thirdly, the SBV reacts to NFA in the benchmark model but not in the proposed one. The SBV manages the inflows of foreign capital by frequently intervening in the foreign exchange market. However, the proposed stance index fails to capture this aspect. In an ordered probit model, the interpretation of the marginal effects is not straightfor- ward as in the standard ordinary least square. The partial effect of one variable depends on the coefficients of all variables
10. INTERNATIONAL CONFERENCE STARTUP AND INNOVATION NATION 307 and their levels. Hence, in Table 6 and 7 we report the marginal probability of the variable at the value of mean for regression 1 and 2, respectively. The first line of Table 6 describes the partial effect of the percentage change in IIP growth on specific probabilities of monetary policy change. One percent increase in IIP growth leads to 0.0054% decrease in the probability of an easing change in monetary policy. By contrast, the probability of a tightening policy change will rise by 0.0068%. The effects on both easing and tightening policy changes are statistically significant at the 10% level. The effects of inflation, NEER, and NFA on policy changes are similar to that of IIP growth in sign but more significant. These policy changes are in line with the conventional knowledge. The partial effects of Vnindex have the expected sign but are insignificant. 5 CONCLUSION This paper constructs the monetary policy stance index of the SBV who implements various policy instruments. This index fails to capture the pressure of policy stance in a given period. We propose a new index that includes only the growth rates of money supply and domestic credit. This proposed index reasonably replaces the benchmark one in the sense that it well captures most of the fundamental changes as well as the pressure given a period of the SBV’s monetary policy. A comparison of the monetary policy reaction functions with two indexes based on the ordered probit model is then used to reinforce our proposal. REFERENCES Bernanke, B. S. and Blinder, A. S. (1992). The federal funds rate and the channels of monetary transmission. The American Economic Review, 82(4):901–921. Bernanke, B. S. and Mihov, I. (1998). Measuring monetary policy. The Quarterly Journal of Economics, 113(3):869–902. Boschen, J. F. and Mills, L. O. (1995). The relation between narrative and money market indicators of monetary policy. Economic inquiry, 33(1):24–44. Camen, U. (2006). Monetary policy in vietnam: the case of a transition country. BIS Working Paper, 31. Eichengreen, B., Watson, M. W., and Grossman, R. S. (1985). Bank rate policy under the interwar gold standard: a dynamic probit model. The Economic Journal, 95(379):725–745. Gerlach, S. (2007). Interest rate setting by the ecb, 1999-2006: Words and deeds. Interna- tional Journal of Central Banking, 3(3):1–46. Gerlach, S. and Schnabel, G. (2000). The taylor rule and interest rates in the emu area. Economics Letters, 67(2):165–171. He, D. and Pauwels, L. L. (2008). What prompts the people’s bank of china to change its monetary policy stance? evidence from a discrete choice model. China & World Economy, 16(6):1–21.
11. 308 HỘI THẢO KHOA HỌC QUỐC TẾ KHỞI NGHIỆP ĐỔI MỚI SÁNG TẠO QUỐC GIA Kim, H., Shi, W., and Hwang, K.-M. (2016). Estimating interest rate setting behaviour in korea: a constrained ordered choices model approach. Applied Economics, 48(23):2199– 2214. Peersman, G. and Smets, F. (1999). The taylor rule: a useful monetary policy benchmark for the euro area? International Finance, 2(1):85–116. Romer, C. D. and Romer, D. H. (1989). Does monetary policy matter? a new test in the spirit of friedman and schwartz. NBER Macroeconomics Annual, 4:121–170. Taylor, J. B. (1993). Discretion versus policy rules in practice. Carnegie-Rochester conference series on public policy, 39:195–214. Taylor, M. P., Davradakis, E., et al. (2006). Interest rate setting and inflation targeting: Evidence of a nonlinear taylor rule for the united kingdom. Studies in Nonlinear Dynamics and Econometrics, 10(4):1–18. Xiong, W. (2012). Measuring the monetary policy stance of the people’s bank of china: An ordered probit analysis. China Economic Review, 23(3):512–533.