Singapore: Selected Issues

International Monetary Fund

doi:10.5089/9781451834284.002.A002

Jump to Content Jump to Main Navigation

Singapore: Selected Issues

Author: International Monetary Fund

Publication Date:: 13 Aug 2008

eISBN:: 9781451834284

Language:: English

Keywords:: ISCR; CR; bank; exchange rate; contractionary monetary policy; contractionary monetary policy shock; inflation shock; NEER appreciation; inflation in Singapore; effects of monetary policy; Nominal effective exchange rate; Structural vector autoregression; Global

Download PDF (312.1 KB)

This Selected Issues paper assesses the stability of Singapore’s banking system in a regional context. It proposes a novel methodology for gauging domestic financial stability. The paper assesses the impact of fiscal measures on macroeconomic activity, and analyzes the effects of monetary policy using structural vector autoregressions. Estimates show that the Monetary Authority of Singapore’s exchange rate-centered framework is well suited to shape monetary decision making, given the large impact that changes in the nominal exchange rate have on activity and prices.

Abstract

II. The Effects of Monetary Policy in Singapore¹

Singapore’s monetary policy is unique. It uses the exchange rate as an intermediate target to achieve low inflation and sustainable growth. Over the past year, rising inflation and slowing growth have posed challenges to the conduct of monetary policy. In this paper we seek to understand better the impact of monetary policy on inflation and output using structural vector autoregressions (SVAR). According to the empirical model, a contractionary monetary policy shock (identified as a nominal effective exchange rate appreciation) has powerful effects on both output and prices, providing support for the exchange rate-centered monetary framework.

A. Introduction

1. Singapore manages its exchange rate against an undisclosed basket of currencies. The Monetary Authority of Singapore (MAS) sets the rate of change of the nominal effective exchange rate (NEER)—its intermediate target—to achieve low inflation and sustainable growth. This framework has been in place since 1981 and over this period annual inflation has been 1 ¾ percent (on average), while GDP growth has averaged 7 percent. Unemployment has also been remarkably low, at less than 3 percent during 1987–2007.

2. Starting in mid-2007, inflation has risen sharply—reaching almost 7 percent in the first quarter of 2008—while growth has remained relatively strong. Several factors have been in play, including a 2 percentage point increase in the sales tax (July 2007), a sizable upward reassessment of property values (January 2008) and, more recently, spikes in global commodity prices. In response, the MAS tightened monetary policy in October 2007 (steepening the slope of the exchange rate band) and April 2008 (recentering the band), despite flagging external demand. Given Singapore’s exceptional trade openness, the current environment poses challenges to Singapore’s exchange rate-centered policy framework.

3. This paper sheds light on Singapore’s unique monetary transmission mechanism. It follows a well-established literature on estimating the effects of monetary policy using structural vector autoregressions (SVARs). The paper uncovers powerful effects of the exchange rate on output and inflation, supporting the rationale for the exchange rate-centered monetary framework. Section B provides an overview of inflation developments in Singapore, in particular its persistence and correlation with the nominal effective exchange rate. Section C motivates the main assumptions underlying the SVAR. Section D presents the main empirical results and briefly discusses alternative specifications, focusing on the impact of monetary policy shocks on output and inflation. Section E concludes. The data and technical estimation details are presented in the Annex.

B. Inflation in Singapore: Some Stylized Facts

4. As noted above, inflation in Singapore has been remarkably low and stable but has recently picked up significantly, reaching a 26-year high in Q4 2007. The recent spike marks a deviation from a declining trend. Inflation has generally been trending downward since the beginning of the 1980s, averaging 2 ½ percent in the 1980s, 1 ¾ percent in the 1990s, and just over 1 percent during 2000–2007. With the exception of the early 1980s, the volatility of inflation has been generally low—ranging between 1 to 2 percent. Volatility has gone up since the second half of 2007 on the wake of rising inflation.

5. Inflation has only been mildly persistent over the last two decades. As measured by the half-life of shocks from simple univariate autoregressive (AR) models, shocks to inflation have been relatively short lived.² The empirical model estimated for quarterly inflation is an AR(1) or AR(2), depending on the sub-sample.³ The sum of the AR coefficients is about 0.5–0.6, indicating that the half-life of inflation shocks is about 1–1 ½ quarter.⁴

6. Inflation in Singapore is correlated with both the nominal effective exchange rate and output. The contemporaneous correlation between quarterly inflation and the NEER is relatively high at minus 0.4, suggesting that NEER appreciations tend to occur in tandem with lower inflation. Granger predictability tests reveal that inflation Granger-causes the NEER, but the evidence that the NEER causes inflation is weaker and not statistically significant. This suggests that while monetary policy (captured by NEER changes) responds to inflation shocks rather quickly, the pass-through from the exchange rate to inflation may be low and that the effect of the NEER on inflation may operate with a long lag. Consistent with a Phillips curve relationship, inflation is positively correlated with deviations of output from its trend.⁵

C. Data and Empirical Model

7. The data are quarterly and span the period 1979–2007. The variables included in the SVAR are the consumer price index (CPI), real GDP, the NEER, the domestic 3-month nominal interbank interest rate (SIBOR), money aggregates (M1 and M2), and foreign variables. The latter includes the “world” oil price (average from the IMF’s WEO), a trade-weighted foreign GDP, and the 3-month LIBOR. In the case of GDP and CPI, the series are seasonally adjusted. The data series are shown in the Annex.

8. The SVAR methodology is applied to identify monetary policy shocks and simulate their impact on output and inflation. The baseline identification assumption is adapted from Kim and Roubini (2000). They present a nonrecursive identification scheme that generate hump-shaped response of output to a contractionary monetary policy shock and has been widely used.⁶ One of the main departures from Kim and Roubini (2000) is that the NEER substitutes for the short-term interest rate in the policy reaction function. More specifically, the SVAR model in this paper assumes that the model economy can be represented by:

B_{0} Y_{t} = k + B_{1} y_{t - 1} + ... + B_{p} y_{t - p} + u_{t}

$where y_{t} = [p_{t}^{oil} i_{t}^{*} x_{t} p_{t} m_{t} {neer}_{t} i_{t}]$ is the n x 1 data vector containing the oil (or commodity) price index (p^oil), foreign interest rate (i^*), real GDP (x), domestic CPI (p), monetary aggregate (m), NEER, and domestic interest rate (i); k is a vector of constants, B_{_k} is an n x n matrix of coefficients (with k = 1, ..., K), and u t is a white-noise vector of structural shocks. All variables enter the VAR in natural logarithms.

9. The SVAR approach is well-suited to the analysis of monetary policy effects in Singapore. One of the SVAR main advantages is its simplicity and the fact that it does not impose potentially restrictive assumptions about behavioral relationships and the dynamics of the economy. In the case of Singapore, the same monetary policy regime since the early 1980s provides a relatively long sample to identify monetary policy shocks without concerns about structural breaks typically associated with changes in the policy regime.

10. The following contemporaneous restrictions are imposed to identify the structural shocks (the details and the associated matrix are presented in the Annex):⁷

The commodity price index is exogenous with respect to all the variables in the system; in contrast, the domestic interest rate (being a financial variable) is affected by shocks to all other variables included in the VAR;
The foreign interest rate and domestic output responds contemporaneously to the oil price (or commodity prices) within a quarter, but the latter is not affected by the former contemporaneously (zero restriction); as in Kim and Roubini (2000), firms adjust output in response to policy shocks or financial market shocks with a lag;
Domestic prices respond contemporaneously to oil price shocks and to output (the second restriction can be relaxed without affecting the results);
Money responds to domestic output and interest rates, consistent with standard money demand theory. The restriction that the coefficient on the interest rate is zero may be imposed without affecting the estimated impulse response functions; and
The NEER responds to output, prices, the oil price, and domestic interest rates. The inclusion of the oil price may account for the pre-emptive nature of monetary policy as it responds to expected price pressures consistent with its medium-term orientation.

D. Main Findings

11. The main results from the SVAR are consistent with a strong effect of monetary policy on output and prices. To evaluate the impact of the NEER on activity and the CPI, the empirical model is used to estimate impulse-response functions. The results may be described as follows:

A contractionary monetary policy shock is described as a NEER appreciation. The appreciation is highly persistent and remains statistically significant up to 8 quarters;
Consistent with the MAS’ own findings, contractionary monetary policy shocks have strong effects on output. A NEER appreciation shock lowers output (with a hump-shaped impulse response function)—the effect is economically and statistically significant after 2 quarters and peaks at 8 quarters. The lagged impact justifies a forward-looking orientation of monetary policy;
A monetary policy contraction has a persistent and strong negative effect on the price level; the effect of the contractionary shock on the CPI becomes economically and statistically significant after 2 quarters and peaks after 8 quarters;
Underscoring the rationale for an exchange rate-centered monetary policy framework, the effect of an interest rate increase on output is small and statistically insignificant at the 5 percent level;
Output shocks have a strong positive impact on the CPI, as implied by a Philips curve relationship (see also Parrado, 2004). The impact is strongest at 4–6 quarters and dissipates after 8–10 quarters as the effect of nominal and other rigidities diminish;
The NEER responds with relatively short lags to shocks to the CPI and output. In particular, the NEER appreciates in response to increases in output (after 2 quarters) and the CPI (after 1 quarter). The effect of the latter on the NEER is generally small and is not statistically significant at the 5 percent level. This is in tune with an empirical characterization of monetary policy decisions in which the MAS targets the rate of change of the NEER according to a Taylor-rule. Parrado (2004) and McCallum (2007) show that in fact a Taylor rule for the NEER provides a good fit to the MAS’ policy reaction function.
The effects of output (positive) and interest rate (negative) on monetary aggregates (M1 or M2) is consistent with a well-behaved money demand curve; and
As expected, the NEER and domestic interest rate respond strongly to foreign interest rate shocks; also, domestic interest rates decline (on impact) following a NEER appreciation.

12. Alternative identification schemes yield broadly similar results.⁸ Applying a slightly modified version of the Eichenbaum and Evans (1995) recursive assumptions, the estimated VAR becomes $where y_{t} = [p_{t}^{oil} i_{t}^{*} x_{t} p_{t} m_{t} {neer}_{t} i_{t}]$ .⁹ As noted above, the effect of an interest rate shock on output is slightly stronger as well as the response of the NEER to CPI shocks. Reversing the ordering of x and p or neer and i does not affect the qualitative results. The results from imposing sign restrictions on the impulse responses, (Uhlig, 2005) are also consistent with those of the baseline model, but are less robust to the changes in the underlying assumptions. For example, the negative response of the CPI to a contractionary monetary policy shock is in line with results generated by standard monetary models, but the impulse response (and its shape) depends largely on the assumed lagged effect of NEER shocks on output.

E. Concluding Remarks

13. This paper assesses the effects of monetary policy on economic activity and inflation. The findings suggest an important role for monetary policy in delivering low and stable inflation, a salient feature of Singapore’s recent monetary history.

14. The results provide support for the exchange rate-centered monetary framework. The main findings confirm that the effects of the interest rate shocks on output and prices are significantly less important than those of the nominal exchange rate. In addition, according to the estimated models, monetary policy can be empirically characterized as a Taylor rule in which the NEER responds to output and inflation shocks. The powerful effects of monetary policy combined with the credibility of the framework may explain the relatively low inflation persistence.

Annex II.1. Estimation Details

15. The reduced form model is estimated with six lags in log-levels, except for the domestic and foreign interest rate.¹⁰ While all variables can be characterized as nonstationary (or near-nonstationary as in the case of interest rates) according to standard unit roots tests, most findings are robust to first differencing and inference can still be conducted with the estimated model in levels (Canova, 2007, page 125). The structural model can be rewritten in reduced form as:

y_{t} = c + C_{1} y_{t - 1} + ... + C_{p} y_{t - p} + e_{t}

where $e_{t} = B_{0}^{- 1} u_{t}$ is also white-noise vector process, with variance-covariance matrix given by ${Ω = B_{0}^{- 1} D (B_{0}^{- 1})}^{'},$ where D is the variance-covariance matrix of the structural shocks. The matrix Ω can be rewritten as Ω = ADA’ where D is diagonal. In this case, since $u_{t} = A^{- 1} e_{t},$ With $A = B_{0}^{- 1}, then E (u_{t} u_{t}^{'}) = E (A^{- 1} e_{t} e_{t}^{'} (A^{- 1})^{'}) = A^{- 1} (AD A^{'}) (A^{- 1})^{'} = D, i . e .,$ the vector u_{_t} is orthogonal and can now be interpreted as “structural”shocks.¹¹ In practical terms, identification amounts to imposing restrictions on the matrix $B_{0}^{- 1}$ that orthogonalizes the reduced form errors, eliminating their contemporaneous correlation.¹² A widely used identification scheme is the recursive ordering (Cholesky decomposition) proposed by Sims (1980), which assumes that A has a lower triangular structure. This is equivalent to a hierarchical ordering of the variables, with the most exogenous variable ordered first.

16. Statistical inference can be conducted directly based on the estimated log-likelihood. If there are n^* estimated parameters in B_₀, the number of over-identifying restrictions (r) is given by r = (n(n -1) / 2) - n^*. The test for over-identifying restrictions is based on the maximized value of the log-likelihood and has a chi-square distribution with r degrees of freedom.¹³

Identifying Restrictions

17. The restrictions described in the main text can be written as:

[\begin{matrix} 1 & 0 & 0 & 0 & 0 & 0 & 0 \\ a_{21} & 1 & 0 & 0 & 0 & 0 & 0 \\ a_{31} & 0 & 1 & 0 & 0 & 0 & 0 \\ a_{41} & 0 & a_{43} & 1 & 0 & 0 & 0 \\ 0 & 0 & a_{53} & a_{54} & 1 & 0 & a_{57} \\ a_{61} & 0 & a_{63} & a_{64} & 0 & 1 & a_{67} \\ a_{71} & a_{72} & a_{73} & a_{74} & a_{75} & a_{76} & 1 \end{matrix}] [\begin{matrix} p_{t}^{oil} \\ i_{t}^{*} \\ x_{t} \\ p_{t} \\ m_{t} \\ {neer}_{t} \\ i_{t} \end{matrix}] = B (L) [\begin{matrix} p_{t}^{oil} \\ i_{t}^{*} \\ x_{t} \\ p_{t} \\ m_{t} \\ {neer}_{t} \\ i_{t} \end{matrix}] + [\begin{matrix} u_{t}^{poil} \\ u_{t}^{t *} \\ u_{t}^{x} \\ u_{t}^{p} \\ u_{t}^{m} \\ u_{t}^{neer} \\ u_{t}^{i} \end{matrix}]

where $B (L) = \sum_{i - 1}^{p} B_{i} (L^{i})$ is a matrix polynomial in the lag operator (L) and u_{_t} is the vector of “structural” shocks. In this case, the over-identifying restrictions test is distributed as a chi-square with four degrees of freedom. For instance, according to the model above, the empirical policy reaction function is given by:

{neer}_{t} = - a_{63} x_{t} + a_{- x}^{'} y_{t} + B (L) y_{t} + u_{t},

where the impact of output shocks on the NEER is given a₆₃ and a^‘_-x is the vector of coefficients excluding that on x. In this baseline specification there are four over-identifying restrictions. Additional zero restrictions are also imposed on a₆₁, a₄₃, and a₆₇. In some cases, the impact of the interest rate on output is larger but is only marginally significant (e.g., when only a₄₃ = 0 is imposed).

References

Canova, Fabio (2007). Methods for Applied Macroeconomic Research. Princeton University Press.
- Search Google Scholar
- Export Citation
Eichenbaum, Martin and Evans, Charles (1995). “Some Empirical Evidence on the Effects of Monetary Policy Shocks on Exchange Rates,” Quarterly Journal of Economics, 110(4), 975–1009.
- Search Google Scholar
- Export Citation
Kim, Soyoung and Roubini, Nouriel, (2000). “Exchange rate anomalies in the industrial countries: A solution with a structural VAR approach,” Journal of Monetary Economics, 45(3), pages 561-586.
- Search Google Scholar
- Export Citation
McCallum, Bennett (2007). “Monetary Policy in East Asia: the case of Singapore,” IMES Discussion Paper 2007 E10.
- Search Google Scholar
- Export Citation
Parrado, Eric (2004). “Singapore’s Unique Monetary Policy Framework: How Does It Work?” IMF Working Paper.
- Search Google Scholar
- Export Citation
Reis, Ricardo and Pivetta, Frederic. (2007). “The Persistence of Inflation in the United States,” Journal of Economic Dynamics and Control, 31 (4), 1326-1358.
- Search Google Scholar
- Export Citation
Sims, Christopher, and Zha, Tao (1999). “Error Bands for Impulse Responses,” Econometrica, 67 (5), 1113-1155.
- Search Google Scholar
- Export Citation
Uhlig, Harald (2005). “What Are the Effects of Monetary Policy? Results from an Agnostic Identification Procedure,” Journal of Monetary Economics, 52 (2), 190-212.
- Search Google Scholar
- Export Citation

Prepared by Roberto Guimaraes-Filho.

The estimated degree of persistence at time t reflects what inflation is expected to be at time t + s, conditional on all the present and past inflation up to time t.

The model is chosen according to the Bayesian information criterion. Low-order autoregressive dynamics are present in the data, with one lag (or two, depending on the effective estimation sample) providing a good fit.

This is much lower than the estimated persistence calculated by Reis and Pivetta (2007) using post-WWII U.S. data.

This result is robust to at least two different measures of the trend (i.e., applying the HP and band-pass filters).

Other identification schemes are applied to assess the robustness of the results. The recursive identification of Eichenbaum and Evans (1995) and the sign approach proposed by Uhlig, (2005) are briefly discussed below.

No restrictions are imposed on the lagged structural parameters of the model.

In addition, there is no evidence of significant structural instability in the reduced-form VAR. For each equation of the reduced-form VAR, Andrew’s sup-Wald test is applied to test jointly for the stability of all the coefficients on the lags of a given variable. In this regard, the impulse response functions for the interest rate and NEER based on a reduced form estimated over the 1991–2007 period are broadly similar to those reported above, suggesting that there have been no major changes in the monetary transmission mechanism. (This may change with the rising importance of domestic demand and interest rate-sensitive sectors). Interestingly, the impact of interest shocks is larger (but is only marginally significant) when additional over-identifying restrictions are imposed (see Annex).

The baseline recursive structure in Eichenbaum and Evans (1995) does not include the oil price but incorporates the ratio of U.S. nonborrowed reserves to total (banking) reserves to identify the monetary policy shocks.

The estimated reduced form has 4 lags and a time trend also yields a good fit, with the reduced form passing the standard specification tests for autocorrelation and heteroskedasticity. Regarding the normality of the residuals, there is some excess kurtosis as indicated by the Jarque-Bera test. The structural parameters are estimated by maximum likelihood, but it may also be estimated by solving the nonlinear system given by $Ω = B_{0}^{- 1} D {(B_{0}^{- 1})}^{'}$ .

Since the equality $e = B_{0}^{- 1} u and u = A^{- 1} e, the equality A = B_{0}^{- 1}$ follows immediately.

Alternatively, note that the matrices B₀ and D cannot have more unknowns than Ω. In this case, since D has n parameters (it is diagonal) and Ω has n(n+1)/2 parameters (it is symmetric), this constrains B₀ to have at most n(n-1)/2 free parameters.

The standard errors of the impulse responses are calculated by Monte Carlo simulation. They are broadly similar to the probability bands are calculated from a Bayesian method that employs a Gaussian approximation to the posterior of the matrix A (recommended by Sims and Zha (1999) for overidentified models).

Other Publishers

Save
Cite
Email this content

Share Link

Copy this link, or click below to email it to a friend
Email this content
or copy the link directly:

https://www.elibrary.imf.org/view/journals/002/2008/281/article-A002-en.xml
The link was not copied. Your current browser may not support copying via this button.

Link copied successfully