Financial stress, economic activity

and monetary policy in the

ASEAN-5 economies

Boon Hwa Tnga,* and Kian Teng Kwekb

aBank Negara Malaysia, Jalan Dato’ Onn, 50480 Kuala Lumpur,MalaysiabFaculty of Economics and Administration, University of Malaya, 50603Kuala Lumpur, Malaysia

This article uses a structural vector autoregression approach to analyse theimpact of financial stress on the economy and the relationship betweenmonetary policy and financial stress in the ASEAN-5 economies(Indonesia, Malaysia, Philippines, Singapore and Thailand). We find thatan increase in financial stress leads to tighter credit conditions and lowereconomic activity in all five countries. The estimated impact on the realeconomy displays an initial rapid decline followed by a gradualdissipation. In Malaysia, the Philippines and Thailand, the central bankstend to reduce policy interest rates (IRs) when financial stress increases,although there is substantial cross-country variation in the magnitude andtime dynamics. The lower policy IRs are found to have little significanteffects in lowering financial stress, but are still effective in stimulatingeconomic activity through other channels. These findings imply thateasing monetary policy is likely necessary but insufficient to addressgrowth slowdowns associated with financial stress. Monetary easingshould instead be complemented with other policy measures which aretargeted at restoring financial stress to normal levels.

Keywords: financial stress; monetary policy; small open economy

JEL Classification: E44; E50; E58

I. Introduction

There has been a resurgence of interest in the inter-connections between macroeconomic and financialstability since the 2008 Global Financial Crisis(GFC) and euro debt crisis. This is no doubt attribu-table to the large scale and depth of the crises. The

large adverse growth effects in economies at theepicentre of these crises were consistent with earliercrisis experiences (Reinhart and Rogoff, 2009). Inother economies, particularly small-open economies,these episodes were stark reminders that their growthand financial stability prospects are susceptible toboth domestic imbalances and external spillovers.

*Corresponding author. E-mail: boonhwa@bnm.gov.my

Applied Economics, 2015Vol. 47, No. 48, 5169–5185, http://dx.doi.org/10.1080/00036846.2015.1044646

© 2015 Taylor & Francis 5169

Most of such economies were significantly affectedthrough weak exports and the financial spillovers.Although the policy responses during the GFC

period were tailored largely to country-specificconditions, a notable observation was the moveby central banks globally to reduce their policyinterest rates (IRs) during this period. This heldtrue irrespective of the respective central banks’monetary policy mandates (inflation targeting ornot), which raises the questions of whether mone-tary policy was influenced directly by the financialstress and if monetary policy was effective infacilitating the economic recovery.In this context, this article aims to shed light on

these issues for five small-open economies in Asia –Indonesia, Malaysia, Philippines, Singapore andThailand (henceforth, ASEAN-5). Using a structuralvector autoregression (SVAR) approach, we attemptto give insight to four questions: first, what is theimpact of financial stress on economic activity?While the spillover to growth from lower exports iswell understood, relatively less is known of thegrowth effects from the financial spillovers. This isespecially true for economies with few past inci-dences of financial crisis, such as the ASEAN-5, toinfer the growth effects from. Second, does monetarypolicy respond systematically to increases in finan-cial stress? Third, is monetary policy effective inalleviating financial stress? Finally, do changes infinancial stress levels alter the transmission of finan-cial stress to the real economy?The model builds from the existing open-economy

VAR literature by using financial stress indices(FSIs), continuous indicators of stress in financialmarkets, to reflect financial cycles in global financialmarkets and in the ASEAN-5 economies. Throughthe FSIs, the VAR models capture in a parsimoniousmanner distinct features of financial episodes, such asunderlying risk appetite and uncertainty. Using theFSIs offers two advantages: first, they facilitateanalysis of macro-financial linkages during tranquiland stressful periods in financial markets, as they arecontinuous measures of financial stress. FSIs are thususeful for analysing issues pertaining to the financialcycle in countries with few historical incidences ofsevere financial episodes. Second, the FSIs summa-rize financial conditions in all major asset markets,hence sidestepping potential pitfalls from analysingspillovers within specific asset markets.

We find that an increase in financial stress leads totighter credit conditions and lower economic activityin all five countries. The estimated impact on the realeconomy displays an initial rapid decline followed bya gradual dissipation. In Malaysia, the Philippinesand Thailand, the central banks reduced policy IRswhen financial stress increases, although there issubstantial cross-country variation in the magnitudeand time dynamics. The lower policy IRs are foundto have little significant effects in lowering financialstress, but are still effective in stimulating economicactivity through other channels. Overall, this result isconsistent with these central banks acting to achievemacroeconomic stability, as lower policy IRs act tooffset the contractionary effects of higher financialstress on economic activity.The remaining article is organized as follows.

Section II reviews the transmission channels offinancial stress to the real economy and the relatedmonetary policy issues. The SVAR model and dataused are detailed in Section III. Section IV presentsthe results. Section VI concludes.

II. How Financial Stress Affects EconomicActivity

Access to financing

A main channel in which financial stress affects realeconomic activity is through access to financing.Higher financial stress can lead to lower access tofinancing by firms and households as the economicoutlook deteriorates and asset prices decline. Thisoccurs through several mechanisms. From bor-rowers’ perspective, the financial acceleratormechanism posits that external finance premiums1

increase when an adverse financial shock leads to adecline in net worth as asset prices fall and theeconomic outlook deteriorates (Bernanke andGertler, 1989; Garber and Grilli, 1989). This happensbecause lenders perceive investments as more riskyand have lower expected profits. The higher cost offunds then reduces access to desired financing andcauses a decline in spending that is more persistentcompared to the size of the initial shock. Meanwhile,the bank capital and bank lending channels empha-size the role of lenders. Adverse financial shockserode banks’ capital base through lower profits,

1Defined as the difference in cost of financing an investment between internally and externally sourced funds.

5170 B. H. Tng and K. T. Kwek

losses on existing loans and other assets on theirbalance sheets, forcing them to reduce lending(Bernanke and Blinder, 1992; Kashyap and Stein,1995; Van Den Heuvel, 2002). This leads firmsto reduce capital expenditures and households toreduce spending.2

In equity markets, the Tobin’s q mechanismdepicts how financial stress affects the cost of equityand suppresses economic activity (Tobin, 1969). Thismechanism establishes a positive link between equityprices and capital investments by relating the marketvalue of firms to the replacement cost of capitalgoods. Since equity prices decline during stressepisodes, the market value of firms relative to theircost of capital goods also declines. Firms thereforeneed to issue more equity relative to periods whentheir market value is higher. This depresses fundraising in equity markets and leads to a decline ininvestment expenditure.

Uncertainty

Financial stress is also transmitted to the economythrough higher uncertainty in financial markets andthe economic outlook. Bloom (2009) studies thetransmission of uncertainty through a reduced formVAR model and a structural firm-level model ofinvestment. Firms hire and invest when businessconditions are above a certain level and fire anddisinvest when business conditions are below athreshold. There is a range of business conditionswhere firms find it optimal to take no action. Thisregion of inaction increases with the level ofuncertainty. Bloom (2009) finds a sharp fall, arebound and an overshoot in employment, outputand productivity.3 He explains that hiring and invest-ment initially fall rapidly as firms hold back onplanned projects and adopt a wait-and-see approach.Lower employment and investment by higherproductivity firms then cause a fall in productivity.As the uncertainty dissipates, firms react to pent-updemand for capital and labour, hence causing anovershoot in investment, employment and

productivity. Consumer spending is also affected byuncertainty, as consumers delay spending amiduncertain employment and wealth statuses. Leeet al. (2010) estimate a three-variable VAR andfinds that higher uncertainty leads to a hump-shapeddecline in household wealth and consumptionover approximately 2 years. Carrière-Swallow andCéspedes (2013) analyse the impact of uncertaintyshocks on investment and private consumption indeveloped and emerging markets using a VARmodel. The authors find notable differences betweendeveloped and emerging economies. In developedeconomies, they find that investment displays asimilar dynamic as Bloom (2009). However, theresponse of investment in emerging economies islarger and there is no subsequent overshoot. Forprivate consumption, the authors find that the impactin emerging economies is larger compared todeveloped economies.

The relationship between monetary policy andfinancial stress

What is the role of monetary policy when financialstress increases and the real economy slows? There isno conceptual agreement yet on whether a monetarypolicy regime that best promotes price and outputstability should respond to financial stability. Thequestion of whether financial factors should enterthe monetary policy reaction function is still beingdebated.4

One literature analyses the desirability for mone-tary policy to respond to asset prices and creditthrough dynamic stochastic general equilibrium(DSGE) models. In a DSGE model with equitymarket cycles, Bernanke and Gertler (1999, 2001)find that a monetary policy rule based on inflationtargeting is optimal for stabilizing inflation and out-put. This arises because stock market booms lead tostronger demand and higher inflation. It is thereforesufficient to consider the inflation forecast alone toset monetary policy once the informational content ofasset prices in predicting inflation is incorporated.5

2 See Dell’Ariccia et al. (2008) and Mendoza and Terrones (2008) for other selected examples of empirical studies thataddress the relationship between credit and real economy.3 For instance, industrial production falls rapidly for 4 months, rebounds after 7 months and subsequently overshoots beforeits effects gradually dissipates approximately 3 years after the uncertainty shock.4 See Baxa et al. (2013) for a more extensive review of this literature.5Despite their strong stance against systematic reactions to asset prices, Bernanke and Gertler (2001) caveat that this doesnot preclude short-term monetary policy interventions during periods of financial instability.

Financial stress, economic activity and monetary policy in the ASEAN-5 5171

Cecchetti et al. (2000) find, in contrast, that it isoptimal for central banks to include equity prices intheir policy reaction function. A key departure in theunderlying assumptions from Bernanke and Gertler(1999, 2001) is that the central bank has informationon whether the equity prices are driven by funda-mentals and the timing of the bubble burst. Morerecently, Christiano et al. (2010) find that there arewelfare gains from expanding the Taylor rule toinclude credit. Cúrdia and Woodford (2010) analysethe benefits of adding credit and credit spreads to theTaylor rule. They show that there are economicbenefits to augmenting the Taylor rule with creditspreads and, to a smaller extent, credit as well.One of the highlighted pitfalls of a monetary

policy approach that responds only to inflation isthat past experiences reveal that asset price boomsare not always inflationary. This is pointed out,among many others, by Borio and Lowe (2002),Bordo and Wheelock (2004) and Christiano et al.(2010). For example, Borio and Lowe (2002) findthree stylized features of financial imbalances – rapidasset price increases, fast credit expansions andabove average capital accumulation. The authorsalso provide evidence from a large number of finan-cial crises that inflation does not systematicallyincrease during the build-up to financial crises orunwinding of lending booms, but are deflationarythereafter. This feature induces an asymmetryamong the financial cycle, inflation and monetarypolicy. Specifically, monetary policy staysunchanged during the build-up of financial im-balances because there is no inflation, but is loosenedaggressively after the onset of the crisis due to defla-tionary pressures. A major pitfall is that becausemonetary policy was not tightened earlier in thefinancial cycle, there is subsequently less space inhow much monetary easing the central bank can do,at least in its conventional instrument. Borio andLowe (2002, 2004) thus advocate explicit considera-tion of financial imbalances in setting monetarypolicy.Despite the lack of intellectual consensus, there is

evidence that many central banks do respond tofinancial factors in practice. A survey of over 90central banks in both advanced and emerging econo-mies revealed a significant correlation betweenmonetary policy and financial stability concerns,including financial sector solvency, credit rationingand asset price volatility (Mahadeva and Sterne,

2000). Studies have also estimated the monetarypolicy reaction functions of central banks to searchfor indications of explicit attention to financial fac-tors. Borio and Lowe (2004) estimate severalpermutations of the monetary policy reaction func-tions for the United States, Germany, Australia andJapan. They start with a standard Taylor rule specifi-cation and gradually add three measures of financialimbalances – the credit gap, equity price gap and adummy variable capturing banking sector stress.Their results reflect variations in the reaction func-tions across countries. The German central bank paidlittle attention to financial imbalances in its monetarypolicy decisions. In Australia, the equity and creditgaps were jointly significant predictors of monetarypolicy movements. In Japan, there is evidence thatmonetary policy responded asymmetrically to creditand equity gaps, more when the gaps were negative.In the United States, the study also finds evidencethat the Federal Reserve responded asymmetricallyto financial imbalances. Policy IRs are more respon-sive to negative credit and equity gaps than positivegaps. More recently, Baxa et al. (2013) test thesignificance of financial stress in IR decisions usinga time-varying specification of monetary policy infive advanced economies (United States, UnitedKingdom, Australia, Canada and Sweden). Theauthors find that central banks were unresponsive tofinancial stress at low and normal levels, but ofteneased their policy rates in response to higher financialstress, in particular, to equity and bank-related finan-cial stress.

III. Methodology

In this study, we use an SVAR approach to assessthe impact of financial stress on the economy andthe relationship between financial stress and mone-tary policy. This modelling approach draws frommore recent efforts to study linkages betweenfinancial conditions and economic activity usingFSIs and VAR models. Representative studies areLi and St-Amant (2010), Davig and Hakkio (2010),Hollo et al. (2012), Mallick and Sousa (2013) andRoye (2011). Although the indices used vary withstudies, all reflect stress in financial marketsthrough declining and volatile asset prices, andhigher bond yields/spreads. The existing analyses

5172 B. H. Tng and K. T. Kwek

have thus far tended to focus on developed econo-mies, particularly euro area economies and theUnited States, which is unsurprising given therecent financial episodes there. We contribute tothis literature by adapting the model structure tobe more suited for small-open economies, byincluding external variables to account for largeexposures to the foreign environment.

Data

The sample consists of the ASEAN-5 countries ofIndonesia, Malaysia, Philippines, Singapore andThailand. The variables are in monthly frequencyand range from January 1997 to December 2013. Asummary of the variables is presented in Table 1.Three variables characterize the external environ-

ment: a global commodity price index (GCP), aworld industrial production index (IPIw) and a FSIfor the US economy (FSIus). GCP captures globalprices of food, fuel and metal commodities. IPIwcaptures global real economic conditions. Thisglobal measure is preferred to the more commonlyused US focused indicator, as it abstracts from tradediversification away from the United States. Inaddition, focusing on US demand alone risks mis-identification of commodity price shocks, as com-modity price movements are increasingly beingattributed to demand from emerging markets, suchas China. The final external variable is an index offinancial stress for the US economy, FSIus, whichproxies for global financial conditions. To be sure,

financial episodes occur in other countries as well,especially in emerging markets. However, Kaminskyand Reinhart (2003) find that financial episodes tendto remain confined within their regions unless theyspread to major financial centres. This suggests thatASEAN-5 financial markets are unaffected by finan-cial episodes that originate outside the region and thatfinancial spillovers to the region only occur whenmajor financial markets are affected. Therefore, wedo not attempt to measure global financial stress andassume that US financial stress aptly reflect globalfinancial conditions.Six variables characterize the domestic environ-

ment: the industrial production index (IPI) capturesreal economic activity; the consumer price index(CPI) reflects the price level; the short-term IR isthe monetary policy instrument in Indonesia,Malaysia, Philippines and Thailand, and a floatingshort-term money market IR in Singapore; credit (C)is claims from the domestic banking system, and theexchange rate (EX) is the nominal effective exchangerate. The last variable, an index of financial stress(FSI), is a summary indicator of stress in financialmarkets from Tng et al. (2012).6 This indexcomprises stress indicators in four segments ofdomestic financial markets: the banking sector,foreign exchange market, bond market and equitymarket. The market-specific indicators of stress areweighted to their markets’ relative sizes, as reflectedby the amount of outstanding financing in the marketsegments. The ASEAN-5 and US FSIs are shown inFig. 1.

Table 1. Summary of variables used in the estimations

Variable Abbreviation Definition Source

ExternalCommodity prices GCP Commodity price index (sa, log) International Monetary FundWorld output IPIW World industrial production index

(sa, log)CPB Netherlands Bureau for EconomicPolicy Analysis

US financial stress FSIUS US Financial stress index Hakkio and Keeton (2009)

DomesticOutput IPI Industrial production index (sa, log) International Financial Statistics (IFS)Prices CPI Consumer price index (sa, log) IFSInterest rate IR Short-term interest rate IFSCredit C Bank credit, deflated by CPI (sa, log) IFSExchange rate EX Nominal effective exchange rate (log) Bank for International SettlementsFinancial stress FSI Financial stress index Tng et al. (2012)

Notes: ‘sa’ refers to seasonally adjusted. ‘Log’ refers to the natural log.

6 The only difference is the weights are now updated every quarter instead of annually.

Financial stress, economic activity and monetary policy in the ASEAN-5 5173

The variables are standardized prior to aggrega-tion. Avalue of 0 reflects neutral financial conditions,high values reflect stress and low values reflect buoy-ance in financial markets. The FSIs indicate thathigher stress in ASEAN-5 is observed during threeperiods. In order of severity, they are the AsianFinancial Crisis (AFC) (1997–1998), the technologybubble burst in the US (2000–2001) and the recentGFC (2008–2009). The FSIs suggest that, for theASEAN-5, the latter two episodes originated exter-nally while the AFC was a domestic and regionalepisode.7

The SVAR model8

A schematic summarizing the assumptions of theSVAR model is illustrated in Fig. 2. Domestic outputand prices are influenced by two groups of variables:the first is the external environment, consisting ofcommodity prices, world demand and global finan-cial conditions. The second group characterizes

domestic financial markets with a short-term IR, theexchange rate, credit and the FSI. The sample coun-tries are modelled as small-open economies and arethus affected by but do not affect external conditions.The external variables directly affect domestic outputand prices through trade and price channels andindirectly through domestic financial markets.

External Environment Commodity prices Global output External financial stress

Domestic Economy Output and Prices

Domestic Financial Market Interest rate Exchange rate Credit Domestic financial stress

Fig. 2. Schematic illustration of causality assump-tions in the VAR model

–2

0

2

4Indonesia

–2

0

2

4Philippines

–2

0

2

4

January1997

January2000

January2003

January2006

January2009

January 2012

January1997

January2000

January2003

January2006

January2009

January 2012

January1997

January2000

January2003

January2006

January2009

January 2012

January1997

January2000

January2003

January2006

January2009

January 2012

January1997

January2000

January2003

January2006

January2009

January 2012

January1997

January2000

January2003

January2006

January2009

January 2012

Thailand

–2

0

2

4Malaysia

–2

0

2

4Singapore

–2

3

8United States

Fig. 1. Financial stress in the ASEAN-5 economies (1997–2013)Source: Authors’ calculation and Hakkio and Keeton (2009).

7 See Tng et al. (2012) for a discussion of financial stress in the ASEAN-5 economies during these three financial episodes.8 This model also applied in Tng (2013) to analyse the exposure of the ASEAN-5 economies to external shocks. Hence, thedescription of the VAR model is largely similar to that article.

5174 B. H. Tng and K. T. Kwek

External conditions may influence monetary policy.This in turn affects domestic financial conditions,output and prices. External conditions also influencethe exchange rate and domestic asset prices throughcross-border capital flows. This consequently affectsthe terms of trade, wealth and financing conditions,which in turn affect output and prices. The financialaccelerator mechanism may also amplify the direct

effects of external shocks through a feedback effectfrom interactions between the real economy andfinancial markets. For instance, when faced with anadverse external demand shock, lower profits anddeteriorating balance sheet positions of export-oriented companies may cause an increase in borrow-ing premiums and lower access to financing. Thisleads to moderating investment and credit-financedtrade.To characterize these channels, consider the

following SVAR model for each economy:

AXt ¼ B Lð ÞXt�1 þ εt (1)

X is a vector of variables with a similar ordering asTable 1. A is a matrix of contemporaneous coeffi-cients in structural form. B Lð Þ is a matrix polynomialin the lag operator, L. εt is a vector of structuraldisturbances, such that

εt ¼ Aet (2)

et is a vector of residuals from the correspondingreduced-form VAR. The equations can be orga-nized into external and domestic blocks. Structuralshocks are identified using the approach suggestedby Sims (1986), Bernanke (1986) and applied bymany others thereafter, by placing restrictions onthe contemporaneous coefficients. The assumptionsmade on A are

Commodity prices are contemporaneously exo-genous to all other variables. World output and USfinancial stress are identified recursively by assumingthe former is contemporaneously affected by com-modity prices, while US financial stress is contem-poraneously affected by commodity prices and worldindustrial production. The external variables are con-temporaneously unaffected by the country-specificvariables. The first four variables in the domesticblock are ordered recursively in the following order– IPI, CPI, IR, C, EX and FSI. The short-term IRbroadly follows a ‘Taylor rule’ principle, as it reactscontemporaneously to economic activity (IPI) andprices (CPI).9 The exchange rate is ordered beforefinancial stress to model the narrative that a financialshock may trigger capital outflows and affect theexchange rate with a lag.Block-exogeneity restrictions are also imposed on

the domestic variables in the external equations toimpose the small-open economy assumption. Thismeans that the external variables affect each other inlags, but are unaffected by the domestic variables bothin lags and contemporaneously. This approach follows

εGCPεIPIwεFSIusεIPIεCPIεIRεCεExεFSI

26666666666664

37777777777775

¼

a11 0 0 0 0 0 0 0 0a21 a22 0 0 0 0 0 0 0a31 a32 a33 0 0 0 0 0 0a41 a42 a43 a44 0 0 0 0 0a51 a52 a53 a54 a55 0 0 0 0a61 a62 a63 a64 a65 a66 0 0 0a71 a72 a73 a74 a75 a76 a77 0 0a81 a82 a83 a84 a85 a86 a87 a88 0a91 a92 a93 a94 a95 a96 a97 a98 a99

26666666666664

37777777777775

eGCPeIPIweFSIuseIPIeCPIeIReCeExeFSI

26666666666664

37777777777775

(3)

9This reaction function is not exactly the same as the one originally suggested in Taylor (1993) as other variables enter thefunction in lags.

Financial stress, economic activity and monetary policy in the ASEAN-5 5175

from Cushman and Zha (1997), Maćkowiak (2007),Genberg (2005) and Raghavan et al. (2012). Theblock-exogeneity restrictions translate to the coeffi-cient matrix for the lag structure, Bi, where i representsthe lags, with the variables ordered similar to Table 1:

B ¼

b11 b12 b13 0 0 0 0 0 0b21 b22 b23 0 0 0 0 0 0b31 b32 b33 0 0 0 0 0 0b41 b42 b43 b44 b45 b46 b47 b48 b49b51 b52 b53 b54 b55 b56 b57 b58 b59b61 b62 b63 b64 b65 b66 b67 b68 b69b71 b72 b73 b74 b75 b76 b77 b78 b79b81 b82 b83 b84 b85 b86 b87 b88 b89b91 b92 b93 b94 b95 b96 b97 b98 b99

26666666666664

37777777777775

(4)

The estimations are carried out using four lags.Table 2 presents results from the AkaikeInformation Criterion (AIC) and the SchwarzInformation Criterion (SIC), with six lags set as themaximum length given the relatively short sample.The AIC chose a longer lag length with a wide rangefrom 3 to 6, while the SIC selected one lag for allcountries. Given these differing results, we use thesetests as guides rather than a hard-and-fast rule andchoose the average of the AIC lags of 4.

Estimation issues

Two issues arise from the estimation of the SVARmodel. The first is addressing potential structuralbreaks during the AFC period. Many studies tacklethis by splitting their sample into pre-AFC and post-AFC subsamples.10 In this study, the sample period

before the AFC is relatively short since the data startfrom 1997. This partially mitigates the need to differ-entiate pre- and post-AFC conditions. In addition, abenefit to utilizing the full sample is that it includesthe AFC episode, which for the ASEAN-5 economiesis the only major domestic financial episode to haveoccurred during the sample period. Having this var-iation in domestic financial stress during the AFC isimportant to differentiate domestic and foreign finan-cial shocks, and hence their impact on the economies.In addition, a disadvantage of using a post-AFC sub-sample is that the smaller sample results in lowerefficiency. We attempt to balance these trade-offs byusing the full sample to estimate the interactionsamong financial stress (FSI), real economic activity(IPI) and policy IRs, but use data from 2000 onwardswhen studying the impact of monetary policy shockson the real economy. The former is done since havingthe AFC episode is important to properly identifyfinancial shocks, while the latter is done to avoidwell-known instabilities in the monetary policy reac-tion function during the AFC period.Another issue is whether to difference or de-trend

the nonstationary variables. Many studies, followingSims (1980), Sims et al. (1990) and Ramaswamy andSloek (1997), estimate their VARs with nonstation-ary variables in levels under the premise that theirinterest is not in the parameter estimates but rather inthe interrelationships.11 Moreover, parameter esti-mates are usually not focused on in VARs sincethey are often over-parameterized. These studiesinstead analyse the time dynamics of interest fromthe impulse response functions. We choose to esti-mate the VAR models in levels. While acknowled-ging the pitfalls associated with estimating VARs inlevels, this is nonetheless a revealing way to examinethe interrelationships of interest.

IV. Estimation Results

This section presents the results from the SVARmodel. The impulse responses are plotted over 60months with the 95th percentile confidenceintervals.12

Table 2. Lag length selection from Akaike Informationand Schwarz Criterions

Akaike InformationCriterion

SchwarzCriterion

Indonesia 6 1Malaysia 3 1Philippines 4 1Singapore 3 1Thailand 4 1

10 See Fung (2002), Disyatat and Vongsinsirikul (2003), Hesse (2007) and Raghavan et al. (2012) for some referenceswithin the VAR literature.11More recent studies in this vein are Disyatat and Vongsinsirikul (2003) and Raghavan et al. (2012).12 The bootstrap methodology applied is from Hall (1992) using 100 replications. Increasing the number of replications to500 does not materially change the results.

5176 B. H. Tng and K. T. Kwek

The impact of financial stress

Figure 3 illustrates the impulse responses of indus-trial production to a 1 SD unexpected increase infinancial stress.The impulse responses show that higher finan-

cial stress leads to a decline in output. A similarityin the output responses across countries is that thedeclines are initially rapid and followed by a moregradual dissipation. Most of the contractionaryeffects occur within the first year after the shockwith a majority of the effects dissipating by thesecond year. There is nonetheless some cross-country heterogeneity in the time dynamics. InIndonesia and Malaysia, there is a subsequentovershoot in IPI, which is indicative of the pre-sence of an uncertainty channel in which firmssubsequently react to pent-up demand for capitaland labour. The response for the Philippines is themost persistent, with the largest effects feltapproximately 2 years after the shock, followedby dissipation over the subsequent 3 years. IPI inSingapore and Thailand recover quickly with theirIPIs returning to baseline levels approximately 1year after the shock. In general, the time dynamics– a sharp drop and gradual dissipation – are con-sistent with results from other similar studies, for

instance Davig and Hakkio (2010) for the USeconomy and Holo et al. (2012) for euro areaeconomies.As alluded to in Section II, a main conduit in

which financial stress causes a reduction in economicactivity is through lower access to financing frombanks. Figure 4 gives evidence of this channel byillustrating the impulse responses of real credit to a 1SD increase in financial stress. Real credit declines inall cases. Similar to the previous impulse responses,the initial declines in real credit are the sharpestduring the first year after the shock, which is thenfollowed by a more gradual dissipation. While dete-riorating credit conditions contribute to moderatingeconomic activity as financial stress increase, onefactor that may limit the downward pressure on thereal economy is if higher financial stress causeshigher cross-border capital outflows and depreciationin the exchange rate, which then stimulates the exportsector. Figure 5 tries to provide some insight intohow exchange rates tend to move when financialstress increases. The results display substantialcross-country heterogeneity. Exchange rate deprecia-tion is observed in Indonesia, the Philippines andThailand, albeit with differing time dynamics. InMalaysia, there is initially a depreciating effect

–0.015

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Indonesia

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Malaysia

–0.010

–0.005

0.000

0.005

0.010

0 10 20 30 40 50 60

Philippines

–0.020

–0.015

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Singapore

–0.030

–0.020

–0.010

0.000

0.0100.010

0 10 20 30 40 50 60

Thailand

Fig. 3. Response of IPI to a financial stress shockSource: Authors’ estimates.

Financial stress, economic activity and monetary policy in the ASEAN-5 5177

–0.030

–0.020

–0.010

0.000

0.010

0 10 20 30 40 50 60

Indonesia

–0.008

–0.006

–0.004

–0.002

0.000

0.002

0.004

0 10 20 30 40 50 60

Malaysia

–0.010

–0.005

0.000

0.005

0.010

0 10 20 30 40 50 60

Philippines

–0.008

–0.006

–0.004

–0.002

0.000

0.002

0.004

0 10 20 30 40 50 60

Singapore

–0.008

–0.006

–0.004

–0.002

0.000

0.002

0.004

0 10 20 30 40 50 60

Thailand

Fig. 4. Response of real credit to a financial stress shockSource: Authors’ estimates.

–0.060

–0.040

–0.020

0.000

0.020

0.040

0 10 20 30 40 50 60

Indonesia

–0.010

–0.005

0.000

0.005

0.010

0 10 20 30 40 50 60

Malaysia

–0.015

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Philippines

–0.002

–0.001

0.000

0.001

0.002

0.003

0 10 20 30 40 50 60

Singapore

–0.015

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Thailand

Fig. 5. Response of NEER to a financial stress shockSource: Authors’ estimates.

5178 B. H. Tng and K. T. Kwek

followed by temporary appreciation. The estimatedeffect for Singapore is both economically and statis-tically insignificant from 0.13

Overall, the impulse responses indicate that finan-cial stress has negative effects on real economicactivity. It, nonetheless, begs the question of its over-all influence on economic activity. Financial shocksmay have significant negative effects on domesticoutput, but explain only a small fraction of the totalvariation in output if they occur infrequently. Wenext analyse the variance decomposition of IPI toderive the contribution of financial stress to the realeconomy. The decomposition results at the 24- and36-month horizons are presented in Table 3. As pre-viously suggested, the contributions from domesticfinancial stress (FSI) to real economic activity (IPI)are relatively small. Aside from Indonesia as an out-lier with the highest contribution of 39%, the con-tributions in the other four sample countries arebelow 5%. This indicates that, at least for theASEAN-5 economies, financial stress events havehistorically been tail risks to real economic activity.

But such events have significant adverse effectswhen they occur. Meanwhile, a large amount of thevariation in output can be attributed to external fac-tors, which account for an average of 54% and 60%of the total variation in output at the 24- and 36-month horizons. The high external contributionsvalidate the importance of including the foreign vari-ables in VAR models when analysing openeconomies.14

Monetary policy under financial stress

Do the ASEAN-5 central banks alter their monetarypolicy stance when financial stress increases? Is iteffective? We now explore the two-way interactionbetween monetary policy and financial stress. Weexclude Singapore from this analysis because theexchange rate instead of an IR is used to conductmonetary policy. The results for Singapore are there-fore not comparable with the other economies, due todifferences in the policy instrument and identificationof monetary policy shocks in the SVAR.

Table 3. Decomposition of the forecast error variance of output (%)

GCP IPIW FSIUS IPI CPI IR C EX FSI

24 months

Indonesia 1 3 3 25 2 4 7 17 39Malaysia 6 18 49 14 5 1 2 2 4Philippines 34 18 6 25 2 0 10 2 3Singapore 5 39 24 28 0 0 1 1 1Thailand 6 42 17 31 0 0 1 0 2

36 months

Indonesia 1 4 3 23 2 4 13 16 35Malaysia 4 25 49 10 4 0 1 2 3Philippines 37 14 13 18 1 0 8 4 4Singapore 4 44 27 21 0 0 1 1 1Thailand 5 52 16 23 0 0 2 0 1

Source: Authors’ estimates.

13 The large variations in exchange rate responses across countries likely reflect differences in both institutions and policyregimes that are beyond the intended scope of this study. In Singapore, the NEER serves as the monetary policy instrument,indicating essentially that the Monetary Authority of Singapore’s monetary policy stance does not systematically respondto changes in financial stress. In Malaysia, the central bank intervenes to reduce exchange rate volatility. This may explainwhy the depreciation is temporary – for example, upon experiencing sudden capital outflows and exchange rate deprecia-tion as financial stress increases, the central bank intervenes to limit the abrupt exchange rate depreciation and hencereduces the overall exchange rate volatility associated with capital flow movements. Malaysia’s exchange rate dynamics islikely also influenced by changes in the exchange rate regime during the sample period.14 See Tng (2013) for an analysis of the impact of external shocks on output and inflation using the same VAR model andsample.

Financial stress, economic activity and monetary policy in the ASEAN-5 5179

Figure 6 analyses monetary policy behaviourwhen financial stress increases,15 by illustrating theimpulse response of IRs to a 1 SD increase in finan-cial stress. The impulse responses show that the IRsin Malaysia and the Philippines are lowered whenfinancial stress increase.16 Their IRs decline the mostduring the first year after the financial shock. InThailand, the IR displays an initial spike, followedby an easing trajectory similar to Malaysia and thePhilippines. To see if the initial IR spike in Thailand’scase is attributable to the brief period of high IRpolicy during the AFC, we also show the impulseresponse function from the SVAR model estimatedfrom 2000 onwards in Fig. 6. The results show thatremoving the AFC period from the sample eliminatesthe initial spike in the IR, suggesting that the spike isindeed a reflection of monetary policy tighteningduring the AFC period. In Indonesia, the IR initiallyincreases as well. Unlike Thailand, the initialincrease in Indonesia’s IR lasts for a longer durationand does not disappear when the AFC episode isremoved from the sample. However, the magnitude

of the increase becomes largely statistically insignif-icant from zero.A natural follow-up question is whether monetary

policy influences financial stress levels. Figure 7 pro-vides an indication through the impulse responses offinancial stress to IR shocks. The responses of finan-cial stress are heterogeneous across countries, and areoften small and statistically insignificant. Thisreflects a limit in the use of monetary policy toalleviate financial stress and that direct financial sec-tor intervention is likely necessary to restore financialstability during crisis periods. This result, however, isnot a case against monetary policy easing duringperiods of higher financial stress. As shown earlier,higher financial stress adversely affects economicactivity and central banks may still use monetarypolicy to restore macroeconomic (output) stability.A key premise is that lower IRs stimulate output notby restoring financial stability, but through otherchannels.To give insight to this hypothesis, we attempt to

distinguish the effects of IRs on output that is

–2.000

–1.000

0.000

1.000

2.000

3.000

0 10 20 30 40 50 60

Indonesia

–0.150

–0.100

–0.050

0.000

0.050

0.100

0 10 20 30 40 50 60

Malaysia

–0.300

–0.200

–0.100

0.000

0.100

0.200

0 10 20 30 40 50 60

Philippines

–0.400

–0.200

0.000

0.200

0.400

0 10 20 30 40 50 60

Thailand

–0.150

–0.100

–0.050

0.000

0.050

0.100

0.150

0 10 20 30 40 50 60

Indonesia(Sub-sample: '00-'13)

–0.100

–0.050

0.000

0.050

0.100

0 10 20 30 40 50 60

Thailand(Sub-sample: '00-'13)

Fig. 6. Response of interest rate to a financial stress shockSource: Authors’ estimates.

15 Singapore is excluded from this analysis because its central bank uses the exchange rate instead of a short-term interestrate as its policy instrument to conduct monetary policy. The results for Singapore are therefore not comparable with theother economies, due to differences in the policy instrument and identification of monetary policy shocks in the SVAR.16The initial spike in Malaysia’s case is small and statistically insignificant and is thus discounted for inference.

5180 B. H. Tng and K. T. Kwek

attributable to domestic financial stress as a transmis-sion channel. We achieve this by comparing theimpulse response functions from the baseline modelto those from a restricted model. The restricted modelis similar to the baseline model, except that domesticfinancial stress is exogenous. Doing so blocks off theresponses of output to a change in the IR that passesthrough financial stress. The differences in impulseresponses between the baseline and restricted SVARsreflect the degree of pass-through via domestic finan-cial stress. This method of analysing the transmissionchannels of monetary policy follows from Morsinkand Bayoumi (2001), Chow (2004) and Raghavanet al. (2012). To avoid specification issues due towell-known instabilities in the monetary policy reac-tion function during the AFC period, the impulseresponses for this analysis are estimated using dataonly from 2000 onwards.Figure 8 shows impulse responses of IPI to IR

shocks from the baseline and restricted models. Inall cases, the impulse responses from both models arelargely similar and fall within the error bands fromthe baseline model. Thus, the analysis of monetarypolicy shows that although lowering IRs generallyhave limited effects in restoring financial stability,such policy moves are effective in stimulating

economic activity through other channels. Easingmonetary policy in the midst of financial episodes istherefore a desirable policy strategy to offset thecontractionary effects of higher financial stress onoutput.

V. Robustness

The assumptions made on the exogeneity of thedomestic variables in the foreign equations are intui-tive and common practice in existing literature. Assmall-open economies, it is reasonable to assume thatthey are affected by and cannot affect external devel-opments. It also seems reasonable to assume thatoutput and prices are affected by the financial vari-ables in lags, given that changing them are oftentime-consuming activities and entail additionalcosts. But the ordering of the FSI variable withinthe financial block is not as self-evident. Financialstress can have contemporaneous effects on theexchange rate since its value is partly determinedby cross-border capital flows which can react quicklyto changes in financial conditions. Monetary policymay also react contemporaneously to financial stress

–0.150

–0.100

–0.050

0.000

0.050

0.100

0 10 20 30 40 50 60

Indonesia

–0.050

0.000

0.050

0.100

0.150

0 10 20 30 40 50 60

Malaysia

–0.100

–0.050

0.000

0.050

0.100

0 10 20 30 40 50 60

Philippines

–0.150

–0.100

–0.050

0.000

0.050

0 10 20 30 40 50 60

Thailand

Fig. 7. Response of financial stress to an interest rate shockSource: Authors’ estimates.

Financial stress, economic activity and monetary policy in the ASEAN-5 5181

if central banks take it as a forward-looking signal ofmacroeconomic prospects. To test the sensitivity ofthe baseline findings, we estimate the SVAR modelswith alternative orderings of the FSI within the finan-cial block and replicate the impulse responses fromthe main findings of this article (Figs 3 and 8) – in theAppendix. The responses generated from alternativespecifications are broadly in line with the baselinemodel. The impact of FSI shocks on the IPI is alsobroadly similar. The initial fall in IPI is steep, fol-lowed by a gradual tapering off. Similarly, theresponses of IPI to IR shocks are robust to changesin the ordering of the FSI variable.

VI. Concluding Remarks

The goal of this article has been to use a SVARapproach to contribute to the understanding of howfinancial stress affects the economy and monetarypolicy transmission. We find that financial stresshas negative effects on real economic activity, creditand, in some cases, the exchange rate. Although thereis some heterogeneity in the responses, an empiricalregularity in the responses of output is that the largest

effects are felt within the first year of the shock.However, financial stress contributes a small shareof the overall variation in output, which is likelyattributable to the low frequency of high financialstress episodes. We also find that central banks inMalaysia, Thailand and the Philippines tend toreduce their policy IRs when financial stressincreases. Although lower IRs have mixed results intheir ability to reduce financial stress, they are stillable to stimulate economic activity through otherchannels. The estimations also find that althoughlower IRs have limited results in their ability toreduce financial stress, they are still effective in sti-mulating economic activity through other channels.More generally, these findings suggest the neces-

sity for monetary policy easing to help offset thecontractionary effects of adverse financial shockson the real economy. But it likely also needs to beaccompanied by direct financial sector interventionsto restore financial stability. This may include, forexample, short-term loans to alleviate liquidityshortages, direct equity injections to financial institu-tions to reduce solvency concerns and ensuring thesufficiency of trade credit to facilitate continued tradeactivities. In addition to achieving a higher

–0.008

–0.006

–0.004

–0.002

0.000

0.002

0 10 20 30 40 50 60

Indonesia

–0.010

–0.005

0.000

0.005

0 10 20 30 40 50 60

Malaysia

–0.010

–0.005

0.000

0.005

0.010

0 10 20 30 40 50 60

Philippines

–0.030

–0.020

–0.010

0.000

0.010

0.020

0 10 20 30 40 50 60

Thailand

Fig. 8. Response of IPI to an interest rate shockSource: Authors’ estimates.Notes: The blue line and dotted lines are the responses and error bands from the baseline model. The red line is the responsefrom the restricted model.

5182 B. H. Tng and K. T. Kwek

effectiveness in restoring financial stability, anotherbenefit of a targeted policy approach to restore finan-cial stability is that it reduces time-lag issues betweenthe policies’ effects on output and the effect thathigher financial stress has on output. While there ispotentially such a timing mismatch for monetarypolicy, policy instruments that directly restore finan-cial stress to normal levels reduce this pitfall.

AcknowledgementsThe views expressed here do not represent those ofBank Negara Malaysia. Earlier drafts were presentedat the Bank of Thailand and Bank for InternationalSettlements (BoT-BIS) 8th Annual Workshop of theAsian Research Networks 2015, the ISI RegionalStatistics Conference 2014 in Malaysia and thejoint meetings of the Econometric SocietyAustralasian Meeting (ESAM) and AustralianConference of Economists (ACE) 2014. This articlebenefitted from comments made by the conferenceparticipants. This article also benefitted from helpfulcomments from Mala Raghavan, Fraziali Ismail,Mohamad Hasni Sha'ari and Ahmad Othman. Allerrors and omissions are ours.

Disclosure StatementNo potential conflict of interest was reported by theauthors.

References

Baxa, J., Horváth, R. and Vašíček, B. (2013) Time-varyingmonetary-policy rules and financial stress: doesfinancial instability matter for monetary policy?,Journal of Financial Stability, 9, 117–38.doi:10.1016/j.jfs.2011.10.002

Bernanke, B. S. (1986) Alternative explanations of themoney-income correlation, Carnegie-RochesterConference Series on Public Policy, 25, 49–99.doi:10.1016/0167-2231(86)90037-0

Bernanke, B. S. and Blinder, A. (1992) The federal fundsrate and the channels of monetary transmission, TheAmerican Economic Review, 82, 901–21.

Bernanke, B. S. and Gertler, M. (1989) Agency costs, networth, and business fluctuations, The AmericanEconomic Review, 79, 14–31.

Bernanke, B. S. and Gertler, M. (1999) Monetary policyand asset price volatility, in Proceedings at theEconomic Policy Symposium, Federal Reserve Bankof Kansas City, Kansas City, MO, pp. 77–128.

Bernanke, B. S. and Gertler, M. (2001) Should centralbanks respond to movements in asset prices?, TheAmerican Economic Review, 91, 253–57.doi:10.1257/aer.91.2.253

Bloom, N. (2009) The impact of uncertainty shocks,Econometrica, 77, 623–85. doi:10.3982/ECTA6248

Bordo, M. D. and Wheelock, D. C. (2004) Monetarypolicy and asset prices: a look back at past USstock market booms, National Bureau of EconomicResearch Working Paper Series No. 10704, NationalBureau of Economic Research, Cambridge, MA.

Borio, C. and Lowe, P. (2002) Asset prices, financial andmonetary stability: exploring the nexus, BISWorking Papers No. 114, Bank for InternationalSettlements, Basel.

Borio, C. and Lowe, P. (2004) Securing sustainable pricestability: should credit come back from the wilder-ness?, BIS Working Papers No. 157, Bank forInternational Settlements, Basel.

Carrière-Swallow, Y. and Céspedes, L. F. (2013) Theimpact of uncertainty shocks in emerging economies,Journal of International Economics, 90, 316–25.doi:10.1016/j.jinteco.2013.03.003

Cecchetti, S., Genberg, H., Lipsky, J. et al. (2000) AssetPrices and Central Bank Policy, International Centrefor Monetary and Banking Studies, London.

Chow, H. K. (2004) AVAR analysis of Singapore’s mone-tary transmission mechanism, SMU Economics andStatistics Working Paper Series No. 19-2004,Singapore Management University, Singapore.

Christiano, L., Ilut, C. L., Motto, R. et al. (2010)Monetarypolicy and stock market booms, National Bureau ofEconomic Research Working Paper Series No.16402, National Bureau of Economic Research,Cambridge, MA.

Cúrdia, V. and Woodford, M. (2010) Credit spreads andmonetary policy, Journal of Money, Credit andBanking, 42, 3–35. doi:10.1111/jmcb.2010.42.issue-s1

Cushman, D. O. and Zha, T. (1997) Identifying monetarypolicy in a small open economy under flexibleexchange rates, Journal of Monetary Economics,39, 433–48. doi:10.1016/S0304-3932(97)00029-9

Davig, T. and Hakkio, C. (2010) What is the effect offinancial stress on economic activity, EconomicReview, Second Quarter 2010, Federal ReserveBank of Kansas City, Kansas City, MO.

Dell’Ariccia, G., Detragiache, E. and Rajan, R. (2008)The real effect of banking crises, Journal ofFinancial Intermediation, 17, 89–112. doi:10.1016/j.jfi.2007.06.001

Disyatat, P. and Vongsinsirikul, P. (2003) Monetary policyand the transmission mechanism in Thailand,Journal of Asian Economics, 14, 389–418.doi:10.1016/S1049-0078(03)00034-4

Fung, B. S. (2002) A VAR analysis of the effects ofmonetary policy in East Asia, BIS WorkingPapers No. 119, Bank for International Settlements,Basel.

Financial stress, economic activity and monetary policy in the ASEAN-5 5183

Garber, P. M. and Grilli, V. U. (1989) Bank runs in openeconomies and the international transmission ofpanics, Journal of International Economics, 27,165–75. doi:10.1016/0022-1996(89)90083-4

Genberg, H. (2005) External shocks, transmissionmechanisms and deflation in Asia, BIS WorkingPapers No. 187, Bank for International Settlements,Basel.

Hakkio, C. S. and Keeton, W. R. (2009) Financial stress:what is it, how can it be measured, and why does itmatter?, Economic Review, 94, 5–50.

Hall, P. (1992) The Bootstrap and Edgeworth Expansion,Springer, New York.

Hesse, H. (2007) Monetary policy, structural break and themonetary transmission mechanism in Thailand,Journal of Asian Economics, 18, 649–69.doi:10.1016/j.asieco.2007.03.009

Hollo, D., Kremer, M. and Duca, M. L. (2012) CISS – acomposite indicator of systemic stress in the financialsystem, ECB Working Paper No. 1426, EuropeanCentral Bank, Frankfurt.

Kaminsky, G. L. and Reinhart, C. (2003) The center andthe periphery: the globalization of financial turmoil,National Bureau of Economic Research WorkingPaper Series No. 9479, National Bureau ofEconomic Research, Cambridge, MA.

Kashyap, A. and Stein, J. (1995) The impact of monetarypolicy on bank balance sheets, Carnegie-RochesterConference Series on Public Policy, 42, 151–95.doi:10.1016/0167-2231(95)00032-U

Lee, J., Rabanal, P. and Sandri, D. (2010) US con-sumption after the crisis, IMF Staff PositionNotes No. 10/01, International Monetary Fund,Washington, DC.

Li, F. and St-Amant, P. (2010) Financial stress, monetarypolicy and economic activity, Bank of CanadaWorking Paper No. 2010-12, Bank of Canada,Ottawa, ON.

Maćkowiak, B. (2007) External shocks, U.S. monetarypolicy and macroeconomic fluctuations in emergingmarkets, Journal of Monetary Economics, 54, 2512–20.

Mahadeva, L. and Sterne, G. (Eds) (2000) MonetaryPolicy Frameworks in a Global Context, 1st edn,Routledge, London.

Mallick, S. K. and Sousa, R. M. (2013) The real effects offinancial stress in the Eurozone, International Reviewof Financial Analysis, 30, 1–17. doi:10.1016/j.irfa.2013.05.003

Mendoza, E. G. and Terrones, M. E. (2008) Ananatomy of credit booms: evidence from macro

aggregates and micro data, National Bureau ofEconomic Research Working Paper Series No.14049, National Bureau of Economic Research,Cambridge, MA.

Morsink, J. and Bayoumi, T. (2001) A peek inside theblack box: the monetary transmission mechanism inJapan, IMF Staff Papers No. 48, InternationalMonetary Fund, Washington, DC, pp. 22–57.

Raghavan, M., Silvapulle, P. and Athanasopoulos, G.(2012) Structural VAR models for Malaysian mone-tary policy analysis during the pre- and post-1997Asian crisis periods, Applied Economics, 44, 3841–56. doi:10.1080/00036846.2011.581360

Ramaswamy, R. and Sloek, T. (1997) The real effects ofmonetary policy in the European Union: what are thedifferences?, IMF Working Papers No. 97/160,International Monetary Fund, Washington, DC.

Reinhart, C. M. and Rogoff, K. S. (2009) The aftermath offinancial crises, The American Economic Review, 99,466–72. doi:10.1257/aer.99.2.466

Roye, B. V. (2011) Financial stress and economic activityin Germany and the euro area, Kiel Working PaperNo. 1743, Kiel Institute for the World Economy, Kiel.

Sims, C. A. (1980) Macroeconomics and reality,Econometrica, 48, 1–48.

Sims, C. A. (1986) Are forecasting models usable forpolicy analysis?, Quarterly Review, FederalReserve Bank of Minneapolis, 10, 2–16.

Sims, C. A., Stock, J. H. and Watson, M. W. (1990)Inference in linear time series models with someunit roots, Econometrica: Journal of theEconometric Society, 58, 113–44.

Taylor, J. B. (1993) Discretion versus policy rules inpractice, Carnegie-Rochester Conference Series onPublic Policy, 39, 195–214. doi:10.1016/0167-2231(93)90009-L

Tng, B. H. (2013) External risks and macro-financiallinkages in the ASEAN-5 Economies, Bank NegaraMalaysia Working Paper Series No. 1/2013, BankNegara Malaysia, Kuala Lumpur.

Tng, B. H., Kwek, K. T. and Sheng, A. (2012) A creditapproach to measuring financial stress in Malaysia: acomparison from the Asian crisis to the global crisis,Singapore Economic Review, 57, 1250013-1–1250013-24.

Tobin, J. (1969) A general equilibrium approach to mone-tary theory, Journal of Money, Credit and Banking, 1,15–29.

Van Den Heuvel, S. J. (2002) Does bank capital matter formonetary transmission?, Economic Policy Review, 8,259–65.

5184 B. H. Tng and K. T. Kwek

Appendix. Robustness of Impulse Reponses to Alternative Ordering Assumptions

Response of IPI to FSI shocks Response of IPI to interest rate shocks

–0.02

–0.01

–0.01

0.00

0.01

0 5 10 15 20 25 30 35 40 45 50 55 60

Indonesia

B 2

3 4

–0.01

–0.01

0.00

0.01

0 5 10 15 20 25 30 35 40 45 50 55 60

Malaysia

B 2

3 4

–0.01

–0.01

0.00

0.01

0 5 10 15 20 25 30 35 40 45 50 55 60

Philippines

B 2

3 4

–0.01

–0.01

0.00

0.01

0 5 10 15 20 25 30 35 40 45 50 55 60

Singapore

B 2

3 4

–0.02

–0.01

–0.01

0.00

0.01

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Thailand

B 2

3 4

–0.01

0.00

0.00

0.00

0.00

0 5 10 15 20 25 30 35 40 45 50 55 60

Indonesia

B 2

3 4

0.00

0.00

0.00

0.00

0 5 10 15 20 25 30 35 40 45 50 55 60

Malaysia

B 2

3 4

–0.01

0.00

0.00

0.00

0.00

0 5 10 15 20 25 30 35 40 45 50 55 60

Philippines

B 2

3 4

–0.02

–0.01

–0.01

0.00

0.01

0 4 8 12 16 20 24 28 32 36 40 44 48 52 56 60

Thailand

B 2

3 4

Fig. A1. Impulse responses from alternative ordering assumptionsSource: Authors’ estimates.Notes: B refers to impulse responses from the baseline model. The numerals 2, 3 and 4 are impulse responses fromspecifications with the FSI ordered, respectively, before the NEER, the NEER and real credit, and the NEER, real credit andthe interest rate. Other assumptions remain similar to the baseline model.

Financial stress, economic activity and monetary policy in the ASEAN-5 5185