Measuring the effectiveness of monetary policy during the global commodity boom : a case of Indonesia

(1)

Measuring the Effectiveness

of Monetary Policy During the

Global Commodity Boom

A case of Indonesia

ALFONSO D.D. HUTAGALUNG

ANR 11384832 alfonso.danielsaut@gmail.com

MSc Economics track International Economics & Globalization

SUPERVISOR

Peter Foldvari

SECOND READER

(2)

1

Statement of Originality

This document is written by Alfonso Danielsaut D’Albuquerque Hutagalung who declares to take full responsibility for the contents of this document. I, Alfonso D.D. Hutagalung, declare that the text and the work presented in this document is original and that no sources other than those mentioned in the text and its references have been used in creating it. The Faculty of Economics and Business is responsible solely for the supervision of completion of the work, not for the contents

Alfonso Hutagalung 15 August 2017

(3)

2

Abstract

This paper investigates the effectiveness of monetary policy by Bank Indonesia in stabilizing inflation during the period of global commodity boom. We adopt the use of narrative approach to identify the monetary stance of Bank Indonesia during the period of 1999-2016. The result from the narrative analysis is transformed into dummy variable and evaluated using a structural vector auto regression (SVAR) model. The dummy variable helps us to avoid simultaneity issue in monetary literature, known as the price puzzle (Sims, 1986). The empirical exercise suggests three views regarding the conduct of monetary policy. First, inflation in Indonesia is relatively dependent to movement in global commodity price and exchange rates which complicate monetary conduct. It is unclear whether contractionary monetary policy is effective in containing inflation. Second, production activity of manufacturing sector deteriorates during periods of high inflation. Thirdly, monetary policy has less of an impact on real production activity compared to evidences from developed countries such the United States and United Kingdom.

(4)

3

1. Introduction

After the Asian crisis in 1998, the performance of manufacturing sector in Indonesia deteriorates despite periods of high economic growth. Concurrently, Indonesia encountered a global commodity boom which began in 2000. The cycle was triggered by rising demands in emerging markets particularly China and India. The global price fluctuations brought significant impact to commodity exporters such as Indonesia. Inflationary pressure from commodity prices as well as increasing domestic economic activities fostered an environment of high inflation in Indonesia. Bank Indonesia, the monetary authority of Indonesia, suddenly run into an unfamiliar task of containing global spillover. As a response, Bank Indonesia seek to control domestic inflation and instigated a contractionary monetary policy. It is debatable whether the prescribed action is successful in containing global commodity boom. Frankel (2006) suggests that monetary policy by the U.S. Federal Reserve have the capability to influence aggregate prices at the global level. However, does Bank Indonesia possess similar force to control spillovers from the global commodity boom?

Figure 1.1 Percentage share of Indonesian merchandise exports from 1990-2015

Source: Author’s aggregation based on World Development Indicator

Monetary policy has its virtue in stabilizing the domestic inflation. However, it is prone to side-effects in an economy with heterogeneous production sectors. Monetary policy, at first best, performs identically to antibiotic, treats and prevents bacterial infection. Though, prolonged reliance over the treatment reduces the marginal effect of the medicine. It may cure on the expense of other mutual substances within our body. Monetary policy conveys similar attributes to an antibiotic. Potentially, it stabilizes the price level during economic boom which are desirable for economic growth. Yet, there are asymmetric effect of monetary policy which might depress the non-booming sector in order to maintain the aggregate economy.

0 10 20 30 40 50 60 70 1990 1995 2000 2005 2010 2015 % o f sh ar e Commodity exports Manufacturing exports

(6)

5

Central Bank should consider the trade-offs of pursuing domestic price stability. A narrow focus on a single policy rate is questionable since it ignores features such as heterogeneity within an open economy. Repercussions of a stabilization act could prove short-lived and reversible in the future or persist as some sort of hysteresis. Krugman (1987) argued that contractionary monetary policy is desirable to curb overheating but it might depress the non-booming sectors. Indonesia’s dependence on commodity productions poses a compelling case to analyze monetary policy in a small economy with the presence of heterogeneous firms.

Commodity boom generates asymmetric spillover to the domestic economy. For commodity exporters, positive demand shock increases revenue and domestic absorption through higher spending. The demand shock pushes wages upwards and spreads to other sectors, for instance, manufacturing. Simultaneously, higher revenue causes the exporter’s currency to appreciate. The currency appreciation increases the relative price of domestic goods in the foreign market, making other exporting sectors less competitive. This causal relationship is known as the Dutch Disease.

In a span of 12 years from 2001, the real effective exchange rate (REER) of Indonesia grew by more than 50% until it phased out in 2013 (Cali and Nedeljkovic, 2016). The financial crisis in 2008 produced a sharp downfall in REER although the value recovered instantly afterward. Indonesia was considered one of the most resilient countries during the financial crisis with stable GDP growth relative to other emerging countries. Subsequently, the Rupiah has weakened partly due to the global economic slowdown and the expiration of monetary easing by the U.S. Federal Reserve. Periods of strong REER appreciation demonstrates a familiar symptom of Dutch Diseases. The following currency depreciation does not necessarily mean a boost in competitiveness for manufacturing export. Recovery remains sluggish and vulnerable to the volatility of commodity prices.

Figure 1.2 Indonesia REER vis-à-vis USD, January 2002-December 2015

Source: Federal Reserve of St. Louis

30,00 40,00 50,00 60,00 70,00 80,00 90,00 100,00 110,00 120,00

(7)

6

It is important to note that interest rates of Indonesia’s major trading partners, such as the United States, the European Union, and Japan, were falling until the financial crisis in 2008. The higher domestic interest rate, ceteris paribus, is associated with a currency appreciation which could augment the spillover of the commodity boom and strengthen the overvaluation of the Rupiah. The interest rate gap caused a stronger currency appreciation based on the uncovered interest parity (UIP) condition. Furthermore, interest rate fluctuations create uncertainty in the real economy which restraints investment decision by firms. Interest rate hike could increase the cost of capital through higher borrowing cost or deferred investment decision to the future.

What could have happened to the manufacturing sector during the global commodity boom? Did monetary policy exacerbate the decline of the manufacturing sector? To a certain extent, Bank Indonesia’s effort to stabilize macroeconomic fundamentals is highly desired in the midst of the commodity boom. However, it is naïve to assume that monetary policy has performed immaculately to stabilize inflation. Small economy is considered as a price taker in the global market. Therefore, I assume that the monetary authority has less of an influence on global prices.

To answer these questions, I will employ an empirical analysis using a structural vector auto regression (SVAR) model. The dataset consists of monthly observations from 1990-20161_{. As common in monetary}

literature, identification remains a challenge in disentangling the actual effect of monetary policy. I adopt a narrative approach to solve empirical irregularity such as the price puzzle. Using minutes from the monthly Governor’s meeting, I assigned a dummy variable to indicate the monetary stance of Bank Indonesia. Subsequently, this qualitative study will be integrated into a statistical exercise. First, I examine the response of macroeconomic indicators, such as inflation and manufacturing production, to shocks from monetary policy, nominal exchange rate and global commodity price. The result from the impulse response functions serves as a descriptive relationship between the variables of interest. Finally, I analyze the contribution of each impulse variables in explaining the variance of inflation and manufacturing productions.

Based on the empirical exercise, I formalized three findings to approach our research questions. First, inflation in Indonesia is relatively susceptible to movements in commodity prices, compared to monetary policy. Second, quantitively the impact of a contractionary monetary policy is not significant to the performance in the manufacturing sector. This finding challenges our earlier hypothesis that monetary

1_{Despite the extensive number observations on macro-economic variables, I decided to run the empirical analysis}

(8)

7

policy is crucial to the deterioration of manufacturing. Taken together, these findings question the effectiveness of Bank Indonesia’s monetary policy. However, we should not infer that the monetary authority is incompetent since there are structural problems that limit the transmission of monetary policy. Thirdly, the narrative approach is relatively effective in avoiding empirical irregularities such as the price puzzle compared to the nominal interest rate. Impulse response functions also show consistent patterns that monetary policy works with an inertia.2

The rest of the paper is organized follows. Section two contains previous literature regarding the conduct of monetary policy. In section three, we begin with a narrative description of the monetary stance of Bank Indonesia during the commodity boom. Section four integrated the narrative approach into an SVAR model. Chapter five will discuss the findings and limitation of the study. In the last section, we will summarize our view regarding monetary policy in Indonesia.

2. Literature review

On the use of narrative approach

By narrative, this paper suggests the use of epidemiology in the analysis of economic policy. What motivates the Central Bank to decide upon a certain policy? The reasoning behind a certain economic policy is important to measure the actual impact of the policy in the real economy. Central Banks regularly documented minutes of their internal meetings before the announcement of the upcoming monetary stance. The narrative approach makes use of this resources and evaluate the motivations of policymakers regarding their action. Studies using narrative can be traced back to Romer and Romer (1989). The authors implemented a classic narrative approach, pioneered by Friedman and Schwartz, to obtain evidence about the real effects of monetary contraction. The use of dummy variable is successful in terms of solving simultaneity issue in monetary studies such as the price puzzle. Moreover, the use of narrative improves identification of structural model which meaningful inference from empirical analysis. Other literature supports alternatives to evaluate monetary policy such as historical analysis, apart from structural models (Taylor, 1997). Shiller (2017) put forth this issue in his American Economic Association Presidential Address.

Bernanke and Mihov (1998) argued that the analysis by Romer and Romer was successful in terms of

(9)

8 disentangling shocks by money supply and money demand. However, their analysis treats every monetary shock as comparable due to the use of the dummy variable. Boschen and Mills (1991) attempted to extend the variation in shocks by assigning a scale to each contractionary policy. While Romer and Romer used a dummy of one for contractionary policy, Boschen and Mills assigned different scales for both contractionary policy and expansionary policy. Despite the narrow focus, the Romer dummy is consistently evaluated in recent literature as a benchmark for the impact of monetary policy.

Global commodity price and monetary policy

The origin of a monetary tightening could be as important as the development of the asymmetric effect. Bernanke, Gertler, Watson, Sims, and Friedman (1997) argue that the impact of oil and energy prices increase are relatively small to total production costs, hence not significant to account for a decline in output. An appropriate explanation, therefore, is that oil price shock triggers a tightening policy which prompts a recession. For developed countries, this statement might be feasible since commodities were relatively cheaper. However, for developing countries, such as Indonesia, the oil price has a significant influence on economic activity. Oil shocks might matter more than monetary tightening.

Bernanke et al. (1997) also explore the relationship between a commodity price shock and endogenous monetary policy through impulse response analysis. The result shows that the US federal funds rate rises following a commodity price shock which induces a fall in output. In the absence of monetary policy response, the recessionary impact is eliminated and inflationary impact is aggravated. Consequently, U.S. Federal Reserve is able to decide whether a recession is desired during an inflationary period. It is important to distinguish the feature of the U.S. Federal Reserve and Bank Indonesia when dealing with global commodity shocks. The U.S. Federal Reserve have the capability to influence global shocks which are lacking from Bank Indonesia.

Period of Great Moderation was characterized by declining interest rate in the United States and European countries due to the reduction in business cycle fluctuation. Frankel (2006) shows that commodity price is counter-cyclical to monetary policy. If major economies such as the United States and European countries employ a loose monetary policy, commodity prices will soar. Monetary easing by developed economies is considered as one of the causes of the global commodity boom. If so, then how should monetary authority in emerging countries react to such movement?

(10)

9

How do monetary shocks affect heterogeneous sectors?

In practice, there are various types of heterogeneity in the economy. Gertler and Gilchrist (1994) examines the impact of monetary policy to manufacturing firms with different sizes. Do small firms and big firms behave differently to monetary shocks? The credit market is one of the transmission channels of monetary policy. If the Central Bank wants to reduce the money supply, an interest rate hike could reduce the amount of credit in circulation and attempt to manage inflation. This objective could be followed by a credit rationing or higher lending rate by commercial banks to creditors, in this case, firms. In the first scenario, banks are reluctant to borrow from the interbank market due to higher interest rate and reduce credit supply. On the second scenario, banks raise loan rates and increase the cost of borrowing. The higher borrowing cost lowers the demand for credit by firms. Both set-ups are aimed at decreasing the amount of money in circulation but differ in the process.

Small firms could suffer more compared to large firms since they lack the capability to adjust during recessions. Under credit rationing, banks prefer to allocate more credits to large firms since they are capable to smooth-out production with larger inventories and market power. This sequence is known as the “flight to quality”. If borrowing rates increase, small firms will suffer high borrowing cost as they are deprived of other financing options. Additionally, large firms are able to explore alternative options such as foreign credit, bond issuance or equity financing.

Stiglitz (2000) further emphasizes the relationship between capital constraint and theory of firms. Contractionary monetary policy during the East Asian crisis provides evidence that higher interest rate had a large impact on firm’s production activity. The effect is amplified for financially constrained firms which depressed trade exports (World Bank, 2000). The Asian financial crisis began with a slump in the credit market. Speculative currency attack attempted to test the resilience of the fixed exchange rate regime in Indonesia. At the same time, the government was running unsustainable fiscal deficits which depleted foreign reserves. In the 1990’s, Indonesian firms and banks receive abundant capital flow from foreign banks due to financial deregulation. As soon as the peg was abandoned, currency mismatch caused a sudden stop to financial resources. Firms and banks suffered liquidity constraint which depressed the real economy. Monetary tightening by the Central Bank constrained the credit market even further.

The distortionary effect could also occur between traded sectors and non-traded sectors. Krugman (1987) provides a classic literature on monetary policy and the ensuing distortion in international trade. He argued that monetary tightening causes a currency appreciation which, consequently, leads to a loss of

(11)

10 competitiveness in traded goods. If the currency appreciation persists, then the performance of the traded goods sector will deteriorate and cause a trade deficit. For export-led growth this situation is not ideal since it could affect the business cycle. Krugman uses United Kingdom trade performance during Margaret Thatcher’s ear to illustrate his proposition. Bank of England adopted a more contractionary stance than their trading partners to control inflationary pressure from the non-tradeable sectors. As a result, the Pounds strengthens relative to U.K.’s trading partner and reduce the competitiveness of traded sectors. Krugman questions the implication of monetary policy to the traded goods sector. He argue that the contraction should not be only occur in the tradeable sector, as illustrated in United Kingdom’s case.

Contractionary monetary policy by Bank Indonesia could yield similar results and reduce the competitiveness of the non-booming sector. It could induce a currency appreciation which deteriorates the relative price of Indonesian exports. Under such currency pressure, the government should devise their policy and preserve trade competitiveness. Krugman (1987) suggests alternative measures, such as capital control or trade policy instruments, to insulate the tradable sectors from distortions. If monetary tightening is unavoidable, policy makers should compensate the non-booming sector to minimize the loss.

Monetary policy has its virtue which allows Central Banks to pursue inflation stability independently. However, critics assert that monetary policy is responsible, partially, for their adherence to inflation targeting (Woodford, 2012). Dependence on a single policy rate for the aggregate economy exploits the principle of one size fits for all. Svensson (2010), on the other hand, insists that monetary policy should seek price stability and oversee financial development through macro-prudential policy. In response to this view, this study will limit their scope to explaining the actual impact of monetary policy rather than pointing retrospective judgment to Bank Indonesia’s conduct.

Monetary policy literature in Indonesia

Literature regarding the actual impact of monetary policy in Indonesia is scarce. Most of the studies related to monetary policy focus on the transmission channel and the behavior of monetary policy in Indonesia. Goeltom (2008) explains the various transmission channels of monetary policy in Indonesia. She argued that monetary policy has been able to influence interest rate and bank lending but less for exchange rate mechanism. However, there is no clear consensus amongst previous studies regarding the capability of monetary policy in managing inflation. The interest rate and credit channel in Indonesia have been evaluated empirically by several studies using factor-augmented vector auto regression (FAVAR) as well global vector auto regression (GVAR) approach (Harahap, Nurliana, Ariyanti, Khasananda, 2013, Agung,

(12)

11 Kusmiarso, Pramono, Hutapea, Prasmuko & Prastowo, 2013, Juhro, 2014). They conclude that interest rate has a huge influence on the lending and deposit rate after the global financial crisis. Global shocks such as oil price, U.S. Federal Reserve rate and economic growth of trading partners appeared to have significant spillovers on domestic output. Kusmiarso (2002) try to cover the period before the Asian financial crisis. Interest rate on treasury bond appears to be significant to lending rate although changes appeared to be sluggish before the Asian financial crisis. Despite the availability of evidence on monetary transmission, I realize the need for more empirical evidence regarding the role of policy in stabilizing inflation. After all, the Central Bank of Indonesia has its mandate in managing domestic inflation.

3. Identification of monetary stance in Indonesia

The use of narrative approach offers a different perspective to analyze the impact of monetary policy. Policy rates, the main instrument of monetary policy by Bank Indonesia, is not purely considered as a knife-edge decision to achieve a certain target. It operates as a benchmark for lending rates by banks or treasury bonds and also as an anchor of future expectation. Apart from nominal anchor, Central Bank’s policy rate is commonly used as a predictor to the future economic environment. Firms as well households will take note of the Central Bank’s stance before executing their business decision or personal investment. Thus, analyzing the Central Bank’s motivation could help us to analyze the actual impact of monetary policy.

In order to simplify the identification process, we will restrict our analysis to periods of contractionary policy in the similar spirit of Romer and Romer (1989). The isolation will help us to focus on actions which are intended to counteract inflation. In the past 15 years, Indonesia experienced a high level of economic growth which allows inflation to be moderately high compared to developed countries. In contrast, monetary authority in advance economy has been pursuing unconventional measures to thrust inflation above its zero-lower bound ever since the global financial crisis. In principle, a contractionary policy which is designed to cure inflation is neutral from the influence of output to a certain extent. On the other hand, expansionary monetary policy is endogenous to both inflation and output motives.

I use historical records of the Bank Indonesia’s Governor meeting to describe the reasoning behind the shift in monetary stance. In practice, Bank Indonesia doesn’t explicitly implement an unconventional monetary policy such as forward guidance. However, they regularly publish a summary of the Governor’s meeting to the public on a monthly basis. The records contain Bank Indonesia’s monetary stance based on the ongoing macroeconomic development and projected inflation in the medium term. The minutes

(13)

12 also detailed the required measures to maintain inflation within the target. The earliest public record of the Governor’s meeting could be traced back to April 2005, after the official adoption of inflation targeting. However, I have been able to obtain annual reports of the Bank’s monetary stance, dating back from 1990 until 2005. These records are available in printed documents at Bank Indonesia’s internal library. The annual reports lack monthly evidence of the Bank’s projection. Yet, it complements our analysis with sufficient description and reasoning of the Bank’s decision throughout the year as well as policy projections for the upcoming year.

Apart from changes in the policy rate, it is essential to identify the framework of Bank Indonesia during the span of 1990-2005. There are two major shifts that affect Bank Indonesia’s monetary policy framework. The first shift occurred in 1999, following the Asian financial crisis. The crisis in 1998 brought an end to the fixed exchange rate regime and marked the adoption of floating exchange rate. The Asian financial crisis forced Indonesia to leave its peg due to the depletion of foreign currency reserves. Under unsustainable fiscal deficits and speculative currency attack, the Rupiah depreciated dramatically and inflation soar due to exchange rate pass through. Consequently, Bank Indonesia gained independence in conducting monetary policy under Law no.23/1999. The initial target for the newly-minted authority is to focus on a single objective of price stability. Moreover, Bank Indonesia also assumed the responsibility to maintain the foreign exchange stability after abandoning the peg. The second event came in July 2005 with the official introduction of inflation targeting framework. Under the new framework, Bank Indonesia explicitly announce a target for inflation in the medium term. Additionally, the Central Bank introduced Bank Indonesia (BI) rate as the main instrument for monetary policy, replacing treasury bonds as interest rate anchor. In August 2016, BI rate was effectively replaced by BI 7-days repo rate a nominal anchor.

As a proposition, I illustrate the timeline of monetary policy in Indonesia and compare it to historical records of recession by the Federal Reserve of St. Louis. The OECD based recession indicator is the only source for identification which is available for Indonesia on monthly basis. Federal Reserve of St. Louis constructed an indicator of recession based on Organization for Economic Co-operation and Development (OECD) Composite Leading Indicators. Business cycles and turning points are identified based on the growth cycle approach. A dummy variable is then created to represent periods of economic expansion and recession. Recessionary period is given a value of one and zero otherwise. A recession occurs for the entire period of the peak until the through. The data suggests that there are five recessions in Indonesia that occurred during the period of 1990-2015.

(14)

13 Figure 3.1 Bank Indonesia’s monetary stance to Fed of St. Louis recession indicator from 1990-2015

Source: Federal Reserve of St. Louis using OECD Composite Leading Indicators data

Based on the records of Bank Indonesia, we identify six contractionary attempts by the authority to stabilize inflation. The actions started in April 1994, March 2001, March 2005, May 2008, February 2011 and June 2013. In each period, Bank Indonesia emphasized the need for a contractionary period and acted accordingly. Amongst the contractionary periods, the contraction in 1997 is debatable since inflation was not the main concern of Bank Indonesia. However, we will include this episode as a comparison for suspected periods of contraction. The upcoming section will describe the chronology and objectives of the shift in policy by Bank Indonesia.

May 1995

Following the recovery from oil recession, the government arranged a set of financial deregulations in order to stimulate non-oil export sectors in Indonesia. The financial deregulation attracted new banks to open up and increased the supply of credit to the economy. Credits grew at an annual rate of 22% from 1990-1996 with private banks receiving huge capital inflow from foreign financial intermediaries. Investment and consumption in construction and manufacturing sectors were booming. As a result, inflation rose up to 8.9% and Bank Indonesia decided to tighten up the money supply. Interest rates were adjusted monthly from 8.72% in April 1994 up to 14.74% by May 1995. Domestic credit was put in strain but control on foreign debt remained lenient which allow private firms to access fund directly from foreign intermediaries. This decision was highly disputed since it nurtured the origin of the Asian financial crisis. Domestic banks saw increases in non-performing loans while exchange rate risk developed as a threat for private firms.

0 1

1990-01-01 1995-01-01 2000-01-01 2005-01-01 2010-01-01 2015-01-01

BI policy

OECD based Recession Indicators for Indonesia from the Peak through the Trough, +1 or 0, Monthly, Not Seasonally Adjusted

(15)

14 July 1998

Following the depletion of foreign reserves, Bank Indonesia decided to abandon the pegged exchange rate in August 1997. Subsequently, the Asian financial crisis forced Bank Indonesia to employ contractionary policy under threat of bank run and currency attack. The International Monetary Fund (IMF) suggested the use of interest rates to in response to the macroeconomic development. Several macroeconomists argued that the proposition could exacerbate the domestic financial market (Radelet and Sachs, 1998). Sugisaki (1998), IMF deputy managing director at the time, addressed the need for a higher interest rate as part of IMF rescue measures. Higher interest rates could attract funds and create a currency appreciation to offset the devaluation of Rupiah. The IMF claimed that the proposition is justified by the success of analogous action in the pre-crisis period. The desired currency appreciation would lessen foreign debt burdens of private firms and pass through from import goods. However, the exchange rate effect only lasted momentarily before placing a heavy constraint on the fragile financial sector. The IMF and Bank Indonesia miscalculated the impact the of the action. Interest rates on Surat Bank Indonesia went upward from 20% to 70.81% while interbank rates responded similarly, increased from 11.5% to 30%. Consequently, domestic banks were unable to provide liquidity and meet their obligations. The tightening reached the real economy with exporters, especially manufacturers, failed to receive sufficient credit for working capital. Total manufacturing production plummeted by 23.7% in the first nine months of the crisis. External funding was not an option since foreign banks have stopped accepting a letter of credit. This situation is worsened by the loose of confidence by foreign investor. Although Indonesia and the IMF managed to secure an arrangement by January 1998, exchange rate didn’t regain their value since higher interest rate was anticipated as high-risk premia on domestic bonds. The monetary contraction added further strain to the evolution of the Asian financial crisis. The measure could be seen as a typical austerity measure implemented by the IMF when dealing with the economy. This episode doesn’t demonstrate pure intention to cure inflation. Yet, the records show that lower inflation and a currency appreciation is desirable to sustain mounting debt.

August 2001

Indonesia demonstrated substantial growth in the aftermath of the Asian financial crisis. Actual economic growth reached 4.8%, surpassing Bank Indonesia’s prediction of 3-4%. Furthermore, the Government of Indonesia managed to secure a loan package with IMF and provide sufficient liquidity to insolvent banks. Banks began to roll out more credit to the economy even though there are increases in Capital Adequacy Ratio to 8%, a 4% increase. There were high expectations of an upcoming recovery after the slump of

(16)

15 2000 which contributes to wage and pricing expectation. Apart from domestic pressure, there were suggestions that the Federal Reserve is going to lift the Fed rates. Apparently, the attempt didn’t materialize as policy rates began falling which marked the beginning of the Great Moderation. However, domestic political instability was perceived as a risk by the global market which causes the depreciation of Rupiah. This factor contributed to the surge in exchange rate pass through on import goods. As a result, inflation by the end 2000 rose to 9.35%, far above the 3-5% target inflation. Bank Indonesia was confronted by the responsibility to bring down inflation when the economy was recovering. The Bank realized that monetary transmission was not in force but decided to tighten up excess liquidity. Interest rate on treasury bonds was adjusted several times since rates on deposits remained sluggish to changes.

December 2005

Based on the record of the Governor’s meeting, Bank Indonesia decided to tighten up the monetary stance due to internal and external pressure. Target inflation for 2005 was set at 6±1% but Bank Indonesia was forced to adjust policy rates for nine straight months in order to contain inflationary pressure. Real economy grew moderately from 5.5% to 6.1%, supported by investment growth and exports in natural resource-based commodities and agriculture-industrial products. Concurrently, import rose at a higher rate in line with the expansion of domestic demand. The higher influx of import goods allows exchange rate pass through to enhance the impact of imported inflation to the heated domestic economy. This situation was aggravated with multiple hikes of domestic fuel price fuel price due to global oil price shock. Fuel price in Indonesia was subsidized by the government which might cause a distortionary effect on consumption. The subsidy was targeted for lower income households although in reality the subsidy was misallocated to middle income and higher income households (Dartanto, 2013). Oil price rose as high as 100% and caused domestic inflation to jump from 7,5% to 17.11% in a span of 6 months. At the same time, the Federal Reserve just announced a series of interest rate hike, dating back to June 2004, on the back of asset price bubble and high inflation. Minutes from the Government’s meeting insisted that the multiple fine-tuning decisions were necessary amidst mounting inflationary pressure.

October 2008

Indonesia went through a period of global commodity boom which spiked in 2008. The upswing was caused by high demand from emerging market countries such as China and India. Simultaneously, advanced countries were suffering from the global financial recession due to the subprime mortgage crisis which contaminated European financial market. Interest rates in developed countries were slashed to

(17)

16 lower zero bound as unconventional monetary policies were rolled out. The turmoil was severe but Indonesia managed to withstand the pressure. Economic growth remained resilient around 6.1% which was recorded as the highest annual growth since the crisis in 1997. Growth was boosted by strong export performance, particularly commodities export, household’s consumptions and rising private investment. Indonesia managed to diversify their trading partner and expanded their trade flow to China in the midst of the subprime crisis. Policy rate remained steady at 8% for six months in 2008 before Bank Indonesia recognized mounting pressure on inflation level. Annual inflation began at 7.4% for the first months which was within the target of 5±1% for 2008. In March 2008, inflation surged 8.17% and continue to rise up to 12.14% by September. The appreciation of Rupiah didn’t manage to hold the surge in global commodity price which passed through the domestic economy. Bank Indonesia attempted to counter the rising inflation by raising the BI rate in six successive months. Each of the records of the Governor’s meeting insisted that global commodity boom gave a significant contribution to the surge in domestic inflation. The Governor of Bank Indonesia, Boediono, emphasized that the institution would make use of all the monetary tools at their disposal to curb inflation. Inflation did fell gradually after September and reached 2.71% in July 2009. However, the actual effect of a contractionary monetary policy is ambiguous considering the dramatic fall of global commodity price in the third quarter of 2008. Domestic inflation responded simultaneously to the change in global commodity prices.

Figure 2.2 Global commodity price index. Index 2005=100 from January 1990-December 2015

Source: Federal Reserve of St. Louis 0 50 100 150 200 250

2002-01-01 _{Global Price Index of All Commodities©, Index 2005 = 100, Monthly, Not Seasonally Adjusted}2004-01-01 2006-01-01 2008-01-01 2010-01-01 2012-01-01 2014-01-01

Global price of Industrial Materials index©, Index 2005 = 100, Monthly, Not Seasonally Adjusted

(18)

17 February 2011

The effect global financial recession began to evaporate and spell out some improvement going forward. Bank Indonesia misjudged the ability of advance countries such as the United States to overcome the recession as well as China’s steady demand. Global commodity prices returned to its pre-crisis level in the back of a steady demand from China and India. This strong demand boosts Indonesia’s export, particularly in commodities, and increased optimism in economic growth. The Central Bank predicted that economic growth will remain within six 6-6.5% in 2011. Core inflation in January rose to 0.92% (mtm) and 7.02%(yoy) which surpassed the target of 5±1%. In order to stabilize the level of inflation, Bank Indonesia decided to lift the interest rate by 25 bps. Based on the minutes of the Governor’s meeting, rising global commodity prices were mentioned multiple times as the forces behind the high inflation expectation. Results from the survey of consumers, producers, and financial assets prices also reinforced the prospect of higher inflation in the future3_{. It brought significant influence to volatile food prices and fuel which was}

vital to inflation dynamics.

November 2014

With the end of quantitative easing by U.S. Federal System in sight, Bank Indonesia decided to tighten its monetary stance. The shift began in June 2013 with an increased by 25 bps. Inflation began to scratch the upper boundary of target inflation which is set at 4.5±1%. Initially, increases in inflation were influenced by administered price such as electricity, gas price, and volatile foods. Global commodity price went through a rebound after the global financial recession. Bank Indonesia observed the spillover of the global recession and the recent slowdown in China. These external factors affected projections of Indonesia’s economic performance downward. At the same time, signals from the U.S. Federal Reserve suggested the conclusion of the monetary easing. Minutes of Board meeting from June 2013 indicated that a normalization of monetary policy should be taken in the near future. By this this time, the depreciation of Rupiah was underway which deteriorated import pass through to inflation. Inflation spiked in July 2013 due to the removal of fuel subsidy and Bank Indonesia decided to lift the policy rate from 5.75% to 7.5% by November 2013. The U.S. tapering began in January 2014 with the end of the asset purchase and interest rate began to increase by the end of the year. Simultaneously, Joko Widodo, the new President of Indonesia, decided to re-allocate expenditure budgets from fuel subsidy to productive sectors. This step was part of the Government’s agenda in reducing consumption distortion of oil in the economy. The Government realized that the influence global commodity price was significant. Thus, in order to cope

(19)

18 with the volatility of commodity prices, both the Government and Bank Indonesia mutually consented to shrink fuel subsidy. Bank Indonesia maintained its tight stance and decided to raise interest rate by a further 25 bps in ensuing month. In its November 2014 meeting, the record stated that Bank Indonesia is committed to bringing inflation down in order to achieve their inflation target. However, a temporary spike in inflation is inevitable after the removal of the fuel subsidy. The reduction in fuel subsidy could arguably reduce the excessive consumption of fuel and change the inflation behavior in the future. Bank Indonesia opted to tighten its monetary stance despite the resulting fall in output.

Findings of the narrative approach

Based on the narrative approach, we could point out several findings from the six contractionary periods. First, the nature of inflation that occurred preceding the contractionary stance was influenced by global factors apart from the episode in 1995. The episodes in 1998, 2001, 2005, 2008, 2011 and 2013 were caused by volatility in the global economy. This symptom showed that, as a country, Indonesia is vulnerable to global shocks despite its emergence as one the of the worlds’ largest market and population. Dependency to advance economies and commodities caused spillovers to the domestic economic cycle to a certain degree. Ever since the Asian financial crisis, Indonesia has put some effort to diversify its trading partner with emerging countries such China and India as well strengthening economic cooperation with its ASEAN counterpart. However, the enlargement of trading partners is not concurrently followed by diversification in production sectors.

The second finding is related to the volatility of global commodity price. Out of the six contractionary episodes, four episodes were caused by volatility in commodity prices. Indonesia is known as an exporter of commodity products such as palm oil, coal, agricultural products and crude oil. Apart from production, Indonesia also imports various commodities such as refined oil, food and other processed commodities which are essential inputs for the industrial production. Movements in commodity prices produces spillovers to the domestic economy. When oil price increase, inflation immediately surged through the supply side as well as the demand side. In the 1990’s, oil production was a huge part of the country’s GDP. Thus, an upswing in price causes output to substantially increase and boost the economy. Increases in oil price also affect the production side through higher administered price. The Government of Indonesia spent around 20-25% of their fiscal expenditure for fuel subsidy to offset the volatility of oil price. Initially, fuel subsidy is designed to protect the poor and stimulate industrial production. In reality, the subsidy has produce distortionary effect on consumption which is contradictory to the intended purpose.

(20)

19 The shocks in 2008 and 2013 were influenced by increased demand for commodities by emerging market economy. Exports of commodities were the motor of growth, accompanied by strong domestic consumption and non-tradable services. Indonesia was able to maintain high economic growth, ranging from 5.5% to 6.3%, despite the recession that plagued advance economies. During this period, Indonesia continued to grow on the back of strong commodity price and suffered periods of high inflation. If the contractionary measure is intended to suppress output, the effect could, ambiguously, be more potent on manufacturing sectors. Manufacturing is a capital-intensive sector which requires substantial investment to maintain its production4_{. Moreover, higher commodity prices cause production cost to increase. Thus,}

in order to stabilize the aggregate inflation, a contractionary policy may tend to depress manufacturing sector in the back of the commodity boom. Under the credit channel, there are notions that loans will be preferably given to sectors which are flourishing, flight to quality, with better growth prospect. Otherwise, lending rates will increase and create financial constraints to the non-booming sector.

The narrative part has been able to suggest patterns in Bank Indonesia’s monetary policy. Each contractionary action has been followed by a recessionary period or, to a certain extent, further deterioration of output. In the next section, we will supplement our analysis by translating our narrative approach into a statistical test.

4. Empirical approach

In this section, the findings from the narrative approach will be integrated into a structural vector auto regression (SVAR) model. The use of VAR model to predict relationship among macroeconomic variables such as monetary shock, GDP, and inflation is widely documented. VAR models allow variables to be correlated with its own lagged value, as well as other predictor variables, and causality to run in both directions (Sims, 1980). The first part will describe the quantitative interpretation of the narrative analysis. The second part contains the specification of the empirical model of the narrative approach based on SVAR approach. Additionally, the results from the narrative approach will be compared to the interest rate indicator.

4_{Based on author’s calculation, approximately 73-78% of credit in manufacturing sector is utilized as working capital.}

(21)

20

The Narrative Dummy

Results from narrative approach have been taken into statistical approach by Romer and Romer (1989) and Boschen and Mills (1991). The former assigned contractionary periods with a value of one for peak months and zero for otherwise. A dummy variable is practical for the empirical approach but it represents a narrow definition of contractionary policy. Every contractionary policy will be treated with similar intensity which restricts the extent of monetary policy implementation. Boschen and Mills (1991) extended this analysis by assigning a different intensity to each monetary stance. Contractionary episodes were clustered into five categories; very contractionary, contractionary, neutral, expansionary and very expansionary. Although Boschen and Mills offer a richer interpretation of contractionary episodes, the assignment of value is prone to tendentious interpretation compared to the classic Romer and Romer (Bernanke and Mihov, 1998).

Following the previous narrative analysis, this paper will focus on the effect of a contractionary policy, in the spirit of Romer and Romer (1989). Bank Indonesia represents their monetary stance based on the changes in their interest rate as their main instrument. To replace nominal interest rate, a dummy of one represents contractionary period. A tightening period is valid if increases in interest rate last for a minimum of six months. A value of one will then be assigned to the period with the highest interest rate, presenting the full scale of monetary contraction.

To understand the mechanism behind the dummy variable, Woodford (1999) provides a solid argument regarding monetary policy inertia. He suggests that monetary policy worked with an inertia and there is an optimal set of action for changes in interest rate. For example, changes in policy rate could move the yield on lending rate simultaneously or with a lagged response, in the case of stickiness. Frequent changes could spur an environment of high volatility and produce uncertainty in the economy. Under inflationary pressure, Central Banks might be forced to contemplate their predictions repeatedly. This situation is much severe for a small economy which is vulnerable to multiple external shocks. Interest-rate smoothing is desirable from the monetary authority perspective to stabilize the economy. Gradual movement in monetary stance could act as an anchor for the real economy, displaying certainty of future projection by the monetary authority. Naturally, changes in interest rate are likely to be followed by another change in the same direction rather than the opposite direction (Rudebusch, 1995, Goodhart, 1999). Hence, in order to examine the actual impact of interest rate changes, we need to observe the timeframe of the monetary tightening. This treatment could help us solve simultaneity problem that occurs in monetary studies.

(22)

21 Table 4.1 Summary of contractionary episodes by Bank Indonesia

Episodes Period Changes Initial position Peak position Changes

APRIL 1994 - MAY 1995 14 12 8.72% 14.74% 6.02% FEBRUARY 1998 - JULY 1998 6 5 10.87% 70.81% 59.94% JUNE 2000 - AUGUST 2001 15 10 11.74% 17.67% 5.87% APRIL 2005 - DECEMBER 2005 9 9 7.7% 12.75% 5.05% MAY 2008 - OCTOBER2008 6 6 8.25% 9.50% 1.25% FEBRUARY 2011 - JULI 2011 6 1 6.5% 5.75% 0.25% JUNE 2013 - NOVEMBER 2014 18 6 6% 7.75% 1.75%

Source: Author’s calculation using Bank Indonesia data

A summary of all the six contractionary episodes is presented in table 4.1. Contractionary periods appear to be protracted before the introduction of BI rate in 2005, bearing the recent episode. The use of yield on Surat Bank Indonesia (SBI) and open market operation as the main instrument of monetary policy present a different characteristic to the era of BI rate. Accumulated changes on interest rate average within the range of approximately 1,5% to 5,5% except for the episode in February 1998 – July 1998. The interest rate on Surat Bank Indonesia, government bond, fluctuate over 59% in a span of six months. The contractionary episode in 1998 will be treated as an outlier in our sample due to motives not solely dependent on inflation. Apart from the massive increase in interest rate, this episode also suffered changes in exchange rate regime which creates uncertainty in the conduct of monetary policy. Contractionary policy during 1998 was not solely independent on inflationary pressures but also debt consideration alongside increases in the risk premium. Thus, disentangling inflation as the sole motive of contractionary measure is problematic for this episode.

Data Specification

The dataset covers monthly observations of economic indicators from 1990-2016. These indicators consist of global commodity prices, manufacturing production index, consumer price index (CPI), credit utilized as working capital in the manufacturing sector, Bank Indonesia’s policy rate and two optional variables,

(23)

22 nominal exchange rate and Brent oil price index. Global commodity prices, manufacturing production index, CPI, nominal exchange rate and Brent oil price are available in Federal Reserve of St. Louis Economic Data (FRED). Domestic indicators such as credit variables are taken from Bank Indonesia’s Banking Statistics. There are two different resources for Central Bank’s interest rate. The first one is Indonesian Financial Statistics (SEKI) which provide the data for BI rate for 2002-2016. Secondly, Bank Indonesia annual report for periods before 2002. I checked the consistency of interest rate from the annual report compared to SEKI and FRED. SEKI first publication began in June 2005 while Bank Indonesia’s annual report is available since 1984. Bank Indonesia’s annual report is accessible in the Bank’s library. There are missing observations and inconsistency in the time series data which limit the scope of analysis to periods after 1999.

As common in time series analysis, I check the presence of unit roots using Dickey-Fuller test to ensure the stationarity of each variable. I transform the variables into logarithmic values and use Hodrick-Prescott (HP) filter to separate the cyclical component from the trend. HP filter is common in real business cycle literature albeit prone to critics. Hamilton (2017) argue that HP filter is prone to spurious dynamic relations which might produce false inference. To avoid false results, I transformed the variables into first difference and find similar results from both approaches. Bank Indonesia also the HP filter as a tool for their data analysis. Thus, I decided to proceed the HP filter approach. The Dickey-Fuller test rejects the presence of unit root at 1% critical value for each variable.

The cyclical components of each macroeconomic variables are compared to the manufacturing production index in Indonesia in figure 4.1. Global commodity price appears to be procyclical with movement in manufacturing production. On the other hand, Brent oil price and consumer price index appear to be counter-cyclical. I decided to the exclude Brent oil price from the model since fluctuations of oil price is parts of the global commodity price. I run a linear regression and find that cyclical component of oil price is significantly correlated to global commodity price. Credit in the manufacturing sector also displays counter-cyclical behavior to manufacturing production throughout the whole period. Is it possible that credit increase during sluggish production activity as a buffer? Nominal exchange rate illustrates no distinct patterns although there are periods where it shows counter-cyclicality, particularly during recessions.

The Structural Vector Auto Regression Model

The empirical model is based on the representation of the structural VAR model which incorporate domestic macroeconomic variables and foreign exogenous variables. In order to examine, the dynamics

(24)

23 of inflation and monetary policy in Indonesia, we will use a system of five variables. This consideration is not optimal considering the presence of multiple relationships that affect macroeconomic variables. However, the five variables that are included in the model are essential to investigate our research question. The empirical model is expected to capture the dynamics in the international environment, apart from the relationship between the Indonesian economy and monetary shocks, and the financial transmission channel of monetary policy. Interpretations of structural VAR model require identifying assumption based on institutional knowledge such as country characteristic or economic theory. These assumptions will be assigned as restrictions to the model in order to identify the appropriate response of dynamics to each variable.

In its standard form, a VAR model represents an equation in linear function which includes lagged values of dependent variable, lagged values of other variable of interest, a serially uncorrelated error term and additionally a constant. The standard model in its structural form:

𝐵₀𝑥_𝑡 = µ_𝑜 + 𝐵₁x_t−1 + 𝐵₂x_t−2+ ⋯ + 𝐵_𝑛x_t−n+ ε_𝑡 where 𝑥_𝑡 = [ 𝑐_𝑡 𝑦_𝑡 𝑖𝑡 𝑚_𝑡 𝑟_𝑡] 𝑥𝑡−𝑛= [ 𝑐_𝑡−𝑛 𝑦_𝑡−𝑛 𝑖𝑡−𝑛 𝑚_𝑡−𝑛 𝑟_𝑡−𝑛] 𝜀_𝑡= [ 𝜀𝑐,_𝑡 𝜀𝑦,_𝑡 𝜀𝑖,_𝑡 𝜀𝑚_,𝑡 𝜀𝑟,𝑡] µ₀ = [ 𝑎1 𝑎₂ 𝑎3 𝑎4 𝑎5] and 𝐵0 = [ 1 . . . 𝑎₁₅ . 1 . . . . . . . . . . . 1 . 𝑎₅₁ . . . 1 ] 𝐵1 = [ 𝛿51 . . . 𝛿15 . . . . . . . . . . . . . . . 𝛿₅₁ . . . 𝛿₅₁] 𝐵𝑛 = [ 𝜔51 . . . 𝜔15 . . . . . . . . . . . . . . . 𝜔₅₁ . . . 𝛿𝜔₅₁]

𝑥𝑡 is a vector of macroeconomic variables which consist of one external variable, the global commodity

price, 𝑐𝑡. The domestic indicators consist of total manufacturing production 𝑦𝑡, consumer price index, 𝑖𝑡

and Bank Indonesia monetary stance, 𝑟𝑡. The price variables and manufacturing production are presented

in index and monetary policy is denoted in percentage points. In the second model, nominal exchange rate 𝑒_𝑡 will replace credit in manufacturing sector. 𝜀𝑡 is a vector of zero-mean structural shocks or structural

innovations, that is serially uncorrelated. 𝑥𝑡−𝑛 is lagged value for all the macroeconomic variables, whereas

matrix 𝐵0 comprises of contemporaneous coefficients and 𝐵1,…, 𝐵𝑛 represents the lagged coefficients of

(25)

24 The model could be rewritten as

𝐵(𝐿)𝑥𝑡= 𝜀𝑡

where 𝐵(𝐿) = 𝐵0−𝐵1𝐿 − 𝐵2𝐿 2 − ⋯ − 𝐵𝑛𝐿𝑛 is the autoregressive lag polynomial. The

variance-covariance matrix of the structural error term is normalized such that

𝐸{𝜀_𝑡, 𝜀_𝑡} = ∑ = 𝐼_𝜀 _𝑘.

First, it means that there are equal numbers of structural shocks and variables in the model. These structural shocks are, by definition, mutually uncorrelated which implies that ∑ _𝜀 is diagonal. Moreover, the optimal number of lag in the VAR model will be determined using different information criteria which are Akaike Information criterion (AIC), Hannah-Quin information criterion (HQIC) and Schwarz Bayesian information criterion (SBIC).

𝐴𝐼𝐶(𝑚) = log 𝑑𝑒𝑡 (𝐸_𝑢(𝑚)) +2 𝑇𝑚𝑘 2 𝐻𝑄𝐼𝐶(𝑚) = log 𝑑𝑒𝑡 (𝐸𝑢(𝑚)) + 2 log 𝑙𝑜𝑔𝑇 𝑇 𝑚𝑘 2 𝑆𝐵𝐼𝐶(𝑚) = log 𝑑𝑒𝑡 (𝐸_𝑢(𝑚)) +log 𝑇 𝑇 𝑚𝑘 2

Where 𝑚 = 0, 𝑘 is the number of variables and 𝑇 is the sample size. Subsequently, we present the reduced-form representation of the structural VAR model. In the reduced-form model, 𝑥𝑡 is expressed as

a function of lagged values of 𝑥𝑡. This equation is acquired by multiplying both sides of the structural VAR

model by 𝐵₀−1

𝐵₀−1𝐵₀𝑥_𝑡 = 𝐵₀−1𝐵₁𝑥_𝑡−1− 𝐵₀−1𝐵₂𝑥_𝑡−2− ⋯ − 𝐵₀−1𝐵_𝑛𝑥_𝑡−𝑛− 𝐵₀−1𝜀_𝑡

which yield

𝑥_𝑡 = 𝐴₁x_t−1 + 𝐴₂x_t−2+ ⋯ + 𝐴_𝑛x_t−n+ µ_𝑡

Where 𝐴𝑛 = 𝐵0−1𝐵𝑛 , i = 1, …., n and µ𝑡= 𝐵0−1𝜀𝑡 . µ𝑡 = 𝐵0−1𝜀𝑡 expresses the relationship between

structural exogenous shocks 𝜀𝑡 in its reduced form µ𝑡that are zero-mean white noise process with

(26)

25 a unique solution for the coefficients in 𝐵−1and run an impulse response functions. The most common approach is to apply the recursive ordering approach or the Cholesky decomposition which is advocated by Sims (1980).

Identification and Estimation

Structural VAR model requires identifying assumptions to enable meaningful structural interpretations. This assumption could be based on institutional knowledge, economics theory or other extraneous constraints on the model responses. Identification allows us to determine the nature of contemporaneous interactions among the selected variables. The rule of thumb for identification requires us to impose (𝑛 ∗ (𝑛+1

2 ) restrictions on the covariance matrix where 𝑛 is the amount of endogenous variable in the system.

Based on our proposed variables, we need to identify 15 restrictions on the structural VAR model in order to achieve exact-identification. The normalization of matrix ∑ = 𝐼 produces five restrictions which allows standard deviations of shocks to be normalized to one. This assumption tells us that structural shocks come from distinct sectors from the economy. Hence, we only need to identify 10 additional restrictions to exact identify the model.

Sims (1980) introduced a recursive scheme by allowing 𝐵0 as a lower triangular. Under this scheme, the

first variable in the ordering could influence all other variables contemporaneously without allowing simultaneous feedback from other variables. The Cholesky decomposition of our model could be presented as follows: 𝐵0 = [ 1 0 0 0 0 𝑎₂₁ 1 0 0 0 𝑎₃₁ 𝑎₃₂ 1 0 0 𝑎₄₁ 𝑎₄₂ 𝑎₄₃ 1 0 𝑎51 𝑎52 𝑎53 𝑎54 1][ 𝑐𝑡 𝑦𝑡 𝑖_𝑡 𝑚_𝑡 𝑟_𝑡] 𝜀𝑡 = [ 𝜀𝑐,_𝑡 𝜀𝑦,_𝑡 𝜀𝑖,_𝑡 𝜀𝑚_,𝑡 𝜀𝑟,_𝑡]

Based on this identification, global commodity price is exogenous to other variables in the system. Global commodity price is included since commodity is one of the main export sectors of Indonesia. Around 35% of Indonesia’s exports, which is valued at approximately $161 billion, and 22% of imports were commodities in 2015. Commodities are a significant part of the Indonesian economy hence, changes in commodity prices could directly influence the Indonesian economy. However, Indonesia is considered as a small economy in the global commodity market due to the small share compared to the total world production. Thus, changes in domestic variables should not influence aggregate global indicators, 𝑎₁₂=

(27)

26 𝑎₁₃= 𝑎₁₄ = 𝑎₁₅= 0.

Manufacturing production index contemporaneously affected by global commodity price through changes in domestic prices. Indonesia is a raw commodity exporter, yet imports of processed commodities are significant to the manufacturing sector. Higher global commodity prices pass through the domestic economy which affects goods such as food prices, fuel, and production materials which are essential to real production. In our baseline model, we restrict manufacturing production from CPI and monetary policy since both variables tend to work with an inertia (Woodford, 1999). We could argue that real production is not instantly affected by changes in monetary policy since production decision is inelastic to sudden changes. Thus, we will restrict contemporaneous response of manufacturing production from other endogenous variables, 𝑎₂₃ = 𝑎₂₄ = 𝑎₂₅ = 0.

CPI is affected simultaneously by commodity prices since commodities form a significant portion of the price level basket. Apart from commodity price, industrial production could produce a contemporaneous effect on inflation. Higher production activity could produce a demand shock which increases wage and spending. As a result, price level could go up simultaneously if prices are flexible. We could argue that nominal rigidities exist but higher economic activity could affect goods with higher elasticity such food prices. The traditional market characteristic in Indonesia allows particular goods such as food to be more elastic compared to other goods. Moreover, food prices form a significant portion of Indonesia’s consumer price basket. Apart from commodity prices and manufacturing production, there is no contemporaneous effect from other variables, 𝑎34= 𝑎35= 0.

Credit as an endogenous variable is contemporaneously affected by changes in production activity. Manufacturing sector use credit, broadly, for working capitals. Thus, changes in production activity will affect the demand for credit. Global commodity price and consumer price index also affect credit contemporaneously. This relationship is comparable to price and money supply. Since credit in nominal value, an increase in price will increase the value of credit in the economy. Lastly, I restrict the credit from Central Bank’s interest rate to allow stickiness in lending rate. Lending rate is the transmission channel of monetary policy to the credit market. Thus, only 𝑎₄₅ = 0

Lastly, interest rate by the Indonesian central bank is the policy variable within the system. Since its independence in 1999, the Central Bank is committed to stabilizing the level of inflation and exchange rate mechanism in Indonesia. In reality, Bank Indonesia also consider policies to support economic growth through accommodative monetary policy. Bank Indonesia could only observe economic growth using

(28)

27 quarterly data. However, manufacturing production index is available on a monthly basis which allows policy makers to use the manufacturing index as a reference. Hence, apart from CPI and global commodity prices, production index and credit could influence monetary policy simultaneously as indicators for the real economy. Alternatively, I will replace the nominal interest rate with the dummy of monetary stance.

In the second model, I replace credit with nominal exchange rate. Goeltom (2008) emphasize the role of exchange rate channel in monetary policy. Exchange rate has a significant influence to inflation which complicate monetary conduct, especially for small economy. Bank Indonesia is also responsible in maintaining the fluctuation of exchange rate after the adoption of flexible exchange rate in 1999. Nominal exchange rate is included in the model as the most endogenous variable in the system.

𝐵0 = [ 1 0 0 0 0 𝑎₂₁ 1 0 0 0 𝑎31 𝑎32 1 0 0 𝑎₄₁ 𝑎₄₂ 𝑎₄₃ 1 0 𝑎51 𝑎52 𝑎53 𝑎54 1][ 𝑐𝑡 𝑦_𝑡 𝑖_𝑡 𝑟_𝑡 𝑒_𝑡] 𝜀𝑡= [ 𝜀𝑐,_𝑡 𝜀𝑦,_𝑡 𝜀𝑖,_𝑡 𝜀𝑟_,𝑡 𝜀𝑒,_𝑡]

In general, there are literature that argued the consistency of Cholesky decomposition (Cooley and Leroy, 1985, Sousa and Zaghini, 2008). Authors argue that the composition of variables in the SVAR model is sensitive to ordering mechanism. Moreover, this ordering mechanism tends to expose informal relationships between variables which disregard economic principles. In principle, for 𝑘 variables, there are 𝑘! permutations of ordering. A higher number of variables could yield more combinations of ordering. In order to avoid this ordering issue, we try to limit our model with five variables, including one exogenous variable. Additionally, I provide the sensitivity analysis for the ordering of the variables and compare the results from different ordering.

Lag Selection and Model Stability

The number of lags that are included in the model are based on the different information criterions. We limit the maximum number of lags into 12 months as a benchmark for the model. Based on the lag criterion analysis in STATA, the results were recorded as follows:

(29)

28 Table 4.2 Lag selection criterion for baseline SVAR model

Ivanov and Killian (2005) provides a comparison of different lag criterion as a benchmark for VAR analysis. Based on their studies, AIC relatively produces the most accurate impulse response estimates for monthly VAR models, compared to HQIC and SBIC. The AIC selection yields two lagged values to be included in the model. In contrast to other empirical VAR papers, our model tends to be smaller for a monthly VAR model. Strongin (1995), Bernanke-Getler (1995) and Leeper (1997) use twelve lagged values while smaller models, for example, while Eichenbaum and Evans (1995), use six lagged values in their model. Regardless of the result of other studies, I use the result from the lag selection criterion to maintain consistency with the empirical exercise.5

In VAR analysis, it is important to check whether the of the variables are covariance stationary. Covariance stationary is fulfilled if their first two moments exist and are independent of time. Lutkepohl (2005) show that if each eigenvalue of matrix A is strictly less than one, then the VAR model is stable. This condition is required to ensure that the impulse-response functions and forecast error variance decomposition have meaningful interpretation. I run the test in Stata and find the highest eigenvalue to be 0.935 which is less than one. Thus, I conclude that the model is stable. Additionally, I also check the presence of autocorrelation in the residuals using the Lagrange multiplier (LM) test. SVAR models assume that the residuals are not correlated. The result of the LM test shows that the p-value at lag two is higher than 0.05. Thus, we can’t reject the null hypothesis of no auto correlation in the residuals. After evaluating the

5_{I run an experiment with 12 lagged value as a comparison to the lag selection criterion. Additional lag increases the}

p-value of the estimates for commodity price to price level and monetary stance, also price level to monetary stance. However, impulse responses from bigger model show higher volatility from high-frequency data. Thus, I favor the smaller lag order to smooth the impulse response function.

Lag LR FPE AIC HQIC SBIC

0 … 1.3e-12 -13.1595 -13.1218 -13.0665 1 1098 2.6e-15 -19.3977 -19.1713* -18.8399* 2 75.978 2.2e-15 -19.5524* -19.1373 -18.5296 3 43.93 2.3e-15 -19.5162 -18.9125 -18.0286 4 25.415 2.7e-15 -19.3699 -18.5775 -17.4174 5 15.532 3.3e-15 -19.1647 -18.1836 -16.7474 6 52.047 3.3e-15 -19.1769 -18.0072 -16.2947 … 12 72.73* 4.7e-15 -18.9716 -16.6699 -13.3002

(30)

29 residuals and stability, we will proceed with the interpretation of the SVAR model results.

5. Results

Simulation of Impulse Response Functions

The impulse response functions of the Indonesian economy under global commodity price shock are presented in Figure 4.1. Impulse responses were derived using the recursive Cholesky decomposition under exact identification of restrictions and estimated coefficients. It is important to note that each variable is measured in level and checked for stationarity in the previous section. The response of each domestic variables is based on one standard deviation shock to global commodity prices, holding other variables constant. This section will focus on the consistency of the impulse response function to economic theory

Figure 4.1 Impulse response functions of one standard deviation in commodity prices using the dummy variable as monetary stance.

(31)

30 Response of consumer price index Response of manufacturing production

A shock to global commodity prices produces contemporaneous effects to domestic variables. Shocks last for approximately 20-30 months before it dies out. We could observe that CPI is positively affected by commodity shocks which peak within 10 months. This surge in price level is followed by a monetary tightening. Apart from CPI, global commodity price could influence the monetary authority to response to price changes. Global commodity prices could be treated as a proxy for future inflation. Sims (1998) suggests that monetary policy makers could react immediately to commodity prices since the data is available on a daily basis. This proposition is illustrated in the records of Bank Indonesia’s Governor meeting. The documents describe that inflationary pressure comes from movement in commodity prices. In period after 2005, this pattern becomes noticeable. Consequently, monetary policy shifts their response to movement in commodity price.

On the sectoral level, manufacturing production display ambiguous response with a negative drop, followed by positive recovery within the first six months. Subsequently, manufacturing production suffers a negative effect, particularly after nine to twelve months, and disappears after approximately two years. The negative effect is expected since higher commodity price cause input for manufacturing production to be more expensive. On the other hand, credit in manufacturing experienced a positive effect after a positive commodity shock. Credit in manufacturing appears to be countercyclical to manufacturing production as evident from figure 4.1 in appendix.

Figure 4.2 Impulse response functions of one standard deviation in commodity prices using Bank Indonesia interest rate as monetary stance

Measuring the effectiveness of monetary policy during the global commodity boom : a case of Indonesia