Resilience and Growth in the Small States of the Pacific
Chapter

Chapter 8. Drivers of Inflation in the Pacific Island Countries

Author(s):
Hoe Khor, Roger Kronenberg, and Patrizia Tumbarello
Published Date:
August 2016
Share
  • ShareShare
Show Summary Details
Author(s)
Xuefei Bai, Patrizia Tumbarello and Yiqun Wu 

The Pacific island countries (PICs) are highly exposed to commodity price shocks and exchange rate volatility. This is the result of the structure of these economies: narrow production and export bases and, hence, heavy reliance on imported food and fuel. This makes inflation volatile, posing a challenge for monetary policy. Some countries have also a de facto peg vis-à-vis the U.S. dollar, which makes the domestic currency volatile vis-à-vis the Australian dollar.

During 2000–14 average consumer price index (CPI) inflation in PICs was higher and more volatile than in other developing Asia and Pacific economies (Figure 8.1). Although more volatile, before 2007, average headline inflation in the Pacific region was nevertheless in single digits, in line with global inflation developments (Ciccarelli and Mojon 2010). The large increase in global food and fuel prices, a strong Australian dollar driven by high terms of trade in Australia, and several natural disasters drove headline inflation to a peak of almost 13 percent in the third quarter of 2008 (Figure 8.1). Since then, the global financial crisis put significant downward pressure on inflation through various channels, including lower commodity prices, and exchange rate depreciation exacerbated these pressures.1

Figure 8.1Inflation Development

Sources: Country authorities; and IMF, International Financial Statistics database.

Note: PICs = Pacific island countries.

1 Includes Fiji, Kiribati, Marshall Islands, Micronesia, Palau, Papua New Guinea, Samoa, the Solomon Islands, Timor-Leste, Tonga, Tuvalu, and Vanuatu.

2 Average.

As most PICs are low-income countries with relatively high weights of food and transportation in their CPI baskets (Annex Table 8.1.1) and large shares of food and oil in total imports (Figure 8.2), inflationary spillovers have been challenging for policymakers.2 Although there is extensive literature on the effects of external factors on inflation, there are still few studies on the PICs. Research is needed to answer such questions as the extent to which recent inflation is attributable to changes in international commodity prices and to exchange rate volatility, and how commodity prices and exchange rate pass-through to inflation can be quantified.

Figure 8.2Pacific Island Countries: Import Composition, 2009–13

(Percent)

Source: IMF staff estimates.

To shed some light on these questions, this chapter uses two methodologies to study the inflation pass-through of international commodity prices and exchange rates. We first use fixed-effects panel regressions to examine the inflation dynamics from a regional point of view, and then estimate vector autoregression (VAR) models for each country, taking country-specific characteristics into account.

Inflation Pass-Through: Panel Regressions

A cross-country dynamic panel model was estimated to cast light on the sources of inflation in the six countries in our sample. With the exception of Papua New Guinea, which has a de jure floating exchange rate regime and a de facto crawling-like arrangement since April 2014, five countries in our sample (Fiji, Samoa, Solomon Islands, Tonga, Vanuatu) peg their currencies to a basket of major trade partners’ currencies, including the Australian dollar, U.S. dollar, Japanese yen, and others.

Our model incorporates the following variables: current CPI inflation (dependent variable); lagged inflation, which captures the inflation-inertia component; international food prices; international oil prices; current and lagged exchange rates against the Australian dollar (or the nominal effective exchange rate), which is intended to capture the degree of pass-through from the exchange rate to domestic prices; and output gap, as a measure of excess demand pressures. The output gap is measured as the deviation of actual output from potential output (defined as 1 – GDP/potential GDP). Potential output is estimated as a measure of the GDP trend using the Hodrick-Prescott filter. A positive output gap indicates that output is below trend. Our analysis focuses on the Australian dollar exchange rate. Australia remains the most important source of imports in the region,3 and the nominal effective exchange rate of all six countries has moved closely with their nominal exchange rate against the Australian dollar, especially since the global financial crisis. We use quarterly data from the second quarter of 2003 to the first quarter of 2013. Annex Table 8.1.3 lists the details for data sources.

Specifically, we estimate the following specification using a fixed-effects panel regression:

where CPI is the consumer price index, Pfood is the international food price measured in the domestic currency, Poil is the international oil price measured in the domestic currency, AUD is the bilateral exchange rate against the Australian dollar (local currency per Australian dollar), YGAP is the output gap, and Credit is credit to the private sector measured in the local currency. We also run a separate set of models using the nominal effective exchange rate in the place of AUD. The model was estimated using a differenced generalized method-of-moments estimator that allows past actual values of the dependent variable (in this case, CPI inflation) to affect its current level without being subject to the problems of endogeneity present in dynamic panel analysis.

The econometric results are shown in Annex Table 8.1.2. Group 1 regressions use the base dynamic generalized method-of-moments model, while group 2 regressions estimate the same model but allow for cross-sectional fixed effects. The results show a high degree of persistence in inflation in PICs. In particular, 1 percentage point of past inflation is associated with an inflation increase in the current period in a range of 0.4 to 0.6 percentage point, depending on model specifications.

As expected, headline inflation depends on external factors, including international food prices, oil price inflation, and the Australian dollar exchange rate (or the nominal effective exchange rate). In general, an increase of 1 percentage point in international food price inflation, oil price inflation, and the Australian dollar exchange rate (depreciation) is associated with an increase in inflation of about 0.03, 0.01, and 0.10 percentage point, respectively. In a separate set of models, a 1 percentage point increase in the nominal effective exchange rate (appreciation) is roughly associated with a decrease in inflation of about 0.14 percentage point. Not surprisingly, the impact of international food prices is stronger than that for oil prices, since food is the largest category in the CPI basket for all six PICs.

As for domestic factors, the econometric results confirm that both output gap and credit to the private sector are important determinants of inflation. In general, a 1 percentage point increase in output gap (less demand) is associated with disinflation in a range of 0.02 to 0.07 percentage point, depending on model specifications. A 1 percentage point increase in private credit growth is associated with an inflation increase of about 0.04–0.05 percentage point.

Vector Autoregression Estimation

While the cross-country panel regression results have shed light on the inflation pass-through of global commodity prices and exchange rates in the region, it is also helpful to identify country-specific responses of headline inflation to external shocks. In this section, we conduct the VAR analysis for each of the six PICs, taking country-specific characteristics into account. The VAR system expresses current values of each series as a weighted average of the recent past of all the series, plus a term that contains all the other influences on the current values:

with Bl=A01Al and μt=A01et

where yt denotes the m*1 vector of variables included in the VAR for quarter t, and the error vector ut measures the extent to which yt cannot be determined exactly as a linear combination of the past values of y, with weights given by the constant coefficients v and Bl, l = 1,2,…,p.

The selection of variables in the VAR models is based on the results of the panel regressions and country-specific ordinary least squares regressions. For each country, international oil price (Δlog (Poil)), international food price (Δlog (Pfood)), the Australian dollar exchange rate change against the local currency (Δlog (AUD)), headline inflation (Δlog (CPI)), output gap (YGAP), and credit to the private sector (Δlog (Credit)) are always included in the VAR system. Besides these common variables, we also consider country-specific factors.

For Fiji, since sugar is one of its main exports, and changes in sugar prices have a significant impact on headline inflation in our ordinary least squares regression for the country, we include Δlog (Psugar) in its VAR system. For Papua New Guinea, as Figure 8.3 shows, the inflation pass-through of the U.S. dollar exchange rate is more significant than that of the Australian dollar given the important role played by the U.S. dollar in the global oil market.

Figure 8.3Papua New Guinea: CPI Inflation and the U.S. Dollar Exchange Rate

(Percent)

Source: IMF staff estimates.

Note: CPI = consumer price index.

For this reason, Δlog (USD) is also included in its VAR system. The models are estimated from the first quarter of 2003 to the first quarter of 2013. The number of optimal lags in each VAR is determined by the Schwarz criteria. Based on VAR results, we plot the impulse responses of headline inflation to the shocks of global commodity prices and exchange rates on a two-year horizon.

The identification strategy assumes that shocks in food prices, oil prices, the Australian dollar exchange rate, and domestic credit growth affect inflation in the current and subsequent periods. This is implemented using a Choleski decomposition with the variables in the following order: Δlog(Pfood), Δlog(Poil), Δlog(AUD), YGAP, Δlog(Credit), and finally Δlog (CPI).4 The relationship between reduced errors (μt) and structural errors (et) is presented in A01—an assumed matrix, whose coefficients are estimated.

Responses to International Commodity Price Shocks

The responses of headline inflation to the change in international food and oil prices for each country are presented in Annex Figures 8.1.1 and 8.1.2. The results suggest that the response of headline inflation to the international food price shocks is positive for all countries, as expected. Generally, the response peaks within one year, then tapers off and eventually disappears in two years. Take Fiji, for example: a positive shock to international food price increases inflation with a one-quarter lag and it lasts for two quarters. An increase of the international oil price also has a positive effect on inflation for all six countries.5 Compared with the food price shock, the effect of the international oil price shock on inflation is overall smaller but lasts longer. As the only oil producer in our sample, Papua New Guinea has more resources to fight inflation when oil prices are rising. While the inflation effect of the oil price shock lasts in Papua New Guinea for almost one and a half years, the shock is mainly from the demand side, as more income with a high oil price leads to more consumption and investment. Among nonoil producers, the inflationary impact of the oil price shock lasts only two to three quarters in Fiji, Samoa, and the Solomon Islands, and one year or more in Tonga and Vanuatu.

Responses to the Exchange Rate Shocks

Annex Figure 8.1.3 presents the responses of headline inflation to the movement of the Australian dollar. In general, a depreciation of the currencies of the seven PICs against the Australian dollar does put pressure on headline inflation. However, Papua New Guinea appears to be an exception. There are two possible reasons. First, the country has a relatively flexible exchange rate regime, so the kina exchange rate has been more likely adjusted to accommodate the rising inflation pressure. Second, as the U.S. dollar is the primary oil currency, the inflation pass-through of the U.S. dollar exchange rate is more significant than that of the Australian dollar in Papua New Guinea. This point is supported by the synchronization between inflation and the U.S. dollar exchange rate change, as well as the VAR impulse response results (see Figure 8.3).

Besides the shocks from international commodity prices and exchange rates, credit to the private sector and output gap are also included in the VAR for each country. As expected, the response of headline inflation to a domestic credit shock is generally positive, while the response to an output gap shock is generally negative.

Conclusion

Given the relatively weak monetary transmission mechanism (see Chapter 12), the pass-through of international commodity prices (especially food prices) and exchange rates has been rapid and large in the PICs (see Chapter 13). However, our analysis indicates that these countries have been more resilient to external inflationary shocks since the global financial crisis. This is probably due to price controls and subsidies to food and transportation prices, the implementation of more prudential macroeconomic policies, and more diversified import sources.

Annex 8.1. Responses to External Shocks in Pacific Island Countries

Annex Figure 8.1.1CPI Response to One Standard Deviation Shock from International Food Price

Source: IMF staff estimates.

Note: CPI = consumer price index. Figure shows accumulated response of inflation to international food prices, measured as the ratio of Δlog(CPI) to Δlog(FOOD). The blue lines represent the impulse response function; red dotted lines the upper and lower bounds of the confidence interval.

Annex Figure 8.1.2CPI Response to One Standard Deviation Shock from International Oil Price

Source: IMF staff estimates.

Note: CPI = consumer price index. Figure shows accumulated response of inflation to international oil prices, measured as the ratio of Δlog(CPI) to Δlog(OIL). The blue lines represent the impulse response function; red dotted lines the upper and lower bounds of the confidence interval.

Annex Figure 8.1.3CPI Response to One Standard Deviation Shock from Australian Dollar Exchange Rate

Source: IMF staff estimates.

Note: CPI = consumer price index. Figure shows accumulated response of inflation to Australian dollar (AUD) exchange rate, measured as the ratio of Δlog(CPI) to Δlog(AUDRATE). The blue lines represent the impulse response function; red dotted lines the upper and lower bounds of the confidence interval.

Annex Table 8.1.1Selected Indicators in Pacific Island Countries
CountryPer Capita GDP (2012, US$)Food’s Weight in CPI BasketTransportation’s Weight in CPI BasketCPI Inflation Central Bank Reference1
Fiji4,57840.316.23 percent
Kiribati1,54651.28.1
Marshall Islands3,23635.913.7
Micronesia3,21537.19.3
Palau14,02225.416.8
Papua New Guinea2,09840.913.05 percent
Samoa4,16550.39.33 percent
Solomon Islands1,93646.89.9No specific reference value
Timor-Leste4,14264.36.4No specific reference value
Tonga4,57245.711.96–8 percent
Tuvalu3,57542.08.8
Vanuatu2,99638.411.80–4 percent
Source: IMF staff estimates.Note: … = not applicable.1 All the Pacific island country central banks have price stability as one of their monetary policy objectives, and some set the inflation reference value. However, none of them has a formal inflation target.
Source: IMF staff estimates.Note: … = not applicable.1 All the Pacific island country central banks have price stability as one of their monetary policy objectives, and some set the inflation reference value. However, none of them has a formal inflation target.
Annex Table 8.1.2Drivers of Inflation in Pacific Island Countries
Basic GMMGMM with Fixed EffectsGroup 3 Regression
123412341234
Constant0.005***0.004**0.005***0.004*0.006***0.005**0.007***0.0040.006***0.007***0.006***0.006***
Past inflation [t − 1]0.559***0.495***0.580***0.544***0.447***0.411**0.377***0.514***0.458***0.398***0.527***0.482***
Global oil price inflation [t − 1]0.011**0.005*0.011**0.010***0.010**0.005**0.010**0.01***0.0000.005−00.003
Global food price inflation0.027***0.015**0.024***0.0100.028***0.016***0.026***0.011*0.027***0.025***0.023***0.022***
Depreciation against Australian dollar0.038***0.027***0.045***0.037***
Depreciation against Australian dollar [t-1]0.065***0.093***0.064**0.090***0.117***0.117***
NEERchange−0.046*−0.028−0.05−0.04
NEER change [t − 1]−0.087***−0.165***−0.096***−0.16*** -0.07***−0.12***−0.11***
Output gap−0.021**−0.061***−0.015*−0.071***−0.027**−0.054***−0.024*−0.04***−0.03***
Credit growth0.046**0.037**0.050**0.04**
Source: IMF staff estimates.Note: GMM = generalized method of moments; NEER = nominal effective exchange rate.1 Includes Fiji, Papua New Guinea, Solomon Islands, Samoa, Tonga, and Vanuatu.*p < 0.1, **p < 0.05, ***p < 0.01.
Source: IMF staff estimates.Note: GMM = generalized method of moments; NEER = nominal effective exchange rate.1 Includes Fiji, Papua New Guinea, Solomon Islands, Samoa, Tonga, and Vanuatu.*p < 0.1, **p < 0.05, ***p < 0.01.
Annex Table 8.1.3List of Variables and Sources
VariableContentSource
CPIQuarterly consumer price indexPICs authorities
PfoodAverage-quarter international food price by domestic currencyIMF IFS database and authors’ calculation
PoilAverage-quarter international oil price by domestic currencyIMF IFS database and authors’ calculation
AUDQuarterly bilateral exchange rate against Australia dollar (local currency per AUD)IMF IFS database
YGAPQuarterly output gap, defined as 1-GDP/Potential GDPGDP data is from PICs authorities, potential GDP is estimated by applying the HP filter to actual GDP
CreditEnd-quarter level of credit to private sector (in local currency)PICs authorities and authors’ estimates
FweightWeight of food in CPI basket for each economyPICs authorities
ImratioAnnual imports/GDP, which shows a country’s reliance on imported goods, is used as a proxy for domestic production capacity (higher ratio means weaker capacity)PICs authorities and authors’ calculation
Note: CPI = consumer price index; IFS = IMF, International Financial Statistics; HP = Hodrick-Prescott; PICs = Pacific island countries.
Note: CPI = consumer price index; IFS = IMF, International Financial Statistics; HP = Hodrick-Prescott; PICs = Pacific island countries.
References

    CiccarelliMatteo and BenoitMojon. 2010. “Global Inflation.” The Review of Economics and Statistics92 (3): 52435.

    De GregorioJoseOscarLanderretche and ChristopherNeilson. 2007. “Another Pass-Through Bites the Dust? Oil Prices and Inflation.” Working Paper 417 Central Bank of Chile Santiago.

    • Search Google Scholar
    • Export Citation

    HookeMark. 2002. “Are Oil Shocks Inflationary? Asymmetric and Nonlinear Specifications versus Changes in Regime.” Journal of Money Credit and Banking34 (2): 54061.

    • Search Google Scholar
    • Export Citation

    TumbarelloPatrizia. 2006. “What Drives Inflation in Vietnam? A Regional ApproachSelected Issues Paper. IMF Country Report 06/422 International Monetary Fund Washington.

    • Search Google Scholar
    • Export Citation
1

In April 2009 Fiji devalued its currency by 20 percent to increase trade competitiveness and counter the rapid decline in its international reserves. The Solomon Islands operated a de facto peg against the U.S. dollar before 2009, and has gradually allowed greater exchange rate flexibility to safeguard reserves.

2

The potential unfavorable effects of inflation on poverty and growth are well documented in the literature. Rising inflation can increase income inequality because it is essentially a regressive tax with an adverse impact on the poor. And rapid inflation can quickly erode the purchasing power of the poor because they have limited cash and bank deposits, and generally lack job security and financial assets to provide an adequate hedge against inflation (Tumbarello 2006). As a result, this could lead to severe economic and social problems.

3

It should also be noted that the import sources of PICs have become more diversified over time. In 2012 some 30 percent of import goods came from Australia, down from 43 percent in 1995.

4

The Fiji model includes the price of sugar; the Papua New Guinea model includes the U.S. dollar exchange rate.

5

For the pass-through from the price of oil to inflation, see De Gregorio, Landerretche, and Neilson (2007) and Hooke (2002).

    Other Resources Citing This Publication