Journal Issue

Philippines: Selected Issues

International Monetary Fund
Published Date:
May 2006
  • ShareShare
Show Summary Details

IV. Second-Round Effects of the Oil Shock On Inflation35

A. Introduction

1. Since early-2004, a series of supply shocks has increased inflation above target in the Philippines and threatened to drive inflation expectations higher. Foremost among these shocks has been the surge in international oil prices, which rose 41 percent in 2005, after rising 31 percent in 2004 and 16 percent in 2003. In responding to these shocks, the Bangko Sentral ng Pilipinas (BSP) has followed the standard inflation targeting approach, namely, accommodating immediate effects on the price level, but increasing policy rates on indications of a possible rise in expectations of inflation. While the BSP has been able to strike a balance—repeated tightening over the past year has helped put inflation on a downward trajectory—it is in practice often difficult to distinguish the immediate impact of supply side shocks on inflation from a more fundamental shift in medium-term inflation expectations. This distinction is especially hard to make in the Philippines, where a series of supply side shocks has pushed inflation substantially above target in both 2004 and 2005.

2. This chapter assesses the impact of the recent surge in international oil prices on medium-term inflation expectations in the Philippines. Conceptually, the impact of an oil shock on inflation can be thought of as consisting of first-, second-, and third-round effects. First-round effects consist of (i) direct pass through from international oil prices to domestic fuel prices, where the speed of this channel depends on whether there are fuel subsidies and on the competitive structure of the energy sector; and (ii) indirect effects which capture the role of fuel products as intermediate inputs in production. Second-round effects occur when the first-round rise in inflation “spills over” into perceptions of medium-term inflation, as households and firms adjust their price and wage setting behavior. Third-round effects can arise from a country being a net importer of oil, so that a rise in oil prices reduces domestic incomes and spending. In addition, higher oil prices may also lower growth in key trading partners. Lower activity may in turn reduce medium-term inflation pressures.36

3. To assess the impact of the oil shock on inflation expectations, this chapter uses core inflation to capture the second-round effects of the recent surge in oil prices. Since survey data on inflation expectations in the Philippines are limited, a proxy for second-round effects is needed. Measures of core inflation are an obvious candidate, as they aim to strip out the first-round effects of supply shocks. However, there exist many different ways to measure core inflation, so this chapter first reviews a widely-used list of core inflation measures and compares their ability to predict changes in headline inflation. It then uses the most successful measure of core inflation to quantify the second-round effects of the recent surge in oil prices on inflation.

4. This chapter is structured as follows. Section B reviews recent inflation developments and identifies a series of supply shocks that have pushed inflation above target in the last two years. Section C introduces some widely-used measures of core inflation and compares them to headline inflation in the Philippines. Section D assesses the relative performance of these measures in predicting changes in headline inflation, while section E uses the most successful measure to quantify second-round effects from the oil shock in the Philippines. As a comparison, similar calculations are performed for Indonesia and Thailand. Section F concludes.

B. Recent Inflation Developments

5. Since 2004, inflation developments in the Philippines can be characterized by three discrete episodes. Figure 1 plots headline inflation in year-on-year (y/y) terms, month-on-month (m/m) terms (annualized, seasonally adjusted), and the inflation trend—the simple mean of m/m inflation over a particular period. The trend measure is a more timely measure of changes in inflation pressures, since y/y inflation is a backward-looking indicator. Figure 1 shows that the recent path of inflation can be broken down into three discrete episodes. Episode 1 saw trend inflation rise from 3.8 percent in 2003 to 6.5 percent from January to May 2004. Episode 2 saw the inflation trend rise still higher, to 9.5 percent from June to December 2004, while Episode 3 encompasses all of 2005, when trend inflation fell back to 6.4 percent, just above the inflation target of 5–6 percent.

Figure 1:Headline Inflation

(Jan. 2003 - Dec. 2005, in percent)

6. These episodes can be traced to a series of supply side shocks, the most recent of which is the surge in international oil prices. Food price inflation accelerated sharply in early 2004, as higher cost of animal feed drove meat prices higher. In addition, a shift in demand away from poultry due to the Avian Flu virus pushed fish prices up as well. The rise in international oil prices in 2004 led to fare increases in the transport sector in mid-2004. These fare hikes caused the land transport component in the CPI to jump sharply, in part a catch-up effect after being flat for several years. Hikes in electricity tariffs caused the light component in the CPI to jump in late 2004. In 2005, food price inflation moderated as meat imports rose and planting conditions improved, while first-round effects from the surge in oil prices caused the energy component in the CPI to rise sharply towards the end of the year, with knock-on effects on fares in the transport sector. Starting in November, fuel prices began to fall, however, even with implementation of the first stage of the VAT reform law on November 1. As a result, the contribution of energy-related components to CPI inflation was negative in December. Figure 2 shows the contributions to annualized m/m inflation of key CPI items, along with their CPI weights, and the USD price per barrel of Dubai Fateh crude.

Figure 2.Contributions to Headline Inflation

(Jan. 2004 - Dec. 2005, annualized m/m, in percent)

7. With the rapid succession of supply side shocks, expectations of inflation one-year-ahead rose in the course of 2004 and have remained high. Figure 3 plots inflation expectations one-year-ahead, using survey data from Consensus Economics. It shows that inflation expectations moved up sharply in the course of 2004 and remained high in 2005. Unfortunately, these data are only available from 2000 and thus do not lend themselves to a systematic exploration of how the oil shock might affect medium-term inflation expectations. In addition, the consensus forecasts tend to lag actual inflation data, so that they may not have much forward-looking content. As a result, this chapter uses core inflation to assess the second-round effects from the surge in oil prices on inflation.37

Figure 3.One-Year-Ahead Inflation Expectations

(Jan. 2002 - Dec. 2005, in percent)

Source: Consensus Economics.

C. Measures of Core Inflation

8. Measures of core inflation aim to help policy makers identify shifts in medium-term trend inflation. The common rationale underlying all measures of core inflation is to eliminate or discount sharp, quickly reversed movements in prices or one-off shocks that create short-term volatility in headline inflation. Core inflation measures thus tend to be less volatile than headline inflation. There are two basic approaches to constructing core inflation measures. The first, the exclusion-based approach, eliminates certain items from the CPI that are thought to be subject to frequent supply shocks that are in many cases self-reversing. The official measure of core inflation in the Philippines, which eliminates certain food and energy components in the CPI, falls into this category. A second approach is to eliminate outliers or re-weight the CPI basket according to statistical criteria. These statistics-based measures of core inflation include the trimmed mean, weighted median, both of which are also monitored by the BSP, but there are also many other possible approaches.

9. This chapter constructs five measures of core inflation for the Philippines, all of which exhibit lower volatility than headline inflation. These measures are:

  • Official measure of core: this exclusion-based measure eliminates rice, corn, fruits and vegetables, LPG, kerosene, and oil, gasoline and diesel from the CPI basket, with the total weight of excluded CPI items amounting to 18.4 percent in the 2000-based CPI. This measure has the advantage that it is easily communicated to the public, since unlike some of the other measures considered below, the excluded items remain constant over time.
  • Trimmed mean: this measure eliminates CPI components with unusually large or small inflation rates in a given month. The chapter follows the BSP, which measures the trimmed mean as the central 70 percent of the distribution of CPI components, weighted according to their weights in the CPI basket. This measure implicitly assumes that outlying price changes are more likely to reflect relative price changes with no impact on medium-term trend inflation.
  • Weighted median: this measure is an extreme case of the trimmed mean, where only the 50th percentile of the weighted distribution of CPI constituent inflation rates is used. The chapter follows the BSP in constructing this measure, where the median inflation rate corresponds to the cumulative CPI weight of 50 percent.
  • Persistence-based measure:Blinder (1997) argues that core inflation is the “persistent” part of inflation that is correlated with future inflation. Cutler (2001) was the first to construct such a measure for the U.K. This chapter constructs a similar measure for the Philippines, estimating an autoregressive (AR) process with 12 lags based on m/m inflation rates for all CPI components and then double weighting CPI constituents according to their weights in the consumption basket and the combined size of the AR estimates, with items with insignificant or negative serial correlation receiving a weight of zero. Items with high persistence include rentals, kerosene, land transport, and oil, gasoline and diesel, while items with low or insignificant persistence include fruits and vegetables, light and water charges.
  • Adaptive expectations:Cogley (2002) advocates a measure of core based on adaptive expectations. Cogley shows that this measure of core is much less volatile than standard measures of core for the U.S. and finds that it outperforms other measures in forecasting changes in headline inflation. This measure is given by:

where πt* is core inflation in period t, πt is headline inflation in period t, and g0 is a smoothing parameter that determines the half-life of inflation shocks. Since g0 is an unobserved parameter, it is assumed that the half-life of inflation shocks, based on m/m inflation data (annualized, sa) lies between 1 (g0 = 0.69) and 6 months (g0 = 0.12), where the half-life is given by In(2)/g0.

Table 1 compares the means and standard deviations for each of these measures of core to headline inflation in the Philippines. It shows that each of these measures exhibits lower volatility than headline inflation, a desirable property for measures of core inflation. However, for some measures the reduction in volatility is larger than for others. At one extreme, the persistence-based measure has a standard deviation of 3.8 percent, while at the other extreme the adaptive expectations measure with g0 = 0.12 has a standard deviation less than half that of headline inflation. In addition, Table 1 shows that some measures of core tend to be biased relative to headline inflation. The trimmed mean and weighted median measures tend to be biased downward, due to the right-skewness of the underlying inflation data, a common problem associated with these measures. Meanwhile, the persistence-based measure and the adaptive expectations measure with g0 = 0.12 are biased upward.38

Table 1.M/M Inflation in Percent (annualized, sa): Feb. 1994 - Oct. 2005.
MeanStandard Deviation
Official Core6.03.6
Trimmed Mean4.32.9
Weighted Median4.53.0
Persistence-Based Core6.63.8
Adaptive Expectations (g0 = 0.69)6.13.5
Adaptive Expectations (g0 = 0.12)6.52.1

D. Choosing Among Measures of Core Inflation

10. The usefulness of a core inflation measure depends on its ability to predict changes in headline inflation. The inflation target in the Philippines is formulated in terms of headline inflation. As such, the primary role of core inflation is to inform policy makers about the future path of headline inflation. To compare how well the various measures of core inflation perform this task, the following equation from Cogley (2002) is estimated:

where πt* represents core and πt stands for headline inflation. This equation assesses whether deviations of headline above core inflation (πtπt*)>0 in period t are on average followed by a deceleration in headline inflation H periods ahead (πt+Hπt). If this is the case, then βH should be negative and significantly different from zero. Indeed, the definition of Bryan and Cecchetti (1994) for core inflation allows one to go further. They define core inflation as “the component of price changes that is expected to persist over medium-run horizons of several years.” In other words, core inflation can be written as πt* =Etπt+H where H is a suitably long forecast horizon. This means that if core is unbiased relative to headline inflation, then it should be the case that αH = 0 where this restriction follows from the fact that (πt+Hπt) and (πt-πt*) are (approximately) mean zero. In addition, if πt* =Etπt+H, then it should be the case that βH = -1, where this restriction measures whether core inflation correctly filters out the transient components in headline inflation. In the regressions below, both parameters are freely estimated and tests are then performed to assess whether the restrictions hold. In order to decide between two measures of core inflation that are similar in terms of αH and βH, the competing measures are also assessed in terms of the R2 in these regressions. Core measures that account for a greater percentage of subsequent changes in inflation filter out more transient variation and are thus preferable to other measures.39

11. The adaptive expectations measure of core outperforms other measures in terms of explaining subsequent variation in headline inflation. For a sample from February 1994 to October 2005, most candidate measures of core tend to exhibit some bias relative to headline as the prediction window lengthens from one to 36 months ahead (αH ≠ 0), while predictive content is relatively good across all measures (βH = -1). The adaptive expectations measure of core with g0 = 0.12 stands out, however, in terms of explaining subsequent variation in headline inflation. Figure 4 shows that the R2 of this measure peaks 24 months ahead at around 60 percent, substantially above any other measure. This suggests that the adaptive expectations measure with g0 = 0.12 does a better job than other measures at filtering out transient components from headline inflation. A technical annex reports αH and βH estimates for the various measures of core inflation.

Figure 4.Explanatory Power (R2) of Core Inflation Measures

(One to thirty-six months ahead)

E. Second-Round Effects from the Oil Shock

12. Second-round effects from rises in international oil prices have historically been large in the Philippines by regional standards. This section estimates a VAR in the In differences of the USD price per barrel of Dubai Fateh crude, the nominal exchange rate and the adaptive expectations measure of core. This estimation is performed for Indonesia, the Philippines and Thailand for a sample from 1970 Q2 to 2005 Q3 using quarterly data. The smoothing parameter is assumed to be consistent with a half-life for inflation shocks of 4 quarters (g0 = 0.17), the same value as used by Cogley (2002) for quarterly data. Impulse responses are then generated for a 10 percent permanent shock to the price of Dubai Fateh crude, using this ordering. Table 2 summarizes the results, showing the additions to y/y headline and core inflation rates resulting from this shock. Table 2 shows that the oil shock adds up to 0.6 percentage points to y/y core inflation, substantially above Indonesia and Thailand, in part a reflection of fuel subsidies in the latter two countries, which have only recently been reduced or eliminated.

Table 2.Additions to y/y Headline and Core Inflation Rates (in ppts).

13. Given the long sample period, it is likely that these estimates for second-round effects are an upper bound for the Philippines. In particular, it is likely that increased competition, greater fiscal discipline and the BSP’s enhanced inflation fighting credentials have in recent years reduced the scope for second round effects. Moreover, the greater pass-through into inflation from changes in international oil prices also points to a strength of the Philippines which has been highlighted recently: the absence of fuel subsidies.

F. Conclusions

14. A 10 percent permanent rise in the price of Dubai Fateh crude is estimated to add up to 0.6 percent to y/y core inflation in the Philippines. This estimate uses the adaptive expectations measure of core to proxy for medium-term inflation expectations, as this measure outperforms other measures in terms of explaining subsequent variation in headline inflation. An added advantage of this measure is that it is easy to construct across countries. Estimated second-round effects are shown to be substantially smaller in Indonesia and Thailand, although these estimates may be biased downward by fuel subsidies, which have only recently been reduced or eliminated. Indeed, going forward, the estimated magnitude of second-round effects for the Philippines is likely an upper bound, as greater competition, ongoing fiscal consolidation and enhanced monetary policy credibility will likely reduce the scope for second-round effects over time.

ANNEX I Technical Annex

Figure A1.Predictive Power of Official Core Measure: βH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A2.Predictive Power of Trimmed Mean: βH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A3.Predictive Power of Weighted Median: βH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A4.Predictive Power of Adaptive Expectations: βH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A5.Predictive Power of Persistence Measure: βH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A6.Bias of Official Core Measure: αH.

(Point estimate bolded, 90 percent confidence interval dotted)

Figure A7.Bias of Trimmed Mean: αH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A8.Bias of Weighted Median: αH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A9.Bias of Adaptive Expectations Measure: αH.

(Point estimates bolded, 90 percent confidence interval dotted)

Figure A10.Bias of Persistence-Based Measure: αH.

(Point estimates bolded, 90 percent confidence interval dotted)

    BlinderA.“Commentary,”Federal Reserve Bank of St Louis ReviewMay/June1997.

    BryanM. and S.Cecchetti“Measuring Core Inflation” In Monetary Policyedited byN. GregoryMankiw pp. 195215Chicago: University of Chicago Press, 1994.

    • Search Google Scholar
    • Export Citation

    CogleyT.“A Simple Adaptive Measure of Core Inflation,”Journal of Money Credit and BankingVol. 34No. 1 (February2002) pp. 94113.

    • Search Google Scholar
    • Export Citation

    CutlerJ.“A New Measure of ‘Core Inflation in the U.K.,’”MPC Unit Discussion Paper No 3 Bank of EnglandMarch2001.

    GuinigundoD.“An Official Core Inflation Measure for the Philippines,”Bangko Sentral ReviewJuly2004.

    Reserve Bank of New Zealand“Monetary Policy Statement,”September2005 (


Prepared by Robin Brooks (


This classification draws on the September 2005 Monetary Policy Statement of the Reserve Bank of New Zealand.


The BSP’s inflation report, which was first published in Q4 2001, also includes a quarterly survey of private sector inflation forecasts.


For a survey of the BSP’s measures of core inflation, see Guinigundo (2004).


To ensure that these estimates are robust to a trend decline in inflation, the regressions include a linear time trend.

Other Resources Citing This Publication