This Selected Issues paper analyzes the Ukrainian business cycle. It focuses on the price of steel—Ukraine’s major export—and its relation to the economic performance. It establishes a forecasting model for steel prices, which points to downside risks to steel prices. It discusses the effects of steel price fluctuations on economic activity and how policies might then mitigate them. The paper also develops core inflation indicators for Ukraine to better distinguish underlying from temporary influences on inflation.

Abstract

This Selected Issues paper analyzes the Ukrainian business cycle. It focuses on the price of steel—Ukraine’s major export—and its relation to the economic performance. It establishes a forecasting model for steel prices, which points to downside risks to steel prices. It discusses the effects of steel price fluctuations on economic activity and how policies might then mitigate them. The paper also develops core inflation indicators for Ukraine to better distinguish underlying from temporary influences on inflation.

I. Two Aspects of the Ukrainian business cycle

Core Questions, Issues, and Findings

Why is it important to strengthen the analysis of cyclical developments in Ukraine? Since 2000, Ukraine has experienced pronounced and largely unexpected fluctuations in growth and inflation. A better understanding of these events is key to improving and communicating macroeconomic policy responses, including monetary policy reactions as Ukraine moves to more flexible exchange rates and, ultimately, inflation targeting.

Why are steel prices developments important for Ukraine? The production and export of steel is an important pillar of the Ukrainian economy and steel price developments seem closely linked to fluctuations in economic activity. VAR-based estimates suggest that a 10 percent decline in steel prices reduces annualized GDP growth by 1½ percentage points in the quarter the shock occurs, although the effect dissipates relatively quickly.

What is the outlook for steel prices? What can be done to mitigate steel price risks? During the global economic upswing steel prices have risen to well above their long term trend. But with the global business cycle now turning, a model-based forecast suggests that steel prices could come down in the years ahead by a cumulative 35 percent. A flexible exchange rate and counter-cyclical fiscal policy would help to limit the effects of steel price fluctuations on the economy.

Why are core inflation indicators useful for Ukraine? Core inflation indicators decompose large monthly movements in inflation into underlying and temporary factors. This helps clarify inflation trends, allows better forecasts, and underpins sound policy formulation: while disturbances having a lasting effect on inflation call for a policy adjustment, this is not necessarily the case for temporary fluctuations. Core inflation indicators can also be used to motivate and communicate policy decisions.

How has core inflation evolved in Ukraine in recent years? Core inflation indicators calculated for Ukraine, including by the Ukrainian authorities, suggest that, while adverse price shocks have played a role, the acceleration of prices observed in the last two years mainly reflects underlying inflationary pressures.

A. Introduction

1. As Ukraine has emerged from its turbulent initial transition phase, the volatility of inflation and growth has remained high. GDP growth accelerated to 12.1 percent in 2004, fell to 2.7 percent in 2005, before rebounding to 7.6 percent in 2007. Over the same period, inflation has twice reached the mid teens (and now stands at 26 percent), and has dipped as low as 6½ percent (Figure I.1). Understanding these fluctuations, and better forecasting them, is a key challenge in Ukraine’s efforts to gradually move to an inflation-targeting monetary regime.

Figure I.1.
Figure I.1.

Ukraine: Growth and Inflation

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

2. The first section of this paper focuses on the price of steel—Ukraine’s major export—and its relation to the economic performance. Steel prices and growth have evolved in parallel of late (Figure I.2). The paper establishes a forecasting model for steel prices, which points to downside risks to steel prices. It then discusses the effects of steel price fluctuations on economic activity and how policies might mitigate them.

Figure I.2.
Figure I.2.

GDP growth and Steel Prices

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

3. The second section develops core inflation indicators for Ukraine, to better distinguish underlying from temporary influences on inflation. Ukraine’s inflation has increased markedly since the summer 2006, reflecting both underlying pressures and transitory shocks. This section discusses how core inflation indicators may be of use to policymakers and a variety of core inflation approaches. It calculates core inflation indicators for Ukraine using various methodologies, and assesses their properties against the properties of the authorities’ newly constructed measure.

B. Long-term trends in world steel prices and implications for Ukraine 1

4. The production and export of steel is an important pillar of the Ukrainian economy. With steel accounting for more than a third of total goods exports (equivalent to some 12 percent of GDP), real growth has been closely linked to steel prices. During the global economic upswing of the past few years, along with a wider surge in metals valuations, steel prices have risen dramatically, thereby underpinning Ukraine’s mostly favorable export performance and impressive GDP growth. Although steel prices have been holding up, the current global business cycle downturn raises questions about how long this will last. A decline in steel prices would have significant adverse effects on growth and export receipts, and a key issue is how Ukraine’s economy can be made more robust to such global price volatility.

5. This section assesses the risks to steel prices and the policy implications for Ukraine. To this end, section B discusses long-term trends in steel prices and develops a simple forecasting model. Section C analyzes the macroeconomic implications of a sustained decline in steel prices. Section D discusses policy options.

The Evolution of Steel Prices

6. To shed light on the possible direction of future developments, this section analyses past steel price behavior and develops a simple forecasting model. Few useable steel price forecasts are readily available. While futures prices provide an implicit market forecast for several commodities, the futures market for steel is still in its infancy. Steel futures have started trading on the London Metals Exchange this year, but market volumes are still low, and do not yet provide a solid basis for steel price forecasting. Therefore, to get a sense for the outlook for steel prices, past steel price developments are analyzed and a simple model is estimated.

7. A first step in the analysis is to select a steel price series that can serve as a reasonable proxy for the market in which Ukraine operates. There are many different steel prices—the steel market is segmented along product, quality, and geographical dimensions—which do not always move together. A specific price index for Ukraine is not available. To identify a relevant price series, various steel price indices available in Bloomberg were compared with the actual deflator for metals exports in the Ukrainian balance of payments. In the event, the average of CIS hot- and cold-rolled prices matches relatively closely the actual development of Ukrainian steel export prices over the past six years (Figure I.3). However, CIS prices are available only from 1994 and for longer-run comparisons the U.S. series of scrap metals prices also seems to be a reasonable match (although at a monthly frequency it is far more volatile).

Figure I.3.
Figure I.3.

Ukraine: Steel Export Deflator and Selected International Steel Prices 1995-2006

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: NBU, Bloomberg, and staff calculations.

8. Current steel prices are high relative to the long-term downward trend. The scrap metal price series, which is available from the late 1950s, allows an analysis of long-term trends in steel prices. In Figure I.4, the scrap price is shown from 1960 to present, deflated by the U.S. consumer price index. Two points are worth noting. First, despite the recent surges, from 1960 to present this real steel price has trended down, likely reflecting improved production methods and falling costs of steel production. Second, current prices are some 35 percent higher than the trend, although such protracted deviations have also occurred in the past.

Figure I.4.
Figure I.4.

Real Steel Prices 1/, 1960–2006

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: Bloomberg and staff calculations1/ US PPI By Processing Stage: Iron & Steel Scrap NSA
A simple model of steel prices

9. A basic model of steel prices is estimated, relating them to inputs in steel production. For the steel price, the simple average of nominal CIS hot-rolled and cold-rolled steel prices is used. Input variables are nominal price series for iron ore, coal, nickel, and tin, as well as energy prices (using the price of crude oil as proxy). The source for the steel price series is Bloomberg; the other series were taken from the commodities data base that is maintained by the Fund’s Research Department. The data, shown in Figure I.5, have a monthly frequency and cover the period 1994M9–2007M2.2 All data series contain a unit root in their levels, but not in their first differences (Appendix 1).

Figure I.5.
Figure I.5.

Steel Prices and Selected Commodities: 1994M9, 2007M12

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: Bloomberg, IMF WEO Database.

10. Consistent with data properties, an error-correction model is employed, which distinguishes between short-term dynamics and long-run relationships that might exist between steel prices and key production inputs. The broad framework is given by:

ΔZt=ΠZt1+i=1k1ΓiΔZti+μ+εt(1)

where Z is a vector containing the n variables of the system, matrix Π captures information on the long-run relationships among the variables in Z, matrix Γ contains information on the lagged variables (dynamics), and μ is a vector with constants.

11. The model was narrowed down applying a general-to-specific approach. According to the Johansen cointegration test, steel prices are cointegrated with the prices of coal, iron ore and tin, but not with the prices of nickel or energy (Appendix 2, which shows only the final system). As might already be inferred from the absence of evidence for cointegration for these variables, the prices of oil and nickel do not significantly contribute to the explanatory power of the model and are therefore eliminated. It also appears that tin prices can be dropped from the model without significant loss of explanatory power, and they too are therefore eliminated.

12. The remaining empirical model includes only the prices of steel, coal, and iron ore, each with 2-lags.3 The estimated 3-equation system has the following form:

Δs=0.090.04*(s1.44it10.67ct134.44)+i=12β1iΔsti+i=12β2iΔiti+i+12β3iΔct1(2)
Δi=0.460.03*(i0.69st1+0.46ct1+23.90)+i=12β1iΔsti+i=12β2iΔiti+i+12β3Δct1(3)
Δc=0.070.002*(c1.50st1+2.16it1+51.64)+i=12β1iΔsti+i=12β2iΔiti+i+12β3iΔct1(4)

where s denotes the steel price, i the price of iron ore, and c the price of coal (see Appendix 3 for the lag coefficients and the usual statistics).

13. The model’s long-term relationship indicates that steel prices are positively related to iron ore and coal prices, as expected given their role in the steel manufacturing process. In equations 24, the term between parentheses is the error correction term (ECT) that represents the long-run (or equilibrium) relationship between the three variables. Note that this relationship is the same across the system (that is, there is only one cointegrating equation), but that it is normalized on the left-hand-side variable in each of the individual equations. The coefficient of the error correction term (0.04) can be interpreted as the speed of adjustment to this long-run relationship, and implies that it takes about two years for steel prices to reach their equilibrium value.

What the model says about the outlook for steel prices

14. The model is used to project steel prices for 2008–2013. The inputs are the IMF’s WEO projections for iron ore and coal prices. For iron ore, these projections are mostly based on contracted and futures prices, while the outlook for coal is derived from an ARIMA model, complemented with information on long-term trends, developments in related markets, and inventory levels. Since the main concern here is with the broad direction of steel prices (rather than the short-term dynamics), the analysis focuses on the ECT.

15. The model projections indicate a substantial medium-term downward risk to steel prices. Figure I.6 shows the ECT, denoted as equilibrium price, together with the actual price. It is worth noting that, given (historically high) current coal and iron ore prices, the current steel price level is only marginally above its long-run level, as given by the ETC (in contrast with the historical trend in Figure I.4). In other words, current steel prices, though high by historical standards, are in line with inputs. Against this background, it is reasonable to expect that the steel price developments in the near future will be closely tied to those of the inputs. Using the WEO forecasts for the latter, the model shows a further rise of steel prices in 2008, followed by a steady cumulative decline of about 35 percent by 2013. This forecasted price decline broadly coincides with the measure of disequilibrium that was established above on the basis of an extrapolation of the long-term trend.

Figure I.6.
Figure I.6.

Ukraine: Actual and Estimated Equilibrium Steel Price, 1995-2013

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: Bloomberg, WEO, and staff calculations.

Macroeconomic Effects of Steel Price Changes

16. Empirical research suggests that terms-of-trade shocks—such as large steel price changes—contribute to macroeconomic volatility, both directly and because of a strong association between terms-of-trade swings and volatility in fiscal and monetary outcomes (Hausmann, 1999). Large terms-of-trade shocks may also undermine fixed exchange rate regimes. In this section, the potential impact of falling steel prices on key Ukrainian macroeconomic variables is explored.

Channels of impact

17. Steel price changes—or terms of trade shocks more generally—affect the economy via various channels, and affect several key economic variables. In particular, steel price changes are likely to have an impact on:

  • Balance of Payments. In the traditional theory of terms-of-trade shocks, a deterioration of a country’s terms-of-trade reduces its real income and national savings, which, under the assumption of unchanged investments, implies a deterioration of the current account (Harberger, 1950; and Laursen and Metzler, 1950). However, as has been demonstrated in later intertemporal models, the extent to which this mechanism applies depends on whether the terms-of-trade shock is (perceived as) transitory or persistent (Obstfeld, 1982; and Svensson and Razin, 1983). If a terms-of-trade deterioration is perceived as long-lasting or permanent, households and firms may reduce consumption and investment, which will tend to improve the current account, leaving the sign of the effect of the terms-of-trade shock uncertain.

  • Output. As illustrated in the introduction of this chapter, steel price changes also affect real GDP. They do so directly, as steel production, in the case of a price fall, is reduced in response to lower prices and steel demand. But they also affect output indirectly via the negative effects of lower export earnings on incomes and domestic demand. The size, but not the direction, of these effects will depend on whether the shock is deemed permanent or temporary.

  • Inflation. Lower steel prices, by lowering demand relative to potential, will tend to reduce inflation. In addition, inflation may be affected via changes in the real exchange rate, but this depends in part on the exchange rate regime. In the case of a fixed rate regime, a decline in steel prices would leave the nominal exchange rate unchanged and would therefore lower the real exchange rate only gradually and to a limited degree via the reduction in domestic inflation brought about by the change in demand. In the case of a flexible exchange rate, in contrast, a steel price fall would cause an immediate nominal depreciation, thereby raising import prices and inflation. The size of this effect will depend on the extent of the exchange rate adjustment.

Estimation of impact on key variables

18. To assess the likely impact of a steel price decline within the current policy regime, two quarterly vector autoregression (VAR) models were estimated, for the period 1999–2007.4 The first VAR contains four endogenous variables: CIS steel price changes; real GDP growth; changes in the real exchange rate; and CPI inflation—each with two lags. In the second VAR, the same variables are included, again with two lags, but the trade balance (in percent of GDP) is used in place of the real exchange rate.5 The separate VARs are used to isolate the analysis of the impact on the trade balance from that on GDP and inflation. This approach is preferred because Ukraine’s trade balance, even after seasonal adjustment, has shown a very bumpy pattern over the sample period, affecting the responses of the other variables in the model. However, the choice for the two-model approach is mostly presentational; in terms of both magnitude and direction the responses of GDP and inflation are broadly comparable across the two models, and the results are remarkably robust. The models are stationary and have relatively good statistical properties (Appendix 4 and 5).

19. The estimated effects on the Ukrainian economy are substantial. According to the impulse response simulations from the VARs, a 10 percent decline in steel prices (which equals a shock of roughly half a standard deviation) reduces annualized real GDP growth by 1½ percentage points in the quarter of impact (Figure I.7). Thereafter, the effect dissipates quite quickly, to disappear by the fourth quarter, leaving GDP a cumulative ½ to ¾ percentage point lower over the full year. The effect on inflation becomes apparent only in the second quarter after the shock and is largest after three quarters, at which point a 10 percent steel price decline would bring inflation about 2 percentage points lower. The effect peters out after about six quarters. Finally, the effect on the trade balance is largest in the second quarter after the shock, when the trade balance is about 1 percentage point of GDP lower. Thereafter, the negative effect on the trade balance diminishes gradually, with some effect remaining for 2½ years.

Figure I.7.
Figure I.7.

Ukraine: Responses to a One Standard Deviation Shock in Steel Prices

(Estimated over 10 Quarters from Shock)

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: IMF Staff Caluclations

20. And with high steel price volatility, effects could be very pronounced. It needs to be born in mind that the typical steel price shock tends to be significantly larger than the 10 percent that is used in the simulations. Downward price corrections of 30–40 percent took place, for instance, during 1996–99 and 2000–01. And in 2005, CIS steel prices fell by over 30 percent in just five months, even though on that occasion they recovered quickly.

21. These macroeconomic effects of a large fall in steel prices would pose considerable challenges to policy makers in Ukraine.

  • The fiscal position would likely deteriorate as a result of the negative effect on output—an effect that can be countered by contracting public spending, but only at the expense of further weakening economic growth.

  • The sharp deterioration of the trade balance, and associated reserves losses, could potentially lead to pressures on the hryvnia that might make the exchange rate peg unsustainable. An abrupt exit from the peg may in turn trigger adverse effects in the balance sheets of households, corporates, and the banking system.

  • Sharp steel price changes may also directly affect the banking system. This is because terms-of-trade booms are often accompanied by strong increases in domestic deposits that may fuel a domestic lending boom. If the terms-of-trade boom subsequently reverses, a sharp contraction in deposits can have a destabilizing effect on banks (Hausmann, 1999). While Ukraine’s credit boom of recent years appears to have been fueled in large part by foreign inflows, the possibility of destabilizing effects on banks of a slowdown in domestic deposit growth should not be discounted.

Coping with Steel Price Changes

22. Steel price volatility is set to remain a key factor for Ukraine for the foreseeable future. One obvious solution to the dependence on steel is diversification. However, this will inevitably be a long and gradual process, and should primarily be driven by market forces because policy intervention may only frustrate the workings of comparative advantage, thereby lowering Ukrainian welfare. Structural reforms, however, and a further liberalization of markets (notably in agriculture) would allow market forces to work better and thereby promote greater diversification, eventually. More effective hedging against price risks by the steel industry, as hedging instruments become available, may also potentially help smooth adjustments (text box). In the short- to medium term, however, strong exposure to steel price volatility is likely to remain and the macroeconomic consequences have to be managed. The following areas are of particular interest:

Fiscal policy response

23. Counter-cyclical fiscal policies can help cushion the impact of steel price shocks. In the face of transitory shocks to steel prices, countercyclical fiscal policies, that imply some fiscal expansion when steel prices are low, and a contractionary stance when they are high, help mitigate the effects on the economy. Under a nominal balance target—as has effectively been applied in Ukraine in recent years—the government needs to cut spending whenever lower steel prices and the associated lower output growth reduce revenues; symmetrically it tends to step up spending in times of high steel prices and growth. Such policies are sound from a strict fiscal sustainability perspective, but tend to amplify the business cycle and the impact of steel price volatility.

24. A fiscal stabilization fund is a less suitable option. A fiscal stabilization fund is an established way to make room for countercyclical policies in the context of volatile export receipts.6 In such a scheme, a part of government revenues are saved during periods of high export prices and revenues, to be drawn upon in times of low revenues. Several commodities exporters have opted for such stabilization funds, including, Chile (Copper Stabilization Fund) and Russia (Oil Reserve Fund). But stabilization funds are usually implemented in cases where the government is the direct beneficiary of commodity export receipts. In Ukraine, however, where iron ore extraction and steel production are the domain of the private sector, the route of a stabilization fund would be less straightforward, and probably less suitable.

25. Permanent steel price shifts may require structural fiscal adjustment. A permanent shift in real steel prices would lower the net present value of government revenue receipts. If this put fiscal sustainability into question, fiscal adjustment could not be avoided. In such a case, a medium-term expenditure framework may help smooth the adjustment over a number of years and minimize its procyclical effect. Of course, this raises the question of how one might identify permanent steel price shifts (or shocks that decay very slowly). A reasonable indication might be a lack of reversion in input prices toward their real long terms trends over a 5–7 year horizon (i.e. as indicated by futures markets and or long-term contracts).

Exchange rate policy

26. The exchange rate regime is another important factor determining how steel price changes affect the domestic economy. In general, flexible exchange rate regimes are thought to be better placed to deal with terms-of-trade shocks than fixed rate regimes (see e.g., Broda, 2001; and Edwards and Levy Yeyati, 2003).

27. Currency pegs may hinder adjustment to terms-of-trade shocks. Under a fixed exchange rate—as Ukraine de facto applied in recent years—the negative effects on output of a deterioration in the terms of trade will have to be counteracted fully by adjustment of domestic prices and wages—which tends to be a slow process when domestic prices are sticky. In addition, to prevent the currency from depreciating after a negative terms-of-trade shock, the central bank will typically have to use foreign currency reserves to buy domestic currency, thereby reducing the money supply with a further contractionary effect on the economy.

28. Under a flexible exchange rate, in contrast, movements in the exchange rate can help absorb the shock. That is, the required change in the real exchange rate can, at least in part, be borne by the nominal exchange rate, rather than nominal domestic prices and wages. Indeed, with a flexible exchange rate, deterioration in the terms of trade, via its negative effect on output and domestic money demand, leads to lower nominal interest rates and a depreciation of the real exchange rate. This improves the competitiveness of exports, and reduces that of imports, thus providing stimulus to production that helps to offset (part of) the negative effects of the fall in export prices.

29. Gradually allowing for more exchange rate flexibility—and eventually moving to a float and inflation targeting—would help make Ukraine more resilient to steel price fluctuations. In addition to the helpful role that adjustments in the nominal exchange rate can play, a floating exchange rate would also allow for an independent monetary policy that can be loosened when the economy is hit by a negative terms-of-trade shock and output and inflation fall, thus increasing the room for policies to respond to steel price shocks. An inflation target would be needed to provide a nominal anchor.

Financial sector policies

30. Ensuring soundness of the banking sector is also key. In light of the risks that strong cycles in domestic deposit growth pose to the banking system, coping with terms-of-trade volatility also requires a sound and well-regulated financial sector that is well placed to deal with fickle inflows of foreign exchange earnings. Adequate supervision, strong prudent lending standards, and high capital-adequacy and liquidity ratios can foster the resilience of banks and reduce the risk that periodic export-related contractions in deposits lead to a sharp rationing of credit, with a further negative effect on domestic demand. Internationalization of the banking sector—as is taking place in Ukraine—also helps to diversify risks and secure the availability of credit in times of low steel prices.

Hedging Against Steel Price Fluctuations—Steel Futures

On the micro-level, a potential way of insuring against steel price volatility is to buy or sell steel futures. In principle, this technique may be used by both the public sector and by private sector entities. Recently, possibilities for hedging in international markets have been increased with the introduction of steel futures at the London Metals Exchange. And reportedly other exchanges, including the New York Mercantile Exchange, have plans to introduce similar steel-related products. Of course, steel producers can also hedge without the use of derivatives, via long-term sales contracts.

Whether futures will prove useful for Ukrainian steel producers remains to be seen. In terms of government policy, however, it may be useful to consider removing any constraints facing private sector entities who might wish to use them.

C. Core Inflation Indicators in Ukraine 7

31. The key purpose of core inflation indicators is to help policymakers distinguish underlying inflationary pressures, which require a policy response, from transitory effects, which may not. Applied to Ukraine, such indicators point to a significant increase in underlying inflation since the beginning of 2007. They confirm that the recent acceleration of prices (26.2 percent in March-08) mainly reflects mounting underlying inflationary pressures. If not tackled rapidly, the increase in inflation might significantly affect inflation expectations, which seem to be currently drifting up, and feed a vicious inflation-wage spiral potentially harmful for the competitiveness of the Ukrainian economy.

32. This section first reviews recent inflation developments and discusses why core inflation indicators are important and how they are used by central banks. Then, it constructs core inflation indicators for Ukraine. Given data constraints, the uncertainties inherent to the estimation of unobservable variables, and the shortcomings of individual approaches, the assessment relies on a broad range of methods, discussing the pros and cons of each approach. The last section analyzes the properties of the different indicators.

Inflation Developments in Ukraine

33. Inflation has trended up since 2003. Like most other CIS economies, Ukraine experienced a sharp increase in the price level in the early years of the transition. This resulted from waves of price liberalization while conditions for a parallel development in the supply of goods and services were not in place. The structural reforms in the second part of the nineties and the tighter monetary policy stance that supported the introduction of the Hyrvnia helped anchor price expectations. Inflation declined gradually and fell briefly to zero in 2002. However since then it has increased almost continuously and reached 26.2 percent in March 2008 (Figure I.8).

Figure I.8.
Figure I.8.

Inflation trends in Ukraine

CPI inflation (year over year) and 36-months moving average

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

34. The higher inflation rates since 2003 partly reflect a succession of unfavorable shocks to energy and, more recently, food prices (Figure I.9). The price of natural gas imported by Ukraine more than doubled over the period 2005–2007. In addition, since the beginning of 2007, food prices, which represent around 55 percent of the CPI basket in Ukraine, have risen along with global food prices, the effect of which was magnified in Ukraine by a contraction of agriculture production due to unfavorable climatic conditions.

Figure I.9.
Figure I.9.

Inflation components

Year over year rates

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

35. The increase of inflation since 2003 may also be due to mounting underlying price pressures. There are converging signs that the economy is overheating. The acceleration of real wage growth well beyond productivity trends, the increase in capacity utilization rates, and the fast increase in producer prices support this view. The marked shift in the composition of growth toward domestic components and the sharp deterioration of the current account deficit since 2004 (from a surplus of 10 percent to a deficit of 4 percent of GDP) point in the same direction.

36. Distinguishing between underlying and erratic, temporary influences on inflation are particularly difficult in Ukraine. The analysis of Ukrainian CPI data is made complicated by the large volatility of the CPI, including in seasonally adjusted terms. This reflects, inter alia, the significant share of products in the basket subject to climatic developments (energy and food items), and frequent and large changes in administered prices.8 When developments in inflation are strongly influenced by a small number of large changes in a few CPI components, the risk exists that the aggregate measure deviates significantly from its underlying trend (Pujol and Griffiths, 1998).9 The non-normality of the cross-sectional distribution of the monthly price changes in Ukraine (it is leptokurtic and skewed to the right), suggests that this risk might be significant in Ukraine.10

Functions of Core Inflation Indicators

37. Central banks use core inflation indicators for three purposes:

  • Policy formulation. Core inflation indicators help policymakers determine whether movements in consumer prices reflect short-term disturbances or a deeper trend. This is an important input for the formulation of policies, which should focus on the control of underlying price movements and the anchoring of price expectations. Attempts to control relative price movements and to address temporary changes in inflation would likely result in larger inflation and output volatility.11

  • Explaining policies. The monetary authorities use core inflation indicators to motivate their policy decisions and explain them to the public. Clear communication on core inflation developments also helps preserve the credibility of the monetary authorities when headline inflation deviates from its core level in the event of a temporary shock.

  • Limiting second-round effects of temporary disturbances. In the event of unfavorable supply shocks (e.g. to energy or food prices), a greater emphasis by the authorities on developments in core inflation helps focus economic agents’ attention on the underlying trend in inflation. This minimizes the risk of the temporary increase in inflation influencing inflation expectations.

38. Core inflation indicators play an important role in inflation-targeting countries. In almost all inflation targeting countries, the target measure of inflation is based on the CPI. However, most inflation-targeting countries have specified “escape clauses” for their targets, to prevent inappropriate monetary policy reactions to supply shocks. To this aim, virtually all of them have developed and monitor various measures of core inflation (see Roger and Stone, 2005).

39. Core inflation does not have a unique widely accepted definition, and several concepts have been developed (Silver, 2006). All approaches intend to remove the impact of transitory (often supply-driven) shocks on inflation, but the definition of shocks differs across approaches. Some methods rely on purely statistical techniques for the identification of shocks while others appeal to theory-based economic models. Another key difference concerns the degree of complexity: some indicators are fairly simple and understandable by the public; others are considerably more complex.

40. The literature stresses a number of desirable properties for core inflation indicators. According to Roger (1997), core inflation indicators should be (i) significantly less volatile than headline inflation; (ii) available on a timely basis; (iii) unbiased, in the sense that the difference between the historical average of headline and core measure should be small; and (iv) easily reproducible. Wynne (1999) argues that, in addition, core inflation indicators should be (v) understandable by the public; (vi) have some theoretical basis; (vii) not be subject to revisions; (viii) be forward-looking, in the sense of helping to project inflation developments; and (ix) have a track record of some sort, i.e. they should historically not deviate excessively from a trend measure of inflation.

41. There is obvious trade-off among these properties. Namely, there is an incompatibility between those properties related to simplicity (easy to understand and reproduce, not revised) and those calling for an approach consistent with economic theory, which will imply the use of more sophisticated techniques and frequent revisions to the core inflation series. Similarly, sophisticated statistical techniques can be expected to improve core inflation indicators performance on statistical criteria (unbiasedness, volatility, forward-looking orientation), at a significant cost in terms of complexity and reproducibility.

42. The consensus, and actual practice, is to use more than one measure for analytical purposes, but to rely on simple indicators for communication purposes. Most authors (see Roger (2000) and Mankikar and Paisley (2004)) argue that the authorities should rely internally on a wide range of core inflation measures, independently of their degree of complexity, for analytical purposes and the definition of policies. The authorities should however carefully select the core inflation indicators used in their communication with the public. These indicators should be simple and straightforward to explain. This corresponds to the actual practice of central banks typically relying on simple indicators (e.g. exclusion measures) in their publications, but monitoring a wider range of core inflation indicators for analytical purposes (Table I.1).

Table I.1.

Core Inflation Indicators Used by Central Banks

Selected inflation-targeting countries

article image
Sources: Publications of national central banks (inflation reports, monthly bulletins, research papers), Staff.

This list is not exhaustive and several central banks most likely rely on other measures than those mentioned in the table.

Core Inflation Techniques

43. Core inflation indicators are calculated for Ukraine with four popular techniques: exclusion indexes, trimmed mean indexes, generalized dynamic factor model, and the Quah and Vahey (1995) structural VAR approach. The general principles and pros and cons of each technique are discussed in this section. The following sections apply the techniques to Ukraine and assess the properties of the core inflation indicators.

i) Exclusion indexes.

44. This is the most common way to calculate core inflation. The principle of this method is to exclude from the (n) components of the CPI basket a few prices (n-p) which have historically been particularly volatile. The (p) remaining non-zero weight items are re-scaled so that their cumulated weight adds up to one. A key choice concerns the elements to be excluded for the calculation of core inflation. Formally, the measure of underlying inflation is a weighted sum of the monthly price changes in the p non-excluded components:

πtc=i=1pwiπi,t,withp<n,andi=1pwi=1(5)

45. The main advantages of exclusion indicators are that they are simple, available on a timely basis, easily replicable, and understandable by the public. These features make such measures well suited for communication on core inflation developments. Their main shortcoming is that some of the components that are found to be volatile over the whole period may become relatively stable over time while, symmetrically, some components established as not being volatile may become so. There may also be, when the core inflation indicator is used for operational purposes, some public sensitivity to the systematic exclusion of important consumption items, such as food and energy.

ii) Trimmed mean and weighted median.

46. Trimmed indexes and weighted median are constructed by neutralizing the influence of selected CPI items based on whether their price changes are outliers in a given month (Bryan, Cecchetti and Wiggins, 1997). The resulting index is based on the central part of a distribution of “non-extreme” monthly price changes. Contrary to exclusion measures, the goods that are “trimmed” can change every month. Two key choices have to be made in the construction of these indicators: (i) the proportion of the sample to be trimmed; and (ii) whether to trim symmetrically, i.e. whether to remove an identical proportion of the CPI basket in both ends of the distribution of monthly price changes. Formally, core inflation is calculated as in equation (6), with a and b the percentage trimmed in each tail of the distribution of monthly price changes:

πtc=11a+b100*i=1Ia,bwi,tπi,t(6)

47. Trimmed mean indicators share some of the appealing characteristics of exclusion measures. Underlying inflation estimates are available on a timely basis and their replication is easy. But they are not without problems. An implication of (6) is that large price shocks (e.g. large increases in seasonal food prices) will be trimmed while subsequent small and gradual adjustments would not be. The result may be a bias in the core inflation estimates. For this reason, trimmed means and weighted median indicators should be carefully designed, and used at more than one level of trim and in conjunction with other indicators of underlying inflation.

iii) Generalized Dynamic Factor Models (GDFM).

48 This technique extracts the signal from a large number of time-series. It takes into account the information in both the cross-sectional and time dimensions of the sample. It was developed by Forni and others (2000), as an extension of the initial work by Stock and Watson (1989). The model decomposes each CPI component into two sets of unobservable components: a common (principal) component and an idiosyncratic component. Underlying inflation is proxied by the common component of the CPI series, which is estimated as a linear combination of a small number (p) of common factors (u) driving the evolution of the various components of the CPI :12

πi,t=πi,tc+εi,t=k=1pai,k(L)uk,t+εi,t,whereai,kisapolynomialinthelagoperator(7)

49. The GDFM approach has strong advantages. Several studies have demonstrated the good properties of core inflation indicators using this technique (Stavrev, 2006). The main drawback of the GDFM approach is that it is fairly complex, which makes the results difficult to replicate and explain to the public. Core inflation indicators based on factor models are frequently subject to criticism that they are derived essentially from a “black box.” Moreover, the results are sensitive to the choice of the series included in the database and to the number of common factors selected in the model.

iv) Core inflation according to Quah and Vahey (1995).

50. Quah and Vahey (1995) proposed a technique for measuring underlying inflation based on time series analysis of the inflation and output dynamics. The key assumption is that inflation and output are driven by two types of shocks (u1 and u2): respectively, those which have no permanent impact on output (“nominal” shocks) and those associated with a persistent effect on output (‘real’ shocks). A structural VAR is estimated to derive a measure of core inflation using the restriction that core inflation is the component of inflation that has no long-run effect on output (consistent with a vertical long-run Phillips curve).

51. In practice, the estimation of core inflation is made in two steps. First, a bivariate output-inflation VAR model is estimated. Output growth is proxied by the monthly change in industrial production and inflation by the monthly change of the headline CPI. Both series are seasonally-adjusted. The matrix of the VAR residuals is used to estimate the two shocks u1 and u2 mentioned above. The second step consists in computing the core inflation series, by applying the VAR coefficients to the vector of nominal shocks. Formally, the VAR and equation to derive the core inflation indicator can be written as:

Xt=(Δytπt)=D(L)*(ut1ut2)andπtc=i=0Id21(i)ut11(8)

52. The Quah and Vahey approach has economic foundations. Contrary to the first three approaches, which are purely statistical techniques, it uses an economic restriction in the computation of core inflation. Its main drawback is that the economic model used to derive core inflation might be overly simple, in the sense that the assumption of only two types of structural innovations is very restrictive.13 Moreover, the results are sensitive to the choice of the variable that proxies output (industrial production vs. monthly GDP), and to the lag length in the VAR.

Estimating Core inflation Indicators for Ukraine

53. The Ukrainian authorities recently started to publish an exclusion measure of core inflation for Ukraine. The indicator is based on the exclusion of fresh fruits and vegetables; other unprocessed food products (selected cereals, meats and dairy products); housing and communal services (including energy); selected administratively priced bakery products; fuel products and derivatives; and selected administratively priced transport and communication services. The items excluded represent about 45 percent of the CPI basket, a relatively large proportion (Figure I.10).

Figure I.10.
Figure I.10.

Ukrainian’ authorities core Inflation indicator

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

54. This measure is compared to the core inflation indicators calculated using the four techniques described above, with the following specifications:

  • Exclusion measures: Two indicators are constructed: the first one excludes energy, a particularly volatile CPI component in Ukraine; the second one excludes the 20 percent most volatile items of the basket over the period 2001–2007, which is equivalent to excluding energy items, most of fresh food products, and several administered prices.

  • Trimmed indices and weighted median. Two indexes, with extreme choices concerning the proportion of items trimmed, are constructed: (i) a 10–15 percent trimmed index, which removes a small share of the basket; and (ii) an indicator corresponding to the 65th percentile of the distribution of monthly price changes. In both cases, asymmetric trims are used: given the distribution of monthly price changes in Ukraine, any symmetric trim would result in a measure of core inflation that would almost systematically be lower than headline inflation.

  • Generalized dynamic factor model. The GDFM is applied to the series of the monthly changes of the CPI components from January 2000 through December 2007. The low-weight, non-stationary items were dropped from the dataset.14 Based on the principal component analysis of the spectral density matrices of the data, and given the relatively low number of series used, only the first common factor was kept.15

  • Quah and Vahey. The VAR is estimated over the period 2001–2007. Output growth is proxied by the monthly change in industrial production and inflation by the monthly change of the headline CPI. Both series are seasonally-adjusted and stationary. The core inflation series is derived by applying the VAR coefficients to the vector of nominal shocks.

55. While adverse price shocks have played a role in the recent increase in inflation, so have mounting underlying inflationary pressures—that is, core inflation has also risen. Despite the diversity of approaches, the fluctuations in the core inflation series show clear similarities (the cross correlations range from 0.5 to 0.9). All indicators confirm that underlying price pressures have increased significantly since the beginning of 2007 (Figure I.11). According to most measures, core inflation reached a trough in the second half of 2006 in year-on-year terms. Since then, it has increased significantly and is now significantly above its average of the last six years according to all measures.

Figure I.11.
Figure I.11.

Measures of Underlying Inflation in Ukraine 1/

Year-on-year

Citation: IMF Staff Country Reports 2008, 228; 10.5089/9781451839135.002.A001

Source: State statistics Committee of Ukraine and Staff Calculations.1/ This figure shows only four of the six core inflation indicators calculated in the paper. The 10–15 percent trimmed index indicator shows developments very similar to those of the core inflation indicator based on the 65th percentile of the distribution of the monthly price changes. Developments in the core inflation indicator excluding energy have followed very closely the CPI over the recent period.

Properties of the Core Inflation Indicators

56. This sub-section discusses the statistical properties of the core inflation indicators, based on the Roger and Wynne criteria. The core inflation indicators have acceptable statistical properties (Table I.2):

  • They generally do not exhibit a significant bias compared to the CPI. While in the case of trimmed mean indicators the absence of bias is imposed ex-ante, this is not the case for core inflation indicators calculated using the exclusion measures, the GDFM, and the structural VAR approach. Exclusion measures perform worse than the other approaches according to this dimension, reflecting notably the protracted nature of energy and food prices shocks over the period.

  • Most measures are substantially less volatile than the CPI. The trimmed means, weighed median and GDFM indicators perform particularly well, with a reduction in volatility (measured by the standard deviation of the monthly change of the core inflation indicator) approaching 50 percent compared to the headline CPI. 16 The Quah and Vahey approach leads to a volatile core inflation indicator (more volatile than the seasonally-adjusted CPI). This also applies to some exclusion measures. The core inflation indicator recently introduced by the Ukrainian authorities performs well in this dimension.

Table I.2.

Core Inflation Measures: Summary Statistics

Period from Jan-02 to Dec-07 (m-o-m rates)

article image
Source: Staff estimates.

57. Core inflation indicators track more or less closely fluctuations in the CPI. Correlations between the core inflation indicators and the seasonally-adjusted CPI range between 0.5 and 0.9. These numbers should be interpreted with care: while core inflation indicators should show some correlation with the CPI, a very strong correlation would imply that core inflation has little additional content. The correlation between the CPI and core inflation indicators is, however, an important information for the analysis of fluctuations in the core inflation indicators: a given discrepancy between core and headline inflation should not be interpreted in the same way for strongly and poorly correlated core inflation measures.

58. A comparison of core inflation indicators to long-term trend measures of inflation shows significant differences between the indicators. Following Bryan and Cecchetti (1994) and Bakhshi and Yates (1999) the underlying inflation indicators are compared to a moving average of actual inflation over a given period. The deviation is measured by the root mean squared error (RMSE) and the mean absolute deviation (MAD) of the core inflation indicator πct relative to the 24-month centered moving average of headline inflation π*;t (formulas 9). The last two rows of Table I.2 confirm the good statistical properties of trimmed means, weighted median, and GDFM indicators. They also show that the core inflation indicator introduced by the Ukrainian authorities performs better in this dimension than other, more simple, exclusion measures calculated in this paper.

RMSE=t=1T(πtcπt*)2TandMAD=ΣAbs(πtcπt*)T(9)

59. The last step is to assess if the gap between core inflation indicators and headline inflation helps predict the future direction of inflation. To investigate whether core inflation developments include information on the future path of inflation, equation 10 is estimated on quarterly data over the period 2001–2007. In the equation, π denotes the headline inflation rate and πc the core inflation rate. The regression results provide information on how the gap between core and headline inflation is closed over the h following quarters. A positive b parameter implies that headline inflation converges towards core inflation. The regression is estimated with h taking values from 4 to 8 quarters, which corresponds to the relevant horizon for monetary policy.

πt+hπt=a+b(πtcπt)+ε(10)

60. Results show significant differences in the predicting power of the various indicators (Table I.3). The best results are obtained with the GDFM indicators, for all the time horizons considered. Headline inflation also tends to converge towards trimmed mean and weighed median indicators, especially for relatively long horizons (18 to 24 months). Exclusion measures also include some information content of on future developments in headline inflation. This however applies only to indicators excluding a significant proportion of the CPI basket.17

Table I.3.

Regression Results

Estimation period: 2001:Q4 – 2007-Q4

article image
Source: Staff estimates.

61. Overall, the analysis confirms the appealing features of trimmed mean indicators. Table I.4 summarizes the performance of core inflation indicators relative to the Roger and Wynne criteria. While exclusion measures perform very well in the first four criteria, which are mainly related to simplicity, their statistical properties tend to be poorer than for most of other measures. The opposite applies to the GDFM-based core inflation indicator, which has sound statistical properties, but scores low on several dimensions due to its complexity. This makes this indicator clearly better-suited for internal use. The table highlights the limitations of the core inflation estimates derived from the model-based Quah and Vahey approach which, notwithstanding its theoretical underpinnings, tends to have below-average statistical properties. Overall, the comparison highlights the good properties of the trimmed mean and weighted median indicators, which perform well on many dimensions.

Table I.4.

Properties of the core inflation indicators

article image
Source: Staff estimates.

absence of bias is imposed ex ante for trimmed mean and weighed median

Conclusions

62. The analysis points to a significant increase in core inflation since the beginning of 2007. This paper calculates core inflation in Ukraine using various techniques. It shows that the application of standard techniques to Ukrainian data leads to core inflation indicators with acceptable statistical properties. The core inflation indicators confirm that while adverse price shocks have played a role, the acceleration of prices observed in the last two years mainly reflects mounting underlying inflationary pressures.

63. The construction and publication of a core inflation indicator in Ukraine is a step in the right direction. This indicator, based on the exclusion of selected volatile CPI items, has sound statistical properties. The authorities should use it to explain inflation developments and policies. Given the limitations of exclusion-based measures of core inflation, the authorities should develop other indicators for internal analytical purposes. More generally, and considering that core inflation indicators are only one of many instruments that can be used to analyze inflation, other tools (models, indirect indicators of inflation) should be used in conjunction with core inflation measures to get a better gauge of inflation pressures.

Appendix I Unit Root Tests

article image
*, **, *** denotes rejection of the null hypothesis that the series contain a unit root at the 10, 5 and 1 percent significance level, respectively.

Appendix II Johansen Cointegration Test for System Comprising CIS Steel, Coal, and Iron Ore

article image

Appendix III VEC Estimation Output

article image

Appendix IV var Estimation Output (Model including REER)

article image

Appendix V

Appendix 5. VAR Estimation Output

(Model Including Trade Balance)

article image

References

  • Bagliano, F. C., and Morana, C. (2003): “A Common Trends Model of UK Core Inflation,” Empirical Economics, Springer, vol. 28(1), pages 157 -172, January.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Bakhshi, H., and Yates, A. (1999): “To Trim or Not to Trim? An Application of a Trimmed Mean Inflation Estimator to the United Kingdom,” Working Paper No. 97 (London, Bank of England).

    • Search Google Scholar
    • Export Citation
  • Ball, L. and Mankiw, N. G. (1994): “Asymmetric Price Adjustment and Economic Fluctuations,” Economic Journal, 104 (423), pp. 247 -261.

  • Ball, L. and Mankiw, N. G. (1995): “Relative-Price Changes as Aggregate Supply Shocks,” Quarterly Journal of Economics, 110 (1), pp. 161 -193.

  • Brischetto, A., and Richards, A.: (2006): “The Performance of Trimmed Mean Measures of Underlying Inflation,” Reserve Bank of Australia Research Discussion Paper 2006-10. Sydney: Reserve Bank of Australia.

    • Search Google Scholar
    • Export Citation
  • Broda, Christian, 2001, “Coping with Terms-of-Trade Shocks: Pegs versus Floats,” American Economic Review, Vol. 91, No. 2, pp. 376 -380.

    • Search Google Scholar
    • Export Citation
  • Bryan, M. F., and Cecchetti, S.G. (1994): “Measuring Core Inflation,” in Monetary Policy, ed by N. Gregory Mankiw (Chicago, University of Chicago Press for National Bureau of Economic Research), pp. 195 -215.

    • Search Google Scholar
    • Export Citation
  • Bryan, M. F., Cecchetti, S. G., Wiggins, R. L. (1997): “Efficient inflation estimation,” Working Paper 9707, Federal Reserve Bank of Cleveland.

    • Search Google Scholar
    • Export Citation
  • Edwards, Sebastian and Eduardo Levy Yeyati, 2003, “Flexible Exchange Rates as Shock Absorbers,” NBER Working Paper 9867.

  • Forni, M., Hallin, M., Lippi, M., and Reichlin, L. (2000): “The generalized dynamic factor model: identification and estimation,” The Review of Economics and Statistics 82, 540 -554.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Harberger, Arnold, 1950, “Currency Depreciation, Income and the Balance of Trade,” Journal of Political Economy, Vol. 58, pp. 47 -60.

  • Hausmann, Ricardo, 1999, “Managing Terms of Trade Volatility,” PREM Economic Policy Note No. 18, Washington DC: World Bank.

  • International Monetary Fund, 2001, “Stabilization and Savings Funds for Nonrenewable Resources, Experience and Fiscal Policy Implications,” IMF Occasional Paper 205. Washington: IMF. 2001

    • Search Google Scholar
    • Export Citation
  • Laursen, Sven and Lloyd Metzler, 1950, “Flexible Exchange Rates and the Theory of Employment,” Review of Economics and Statistics, Vol. 32, pp. 281 -299.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Mankikar, A., and Paisley, J. (2004): “Core Inflation: A Critical Guide,” Bank of England Working Paper No. 242.

  • Obstfeld, Maurice, 1982, “Aggregate Spending and the Terms of Trade: Is There a Laursen-Metzler Effect?,” Quarterly Journal of Economics, Vol. 97, pp 251 -270.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Pujol, T., and Griffiths, M. (1998): “Moderate Inflation in Poland: A Real Story,” in: C. Cottarelli and G. Szapáry, eds., 197 -229.

    • Search Google Scholar
    • Export Citation
  • Quah D. et Vahey S. P. (1995): “Measuring Core Inflation,” The Economic Journal, September, 105, pp. 1130 -1144.

  • Roger, S. (1997): “A robust measure of core inflation in New Zealand, 1949-96,” Reserve Bank of New Zealand Discussion Paper Series, No. 7 (March).

    • Search Google Scholar
    • Export Citation
  • Roger, S. (2000): “Relative Prices, Inflation and Core Inflation,” IMF Working Paper, No. 58 (March), International Monetary Fund.

  • Roger, S. and Stone, M. (2005): “On Target? The International Experience with Achieving Inflation Targets,” IMF Working Paper, No. 163 (August), International Monetary Fund.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Sédillot, F. and Le Bihan, H. (2002): “Implementing and interpreting indicators of core inflation: the case of France,” Empirical Economics, Springer, vol. 27(3), pp. 473 -497.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Silver, M. (2006): “Core Inflation Measures and Statistical Issues in Choosing Among Them,” IMF Working Paper WP/06/97, International Monetary Fund.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stavrev, E. (2006): “Measures of underlying inflation in the euro area: Assessment and role for monetary policy,” IMF Working Paper WP/06/197, International Monetary Fund.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Stock, J. H. and M. W. Watson (1989): “New indexes of coincident and leading economic Indicators,” NBER Macroeconomic Annual, 351 -393.

  • Svensson, Lars and Assaf Razin, 1983, “The Terms of Trade and the Current Account: The Harberger-Laursen-Metzler Effect,” Journal of Political Economy, Vol. 91, No. 1, pp.97 -125.

    • Crossref
    • Search Google Scholar
    • Export Citation
  • Wynne, Mark A. (1999): “Core inflation: a review of some conceptual issues,” European Central Bank Working Paper # 5.

1

Prepared by David Hofman.

2

The sample size is constrained by the availability of CIS data.

3

The lag order was selected on the basis of the Schwarz information criterion.

4

The current policy regime involves a fixed exchange rate, and a low and relatively stable target for the fiscal deficit. See Chapter II for further information on the fiscal framework.

5

All variables are seasonably adjusted, except for steel prices, which show no seasonal pattern.

6

For further analysis on the role and conditions for the effectiveness of Fiscal stabilization funds, see IMF (2001).

7

Prepared by Laurent Moulin.

8

Food and energy-related products account for about 60 percent of the CPI basket in Ukraine.

9

High variability of price movements across categories can also imply a bias toward higher inflation because of menu effects. While large increases in costs are likely to be passed along to consumers, slight declines in other sectors may not lead to price declines because of the cost of re-printing menus (Ball and Mankiw, 1994, 1995).

10

This study uses a breakdown of Ukraine’s CPI into 53 items, in the COICOP classification. Such data was recently made available on the website of the State Statistics Committee of Ukraine. For the early years of the period some CPI components were estimated using the monthly price changes in the previous classification.

11

The reason is that monetary policy affects inflation with long and variable lags and, hence, is not suited for addressing short-term and temporary fluctuations in inflation.

12

Each inflation component is allowed to react differently to the common shocks.

13

Several authors have extended the Quah and Vahey (1995) model to multivariate common trends model. See notably Bagliano and Morana (2003) and Sédillot and Le Bihan (2002).

14

The low-weight items are excluded because of their possible disproportionate impact on the common factors.

15

The choice of the number of common factors involves some arbitrariness. It was, in line with existing literature on the use of such models, decided to stop at the factor that allows to explain about 50 percent of total data variability.

16

For trimmed means indicators, most of the gain in terms of volatility reduction is obtained after a relatively modest trim. Brischetto and Richards (2006) reached a similar conclusion when studying the properties of core inflation indexes for Australia, the United States, Japan and the euro area.

17

Granger causality tests between headline CPI inflation and the core inflation indicators for various time horizons show evidence of Granger causality from trimmed mean and GDFM core inflation indicators to CPI developments. There is no evidence of Granger causality running from the CPI to core inflation indicators.

Ukraine: Selected Issues
Author: International Monetary Fund