Causality Among PPI-CPI-Exchange Rates in Romania

Causality among PPI-CPI-Exchange Rates in Romania:

A Frequency Domain Approach

Aviral Kumar Tiwaria, b

a Montpellier Business School

2300 Avenue des Moulins, 34185, Montpellier, Hérault, LR, France

and

bRajagiri Centre for Business Studies

Rajagiri Valley Campus, Kochi, India

Email:

Faridul Islam

Department of Economics, Morgan State University

1700 Cold Spring Lane, Baltimore, MD 20251-0001, US

Email:

Mihai Mutascuc, d

c ESCE International Business School

10 rue Sextius Michel, 75015 Paris, France

Email:

and

d Laboratoire d'Economie d'Orléans, UMR7322

Faculté de Droit d'Economie et de Gestion, Université of Orléans

Rue de Blois - B.P. 6739, 45067, Orléans, France

Email:

Abstract

This paper applies the frequency domain approach to monthly Romanian data to examine the direction of causality among Consumer Price Index (CPI), Producer Price Index (PPI) and exchange rate. Theoretically, exchange rate can be an important determinant of domestic inflation in a globalized world. And yet, the topic has not received much academic scrutiny. The methodologyused here is more general and can capture nonlinear effect and cyclical nature in the cause and effect relationship. We find that the recent information on CPI contains useful information in improving forecasts performance of PPI in short-, medium-and long-run cycles. This reconfirms that PPI is not a useful predictor of CPI movements. However, we observe a very close peak in the business cycles in CPI and PPI. Medium run business cycles emanate from nominal effective exchange rate (NEER); and short-medium and long-run cycle from CPI. Evidence shows bi-directional long-run causality between CPI and NEER; and uni-directional short- and medium run causality from CPI and PPI to NEER. The peak of business cycles is observed in both cases at the same frequencies.

Keywords: Exchange Rate, PPI, CPI, Forecast, Romania

JEL Codes: E31, E37, F31, C32, C53

1. Introduction

Does exchange rate contain valuable predictive power over consumer prices? This question has been gaining momentum in as globalization, and with it trade, takes a center stage in the pursuit for economic growth. While a sizeable literature on the CPI-PPI nexus exists, the role of exchange rate as a player in the domestic inflation has received little attention from the academia. The paper examines the dynamic interaction among Consumer Price Index (CPI), Producer Price Index (PPI) and exchange rate by implementing the frequency domain approach to the Romanian monthly data. Our results show that recent information on CPI is useful in improving forecasts performance of PPI inflation in the short-, medium-and long-run cycles. This reconfirms that PPI is not a useful predictor of CPI movements. Further, evidence suggests bi-directional long-run causality between CPI and NEER and uni-directional short- and medium run causality from CPI and PPI to NEER.

In a competitive market, product prices are determined by the marginal cost, mainly wage costs although the final price, inclusive of any mark up above this cost, is set by the demand factors.A surge in the mark-up (reflected in higher price) also influences wage cost because workersare interested inreal, as opposed to nominal wages. In this situation, causality likely runs from consumers’ to produces’ prices. Economic theory predicts that movement in producers’ prices eventually spills over to consumers’ prices. Ashock to supply or production side can also lead to higher product prices through increasedinput prices, i.e.PPI may also Granger causes CPI – the reverse causality.

Theprice of input can rise due to an increase in the cost of (a) domestically suppliedinputs for any reason; and (b) imported inputs,due to exchange rate shock,i.e. an unfavorable terms of trade, inter alia; and (c) a combination of the two. Thus, it is through trade that the exchange rate can become important in domestic inflation. Against the backdrops of a potential links, it is plausible that information on producers’ price and exchange rate contains valuable predictive power over consumer prices.

Several reasons justify causality among the series. (a) Production process in the retail sector uses current domestic or imported materials as input, but adds value with lag(s). (b) The input prices are determined not only by the interaction of domestic demand and supply, but also by the importcontents, which depends on the nominal exchange rate. (c) The level of indirect taxes, marginal cost of retail producers, interest rates etc., play their respective roles. Experience suggests that the relationship between producers’ and consumers’ price is not as obvious as it might appear on the surface. Studies based on United States data suggest that the PPI do not have much predictive power over the CPI (Clark,1995;Blomberg and Harris, 1995). However, Dion (1999)found some support for a relation between the series for Canada, although the paper did not focus onthe predictive power of PPI.

On the methodological front, the literature on the CPI and PPI nexus shows that due consideration was not given to thetime series properties of the data. Most of the earlier studies have examined Granger-causality between the PPI and the CPI by using Vector Autoregressive (VAR) models in first differences. The procedure assumes that price level series isI(1) such that inflation rates are stationary; and PPI and the CPI are not cointegrated.[1]If the assumptions are not met, the VAR model in first difference is not the appropriate. If the pricelevel series are I(2), then causality analysis should take this property into account, which adds to complexity. Also, in the presence of cointegration, the VAR model in first or second difference of series may be inappropriate due to omitted-variable bias caused by the exclusion of the Error Correction Mechanism (ECM) term (Engle and Granger, 1987). The linear specification may also cause bias in estimatesand producemisleading conclusions.

Cushing and McGarvey (1990) developed a theoretical model to examine causality between wholesale prices and consumers’ prices. They argue that wholesale prices will lead consumers’ prices independently because primary goods are used as inputs of the consumption goods with lag. Shahbaz et al. (2009) points out that the production of final goods in each period uses the primary inputs with lags. They argue that the link works from supply-side shock to primary goods market and thus influences wholesale and consumer prices in subsequentperiods. Colclough and Lange (1982) developed a modelbased on derived demand. Theypoint out that the opportunity cost of resources and intermediate materials is captured by the production cost, whichinfluences the demand for final goods and services.This implies that consumers’ prices lead producers’ prices. The process bywhich consumers’ prices might determine producers’ prices was further examined and illustrated by Caporale et al. (2002). They document that consumers’ prices may cause producers’ prices through labor supply channel, which may also work through supply shocks in the labor market, provided wage earners in the wholesale sector want to preserve the purchasing power of their incomes. This can happen due to lags, as noted above. Perhaps, the wage-setting process and the mechanism by which expectations are formed have some role. The argument, in part explains why theoretically, there may be unidirectional or bidirectional Granger-causality between producers’ and consumers’ price. Therefore, either producers’ or consumers’ price, or both may contain information to predict each other, and are clearly left to empirical determination.

The objective of the paper is to examine the direction of causality among CPI, PPI and exchange rate in Romania, an emerging economy. The paper usesmonthly data on the series from 1990:01 to 2009:03. We decomposecausality by time horizons to demonstrate intermediate Granger-causality. The findings have important bearings not only on how Romania might craft inflation related policies, but also forother emerging nations opening up their economies. If CPI Granger-causes PPI at intermediate frequencies, CPI should be a leading indicator in policy discussions;particularly,in setting intermediate macroeconomic targets pertaining to monetary or fiscal policies. The policymakers may use the knowledge to stay better prepared toaddress the negative consequences of producers’ price inflation at the intermediate horizons.

This study contributes to the existing literature in several ways. First, methodologically,the approach is superior to other approachesbecauseit can capture the nonlinear effect and the cyclical nature of the cause and effect relationship among the series. The authors are unaware of any similar study in the context of Romania. Second, in departure from the conventional one-shot measure based on linear models a la, Toda and Yamamoto (1995), VAR, VECMor nonlinear models ofBaek and Brock (1992), the approach can detect both strength and direction of causality. Thirdthe paper shows the cyclical nature of causalityamong the series over the frequency bands (i.e., period of cycles are defined with frequency bands), which also may be due to market imperfections and/or frictions in the economy.[2]Fourth, results of this study show the weakness of the simple supply-demand relationship and point to the need for a framework that can capture cyclical relationship. Cycles in the relationship between CPI and PPI in Romania appear to originate from supply and demand. Fifth, from the policy perspective, the results show that shocks originating from the labor supply channel and wage-setting process as well as the mechanism by which expectations are formed play major role in aneconomy. This information can helpRomanian policymakers todefine inflation targets for anintermediate range in terms of CPImeasure.Knowledge on the relation between producer prices and exchange rates can provide the monetary policy authorities and the central banker early signal on the dynamics of “cost-push” shocks, and help to improve the forecast performance of CPI-a major macroeconomic policy goal for an economy.

The rest of the paper is organized as follows. Section 2 reviews the extant literature. Section 3 describes the data sources and the methodological framework. Results are reported in Section 4and conclusions and policy implications in Section 5.

2. Literature review

During the past decades, several authors have examined relationship between exchange rate and inflation using data from the industrialized, G-7, the European Union (EU), theCentral and Eastern European (CEEC), and the Organization for Economic Co-operation and Development (OECD) economies. Othershave investigated the inflation-exchange rate nexus for economies characterized by fluctuations in the socio-economic variables.

Andrés et al. (1996)applied theBalassa-Samuelson effect to investigate the inflation-exchange rate nexus in some industrialized countries, using data from 1960-73for the fixed exchange rate regimes. The authors demonstrate that in the long-run, the costs of inflation are underestimated in the countries and the periods,under fixed exchange rates regime.Egert (2002) exploredthe same for transition economies found that the movement in price and real exchange rate is weaker than expected. Holman and Rioja (2001) extendedthe work to explore the international transmission of anticipated inflation under alternative exchange-rate regimes, bothflexible and fixed. The authorsexplore the beggar-thy-neighbor policy of inflation in a flexible nominal exchange rate. For inflation under nominal exchange rate regime, each country sustains a welfare loss. Jeong and Lee (2001) focus on the G-7 countries (four from EU, two from North Americaand Japan.The authors find that the price levels in several countries move together under fixed and flexible regimes.In investigating nominal convergence process of Central and Eastern European Countries Lewis (2009, p.508) concludes, “The key result, robust across all scenarios, is that countries with fixed exchange rates will find it much harder to simultaneously meet the criteria than inflation targets. The ability of a country to get inflation below the reference value under a fixed exchange rate show a strong effect for the relative price level.”

Kamin (2001) examines thelink between real exchange rates and inflationin Mexicofrom 1988-94using an error-correction model. The results suggest that the real appreciation of the peso over the period is the result of growing domestic demand and backward-looking inflation.Bleaney and Fielding (2002) apply a model to a sample of 80 developing countries and find thatthesecountriescould reduce inflationary expectations by pegging exchange rate. Arize et al. (2004) examine data from 1973.q2–1998.q1 from 82 developed, developing and transitioneconomies with floating exchange rates. The results suggest that nominal exchange-rate variability is positivelycorrelated with inflation variability.

Using Hungarianmonthly data from 1991.06-2002.02,Barlow (2005) shows that inflation acts as a good control over depreciation of the exchange rate in the pre announced crawl in 1995.Celasun (2006) tests the forward-looking pricing hypothesis using data from four exchange rate based stabilization episodes. The results show that the level of inflation increases the value of the real exchange rate appreciation. This works through imperfect credibility model of exchange rate based stabilization episodes. Carranza et al. (2009) usevarying degrees of dollarization using data fromover hundred countries. Based on panel model, they find that countries with the high dollarization degree register high inflation pass-through. Civcir and Akçağlayan (2010)examine the monetary policy reaction functions for the Central Bank of Turkey over the intervals: 1987:01–2001:12 and 2002:01–2009:05. They find that under inflation targeting, exchange rate developments continue to have strong effect on monetary policy. Reboredo (2012) studies the comovement between oil prices and exchange ratesfor different countries but using daily data, from 4 January 2000 to 15 June 2010. Heunderlines the weak connection between oil price and exchangerate.

Additional evidences on inflation-exchange rate nexusis provided by Pavasuthipaisit (2010), Prasertnukul et al. (2010),Pourroy (2012), and Berganza and Broto (2012). Following Taylor (2000), Junttila and Korhonen (2012) explore the exchange rate pass-through into aggregate import prices for nine OECD countries. They find that for incomplete exchange rate pass-through,it is positively correlated with the inflation in the importing country. In the same note, Jiménez-Rodríguez and Morales-Zumaquero (2016) analyze the link between exchange rate pass-through to domestic prices and import prices for the G-7 countries. They validate the Taylor’s hypothesis by showing there is a robust and significant relationship between exchange rate pass-through and inflation volatility, with the same sign.

As part of the post-Soviet emerging nation, a few studies have examined the Romaniancase. Gueorguiev (2003) explores the exchange rate pass-through to prices. Using recursive VAR model, he finds that the pass through is, “large and relatively fast, accounting for a sizable fraction of inflation” (p.1).Budina et al. (2006) use error correction model and show that inflation is largely a monetary phenomenon.Dumitru and Jianu (2009) use data from 1998 to 2006 to examine the Balassa-Samuelson effect, which is useful in understanding the inflation-exchange rate nexus. Some of the results suggest that the effect is expected to determine higher inflation with future price liberalization for the non-tradable.Analyzing six emerging European countries that target inflation, including Romania, Nojkovic and Petrovic (2015) find that in Romania, Serbia and Albania the exchange rate is a goal for itself. For Czech Republic, Poland and Hungary, the exchange rate is a useful instrument to achieve inflation target.

On this literature ground, the paper investigates the direction of causality between the CPI, PPI and exchange rate.

3. Data and Methodology

3.1 Data

We use consumers’ price index (CPI), producers’ price index (PPI) to measure consumers’ price and producers’ price, respectively; and exchange rate is measured by nominal effective exchange rate (NEER). All data is monthly from 1990m1-2009m03,obtained from IMF-CD ROM (2010).

3.2 Methodology

As part of empirical strategy, we implement the frequency domain approach to examine causality among PPI, CPI and exchange rate for Romania. We do this in departure from that old traditionwhere time domain approach has been the mainstay inexamining relationships among economic time series.

In addition tothe direction, we are also interested in the strength of the relationships involving the nominal magnitudes like CPI, and PPI, and NEER. A relevant question is, whetherthe direction of the relationship stays constant amongthe trio or itexhibits some cyclical variation over the study period and across frequencies. Some of these questions have been partly addressed by a few authorsusingthe Granger testin the time domain, whichis a one-shot test showing evidence of causality. However, the testcannot detect direction and strength of causality over the frequencies – the cyclical causality.The advances in time series methodologyenable us to decompose causality by frequency andanalyze the cyclical nature and the strength of Granger causality. The differing theoretical perspectives producevariation in predictions about the short and long-term trade-offs among CPI, PPI and NEER.