The Impact of Oil Prices on Soybeans Commodity Prices: Asymmetric Cointegration Evidence

The Empirical Economics Letters, 15(1): (January 2016) ISSN 1681 8997 The Impact of Oil Prices on Soybeans Commodity Prices: Asymmetric Cointegration Evidence R. Balach, B.T Matemilola *, Lee Chin and Amdadullah Baloch Department of Economics, Universiti Putra Malaysia Selangor, Malaysia Abstract: The increases in soybean commodity prices are attributed to increase in oil prices which have affected the prices of agricultural grains commodity products such as soybean oil price. The study investigates asymmetric relationship between soybeans commodity price and crude oil price using most recent data. In order to account for possible asymmetric effect of oil price on soybeans commodity price, the paper uses momentum threshold autoregressive (MTAR) models, developed by Enders and Siklos (2001). Empirical results reveal that soybeans commodity prices and crude oil prices are cointegrated and the adjustment process is asymmetric in MTAR-consistent result. Our findings confirm that increase in crude oil prices affects the soybeans commodity prices in an asymmetric ways. Specifically, soybeans commodity prices rise faster when crude oil price increases but there is no immediate offsetting effect when the crude oil prices decrease. In other words, the asymmetric effect of oil price-shocks is greater when oil price increases than when oil price decreases. Policymakers may use these results to design appropriate policies to curb the inflationary consequences of oil prices. Keywords: Global Soybeans Price; Price Index; Crude Oil Price; Asymmetric Cointegration; Asymmetric Error Correction JEL Classification Number: Q00; L71; Q430 1. Introduction Oil price shock presents policymakers with difficult choices because it simultaneously increase inflation and reduce output (Jimenez-Rodriguez, 2011). Oil is used as a direct input for various consumer items and it is generally believed that higher oil prices lead to inflation which in turn translates into higher prices, for at least some commodity. In the literature, the effect that oil price shocks have on inflation which causes rising commodity prices is called pass-through effect (Ibrahim and Chancharoenchai, 2013). The degree of oil price pass-through into inflation has attracted the interest of researchers focusing on oil price shock and its effects on prices of goods, following the first OPEC oil embargo in 1973. Recently, the degree of oil price pass-through into commodity prices has received considerable attention among researchers due to occurrences of drastic fluctuations in global oil prices. * Corresponding author. Email: matemilolabt@gmail.com

The Empirical Economics Letters, 15(1): (January 2016) 16 In the literature, there are various channels through which oil price changes affect inflation, leading to higher commodity prices. Firstly, it is often cited that oil price increases reduce availability of basic inputs into production which reduces potential output (Brown and Yucel, 1999). Accordingly, there is a backward aggregate supply function through production cost increase and productivity decline, which leads to the increase in the aggregate price(brown and Yucel 2002).Afterwards, it may even generate a wageprice spiral that leads to further rise in the commodity prices. Furthermore, oil price increases could lower the nation s aggregate demand through real balance effect (Mork, 1994) and consumption and investment effects (Lardic and Mignon 2008). Moreover, Hanson et al (1993) note the transmission channel of oil prices to inflation through exchange rate changes. But, the influence of exchange rate channel depends on whether the nation is an oil-importing or oil-exporting country and the nation s dependence on international trade (Mansor and Rusmawati 2012). Empirical evidence on the effect of oil price pass-through to commodity prices shows contradictory results which justify the need for further studies to clarify misconceptions. For instance, in an early study, Burbidge and Harrison (1984) use five developed economies namely, Canada, Germany, Japan, the UK and USA to investigate the effect of oil price pass-through to food prices. The authors find considerable oil price effects on Canadian and the United States price levels and lesser effects on other countries. Similarly, LeBlanc and Chinn (2004) report modest effects of oil price fluctuation on inflation for the United States, United Kingdom, France, Japan, and Germany. Conversely, Hooker (2002) results show breakdown in the oil price pass-through into core US inflation. Barsky and Kilian (2004) argue that there are numerous factors in the US inflation that are not related to oil price shocks and that inflation does not follow the oil shocks. Recently, Chen (2009) findings reveal that the oil price pass-through effect into food prices is declining. Following the above inconsistencies in previous empirical findings in the literature, this paper focuses on the long-term relationship between soybeans oil price and crude oil prices on the global economy within an asymmetric cointegration framework. There are several evidence that support asymmetric relationship between the oil price and macroeconomic variables (see for example, Cuñado and Pérez de Gracia, 2005; Hooker, 2002; Mork et al., 1994; Ferderer, 1996). The common findings of these studies support nonlinear relationship between oil prices and macroeconomic variables. More specifically, some studies find that consumer price indexes responds asymmetrically to oil price shocks (see, Mansor and Chancharoenchai, 2013; Hooker, 2002). The results mostly confirm stickiness of nominal prices, especially in the downward direction, which provides support for asymmetric adjustments in consumer price indexes. The contribution of this paper is

The Empirical Economics Letters, 15(1): (January 2016) 17 twofold. Firstly, this present paper investigates asymmetric relationship between global soybeans commodity price and global crude oil price using most recent data. Secondly, in order to account for possible asymmetric effect of global oil price on global soybeans commodity price, this paper applies Enders and Siklos (2001) threshold autoregressive (TAR) and momentum threshold autoregressive (MTAR) models. The paper is organized as follows. Section 2 describes the empirical methodology and data. Section 3 presents the results of the empirical analysis while concluding remarks are given in the final section. 2. Data and Methodology This study analyses monthly data of the crude Oil price (WTI) and soybeans commodity price, from April 1984 to March 2014. The data of both variables are collected from Mandi index. The aim of study is to analyse whether there exist asymmetric cointegration between the two variables. The long run relationship between the variables is as follows: LCP t = β + β 1 LCO t + ε t (1) In above long run relationship the LCP is the natural log of soybeans commodity price and LCO is the natural log of crude oil price (WTI).The residuals specification in the Engel and Granger (1987) co-integration assumes symmetric adjustment in the long-run. However, asymmetric adjustment may exist. Enders and Siklos (2001) momentum threshold autoregressive (M-TAR) model modify the Engel and Granger (1987) test strategy and it has good power and size properties relative to the assumption of symmetric adjustment. In asymmetric adjustment, we established two partitions for the residuals by classifying them as above threshold and below threshold (Enders and Siklos, 2001). To allow asymmetric adjustment in the residuals, the paper follows Enders and Siklos (2001). Enders and Siklos (2001) have developed two different nonlinear cointegration models, which allow for tests of asymmetry, the threshold autoregressive (TAR) and the momentum threshold autoregressive (MTAR). TAR model can be described by the following equation: μ t = ρ + I t μ t 1 + ρ (1 I t )μ t 1 + Where v t ~iid 0, σ 2 and I t is the Heaviside indicator such that I t = 1 if μ t 1 τ 0 if μ t 1 < τ. k i=1 λ i μ t i + v t (2) Enders and Siklos (2001) considered an alternative rule for setting the Heaviside indicator as follows: M t = 1 if μ t 1 τ (4) 0 if μ t 1 < τ. (3)

The Empirical Economics Letters, 15(1): (January 2016) 18 Equations 2 and 4 form Momentum Threshold Autoregressive Model (MTAR) while equation 2 and 3 form the Threshold Autoregressive Model (TAR).Enders and Siklos (2001) used Chan s methodology to a Monte Carlo study to get F- statistics for the null hypothesis of p 1 = p 2 = o, then, we search for the threshold t using Chan s procedures. The recommendation is to select the adjustment mechanism using a model selection criterion such as Akaike Information Criteria (AIC) or Schwartz Information Criteria (SIC). In this study, we use AIC in our estimation. The MTAR model is more important when the adjustment is believed to exhibit more momentum in one direction than the other (Payne and Waters, 2008). The model above assumes that threshold value (τ) is unknown but it has to be estimated by a grid search (Matemilola et al, 2014; Al-Gudhea et al, 2006; Enders and Siklos, 2001). Firstly, the residual from the TAR (μ t ) and MTAR ( μ t ) model is sorted in sequence, in an ascending order. Secondly, to ensure that the number of observation in each regime is reasonable, each (μ t ) that falls between the lowest 15 percent and highest 85 percent of the series are considered as potential threshold. Third, we run regression on Equation (1) and use each (μ t ) as a potential value of the threshold. Finally, the value that has the lowest residual sum of squares is taken as the consistent estimate of the threshold. MTAR model is the choice model of our analysis and in the presence of asymmetric co-integration; it is used to estimate asymmetric error correction models for soybeans commodity prices. 3. Empirical analysis Table 1 shows the tests of ADF and PP unit root. The ADF and PP tests have both trend and intercept terms. In ADF test the Schwarz Information Criterion (SIC) is used to select the optimal lag. The results of the ADF and PP unit root tests are consistent where we fail to reject the null hypothesis at level, except soybeans commodity prices (LCP) and crude oil prices (LCO). However, all the variables are stationary after first differencing where we reject the null hypothesis at the 1% significant level. The tests indicate that all variables are integrated of order I (1). Therefore, the paper proceeds to co-integration test. Table 1: ADF and PP Unit Root Tests Variables Level First difference ADF PP ADF PP LCP -3.501** -2.863-13.769*** -13.872*** LCO -3.459** -3.150* -13.933*** -13.386*** Note: *** indicates 5% level of significance. LCP= Natural logarithm of the specific commodity (soybeans oil). LCO= Natural logarithm of crude oil price (WTI).

The Empirical Economics Letters, 15(1): (January 2016) 19 Table 2: Symmetric Cointegration Tests JJ test Commodity Price EG Test Statistics Null hypothesis r = 1 r 1 LCP -3.6825*** Trace 15.640** 1.047 Critical values Max Eigenvalue 14.593** 1.047 1% -2.571 5% 10% -1.941-1.616 Trace 15.494** 3.841 Max Eigenvlaue 14.264** 3.841 Note: *** and ** denote significance at 1% and 5% levels, respectively. SIC is used to select the optimal lag order. Table 2 shows the results of Engel Granger (EG)and Johansen Juselius (JJ) cointegration tests. The VAR lag order for the JJ test is based on non-auto correlated errors. All residuals are not auto-correlated, so lag order 1 is chosen. From the OLS estimation, we extract residual, and the EG results shows that residual is stationary which fulfil the necessary condition for cointegration. In Table 2, the results of EG shows there is cointegration at 1% significant level. Therefore, we conclude there is cointegration between the two variables. For JJ, both trace and Max Eigen statistics reject the null hypothesis of no cointegration equation, therefore the paper concludes that there is long run relation between soybeans commodity price and crude oil price. Table 3: Estimates for TAR and MTAR Co-integration TAR (zero) TAR-Consistent MTAR(zero) MTAR-consistent 1-0.048 (-2.440) -0.077(-3.155) -0.027(-1.421) 0.034 (1.178) 2-0.037(-2.098) -0.027(-1.761) -0.053(-3.011) -0.061(-4.35) c 0.000 0.274 0.000 0.043 F-joint stat. (Ф) 5.063[5.869] 6.540 [6.902] 5.466[6.361] 9.701**[8.142] F-equal: 0.197[2.728] 3.071[6.463] 0.981[3.863] 9.219**[8.270] Note: ** indicates significance at 5% level, and numbers in parenthesis are test statistics. Numbers in brackets are simulated critical values obtained from Monte Carlo simulation. C indicates the threshold value of. Table 3 report the results of the autoregressive (TAR) model and the momentum threshold autoregressive (MTAR) model. The paper uses the maximum lag automatically selected by the system. However, the Monte Carlo experiment is used to search for the critical value at 5%. For the TAR model with zero threshold value, the value of Ф is (5.063)

The Empirical Economics Letters, 15(1): (January 2016) 20 which is less than (5.869) at 5% critical value. However, F-equality statistics (0.197) is less than (2.728) at 5% critical value; therefore we cannot reject the null of symmetric adjustment in TAR with zero thresholds. For the TAR consistent model, the threshold value is 0.274. The value of F-joint (6.540) which is less than (6.902) at 5% critical value, thus in this case, we also cannot reject the null. F-equality statistics (3.071) is less than the critical value (6.463) at 5%. Likewise, we cannot reject the null hypothesis of symmetric adjustment. In Table 3 the momentum threshold autoregressive (MTAR) model with zero threshold value, the value of Ф (5.466) is less than (6.361) at 5 % critical value provided by Monte Carlo simulation. Similarly, in the MTAR consistent model, the value of Ф (9.701) is greater than the critical value (8.142) provided by Monte Carlo simulation at 5%. Hence, we reject the null hypothesis of no co-integration. With the MTAR consistent, F- equal (9.219) exceed F-critical value (8.270) at 5% level. Thus, the null hypothesis of symmetric adjustment is rejected at the 5% level. This result indicates that soybeans commodity prices and crude oil prices are co-integrated and the adjustment process is asymmetric in MTAR consistent result. Table 4: Estimation Results of Asymmetric Error Correction Model Dynamic equation of commodity price (Soybeans Oil) DLCP t = -0.0002+ 0.004Z + t 1-0.053Z t 1 + 0.333DLCP t 1-0.023DLCO t 1 (0.948) (0.8585) (0.000) (0.000) (0.506) Adj- R 2 = 0.133JB= 6.836(0.033) ARCH(1)=0.176(0.674) F- Statistics= 14.725(0.000) LM(1) = 5.386(0.021) RESET(1)= 2.415(0.016) Note: Numbers in brackets are p-values. JB= Jargue- Bera test for normality. LM= Test for serial correlation ARCH= Test for autoregressive conditional Heteroskedasticity. RESET= Ramsey s misspecification test with fitted terms set to 1. The results on Table 4 are the estimated results of asymmetric error-correction models for LCP (commodity prices) and LCO (crude oil prices) with MTAR consistent model. The result of adjusted R 2 is adequate and F- statistics is significant (0.000). So the both results are satisfactory. However, to check the results, we test various diagnostic tests. In the test of autocorrelation, we notice significance at lag 1. We also test for ARCH test for + Heteroskedasticity which is not significant. Where Z t 1 and Z t 1 are the residuals from + the equation. The lag length is based on AIC. The adjustment of speed of Z t 1 is insignificant. Subsequently, we find significant long run relation between commodity prices (LCP) and crude oil prices (LCO). The error-correction term coefficient is significant in below their long run value Z t 1 0.053 (0.000). Our results are consistent with prior studies in the literature that support asymmetric relationship between the oil

The Empirical Economics Letters, 15(1): (January 2016) 21 price and macroeconomic variables (e.g. Cunado and Perez de Gracia 2005; Hooker 2002; Mork et al., 1994, and Ferderer 1996). The common findings of these studies support nonlinear relationship between oil prices and macroeconomic variables. 4. Conclusions The increases in soybean commodity prices are attributed to increase in oil prices which have affected the prices of agricultural grains commodity products such as soybean oil price. This paper investigates asymmetric relationship between soybean commodity price and crude oil price. In order to account for possible asymmetric effect of crude oil price on soybeans commodity price, the paper applies threshold autoregressive (TAR) and momentum threshold autoregressive (MTAR) models develop by Enders and Siklos (2001). Most of the studies claim that commodity price and global oil price are cointegrated. Our study extends the analyses to allow for asymmetric adjustment mechanism of the soybeans commodity price to changes in the crude oil prices using most recent data sets. This result indicates that soybeans commodity prices and crude oil prices are cointegrated and the adjustment process is asymmetric in MTAR consistent result. Our findings serve as a confirmation that the increasing global crude oil price does affects the global soybeans commodity prices in an asymmetric ways. In order words, the global crude oil price inflation is transmitted into global soybeans commodity price. Moreover, due to the asymmetric adjustment of the soybeans commodity price, the soybeans commodity prices rises faster when crude oil price increases. Conversely, there is no immediate offsetting effect when the crude oil prices decline. Thus, the commodity prices tend to remain high. This study documents the possibility of asymmetric effect of oil price-shocks which are greater when oil price increases than when oil price decreases. Policymakers could use these results to design appropriate policies in curbing the inflationary consequences of oil prices. The paper makes two contributions. Firstly, the paper investigates asymmetric relationship between global soybeans commodity price and global crude oil price using most recent data. Secondly, in order to account for possible asymmetric effect of global oil price on global soybeans commodity price, this paper applies Enders and Siklos (2001) threshold autoregressive and momentum threshold autoregressive models.

The Empirical Economics Letters, 15(1): (January 2016) 22 References Al-Gudhea, S., Kenc, T., and Dibooglu, S., 2006, Do retail gasoline prices rise more readily than they fall? A threshold co-integration approach, Journal of Economics and Business, 560-574. Barsky, R. B., and Kilian, L. 2004,Oil and the Macroeconomy since the 1970s, Journal of Economic Perspectives, 18(4), 115 34. Brown, S. and Yucel, M., 2002, Energy prices and aggregate economic activity: an interpretative Survey, Quarterly Review of Economics and Finance, 42 (2), 193 208. Brown, S. P. A., and Yücel, M. K., 1999, Oil Prices and U.S. aggregate economic activity: A question of neutrality, Economic and Financial Review, Federal Reserve Bank of Dallas, Second Quarter: 16-23. Burbidge, J. and Harrison, A., 1984, Testing for the effects of oil-price rises using vector auto regressions, International Economic Review, 25 (2), 459-484. Cunado, J. and de Gracia, F. P., 2005, Oil prices, economic activity and inflation: evidence from some Asian countries, Quarterly Review of Economics and Finance,45(1), 65-83. Chen, S 2009, Oil price pass-through into inflation, Energy Economics, 31, 126-133. Engel, R.F. and Granger, CWJ., 1987, Co-integration and error correction: Representation, estimating, and testing, Econometrica, 55, 251-276. Enders, W. and Siklos, P. 2001, Cointegration and threshold adjustment, Journal of Business and Economic Statistics, 19, 166 176. Federer, J. Peter, 1996, Oil price volatility and the macro economy: A Solution to the asymmetry puzzle, Journal of Macroeconomics, 18:1-16. Hanson, K., Robinson, S. and Schluter, G., 1993, Sectoral effect of world oil priceshock: economy-wide linkages to the agricultural sector, Journal of Agricultural and Resource Economics, 18(1), 96-116. Hooker, M. A., 2002, Are oil shocks inflationary? Asymmetric and non-linear specifications versus changes in regime, Journal of Money, Credit and Banking, 34 (2), 540-561. Jimenez-Rodriguez, R., 2011, Macroeconomic structure and oil price shocks at the industrial level, International Economic Journal, 25 (1), 173-189.

The Empirical Economics Letters, 15(1): (January 2016) 23 Lardic, S. and Mignon, V., 2008, Oil prices and economic activity: an asymmetric cointegration approach, Energy Economics, 30 (3), 847 855. LeBlanc, M. and Chinn, M.D., 2004, Do high oil prices presage inflation? The evidence from G5 countries, Business Economics, 34, 38-48. Mansor, H.I and Chancharoenchai, K., 2013, How inflationary are oil hikes? A disaggregated look at Thailand using symmetric and asymmetric cointegration models, Journal of Asia Pacific Economy, 19:3,409-422. Mansor H. Ibrahim and Rusmawati Said, 2012, Disaggregated consumer prices and oil price pass-through: evidence from Malaysia, China Agricultural Economic Review, 4(4), 514-529. Mork, K.A., 1994, Business Cycles and the Oil Market, The Energy Journal, 15, 15-38. Matemilola B.T, Bany-Ariffin, A.N and Muhtar, F.E., 2015, Impact of monetary policy on bank lending rate in South Africa, Borsa Istanbul Review, 15, 53-59. Appendix Cusum Test for Asymmetric 60 40 20 0-20 -40-60 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 CUSUM 5% Significance

The Empirical Economics Letters, 15(1): (January 2016) 24 Cusum Test for Symmetric 60 40 20 0-20 -40-60 84 86 88 90 92 94 96 98 00 02 04 06 08 10 12 CUSUM 5% Significance Table 5: OLS Estimation Dependent Long-run coefficients Variables Constant LCO LCP 4.869205 (0.0000) 0.424308 (0.0000)