Bio -based and A ppl ied Economics BAE Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6e172 | DOI: 10.36253/bae-10896 Copyright: © 2022 H. Uçak, E. Yelgen, Y. Arı. Open access, article published by Firenze University Press under CC-BY-4.0 License. Firenze University Press | www.fupress.com/bae Citation: H. Uçak, E. Yelgen, Y. Arı (2022). The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey. Bio-based and Applied Economics 11(1): 37-54. doi: 10.36253/bae-10896 Received: May 19, 2021 Accepted: January 17, 2022 Published: July 22, 2022 Data Availability Statement: All rel- evant data are within the paper and its Supporting Information files. Competing Interests: The Author(s) declare(s) no conflict of interest. Editor: Fabio Gaetano Santeramo. ORCID HU: 0000-0001-5290-5846 EY: 0000-0003-4506-377X YA: 0000-0002-5666-5365 The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Harun Uçak*, Esin Yelgen, Yakup Arı Alanya Alaaddin Keykubat University, Turkey *Corresponding author. E-mail: harun.ucak@alanya.edu.tr Abstract. In agricultural economics, fluctuations in food prices and the factors affect- ing these fluctuations have always been an important research topic. From produc- tion to delivery to consumers, the supply chain of agricultural products has a dynamic structure with continuous changes. In this dynamic process, analyzing the intensive use of energy at each stage has gained more importance with its deepening effects in com- parison to the past. This study will empirically explore the volatility spillovers between energy price index and fruit-vegetables price index in the period of 2007-2020 in Tur- key using the Kanas and Diebold-Yilmaz approaches. According to the results obtained from the Kanas approach in the study, it has been observed that there is a statistically significant volatility spillover from the energy price index to the vegetable price index, whereas there is no statistically significant volatility spillover to the fruit price index. This finding was supported by the results obtained from the Diebold-Yilmaz approach showing that there is a volatility spillover of 13.52% to the vegetable price index and 0.86% to the fruit price index from the energy price index. Keywords: volatility spillover, energy, agricultural prices, EGARCH, agricultural mar- kets. JEL codes: Q11, Q18, Q41, Q47, C32. 1. INTRODUCTION Volatility in food prices and the reasons behind this volatility have recently become a trending topic of discussions throughout the world, while they are often discussed in literature as well. In this regard, pricing process of sub-product groups must also be analyzed in addition to general food prices. Indeed, due to the difficulties in storing these products for a long period, changing vegetable and fruit prices might well cause producers and consumers to be deeply affected by price volatility. On the other hand, it is also highly important to examine the reasons that may affect the price fluc- tuations of these products. Fresh fruit and vegetables sector is considered one of the most essential sectors in the agricultural industry as it is vital for sustaining human life. In this context, the United Nations declared the year of 2021 as the “Interna- tional Year of Fruits and Vegetables”, highlighting the importance of fruits 38 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı and vegetables in nourishment, the problems experi- enced in the process from production to consumption, food wastes and losses, the importance of farming in the fight against famine and small family businesses gen- erating incomes. Thus, the factors that underlie price changes in agricultural markets is currently a hot topic. Prices in agricultural markets have recently been affect- ed by macroeconomic factors such as exchange rate, inflation (Algieri, 2016), interest rates, energy prices and demand for biofuels, monetary policies, financial invest- ments and speculations, sudden trade restrictions or lack of information, transaction costs, agricultural policies and international prices (Kalkuhl et al., 2016; Algieri, 2016; Kornher and Kalkuhl, 2013). This study will focus on Turkey from an empirical perspective within its scope. While the country stands out in fruit and vegetable production across the world, Turkey is experiencing frequent price volatility at recent times. According to the World Food Organization’s 2019 statistics, Turkey is the 4th largest producer of fresh veg- etables in the world (Statista, 2021a). In addition, it is the 6th largest producer of fresh fruits in the world (Statista, 2021b). Therefore, Turkey is one of the most important agricultural producers in the world. However, Turkey’s currency is one with the highest volatility among emerg- ing market markets and this causes fluctuations in the fruit and vegetable price indices. Besides, fluctuations in energy prices due to the volatility of the exchange rate and global markets has become significant as energy is an input item in production processes. Consider- ing upward fluctuations in particular, the practices for direct sale points and mediators in the supply chain have been heavily discussed in recent years. In the same vein, the fluctuations in food prices have been the hot topic in Turkey too due to the recent global crises, the climate change and foreign-source dependency on ener- gy. It is stated that the reason behind these fluctuations in agricultural product prices is the increasing produc- tion input prices by farmers. Besides seasonal effects on the price fluctuations in agricultural commodities, it can be observed that increasing energy prices have a direct or indirect aggravating effect on the costs of agri- cultural inputs such as fertilizers, chemicals, irrigation, production, storage and transportation (Fasanya and Akinbowale, 2019; Tadasse et al., 2016; Algieri, 2016). Moreover, the use of modern technology applications in agriculture also increases energy consumption. The use of agricultural machinery and pesticides requires the consumption of fossil fuels, and indeed, intense energy consumption is particularly observed in the field of pes- ticide production (Öztürk et al., 2010). Besides, price volatility in the categories of electricity, coal, petroleum products and natural gas has an extremely deep nega- tive impact on the economic performance of Turkey, as an energy importer. As a matter of fact, oil and natural gas reserves are limited in Turkey leading to foreign- source dependence in the field of energy. Thus, it is observed that Turkey has been the country with the fast- est increase in energy demand among the Organization for Economic Cooperation and Development (OECD) countries in the past 20 years. Within this framework, Turkey ranks second in the world after China in the increase in electricity and natural gas demands. Exist- ing energy sources cannot unfortunately meet Turkey’s increasing energy needs and thus, the country meets nearly 74% of its energy needs via imported sources (MFA, 2020). Considering that Turkey is a country dependent on imports of oil in its consumption, there is an urging need to address the effects of changing energy prices on the performance of several sectors and indus- tries (Algan et al., 2017). On the other hand, the increase in energy prices in recent years is one of the most crucial cost items threatening agricultural production (Yıldırım, 2020). Hence, the fluctuations in these costs reflect on product prices and cause difficulties in production plans (Fasanya and Akinbowale, 2019: 186; Tadasse et al., 2016: 63; Algieri, 2016: 210). For the reasons mentioned above, this study aims to investigate the effects of changes in energy prices on other price indices for Turkey. In this regard, we ana- lyzed the volatility spillover between the Energy Price Index (EPI), the Fruit Price Index (FPI) and the Vegeta- ble Price Index (VPI) using monthly data sets from Jan- uary 2007 to December 2019 by two different methods: The Kanas (1998) Approach for volatility spillover effect and the Diebold-Yilmaz (2009, 2012) spillover index, analyzed respectively. As for the content of the study, the second section consists of an extensive literature review. This part is followed by a detailed description of the methodology. The fourth section summarizes the data set used in the study. In the fifth part, empirical results of the analyses are given in two subsections. Finally, the last section covers comments, discussions and policy recommendations based on the study results. 2. BACKGROUND AND LITERATURE REVIEW Energy consumption is one of the main determi- nants of the socio-economic development of countries. More specifically, oil and its derivatives are considered one of the main production factors in an economy. They are used in the energy supply of various sectors includ- ing agriculture, transportation, industry and households, 39The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 in addition to their extensive use as raw materials in the production of other energy products like electricity and petrochemistry. Thus, oil and its derivatives have a vast impact on other commodities (Sarwar et al., 2020; Taghizadeh-Hasery et al., 2019). At recent times, agricultural products and energy markets have been growingly intertwined (Koirala et al, 2015: 431). From this perspective, energy consump- tion in agriculture can be evaluated in two categories: (1) Direct energy use: Energy inputs such as electric- ity, fuel, oil, coal, petroleum products, natural gas, biomass can be used in agricultural activities. (2) Indi- rect energy use: The amount of energy consumed in human and animal labor, agricultural tools or machin- eries, fertilizers, pesticides, irrigation or seed produc- tion. In this regard, energy prices affect the costs of inputs necessary for faming including inorganic ferti- lizers and fuel for agricultural machinery. Moreover, it is commonly observed that energy prices increase transportation costs and therefore, affect food trans- portation and distribution costs. The primary energy products directly consumed in agricultural production include fuels such as coal, petroleum products, natu- ral gas and biomass. Also, electricity is widely used as power carrier in farming and particularly irrigation operations. It is a source commonly benefited in the agricultural industry (Akder et al., 2020: 9; Radmehr and Henneberry, 2020: 2; Sarwar, 2020: 1; Öztürk et al., 2010: 2; Mawejje, 2016: 2; Nwoko et al., 2016: 2; Gilbert and Mugera 2014: 201). Yet, the history is marked by many crises relat- ed to food supply and demand. In this vein, it can be observed that the recent price volatility in food has had a destructive effect. The increased volatility in prices in this field can be associated with the transition from the labor-intensive to a more capital-intensive agricul- tural production in recent years as well as the regional and national differences in terms of farming. The use of energy is naturally essential in agricultural produc- tion. Today’s technology enables growing even tropical products in cold regions thanks to the heat provided by energy sources. Hence, technology allows countries that are rich in energy sources to produce fruits and vegeta- bles despite their cold climate. On the other hand, espe- cially developing countries that import energy seem to have hardship in their agricultural operations due to the high energy prices increasing the costs of inputs. This leads to an intricate relationship between energy and prices of agricultural products. From this per- spective, various studies analyze the effects of oil and other energy prices on agricultural product prices. For example, Hau et al. (2020) and Koirala et al. (2015) dis- cuss the relations between oil and agricultural prices in terms of futures. Sarwar et al. (2020), Hesary et al. (2019), Alghalith (2010) and Zhang et al. (2010) examine the effects of the changing crude oil prices on agricul- tural products. On the other hand, Radmehr and Hen- neberry (2020), Balcılar and Bekun (2019) and Huchet- Bourdon (2011) scrutinize the effects of energy and exchange rates on agricultural products’ prices. Mawe- jje (2016) further dwells upon the importance of energy and climate shocks in the case of Uganda and the food prices in this country. In their study, Volpe et al (2013) also investigate how fuel prices in the USA affect the prices of whole- sale products and their transportation costs. Since agri- cultural products themselves have been used for energy production at recent times, Baffes (2011) examines the relations between oil, biofuel and prices of agricultural products. The literature in this field contains many other simi- lar studies analyzing the volatility in the prices of energy and agricultural product using the econometric tech- niques that are also benefited in this study. Table 1 sum- marizes these studies in detail: 3. METHODOLOGY 3.1 The Kanas approach for the volatility spillover effect Engle (1982) developed a new method to measure the volatility in a time series by modeling conditional variance. He revealed that the conditional variance is a function of the lagged values of the error term squares and modeled the change of the error term squares with respect to time using the ARCH process. Thanks to the introduction of the GARCH model in the literature, many other conditional variance models started to be widely used (Bollerslev, 1986). Although the standard GARCH model captures various features of financial series such as excess kurtosis and volatility cluster- ing, they are not successful in capturing the leverage effect of financial time series. Standard GARCH models tend to ignore the negative correlation between current return and future return volatility. Further, the con- straints on parameters to ensure the stationarity of the GARCH process can make parameter estimation dif- ficult. Lastly, another difficulty is to interpret whether shocks persist on the conditional variance in the stand- ard GARCH model. An alternative model developed by Nelson (1991) is the EGARCH model that removes these defects in the standard GARCH modeling of the financial time series, prevents the model from giving symmetrical responses in cases of positive and nega- 40 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı Ta bl e 1. S um m ar y Li te ra tu re R ev ie w. Au th or s G oa l M et ho do lo gy In du st ry Re gi on Re su lts H au e t a l. (2 02 0) To e xp lo re th e vo la til ity b et w ee n gl ob al cr ud e oi l a nd C hi na ’s ag ric ul tu ra l fu tu re s. - Ti m e- de pe nd en t p ar am et er st oc ha st ic v ol at ili ty - C on di tio na l v ol at ili ty A gr ic ul tu re , E ne rg y, Fu tu re s a nd O pt io ns M ar ke t C hi na Th er e is a he te ro ge ne ou s d ep en de nc e be tw ee n th e vo la til ity o f a gr ic ul tu ra l f ut ur es a nd th e vo la til ity o f c ru de oi l. W hi le c ru de o il vo la til ity d oe s n ot a ffe ct a gr ic ul tu ra l vo la til ity in th e no rm al m od e of th e cr ud e oi l m ar ke t, oi l vo la til ity a t h ig h or lo w a m ou nt s h as b ee n fo un d to h av e an e xt re m el y sig ni fic an t e ffe ct . Ti w ar i e t a l. (2 02 0) To a na ly ze th e pr og re ss io n- re gr es sio n re la tio ns hi p be tw ee n th e pr ic e in di ce s o f en er gy fu el s a nd fo od , i nd us tr ia l i np ut s, ag ric ul tu ra l r aw m at er ia ls, m et al s, an d be ve ra ge s ( le ad -la g re la tio n) in th e tim e- fr eq ue nc y do m ai n. - W av el et c oh er en ce a nd p ha se di ffe re nc es , - D ie bo ld & Y ilm az (2 01 2) a nd Ba ru ni k & K re hl ik (2 01 7) vo la til ity sp ill ov er in di ce s Fo od & En er gy G lo ba l Th e re su lts re ve al ed si gn ifi ca nt re la tio ns hi ps b et w ee n fu el an d fo od p ric es , f ue l a nd in du st ry p ric es , a nd fu el a nd m et al p ric es . Th e re su lts a lso sh ow ed th at th er e ar e ph as e re la tio ns hi ps be tw ee n th es e pa ire d pr ic es . Th e vo la til ity sp ill ov er re su lts sh ow ed th at th e ag ric ul tu ra l i nd us tr y w as th e m os t a ffe ct ed in du st ry b y sh oc ks fr om o th er m ar ke ts . Ba lc ıla r a nd B ek un (2 01 9) To e xa m in e th e st ru ct ur e of th e in te rc on ne ct ed ne ss b et w ee n th e re tu rn s of o il an d fo re ig n ex ch an ge p ric es w ith se le ct ed a gr ic ul tu ra l c om m od ity p ric es . - D ie bo ld & Y ilm az (2 01 2) vo la til ity sp ill ov er in de x Fo od , E ne rg y & Fi na nc e N ig er ia A na ly se s d em on st ra te th at b an an a, c oc oa , p ea nu t, co rn , so yb ea n, a nd w he at a re n et tr an sm itt er s o f s pi llo ve r. M or eo ve r, th er e is w ea k sp ill ov er b et w ee n th e va ria bl es o f ric e an d so rg hu m , i n ad di tio n to p ric e in fla tio n, n om in al eff ec tiv e ex ch an ge ra te , a nd o il pr ic es . Fa sa ny a an d A ki nb ow al e (2 01 9) To a na ly ze th e re tu rn s a nd v ol at ili ty o f cr ud e oi l a nd fo od p ric es . - D ie bo ld & Y ilm az (2 01 2) vo la til ity sp ill ov er in de x Fo od & En er gy N ig er ia Ev id en ce o f t he in te rc on ne ct ed ne ss b et w ee n cr ud e oi l an d fo od p ric es w as fo un d ba se d on sp ill ov er in di ce s. A dr an gi e t a l. (2 01 7) To e xa m in e th e da ily v ol at ili ty sp ill ov er s be tw ee n cr ud e oi l p ric es a nd a se le ct ed gr ou p of b as ic a gr ic ul tu ra l p ro du ct s. - Jo ha ns en -J us el iu s co in te gr at io n te st , - D yn am ic c on di tio na l co rr el at io ns , - Sp ec tr al a nd c ro ss sp ec tr al an al ys es , - Th e Bi va ria te E G A RC H m od el a nd G ra ng er c au sa lit y te st s. Fo od & En er gy U SA Th e Jo ha ns en -J us el iu s c oi nt eg ra tio n te st re ve al s t ha t th e lo ng -r un e qu ili br iu m re la tio ns hi ps b et w ee n cr ud e oi l p ric es a nd th e co m m od iti es in q ue st io n ha ve di sa pp ea re d. Th e dy na m ic c on di tio na l c or re la tio ns sh ow th at th e re la tio ns hi p be tw ee n ag ric ul tu ra l p ro du ct s a nd c ru de o il ch an ge s o ve r t im e. Th e sp ec tr al a nd c ro ss -s pe ct ra l a na ly se s c on fir m th at vo la til ity in c ru de o il pr ic es is a ss oc ia te d w ith v ol at ili ty in ag ric ul tu ra l p ro du ct s g iv en in th e sa m pl e. Th e Bi va ria te E G A RC H m od el a nd G ra ng er c au sa lit y te st s c on fir m th is re la tio ns hi p. A na ly se s a lso c on fir m th at th e flu ct ua tio ns in c ru de oi l p ric es a re re la te d w ith th e vo la til ity o f a gr ic ul tu ra l pr od uc ts g iv en in th e sa m pl e. Ju di th e t a l. (2 01 7) To in ve st ig at e th e ca us e of in cr ea se d fo od p ric e vo la til ity . - D es cr ip tiv e st at ist ic s - C or re la tio n an al ys is Fo od U SA Th e re su lts sh ow th at th e m ai n so ur ce o f f oo d pr ic e vo la til ity is m ai nl y th e oi l p ric e sh oc k. 41The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Au th or s G oa l M et ho do lo gy In du st ry Re gi on Re su lts C ab re ra a nd S ch ul z (2 01 6) To in ve st ig at e th e ris k of p ric e an d vo la til ity a ris in g fr om th e co rr el at io n be tw ee n en er gy a nd a gr ic ul tu ra l co m m od ity p ric es a nd to e xa m in e th ei r ch an gi ng d yn am ic s o ve r t im e. - Th e as ym m et ric d yn am ic co nd iti on al c or re la tio n G A RC H m od el , - Th e m ul tiv ar ia te m ul tip lic at iv e vo la til ity m od el Fo od & En er gy G er m an yIt is re ve al ed th at p ric es m ov e to ge th er in th e lo ng r un an d bo th ra pe se ed a nd b io di es el p ric es re ac t t o de vi at io ns fr om th e eq ui lib riu m . In th e sh or t r un , b io di es el p ric es d o no t a ffe ct ra pe se ed an d cr ud e oi l p ric e le ve ls, b ut ra th er re ac t t o pr ic e ch an ge s i n th e ot he r t w o m ar ke ts . M or eo ve r, th e vo la til ity o f b io di es el is re la te d to th e vo la til ity o f c ru de o il an d ra pe se ed , w hi le th e co rr el at io n be tw ee n th e flu ct ua tio n of ra pe se ed a nd c ru de o il ha s be en in cr ea sin g in re ce nt y ea rs . Lu co tte (2 01 6) To a na ly ze th e dy na m ic s o f co -m ov em en ts in c ru de o il an d fo od pr ic es . - C or re la tio ns o f V A R es tim at io n er ro rs (v ar ia nc e de co m po sit io ns ) Fo od & En er gy G lo ba l Em pi ric al re su lts m an ife st th at a fte r t he sp ik es in th e co m m od ity p ric es , s tr on g po sit iv e co -m ov em en ts w er e ob se rv ed b et w ee n th e cr ud e oi l p ric e an d fo od p ric e in di ce s, w hi le si gn ifi ca nt c or re la tio n co effi ci en ts w er e no t o bs er ve d in th e pe rio d be fo re th e sp ik es in th e co m m od ity p ric es . N w ok o et a l. (2 01 6) To e xa m in e th e lo ng -r un a nd sh or t- ru n re la tio ns hi ps b et w ee n oi l p ric e an d fo od p ric e vo la til ity a nd th e ca us al ity re la tio ns hi ps b et w ee n th em . - Jo ha ns en a nd Ju se liu s co -in te gr at io n te st , - Th e ve ct or e rr or c or re ct io n m od el , - G ra ng er c au sa lit y te st Fo od & En er gy N ig er ia A na ly se s r ev ea l t ha t t he re is a lo ng -r un re la tio ns hi p be tw ee n oi l p ric e an d lo ca l f oo d pr ic e vo la til ity a nd th at ca us al ity is o ne -w ay fr om o il pr ic e vo la til ity to fo od p ric e vo la til ity . G ilb er t a nd M ug er a (2 01 4) To in ve st ig at e th e ro le o f b io fu el s i n ex pl ai ni ng th e in cr ea se d vo la til ity in fo od p ro du ct s. - Th e m ul tiv ar ia te G A RC H m od el - Th e D yn am ic C on di tio na l C or re la tio n m od el Fo od & En er gy U SA A na ly zi ng a sa m pl e be tw ee n 20 00 a nd 2 01 1, th e st ud y fo un d in cr ea se s i n th e co rr el at io n an d jo in t m ov em en ts be tw ee n gr ai n an d cr ud e oi l p ric es a fte r 2 00 6 an d es pe ci al ly in 2 00 8 w he n cr ud e oi l p ric es w er e hi gh . Re se ar ch er s c on cl ud ed th at th e in cr ea se d vo la til ity in gr ai ns d ur in g th e 20 08 -2 00 9 in cr ea se w as la rg el y du e to sh oc ks tr an sf er re d fr om c ru de o il to g ra in s, pa rt ic ul ar ly co rn , w he at , a nd so yb ea n pr ic es . Ta de ss e et a l. (2 01 4) To se ar ch fo r e m pi ric al e vi de nc e on th e qu an tit at iv e sig ni fic an ce o f su pp ly, d em an d, a nd m ar ke t s ho ck s fo r p ric e ch an ge s i n in te rn at io na l f oo d co m m od ity m ar ke ts . To e xp lo re th e m ai n dr iv er s o f f oo d pr ic e sp ik es a nd v ol at ili ty fo r w he at , c or n, a nd so yb ea ns a nd sh ow h ow th es e fa ct or s tr ig ge re d th e cr isi s i n ex tr em e pr ic e sw in gs . - Th e pr ic e sp ik e m od el b y di ffe re nt ia te d re gr es sio n - Th e vo la til ity m od el b y pa ne l re gr es sio n - Th e pr ed ic tio n of e xt re m e vo la til ity b y qu an til e re gr es sio n Fo od G lo ba l Th ey c on cl ud ed th at o il pr ic es a re a st at ist ic al ly sig ni fic an t f ac to r i n ex pl ai ni ng th e in cr ea se s a nd v ol at ili ty in fo od p ric es . 42 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı tive shocks in volatility, and thus is more convenient for modeling conditional variance. In this model, the loga- rithmic conditional variance depends on both the size and the sign of the residuals (Nelson, 1991; Bollerslev et al., 1994). EGARCH (p, q) is: ln(σ2 t)=ω+∑p i=1[αizt-i+(γi|zt-i|-E[|zt-i|])]+∑q i=1βiln(σ2 t-i) (1) where zt=εt ⁄σt and the coefficient αi captures the sign effect and γi captures the size effect. So, the EGARCH (1, 1) model can be expressed as follows ln(σ2 t)=ω+α1zt-i+(γi|zt-i|-E[|zt-i|])]+β1ln(σ2 t-i) (2) where γi is also referred to as the asymmetry coefficient and β1 indicates volatility persistence. It can be said that there is a leverage effect on the conditional variance when has a value other than 0. In this study, the Kanas (1998) approach is taken as basis in determining the volatility spillover. Before volatility modeling, it is first necessary to determine the most convenient Autoregressive Moving Average (ARMA) models for the conditional mean process. By testing the ARCH effect on the residuals obtained from these models, the most convenient EGARCH (1, 1) model is determined according to the information cri- teria and likelihood value. The assumed distributions for EGARCH models are Normal Distribution (norm), Skewed-Normal Distribution (snorm), Student-t Distri- bution (std), Skewed-Student-t Distribution (sstd), Gen- eralized Error Distribution (ged), Skewed-Generalized Error Distribution (sged), Normal Inverse Gaussian Distribution (nig) and Johnson’s SU Distribution (jsu). EGARCH (1,1) models with different distributions are compared according to Akaike Information Criteria (AIC), Bayes Information Criteria (BIC), Shibata Infor- mation Criteria (SIC), Hannan-Quinn Information Cri- teria (HQIC) and likelihood values. Kanas (1998) defines the residual squares of other variables obtained from the conditional variance model as exogenous variables and made parameter estimates in order to determine the volatility spillover. Accordingly, the EGARCH (1,1) model to be estimated is as follows: ln(σ2 t)=ω+α1zt-i+(γi|zt-i|-E[|zt-i|])]+β1ln(σ2 t-i)+τ1ln(u2 t-i) (3) In the above equation, ut is the residuals obtained from the conditional variance model, and τ1 is the coef- ficient showing the volatility spillover. If the coefficient τ1 is statistically significant, it is concluded that there is a volatility spillover. Au th or s G oa l M et ho do lo gy In du st ry Re gi on Re su lts G ar de br oe k an d H er na nd ez (2 01 3) To e xa m in e th e vo la til ity sp ill ov er s i n oi l, et ha no l, an d co rn p ric es . - M ul tiv ar ia te G A RC H m od el sF oo d & En er gy U SA In th e st ud y, sig ni fic an t v ol at ili ty sp ill ov er is o bs er ve d on ly fr om c or n an d no t v ic e ve rs a. A lso , r es ea rc he rs do n ot d et ec t c ro ss -v ol at ili ty e ffe ct s f ro m o il to c or n m ar ke ts . Th e re su lts d o no t p ro vi de a ny e vi de nc e th at vo la til ity in e ne rg y m ar ke ts h as a si gn ifi ca nt e ffe ct o n pr ic e vo la til ity in th e U S co rn m ar ke t. N az lıo ğl u et a l. (2 01 3) To e xa m in e th e vo la til ity tr an sm iss io n be tw ee n oi l a nd se le ct ed a gr ic ul tu ra l co m m od ity p ric es w hi ch a re w he at , c or n, so yb ea n, a nd su ga r. - Th e va ria nc e ca us al ity te st Fo od & En er gy G lo ba l D at a ar e an al yz ed in tw o pe rio ds a s t he p re -c ris is pe rio d (J an ua ry 1 98 6 - D ec em be r 2 00 5) a nd th e po st - cr isi s p er io d (J an ua ry 2 00 6- M ar ch 2 01 1) to d et er m in e th e im pa ct o f t he fo od p ric e cr isi s. Th e re su lts sh ow ed th at a lth ou gh th er e w as n o ris k of sp ill ov er b et w ee n oi l an d ag ric ul tu ra l c om m od ity m ar ke ts in th e pr e- cr isi s pe rio d, th er e w as a ct ua l o il m ar ke t v ol at ili ty sp ill ov er to ag ric ul tu ra l m ar ke ts - ex cl ud in g su ga r - in th e po st -c ris is pe rio d. Se rr a (2 01 1) To in ve st ig at e pr ic e re la tio ns b et w ee n cr ud e oi l, et ha no l a nd su ga r - a se m ip ar am et ric G A RC H m od el Fo od & En er gy Br az il Th e re su lts re ve al th at in th e lo ng r un , e th an ol p ric es in cr ea se a lo ng w ith th e in cr ea se in b ot h cr ud e oi l a nd su ga r p ric es . 43The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 3.2 The Diebold-Yilmaz approach for the volatility spillover effect Diebold and Yilmaz (2009) describe the return and volatility spillover on the basis of the Vector Autore- gressive (VAR) model. Here, the total spillover index is measured based on the Cholesky decomposition. Nev- ertheless, Diebold and Yilmaz (2012) developed a meth- odology in a later study to evaluate directional spillover in a generalized VAR framework. This VAR framework approach offers variance decomposition that is invariant to the order of variables after that of Koop et al. (1996) and Pesaran and Shin (1998). In the N-component standard VAR model, each entity xi with = 1,…, N is expressed as follows: (4) where yt is Nx1 matrix of dependent variables and φi are NxN matrix of coefficients. εt is the vector of indepen- dently and identically distributed innovations (iid) and follows εt~N(0,Σ) where Σ is variance-covariance matrix. The moving average representation of the VAR model is as follows: (5) where Ai are NxN matrix of moving average coefficients and Ai=φ1Ai-1+φ2Ai-2+…+φpAi-p. Then, given the VAR framework, H-step-forecast error-variance decomposi- tions are defined as follows: (6) where σij represents the standard deviation of the error term, Σ is variance-covariance matrix and ∆i is the selection vector of which ith element is equal to 1 and the other elements are 0. If each element of the decom- position matrix is divided by row sums, each forecasting error decomposition variance will be normalized, thus using the available information in the decomposition matrix to compute the spillover effects as follows: (7) with (H)=1 and (H)=N. In the light of the above definitions and equations from 4.4 to 47, Diebold and Yilmaz (2012) defined total, directional and net spillovers as described below: The total volatility spillovers index based on h-step- ahead forecasts with the following equation: (8) Directional volatility spillovers to i market from other j markets: (9) Directional volatility spillovers from market i to other j markets: (10) The net spillover index is obtained using Equations 4.9 and 4.10 as follows (11) 4. DATA ANALYSIS As signified in the introduction, this study aimed to analyze the relationship between the fruit and vegetable price volatility and the energy price volatility in Turkey. Both energy and product prices consist of the data sets obtained from Eurostat within the scope of the Harmo- nized Index of Consumer Prices (HICP). The scope of energy index includes “electricity, gas and other fuels”. The energy price index is a variable with broader content than the crude oil price, which is widely cited in the lit- erature. It is considered noteworthy to refer to this ener- gy price index in this analysis. The monthly data set obtained from Eurostat con- sists of the Energy Price Index (EPI), the Fruit Price Index (FPI) and the Vegetable Price Index (VPI) between January 2007 and December 2020. Appendix- A, Table-A1 and Table-A2 demonstrate the descriptive statistics and Augmented Dickey-Fuller Unit Root Test results for the data set of these indexes and their loga- 44 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı rithmic returns. Figure 1 shows the time-series plot of the variables. 5. EMPIRICAL RESULTS 5.1 Empirical results for Kanas Approach The convenient conditional mean models for EPI, FPI, and VPI were found to be AR (1), ARMA (2,2), and MA (1), respectively. The output of conditional mean models and ARCH test results are given in Table A3 in Appendix-A. The evaluation of the volatility models is given in Table 2. The results1 in Table 2 manifest that the most ade- quate models are as follows: Sged-EGARCH (1,1) for 1 EGARCH-type volatility models were estimated using “rugarch” R package developed by Ghalanos (2020a, 2020b). EPI; std-EGARCH (1,1) for FPI and norm-EGARCH (1,1) models for VPI. Table A2 points out to the param- eter estimation results and diagnostic test results of the models. It is evident in all three models that all parameters are statistically significant. According to the diagnos- tic test results, the results of Ljung-Box (LB) and Lan- grange-Multiplier (LM) tests indicate that there are no autocorrelation problems in the residuals and heterosce- dasticity problem in the residual squares. The Nyblom Stability Test (NST) results show that there is no struc- tural break according to the NST critical value of 1.49 at 10% confidence level. As in NST, common statistical values calculated for Sign Bias Test (SBT) are given and according to these test statistics, there is no functional error in the conditional volatility model. Looking at the results of the Pearson Goodness of Fit (GoF) test, it can be understood that the empirical distribution of stand- ard residuals and the theoretical distribution are aligned. Figure 1. Time-series Plot of Indexes and Log-returns. 45The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Negative values for EPI and VPI can be found by analyzing the values of “gamma1” parameters that show the leverage effect. In this case, it can be concluded that the effect of bad news on EPI and VPI volatility is higher the effect of good news and increases the volatility per- sistence. The persistent values indicate that the volatility persistence is high for EPI and VPI variables. It is also found that the half-life of persistence in VPI was 9.94 days. Thus, the effect of good news on the volatility is higher for FPI, while the volatility persistence and half- life are lower. This is an indication that good news has a less impact than bad news in the leverage effect. The time-series graph of the volatilities obtained from the models is as described in Figure 2. Table 2. EGARCH(1,1) Model Evaluation depending on Information Criteria and Likelihood Values. dist EPI FPI VPI AIC BIC SIC HQIC L AIC BIC SIC HQIC L AIC BIC SIC HQIC L norm -5.18 -5.10 -5.18 -5.15 436.5 -2.54 -2.44 -2.54 -2.50 216.9 -1.91 -1.81 -1.91 -1.87 164.2 snorm -5.33 -5.23 -5.33 -5.29 449.8 -2.56 -2.45 -2.56 -2.52 219.9 -1.90 -1.78 -1.90 -1.85 164.3 std -5.63 -5.53 -5.63 -5.59 474.7 -2.71 -2.59 -2.71 -2.66 232.0 -1.87 -1.76 -1.87 -1.83 162.3 sstd -5.63 -5.53 -5.63 -5.59 474.7 -2.71 -2.59 -2.71 -2.66 232.0 -1.87 -1.76 -1.87 -1.83 162.3 ged -5.54 -5.44 -5.54 -5.50 467.3 -2.62 -2.51 -2.62 -2.57 224.7 -1.90 -1.79 -1.90 -1.86 164.9 sged -6.17 -6.06 -6.17 -6.13 521.2 -2.65 -2.52 -2.66 -2.60 228.6 -1.88 -1.75 -1.89 -1.83 164.2 nig -6.14 -6.03 -6.15 -6.10 519.0 -2.68 -2.55 -2.68 -2.63 230.8 -1.88 -1.75 -1.88 -1.83 163.9 jsu -6.16 -6.04 -6.16 -6.11 520.1 -2.69 -2.56 -2.70 -2.64 232.0 -1.87 -1.74 -1.88 -1.82 163.5 Normal Distribution (norm), Skewed-Normal Distribution (snorm), Student-t Distribution (std), Skewed-Student-t Distribution (sstd), Generalized Error Distribution (ged), Skewed-Generalized Error Distribution (sged), Normal Inverse Gaussian Distribution (nig) and John- son’s SU Distribution (jsu), Akaike Information Criteria (AIC), Bayes Information Criteria (BIC), Shibata Information Criteria (SIC), Han- nan-Quinn Information Criteria (HQIC), Llikelihood (L). Table 3. The Parameter Estimation of EGARCH(1,1) Models for Price Indices. Parameters sged-EGARCH(1,1) for EPI std-EGARCH(1,1) for FPI norm-EGARCH(1,1) for VPI est Std.Err t-stat sig est Std.Err t-stat sig est Std.Err t-stat sig omega -1.49 0.01 -194.71 0.00 -2.47 1.14 -2.16 0.03 -0.33 0.00 -3793.40 0.00 alpha1 0.35 0.03 11.90 0.00 0.12 0.11 1.05 0.29 0.26 0.00 2136.50 0.00 beta1 0.81 0.00 1176.69 0.00 0.56 0.21 2.71 0.01 0.93 0.00 4454.40 0.00 gamma1 -0.08 0.00 -16.82 0.00 0.39 0.15 2.63 0.01 -0.30 0.00 -2522.70 0.00 shape 0.47 0.01 76.08 0.00 5.69 1.84 3.08 0.00 skew 1.44 0.01 163.05 0.00 stat sig stat sig stat sig LB on SR 1.48 0.75 3.71 0.29 0.25 0.82 LB on SSR 1.19 0.82 0.13 1.00 3.59 0.31 ARCH LM 1.15 0.69 0.10 0.99 2.04 0.46 SBT Joint 0.12 0.99 3.26 0.35 0.60 0.90 Perason GoF 47.67 0.53 42.88 0.72 35.10 0.93 NST Joint 2.41 1.57 1.48 Persistence 0.81 0.56 0.94 Half-life 3.36 1.19 9.94 LB: Ljung-Box SR: Standardized Residuals SSR: Standardized Squared Residuals LM: Langrange Multiplier SBT: Sign Bias Test NST: Nyb- lom Stability Test GoF: Goodness-of-Fit. “omega” is the constant term. “alpha1”is the the ARCH coefficient that is a measure of sign effect. “beta1” is the the ARCH coefficient that is a measure of volatility persistence “gamma1” is the asymmetry coefficient that is a measure of leverage effect. “Normal Distribution (norm), Student-t Distribution (std), Skewed-Student-t Distribution (sstd), Skewed-Generalized Error Distribution (sged). 46 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı It can be said that there was a fluctuation in FPI vol- atility in May 2011 similar to a big shock effect. In this regard, the Iterative Cumulative Sum of Squares (ICSS) introduced by Inclan and Tiao (1994) was applied to all three indexes to locate any structural break in the vari- ance. However, the results showed no break in the vari- ance. To test the volatility spillover of EPI on other vari- ables in this study, the residual squares obtained from the sged-EGARCH (1, 1) model (given in Table 3) were added as an exogenous variable to the volatility models. This step was followed by the parameter estimation. The results are given in Table 4. The diagnostic test results in Table 4 indicate that the models support the hypotheses. According to the results of FPI parameter estimation, it is understood that the “tau1” coefficient (which shows the volatility spillo- ver from EPI to FBI) is not statistically significant, and therefore there is no volatility spillover from EPI to FBI. On the other hand, according to the VPI parameter esti- mations, the “tau1” coefficient is found to be statistically significant leading to the understanding that there is a volatility spillover from EPI to VPI. Hence, it can be concluded that the volatility in the EPI negatively affects the VPI volatility. 4.2 Empirical Results for the Diebold-Yilmaz Approach Table 3 demonstrates the most suitable volatil- ity models determined for EPI, FPI and VPI indexes. Derived from volatility data obtained from these mod- els, the lag value of the VAR model was found to be 1. In addition to this calculation, the VAR (1) model param- eter was estimated. The results of the model estimated by the lag value of selection criteria are respectively pre- sented in Appendix-B, Table B1 and Table B2. The Die- Figure 2. Time-Series Plot of Volatilities Obtained from EGARCH Processes. 47The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 bold-Yilmaz approach results2 obtained on the basis of the VAR model can be seen in Table 5. Before moving on to the results, it is worth reiter- ating that the spillover index shows how much of the total variance that occurs in the variables themselves is caused by other variables. In other words, the Diebold- Yilmaz spillover index demonstrates the contribution of the volatility in price indices to the forecasting error variance. Thus, the results of the total volatility spillo- vers index are based on a 10-step-ahead approach. As these results suggest, it is observed that the volatility spillover from EPI index to other indexes is higher than the others. Furthermore, the VPI is the index that is exposed to the highest volatility trans- fers. The total spillover from EPI to the other indexes is 14.38% and 13.52% of this value belongs to the VPI and the rest belongs to the FPI index. This case points out to shocks in energy prices exhibiting a higher possibility to affect the pattern of other prices in the investigated area. Here, the EPI can be defined as a volatility transmitter. It can be deduced that the risk that the FPI index is exposed to from the outside is low. Indeed, only 2.68% of its current volatility results 2 Diebold-Yılmaz analysis was performed using “Spillover” R package developed by Urbina (2020). from other indexes. On the other hand, it is seen in the VPI index that the externally exposed volatility spillo- ver is 17.97%, and 75.23% of it (13.52%) is due to the EPI. These results also support the outputs obtained from the Kanas (1998) approach. The fact that the total spillover index value is 9.62% points out to a low con- nectedness between these indexes. Nevertheless, it can be seen that the risk in energy prices is transferred to vegetable prices. Due to the high energy prices in Tur- key, for instance, people can only heat their greenhous- es only to protect them from frost rather than proper Table 4. The Parameter Estimation of EGARCH(1,1) Models for Spillover from EPI to FPI and VPI with Diagnostics Tests. Parameters std-EGARCH(1,1) for FPI norm-EGARCH(1,1) for VPI est Std.Err t-stat sig est Std.Err t-stat sig omega -2.52 1.08 -2.33 0.02 -0.28 0.00 -7553.23 0.00 alpha1 0.13 0.11 1.11 0.27 0.28 0.00 5849.33 0.00 beta1 0.56 0.19 2.90 0.00 0.94 0.00 7751.43 0.00 gamma1 0.38 0.15 2.57 0.01 -0.28 0.00 -10783.22 0.00 shape 5.74 1.85 3.10 0.00 tau1 (EPI spillover) 14.36 115.70 0.73 0.47 -7.75 0.01 -686.85 0.00 stat sig stat sig LB on SR 9.76 0.01 3.54 0.32 LB on SSR 3.21 0.37 0.12 1.00 ARCH LM 1.87 0.50 0.08 0.99 SBT Joint 0.40 0.94 3.17 0.37 Perason GoF 48.87 0.48 36.89 0.90 NST Joint 1.60 1.53 Persistence 0.56 0.94 Halflife 1.19 11.68 LB: Ljung-Box SR: Standardized Residuals SSR: Standardized Squared Residuals LM: Langrange Multiplier SBT: Sign Bias Test NST: Nyb- lom Stability Test GoF: Goodness-of-Fit. “omega” is the constant term. “alpha1”is the the ARCH coefficient that is a measure of sign effect. “beta1” is the the GARCH coefficient that is a measure of volatility persistence “gamma1” is the asymmetry coefficient that is a measure of leverage effect. “tau1” is the coefficient showing the volatility spillover Normal Distribution (norm), Student-t Distribution (std), Skewed- Student-t Distribution (sstd), Skewed-Generalized Error Distribution (sged). Table 5. Diebold-Yilmaz Generalized Directional Spillover Output. EPI FPI VPI Contribution from others EPI 91.80 0.27 7.92 8.20 FPI 0.86 97.32 1.82 2.68 VPI 13.52 4.45 82.03 17.97 Contribution to others (spillover) 14.38 4.72 9.75 9.62 Contribution to others including own 106.18102.04 91.78 300.00 Net Spillover 6.18 2.04 -8.22 Total Spillover Index 9.62% 48 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı heating. Despite this widespread use of limited energy, volatility in energy prices affects greenhouse costs. 31 million tons of vegetables were produced in Turkey in 2019 as the world’s 4th largest producer of fresh vege- tables. 23.2 million tons of these crops were grown in agricultural or open areas, and 7.8 million tons were produced in greenhouses. As a matter of fact, around 0.6 million tons of fruits are produced in greenhouses (MAF, 2021). According to the results of the analysis, this explains the reason why the vegetable price index is subject to volatility from the spillover of the fluctuat- ing energy prices. Within this framework, the average spillover effects over the full sampling period are obtained by gener- alized spillover analysis. Diebold and Yilmaz (2009, 2012) stated that full sample spillover measurements cannot clearly ref lect the important sustained and cyclical movement in spillovers. Thus, they developed a rolling window framework that allows time-varying spillover indices to overcome their shortcomings in the spillover index, using a 48-month subsample. In this line, the following graphs show the estimation of the dynamic net and total spillover indexes. These rolling windows were obtained using the 10-step-ahead fore- casting spillovers. The date that stands out at first glance in the roll- ing net spillover index is May 2011, when consumer prices increased by 2.42% and annual inflation rose to 7.17 %. Coupled with the base effect, the high increases in fresh fruit prices due to seasonal transitions marked the rationale behind this rise. In this period, fresh fruit prices increased by 76.12% on a monthly basis, well above the average of the previous period (TCBM, 2011). Therefore, the FPI became the volatility trans- mitter in May 2011 and created a net volatility spillo- ver of 40.05% on the forecasting error variances of other indices. Thus, the total spillover index was esti- mated as 44.33%. Figure 3. The Top-Down Rolling Net Spillovers Indexes for EPI, FPI and VPI. 49The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 6. CONCLUSION Input costs have a significant share in setting the prices of agricultural products and ensuring sustainable production. Increases especially in energy prices may have an effect on many items from production to deliv- ery of products to final consumers. These items include but are not limited to fertilizers, chemicals, irrigation, production, storage and transportation costs. In this context, stable pricing in the field of energy is essential for the price stability of agricultural products. Howev- er, energy prices are not reflected on every agricultural product at the same level. Thus, this study analyzed the prices of fruits and vegetables as the category containing the highest price fluctuations compared to other agricul- tural products. Two different analysis methods, Kanas (1998) and Diebold-Yilmaz (2012), were used in the study and it is concluded that the results obtained from both meth- ods support each other. After the parameter estima- tion of the relevant ARMA models for logarithmic changes of energy, fruit and vegetable price indices, the ARCH effect was determined in the residuals of condi- tional mean models. To identify the residuals of condi- tional mean models, volatility modelling was performed through the EGARCH conditional variance model introduced to the literature by Nelson (1991). Param- eter estimations were made for the EGARCH models by assuming eight different conditional probability dis- tributions. In this regard, sged-EGARCH, std-EGARCH and norm-EGARCH were found to be the most com- patible models for EPI, FPI and VPI, respectively. Con- sidering the outputs of these models indicating the lev- erage effect, it can be seen that the volatility of energy and vegetable price indexes is more affected by bad news in the market. On the other hand, the volatility of fruit price index appears to be mostly affected by good news. At the same time, it can be understood that the volatil- ity persistence and half-life of energy and vegetable price indexes are higher according to the fruit price index. As an exogenous variable in other variables’ volatility modelling, we used the residual squares obtained from the volatility model estimated for the energy price index on the basis of the Kanas (1998) approach. Consequent- ly, it is concluded that there is a statistically significant volatility spillover from the energy to the vegetable price index, while not from the energy index to the fruit price index. This clarifies that the fluctuations in energy pric- es increase the risk and uncertainty in vegetable prices. In the Diebold-Yilmaz (2012) approach, the volatility spillover index results were obtained by using the VAR model for the volatilities attained from the EGARCH models, which were found to be most compatible for the indexes. Accordingly, it is understood that the volatility Figure 4. The Rolling Total Spillovers Index. 50 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı spillovers from the energy to the vegetable price index and the fruit price index are 13.52% and 0.86%, respec- tively. In addition, these calculations show that the risk that the fruit price index is exposed to from the outside is rather low, and only 2.68% of the current volatility are due to other indexes. In the case of the vegetable price index, however, it is found that 75.23% of the net vola- tility index is from energy prices. These results are well overlapping with the results obtained by applying the Kanas (1998) approach. The fact that the total spillover index value is 9.62% points out to a low connectedness between these indexes. As we mentioned in the find- ings section, the share of greenhouse cultivation in veg- etable production is considerably higher than in fruit production. At the same time, vegetable production is higher than fruit production in Turkey. In this case, the amount of energy input needed in vegetable production is naturally higher than fruit production. In addition to these, Turkey’s dependence on foreign energy, increases in the exchange rate, and price increases in the global energy market are other factors to be considered. Thus, it is an expected result that the spillover effect of the energy price index volatility on the vegetable price index is greater than the fruit price index. Another production input that has an indirect effect on energy prices (which, in turn, affect vegetable and fruit prices) is the price of fertilizers used in farming. Indeed, it may well be observed that fertilizer produc- tion is decreasing due to the increasing costs of natural gas and electricity all over the world. This is the indirect factor that causes the upward volatility trend of fruit and vegetable price indices in Turkey. In other words, the volatility of energy prices is quite high in the country. Elaborated in this study from a scientific perspec- tive, the increasing energy prices can be associated with expensive foods due to the increasing costs of processing, transportation, and distribution of agricultural products. In addition, the effect of energy prices on food prices also varies depending on the distance traveled by road. Largely focusing on the fluctuating energy prices and their impact on agricultural products, the results of this study provide important implications for poli- cymakers. In this sense, policymakers should urgently do make improvements in their exchange rate policies and the oil reserve system in order to reduce the nega- tive impact of fluctuations in oil prices on the agricul- tural sector in Turkey, which is an oil importer country. They should also pay as much attention as possible to the global oil markets and their impact on transporta- tion costs. In parallel with the developments in the ener- gy industry, there is also a need to design preventive/ protective regulations to mitigate the agricultural price risks and stabilize the market. In addition, policymakers should take measures to prevent speculative behaviors in the markets in an attempt to prevent price increases of food. In addition to these measures and regulations, governments must support farmers so that they main- tain their resilience, while also protecting consumers against price changes. On the other hand, it is necessary to expand the use of alternative energy sources such as biofuels, wind, and solar energy in order to reduce Tur- key’s dependence on foreign-sourced oil consumption. Similar to the rest of the world, Turkey can grow fruits for a much longer time period than vegetables. According to the results obtained from our study, the time-wise conclusion is that that energy prices have a greater effect on agricultural products grown in a short- er time. Also, the study results are reasonable in the sense that vegetable production in greenhouses is often in greater amounts than fruit production, while requir- ing a high amount of energy consumption. REFERENCES Adrangi, B., Chatrath, A., Macri, J., & Raffiee, K. (2017). Crude oil price volatility spillovers and agricul- tural commodities: a study in time and frequency domains. Review of Economics & Finance, 9(3), 42-56. Akder, A. H., Çakmak, E. H., Sürmeli, B., & Veziroğlu, S. (2020). Piyasa yapısı, aracılık faaliyetleri ve tarımsal örgütlenme. No: TÜSİAD-T/2020-03/614, İstanbul: TUSIAD. Algan, N., et al., (2017). Enerji fiyatlarındaki volatilitenin makroekonomik performans üzerine etkisi. Interna- tional Conference On Eurasian Economies, (291-300. ss.). İstanbul. Alghalith, M. (2010). The interaction between food prices and oil prices. Energy Economics, (32), 1520-1522. Algieri, B. (2016). A roller coaster ride: An empirical investigation of the main drivers of wheat price. M. Kalkuhl, J. Von Braun, & M. Torero (Ed.), Food price volatility and its implications for food security and pol- icy, (207-237. ss.). Springer International Publishing AG Switzerland. Baffes (2011). The energy/non-energy price link: Chan- nels, issues and implications. I. Piot-Lepetit, R. M’Barek (Ed.), Methods to Analyse Agricultural Commodity Price Volatility, (31-. ss.). Springer Science+Business Media. Balcılar, M., & Bekun, F.V. (2019). Do oil prices and exchange rates account for agricultural commodity market spillovers? Evidence from the Diebold and Yilmaz Index. Agrekon, 59(3), 366–385. 51The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Bollerslev, T. (1986). Generalized Autoregressive Con- ditional Heteroskedasticity. Journal of Econometrics, 31(3), 307–327. doi:10.1016/0304-4076(86)90063-1 Bollerslev, T., Engle, R.F. and Nelson, D.B. (1994). ARCH Models, Handbook of Econometrics, 4, 2959-3038. Cabrera, B.L., & Schulz, F. (2016). Volatility linkages between energy and agricultural commodity prices. Energy Economics, (54), 190-208. Diebold, F. X., & Yilmaz, K. (2009). Measuring financial asset return and volatility spillovers, with applica- tion to global equity markets. The Economic Journal, 119(534), 158-171. Diebold, F. X., & Yilmaz, K. (2012). Better to give than to receive: Predictive directional measurement of vola- tility spillovers. International Journal of Forecasting, 28(1), 57-66. Engle, R. F. (1982). Autoregressive Conditional Het- eroskedasticity with Estimates of The Variance of United Kingdom Inflation. Econometrica, 50:4, pp. 987–1007. Fasanya, I., & Akinbowale, S. (2019). Modelling the return and volatility spillovers of crude oil and food prices in Nigeria. Energy, (169), 186-205. Gardebroek, C., & Hernandez, M.A. (2013). Do energy prices stimulate food price volatility? Examining vol- atility transmission between US oil, ethanol and corn markets. Energy Economics, (40), 119-129. Ghalanos, A. (2020a). rugarch: Univariate GARCH mod- els. R package version 1.4-2. Available at: https:// cran.r-project.org/web/packages/rugarch/rugarch.pdf Ghalanos, A. (2020b).  Introduction to the rugarch pack- age. Technical Report Available at: https://cran.r-pro- ject.org/web/packages/rugarch/vignettes/Introduc- tion_to_the_rugarch_package.pdf Gilbert, C.L., & Mugera, H.K. (2014). Food commodity prices volatility: the role of biofuels. Natural Resourc- es, (5), 200-212. Hau, L., Zhu, H., Huang, R., & Ma, X. (2020). Heteroge- neous dependence between crude oil price volatility and China’s agriculture commodity futures: Evidence from quantile-onquantile regression. Energy, (213), 1-19. Huchet-Bourdon, M. (2011). Agricultural commodity price volatility: An overview. OECD Food, Agriculture and Fishing Working Paper, (52), 1-51. Inclan, C., & Tiao, G. C. (1994). Use of cumulative sums of squares for retrospective detection of changes of variance. Journal of the American Statistical Associa- tion, 89(427), 913-923. Kalkuhl, M., Braun, J., & Torero, M. (2016). Volatile and extreme food prices, food security, and policy: An overview. Kalkuhl, J. Von Braun, & M. Torero (Ed.), Food price volatility and its implications for food secu- rity and policy, (207-237. ss.). Springer International Publishing AG Switzerland. Kanas, A. (1998). Volatility Spillovers across Equity Mar- kets: European Evidence. Applied Financial Econom- ics. 8, 245-256. Koirala, K.H., Mishra, A.K., D’Antoni, J.M., & Mehlhorn, J.E. (2015). Energy prices and agricultural commod- ity prices: Testing correlation using copulas method. Energy, (81), 430-436. Koop G., Pesaran M.H. & Potter S.M. (1996). Impulse response analysis in nonlinear multivariate models. J Econometr 74(1):119–147 Kornher, L., & Kalkuhl, M. (2013). Food price volatility in developing countries and its determinants. Quar- terly Journal of International Agriculture, 52(4), 277- 308. Lucotte, Y. (2016). Co-movements between crude oil and food prices: A post-commodity boom perspective. Economics Letters, (147), 142-147. MAF, Ministry of Agriculture and Forestry of the Repub- lic of Turkey (2021). The Current Situation in Green- house Production. Retrieved from https://www.tari- morman.gov.tr/Konular/Bitkisel-Uretim/Tarla-Ve- Bahce-Bitkileri/Ortu-Alti-Yetistiricilik Mawejje, J. (2016). Food prices, energy and climate shocks in Uganda. Agricultural and Food Economics, 4(4), 1-18. MFA, Ministry of Foreign Affairs of the Republic of Tur- key (2021). Turkey’s International Energy Strategy. Retrieved from https://www.mfa.gov.tr/turkeys-ener- gy-strategy.en.mfa Nazlıoğlu, S., Erdem, C., & Soytas, U. (2013). Volatil- ity spillover between oil and agricultural commodity markets. Energy Economics, (36), 658-665. Nelson, D.B. (1991). Conditional Heteroskedasticity in Asset Returns: A New Approach, Econometrica, 59, 2, 347-370. Nwoko, I.C., Aye, G.C., & Asogwa, B.C. (2016). Oil price and food price volatility dynamics: The case of Nige- ria. Cogent Food & Agriculture, (2), 1-13. Olah, J., Lengyel, P., Balogh, P., Harangi-Rakos, M., & Popp, J. (2017). The role of biofuels in food commod- ity prices volatility and land use. Journal of Competi- tiveness, 9(4), 81-93. Öztürk, H. H., Yaşar, B., & Eren, Ö. (2010). Tarımda enerji kullanımı ve yenilenebilir enerji kaynakları. Türkiye Ziraat Mühendisliği VII. Teknik Kongresi, (909-932. ss.). Ankara, Türkiye. Pesaran, M. H. and Shin, Y. (1998). Generalized impulse response analysis in linear multivariate models. Eco- nomics Letters, 58(1):17-29. 52 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı Radmehr, R., & Henneberry, S.R. (2020). Energy price policies and food prices: Empirical evidence from Iran. Energy, 13(4031), 1-15. Santeramo F.G., Lamonaca E. (2019): On the drivers of global grain price volatility: an empirical investiga- tion. Agricultural Economics – Czech, 65: 31–42 Sarwar, M.N., Hussain, H., & Maqbool, M.B. (2020). Pass through effects of oil price on food and non-food prices in Pakistan: A nonlinear ARDL approach. Resources Policy, (69), 1-10. Serra, T. (2011). Volatility spillovers between food and energy markets: A semiparametric approach. Energy Economics, (33), 1155-1164. Statista (2021a). Leading producers of fresh vegetables worldwide in 2019. Retrieved from https://www. statista.com/statistics/264662/top-producers-of-fresh- vegetables-worldwide/ Statista (2021b). Leading producers of fresh fruitsworld- wide in 2019. Retrieved from https://www.statista. com/statistics/279164/global-top-producers-of-select- ed-fresh-fruit-worldwide/ Tadesse, G., Algieri, B., Kalkuhl, M., & Braun, J. (2014), Drivers and triggers of international food price spikes and volatility. Food Policy, (47), 117-128. Tadasse, G., Algieri, B., Kalkuhl, M., & Braun, J. (2016). Drivers and triggers of international food price spikes and volatility. Kalkuhl, J. Von Braun, & M. Torero (Ed.), Food price volatility and its implications for food security and policy, (207-237. ss.). Springer International Publishing AG Switzerland. Taghizadeh-Hesary, F., Rasoulinezhad, E., & Yoshino, N. (2019). Energy and Food Security: Linkages through Price Volatility. Energy Policy, (128), 796–806. TCMB (2011). Türkiye Cumhuriyet Merkez Bankası 2011 Mayıs Ayı Fiyat Gelişmeleri. Available at: https:// www.tcmb.gov.tr/wps/wcm/connect/3b08cd4b-0c0d- 4948-b8e4-bfa253342173/AFiyatMayis11.pdf ?MO D=AJPERES&CACHEID=ROOTWORKSPACE- 3b08cd4b-0c0d-4948-b8e4-bfa253342173-m3fBcm6 Tiwari, A.K., Nasreen, S., Shahbaz, M., & Hammoudeh, S. (2020). Time-frequency causality and connectedness between internationalprices of energy, food, industry, agriculture and metals. Energy Economics, (85), 1-18. Urbina, J. (2020). Spillover Index Based on VAR Mod- elling. R package version 0.1. Available at: https:// cran.r-project.org/web/packages/Spillover/Spillover. pdf Volpe, R., Roeger, E., & Leibtag, E. (2013). How trans- portation costs affect fresh fruit and vegetable prices. Economic Research Service, (160), 1-32. Yıldırım, A. E. (2020). Elektrik fiyatı tarımsal üretimi teh- dit ediyor. Available at: https://www.tarimdunyasi. net/2020/02/25/elektrik-fiyati-tarimsal-uretimi-teh- dit-ediyor/. Zhang, Z., Lohr, L., Escalante, C., & Wetzstein, M. (2010). Food versus fuel: What do prices tell us?. Energy Poli- cy, (38), 445–451. 53The Role of Energy on the Price Volatility of Fruits and Vegetables: Evidence from Turkey Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 APPENDIX-A Table A1. Summary Statistics. Variable Mean Median Min Max Std. Dev. Skewness Ex. kurtosis 5% Perc. 95% Perc. IQrange energy 96.50 93.68 45.33 184.46 36.02 0.82 0.07 45.75 175.00 39.66 fruit 94.05 82.58 40.38 217.00 43.83 0.98 0.06 47.88 194.70 56.74 vegetable 96.89 82.71 36.15 253.72 51.55 1.15 0.63 41.24 216.29 68.24 logret(energy) 0.00 -0.01 -0.08 0.09 0.02 1.60 5.25 -0.02 0.06 0.01 logret(fruit) 0.00 0.00 -0.36 0.46 0.08 0.16 7.82 -0.14 0.10 0.08 logret(vegetable) 0.00 0.01 -0.30 0.28 0.10 -0.06 0.42 -0.19 0.19 0.12 Table A2. Augmented Dickey-Fuller Unit Root Test Results. energy fruit vegetable logret(energy) logret(fruit) logret(vegetable) With Constant t-Statistic 1.39 5.28 1.02 -9.08 -6.10 -8.42 Prob. 1.00 1.00 1.00 0.00 0.00*** 0.00*** With Constant & Trend t-Statistic -0.49 2.15 -1.31 -9.05 -6.98 -8.46 Prob. 0.98 1.00 0.88 0.00 0.00*** 0.00*** Without Constant & Trend t-Statistic 3.64 6.46 2.80 -9.11 -6.09 -8.43 Prob. 1.00 1.00 1.00 0.00 0.00*** 0.00*** *** indicates that log-returns of EPI, FPI and VPI has no unit root. Table A3. ARMA Model Outputs for EPI, FPI and VPI. Coefficients AR(1) for EPI ARMA(2,2) for FPI MA(1) for VPI est sig est sig est sig const −3.99718e-05 0.99 −0.000537185 0.71 0.00 0.99 phi_1 0.33 0.00 1.55 0.00 phi_2 −0.795506 0.00 theta_1 −1.76350 0.00 0.41 0.00 theta_2 0.83 0.00 Mean dependent var 0.00 −1.91e-17 0.00 Mean of innovations 0.00 0.00 −0.000061 R-squared 0.11 0.27 0.13 Log-likelihood 417.55 206.90 154.05 Schwarz criterion −819.7365 −383.0893 −292.7434 S.D. dependent var 0.02 0.08 0.10 S.D. of innovations 0.02 0.07 0.10 Adjusted R-squared 0.11 0.26 0.13 Akaike criterion −829.0905 −401.7972 −302.0974 Hannan-Quinn −825.2939 −394.2041 −298.3008 ARCH LM test 56.00 (9.65e-10)*** 51.3 (7.91e-09)*** 15.33 (3.20e-02)** ** and *** indicate that there is an ARCH effect on residuals. 54 Bio-based and Applied Economics 11(1): 37-54, 2022 | e-ISSN 2280-6172 | DOI: 10.36253/bae-10896 Harun Uçak, Esin Yelgen, Yakup Arı APPENDIX-B Table B1. VAR Lag Selection. lags loglik p(LR) AIC BIC HQC 1 1484.04 -18.51* -18.28* -18.42* 2 1485.41 0.97 -18.42 -18.01 -18.26 3 1487.78 0.86 -18.34 -17.76 -18.10 4 1494.17 0.17 -18.30 -17.55 -18.00 5 1499.96 0.24 -18.26 -17.34 -17.89 6 1508.76 0.04 -18.26 -17.16 -17.81 7 1521.01 0.00 -18.30 -17.03 -17.78 8 1526.37 0.30 -18.26 -16.81 -17.67 *The most convenient VAR Lag is selected 1. Table B1. VAR(1) Model Output. Dependent Var Energy Volatility (evol) Fruit Volatility (fvol) VegetableVolatility (vvol) est sig est sig est sig const 0.01 0.00 0.05 0.00 0.01 0.07 evol[-1] 0.77 0.00 −0.381 0.35 0.44 0.03 fvol[-1] −0.0063 0.49 0.30 0.00 −0.0383 0.29 vvol[-1] −0.0263 0.02 −0.0193 0.83 0.82 0.00 Mean dependent var 0.02 0.06 0.06 Sum squared resid 0.00 0.13 0.13 R-squared 0.60 0.10 0.10 F(3, 162) 82.29 5.79 5.79 rho −0.021 −0.004 −0.004 S.D. dependent var 0.01 0.03 0.03 S.E. of regression 0.00 0.03 0.03 Adjusted R-squared 0.60 0.08 0.08 sig(F) 0.00 0.00 0.00 Durbin-Watson 2.04 2.01 2.01 All lags of evol F(1, 162) 241.41 [0.0000] 0.86818 [0.3528] 5.0403 [0.0261]** All lags of fvol F(1, 162) 0.4696 [0.4942] 15.226 [0.0001] 1.1422 [0.2868] All lags of vvol F(1, 162) 5.6895 [0.0182] 0.044003 [0.8341] 354.2 [0.0000] **The test statistics of all lags of evol in vvol model indicates that evol Granger causes vvol.