Energy Price Policies and Food Prices: Empirical Evidence from Iran

During the last decade, the rising trend in energy prices and its potential effect on food prices have become a controversial issue between policy-makers and economists. Therefore, research addressing the relationship between food and macroeconomic variables, such as energy prices, will be useful in providing information for the design of appropriate economic policies. This study uses data from Iran to examine the impacts (shortand long-term) of exchange rate and energy prices on food prices. Iran is a good case study as in recent years its consumers have faced a rapid increase in both fuel and food prices. The variables employed in this study are the prices of ten food products, exchange rate (the value of Iranian rial per US dollar), and petroleum prices. All‘data in this study are from the Statistical Centre of Iran (SCI). We employ the panel unit root test, Pedroni co-integration tests, Pooled Mean Group (PMG), Mean Group (MG), and Dynamic Fixed Effects (DFE) estimation techniques, applied to a panel of monthly prices for ten food products for the period of March 1995 to February 2018. Results show that in both the shortand long-run, food prices would increase in response to an increase in energy prices. Findings also suggest that the appreciation of the United States Dollar (USD) in terms of the Iranian rial exerts a positive and significant impact on food prices in the long run.


Introduction
Iran is one of the world's most energy-rich countries, accounting for about 11% of world oil reserves and 15.3% of world gas resources, and therefore plays an important role in the global energy markets [1]. In Iran, the government's annual budget is heavily dependent on energy resources, as about 60% of the government's income comes from oil and gas resources. Note that in Iran, the oil and gas industry is not private and is owned by the government. Energy prices in Iran are often determined by the government and therefore, they have little dependence on global energy market fluctuations. As a result, energy prices in Iran are often divorced from global prices as the government uses the energy price policy for welfare and income support. Due to direct and indirect energy subsidies, the prices of energy in Iran, especially gasoline, have been some of the cheapest in the world, so that gasoline prices have been in the range of 0.10-0.40 United States Dollar (USD) per liter over the last forty years [2]. The social and environmental problems of these supportive policies, on the one hand, the intensification of economic sanctions, and consequently increase in the budget deficit on the other hand were strong reasons for the government to pursue energy price reform policies. Eliminating energy subsidies officially began in 2010, which is considered the greatest one-shot rise in energy prices anywhere in the world [3]. Given the adverse effects of these policies on household welfare, especially in low-income deciles, the sudden rise in energy prices are often met by widespread A review of the literature indicates that the majority of studies in this area focus on investigating the link between global oil prices and global food prices. In fact, there is a research gap in how domestic energy prices could influence domestic food prices. In addition, to our knowledge, there has been no study examining the impact of energy price policies on food prices in oil-exporting countries with severe trade sanctions. One of the main reasons for the scarce literature available in these countries could be difficulty in accessing data resources. Hence, due to the different economic structure of these countries with free-trade countries, conducting an empirical study in this scientific field in this group of countries can significantly contribute to the ongoing debate on the relationship energy price policies and food prices. Given the importance of this issue, the current study attempts to fill these research gaps by conducting an empirical investigation of the linkage between domestic energy and food prices, using Iran as a case study.
The key target of the current study is to explore the short-and long-term effects of energy prices and the exchange rate on ten individual food prices, using the panel Autoregressive Distributed Lag (ARDL) co-integration technique. This study contributes to the literature in several important ways. First, the impacts of domestic energy prices and exchange rate on a set of domestic food prices are investigated and, second, we use monthly data from March 1995 to February 2018, to assess the shortand long-term linkages between the variables of concern (the impact exchange rate and energy prices on food prices). Therefore, the second contribution of this study is in its use of a long enough time series to examine the long-term impacts in addition to the short-term effects. To the best of our knowledge, this is the first study that evaluates the direct and indirect impacts of energy prices on food prices in Iran. Figure 1 shows the main steps of modeling and empirical analysis for examining the relationship between research variables. In the first step, it is essential to check the level of stationary of series to investigate the long-and short-term relationships among research variables. When the variables are not stationary, it is important to carry out a panel co-integration test. If this test is verified, the ARDL co-integration technique could be employed to analyze the research data. Pooled Mean Group (PMG), Dynamic Fixed Effect (DFE), and Mean Group (MG) estimation techniques have been used to estimate the short-and long-term coefficients. Finally, in order to check the robustness of the long-run coefficients, the Fully Modified Ordinary Least Square (FMOLS) and Dynamic Ordinary Least Squares (DOLS) are utilized. In the continuation of this section, more details of each step are provided. A review of the literature indicates that the majority of studies in this area focus on investigating the link between global oil prices and global food prices. In fact, there is a research gap in how domestic energy prices could influence domestic food prices. In addition, to our knowledge, there has been no study examining the impact of energy price policies on food prices in oil-exporting countries with severe trade sanctions. One of the main reasons for the scarce literature available in these countries could be difficulty in accessing data resources. Hence, due to the different economic structure of these countries with free-trade countries, conducting an empirical study in this scientific field in this group of countries can significantly contribute to the ongoing debate on the relationship energy price policies and food prices. Given the importance of this issue, the current study attempts to fill these research gaps by conducting an empirical investigation of the linkage between domestic energy and food prices, using Iran as a case study.

Materials and Methods
The key target of the current study is to explore the short-and long-term effects of energy prices and the exchange rate on ten individual food prices, using the panel Autoregressive Distributed Lag (ARDL) co-integration technique. This study contributes to the literature in several important ways. First, the impacts of domestic energy prices and exchange rate on a set of domestic food prices are investigated and, second, we use monthly data from March 1995 to February 2018, to assess the shortand long-term linkages between the variables of concern (the impact exchange rate and energy prices on food prices). Therefore, the second contribution of this study is in its use of a long enough time series to examine the long-term impacts in addition to the short-term effects. To the best of our knowledge, this is the first study that evaluates the direct and indirect impacts of energy prices on food prices in Iran. Figure 1 shows the main steps of modeling and empirical analysis for examining the relationship between research variables. In the first step, it is essential to check the level of stationary of series to investigate the long-and short-term relationships among research variables. When the variables are not stationary, it is important to carry out a panel co-integration test. If this test is verified, the ARDL co-integration technique could be employed to analyze the research data. Pooled Mean Group (PMG), Dynamic Fixed Effect (DFE), and Mean Group (MG) estimation techniques have been used to estimate the short-and long-term coefficients. Finally, in order to check the robustness of the longrun coefficients, the Fully Modified Ordinary Least Square (FMOLS) and Dynamic Ordinary Least Squares (DOLS) are utilized. In the continuation of this section, more details of each step are provided.

Empirical Models
According to the aforementioned discussion, we expect food prices to be linked to energy prices and exchange rates. For model specification, we consider a log-log regression as follows: In Equation (1), the subscripts i = 1, 2, . . . , 10, are for ten food products, t = 1, 2, . . . , T, are for months March 1995 to February 2018; FP is the food prices; EP represents the energy prices, and EX depicts the exchange rate (the value of Iranian rial per USD). Nazlioglu and Soytas [22] and Shaari, et al. [26] argue that energy prices are important determinants of the cost of food production and transportation (direct effect). Energy prices also have a notable impact on food prices through exchange rates (indirect effect). Hence, it is expected that rising energy prices would lead to an increase in food prices. Henneberry, et al. [27], Henneberry, et al. [28], and Radmehr and Henneberry [11] indicate that exchange rates are important factors impacting agricultural trade so that a depreciation of the national currency is expected to lead to an increase in agricultural exports (foreign demand). Therefore, the appreciation of the USD in terms of the Iranian rial is expected to contribute significantly to the increase in food prices. It is worth noting that an increase in the exchange rate means a depreciation of the Iranian rial (appreciation of the USD).
In the current study, we employ the econometric technique proposed by Pesaran, et al. [29]. The study investigates the long-and short-term relationships among food prices, exchange rate, and energy prices within the ARDL framework as follows: Here, Z t = (lnEP, lnEX) In Equation (2), Z t is the vector of the independent variables, while α i is the food products-level fixed effects, δ ij is the coefficient of the lagged lnFP it , and ϕ ij represents the coefficients of the lagged independent variables.
The ARDL co-integration technique has attracted a lot of attention due to its capabilities and advantages over traditional co-integration methods. This approach considers the problems of endogeneity, as well as calculates the short-and long-run coefficients individually in a single model. Furthermore, this technique could be used when the integration ordered variables are mixed of I (0) or/and I (1) [30].
The panel ARDL by modifying the above equation into the error correction term (ECT) is as follows: where In Equation (4), ϑ i (lnFP it−1 − θ i Z it ) denotes the coefficient of speed of adjustment to the long-term status; θ i is the estimated long-run coefficients for independent variables.
To estimate the dynamic panel data models, several techniques are developed in the literature. One of the most widely used of these techniques is the PMG estimator introduced by Pesaran, Shin and Smith [29]. In this technique, it is assumed that the short-term coefficients, intercepts, and the Energies 2020, 13, 4031 5 of 15 variance of the errors are different across food groups, while the coefficients of long-term are constrained to be homogeneous across the food groups. The second approach is the MG estimator. In this method, different regressions for each food group are estimated separately and then the unweighted mean of coefficients is calculated for the entire panel. The third approach is DFE. In the DEF technique, the long-and short-term coefficients are assumed to be equal across the food groups, while excluding the intercept coefficient. To select the most appropriate estimation technique, the Hausman tests are employed.

Data
In the current study, we constructed a set of panel data of ten food products (bread and cereals, dairy, aquatic meats, beverages, sugar and sweets, fruits and dried fruits, meats, oils and fats, vegetables, and other) over the monthly period from March 1995 to February 2018. The variables employed in this study are the Iranian Consumer Price Index (CPI) of ten food products which is considered as a proxy of food prices, exchange rate (the value of Iranian rial per USD), and petroleum prices. Required data for analysis is obtained from the Statistical Centre of Iran (SCI). The data used was transformed into natural logarithms form. The summary statistics of the variables are presented in Table 1. Based on the results of this table, the average logarithmic form of food prices was 2.901 and its maximum and minimum were 5.540 and 0.073, respectively. The average petroleum price was 8.418, with a small variability (0.286), and the average logarithmic form of the exchange rate was 5.066. In addition, a comparison between sub-group foods shows that sugar and sweets had the highest average (3.293) and vegetables had the lowest average (2.368).

Unit Root and Co-Integration Tests
The findings of the Levin, Lin, and Chu (LLC) unit root test are reported in Table 2. It is worth noting that the panel data used in this study are balanced, which means that the data for each food group and for every time period is available. Table 2 shows that the null hypothesis of this test can be accepted for the variable of lnFP. In other words, the variable of lnFP is non-stationary at the level, while stationary at first difference. Therefore, it is likely that there is a long-run co-integrating linkage among the variables of concern. Hence, to examine the existence of a long-run relationship between study variables, the Pedroni co-integration test developed by Pedroni [31] and Kao co-integration test advanced by Kao [32] were employed. The Pedroni co-integration test includes seven statistics: four statistics (panel v-statistic, rho-statistic, PP-statistic, and ADF-statistic) related to the within-dimension, and three statistics (group rho-statistic, PP-statistic, and ADF-statistic) related to the between-dimension. The null hypothesis of these two groups of statistics is no co-integration. The null hypothesis for the Kao test is similar to the Pedroni test. The findings of these tests are reported in Table 3. As can be seen in Table 3, in the Pedroni test, all statistics except the panel v-statistic statistically reject the null hypothesis of no co-integration among the research variables. Similarly, the findings of Kao test approve the presence of co-integration between study variables. Above all, we can conclude that there is explicit evidence of a long-term co-integration equilibrium between variables.  Although the findings of this section approve the existence of co-integration among the variables in this study, they do not provide the magnitude of this linkage. Therefore, in the next step, the long-term and short-term parameters are estimated by the ARDL co-integration technique.

Panel ARDL Results
In this section, we employed the three methods of the PMG, DFE, and MG to estimate the short-and long-run effects of the exchange rate and energy prices on food prices. Table 4 indicates the results of the short-term coefficients, speed of adjustment coefficients, and long-term coefficients for these three estimators. The long-term coefficients of energy prices were positive and significant in the PMG and DFE estimators, while insignificant in the MG estimator. However, energy prices significantly contributed to food prices in the short-run based on the results of the PMG, MG, and DFE estimation techniques. In this study, to compare the findings of three estimators and select the most appropriate of the method, we utilized the Hausman test. Based on this test, pairwise comparisons were performed between the MG and the PMG techniques, and the MG and the DFE techniques. The null hypothesis in the comparison between the MG and the PMG is that the PMG technique is efficient and consistent as well as preferable to the MG technique. Accepting the null hypothesis in the comparison between MG and DFE indicates that the DFE technique is efficient and consistent as well as preferable to the MG estimator.   Table 4, the value of the Hausman test in the comparison between the MG and the PMG is statistically insignificant, showing that the PMG method is more efficient than the MG. In addition, in the term of the MG and the DFE, the results of the Hausman test confirm that the DFE estimator is more appropriate than the MG technique to estimate the ARDL panel. Therefore, it can be concluded that the PMG and DFE techniques are more appropriate than the MG technique. However, given that the PMG estimator allows heterogeneity in the short-term, in the continuation of this section, the results of this estimator will be examined.
A quantitative evaluation of the long-term effect of energy prices on food prices indicates that a 1% increase in energy prices causes a 1.499% rise in food prices. Furthermore, the results reveal that this variable exerted a positive and significant impact on food prices in the short-run. The magnitude of 0.012 shows that a 1% increase in energy prices led to a rise of 0.012% in food prices. These findings indicate that although the effect of an increase in food prices can be negligible in the short-run, its long-term effects will be considerable. The findings of this study reveal that the exchange rate has a positive long-term effect on food prices. A 1% rise in the exchange rate caused a 2.874% increase in food prices in the long-run, although the short-term coefficients were negative and insignificant. This implies that a rise in the exchange rate has the greatest long-term impact on food prices.
The coefficients of the error correction term were negative and significant in the three estimators, which shows the existence of a long-term relationship between the variables in this study. The magnitude of −0.005 in the PMG estimator shows 0.5% of deviations from the long-term equilibrium of food prices, which are corrected every month. This result reveals that food prices take a long time to return to equilibrium following the independent variable shocks.

Robustness Analysis
Due to the fact that the food groups selected in this study do not have the same share in household expenditures, the categories may contain heterogeneous properties. In this situation, the estimation problems, including serial correlation, heteroscedasticity, and endogeneity are likely to occur [33][34][35].
Hence, in the current study, to examine the robustness of the long-run coefficients estimated by the PMG estimator, the FMOLS and DOLS were employed. The FMOLS is proposed by Pedroni [36] to estimate the long-term co-integrating vectors for heterogeneous co-integrated panels. This estimator utilizes a non-parametric method to overcome the endogeneity bias and serial correlation. The DOLS Energies 2020, 13, 4031 8 of 15 estimation approach is developed by Kao and Chiang [37]. This method also considers endogeneity problems of variables and is used to check efficiency when there is a mixed order of integration among research variables [38]. Table 5 reports the FMOLS and DOLS estimation results. As seen in Table 5, food prices are positively and significantly affected by energy prices and exchange rate. In the DOLS, the coefficient magnitudes of energy prices and exchange rate were 0.91 and 1.32, respectively. These results imply that a 1% increase in energy prices and exchange rate causes a 0.91% and 1.32% rise in food prices, respectively. In the FMOLS, the coefficient values of energy prices and exchange rate were 0.93 and 1.51, respectively, indicating a 1% increase in energy prices and exchange rate that leads to a rise of 0.93% and 1.51% in food prices, respectively. The results reported in this section are similar to the PMG's findings. Therefore, we can conclude that the PMG results are robust.

Impulse-Response Functions
In order to have a comprehensive understanding of the dynamic relationship between food prices, energy prices, and exchange rate, we estimated the impulse response function (IRF). The IRF provides explicit information about the dynamic response of food prices to exchange rate and energy price shocks. Figure 2 shows the impulse responses for the shock to food prices. The 95% confidence intervals of IRF were calculated by Monte Carlo estimated standard errors, with 500 repetitions. In Figure 2A, the effect one Standard Deviation (SD) shock in lnFP on lnFP is instantaneously positive, but its degree declines in the long-term. A positive one SD shock within energy prices leads to an increase in food prices both in the short-and long-term. Similarly, the results indicate that the effect of one SD shock in the exchange rate on food prices is positive.
For a deeper insight into the dynamic relationship between the price of each food product and energy and currency prices, we complemented our analysis by estimating IRF separately for each food product (see Figure A1 in Appendix A). Figure A1 shows that the response of each food product price to its own shock is positive and significant in mid-term (periods 3 to 5), but these impacts decrease gradually in the long-term. These results are supported by the IRF for panel VAR presented in Figure 2A. The response of each food product price to energy price shocks indicate that it starts from zero and then grows in the long-run in all food products, which is consistent with the results obtained from Figure 2A, Tables 4 and 5. However, the responses of all food product prices to exchange rate shocks are positive in the long-run, which is consistent with the results obtained from Figure 2C, Tables 4 and 5.

Granger Causality
The existence of co-integration between food prices, exchange rate, and energy prices confirms the presence of Granger causality at lease in one direction. Therefore, the panel VAR-Granger causality test was implemented to evaluate the causal linkage between variables of concern (see Table  6). As the results show, there exists two-way causality between food prices and energy prices, twoway causality between food prices and exchange rate, and two-way causality between energy prices and exchange rate. Such findings are consistent with Nazlioglu and Soytas [22] and Gözgör and Kablamacı [39].

Granger Causality
The existence of co-integration between food prices, exchange rate, and energy prices confirms the presence of Granger causality at lease in one direction. Therefore, the panel VAR-Granger causality test was implemented to evaluate the causal linkage between variables of concern (see Table 6). As the results show, there exists two-way causality between food prices and energy prices, two-way causality between food prices and exchange rate, and two-way causality between energy prices and exchange rate. Such findings are consistent with Nazlioglu and Soytas [22] and Gözgör and Kablamacı [39].

Discussion
In summary, the results of this study provide strong support for the role of domestic energy prices in determining food prices in both the short-and long-run periods in Iran. These results are consistent with the findings of other studies confirming that energy prices significantly contribute to the rise in global food prices [14,40,41]. These findings also are in line with the studies that show food product prices in individual countries are positively and significantly affected by the change in energy prices [22,23,42,43]. However, what distinguishes this study's results from prior studies that examined the role energy prices on food prices in other countries is that the intensity of the long-run impact of energy prices on food prices is much higher in Iran compared with other countries. Our estimates suggest that a 1% increase in the price of energy is associated with the increase in the expected food price in the range of 0.91% to 1.49%, while these effects are much weaker in other countries. For example, Ibrahim [10] argues that a 1% increase in energy prices causes a 0.06% increase in Malaysia's food prices. Factually, if the energy price rises, in a stable economy it is expected that the food prices increase as much as the share of energy in the cost of agricultural production, food processing and marketing, and consequently the price of food to consumers. However, in high-inflation countries such as Iran, rising energy prices could lead to increased inflation expectations of households and firms [44]; consequently, increase households' food demand and decrease firms' food supply (prefer to sell produce at a higher price in the future). These findings are confirmed by Figure 2A,B. As results show, a positive one SD shock within energy prices leads to an increase in food prices and it is not only eliminated but also strengthened in the long run.
The results of this study also confirm that energy prices have an indirect (in addition to direct) positive impact on food prices through an appreciation of the exchange rate. This result is consistent with the findings of Nazlioglu and Soytas [12]. It is worth mentioning that the indirect effect of fluctuations in energy prices on food prices through the exchange rate in Iran is different from the experience of other countries. In fact, the stability of the foreign exchange rates is highly influenced by political stability, and any fears about reducing oil revenues (supply of foreign currency) could motivate people to hoard foreign currencies [45]. Hence, rising energy prices with the aim of eliminating the budget deficit can be a negative signal in the foreign exchange market (appreciation of foreign currencies) which can indirectly cause changes (increase) in food prices. As these results are confirmed by Figure 2C. In fact, the impact of exchange rate shocks on food prices will intensify in the long run. To sum up, the findings reveal that food prices take a long time to return to equilibrium following independent variable shocks.
Looking at the impact of energy prices and exchange rates on each individual commodity price, the results in Figure A1 suggest that the effect one SD shock in energy price and exchange rate leads to an increase in each individual commodity price. Although the magnitude of the impact of these shocks on individual commodity prices are not same, which could be related to government financial and regulatory support for the production and supply of some food commodities [46]. For example, due to this support, the energy price and exchange rate shocks have insignificant effects on meat price.
The findings of this study indicate that the policy of increasing energy prices in Iran has serious adverse effects on inflation in food, and consequently the welfare of households and firms. Despite the importance of this topic, to our knowledge, few studies have focused on the potential effects of this policy on food prices in Iran. What emerges from the results of this study is that important factors such as political and social shocks as well as inflation expectations could play key roles in explaining the relationship energy policy and food prices in this country. Hence, future studies could fruitfully explore this issue further by quantifying these factors and incorporating in empirical models.

Conclusions
The current study examines the short-and long-term impacts of the exchange rate (the value of Iranian rial per USD) and energy prices on food prices in Iran, using monthly prices ranging from March 1995 to February 2018. The panel unit root test, Pedroni co-integration tests, PMG, MG, and DFE estimation techniques are applied for a panel of ten food products. We used the Hausman test to choose among these estimation techniques. According to this test, the PMG is selected, as this technique is most efficient. The estimated long-run coefficients of panel ARDL are also evaluated for robustness using the DOLS and FMOLS. Therefore, it can be concluded from the results of this study that the continuation of the current energy policy leading to rising energy prices will cause an increase in food prices and, therefore, it will have a negative impact on consumer welfare, especially in low-income deciles of Iranian population who spend about 44 percent of their income on food products (high-income deciles spend about 17 percent of their income on food). Hence, rising energy prices would increase the share from food in the spending basket of low-income households which is certainly expected to lead to more food costs insecurity. Thus, policy-decision makers should take into account the adverse effects of the changes in energy prices on domestic food prices when designing income support and economic stabilization policies. Furthermore, to overcome the negative effects of rising energy prices on consumer welfare, the government should implement complementary food policies such as price controls, export tariffs and other export restrictive policies, and import expansion policies. It has been demonstrated that the policy of increasing energy prices can indirectly affect food prices through the exchange rate. Given this fact, policy makers should take into account the indirect effect of oil prices on food price stabilization policies. From the manufacturers' point of view, given that a significant portion of the inputs required to produce food are supplied through imports, appreciation of foreign currencies leads to an increase in the cost of food production. In this situation, if the government does not support manufacturing firms, it can be expected to reduce investment, production, demand, and consequently be conductive to the emergence of stagflation. Furthermore, the results indicate that the elimination of energy subsidies can play a crucial role in increasing inflation expectations of households and firms, which in turn could exacerbate the negative effects of this policy on food prices. Hence, in this situation, communication strategies and management of economic expectations that directly targets the beliefs of households and firms could therefore serve not only to reduce the food inflation but also enhance the economic stability. The results of the current study can provide useful information for domestic and international manufacturers, traders, and investors in predicting prices with higher confidence. For example, when energy prices and exchange rate (the value of dollar) rise, manufacturers can expect to raise prices and production costs (especially those who are dependent on imported inputs.).    Figure A1. Impulse response function (IRF) of each food product to energy price and exchange rate shocks.