6 THE RESULTS

Results are reported first for annual information, for which the pairs of price series that show a long run equilibrium, and then for monthly data in a more detailed fashion. Given the scope of the analysis and the amount of information to be processed, much of the evidence has be considered as the basis for more detailed investigations on specific case studies, possibly employing more sophisticated analytical tools. It is also worthwhile mentioning that for several pairs of prices, a priority area for more in-depth analysis is the presence of structural breaks, which in this work are dealt with only for those series in which they appeared more evident, and more directly connected with changes in the economic and policy framework of the countries involved. Given the purpose of the paper, which is to provide evidence on price transmission also in support of policy analysis work, an attempt has been made to comment the results with reference to the policy framework. This has been possible especially in section 6.2, on the basis of the relatively more accurate indications yielded through monthly data.

6.1 The annual price series

The number of observations available for each series (about 32 on average) prevents us from obtaining fully conclusive evidence from the tests on the dynamic properties of the series, and therefore also from the co-integration tests. This is somehow confirmed by the fact that the ADF and the PP test - whose results are not reported in the tables for brevity - show non homogeneous results in more than one case. Where the dynamic properties of the series could not be clearly identified, the evidence on transmission had to be presented as being less clear cut.

Results are organized by country in alphabetical order. They are reported for both the relation of each price with the reference world market reference prices, and for vertical relations within each country. The latter are analyzed only in those cases in which they appeared likely to convey additional information compared to the former, i.e. where a significant difference emerged among domestic prices in the relation with the world reference price, while they where not considered when either all prices behave in a similar way in relation to world prices, or the different responses could be explained by differences in the order of integration of the series.

The key to the names of variables is reported in Table 1. Results of the Unit root tests are reported in Tables A1 to A15 in the Appendix. These were firstly applied to the series of world reference prices; from the ADF test these all resulted I(1), since all regressions in level accept the null of absence of unit roots, while the same test run on the first differences rejects it at least at 5 percent confidence level for the specific distribution.

Annual price data for Brazil covers only one of the major exportable cash crops in the country, which is soybeans, together with bovine meat, and a set of relatively less important products, such as wheat, maize and rice. In general, these prices show a high degree of transmission with world prices, especially for soybeans (Table 2), as can be expected given the importance of the country in that market. A long run equilibrium emerges between the border prices and the world reference price, as the Engle and Granger procedure shows the presence of co-integration, and the long run parameters are both significant. The same does not apply to the domestic producer price reported for the same product, despite a significant long run coefficient. However, this is not a firm indication of absence of transmission, given both the limited number of observations on which the ADF test was performed, and the fact that critical values are not far from the usual significance levels. Also the Brazilian border price for maize provides evidence of co-integration with the world reference, whereas for the export price this same evidence is not supported by a significant long run parameter (Table 2).

Available prices for Chile do not cover horticultural products, which are among the most important in local agriculture, while it includes some of the livestock products. In general, transmission of world price changes on local prices appears fairly complete, as can be expected given the liberal attitude of the economic policy framework, and the long standing attempt to liberalize foreign trade. Results show a significant long run equilibrium with the world reference prices for the rice and pig meat (Table 3). For rice, all the three prices available - the import unit value, the wholesale and the producer prices reported in FAOSTAT - are co-integrated with the world reference price and show significant coefficients. The same applies to the retail price for pork meat, and to the import and producer price for the same product. For wheat, the Chilean wholesale price shows a high degree of common movement with the world reference, whereas evidence is less firm for the retail price (Table 3). For poultry meat the domestic producer price shows a significant long run parameter, but co-integration is rejected. Although this price is strongly related to the import price, it appears not to be significantly related with the wholesale price: in other words, shocks in the world price appear to be transmitted at the producer level, but not at the import price level, although producers are affected by changes in the import price. The result for milk powder (Table 3) is most probably a spurious one, and/or the outcome of an incorrect model specification, given that the domestic milk powder price results I(0) from the ADF test. Also the results for dairy products, indicating a significant degree of integration for butter and cheese with the world reference export unit values, appear questionable for the same reason.

For Costa Rica few price series are available which do not include the main products of local agriculture, and especially the main cash crops such as bananas, coffee, pineapples, oranges, and other horticultural products. Results obtained from the annual data are not directly comparable with those of the monthly data, reported in section 6.2.1, apart from the case of pork meat. The tests on bovine meat border unit values accept co-integration with the world reference price, especially on the export side (Table 4), while for the import unit value the test rejects the null. The FAOSTAT domestic price for this same product, however, shows no evidence of a significant relation either with the world reference prices, or with the border prices. For pig meat as well a long run equilibrium emerges between the export price and the world reference, and between the domestic producer price and export price. The Engle and Granger test for these price pairs, however, indicates absence of co-integration.

In the case of Egypt, available prices are referred to some of the most important products in the local market, including cereals, cotton, and some of the major livestock products. Cotton and rice, however, had to be excluded from the analysis, since the database only reported administered prices, which were kept fixed through time or changed only few times over the sample period; this prevented a meaningful application of the analysis mentioned in section 5. Also for Egypt, most results from the annual data are not directly comparable with those described in section 6.2.2 for the monthly data; they appear consistent however, in the case of wheat and bovine meat. Results for the annual series indicate in general a fairly incomplete transmission between domestic and world reference prices (Table 5). The few products whose price shows a significant long run equilibrium are trade unit values, and particularly the export price of rice, the import price of wheat and maize, and the retail price of butter. In these four cases, the evidence of co-integration is consistent with the estimates of a significant long-run parameter; for wheat and butter, however, such evidence is less firm, as co-integration is rejected. Among the other products, there is only evidence of vertical transmission, particularly for wheat, between the export, producer, wholesale, and retail price series, and for maize, sorghum and bovine meat between the producer and the wholesale levels. Estimated coefficients are not far from unity in all cases, apart from those relative to spatial transmission between maize import prices and rice export prices with international reference prices, which are significantly higher then one.

Available monthly data for Ghana include prices of some important food crops, like maize and sorghum, which constitute the bulk of cereal consumption, but exclude important cash crops, and primarily cocoa. Evidence of significant price transmission arises for the import price of maize and for the import and the wholesale prices of imported rice. For this latter product, also the producer price reported by FAOSTAT for the locally produced varieties, which are priced differently from the imported ones, show a long run equilibrium with the world reference price, although co-integration is rejected. Relatively less firm evidence of transmission also arises for sorghum, maize, and cassava both at the producer and the wholesale level. As it will be clear from section 6.2.4, part of these results are not confirmed by the more detailed analysis of monthly data, particularly for the local price of maize.

Available price data for India show a considerable degree of linkage with world reference prices, despite the overall agricultural policy attitude of the country, which for a long period has been characterized by a relatively high degree of public regulation. At the same time, for this country in particular the average data employed are inevitably inaccurate, given its size and diversity. Estimates indicate evidence of long run equilibrium in the spatial transmission between the domestic and the world reference prices of wheat, maize, cassava, milk powder, and to some extent rice, while coefficients are mostly not significant for meats (Table 7). For wheat, the producer price shows the highest degree of transmission with the world reference price, while results are less clear cut for the wholesale, the retail and the import prices, and for the FAOSTAT producer price: in these cases co-integration is rejected. For rice, the import price shows a common movement with the world reference price, but the relation with the domestic wholesale and producer price is less clear. Nonetheless, within the domestic market transmission appears to be fairly complete between the wholesale and the retail prices, and between the producer price and the export price. For maize, the relation between the world price and the wholesale price shows a long run equilibrium, while, for other prices, the co-integration is rejected.

In the case of Indonesia, prices showing evidence of a long run equilibrium with the world reference are mainly the import unit value of maize and rice, the producer price of cassava, the FAOSTAT producer price of soybeans, and the import unit value of bovine meat (Table 8). For maize and cassava, evidence is quite clear concerning both the import and the producer price reported in FAOSTAT. For rice, however, evidence appears clear for the import unit value, while less so for the FAOSTAT price, for which co-integration is rejected. Of the two meat products reported, bovine meat shows a long run equilibrium with the world reference price, while the same does not apply to the domestic price reported by FAOSTAT. Most of these results are consistent with those of the analysis based on monthly data, reported in section 6.2.5, particularly for rice.

Annual price data available for Mexico cover some of the most important food crops, such as maize, wheat, sorghum, and some livestock products, which show in general a considerable degree of common movement with world reference prices. Important cash crops such as coffee and horticultural products, however, are not included in the dataset. The tests on the annual price data report significant long run relations primarily between local prices of wheat and bovine meat and the corresponding world reference prices (Table 9), both for the domestic and the border prices. For wheat, the export price shows a clearer relationship than the import price, while the opposite is true for bovine meat. Import prices of soybeans and pork meat also show significant long run equilibrium with the world price, while in the case of poultry, the significance is questionable, as the series result I(0) from the unit root test.

Annual prices for Pakistan include major food staples grown in country, such as wheat, maize sorghum and rice, but it excludes important crops, such as cotton. In general, local price appear to be connected to world reference prices to a significant extent. Results of the test indicate the presence of a long run equilibrium between the domestic and the world reference prices primarily for wheat, rice, maize and bovine meat. A long run relation emerges for the export price of the Basmati rice, and for the domestic wholesale price of the Irri rice. For wheat, maize and bovine meat, a long run equilibrium with world prices emerges for the wholesale and the retail prices, although the estimate for the latter is not confirmed by the result of the co-integration tests. A significant long run coefficient also arises for pig meat and for poultry, although the tests reject co-integration with world reference prices.

Estimates for Senegal (Table 11) are referred to products which are not very important in the country’s agricultural sector. For some more important locally produced goods - like groundnuts, rice, maize - a considerable amount of monthly information is available, whose results are described in section 6.2.7. Concerning annual price series, estimates show only one pair of prices for which a significant long run equilibrium emerges in the presence of firm evidence of co-integration; this is the export price of bovine meat, which is definitely a minor item in the country. For the other products, estimates of significant coefficients are obtained in absence of a clear outcome of the co-integration tests. This is the case for the producer price of maize, and for the import and the producer prices of milk powder, which appear somehow tied to the corresponding world reference prices. Similar evidence also arises for poultry and pork meat. In the case of the import price of rice, however, the result is most probably a spurious one, as the corresponding price series results I(0) in level.

For Thailand, available annual data include some the most important products of local agriculture, primarily rice. Most price series show a high degree of transmission with the corresponding world market reference, a result which is consistent with the policy attitude of the country, traditionally pointing towards openness to foreign trade. As expected, for rice and cassava transmission in the long run is almost complete with the corresponding world reference prices (Table 12), as the prices of these two products - respectively the FAOSTAT export unit value and the producer prices - are in fact almost coincident with the world reference reported by the IMF. The degree of integration is high both at the border and at the domestic producer and retail levels. Similar results also emerge at the wholesale level, although in this case the test rejects co-integration with the world reference price. Maize prices also appear to be related to the world reference price at the wholesale and at the producer level, and the same applies to cassava at the producer level, to the retail price of poultry, and to the producer price of pork meat. For bovine meat, the significance of the long run equilibrium is less clear-cut, due to rejection of the test for co-integration.

Transmission parameters for Turkey (Table 13) indicate rice, wheat, maize and poultry as products whose price is moving similarly to the world market reference. For these products, a long run equilibrium emerges at the wholesale, retail and producer levels, and also with the import price in the case of rice. For wheat, evidence is clear at the wholesale level, while for the export price the difference in the order of integration of the series prevents the possibility of estimating a meaningful coefficients. For bovine meat, only the export unit value appears to be significantly related to the world reference price, while the domestic market appears to receive fewer signals from world markets, despite vertical integration arises both between the producer and the wholesale prices, and between the latter and the retail price. For sunflower seeds, transmission between the producer prices and the world reference appears to be fairly complete. In the case of butter, both the wholesale and the retail prices show a common movement with the world reference unit value in the long run, while no transmission appears to take place at the producer level. For crops, these results are consistent with those generated by the analysis of available monthly price data, reported in section 6.2.7.

Despite being quite detailed and comprehensive, the annual price dataset for Uganda does exclude some important agricultural exportables of the country, such as coffee, tea, and cash crops such as cotton. The degree of price transmission observed appears particularly low: parameters are mostly not significant, especially in the relations between the domestic and the world reference prices. Exceptions are wheat, whose producer price reported by the FAOSTAT shows a long run equilibrium with the world price, the producer price of sorghum, and the import unit value of milk powder. Given the Unit roots tests showing that a number of the series involved are I(0), moreover, some of the significant spatial transmission coefficients are most probably spurious: this is the case for the import price of maize, and for the producer price of soybeans. Vertical transmission between domestic prices, instead, appears quite effective: this is the case of sorghum, maize, rice, and cassava, while results for wheat, soybean, poultry and pork meat appear questionable, due also in this case to the series being I(0). Visual inspection, however, indicates quite clearly the existence of common trends between the domestic price series at different levels. Reasons that can explain these results range from the relatively low degree of tradability of most of the products involved, which are probably traded mostly in local markets, to the relatively high degree of self sufficiency of the country, which also may contribute to explain the poor linkages with the world reference prices.

Finally, annual data for Uruguay, include bovine meat among the main export products, and a number of other important items in local production, primarily cereals. The analysis yields significant transmission parameters for the export price of bovine meat, the producer prices of sorghum and sunflower seeds, and for rice, despite co-integration is rejected. These results appear generally consistent with the long standing liberal attitude of the country in the area international trade policy. Results for wheat and soybeans, pointing to a significant long run equilibrium, instead, are more questionable: they may either to be considered spurious, or deriving from incorrect model specifications, given that some of the corresponding series are I(0) in the levels.

6.2 The monthly price series

Also in this case results are reported by country in alphabetical order. Monthly data do not overlap frequently with the annual information, since there are no border or domestic prices available from FAOSTAT on this time scale, and because in for many products monthly data correspond to too few annual observations to run meaningful analysis. Unit root tests for world reference prices and domestic price series are reported in the Appendix, in Tables A16 to A23. The former are mostly I(1) according to the PP test; inconsistency with the results of the ADF tests arises for wheat and palm kernel oil, which appear to be I(0) from this latter test, and for pig meat and poultry, which appear to be I(2). In order to save space, the Tables attached to the following sections only report coefficients that resulted statistically significant at least at 5 percent level.

6.2.1 Costa Rica

Few monthly price series are available for Costa Rica: palm oil, bovine, pork and poultry meats. As already pointed out for annual data, the prices of major cash crops produced in the country - such as bananas, coffee, pineapples, oranges - do not appear in the dataset, which contains information for few food staples. The Unit root tests indicate that most series are I(1), apart from the poultry price which is stationary, and from the bovine meat price, for which the results are inconsistent between the ADF and the PP tests (see Table A.17 in the Appendix).

Following the Engle and Granger (1987 procedure, the Unit roots test was applied to the residuals of the static regression between each series and the corresponding world reference price, and for pork meat only, to residuals of the relation between the wholesale and the retail prices (Table 16). Co-integration emerged for pork meat, both within the domestic market and with the world reference price, and for palm oil. The long run parameter, however, is not significant for the latter product; therefore, despite the existence of some common movement in the two price series, they appear not to be driven by a long run equilibrium. This is not the case for pork meat, whose transmission parameters are significant in both the short and the long run. As can be expected by looking at the series (Figure 1), transmission is far more complete and quicker within the domestic market compared to that with the world market reference price.

Given the minor role played in commercial agriculture by the products whose prices were considered here, part of the explanation for the poor transmission observed may arise from the limited amount of foreign trade; maize, beef and poultry are mostly exchanged in local markets. In any case, reasons for these results appear not to be found in the trade policy setting, as at least in recent years pork meat shows one of the highest applied tariffs, far higher than the one for bovine meat and palm oil.

The causality tests indicate that the world reference pork meat price is Granger-caused by the Costa Rican wholesale price (Table 17), whereas it is inconclusive between the wholesale and the retail pork meat prices. This result can only be explained by the existence of some expectation mechanism by which the local wholesale price is sensitive to expectations about the future evolution of the world reference price.

Finally, concerning asymmetry - which was analyzed only for the relatively firmer relation found between the pig meat wholesale and world reference price - the dummy variable for the positive residuals of the static regression resulted significant; but the ECM coefficient is smaller in size compared to that of the model without such variable. This indicates that positive price shocks are transmitted in the domestic market at a slower pace compared to negative ones. In turn, this can be explained by the presence of some policy provision aimed at smoothing price effects in the local market, stabilizing the domestic market.

6.2.2 Egypt

Results for monthly information are generally consistent with those yielded with annual data. Transmission between the local and the world reference prices appears generally poor, apart from a few products including wheat, but only over the period following the economic reforms. On the contrary, transmission results generally high within the domestic markets. As it was the case for annual data, also for monthly data rice and cotton prices had to be excluded from the analysis, as only the administrative price where reported, showing insufficient variability in the sample period.

Unit root tests indicate that all series are I(1) (see Table A.18 in the Appendix). The same test applied to the residuals of the static regression (Table 19) accepts co-integration in five out of the nine pairs of prices. For wheat there appears to be no evidence of a long run relationship between the wholesale price and the world reference price. Both a visual inspection of the series (Figure 2), and the test for the stability of parameters based on the Cumulative Sum of Recursive Residuals (Figure 3), however, show a structural break in August 1989. In the data, this period corresponds to a change in the exchange rate reported by the IMF International Financial Statistics, employed to convert the local currency prices. As confirmed by FAO (1999), in the late 1980s, the Egyptian Government moved progressively toward restructuring and reorganizing the economy, adopting, among other provisions, the liberalization and unification of the exchange rate regime that was previously set at different levels for different transactions^[7], together with a major reduction in trade barriers, and the liberalization of marketing channels for several commodities, that were previously operated solely under State control.

Accordingly, the sample was divided into two parts, and two separate models were estimated.^[8] The results for the 1969-1989, which are not reported in the Tables, again failed to show evidence of co-integration, while this is not the case over the period 1989-2001 (Table 19), a result confirmed by the significant long run parameter of model (1). This appears consistent with the qualitative notion that transmission has improved after the economic reform, directly affecting the exchange rate.

It is worth observing that the wheat market in Egypt is still a relatively administered one: the Government operates a floor price both at the producer and at the consumer level (for bread). The transmission result found here from the late 1980s on indicates that prices are transmitted and markets are integrated at least to some extent despite the presence of a minimum guaranteed level. In turn, this is an indication that the floor price is probably only affecting the level of risk for farmers, by truncating the probability distribution of price outcomes, rather than directly affecting price formation.

Contrary to the previous case, co-integration emerges between the domestic wholesale and retail wheat prices (Table 19); this is a fairly predictable result, given the plots of the series (Figure 2). For maize and sorghum co-integration with the world reference price is rejected, as already indicated by annual price series (Table 19). However, a long run relationship between the wholesale and retail prices emerges (Figures 4 and 5), and appears to be confirmed by the significance of the parameters of the ECM model. Both these are minor products in the Egyptian market, characterised by low public involvement compared to the wheat market; therefore one would expect a more direct relationship of domestic prices with the world reference prices.

The same applies to bovine meat, whose co-integration test is rejected (Table 19). For this product too the mentioned change in the exchange rate of the late 1980s could have had an effect. Therefore model (1) was also estimated on the two separate sample period employed for wheat. Co-integration, however, emerges for the period prior to the change in the exchange rate regime (Table 19) - something which is difficult to link qualitatively to the change in the policy environment - and moreover, the long run coefficient shows a negative sign, which is meaningless.

The world reference price appears to Granger-cause the domestic wholesale price for wheat - after 1989 - and for bovine meat (Table 20). Within the domestic market, instead, the tests are inconclusive between the wholesale and the retail prices, both for wheat and bovine meat. For sorghum, the retail price appear to be Granger caused by the wholesale price.

The test for asymmetry indicates that a higher speed and a higher degree of transmission arises for all the three pairs of series showing a long run equilibrium with the world reference price: the wheat wholesale price in the 1989-2001 period, and sorghum retail and wholesale prices for the whole sample (Table 21). In all three cases, the dummy variables for the positive residuals in the static regression are significant, and the ECM parameter are larger in size than the corresponding parameters in the models without the dummies (reported in Table 19).

This indicates that the domestic markets smoothed the price reductions taking place in the world reference price, while they fully passed through the increases in world prices. To some extent, this may also be due to the operation of the floor price at the producer level.

6.2.3 Ethiopia

Available prices for this country include some of the main staples, particularly maize and sorghum, together with a few other food crops, but exclude the main cash crops of the country, such as coffee. Monthly data resulted mostly I(1) from the Unit root tests (see Table A19 in the Appendix) with the two tests yielding almost fully consistent results.

The Engle and Granger (1987) procedure indicated that co-integration is verified in four pairs of prices out of seven. Notably, a long equilibrium emerged between the wheat retail price and the corresponding world reference price, for the retail price of sorghum, and for both the retail and the producer prices of maize. No meaningful relation arose, instead, for sunflower seeds, bovine meat, and the rice retail price, rejecting both co-integration and tests for a long run equilibrium. For maize, the producer price appears to react to changes in the world reference price with a four months delay, whereas two more months are required for the wholesale price. For wheat and sorghum, instead, transmission is faster, taking place within two and three months respectively.

The currently applied tariffs appear small and homogeneous among cereals, rice, and sunflower seeds, and therefore they do not seem to provide any explanation for the results. Rather, the size of the market may be more explicative, since wheat, sorghum and maize are respectively the three most important commodities, at least in terms of current import volume. At the same time, a relatively higher tariff applied in recent years may contribute to explain the lack of transmission observed for bovine meat.

Results of the Granger non-causality tests are inconclusive for wheat and for the producer price of maize, which appear not to be caused by the corresponding world reference price, whereas the world reference prices of maize and sorghum are Granger-causing the local retail prices for these products (Table 23). For maize, the discrepancy between the results of the two prices appears somewhat puzzling, given also the very close behaviour of the two series that can be observed from the plots (Figure 8).

Based on these results, the test for asymmetry was only performed for the maize and sorghum retail prices (Table 24). Both products show a higher speed in the response of prices to positive price shocks, given that the coefficient of the model with the dummy variable is higher compared to the corresponding ones without it, reported in Table 22. Whatever the reason lying behind this evidence, it should be noted that it may be harmful in terms of food security, insofar as the two crops are relevant food staples.

6.2.4 Ghana

Also monthly information for this country include prices of basic food crops like maize and sorghum, while exclude important cash crops, primarily cocoa. Results of the Unit root tests indicate that most of them are I(1) (see Table A20 in the Appendix), with the exception of the retail price of cassava, that results I(0) from the ADF test, and I(1) in the PP test.

Co-integration and the related estimation of ARDL models shows the presence of a reliable long run equilibrium with the corresponding world reference prices only for the wholesale prices of groundnut (Table 25). For maize and sorghum, despite the long run parameter is significant, co-integration is rejected, and for maize also the size of the coefficient appears too large to be credible. For imported rice, cassava, palm oil and palm kernel oil, however, co-integration is rejected, and the long run parameter is no significant.

More evidence of transmission arose within the domestic market, where long run parameters are significant between the wholesale and the retail prices of maize, sorghum, palm oil, imported rice, and cassava, whereas this is not the case for groundnut. For maize, these results are consistent with the conclusions of other studies (Badiane and Shively, 1998; Abdulai, 2000), indicating that the domestic market is relatively well integrated.

In principle, the lack of transmission with the world price may be related to the policy framework that the government was operating until the 1980s, which included price support through intervention. After that period, several changes in the institutional framework have been taking place in country, particularly with the implementation of the Economic Recovery Programme 1983, and with the related change in the exchange rate regime that took place over the 1983 to 1985 period (Badiane and Shively, 1998). The inclusion of a dummy variable accounting for a break in that period, and also the break of the sample into two periods did not yield improvements in the relation with the world reference prices, neither before nor after the mid 1980s.

Results of the test for Granger causality also are puzzling for maize (Table 26), as lagged observations of the Ghanaian wholesale price explain the behaviour of the world reference price. Once more, this could be explained by assuming that price formation in the country effectively follows some world price forecasts. For sorghum and groundnuts, instead, the reverse is true, and the domestic price is Granger-caused by the world reference price. Within the domestic market, the maize wholesale price is caused by the retail price, and the same applies to imported rice. Results for sorghum, palm oil and cassava, instead, are inconclusive.

The picture obtained here partially confirms the one derived from the annual data. Particularly, maize and rice import unit values from FAOSTAT showed in that case a significant relation with the world reference price, that appeared not to apply to the domestic wholesale price of both products, and also to the domestic price of cassava. For sorghum, however, the long-run relation with the world reference price emerged in both cases.

Given the results of the previous tests, the ECM specification with the dummy variable for positive residuals was implemented only for three pairs of prices: the domestic and world reference prices of groundnuts, and the relation between the retail and the wholesale prices of maize and imported rice in the domestic market (Table 27). For these last two products, given the evidence that the retail price is causing the wholesale, the model is considered with the wholesale price as a dependent variable. For the groundnut price, the ECM coefficient of the model including the dummy appears slightly larger in size compared to the one without it; both coefficients, however, are very small, indicating in both cases a slow transmission of (positive and negative) shocks. Within the domestic market, instead, this difference results more substantive for rice, while is quite small in the case of maize.^[9] In other words, the degree of asymmetry is fairly strong for the former, and weaker for the latter product.

6.2.5 Indonesia

Monthly price series for this country are mostly local wholesale prices, whose behaviour is analyzed in relation to the world reference prices. Some of the results obtained from the annual data are partially confirmed, particularly those concerning the wholesale prices of rice, while this does not apply to maize and soybeans: annual import and export unit values reported by FAOSTAT for these products show a common movement with the world reference price which is not confirmed by the monthly data for the corresponding wholesale prices.

Most local price series result I(1) from both Unit root tests (see Table A21 in the Appendix), with the exceptions of local rice and sorghum. Evidence on co-integration with the world reference prices arises only for coffee (Table 28). This is a major crop in the country, whose importance as a foreign currency source is still considerable among agricultural commodities, despite reduction of its overall importance since the growth of oil exports in the 1970s.

The long run parameter for coffee appears close to unity, and the results of the test for co-integration - both the ADF and the PP which is not reported in the table - are consistent with it. Moreover, the indications of a common movement in prices is also confirmed by the plots of the series, showing a lagged relation between the two prices (Figure 18). Such qualitative evidence, however, appears also for local rice (Figure 19), although the dynamic properties of this second series do not allow running meaningful correlation analysis. Drawing on this qualitative evidence, the Granger test is run also for local rice: this shows that for both local rice and coffee the world reference price is Granger-causing the local wholesale price, as one would expect (Table 29).

The test for asymmetry, that could be run only for coffee, shows that positive price shocks are transmitted to the domestic market to a different extent compared to negative ones, since the dummy variable for positive residuals is significant, and that positive shocks are transmitted more rapidly than negative ones, since the size of the ECM coefficients is almost double compared to the one of the model without the dummy (Table 30).

It is interesting to note that the results indicate the presence of transmission in the rice sector, although from the causality test only. Being by far the major food staple in the country, the rice market has been extensively regulated by the BULOG, a State trading enterprise which has been holding monopoly power for a long period, and is still in charge of managing important domestic policies, such as floor prices, and foreign trade. The results obtained are indicating that in the long run the BULOG has been following, at least to some extent, the world market price trends.

6.2.6 Senegal

The monthly price dataset for this country covers some of the major crops, particularly maize, rice, sorghum and groundnut. The series for this last product, however, are stationary, therefore they could not be meaningfully studied in relation with the world price. In general, transmission with the reference world prices appears limited, an observation that arose also from the analysis of the annual price series.

Over the 1990s, the country underwent substantial changes in its commodity policy framework. Particularly, the privatization programme reduced the role of parastatal enterprises in the trade of some major cash crops, such as groundnuts and cotton, and the overall level of public intervention in agricultural markets. From the available data, however, these changes do not appear to have caused systematic changes. The 1994 devaluation of the CFA Franc, instead, is shown by the data as a sudden fall in the local price expressed in US dollar; nonetheless the effect of the devaluation seems to vanish within a short period, and prices appear to go back to about their previous trends (Figures 20 and 21).

The unit root tests indicate that maize and sorghum prices are I(1), that groundnuts prices, both those of the oilcrop and for direct consumption, are stationary, while they diverge between the PP and ADF for cassava, palm oil and rice (see Table A22 in the Appendix). For this latter product, the ADF test only indicates that the price series are I(1).

Co-integration tests yield significant results only for the retail price of rice in relation to the world reference price (Table 31); this is consistent with the importance that this commodity gained in recent years among food imports. For maize, despite significant long run and ECM parameters, and although the plots appears to confirm the existence of a relation (Figure 21), the Engle and Granger procedure rejects co-integration, indicating the risk of spurious t tests results. For rice, the response to a shock in the world market price appears to take place over a far longer time period compared to maize. Lack of transmission arises from both the co-integration tests and the estimation of the ARDL models for the other products for which monthly prices are available, including sorghum, palm oil, and cassava.

Concerning causality (Table 32), for both maize and rice the world reference price is Granger-causing the local price, as can be expected given the relatively small size of the Senegalese markets for these products. The ECM with the dummy variables for positive price shocks (Table 33) also yields significant results. For rice, the size of the ECM coefficient is almost four times bigger than in the model without this variable, and more than three times bigger for maize. Dealing mostly with food crops, this evidence - telling that positive price changes affect local price far faster than negative ones - indicates a potential danger in terms of food insecurity for the net buyers of these goods.

6.2.7 Turkey

Results for this country appear consistent with those found in the analysis of annual data: most crop prices appear to be connected with the corresponding world reference prices, despite the relatively long periods of high inflation included in the sample - particularly the beginning and the end of the 1980s and the early 1990s - the repeated devaluations of the lira, and the crop specific support which is operated in the country for some of the major crops. As it is the case for other countries, the dataset does not cover some important products of Turkish agriculture, such as hazelnuts, cotton, olives, sugar, tea, and tobacco.

The Unit root tests indicate that most monthly series are I(1). Exceptions are the wholesale price of maize, soybeans and bovine meat, and the producer price of soybeans, which are stationary (see Table A23 in the Appendix). Co-integration with the world reference prices arises for the wholesale and the producer prices of rice, and for the producer price of soybeans (Table 34). For some other products, although co-integration is rejected, a long run equilibrium with the world reference price emerges, in accordance with the results of the analysis based on annual data; this is the case of wheat - both for the producer and the wholesale prices - and of the producer price of maize and sunflower seeds. Within the domestic market, co-integration arises for wheat and rice (Table 34); for this last product, the long run parameter results significant both between the producer and the wholesale prices, and between the retail and the wholesale prices.

Granger tests for causality indicate that the world reference prices of wheat and sunflower seeds are sources of information for the corresponding domestic wholesale prices (Table 35), while the result is inconclusive both for the wholesale and the retail prices of rice, for the wholesale prices of maize and wheat, and for the producer price of soybeans. A high degree of transmission emerges within the domestic market, instead, for rice, whose wholesale price is found to Granger-cause the corresponding retail price, and also for wheat and soybeans, whose producer prices appears to cause the corresponding wholesale ones.

Finally, the ECM models with the dummy for asymmetric transmission show that the speed is on average higher for positive price shocks than for negative ones (Table 36). The model was estimated in this setting also for the transmission between the prices of wheat and sunflower seeds and the respective world references, as the results of the tests for co-integration were close to the usual level of significance. The same evidence is confirmed within the domestic market, between the producer and the wholesale prices of wheat, and between the retail and wholesale prices of rice.