Threshold effects in price transmission: the case of Brazilian wheat, maize, and soya prices.

The focus of this article is with price transmission onto Brazilian markets for wheat, soybeans, and maize. In the case of wheat, this corresponds to import price transmission; Brazil is among the world's top five importers of wheat. Virtually all this wheat comes from Brazil's neighbor Argentina, making this the most relevant market from which to measure price transmission. In overall terms, the United States is the world's largest exporter of wheat, providing an alternative estimate of a suitable "world" price. Brazil is the second largest exporter of soybeans, a crop that forms Brazil's most important export commodity. Our concern here is principally with the extent to which world market prices are transmitted to Brazilian exporters. Brazil's chief rival in the world market is the United States, so we use U.S. prices as an estimate of the relevant world price. Finally, in the case of maize, Brazil has changed from being a net importer to a net exporter (from the beginning of 2001). However, the volume of maize traded across Brazil's borders has typically been small, never exporting more than 750,000 metric tones (mt) and rarely importing more than 200,000 mt per quarter over the period of study. Since the United States is the world's largest producer and exporter of maize and Argentina is its largest regional trading partner in this commodity, we consider price transmission from both of these important markets.

The data used in this study are monthly Brazilian, U.S., and Argentine prices for wheat, maize, and soya. The monthly frequency has been chosen because it is the highest frequency that is available over the window of interest. Furthermore, as noted by Barrett and Li (2002), monthly data limit the potential problem of aggregation biases associated with the use of lower frequency data. In addition, the expectation that traders are likely to react to price signals, at least partially, within thirty days favors the use of monthly data. However, we note that shipping times, particularly important between the United States and the two South American countries, make higher frequency data less desirable for this purpose.

The domestic wheat, maize, and soya prices for Brazil were as received by producers, (Getulio Vargas Foundation), converted from R$ to U.S. $/mt using free exchange rates (Central Bank of Brazil). The Argentina wheat price is for fob Trigo Pan, U.S. $/mt (International Grains Council). The wheat price data are from May 1988 to May 2001. Further wheat price data exist prior to May 1988. However, there was evidently a large change in the behavior of the Brazilian data prior to this date. Attempts to account for this structural break using dummy variables did little to ameliorate the problems (such as lack of cointegration between the series). However, as the results presented below demonstrate, the wheat price data are broadly consistent with having unit roots and being cointegrated with Argentine prices from May 1988 onwards. The soya price data also run from May 1988 to April 2001. The U.S. price is for No. 1 yellow soya (Chicago Board of Trade). The maize price series run from August 1986 to May 2001. The Argentine maize price is fob Rosario, U.S. $/mt (International Grains Council). Prices were converted to natural logs prior to estimation and testing. This step is consistent with the majority of studies. In addition, trending series also tend to have differences with growing variance through time, whereas when logged they do not (as in the case of the series explored herein).

As such, all prices used herein are expressed in U.S. dollars. They have either been converted here, or were already converted, into U.S. dollar prices. Without using a common currency, the analysis would not be possible since we would not expect there to be any cointegration in prices denominated in different currencies. However, exchange rates themselves may be volatile, and if prices do not quickly adjust to reflect exchange rate changes, then this may have a significant impact on the findings.

Figure 1 plots the rates used to convert both South American price series into U.S. dollar values. These are presented for ease as the natural log of an index based at unity in August 1986. As can be seen, both Argentina and Brazil experienced a period of rapid exchange rate change from the beginning of the period under study until late 1990 and late 1994, respectively. It seems likely that, in both cases, these changes were linked to significant inflationary episodes. From that date, Argentina has experienced a prolonged period of exchange rate stability following a policy of dollar parity for the remainder of the study period. In the case of Brazil, the period following 1994 is characterized by fluctuating but relatively stable exchange rates. We do not have any clear idea about how movements in exchange rates would affect the size or rate of adjustment back into threshold bands relative to other shocks. However, we acknowledge that exchange rates may play some part since it is these rates that international traders use to compare the prices of traded commodities.

[FIGURE 1 OMITTED]

These data series begin some four (for maize) and six years (for wheat and soya) after the potentially disruptive Falklands War of March to June 1982 and should not be influenced by the disruption to commercial shipping at that time nor the general economic sanctions, evoked under the GATT exception Article XX1.b iii, applied by the United Kingdom, the EC (EU), the United States, Canada, and Australia for the duration.

Time plots indicated that the price series (not presented herein) exhibited behavior that would suggest unit root or near unit roots in the polynomials of autoregressive representations of these series (in the sense they were clearly highly positively serially correlated). However, in a time plot, the series tended to return to what (visually) might be considered a "stable mean," and in this sense they did not unambiguously contain trends. Formal tests of stationarity are undertaken in the next section.

Unit Root Tests

The unit root tests and tests for stationarity are presented in table 1. Augmented Dickey Fuller (ADF) and Phillips Perron (PP) tests for unit roots were performed along with the Kwiatkowski-Phillips-Schmidt-Shin (KPSS) and Leybourne-McCabe (LMc) tests, (6) which have the null of stationarity. Seasonal dummies were placed in the unit root regressions, though this made little difference to the results. Twelve lags were used in the initial regressions, and the Schwarz-Bayes and Akaike information criteria were used to select the number of lags (both criteria selected one lag in all cases).

While they are not presented, the differenced results rejected unit roots (with only an intercept in the regression) at very high levels of significance. Similarly, stationarity could not be rejected for all the differenced series. The results in levels were less categorical with differing results for each of the tests. Consequently, they are presented in table 1. Finally, although they are not presented, tests that allow for structural breaks (Zivot and Andrews 1992), did not lead to the rejection of unit roots.

As can be observed in table 1, all price series fail to reject unit roots in levels (both around mean and trend) at a 5% level of significance according to the PP tests. Similarly, all price series reject stationarity according to the LMc tests (again, both around mean and trend). However, while the ADF test fails to reject unit roots around a trend for all series, it rejects a unit root around a mean for Argentine maize. The KPSS test for stationarity, fails to reject trend stationarity for all the series, but rejects stationarity around the mean for five of the eight series at the 10% level of significance. Therefore, the evidence here is rather mixed regarding the existence of unit roots. The substantial difference between the KPSS and the other tests is noteworthy. Nevertheless, there is sufficient evidence to proceed under the assumption that the series contain unit roots even though the tests did not universally support this conclusion.

Cointegration and Causality

The results for the pair-wise comparisons of price series regarding cointegration and causality are presented in tables 2 and 3. Both these tests were performed within a nonthreshold model. Each of the equations contained two lags in the VAR (one differenced lag in the VECM) although the coefficients are not reported herein. The Schwarz-Bayes information criteria eliminated all but one lag within each of the VARs. However, the Akaike criteria suggested longer lag lengths. With only one lag in the VARs, diagnostic tests (not presented) also suggested that the error exhibited serial correlation. In view of this, two lags were included in the VAR. A time trend was specified within the cointegrating vector so as to be reasonably general, but not within the error correction equation since the data appear to contain no deterministic trends when differenced.

The maximum likelihood tests for the cointegrating rank are presented in table 2. These tests are consistent with the wheat, maize, and soya price pairs being cointegrated. The one exception is the Brazil. U.S. maize pair, which rejects no cointegration at just under the 10% level of significance. Consequently, it is reasonable to proceed to estimate threshold effects, with the obvious exception that the Brazil. U.S. maize pair may not be cointegrated. However, before doing so, Granger causality tests between the series are presented in table 3.