Estimating policy effects on spatial market efficiency: an extension to the parity bounds model.

EPBM estimation results for maize and wheat are provided in tables 1 and 2, respectively. Each table contains estimates of trading regime probabilities before the policy change ([lambda]'s), estimates of the change in trade regime probabilities due to the policy change ([delta]'s), estimates of the parameter of the transfer cost function ([alpha]), estimated standard deviations of profit for different trade regimes ([sigma]'s), the estimated lengths of the transition period for each market pair (l), and the chi-square statistics for the LR test of the joint hypothesis of no change in regime probabilities ([chi square]). Numbers in parentheses below the parameter estimates are estimated standard errors, or in the case of [chi square], the p-value of the test statistic. Results are further summarized in table 3 which shows estimated mean price differentials, transfer costs (freight rates plus [??]), and arbitrage profits for each market pair and commodity over the period before the policy change, the period after the policy change, and for the full sample. Finally, figure 2 provides example graphs of spatial price differentials and parity bounds (90% confidence intervals around transfer cost estimates) for two trade routes, Addis Ababa to Dire Dawa for maize and Addis to Mekele for wheat, as an illustration of key results. We discuss the results for maize and wheat separately.

[FIGURE 2 OMITTED]

Results for Maize

Results for maize flowing from the surplus production regions of Jima and Nekemte to Addis are reported in the first two columns of table 1. Before the policy change, both of these trade routes were in regime 2 over 50% of the time, indicating that before the policy change there is a high probability that any trade flow was occurring at a loss. This is consistent with the negative mean arbitrage profit figures for these maize routes for the pre-policy change period (see table 3). However, these two routes are different in that the remaining probability is allocated mainly to regime 3 for Jima-Addis but for Nekemte-Addis it is allocated to regime 1 (see table 1). This suggests that, prior to the policy change, Jima-Addis is frequently in regime 3 where there are unexploited arbitrage opportunities, but Nekemte-Addis is frequently in spatial equilibrium (regime 1). Nekemte is a bigger market than Jima with more maize production in surrounding areas and typically more throughput, which may explain the higher estimated level of spatial efficiency. Nekemte is also slightly closer to Addis than Jima, and is often the first option for sourcing maize into Addis, particularly for the EGTE.

After the commercial reorientation of the EGTE in October of 1999 there was no statistically significant change in regime probabilities for the Jima-Addis route but a marked and highly statistically significant shift in regime probabilities for the Nekemte-Addis route that occurred over an estimated five-month adjustment period (table 1). These results suggest that the policy change had little effect on the Jima-Addis route but Nekemte-Addis moved fairly rapidly to a situation where price differentials rose and were exceeding estimated transfer cost essentially 100% of the time. (8) These results suggest that, as part of the commercial reorientation of the EGTE in October of 1999, the Nekempte-Addis route may have became less spatially efficient (change in EGTE operations led to an increase in price differentials above transfer cost).

Columns 3 and 4 of table 1 provide results for trade between Addis and the northern maize deficit markets of Dese and Mekele, the latter being the most remote maize market investigated in this study located in the far north of Ethiopia (see figure 1). Before the policy change, both of these trade routes are in regime 2 with substantial probability (87 % for Dese and 50% for Mekele), indicating a high probability that any trade flow was occurring at a loss. (9) However, Mekele also has a 50% probability of being in regime 3 and 0% probability of being in regime 1, while Dese has a relatively small probability of being in either regime 3 or regime 1. These results suggest that, prior to the policy change, Addis-Dese trade rarely experienced unexploited arbitrage possibilities (regime 3) while Addis-Mekele trade experienced this regime about 50% of the time. A possible explanation for the apparent unexploited arbitrage opportunities on the Addis-Mekele route could be the barrier to trade resulting from a roadblock raised at Alamata (south of Mekele, north of Dese) and designed to raise taxes and control grain flows to the Mekele region. If the costs associated with this trade restriction are not fully reflected in estimated transfer cost then price differentials will exceed estimated transfer cost with high frequency, leading to a high estimated probability of regime 3.

After the policy change, the Addis-Dese route experienced a highly statistically significant shift in probability from regime 2 to regime 1, with slow adjustment over a twenty-one-month period, while Addis-Mekele experienced a marginally statistically significant (p-value = 0.082) shift in probability from regime 3 to regime 2, with the adjustment occurring over a shorter four-month period. These changes are consistent with a rise in average price differentials and arbitrage profits on the Addis-Dese route and a fall in average price differentials and arbitrage profits on the Addis-Mekele route (see table 3). An obvious potential explanation for the Addis-Mekele result is removal of the Alamata roadblock, which did take place as part of the 1999 policy change. (10) The increase in price differentials on the Addis-Dese route may have occurred as more maize previously destined for Dese was now transshipped through to Mekele, putting upward pressure on Dese prices.

The final three columns of table 1 contain results for three routes for shipping maize to the most eastern food deficit region of Dire Dawa (see figure 1). Results for all three routes are very similar. Before the policy change there is a high probability (70% or more) of being in regime 2 and relatively low probability of being in regimes 1 and 3. This indicates that either maize is not shipping into Dire Dawa and price differentials do not generally encourage such shipments, or that shipping is occurring but the transfer generally occurs at a loss (price differentials do not cover transfer costs). This can also be seen in figure 2 where price differentials for the Addis-Dire Dawa route are generally at or below the parity bound prior to the policy change in October 1999. One possible explanation for the transfer continuing at a loss is that the EGTE may have been subsidizing maize flows into the Dire Dawa region for food security and/or political reasons (i.e., food aid).

With the October 1999 policy change, all three routes shipping to Dire Dawa switched immediately (zero adjustment period--see the bottom of table 2) to having a much higher probability of being in regime 3. In other words, for all three trade routes into Dire Dawa there was a major and statistically significant increase in the probability that price differentials exceed estimated transfer cost. The increases in mean price differentials and arbitrage profits reported in table 3 for the Dire Dawa maize routes after the policy change are consistent with these probability changes, as are the more frequent observation of price differentials above the parity bound for the Addis-Dire Dawa route observed in figure 2 after the October 1999 policy change. A possible explanation is that, with its commercial reorientation, the EGTE pulled back from these trade routes and market forces were unable to immediately fill the gap (perhaps because of poor infrastructure or lack of information and investment). The end result was relatively higher maize prices in Dire Dawa.

Overall, there is evidence of unexploited spatial arbitrage opportunities prior to the policy change in the Jima-Addis and Addis-Mekele routes, but all other routes show price differentials equal to or lower than estimated transfer cost. There are mixed results concerning the effects of the policy change. In particular, the October 1999 policy changes appear to have had the following effects: (a) little impact on trade between Jima and Addis; (b) increased spatial price differentials (and higher probability of unexploited arbitrage opportunities) for maize flowing from Nekemte to Addis, and flowing into the eastern food deficit market of Dire Dawa; (c) increased spatial price differentials (but without higher probability of unexploited arbitrage opportunities) for maize flowing from Addis to Dese; and (d) reduced spatial price differentials (and lower probability of unexploited arbitrage opportunities) for maize flowing north from Addis to Mekele. Most of the adjustments in regime probabilities occurred very rapidly (zero to five months), with the exception being the increased spatial efficiency for Addis-Dese trade, which took twenty-one months. The Addis-Dese route may have taken longer to adjust than other trade routes because it is the one market pair that experienced a substantial increase in the probability of regime 1. This suggests that it may take longer for maize markets to adjust toward full spatial efficiency than it does to adjust toward other regimes, perhaps because of lack of information and feedback about how markets are working in the absence of government participation.

Results for Wheat