The impact of increasing non-agricultural market access on EU demand for imported fish: implications for Lake Victoria chilled fillet exports

1. Introduction

The importance of the EU to the fish exporting industries in Tanzania and Uganda cannot be overstated. Since the lifting of import bans on fish from the Lake Victoria region in 2000, fish exports have grown approximately 115 per cent per year on average for both countries. Growth in Uganda has been so rapid that fish have become the second largest source of export revenue for the country (Abila, 2000; Uganda Export Promotion Board, 2005). Uganda and Tanzania primarily export Nile perch in fillet form, and as noted by Abila (2003), the development of the fish processing sectors in Lake Victoria riparian countries was the direct result of the extensive growth in Nile perch demand in developed countries, particularly the EU. Currently, the EU imports from 600 to 800 tons of chilled Nile perch fillets per week from Lake Victoria's riparian states and accounts for about 80 per cent of all chilled fillet exports from Tanzania and Uganda (Josupeit, 2005; UNCOMTRADE, 2006).¹

In 2005, the 25 member states of the EU were the largest importers of chilled fish fillets in the world, importing 268.5 million kg valued at US $1.6 billion. This represented a 6 per cent increase in quantity and a 16 per cent increase in value when compared with the previous year. When considering individual countries, eight EU member states were among the top ten importers in the world. These included (in declining order of value): France, Germany, Belgium, Italy, the UK, Sweden, Netherlands and Spain, where import values ranged from US $91.7 million to US $315 million (UNCOMTRADE, 2006).

The top suppliers of chilled fillets to the EU are Iceland, Norway, Tanzania and Uganda. Exports from these countries were 17.94, 31.73, 21.98 and 21.07 million kg, respectively, and valued at €39.8, €178.9, €89.0 and €92.8 million, respectively, in 2005. Combined, these countries accounted for 76 per cent of the total quantity of the EU imports and 77 per cent of the total import value (Eurostat, 2006). Although Iceland and Norway are not members of the EU, they receive tariff-free access for fish exports under the European Economic Area (EEA) agreement, which is an agreement between the EU and the European Free Trade Association (EFTA) (EFTA, 2007).²As members of the African, Caribbean and Pacific Group of States (ACP), fish exports from Uganda and Tanzania also have tariff-free access under the Cotonou Agreement (Ponte et al., 2005).

The WTO non-agricultural market access (NAMA) negotiations in the Doha Round have specifically mentioned reductions in tariffs on imported fish. Ponte et al.(2005)noted that African fish exporters have little direct interest in seeing NAMA tariff negotiations succeed because they already enjoy preferential access to the EU, and extending the preferential treatment to non-ACP countries may lead to preference erosion. Currently, the EU tariff for non-ACP countries is 18 per cent for chilled and fresh fish fillets [Technical Centre for Agricultural and Rural Cooperation ACP-EU (CTA), 2006]. Given the potential erosion in preferential treatment for Uganda and Tanzania resulting from multilateral tariff reductions, the purpose of this study is to access the impact of extending preferential treatment to non-ACP countries on EU demand for chilled fillets. Using a Rotterdam-type production model, the import demand for chilled fillets differentiated by country of origin is estimated for the EU. Results are used to assess the competitiveness of Tanzanian and Ugandan fillets and to simulate the impact of tariff reductions on EU imports from Iceland, Norway, Tanzania and Uganda. Specific objectives are (i) to estimate econometrically the demand for imported chilled fillets in the EU by country of origin, (ii) to obtain demand elasticities from the empirically estimated import demand parameters and (iii) to use the estimated parameters to project the impact of multilateral tariff reductions on source-specific imports.

2. Empirical model

Given the intermediate nature of traded products, the differential production model is used to estimate the import demand for chilled fish fillets to the EU.³ Production applications to import demand for origin-specific products are fairly recent. Past studies typically used the AIDS or Rotterdam models, which are based in consumer theory. Production applications to source- or quality-specific import demand include Davis and Jensen (1994), Koo et al.(2001)and Washington and Kilmer (2002). These studies (particularly Davis and Jensen) indicate that the import demand applications based on utility maximisation were conceptually flawed given that most goods entering into international trade require further processing before final demand delivery. Activities such as handling, insurance, transportation, storage, repackaging and retailing still occur even with final goods, resulting in a significant amount of domestic value added. Theoretically, specifying first-stage aggregates is more intuitive with a production approach and unconditional elasticities are more easily derived. Washington and Kilmer (2002)further showed that the unconditional derived demand estimates could differ significantly from unconditional consumer demand estimates.

In this study, it is assumed that fish imports are facilitated by trade intermediaries in the EU where firms import from various exporting countries and then resell the imported product domestically or re-export to other countries. Keizire (2004)noted that Lake Victoria exports are facilitated primarily by firms (particularly in Belgium and the Netherlands) serving as entry points for the EU. The output of these firms is the total quantity of imported fillets resold or re-exported, and the inputs are the factors of production required in the wholesale trade and the imported fillets. If we assume product differentiation across exporting countries, the input demand equations will not only be the demand for value added inputs such as labour, but also the demand for chilled fillets from each exporting country.

Similar to equation (1), f̄_it = (f_it + f_it−12)/2 and f_i = \big(w_ix_i/∑_iw_ix_i\big). Dx_it = log(w_it) − log(w_it−12) and Dw_it = log(w_it) − log(w_it−12), where x_iand w_irepresent the quantity and price of fillets from source country i. DX_tis the Divisia index.θ_iand π_ijare the parameters to be estimated, where θ_iis the marginal factor share coefficient and π_ijmeasures the conditional price effects. u_itis a random disturbance term. The differential derived demand model requires that the following parameter restrictions be met in order for the model to conform to theoretical considerations: ∑_iθ_i = 1 and ∑_iπ_ij = 0 (adding up); ∑_jπ_ij = 0 (homogeneity) and π_ij = π_ji(symmetry).

Equations (1) and (2) form a system in which equation (1) determines total imports and equation (2) depicts the sourcing of total imports across exporting countries. Equation (1) is derived from the first-order condition of the profit maximisation problem where the logarithmic change in output price is equated to the logarithmic change in marginal cost (Theil, 1980: 32, 38). Equation (2) is derived from the cost minimisation problem where costs are minimised subject to a general logarithmic production function (Theil, 1980: 34). The system assumes a two-step approach to profit maximisation suggested by Theil (1980)and Laitinen (1980): first, firms minimise input expenditure subject to the technology constraint (conditional on output and input prices) and, second, they maximise profit by varying output. The derived demand system is comparable to consumer demand systems, and the disturbances of equations (1) and (2) are stochastically independent by construction (Theil, 1980).

3. Empirical results

3.1. Data and descriptive statistics

The External Trade Section of the Statistical Office of the European Communities (Eurostat) provided the data used in this study, which was the 6-digit HS commodity classification ‘fresh and chilled fillets’. Import values were on a cost-insurance-freight basis. Exporting countries were Norway, Iceland, Tanzania and Uganda. The ROW quantities and values were calculated by subtracting the imports for the top four exporting countries from total extra-EU imports. Monthly data were used for estimation and the time period for the data was from September 2000 to May 2006. Import prices were calculated by dividing the value of the commodity by the quantity. As a proxy for output price, a per unit value measure for EU fillet exports was used. The EU exports were on a free-on-board basis. Initially, labour, utilities and energy were considered as value-added inputs in the model. However, given the high degree of multicollinearity between the price indexes for these inputs, labour was the only value-added input included. The output price measure and the wage index were also provided by Eurostat.

Table 1presents descriptive statistics for the variables used. There are considerable differences in average per unit values (import prices) for the import sources shown, with Uganda and Tanzania tending to receive prices that are €100 to €300 lower than the other sources. The average output price for the EU was significantly higher than all import prices. During the data period, Norway had the largest share of imports onto the EU market, and that of the ROW was nearly as large; Iceland and Tanzania had a smaller share of imports (both about 19 per cent), and Uganda's share was the smallest.

3.2. Estimation results

The total import equation and import demand system were estimated using the LSQ procedure in TSP version 5.0. This procedure uses the multivariate Gauss–Newton method to estimate the parameters in the system (Hall and Cummins, 2005a). Given the singularity of the import demand system, the ROW equation was dropped for estimation. Theil (1980: 92–94) showed that if the parameters in equations (1) and (2) are assumed constant and the errors normally distributed, then cov(e,u) = 0, which suggests that the total import equation need not be estimated jointly with the import demand system. The LR tests were used to test for AR(1) disturbances in the differential production model. The autocorrelation parameter was obtained using the full information maximum-likelihood procedure for singular systems found in Berndt and Savin (1975)and Beach and MacKinnon (1979).

The LR tests were also used to test the economic properties of homogeneity and symmetry. Log-likelihood values, test statistics and critical values are given in Table 2. The hypothesis of no autocorrelation was rejected at the 0.05 significance level, whereas the properties of homogeneity and symmetry could not be rejected. Therefore, all final estimated models have AR(1), homogeneity and symmetry imposed. Theory suggests that the matrix of conditional price effects should be negative semi-definite. This property is confirmed when all eigenvalues of the price coefficient matrix are non-positive. In our case, all eigenvalues were found to be non-positive.

Table 3presents estimation results for the output supply equation. The output price parameter estimate (0.085) was positive as expected and significant at the 0.05 significance level. Although wages had a negative impact on total imports (−2.316), this impact was not significant. This may be the result of using monthly data, where monthly changes in wages had little impact on total imports. The impact of source-specific prices on total imports was negative for all countries except Uganda, and significant for Iceland, Norway and the ROW.

The insignificant coefficients for Tanzanian and Ugandan prices indicate that as each export price increased, ceteris paribus, total fillets imported into the EU did not significantly change. It should be recalled that imports from Tanzania and Uganda were on average much cheaper than those from the other sources. Even the maximum prices for Tanzania and Uganda were still comparable to the mean prices of Norwegian and ROW fillets, and were still significantly less than the mean price of Icelandic fillets (Table 1). It may be the case that with rising prices, importers continued to purchase fillets from Tanzania and Uganda because of their relative inexpensiveness. Another possible explanation of these insignificant coefficients could be a high degree of substitutability between Lake Victoria fillets. Note that the conditional cross-price effect for Tanzania and Uganda (0.141, Table 4) was significantly higher than the cross-price effects between any of the other countries. The conditional and unconditional cross-price elasticities were also significantly larger for these two countries (Table 5). This suggests that as the price of Tanzania's fillets increased, the EU increased imports from Uganda and vice versa. The cross-price elasticities being close to unity indicate that substitution may have occurred to the degree that total imports remained relatively unaffected. Close substitutability would imply a high degree of correlation between the two prices, which may itself be the cause of the insignificant coefficients. In fact, the partial correlation coefficient between these two prices was 0.91; however, excluding each price in turn from the total import equation did not change estimates or improve significance. Therefore, it can be concluded that the insignificant coefficients for these prices are not due to multicollinearity.

Conditional import demand estimates for the EU are presented in Table 4. Marginal factor share estimates indicate a positive and significant relationship between fillet imports from all sources and total fillet imports. As the EU increased total demand for imported chilled fillets, imports from Iceland and Norway had the largest absolute increase, whereas the increase for Tanzania and Uganda (0.054 and 0.145) was relatively smaller. The estimated conditional own-price effects were all negative as expected, and similar in size. Cross-price parameter estimates indicate a significant competitive relationship between Iceland, Norway and the ROW, but with the exception of Iceland and Tanzania, no significant price competition existed between the Lake Victoria countries and the other competing exporters. As previously mentioned, a relatively strong competitive relationship existed between fillets from Uganda and Tanzania, whose cross-price elasticity is 0.141. Iceland and Tanzania aside, this suggests that imports from Lake Victoria were to a degree independent of competing exporter prices (conditional on total imports).

3.3. Conditional and unconditional elasticities

Table 5shows the estimates of the conditional elasticities of derived demand for imported chilled fillets (calculated at sample means). The Divisia index elasticities, which measure the responsiveness of source-specific imports to changes in total imports, indicate that a one-percent increase in total imports increased the EU imports of chilled fillets from these countries by their elasticity values. Compared to other exporters, the Divisia index elasticity for Tanzania was relatively small. This is due to the imports from Tanzania not keeping pace with total EU imports. From 2001 to 2005, total EU imports of chilled fillets increased 62 per cent, with imports from Iceland, Norway, Uganda and ROW increasing by 68, 90, 63 and 76 per cent, respectively, whereas imports from Tanzania increased by only 17 per cent. According to CTA (2007), the slower growth in Tanzania's fish exports may be the result of flight decreases out of Mwanza airport in Tanzania due to high taxes and poor airport maintenance. It was also noted that the export procedures and formalities for Nile perch products, including export licenses issued by the Tanzania Ministry of Trade, have driven up processing costs making exporting more difficult when compared with Uganda and Kenya.

The conditional own-price/cross-price elasticities measure the impact of import price changes on source-specific imports, holding total imports constant, and preserve the same signs as the coefficient estimates in Table 4. Again, the cross-price elasticities show a strong competitive relationship between Ugandan and Tanzanian fillets in the EU. Both these elasticities were significantly larger when compared with the other cross-prices elasticities.

Unconditional elasticities of derived demand are reported in Table 6. Although significant, the impact of EU prices (output prices) on source-specific imports was small, particularly for Tanzania. Unconditional own-price elasticities indicate a significant inverse relationship between source-specific prices and quantities. Note that the conditional and unconditional own-price elasticities for Tanzania and Uganda were very close, due to the negligible effect of changes in export prices for Tanzania and Uganda on total imports.

The unconditional cross-price elasticity also includes the impact of source-specific price changes that are induced by a change in the total imported. It is possible that a change in fillet prices from any country could affect total imports so much that the total volume effect might outweigh the direct impact of relative price changes alone. This happens in a number of cases. Taking indirect volume effects also into account, Norwegian and Tanzanian imports (conditional substitutes for imports from Iceland) become price-neutral for Icelandic imports, whereas imports from ROW switch from being conditional substitutes to being unconditional complements for Icelandic imports. For ROW, the volume effect just balances out the conditional gross substitutability of Icelandic and Norwegian imports, so that there is no unconditional price relationship. In the case of Tanzania and Uganda, however, various conditional cross-price elasticites that were not significantly different from zero are replaced by significant and negative unconditional elasticities. This happens for Tanzania with respect to Norwegian and ROW import prices, and for Uganda with respect to Icelandic, Norwegian and ROW import prices. In all these cases, the unconditional elasticities reveal the presence of unconditional complementarity for Lake Victoria imports.

3.4. The impact of tariff reductions on import demand

In this section, we explore the effect of a possible agreement on NAMA in the context of the Doha Development Round. Our assumption is that a new arrangement will involve a reduction of some kind in the EU tariff on chilled fillet imports. As already mentioned, the EU tariff on fresh fillets from non-ACP countries is currently 18 per cent. Given the EEA and Cotonou Agreements, which maintain tariff-free access for Iceland, Norway, Uganda and Tanzania, tariff reductions would only apply to imports from the ROW. Three possibilities are analysed here: (i) an 18 per cent reduction in ROW prices, which assumes that the tariff is assessed on all ROW imports and will be completely abolished under new NAMA rules; (ii) a 10 per cent decrease in ROW prices and (iii) a 5 per cent decrease in ROW prices. The latter two possibilities account for the fact that the import duties are probably not imposed on all ROW imports, and that the NAMA agreement may not result in the complete abolition of tariffs on fish.

Source-specific imports are projected conditional upon 18, 10 and 5 per cent decreases in ROW prices, representing different degrees of tariff reduction on ROW imports. Projections were obtained using an elasticity-based forecasting procedure [see equation (A2) in the appendix] and results are presented in Table 7.

Our results contradict the preference erosion argument that suggests ACP countries will be negatively affected by multilateral tariff reductions. If tariff reductions are such that ROW prices decrease by 18 per cent, then total EU imports are projected to increase by 4.1 per cent and imports from Tanzania and Uganda are projected to increase by 2.27 and 2.13 per cent, respectively. Imports from Iceland are also projected to increase under all three scenarios, although by a lower percentage than those from Lake Victoria, and imports from Norway are projected to decrease.

Given the unconditional complementary relationship between ROW, Uganda and Tanzania, a decrease in ROW prices should result in greater imports from Lake Victoria. The feared negative impact of preference erosion is the result of considering only conditional price effects, which, to the extent that they reveal conditional substitutability, would suggest that lower competitor prices would lead to a decrease in EU fish imports from ACP countries. However, the conditional estimates obtained in this study show that the Lake Victoria and ROW fillets were price independent even without taking the indirect volume effect into account.⁶

4. Summary and conclusions

This paper reports the estimation of import demand functions for origin-specific chilled fish fillets into the EU using a Rotterdam-type production model. Overall, results suggested that chilled fillet exports from Lake Victoria were independent of changes in Iceland, Norway and ROW export prices conditional upon total imports remaining constant. Unconditionally, an increase in the price of fillets from other countries had a negative impact on Lake Victoria exports. Results further showed that Tanzanian and Ugandan fillets were substitutes for each other, conditionally and unconditionally.

Lake Victoria is the primary source of Nile perch to the EU, and it may be the case that the EU importers view Nile perch as a unique product within the chilled fish fillet category. This may explain the high degree of competitiveness between fillets from Tanzania and Uganda, and the absence of conditional price-competitiveness between Lake Victoria and other suppliers. This suggests that an increase in EU imports from one Lake Victoria country due to lower prices will come at the expense of the other (holding total imports constant). This is somewhat disheartening because both countries have identified their fish exporting sectors as a means of economic development (Dijkstra, 2001).

In years past, Nile perch fillets were positioned as cheaper substitutes for cod and cod-like species (Bambona, 2002). Given the recent increase in relatively cheaper imports from Asia, Josupeit (2005)notes that more value addition and promotional campaigns are needed to show the quality and taste advantages of Nile perch in the EU. The conditional price-independence between Lake Victoria fillets and fillets from other countries signifies the uniqueness of Nile perch in the EU and the potential for increased product promotion.

A positive implication of the results is that as other exporting countries become more competitive, Lake Victoria exports to the EU will remain unaffected or will likely increase. This would also be the case if the increased competitiveness of other countries is due to the effect of multilateral tariff reductions on EU import prices. Although it is believed that an overall increase in market access will impact negatively on EU fish imports from ACP countries, the results of this study suggest that this is not necessarily the case. If tariff reductions result in lower prices for competing exporters, total EU imports are projected to increase by more than 4 per cent, resulting in an increase in chilled fillet imports from Lake Victoria of 2.3 and 2.1 per cent for Tanzania and Uganda, respectively. The preference erosion argument, which suggests that lower tariffs will erode the competitive position of ACP countries, only considers the effect of relative price changes holding total import volume constant (conditional price effects). However, in our study, the estimated conditional price effects suggest that EU demand for fillets from Lake Victoria was independent of changes in ROW prices, which indicates that Lake Victoria exports to the EU would remain unaffected by lower prices of competing exporters, even holding import quantity constant.

One might infer from these results that ACP countries need not always argue for special concessions in the Doha Round to counter the effects of preference erosion. For other fish products, such as frozen fillets where competition from Asian exporters is much greater, it could be the case that Lake Victoria and ROW frozen fillets are substitutes, conditionally and unconditionally; however, in the EU market for chilled fillets, expanding market access may actually be good for Lake Victoria exports.

Acknowledgements

I am greatly appreciative of the assistance of Dr. Caleb Tamwesigire, Lecturer in the Department of Finance, Makerere University Business School (Kampala, Uganda). Lastly, this article was greatly improved by the comments and suggestions from the editor, Alison Burrell, and three anonymous reviewers. This study resulted from a trip to Uganda in 2004 funded by the United Negro College Fund-International Development Partnership Program.

References

Appendix: forecasting procedure

where

Equations (1) and (2) were estimated using all except the last 12 months of the sample data (September 2000–May 2005), and the estimates were used to forecast the remaining years (June 2005–May 2006). The root mean square error (RMSE) for each approach was compared. The average RMSE of the five forecast import demands was 3,436 for the unconditional elasticity approach and 5,778 for the unconditional model approach (representing a forecasting error of 4.3 and 7.3 per cent, respectively, relative to the mean for the total sample period). The average RMSE for the unconditional elasticity approach using elasticities calculated at the mean for the last 12 months was even lower at 3,316, and hence this last-named procedure was used to project the impact of expanding market access in the EU.